WIP Add AES-XTS client encryption support

Maybe I should remove all of this and use libwebsockets :)

.gitea/workflows/buildenv.Dockerfile

@@ -1,29 +1,28 @@
-FROM node:16-bookworm
+FROM node:16-bullseye
 
 WORKDIR /root
 
-ADD ./docker/etc/apt/trusted.gpg.d /etc/apt/trusted.gpg.d
+ADD ./docker/vitastor.gpg /etc/apt/trusted.gpg.d
 
-RUN echo 'deb http://deb.debian.org/debian bookworm-backports main' >> /etc/apt/sources.list; \
-    echo 'deb http://vitastor.io/debian bookworm main' >> /etc/apt/sources.list; \
+RUN echo 'deb http://deb.debian.org/debian bullseye-backports main' >> /etc/apt/sources.list; \
+    echo 'deb http://vitastor.io/debian bullseye main' >> /etc/apt/sources.list; \
     echo >> /etc/apt/preferences; \
     echo 'Package: *' >> /etc/apt/preferences; \
-    echo 'Pin: release n=bookworm-backports' >> /etc/apt/preferences; \
+    echo 'Pin: release a=bullseye-backports' >> /etc/apt/preferences; \
     echo 'Pin-Priority: 500' >> /etc/apt/preferences; \
     echo >> /etc/apt/preferences; \
     echo 'Package: *' >> /etc/apt/preferences; \
     echo 'Pin: origin "vitastor.io"' >> /etc/apt/preferences; \
     echo 'Pin-Priority: 1000' >> /etc/apt/preferences; \
-    perl -i -pe 's/Types: deb$/Types: deb deb-src/' /etc/apt/sources.list.d/debian.sources; \
     grep '^deb ' /etc/apt/sources.list | perl -pe 's/^deb/deb-src/' >> /etc/apt/sources.list; \
     echo 'APT::Install-Recommends false;' >> /etc/apt/apt.conf; \
     echo 'APT::Install-Suggests false;' >> /etc/apt/apt.conf
 
 RUN apt-get update
-RUN apt-get -y install etcd qemu-system-x86 qemu-block-extra qemu-utils fio libasan8 \
+RUN apt-get -y install etcd qemu-system-x86 qemu-block-extra qemu-utils fio libasan5 \
     libgoogle-perftools-dev devscripts libjerasure-dev cmake libibverbs-dev libisal-dev
 RUN apt-get -y build-dep fio qemu=`dpkg -s qemu-system-x86|grep ^Version:|awk '{print $2}'`
-RUN apt-get update && apt-get -y install jq lp-solve sudo nfs-common fdisk parted libc-ares-dev udev
+RUN apt-get update && apt-get -y install jq lp-solve sudo nfs-common fdisk parted
 RUN apt-get --download-only source fio qemu=`dpkg -s qemu-system-x86|grep ^Version:|awk '{print $2}'`
 
 RUN set -ex; \

.gitea/workflows/test.yml

@@ -234,6 +234,60 @@ jobs:
           echo ""
         done
 
+  test_etcd_fail_https:
+    runs-on: ubuntu-latest
+    needs: build
+    container: ${{env.TEST_IMAGE}}:${{github.sha}}
+    steps:
+    - name: Run test
+      id: test
+      timeout-minutes: 10
+      run: ETCD_SCHEME=https /root/vitastor/tests/test_etcd_fail.sh
+    - name: Print logs
+      if: always() && steps.test.outcome == 'failure'
+      run: |
+        for i in /root/vitastor/testdata/*.log /root/vitastor/testdata/*.txt; do
+          echo "-------- $i --------"
+          cat $i
+          echo ""
+        done
+
+  test_etcd_fail_https_antietcd:
+    runs-on: ubuntu-latest
+    needs: build
+    container: ${{env.TEST_IMAGE}}:${{github.sha}}
+    steps:
+    - name: Run test
+      id: test
+      timeout-minutes: 10
+      run: ETCD_SCHEME=https ANTIETCD=1 /root/vitastor/tests/test_etcd_fail.sh
+    - name: Print logs
+      if: always() && steps.test.outcome == 'failure'
+      run: |
+        for i in /root/vitastor/testdata/*.log /root/vitastor/testdata/*.txt; do
+          echo "-------- $i --------"
+          cat $i
+          echo ""
+        done
+
+  test_snapshot_https:
+    runs-on: ubuntu-latest
+    needs: build
+    container: ${{env.TEST_IMAGE}}:${{github.sha}}
+    steps:
+    - name: Run test
+      id: test
+      timeout-minutes: 3
+      run: ETCD_SCHEME=https /root/vitastor/tests/test_snapshot.sh
+    - name: Print logs
+      if: always() && steps.test.outcome == 'failure'
+      run: |
+        for i in /root/vitastor/testdata/*.log /root/vitastor/testdata/*.txt; do
+          echo "-------- $i --------"
+          cat $i
+          echo ""
+        done
+
   test_interrupted_rebalance:
     runs-on: ubuntu-latest
     needs: build
@@ -1224,6 +1278,24 @@ jobs:
           echo ""
         done
 
+  test_checksum_xxhash:
+    runs-on: ubuntu-latest
+    needs: build
+    container: ${{env.TEST_IMAGE}}:${{github.sha}}
+    steps:
+    - name: Run test
+      id: test
+      timeout-minutes: 3
+      run: TEST_NAME=xxhash OSD_ARGS="--data_csum_type xxh3_32" /root/vitastor/tests/test_checksum.sh
+    - name: Print logs
+      if: always() && steps.test.outcome == 'failure'
+      run: |
+        for i in /root/vitastor/testdata/*.log /root/vitastor/testdata/*.txt; do
+          echo "-------- $i --------"
+          cat $i
+          echo ""
+        done
+
   test_old_checksum:
     runs-on: ubuntu-latest
     needs: build

.gitea/workflows/tests-to-yaml.pl

@@ -38,6 +38,10 @@ for my $line (<>)
             {
                 $test_name .= '_antietcd';
             }
+            elsif ($1 eq 'ETCD_SCHEME' && $2 eq 'https')
+            {
+                $test_name .= '_https';
+            }
             elsif ($1 eq 'OLD')
             {
                 $test_name =~ s/^test_/test_old_/s;

CMakeLists.txt

@@ -2,7 +2,7 @@ cmake_minimum_required(VERSION 2.8.12)
 
 project(vitastor)
 
-set(VITASTOR_VERSION "3.0.6")
+set(VITASTOR_VERSION "3.0.3")
 
 include(CTest)
 

csi/Dockerfile

@@ -25,8 +25,8 @@ RUN apt-get update && \
         # NFS mount dependencies
         nfs-common netbase \
         # dependencies of qemu-storage-daemon
-        libaio1t64 libc6 libfuse3-4 libglib2.0-0t64 libgmp10 libgnutls30t64 \
-        libhogweed6t64 libnettle8t64 libnuma1 libselinux1 liburing2 libzstd1 zlib1g && \
+        libnuma1 liburing2 libglib2.0-0 libfuse3-3 libaio1 libzstd1 libnettle8 \
+        libgmp10 libhogweed6 libp11-kit0 libidn2-0 libunistring2 libtasn1-6 libpcre2-8-0 libffi8 && \
     apt-get clean && \
     (echo options nbd nbds_max=128 > /etc/modprobe.d/nbd.conf)
 

csi/Dockerfile.local

@@ -1,5 +1,5 @@
 # Compile stage
-FROM golang:trixie AS build
+FROM golang:bookworm AS build
 
 ADD go.sum go.mod /app/
 RUN cd /app; CGO_ENABLED=1 GOOS=linux GOARCH=amd64 go mod download -x
@@ -9,7 +9,7 @@ RUN perl -i -e '$/ = undef; while(<>) { s/\n\s*(\{\s*\n)/$1\n/g; s/\}(\s*\n\s*)e
     CGO_ENABLED=1 GOOS=linux GOARCH=amd64 go build -o vitastor-csi
 
 # Final stage
-FROM debian:trixie
+FROM debian:bookworm
 
 LABEL maintainers="Vitaliy Filippov <vitalif@yourcmc.ru>"
 LABEL description="Vitastor CSI Driver"
@@ -36,8 +36,8 @@ ADD deb /deb
 
 RUN apt-get update && \
     apt-get -y install /deb/vitastor-client_*.deb && \
-    wget https://vitastor.io/archive/qemu/qemu-trixie-9.2.2%2Bds-1%2Bvitastor4/qemu-utils_9.2.2%2Bds-1%2Bvitastor4_amd64.deb && \
-    wget https://vitastor.io/archive/qemu/qemu-trixie-9.2.2%2Bds-1%2Bvitastor4/qemu-block-extra_9.2.2%2Bds-1%2Bvitastor4_amd64.deb && \
+    wget https://vitastor.io/archive/qemu/qemu-bookworm-9.2.2%2Bds-1%2Bvitastor4/qemu-utils_9.2.2%2Bds-1%2Bvitastor4_amd64.deb && \
+    wget https://vitastor.io/archive/qemu/qemu-bookworm-9.2.2%2Bds-1%2Bvitastor4/qemu-block-extra_9.2.2%2Bds-1%2Bvitastor4_amd64.deb && \
     dpkg -x qemu-utils*.deb tmp1 && \
     dpkg -x qemu-block-extra*.deb tmp1 && \
     cp -a tmp1/usr/bin/qemu-storage-daemon /usr/bin/ && \

csi/Makefile

@@ -1,4 +1,4 @@
-VITASTOR_VERSION ?= v3.0.6
+VITASTOR_VERSION ?= v3.0.3
 
 all: build push
 

csi/deploy/004-csi-nodeplugin.yaml

@@ -49,7 +49,7 @@ spec:
             capabilities:
               add: ["SYS_ADMIN"]
             allowPrivilegeEscalation: true
-          image: vitalif/vitastor-csi:v3.0.6
+          image: vitalif/vitastor-csi:v3.0.3
           args:
             - "--node=$(NODE_ID)"
             - "--endpoint=$(CSI_ENDPOINT)"

csi/deploy/007-csi-provisioner.yaml

@@ -121,7 +121,7 @@ spec:
             privileged: true
             capabilities:
               add: ["SYS_ADMIN"]
-          image: vitalif/vitastor-csi:v3.0.6
+          image: vitalif/vitastor-csi:v3.0.3
           args:
             - "--node=$(NODE_ID)"
             - "--endpoint=$(CSI_ENDPOINT)"

csi/src/config.go

@@ -5,7 +5,7 @@ package vitastor
 
 const (
     vitastorCSIDriverName    = "csi.vitastor.io"
-    vitastorCSIDriverVersion = "3.0.6"
+    vitastorCSIDriverVersion = "3.0.3"
 )
 
 // Config struct fills the parameters of request or user input

debian/build-vitastor-ubuntu-resolute.sh

@@ -1,5 +0,0 @@
-#!/bin/bash
-# 26.04 Resolute Raccoon
-
-docker build --build-arg DISTRO=ubuntu --build-arg REL=resolute -t vitastor-buildenv:resolute -f vitastor-buildenv.Dockerfile .
-docker run -it --rm -e REL=resolute -v `dirname $0`/../:/root/vitastor vitastor-buildenv:resolute /root/vitastor/debian/vitastor-build.sh

debian/changelog

@@ -1,4 +1,4 @@
-vitastor (3.0.6-1) unstable; urgency=medium
+vitastor (3.0.3-1) unstable; urgency=medium
 
   * Bugfixes
 

docker/Makefile

@@ -1,4 +1,4 @@
-VITASTOR_VERSION ?= v3.0.6
+VITASTOR_VERSION ?= v3.0.3
 
 all: build push
 

docker/etc/apt/preferences

@@ -1,3 +1,3 @@
 Package: *
-Pin: release n=trixie-backports
+Pin: release n=bookworm-backports
 Pin-Priority: 500

docker/etc/apt/sources.list.d/vitastor.list

@@ -1,2 +1,2 @@
-deb http://vitastor.io/debian trixie main
-#deb http://http.debian.net/debian/ trixie-backports main
+deb http://vitastor.io/debian bookworm main
+deb http://http.debian.net/debian/ bookworm-backports main

docker/etc/systemd/system/vitastor-host.service

@@ -7,7 +7,7 @@ PartOf=vitastor.target
 [Service]
 Restart=always
 EnvironmentFile=/etc/vitastor/docker.conf
-ExecStart=bash -c 'docker run --rm -i -v /etc/vitastor:/etc/vitastor -v /dev:/dev -v /run:/run -e SYSTEMD_IN_CHROOT=0 \
+ExecStart=bash -c 'docker run --rm -i -v /etc/vitastor:/etc/vitastor -v /dev:/dev -v /run:/run \
     --security-opt seccomp=unconfined --privileged --pid=host --log-driver none --network host --name vitastor vitastor:$VITASTOR_VERSION \
     sleep.sh'
 ExecStartPost=udevadm trigger

docker/etc/vitastor/docker.conf

@@ -4,7 +4,7 @@
 #
 
 # Desired Vitastor version
-VITASTOR_VERSION=v3.0.6
+VITASTOR_VERSION=v3.0.3
 
 # Additional arguments for all containers
 # For example, you may want to specify a custom logging driver here

docker/install.sh

@@ -2,7 +2,8 @@
 
 set -e
 
-cp -urv /etc/systemd/system/vitastor* /host-etc/systemd/system/
-cp -urv /etc/udev/rules.d /host-etc/udev/
+cp -urv /etc/default /host-etc/
+cp -urv /etc/systemd /host-etc/
+cp -urv /etc/udev /host-etc/
 cp -urnv /etc/vitastor /host-etc/
 cp -urnv /opt/scripts/* /host-bin/

docs/config/osd.en.md

@@ -38,7 +38,6 @@ with an OSD restart or, for some of them, even without restarting by updating co
 - [journal_io](#journal_io)
 - [journal_sector_buffer_count](#journal_sector_buffer_count)
 - [journal_no_same_sector_overwrites](#journal_no_same_sector_overwrites)
-- [skip_corrupted_meta_entries](#skip_corrupted_meta_entries)
 - [throttle_small_writes](#throttle_small_writes)
 - [throttle_target_iops](#throttle_target_iops)
 - [throttle_target_mbs](#throttle_target_mbs)
@@ -280,19 +279,13 @@ Maximum number of journal flushers (see above min_flusher_count).
 - Type: boolean
 - Default: true
 
-Only for the old store ([meta_format](layout-osd.en.md#meta_format) 2).
-
-This parameter makes Vitastor keep a copy of metadata area in memory as it is
-on disk, in addition to the metadata database. When the option is enabled, every
-metadata entry is effectively stored in RAM twice. It's required for good performance
-because it allows to avoid additional read-modify-write cycles during metadata
-modifications. Metadata area size with the old store is roughly 224 MB per 1 TB
-of data. You can turn the option off to reduce memory usage by this value, but
-it will reduce performance.
-
-For the new store ([meta_format](layout-osd.en.md#meta_format) 3), the option
-may be changed in the future to support operation without loading full metadata
-database in memory.
+This parameter makes Vitastor always keep metadata area of the block device
+in memory. It's required for good performance because it allows to avoid
+additional read-modify-write cycles during metadata modifications. Metadata
+area size is currently roughly 224 MB per 1 TB of data. You can turn it off
+to reduce memory usage by this value, but it will hurt performance. This
+restriction is likely to be removed in the future along with the upgrade
+of the metadata storage scheme.
 
 ## inmemory_journal
 
@@ -371,8 +364,6 @@ blocks. The only situation when you should increase it to a larger value
 is when you enable journal_no_same_sector_overwrites. In this case set
 it to, for example, 1024.
 
-Not applicable to the new store ([meta_format](layout-osd.en.md#meta_format) 3).
-
 ## journal_no_same_sector_overwrites
 
 - Type: boolean
@@ -386,17 +377,6 @@ journal after writing it instead of possibly overwriting it the second time.
 
 Most (99%) other SSDs don't need this option.
 
-Not applicable to the new store ([meta_format](layout-osd.en.md#meta_format) 3).
-
-## skip_corrupted_meta_entries
-
-- Type: boolean
-- Default: false
-
-Only for the new store ([meta_format](layout-osd.en.md#meta_format) 3).
-Allow OSD to start when some metadata entries or blocks are corrupted by
-skipping them. Should be only used as an emergency measure.
-
 ## throttle_small_writes
 
 - Type: boolean

docs/config/osd.ru.md

@@ -39,7 +39,6 @@
 - [journal_io](#journal_io)
 - [journal_sector_buffer_count](#journal_sector_buffer_count)
 - [journal_no_same_sector_overwrites](#journal_no_same_sector_overwrites)
-- [skip_corrupted_meta_entries](#skip_corrupted_meta_entries)
 - [throttle_small_writes](#throttle_small_writes)
 - [throttle_target_iops](#throttle_target_iops)
 - [throttle_target_mbs](#throttle_target_mbs)
@@ -288,19 +287,13 @@ Flusher - это микро-поток (корутина), которая коп
 - Тип: булево (да/нет)
 - Значение по умолчанию: true
 
-Только для старого хранилища ([meta_format](layout-osd.en.md#meta_format) 2).
-
-Данный параметр заставляет Vitastor всегда держать копию области метаданных
-в памяти в том же виде, как она лежит на диске, в дополнение к БД метаданных.
-То есть, с включённой опцией каждая запись метаданных хранится в памяти дважды.
-Это нужно, чтобы избегать дополнительных операций чтения с диска при записи.
-Размер области метаданных в старом хранилище составляет примерно 224 МБ на
-1 ТБ данных. Вы можете отключить опцию, чтобы снизить потребление памяти
-примерно на эту величину, но при этом также снизится и производительность.
-
-Для нового хранилища ([meta_format](layout-osd.en.md#meta_format) 3) опция,
-возможно, будет переработана в будущем для поддержки работы без полной
-загрузки метаданных в памяти.
+Данный параметр заставляет Vitastor всегда держать область метаданных диска
+в памяти. Это нужно, чтобы избегать дополнительных операций чтения с диска
+при записи. Размер области метаданных на данный момент составляет примерно
+224 МБ на 1 ТБ данных. При включении потребление памяти снизится примерно
+на эту величину, но при этом также снизится и производительность. В будущем,
+после обновления схемы хранения метаданных, это ограничение, скорее всего,
+будет ликвидировано.
 
 ## inmemory_journal
 
@@ -383,8 +376,6 @@ fsync небезопасным даже с режимом "directsync".
 нужно менять - это если вы включаете journal_no_same_sector_overwrites. В
 этом случае установите данный параметр, например, в 1024.
 
-Неприменимо к новому хранилищу ([meta_format](layout-osd.en.md#meta_format) 3).
-
 ## journal_no_same_sector_overwrites
 
 - Тип: булево (да/нет)
@@ -400,18 +391,6 @@ fsync небезопасным даже с режимом "directsync".
 
 Почти все другие SSD (99% моделей) не требуют данной опции.
 
-Неприменимо к новому хранилищу ([meta_format](layout-osd.en.md#meta_format) 3).
-
-## skip_corrupted_meta_entries
-
-- Тип: булево (да/нет)
-- Значение по умолчанию: false
-
-Только для нового хранилища ([meta_format](layout-osd.en.md#meta_format) 3).
-Разрешить OSD запускаться, даже если часть блоков или записей метаданных
-повреждена, пропуская их. Опция предназначена для использования только в
-целях аварийного восстановления.
-
 ## throttle_small_writes
 
 - Тип: булево (да/нет)

docs/config/src/osd.yml

@@ -253,33 +253,21 @@
   type: bool
   default: true
   info: |
-    Only for the old store ([meta_format](layout-osd.en.md#meta_format) 2).
-
-    This parameter makes Vitastor keep a copy of metadata area in memory as it is
-    on disk, in addition to the metadata database. When the option is enabled, every
-    metadata entry is effectively stored in RAM twice. It's required for good performance
-    because it allows to avoid additional read-modify-write cycles during metadata
-    modifications. Metadata area size with the old store is roughly 224 MB per 1 TB
-    of data. You can turn the option off to reduce memory usage by this value, but
-    it will reduce performance.
-
-    For the new store ([meta_format](layout-osd.en.md#meta_format) 3), the option
-    may be changed in the future to support operation without loading full metadata
-    database in memory.
+    This parameter makes Vitastor always keep metadata area of the block device
+    in memory. It's required for good performance because it allows to avoid
+    additional read-modify-write cycles during metadata modifications. Metadata
+    area size is currently roughly 224 MB per 1 TB of data. You can turn it off
+    to reduce memory usage by this value, but it will hurt performance. This
+    restriction is likely to be removed in the future along with the upgrade
+    of the metadata storage scheme.
   info_ru: |
-    Только для старого хранилища ([meta_format](layout-osd.en.md#meta_format) 2).
-
-    Данный параметр заставляет Vitastor всегда держать копию области метаданных
-    в памяти в том же виде, как она лежит на диске, в дополнение к БД метаданных.
-    То есть, с включённой опцией каждая запись метаданных хранится в памяти дважды.
-    Это нужно, чтобы избегать дополнительных операций чтения с диска при записи.
-    Размер области метаданных в старом хранилище составляет примерно 224 МБ на
-    1 ТБ данных. Вы можете отключить опцию, чтобы снизить потребление памяти
-    примерно на эту величину, но при этом также снизится и производительность.
-
-    Для нового хранилища ([meta_format](layout-osd.en.md#meta_format) 3) опция,
-    возможно, будет переработана в будущем для поддержки работы без полной
-    загрузки метаданных в памяти.
+    Данный параметр заставляет Vitastor всегда держать область метаданных диска
+    в памяти. Это нужно, чтобы избегать дополнительных операций чтения с диска
+    при записи. Размер области метаданных на данный момент составляет примерно
+    224 МБ на 1 ТБ данных. При включении потребление памяти снизится примерно
+    на эту величину, но при этом также снизится и производительность. В будущем,
+    после обновления схемы хранения метаданных, это ограничение, скорее всего,
+    будет ликвидировано.
 - name: inmemory_journal
   type: bool
   default: true
@@ -398,15 +386,11 @@
     blocks. The only situation when you should increase it to a larger value
     is when you enable journal_no_same_sector_overwrites. In this case set
     it to, for example, 1024.
-
-    Not applicable to the new store ([meta_format](layout-osd.en.md#meta_format) 3).
   info_ru: |
     Максимальное число буферов, разрешённых для использования под записываемые
     в журнал блоки метаданных. Единственная ситуация, в которой этот параметр
     нужно менять - это если вы включаете journal_no_same_sector_overwrites. В
     этом случае установите данный параметр, например, в 1024.
-
-    Неприменимо к новому хранилищу ([meta_format](layout-osd.en.md#meta_format) 3).
 - name: journal_no_same_sector_overwrites
   type: bool
   default: false
@@ -418,8 +402,6 @@
     journal after writing it instead of possibly overwriting it the second time.
 
     Most (99%) other SSDs don't need this option.
-
-    Not applicable to the new store ([meta_format](layout-osd.en.md#meta_format) 3).
   info_ru: |
     Включайте данную опцию для SSD вроде Intel D3-S4510 и D3-S4610, которые
     ОЧЕНЬ не любят, когда ПО перезаписывает один и тот же сектор несколько раз
@@ -430,20 +412,6 @@
     самого сектора.
 
     Почти все другие SSD (99% моделей) не требуют данной опции.
-
-    Неприменимо к новому хранилищу ([meta_format](layout-osd.en.md#meta_format) 3).
-- name: skip_corrupted_meta_entries
-  type: bool
-  default: false
-  info: |
-    Only for the new store ([meta_format](layout-osd.en.md#meta_format) 3).
-    Allow OSD to start when some metadata entries or blocks are corrupted by
-    skipping them. Should be only used as an emergency measure.
-  info_ru: |
-    Только для нового хранилища ([meta_format](layout-osd.en.md#meta_format) 3).
-    Разрешить OSD запускаться, даже если часть блоков или записей метаданных
-    повреждена, пропуская их. Опция предназначена для использования только в
-    целях аварийного восстановления.
 - name: throttle_small_writes
   type: bool
   default: false

docs/installation/docker.en.md

@@ -26,37 +26,13 @@ at Vitastor Kubernetes operator: https://github.com/Antilles7227/vitastor-operat
 The instruction is very simple.
 
 1. Download a Docker image of the desired version: \
-   `docker pull vitalif/vitastor:v3.0.6`
+   `docker pull vitalif/vitastor:v3.0.3`
 2. Install scripts to the host system: \
-   `docker run --rm -it -v /etc:/host-etc -v /usr/bin:/host-bin vitalif/vitastor:v3.0.6 install.sh`
+   `docker run --rm -it -v /etc:/host-etc -v /usr/bin:/host-bin vitalif/vitastor:v3.0.3 install.sh`
 3. Reload udev rules: \
    `udevadm control --reload-rules`
-4. Enable the vitastor-host service: \
-   `systemctl enable --now vitastor-host`
 
-After these steps, you can return to [Quick Start](../intro/quickstart.en.md).
-
-## Podman
-
-If you use Podman, run the following commands as root before installing Vitastor containers:
-
-```
-ln -s podman /usr/bin/docker
-
-mkdir -p /etc/systemd/system/systemd-udevd.service.d
-
-cat >/etc/systemd/system/systemd-udevd.service.d/override.conf <<EOF
-[Service]
-CapabilityBoundingSet=~
-SystemCallFilter=@mount capset
-EOF
-
-systemctl daemon-reload
-
-systemctl restart systemd-udevd
-```
-
-Without it, udev fails to do calls into a Podman container and Vitastor disk detection doesn't work.
+And you can return to [Quick Start](../intro/quickstart.en.md).
 
 ## Upgrading Containers
 

docs/installation/docker.ru.md

@@ -25,39 +25,14 @@ Vitastor можно установить в Docker/Podman. При этом etcd,
 Инструкция по установке максимально простая.
 
 1. Скачайте Docker-образ желаемой версии: \
-   `docker pull vitalif/vitastor:v3.0.6`
+   `docker pull vitalif/vitastor:v3.0.3`
 2. Установите скрипты в хост-систему командой: \
-   `docker run --rm -it -v /etc:/host-etc -v /usr/bin:/host-bin vitalif/vitastor:v3.0.6 install.sh`
+   `docker run --rm -it -v /etc:/host-etc -v /usr/bin:/host-bin vitalif/vitastor:v3.0.3 install.sh`
 3. Перезагрузите правила udev: \
    `udevadm control --reload-rules`
-4. Включите сервис vitastor-host: \
-   `systemctl enable --now vitastor-host`
 
 После этого вы можете возвращаться к разделу [Быстрый старт](../intro/quickstart.ru.md).
 
-## Podman
-
-Если вы используете Podman, перед установкой контейнеров Vitastor выполните следующие
-команды от имени суперпользователя:
-
-```
-ln -s podman /usr/bin/docker
-
-mkdir -p /etc/systemd/system/systemd-udevd.service.d
-
-cat >/etc/systemd/system/systemd-udevd.service.d/override.conf <<EOF
-[Service]
-CapabilityBoundingSet=~
-SystemCallFilter=@mount capset
-EOF
-
-systemctl daemon-reload
-
-systemctl restart systemd-udevd
-```
-
-Без этих настроек udev не может делать вызовы внутрь Podman-контейнеров и определение дисков Vitastor не работает.
-
 ## Обновление контейнеров
 
 Сначала обязательно проверьте раздел [Обновление Vitastor](../usage/admin.ru.md#обновление-vitastor),

docs/installation/packages.en.md

@@ -33,17 +33,15 @@
   - CentOS 7: `yum install https://vitastor.io/rpms/centos/7/vitastor-release.rpm`
   - CentOS 8: `dnf install https://vitastor.io/rpms/centos/8/vitastor-release.rpm`
   - AlmaLinux 9 and other RHEL 9 clones (Rocky, Oracle...): `dnf install https://vitastor.io/rpms/centos/9/vitastor-release.rpm`
-  - AlmaLinux 10 and other RHEL 10 clones: `dnf install https://vitastor.io/rpms/centos/10/vitastor-release.rpm`
 - Enable EPEL: `yum/dnf install epel-release`
 - Enable additional CentOS repositories:
   - CentOS 7: `yum install centos-release-scl`
   - CentOS 8: `dnf install centos-release-advanced-virtualization`
-  - RHEL 9/10 clones: not required
+  - RHEL 9 clones: not required
 - Enable elrepo-kernel:
   - CentOS 7: `yum install https://www.elrepo.org/elrepo-release-7.el7.elrepo.noarch.rpm`
   - CentOS 8: `dnf install https://www.elrepo.org/elrepo-release-8.el8.elrepo.noarch.rpm`
   - RHEL 9 clones: `dnf install https://www.elrepo.org/elrepo-release-9.el9.elrepo.noarch.rpm`
-  - RHEL 10 clones: not required
 - Install packages: `yum/dnf install vitastor lpsolve etcd kernel-ml qemu-kvm`
 
 ## Installation requirements

docs/installation/packages.ru.md

@@ -33,17 +33,15 @@
   - CentOS 7: `yum install https://vitastor.io/rpms/centos/7/vitastor-release.rpm`
   - CentOS 8: `dnf install https://vitastor.io/rpms/centos/8/vitastor-release.rpm`
   - AlmaLinux 9 и другие клоны RHEL 9 (Rocky, Oracle...): `dnf install https://vitastor.io/rpms/centos/9/vitastor-release.rpm`
-  - AlmaLinux 10 и другие клоны RHEL 10: `dnf install https://vitastor.io/rpms/centos/10/vitastor-release.rpm`
 - Включите EPEL: `yum/dnf install epel-release`
 - Включите дополнительные репозитории CentOS:
   - CentOS 7: `yum install centos-release-scl`
   - CentOS 8: `dnf install centos-release-advanced-virtualization`
-  - Клоны RHEL 9/10: не нужно
+  - Клоны RHEL 9: не нужно
 - Включите elrepo-kernel:
   - CentOS 7: `yum install https://www.elrepo.org/elrepo-release-7.el7.elrepo.noarch.rpm`
   - CentOS 8: `dnf install https://www.elrepo.org/elrepo-release-8.el8.elrepo.noarch.rpm`
   - Клоны RHEL 9: `dnf install https://www.elrepo.org/elrepo-release-9.el9.elrepo.noarch.rpm`
-  - Клоны RHEL 10: не нужно
 - Установите пакеты: `yum/dnf install vitastor lpsolve etcd kernel-ml qemu-kvm`
 
 ## Установочные требования

mon/antietcd_adapter.js

@@ -18,7 +18,7 @@ class AntiEtcdAdapter
                 cluster = cluster ? (''+(cluster||'')).split(/,+/) : [];
             cluster = Object.keys(cluster.reduce((a, url) =>
             {
-                a[url.toLowerCase().replace(/^(https?:\/\/)/, '').replace(/\/.*$/, '')] = true;
+                a[url.toLowerCase().replace(/^(https?:\/\/)?(.*?)(\/.*)?$/, (m, m1, m2) => (m1||'http://')+m2)] = true;
                 return a;
             }, {}));
             const cfg_port = config.antietcd_port;
@@ -26,7 +26,8 @@ class AntiEtcdAdapter
             is_local['0.0.0.0'] = true;
             is_local['::'] = true;
             is_local[''] = true;
-            const selected = cluster.map(s => s.split(':', 2)).filter(ip => is_local[ip[0]] && (!cfg_port || ip[1] == cfg_port));
+            // split :, 3 -> <schema>:<//ip>:<port>
+            const selected = cluster.map(s => s.split(':', 3)).filter(ip => is_local[ip[1].substr(2)] && (!cfg_port || ip[2] == cfg_port));
             if (selected.length > 1)
             {
                 console.error('More than 1 etcd_address matches local IPs, please specify port');
@@ -35,12 +36,15 @@ class AntiEtcdAdapter
             else if (selected.length == 1)
             {
                 const antietcd_config = {
-                    ip: selected[0][0],
-                    port: selected[0][1],
-                    data: config.antietcd_data_file || ((config.antietcd_data_dir || '/var/lib/vitastor') + '/mon_'+selected[0][1]+'.json.gz'),
+                    ip: selected[0][1].substr(2),
+                    port: selected[0][2],
+                    cert: config.antietcd_cert,
+                    key: config.antietcd_key,
+                    ca: config.etcd_ca,
+                    data: config.antietcd_data_file || ((config.antietcd_data_dir || '/var/lib/vitastor') + '/mon_'+selected[0][2]+'.json.gz'),
                     persist_filter: vitastor_persist_filter({ vitastor_prefix: config.etcd_prefix || '/vitastor' }),
-                    node_id: selected[0][0]+':'+selected[0][1], // node_id = ip:port
-                    cluster: (cluster.length == 1 ? null : cluster.reduce((a, c) => { a[c] = "http://"+c; return a; }, {})),
+                    node_id: selected[0][1].substr(2)+':'+selected[0][2], // node_id = ip:port
+                    cluster: (cluster.length == 1 ? null : cluster.reduce((a, c) => { a[c.replace(/^(https?:\/\/)/, '')] = c; return a; }, {})),
                     cluster_key: (config.etcd_prefix || '/vitastor'),
                     stale_read: 1,
                     log_level: 1,

mon/etcd_adapter.js

@@ -1,7 +1,9 @@
 // Copyright (c) Vitaliy Filippov, 2019+
 // License: VNPL-1.1 (see README.md for details)
 
+const fs = require('fs');
 const http = require('http');
+const https = require('https');
 const WebSocket = require('ws');
 const { b64, local_ips } = require('./utils.js');
 
@@ -15,11 +17,18 @@ class EtcdAdapter
         this.ws = null;
         this.ws_alive = false;
         this.ws_keepalive_timer = null;
+        this.opts = {};
     }
 
     parse_config(config)
     {
         this.parse_etcd_addresses(config.etcd_address||config.etcd_url);
+        if (config.etcd_client_cert)
+            this.opts.cert = fs.readFileSync(config.etcd_client_cert, { encoding: 'utf-8' });
+        if (config.etcd_client_key)
+            this.opts.key = fs.readFileSync(config.etcd_client_key, { encoding: 'utf-8' });
+        if (config.etcd_ca)
+            this.opts.ca = fs.readFileSync(config.etcd_ca, { encoding: 'utf-8' });
     }
 
     parse_etcd_addresses(addrs)
@@ -39,7 +48,7 @@ class EtcdAdapter
         for (let url of addrs)
         {
             let scheme = 'http';
-            url = url.trim().replace(/^(https?):\/\//, (m, m1) => { scheme = m1; return ''; });
+            url = url.trim().replace(/^(https?):\/\//i, (m, m1) => { scheme = m1.toLowerCase(); return ''; });
             const slash = url.indexOf('/');
             const colon = url.indexOf(':');
             const is_local = is_local_ip[colon >= 0 ? url.substr(0, colon) : (slash >= 0 ? url.substr(0, slash) : url)];
@@ -130,7 +139,7 @@ class EtcdAdapter
                     }
                     ok(false);
                 }, this.mon.config.etcd_mon_timeout);
-                this.ws = new WebSocket(base+'/watch');
+                this.ws = new WebSocket(base+'/watch', this.opts);
                 this.ws_used_url = cur_addr;
                 const fail = () =>
                 {
@@ -272,7 +281,7 @@ class EtcdAdapter
             {
                 throw new Error(MON_STOPPED);
             }
-            const res = await POST(base+path, body, timeout);
+            const res = await POST(base+path, body, timeout, this.opts);
             if (this.mon.stopped)
             {
                 throw new Error(MON_STOPPED);
@@ -298,7 +307,7 @@ class EtcdAdapter
     }
 }
 
-function POST(url, body, timeout)
+function POST(url, body, timeout, opts)
 {
     return new Promise(ok =>
     {
@@ -310,10 +319,10 @@ function POST(url, body, timeout)
             req = null;
             ok({ error: 'timeout' });
         }, timeout) : null;
-        let req = http.request(url, { method: 'POST', headers: {
+        let req = (url.substr(0, 5) == 'https' ? https : http).request(url, { method: 'POST', headers: {
             'Content-Type': 'application/json',
             'Content-Length': body_text.length,
-        } }, (res) =>
+        }, ...(opts||{}) }, (res) =>
         {
             if (!req)
             {

mon/etcd_schema.js

@@ -45,7 +45,10 @@ const etcd_tree = {
             config_path: "/etc/vitastor/vitastor.conf",
             etcd_prefix: "/vitastor",
             // etcd connection - configurable online
-            etcd_address: "10.0.115.10:2379/v3",
+            etcd_address: "http://10.0.115.10:2379/v3",
+            etcd_client_cert: "",
+            etcd_client_key: "",
+            etcd_ca: "",
             // mon
             etcd_mon_ttl: 5, // min: 1
             etcd_mon_timeout: 1000, // ms. min: 0

mon/http_server.js

@@ -16,7 +16,7 @@ async function create_http_server(cfg, handler)
         };
         if (cfg.mon_https_ca)
         {
-            tls.ca = await fsp.readFile(cfg.mon_https_ca);
+            tls.mon_https_ca = await fsp.readFile(cfg.mon_https_ca);
         }
         if (cfg.mon_https_client_auth)
         {

mon/package.json

@@ -1,6 +1,6 @@
 {
   "name": "vitastor-mon",
-  "version": "3.0.6",
+  "version": "3.0.3",
   "description": "Vitastor SDS monitor service",
   "main": "mon-main.js",
   "scripts": {

node-binding/package.json

@@ -1,6 +1,6 @@
 {
   "name": "vitastor",
-  "version": "3.0.6",
+  "version": "3.0.3",
   "description": "Low-level native bindings to Vitastor client library",
   "main": "index.js",
   "keywords": [

patches/cinder-vitastor.py

@@ -50,7 +50,7 @@ from cinder.volume import configuration
 from cinder.volume import driver
 from cinder.volume import volume_utils
 
-VITASTOR_VERSION = '3.0.6'
+VITASTOR_VERSION = '3.0.3'
 
 LOG = logging.getLogger(__name__)
 
@@ -275,7 +275,7 @@ class VitastorDriver(driver.CloneableImageVD,
             LOG.exception('error getting vitastor pool stats: '+str(e))
 
         self._stats = stats
-
+        
     def get_volume_stats(self, refresh=False):
         """Get volume stats.
         If 'refresh' is True, run update the stats first.
@@ -291,14 +291,6 @@ class VitastorDriver(driver.CloneableImageVD,
         else:
             return (1 + resp['kvs'][0]['value'], resp['kvs'][0]['mod_revision'])
 
-    def _cli(self, descr, *args):
-        args = [ 'vitastor-cli', *args, *(self._vitastor_args()) ]
-        try:
-            self._execute(*args)
-        except processutils.ProcessExecutionError as exc:
-            LOG.error("Failed to "+descr+": "+exc)
-            raise exception.VolumeBackendAPIException(data = exc.stderr)
-
     def create_volume(self, volume):
         """Creates a logical volume."""
 
@@ -310,7 +302,7 @@ class VitastorDriver(driver.CloneableImageVD,
 
         LOG.debug("creating volume '%s'", vol_name)
 
-        self._cli('create volume', 'create', vol_name, '--size', size)
+        self._create_image(vol_name, { 'size': size })
 
         if volume.encryption_key_id:
             self._create_encrypted_volume(volume, volume.obj_context)
@@ -354,7 +346,7 @@ class VitastorDriver(driver.CloneableImageVD,
         snap_name = utils.convert_str(snapshot.name)
         if snap_name.find('@') >= 0 or snap_name.find('/') >= 0:
             raise exception.VolumeBackendAPIException(data = '@ and / are forbidden in volume and snapshot names')
-        self._cli('create snapshot', 'snap-create', vol_name+'@'+snap_name)
+        self._create_snapshot(vol_name, vol_name+'@'+snap_name)
 
     def snapshot_revert_use_temp_snapshot(self):
         """Disable the use of a temporary snapshot on revert."""
@@ -367,8 +359,21 @@ class VitastorDriver(driver.CloneableImageVD,
         snap_name = utils.convert_str(snapshot.name)
 
         # Delete the image and recreate it from the snapshot
-        self._cli('delete image', 'rm', vol_name)
-        self._cli('recreate image', 'create', '--parent', vol_name+'@'+snap_name, vol_name)
+        args = [ 'vitastor-cli', 'rm', vol_name, *(self._vitastor_args()) ]
+        try:
+            self._execute(*args)
+        except processutils.ProcessExecutionError as exc:
+            LOG.error("Failed to delete image "+vol_name+": "+exc)
+            raise exception.VolumeBackendAPIException(data = exc.stderr)
+        args = [
+            'vitastor-cli', 'create', '--parent', vol_name+'@'+snap_name,
+            vol_name, *(self._vitastor_args())
+        ]
+        try:
+            self._execute(*args)
+        except processutils.ProcessExecutionError as exc:
+            LOG.error("Failed to recreate image "+vol_name+" from "+vol_name+"@"+snap_name+": "+exc)
+            raise exception.VolumeBackendAPIException(data = exc.stderr)
 
     def delete_snapshot(self, snapshot):
         """Deletes a snapshot."""
@@ -376,7 +381,15 @@ class VitastorDriver(driver.CloneableImageVD,
         vol_name = utils.convert_str(snapshot.volume_name)
         snap_name = utils.convert_str(snapshot.name)
 
-        self._cli('remove snapshot', 'rm', vol_name+'@'+snap_name)
+        args = [
+            'vitastor-cli', 'rm', vol_name+'@'+snap_name,
+            *(self._vitastor_args())
+        ]
+        try:
+            self._execute(*args)
+        except processutils.ProcessExecutionError as exc:
+            LOG.error("Failed to remove snapshot "+vol_name+'@'+snap_name+": "+exc)
+            raise exception.VolumeBackendAPIException(data = exc.stderr)
 
     def _child_count(self, parents):
         children = 0
@@ -414,7 +427,13 @@ class VitastorDriver(driver.CloneableImageVD,
         if src_vref.admin_metadata.get('readonly') == 'True':
             # source volume is a volume-image cache entry or other readonly volume
             # clone without intermediate snapshot
-            self._cli('create clone', 'create', '--parent', src_name, '--size', size, dest_name)
+            src = self._get_image(src_name)
+            LOG.debug("creating image '%s' from '%s'", dest_name, src_name)
+            new_cfg = self._create_image(dest_name, {
+                'size': size,
+                'parent_id': src['idx']['id'],
+                'parent_pool_id': src['idx']['pool_id'],
+            })
             return {}
 
         clone_snap = "%s@%s.clone_snap" % (src_name, dest_name)
@@ -427,12 +446,15 @@ class VitastorDriver(driver.CloneableImageVD,
             clone_snap = dest_name
             make_img = False
 
-        LOG.debug("creating snapshot '%s'", clone_snap)
-        self._cli('create base snapshot', 'snap-create', '--allow-existing', '1', clone_snap)
-
+        LOG.debug("creating layer '%s' under '%s'", clone_snap, src_name)
+        new_cfg = self._create_snapshot(src_name, clone_snap, True)
         if make_img:
             # Then create a clone from it
-            self._cli('create clone', 'create', '--parent', clone_snap, '--size', size, dest_name)
+            new_cfg = self._create_image(dest_name, {
+                'size': size,
+                'parent_id': new_cfg['parent_id'],
+                'parent_pool_id': new_cfg['parent_pool_id'],
+            })
 
         return {}
 
@@ -442,8 +464,7 @@ class VitastorDriver(driver.CloneableImageVD,
         vol_name = utils.convert_str(volume.name)
         snap_name = utils.convert_str(snapshot.name)
 
-        src_snap = 'volume-'+snapshot.volume_id+'@'+snap_name
-        snap = self._get_image(src_snap)
+        snap = self._get_image('volume-'+snapshot.volume_id+'@'+snap_name)
         if not snap:
             raise exception.SnapshotNotFound(snapshot_id = snap_name)
         snap_inode_id = int(resp['responses'][0]['kvs'][0]['value']['id'])
@@ -452,8 +473,12 @@ class VitastorDriver(driver.CloneableImageVD,
         size = snap['cfg']['size']
         if int(volume.size):
             size = int(volume.size) * units.Gi
+        new_cfg = self._create_image(vol_name, {
+            'size': size,
+            'parent_id': snap['idx']['id'],
+            'parent_pool_id': snap['idx']['pool_id'],
+        })
 
-        self._cli('create clone', 'create', vol_name, '--size', size, '--parent', src_snap)
         return {}
 
     def _vitastor_args(self):
@@ -480,7 +505,49 @@ class VitastorDriver(driver.CloneableImageVD,
         """Deletes a logical volume."""
 
         vol_name = utils.convert_str(volume.name)
-        self._cli('delete volume', 'rm', '--matching', vol_name, vol_name+'@*', '--progress', '0')
+
+        # Find the volume and all its snapshots
+        range_end = b'index/image/' + vol_name.encode('utf-8')
+        range_end = range_end[0 : len(range_end)-1] + six.int2byte(range_end[len(range_end)-1] + 1)
+        resp = self._etcd_txn({ 'success': [
+            { 'request_range': { 'key': 'index/image/'+vol_name, 'range_end': range_end } },
+        ] })
+        if len(resp['responses'][0]['kvs']) == 0:
+            # already deleted
+            LOG.info("volume %s no longer exists in backend", vol_name)
+            return
+        layers = resp['responses'][0]['kvs']
+        layer_ids = {}
+        for kv in layers:
+            inode_id = int(kv['value']['id'])
+            pool_id = int(kv['value']['pool_id'])
+            inode_pool_id = (pool_id << 48) | (inode_id & 0xffffffffffff)
+            layer_ids[inode_pool_id] = True
+
+        # Check if the volume has clones and raise 'busy' if so
+        children = self._child_count(layer_ids)
+        if children > 0:
+            raise exception.VolumeIsBusy(volume_name = vol_name)
+
+        # Clear data
+        for kv in layers:
+            args = [
+                'vitastor-cli', 'rm-data', '--pool', str(kv['value']['pool_id']),
+                '--inode', str(kv['value']['id']), '--progress', '0',
+                *(self._vitastor_args())
+            ]
+            try:
+                self._execute(*args)
+            except processutils.ProcessExecutionError as exc:
+                LOG.error("Failed to remove layer "+kv['key']+": "+exc)
+                raise exception.VolumeBackendAPIException(data = exc.stderr)
+
+        # Delete all layers from etcd
+        requests = []
+        for kv in layers:
+            requests.append({ 'request_delete_range': { 'key': kv['key'] } })
+            requests.append({ 'request_delete_range': { 'key': 'config/inode/'+str(kv['value']['pool_id'])+'/'+str(kv['value']['id']) } })
+        self._etcd_txn({ 'success': requests })
 
     def retype(self, context, volume, new_type, diff, host):
         """Change extra type specifications for a volume."""
@@ -500,6 +567,98 @@ class VitastorDriver(driver.CloneableImageVD,
         """Removes an export for a logical volume."""
         pass
 
+    def _create_image(self, vol_name, cfg):
+        pool_s = str(self.cfg['pool_id'])
+        image_id = 0
+        while image_id == 0:
+            # check if the image already exists and find a free ID
+            resp = self._etcd_txn({ 'success': [
+                { 'request_range': { 'key': 'index/image/'+vol_name } },
+                { 'request_range': { 'key': 'index/maxid/'+pool_s } },
+            ] })
+            if len(resp['responses'][0]['kvs']) > 0:
+                # already exists
+                raise exception.VolumeBackendAPIException(data = 'Volume '+vol_name+' already exists')
+            image_id, id_mod = self._next_id(resp['responses'][1])
+            # try to create the image
+            resp = self._etcd_txn({ 'compare': [
+                { 'target': 'MOD', 'mod_revision': id_mod, 'key': 'index/maxid/'+pool_s },
+                { 'target': 'VERSION', 'version': 0, 'key': 'index/image/'+vol_name },
+                { 'target': 'VERSION', 'version': 0, 'key': 'config/inode/'+pool_s+'/'+str(image_id) },
+            ], 'success': [
+                { 'request_put': { 'key': 'index/maxid/'+pool_s, 'value': image_id } },
+                { 'request_put': { 'key': 'index/image/'+vol_name, 'value': json.dumps({
+                    'id': image_id, 'pool_id': self.cfg['pool_id']
+                }) } },
+                { 'request_put': { 'key': 'config/inode/'+pool_s+'/'+str(image_id), 'value': json.dumps({
+                    **cfg, 'name': vol_name,
+                }) } },
+            ] })
+            if not resp.get('succeeded'):
+                # repeat
+                image_id = 0
+
+    def _create_snapshot(self, vol_name, snap_vol_name, allow_existing = False):
+        while True:
+            # check if the image already exists and snapshot doesn't
+            resp = self._etcd_txn({ 'success': [
+                { 'request_range': { 'key': 'index/image/'+vol_name } },
+                { 'request_range': { 'key': 'index/image/'+snap_vol_name } },
+            ] })
+            if len(resp['responses'][0]['kvs']) == 0:
+                raise exception.VolumeBackendAPIException(data = 'Volume '+vol_name+' does not exist')
+            if len(resp['responses'][1]['kvs']) > 0:
+                if allow_existing:
+                    snap_idx = resp['responses'][1]['kvs'][0]['value']
+                    resp = self._etcd_txn({ 'success': [
+                        { 'request_range': { 'key': 'config/inode/'+str(snap_idx['pool_id'])+'/'+str(snap_idx['id']) } },
+                    ] })
+                    if len(resp['responses'][0]['kvs']) == 0:
+                        raise exception.VolumeBackendAPIException(data =
+                            'Volume '+snap_vol_name+' is already indexed, but does not exist'
+                        )
+                    return resp['responses'][0]['kvs'][0]['value']
+                raise exception.VolumeBackendAPIException(
+                    data = 'Volume '+snap_vol_name+' already exists'
+                )
+            vol_idx = resp['responses'][0]['kvs'][0]['value']
+            vol_idx_mod = resp['responses'][0]['kvs'][0]['mod_revision']
+            # get image inode config and find a new ID
+            resp = self._etcd_txn({ 'success': [
+                { 'request_range': { 'key': 'config/inode/'+str(vol_idx['pool_id'])+'/'+str(vol_idx['id']) } },
+                { 'request_range': { 'key': 'index/maxid/'+str(self.cfg['pool_id']) } },
+            ] })
+            if len(resp['responses'][0]['kvs']) == 0:
+                raise exception.VolumeBackendAPIException(data = 'Volume '+vol_name+' does not exist')
+            vol_cfg = resp['responses'][0]['kvs'][0]['value']
+            vol_mod = resp['responses'][0]['kvs'][0]['mod_revision']
+            new_id, id_mod = self._next_id(resp['responses'][1])
+            # try to redirect image to the new inode
+            new_cfg = {
+                **vol_cfg, 'name': vol_name, 'parent_id': vol_idx['id'], 'parent_pool_id': vol_idx['pool_id']
+            }
+            resp = self._etcd_txn({ 'compare': [
+                { 'target': 'MOD', 'mod_revision': vol_idx_mod, 'key': 'index/image/'+vol_name },
+                { 'target': 'MOD', 'mod_revision': vol_mod, 'key': 'config/inode/'+str(vol_idx['pool_id'])+'/'+str(vol_idx['id']) },
+                { 'target': 'MOD', 'mod_revision': id_mod, 'key': 'index/maxid/'+str(self.cfg['pool_id']) },
+                { 'target': 'VERSION', 'version': 0, 'key': 'index/image/'+snap_vol_name },
+                { 'target': 'VERSION', 'version': 0, 'key': 'config/inode/'+str(self.cfg['pool_id'])+'/'+str(new_id) },
+            ], 'success': [
+                { 'request_put': { 'key': 'index/maxid/'+str(self.cfg['pool_id']), 'value': new_id } },
+                { 'request_put': { 'key': 'index/image/'+vol_name, 'value': json.dumps({
+                    'id': new_id, 'pool_id': self.cfg['pool_id']
+                }) } },
+                { 'request_put': { 'key': 'config/inode/'+str(self.cfg['pool_id'])+'/'+str(new_id), 'value': json.dumps(new_cfg) } },
+                { 'request_put': { 'key': 'index/image/'+snap_vol_name, 'value': json.dumps({
+                    'id': vol_idx['id'], 'pool_id': vol_idx['pool_id']
+                }) } },
+                { 'request_put': { 'key': 'config/inode/'+str(vol_idx['pool_id'])+'/'+str(vol_idx['id']), 'value': json.dumps({
+                    **vol_cfg, 'name': snap_vol_name, 'readonly': True
+                }) } }
+            ] })
+            if resp.get('succeeded'):
+                return new_cfg
+
     def initialize_connection(self, volume, connector):
         data = {
             'driver_volume_type': 'vitastor',
@@ -538,9 +697,13 @@ class VitastorDriver(driver.CloneableImageVD,
                     size = int(volume.size) * units.Gi
                     dest_name = utils.convert_str(volume.name)
                     # Find or create the base snapshot
-                    self._cli('create base snapshot', 'create', '--allow-existing', '1', base_vol.name+'@.clone_snap')
+                    snap_cfg = self._create_snapshot(base_vol.name, base_vol.name+'@.clone_snap', True)
                     # Then create a clone from it
-                    self._cli('create clone', 'create', dest_name, '--size', size, '--parent', base_vol.name+'@.clone_snap')
+                    new_cfg = self._create_image(dest_name, {
+                        'size': size,
+                        'parent_id': snap_cfg['parent_id'],
+                        'parent_pool_id': snap_cfg['parent_pool_id'],
+                    })
                     return ({}, True)
         return ({}, False)
 
@@ -607,8 +770,26 @@ class VitastorDriver(driver.CloneableImageVD,
     def extend_volume(self, volume, new_size):
         """Extend an existing volume."""
         vol_name = utils.convert_str(volume.name)
-        size = int(new_size) * units.Gi
-        self._cli('extend volume', 'modify', vol_name, '--resize', new_size)
+        while True:
+            vol = self._get_image(vol_name)
+            if not vol:
+                raise exception.VolumeBackendAPIException(data = 'Volume '+vol_name+' does not exist')
+            # change size
+            size = int(new_size) * units.Gi
+            if size == vol['cfg']['size']:
+                break
+            resp = self._etcd_txn({ 'compare': [ {
+                'target': 'MOD',
+                'mod_revision': vol['cfg_mod'],
+                'key': 'config/inode/'+str(vol['idx']['pool_id'])+'/'+str(vol['idx']['id']),
+            } ], 'success': [
+                { 'request_put': {
+                    'key': 'config/inode/'+str(vol['idx']['pool_id'])+'/'+str(vol['idx']['id']),
+                    'value': json.dumps({ **vol['cfg'], 'size': size }),
+                } },
+            ] })
+            if resp.get('succeeded'):
+                break
         LOG.debug(
             "Extend volume from %(old_size)s GB to %(new_size)s GB.",
             {'old_size': volume.size, 'new_size': new_size}
@@ -681,7 +862,28 @@ class VitastorDriver(driver.CloneableImageVD,
         """
         from_name = self._get_existing_name(existing_ref)
         to_name = utils.convert_str(volume.name)
-        self._cli('rename', 'modify', from_name, '--rename', to_name)
+        self._rename(from_name, to_name)
+
+    def _rename(self, from_name, to_name):
+        while True:
+            vol = self._get_image(from_name)
+            if not vol:
+                raise exception.VolumeBackendAPIException(data = 'Volume '+from_name+' does not exist')
+            to = self._get_image(to_name)
+            if to:
+                raise exception.VolumeBackendAPIException(data = 'Volume '+to_name+' already exists')
+            resp = self._etcd_txn({ 'compare': [
+                { 'target': 'MOD', 'mod_revision': vol['idx_mod'], 'key': 'index/image/'+vol['cfg']['name'] },
+                { 'target': 'MOD', 'mod_revision': vol['cfg_mod'], 'key': 'config/inode/'+str(vol['idx']['pool_id'])+'/'+str(vol['idx']['id']) },
+                { 'target': 'VERSION', 'version': 0, 'key': 'index/image/'+to_name },
+            ], 'success': [
+                { 'request_delete_range': { 'key': 'index/image/'+vol['cfg']['name'] } },
+                { 'request_put': { 'key': 'index/image/'+to_name, 'value': json.dumps(vol['idx']) } },
+                { 'request_put': { 'key': 'config/inode/'+str(vol['idx']['pool_id'])+'/'+str(vol['idx']['id']),
+                    'value': json.dumps({ **vol['cfg'], 'name': to_name }) } },
+            ] })
+            if resp.get('succeeded'):
+                break
 
     def unmanage(self, volume):
         pass
@@ -754,7 +956,7 @@ class VitastorDriver(driver.CloneableImageVD,
         snap_name = self._get_existing_name(existing_ref)
         from_name = vol_name+'@'+snap_name
         to_name = vol_name+'@'+utils.convert_str(snapshot.name)
-        self._cli('rename', 'modify', from_name, '--rename', to_name)
+        self._rename(from_name, to_name)
 
     def unmanage_snapshot(self, snapshot):
         """Removes the specified snapshot from Cinder management."""

rpm/vitastor-el10.spec

@@ -1,11 +1,11 @@
 Name:           vitastor
-Version:        3.0.6
+Version:        3.0.3
 Release:        1%{?dist}
 Summary:        Vitastor, a fast software-defined clustered block storage
 
 License:        Vitastor Network Public License 1.1
 URL:            https://vitastor.io/
-Source0:        vitastor-3.0.6.el10.tar.gz
+Source0:        vitastor-3.0.3.el10.tar.gz
 
 BuildRequires:  gperftools-devel
 BuildRequires:  gcc-c++

rpm/vitastor-el7.spec

@@ -1,11 +1,11 @@
 Name:           vitastor
-Version:        3.0.6
+Version:        3.0.3
 Release:        1%{?dist}
 Summary:        Vitastor, a fast software-defined clustered block storage
 
 License:        Vitastor Network Public License 1.1
 URL:            https://vitastor.io/
-Source0:        vitastor-3.0.6.el7.tar.gz
+Source0:        vitastor-3.0.3.el7.tar.gz
 
 BuildRequires:  gperftools-devel
 BuildRequires:  devtoolset-9-gcc-c++

rpm/vitastor-el8.spec

@@ -1,11 +1,11 @@
 Name:           vitastor
-Version:        3.0.6
+Version:        3.0.3
 Release:        1%{?dist}
 Summary:        Vitastor, a fast software-defined clustered block storage
 
 License:        Vitastor Network Public License 1.1
 URL:            https://vitastor.io/
-Source0:        vitastor-3.0.6.el8.tar.gz
+Source0:        vitastor-3.0.3.el8.tar.gz
 
 BuildRequires:  gperftools-devel
 BuildRequires:  gcc-toolset-9-gcc-c++

rpm/vitastor-el9.spec

@@ -1,11 +1,11 @@
 Name:           vitastor
-Version:        3.0.6
+Version:        3.0.3
 Release:        1%{?dist}
 Summary:        Vitastor, a fast software-defined clustered block storage
 
 License:        Vitastor Network Public License 1.1
 URL:            https://vitastor.io/
-Source0:        vitastor-3.0.6.el9.tar.gz
+Source0:        vitastor-3.0.3.el9.tar.gz
 
 BuildRequires:  gperftools-devel
 BuildRequires:  gcc-c++

src/CMakeLists.txt

@@ -21,7 +21,7 @@ if("${CMAKE_INSTALL_PREFIX}" MATCHES "^/usr/local/?$")
 endif()
 set(ENABLE_COVERAGE false CACHE BOOL "Enable code coverage")
 
-add_definitions(-DVITASTOR_VERSION="3.0.6")
+add_definitions(-DVITASTOR_VERSION="3.0.3")
 add_definitions(-D_GNU_SOURCE -D_LARGEFILE64_SOURCE -D_FILE_OFFSET_BITS=64 -Wall -Wno-sign-compare -Wno-comment -Wno-parentheses -Wno-pointer-arith -fdiagnostics-color=always -fno-omit-frame-pointer -fvisibility=hidden -I ${CMAKE_SOURCE_DIR}/src)
 add_link_options(-fno-omit-frame-pointer)
 if (${WITH_ASAN})
@@ -75,6 +75,11 @@ if (RDMACM_LIBRARIES)
 	add_definitions(-DWITH_RDMACM)
 endif (RDMACM_LIBRARIES)
 
+find_package(OpenSSL)
+if (OPENSSL_FOUND)
+	add_definitions(-DWITH_OPENSSL)
+endif (OPENSSL_FOUND)
+
 if (${WITH_SYSTEM_LIBURING})
 	pkg_check_modules(LIBURING REQUIRED liburing>=2.10)
 	include_directories(${LIBURING_INCLUDE_DIRS})

src/blockstore/CMakeLists.txt

@@ -4,7 +4,7 @@ project(vitastor)
 
 # libvitastor_blk.a
 add_library(vitastor_blk STATIC
-	../util/allocator.cpp ../util/crc32c.c ../util/ringloop.cpp
+	../util/allocator.cpp ../util/crc32c.c ../util/xxhash.c ../util/ringloop.cpp
 	multilist.cpp blockstore_heap.cpp blockstore_disk.cpp
 	blockstore.cpp blockstore_impl.cpp blockstore_init.cpp blockstore_open.cpp
 	blockstore_flush.cpp blockstore_read.cpp blockstore_stable.cpp blockstore_sync.cpp blockstore_write.cpp

src/blockstore/blockstore_disk.cpp

@@ -83,13 +83,17 @@ void blockstore_disk_t::parse_config(std::map<std::string, std::string> & config
     {
         data_csum_type = BLOCKSTORE_CSUM_CRC32C;
     }
+    else if (config["data_csum_type"] == "xxh3_32")
+    {
+        data_csum_type = BLOCKSTORE_CSUM_XXH3_32;
+    }
     else if (config["data_csum_type"] == "" || config["data_csum_type"] == "none")
     {
         data_csum_type = BLOCKSTORE_CSUM_NONE;
     }
     else
     {
-        throw std::runtime_error("data_csum_type="+config["data_csum_type"]+" is unsupported, only \"crc32c\" and \"none\" are supported");
+        throw std::runtime_error("data_csum_type="+config["data_csum_type"]+" is unsupported, only \"crc32c\", \"xxh3_32\" and \"none\" are supported");
     }
     csum_block_size = parse_size(config["csum_block_size"]);
     discard_on_start = config.find("discard_on_start") != config.end() &&
@@ -173,7 +177,9 @@ void blockstore_disk_t::parse_config(std::map<std::string, std::string> & config
     }
     if (data_block_size / bitmap_granularity < 8)
     {
-        throw std::runtime_error("Data block size must be at least bitmap_granularity*8");
+        fprintf(stderr, "Warning: block_size (%u) / bitmap_granularity (%u) = %u bits. "
+            "Consider using larger block_size or bitmap_granularity for better performance.\n",
+            data_block_size, bitmap_granularity, data_block_size / bitmap_granularity);
     }
     if (!data_csum_type)
     {

src/blockstore/blockstore_disk.h

@@ -16,6 +16,7 @@
 #define BLOCKSTORE_CSUM_NONE 0
 // Lower byte of checksum type is its length
 #define BLOCKSTORE_CSUM_CRC32C 0x104
+#define BLOCKSTORE_CSUM_XXH3_32 0x204
 
 #define MOCK_DATA_FD 1000
 #define MOCK_META_FD 1001
@@ -26,14 +27,14 @@ class allocator_t;
 struct blockstore_disk_t
 {
     std::string data_device, meta_device, journal_device;
-    uint64_t data_block_size;
+    uint32_t data_block_size;
     uint64_t cfg_journal_size, cfg_data_size;
     // Required write alignment and journal/metadata/data areas' location alignment
     uint32_t disk_alignment = 4096;
     // Journal block size - minimum_io_size of the journal device is the best choice
-    uint64_t journal_block_size = 4096;
+    uint32_t journal_block_size = 4096;
     // Metadata block size - minimum_io_size of the metadata device is the best choice
-    uint64_t meta_block_size = 4096;
+    uint32_t meta_block_size = 4096;
     // Atomic write size of the data block device
     uint32_t atomic_write_size = 4096;
     // Whether we should set RWF_ATOMIC on atomic writes

src/blockstore/blockstore_heap.cpp

@@ -12,6 +12,7 @@
 #include "blockstore_heap.h"
 #include "../util/allocator.h"
 #include "../util/crc32c.h"
+#include "../util/xxhash.h"
 #include "../util/malloc_or_die.h"
 
 #define BS_HEAP_FREE_MVCC 1
@@ -62,19 +63,19 @@ uint32_t blockstore_heap_t::get_simple_entry_size()
 uint32_t blockstore_heap_t::get_big_entry_size()
 {
     return sizeof(heap_big_write_t) + dsk->clean_entry_bitmap_size*2 +
-        (!dsk->data_csum_type ? 0 : dsk->data_block_size/dsk->csum_block_size * (dsk->data_csum_type & 0xFF));
+        (!dsk->csum_block_size ? 0 : dsk->data_block_size/dsk->csum_block_size * (dsk->data_csum_type & 0xFF));
 }
 
 uint32_t blockstore_heap_t::get_big_intent_entry_size()
 {
     return sizeof(heap_big_intent_t) + dsk->clean_entry_bitmap_size*2 +
-        (!dsk->data_csum_type ? 4 : dsk->data_block_size/dsk->csum_block_size * (dsk->data_csum_type & 0xFF));
+        (!dsk->csum_block_size ? 4 : dsk->data_block_size/dsk->csum_block_size * (dsk->data_csum_type & 0xFF));
 }
 
 uint32_t blockstore_heap_t::get_small_entry_size(uint32_t offset, uint32_t len)
 {
     return sizeof(heap_small_write_t) + dsk->clean_entry_bitmap_size +
-        (!dsk->data_csum_type ? 4 : (dsk->data_csum_type & 0xFF) *
+        (!dsk->csum_block_size ? 4 : (dsk->data_csum_type & 0xFF) *
             ((offset+len+dsk->csum_block_size-1)/dsk->csum_block_size - offset/dsk->csum_block_size));
 }
 
@@ -89,7 +90,7 @@ uint32_t blockstore_heap_t::get_csum_size(heap_entry_t *wr)
 
 uint32_t blockstore_heap_t::get_csum_size(uint32_t entry_type, uint32_t offset, uint32_t len)
 {
-    if (!dsk->data_csum_type)
+    if (!dsk->csum_block_size)
     {
         return 0;
     }
@@ -212,15 +213,24 @@ void heap_entry_t::set_big_location(blockstore_heap_t *heap, uint64_t location)
     big().block_num = location / heap->dsk->data_block_size;
 }
 
-uint32_t heap_entry_t::calc_crc32c()
+uint32_t heap_entry_t::calc_checksum(blockstore_disk_t *dsk)
 {
-    auto old_crc32c = crc32c;
-    crc32c = 0;
-    uint32_t res = ::crc32c(0, (uint8_t*)this, size);
-    crc32c = old_crc32c;
+    auto old_checksum = checksum;
+    checksum = 0;
+    uint32_t res = 0;
+    if (dsk->data_csum_type == BLOCKSTORE_CSUM_XXH3_32)
+        res = (uint32_t)XXH3_64bits(this, size);
+    else
+        res = ::crc32c(0, (uint8_t*)this, size);
+    checksum = old_checksum;
     return res;
 }
 
+uint32_t heap_entry_t::calc_checksum(blockstore_heap_t *heap)
+{
+    return calc_checksum(heap->dsk);
+}
+
 uint64_t blockstore_heap_t::get_pg_id(inode_t inode, uint64_t stripe)
 {
     uint64_t pg_num = 0;
@@ -301,13 +311,12 @@ int blockstore_heap_t::read_blocks(uint64_t disk_offset, uint64_t disk_size, uin
             heap_entry_t *wr = (heap_entry_t*)data;
             if (wr->size > dsk->meta_block_size-block_offset)
             {
-                fprintf(stderr, "Error: entry is too large in metadata block %u at %u (%u > max %ju bytes). ",
+                fprintf(stderr, "Error: entry is too large in metadata block %u at %u (%u > max %u bytes). ",
                     block_num, block_offset, wr->size, dsk->meta_block_size-block_offset);
 corrupted_block:
                 if (allow_corrupted)
                 {
                     fprintf(stderr, "Metadata block is corrupted, skipping\n");
-                    recheck_modified_blocks.insert(block_num);
                     break;
                 }
                 else
@@ -358,7 +367,6 @@ corrupted_object:
                 if (allow_corrupted)
                 {
                     fprintf(stderr, "Entry is corrupted, skipping\n");
-                    recheck_modified_blocks.insert(block_num);
                     block_offset += wr->size;
                     continue;
                 }
@@ -385,12 +393,12 @@ corrupted_object:
                 goto corrupted_object;
             }
             // Verify crc
-            uint32_t expected_crc32c = wr->calc_crc32c();
-            if (wr->crc32c != expected_crc32c)
+            uint32_t expected_checksum = wr->calc_checksum(this);
+            if (wr->checksum != expected_checksum)
             {
-                fprintf(stderr, "Error: entry %jx:%jx v%ju l%ju in metadata block %u at %u is corrupt (crc32c mismatch: expected %08x, got %08x). ",
+                fprintf(stderr, "Error: entry %jx:%jx v%ju l%ju in metadata block %u at %u is corrupt (checksum mismatch: expected %08x, got %08x). ",
                     wr->inode, wr->stripe, wr->version, wr->lsn,
-                    block_num, block_offset, expected_crc32c, wr->crc32c);
+                    block_num, block_offset, expected_checksum, wr->checksum);
                 goto corrupted_object;
             }
             // Verify offset & len
@@ -878,7 +886,11 @@ bool blockstore_heap_t::calc_checksums(heap_entry_t *wr, uint8_t *data, bool set
             len = wr->big_intent().len;
         else
             assert(0);
-        uint32_t real_csum = crc32c(0, data, len);
+        uint32_t real_csum = 0;
+        if (dsk->data_csum_type == BLOCKSTORE_CSUM_XXH3_32)
+            real_csum = (uint32_t)XXH3_64bits(data, len);
+        else
+            real_csum = crc32c(0, data, len);
         if (set)
         {
             *wr_csum = real_csum;
@@ -928,11 +940,26 @@ static uint32_t crc32c_iter(uint32_t prev_crc, const std::function<uint8_t*(uint
     return prev_crc;
 }
 
+static void xxh3_iter(XXH3_state_t* xxh3_state, const std::function<uint8_t*(uint32_t start, uint32_t & len)> & next, uint32_t pos, uint32_t size)
+{
+    uint32_t cur_len = 0;
+    while (size > 0)
+    {
+        uint8_t *data = next(pos, cur_len);
+        assert(data);
+        cur_len = (cur_len < size ? cur_len : size);
+        XXH3_64bits_update(xxh3_state, data, cur_len);
+        pos += cur_len;
+        size -= cur_len;
+    }
+}
+
 bool blockstore_heap_t::calc_block_checksums(uint32_t *block_csums, uint8_t *bitmap,
     uint32_t start, uint32_t end, std::function<uint8_t*(uint32_t start, uint32_t & len)> next,
     bool set, std::function<void(uint32_t, uint32_t, uint32_t)> bad_block_cb)
 {
     bool res = true;
+    XXH3_state_t* xxh3_state = NULL;
     uint32_t pos = start;
     uint32_t block_end = (start/dsk->csum_block_size + 1)*dsk->csum_block_size;
     uint32_t block_crc = 0;
@@ -949,42 +976,89 @@ bool blockstore_heap_t::calc_block_checksums(uint32_t *block_csums, uint8_t *bit
                     pos += dsk->bitmap_granularity;
                 // zero padding at the beginning or at the end of the block is not counted
                 if (pos > prev && prev > 0 && pos < block_end)
-                    block_crc = crc32c_pad(block_crc, NULL, 0, pos-prev, 0);
+                {
+                    if (dsk->data_csum_type == BLOCKSTORE_CSUM_XXH3_32)
+                    {
+                        if (!xxh3_state)
+                        {
+                            xxh3_state = XXH3_createState();
+                            XXH3_64bits_reset(xxh3_state);
+                        }
+                        uint32_t zeropad = pos-prev;
+                        while (zeropad > 0)
+                        {
+                            uint32_t zerolen = zeropad > 4096 ? 4096 : zeropad;
+                            XXH3_64bits_update(xxh3_state, zero_page, zerolen);
+                            zeropad -= zerolen;
+                        }
+                    }
+                    else
+                        block_crc = crc32c_pad(block_crc, NULL, 0, pos-prev, 0);
+                }
                 prev = pos;
                 while (pos < end && pos < block_end && (bitmap[pos/dsk->bitmap_granularity/8] & (1 << ((pos/dsk->bitmap_granularity) % 8))))
                     pos += dsk->bitmap_granularity;
                 if (pos > prev)
                 {
                     isset = true;
-                    block_crc = crc32c_iter(block_crc, next, prev, pos-prev);
+                    if (dsk->data_csum_type == BLOCKSTORE_CSUM_XXH3_32)
+                    {
+                        if (!xxh3_state)
+                        {
+                            xxh3_state = XXH3_createState();
+                            XXH3_64bits_reset(xxh3_state);
+                        }
+                        xxh3_iter(xxh3_state, next, prev, pos-prev);
+                    }
+                    else
+                        block_crc = crc32c_iter(block_crc, next, prev, pos-prev);
                 }
                 prev = pos;
             }
         }
         else
         {
-            block_crc = crc32c_iter(block_crc, next, pos, (end > block_end ? block_end : end)-pos);
+            if (dsk->data_csum_type == BLOCKSTORE_CSUM_XXH3_32)
+            {
+                if (!xxh3_state)
+                {
+                    xxh3_state = XXH3_createState();
+                    XXH3_64bits_reset(xxh3_state);
+                }
+                xxh3_iter(xxh3_state, next, pos, (end > block_end ? block_end : end)-pos);
+            }
+            else
+                block_crc = crc32c_iter(block_crc, next, pos, (end > block_end ? block_end : end)-pos);
             pos = (end > block_end ? block_end : end);
             isset = true;
         }
+        if (dsk->data_csum_type == BLOCKSTORE_CSUM_XXH3_32 && xxh3_state)
+        {
+            block_crc = (uint32_t)XXH3_64bits_digest(xxh3_state);
+            XXH3_64bits_reset(xxh3_state);
+        }
         if (set)
         {
             *block_csums = block_crc;
         }
         else if (isset && block_crc != *block_csums)
         {
+            res = false;
             if (bad_block_cb)
-            {
                 bad_block_cb(blk_start, *block_csums, block_crc);
-                res = false;
-            }
             else
-                return false;
+                break;
         }
         block_end += dsk->csum_block_size;
         block_crc = 0;
         block_csums++;
     }
+    if (dsk->data_csum_type == BLOCKSTORE_CSUM_XXH3_32 && xxh3_state)
+    {
+        block_crc = (uint32_t)XXH3_64bits_digest(xxh3_state);
+        XXH3_freeState(xxh3_state);
+        xxh3_state = NULL;
+    }
     return res;
 }
 
@@ -1336,7 +1410,7 @@ int blockstore_heap_t::add_entry(uint32_t wr_size, uint32_t *modified_block,
     insert_list_item(li);
     li->block_num = block_num;
     new_wr->size = wr_size;
-    new_wr->crc32c = new_wr->calc_crc32c();
+    new_wr->checksum = new_wr->calc_checksum(this);
     return 0;
 }
 
@@ -1396,7 +1470,7 @@ int blockstore_heap_t::add_big_write(object_id oid, heap_entry_t *old_head, bool
             memset(wr->get_ext_bitmap(this), 0, dsk->clean_entry_bitmap_size);
         memset(wr->get_int_bitmap(this), 0, dsk->clean_entry_bitmap_size);
         bitmap_set(wr->get_int_bitmap(this), offset, len, dsk->bitmap_granularity);
-        if (dsk->data_csum_type)
+        if (dsk->csum_block_size)
         {
             memset(wr->get_checksums(this), 0, get_csum_size(wr));
             calc_checksums(wr, (uint8_t*)data, true, offset, len);
@@ -1425,7 +1499,7 @@ int blockstore_heap_t::add_redirect_intent(object_id oid, heap_entry_t **obj_ptr
             memset(wr->get_ext_bitmap(this), 0, dsk->clean_entry_bitmap_size);
         memset(wr->get_int_bitmap(this), 0, dsk->clean_entry_bitmap_size);
         bitmap_set(wr->get_int_bitmap(this), offset, len, dsk->bitmap_granularity);
-        if (dsk->data_csum_type)
+        if (dsk->csum_block_size)
             memset(wr->get_checksums(this), 0, get_csum_size(wr));
         calc_checksums(wr, (uint8_t*)data, true);
         *obj_ptr = wr;
@@ -1463,7 +1537,7 @@ int blockstore_heap_t::add_big_intent(object_id oid, heap_entry_t **obj_ptr, uin
             memcpy(wr->get_ext_bitmap(this), obj->get_ext_bitmap(this), dsk->clean_entry_bitmap_size);
         memcpy(wr->get_int_bitmap(this), obj->get_int_bitmap(this), dsk->clean_entry_bitmap_size);
         bitmap_set(wr->get_int_bitmap(this), offset, len, dsk->bitmap_granularity);
-        if (dsk->data_csum_type)
+        if (dsk->csum_block_size)
         {
             if (checksums)
                 memcpy(wr->get_checksums(this), checksums, get_csum_size(wr));
@@ -1510,7 +1584,7 @@ int blockstore_heap_t::add_compact(heap_entry_t *obj, uint64_t compact_version,
         new_wr->set_big_location(this, compact_location);
         memcpy(new_wr->get_int_bitmap(this), new_int_bitmap, dsk->clean_entry_bitmap_size);
         memcpy(new_wr->get_ext_bitmap(this), new_ext_bitmap, dsk->clean_entry_bitmap_size);
-        if (dsk->data_csum_type && new_csums)
+        if (dsk->csum_block_size && new_csums)
             memcpy(new_wr->get_checksums(this), new_csums, dsk->data_block_size/dsk->csum_block_size*(dsk->data_csum_type & 0xFF));
     });
 }
@@ -1980,21 +2054,24 @@ int blockstore_heap_t::list_objects(uint32_t pg_num, object_id min_oid, object_i
                 return;
             }
             uint64_t stable_version = 0;
-            iterate_with_stable(obj, UINT64_MAX, [&](heap_entry_t* wr, bool stable)
+            auto first_wr = obj;
+            for (auto wr = first_wr; wr; wr = prev(wr))
             {
-                if (stable)
+                if ((wr->entry_type & BS_HEAP_STABLE) || wr->type() == BS_HEAP_COMMIT || wr->type() == BS_HEAP_ROLLBACK)
                 {
                     stable_version = wr->version;
-                    return false;
+                    break;
                 }
-                if (unstable_size >= unstable_alloc)
+                else
                 {
-                    unstable_alloc = (!unstable_alloc ? 128 : unstable_alloc*2);
-                    unstable = (obj_ver_id*)realloc_or_die(unstable, sizeof(obj_ver_id) * unstable_alloc);
+                    if (unstable_size >= unstable_alloc)
+                    {
+                        unstable_alloc = (!unstable_alloc ? 128 : unstable_alloc*2);
+                        unstable = (obj_ver_id*)realloc_or_die(unstable, sizeof(obj_ver_id) * unstable_alloc);
+                    }
+                    unstable[unstable_size++] = (obj_ver_id){ .oid = oid, .version = wr->version };
                 }
-                unstable[unstable_size++] = (obj_ver_id){ .oid = oid, .version = wr->version };
-                return true;
-            });
+            }
             if (stable_version)
             {
                 if (res_size >= res_alloc)
@@ -2291,9 +2368,7 @@ void blockstore_heap_t::apply_inflight(heap_inflight_lsn_t & inflight)
         }
         if (!next)
         {
-            // The last freed entry must be a deletion
             assert(!prev);
-            assert(wr->entry_type == BS_HEAP_DELETE|BS_HEAP_STABLE);
             auto & pg_idx = block_index[get_pg_id(wr->inode, wr->stripe)];
             auto & inode_idx = pg_idx[wr->inode];
             heap_inode_map_t::iterator li_it;
@@ -2397,10 +2472,7 @@ void inode_map_get(void *inode_idx, heap_inode_map_t::iterator & li_it, heap_lis
     size_t map_n = ((size_t)inode_idx & IMAP_MALLOC_LOW_BITS);
     if (!map_n)
     {
-#pragma GCC diagnostic push
-#pragma GCC diagnostic ignored "-Warray-bounds"
         li_it = ((heap_inode_map_t*)inode_idx)->find(list_item_key(&stripe));
-#pragma GCC diagnostic pop
         li = li_it != ((heap_inode_map_t*)inode_idx)->end() ? *li_it : NULL;
     }
     else if (map_n == 1)

src/blockstore/blockstore_heap.h

@@ -43,7 +43,7 @@ struct __attribute__((__packed__)) heap_entry_t
 {
     uint16_t size;
     uint16_t entry_type;
-    uint32_t crc32c;
+    uint32_t checksum;
     uint64_t lsn;
     uint64_t inode;
     uint64_t stripe;
@@ -69,7 +69,8 @@ struct __attribute__((__packed__)) heap_entry_t
     uint32_t *get_checksum(blockstore_heap_t *heap);
     uint64_t big_location(blockstore_heap_t *heap);
     void set_big_location(blockstore_heap_t *heap, uint64_t location);
-    uint32_t calc_crc32c();
+    uint32_t calc_checksum(blockstore_heap_t *heap);
+    uint32_t calc_checksum(blockstore_disk_t *dsk);
 };
 
 struct __attribute__((__packed__)) heap_small_write_t
@@ -80,7 +81,7 @@ struct __attribute__((__packed__)) heap_small_write_t
     uint32_t offset;
     uint32_t len;
 
-    // Also includes 1 bitmap and 1 crc32c after the bitmap if checksums are disabled
+    // Also includes 1 bitmap and 1 checksum after the bitmap if block checksums are disabled
 };
 
 struct __attribute__((__packed__)) heap_big_write_t
@@ -98,7 +99,7 @@ struct __attribute__((__packed__)) heap_big_intent_t
     uint32_t offset;
     uint32_t len;
 
-    // Also includes 2 bitmaps and 1 crc32c if checksums are disabled
+    // Also includes 2 bitmaps and 1 checksums if block checksums are disabled
 };
 
 struct __attribute__((__packed__)) heap_list_item_t

src/blockstore/blockstore_impl.cpp

@@ -193,12 +193,12 @@ void blockstore_impl_t::loop()
             heap->start_block_write(block_num);
             mb.sent = true;
         }
-        pending_modified_blocks.clear();
         int ret = ringloop->submit();
         if (ret < 0)
         {
             throw std::runtime_error(std::string("io_uring_submit: ") + strerror(-ret));
         }
+        pending_modified_blocks.clear();
         if ((initial_ring_space - ringloop->space_left()) > 0)
         {
             live = true;

src/blockstore/blockstore_impl.h

@@ -78,7 +78,6 @@ public:
     // Suitable only for server SSDs with capacitors, requires disabled data and journal fsyncs
     int immediate_commit = IMMEDIATE_NONE;
     bool inmemory_meta = false;
-    bool skip_corrupted_meta_entries = false;
     uint32_t meta_write_recheck_parallelism = 0;
     // Maximum and minimum flusher count
     unsigned max_flusher_count = 0, min_flusher_count = 0;

src/blockstore/blockstore_init.cpp

@@ -145,7 +145,7 @@ resume_1:
             printf(
                 "Configuration stored in metadata superblock"
                 " (meta_block_size=%u, data_block_size=%u, bitmap_granularity=%u, data_csum_type=%u, csum_block_size=%u, meta_area_size=%ju)"
-                " differs from OSD configuration (%ju/%ju/%u, %u/%u, %ju).\n",
+                " differs from OSD configuration (%u/%u/%u, %u/%u, %ju).\n",
                 hdr->meta_block_size, hdr->data_block_size, hdr->bitmap_granularity,
                 hdr->data_csum_type, hdr->csum_block_size, hdr->meta_area_size,
                 bs->dsk.meta_block_size, bs->dsk.data_block_size, bs->dsk.bitmap_granularity,
@@ -225,7 +225,7 @@ resume_4:
         {
             // Handle result
             uint64_t loaded = 0;
-            int r = bs->heap->load_blocks(bufs[i].offset-bs->dsk.meta_block_size, bufs[i].size, bufs[i].buf, bs->skip_corrupted_meta_entries, loaded);
+            int r = bs->heap->load_blocks(bufs[i].offset-bs->dsk.meta_block_size, bufs[i].size, bufs[i].buf, false, loaded);
             if (r != 0)
                 exit(1);
             entries_loaded += loaded;

src/blockstore/blockstore_open.cpp

@@ -28,7 +28,6 @@ void blockstore_impl_t::parse_config(blockstore_config_t & config, bool init)
     throttle_target_parallelism = strtoull(config["throttle_target_parallelism"].c_str(), NULL, 10);
     throttle_threshold_us = strtoull(config["throttle_threshold_us"].c_str(), NULL, 10);
     perfect_csum_update = config["perfect_csum_update"] == "true" || config["perfect_csum_update"] == "1" || config["perfect_csum_update"] == "yes";
-    skip_corrupted_meta_entries = config["skip_corrupted_meta_entries"] == "true" || config["skip_corrupted_meta_entries"] == "1" || config["skip_corrupted_meta_entries"] == "yes";
     if (config["autosync_writes"] != "")
     {
         autosync_writes = strtoull(config["autosync_writes"].c_str(), NULL, 10);

src/blockstore/blockstore_write.cpp

@@ -22,23 +22,21 @@ void blockstore_impl_t::prepare_meta_block_write(uint32_t modified_block)
     ring_data_t *data = ((ring_data_t*)sqe->user_data);
     uint8_t *buf = (uint8_t*)memalign_or_die(MEM_ALIGNMENT, dsk.meta_block_size);
     data->iov = (struct iovec){ buf, (size_t)dsk.meta_block_size };
-    data->callback = [this, modified_block](ring_data_t *data)
+    data->callback = [this, modified_block, buf](ring_data_t *data)
     {
+        free(buf);
         live = true;
         if (data->res != data->iov.iov_len)
         {
             // FIXME: our state becomes corrupted after a write error. maybe do something better than just die
             disk_error_abort("data write", data->res, data->iov.iov_len);
         }
-        auto it = modified_blocks.find(modified_block);
-        assert(it != modified_blocks.end());
-        free(it->second.buf);
-        modified_blocks.erase(it);
+        modified_blocks.erase(modified_block);
         heap->complete_block_write(modified_block);
         ringloop->wakeup();
     };
     io_uring_prep_writev(
-        sqe, dsk.meta_fd, &data->iov, 1, dsk.meta_offset + ((uint64_t)modified_block+1)*dsk.meta_block_size
+        sqe, dsk.meta_fd, &data->iov, 1, dsk.meta_offset + (modified_block+1)*dsk.meta_block_size
     );
     unsynced_meta_write_count++;
     pending_modified_blocks.push_back(modified_block);
@@ -251,12 +249,13 @@ enospc:
             goto enospc;
         assert(res == 0);
         PRIV(op)->lsn = obj->lsn;
+        if (op->len)
+            heap->use_buffer_area(op->oid.inode, loc, op->len);
         prepare_meta_block_write(PRIV(op)->modified_block);
         PRIV(op)->pending_ops++;
         if (op->len > 0)
         {
             // Prepare buffered data write
-            heap->use_buffer_area(op->oid.inode, loc, op->len);
             if (dsk.inmemory_journal)
             {
                 memcpy((uint8_t*)buffer_area + loc, op->buf, op->len);
@@ -349,7 +348,6 @@ resume_12:
     }
 resume_4:
     {
-        BS_SUBMIT_CHECK_SQES(1);
         auto obj = heap->read_entry(op->oid);
         int res = 0;
         if (PRIV(op)->write_type == _REDIRECT_INTENT)
@@ -406,12 +404,11 @@ resume_6:
         if (ref_us > exec_us + throttle_threshold_us)
         {
             // Pause reply
-            PRIV(op)->pending_ops++;
             PRIV(op)->op_state = 7;
             // Remember that the timer can in theory be called right here
             tfd->set_timer_us(ref_us-exec_us, false, [this, op](int timer_id)
             {
-                PRIV(op)->pending_ops--;
+                PRIV(op)->op_state = 8;
                 ringloop->wakeup();
             });
             return 1;

src/blockstore/v1/init.cpp

@@ -189,7 +189,7 @@ resume_1:
             printf(
                 "Configuration stored in metadata superblock"
                 " (meta_block_size=%u, data_block_size=%u, bitmap_granularity=%u, data_csum_type=%u, csum_block_size=%u)"
-                " differs from OSD configuration (%ju/%ju/%u, %u/%u).\n",
+                " differs from OSD configuration (%u/%u/%u, %u/%u).\n",
                 hdr->meta_block_size, hdr->data_block_size, hdr->bitmap_granularity,
                 hdr->data_csum_type, hdr->csum_block_size,
                 bs->dsk.meta_block_size, bs->dsk.data_block_size, bs->dsk.bitmap_granularity,

src/blockstore/v1/read.cpp

@@ -620,7 +620,8 @@ bool blockstore_impl_t::fulfill_clean_read(blockstore_op_t *read_op, uint64_t &
     else if (from_journal)
     {
         // Don't scan bitmap - journal writes don't have holes (internal bitmap)!
-        uint8_t *csum = !dsk.csum_block_size ? 0 : (clean_entry_bitmap + dsk.clean_entry_bitmap_size);
+        uint8_t *csum = !dsk.csum_block_size ? 0 : (clean_entry_bitmap + dsk.clean_entry_bitmap_size +
+            item_start/dsk.csum_block_size*(dsk.data_csum_type & 0xFF));
         if (!fulfill_read(read_op, fulfilled, item_start, item_end,
             (BS_ST_BIG_WRITE | BS_ST_STABLE), 0, clean_loc + item_start, 0, csum, dyn_data))
         {

src/client/CMakeLists.txt

@@ -13,10 +13,10 @@ if (RDMACM_LIBRARIES)
 endif (RDMACM_LIBRARIES)
 add_library(vitastor_common STATIC
 	../util/epoll_manager.cpp etcd_state_client.cpp messenger.cpp ../util/addr_util.cpp
-	msgr_stop.cpp msgr_op.cpp msgr_send.cpp msgr_receive.cpp ../util/ringloop.cpp ../../json11/json11.cpp
+	msgr_encrypt.cpp msgr_stop.cpp msgr_op.cpp msgr_send.cpp msgr_receive.cpp ../util/ringloop.cpp ../../json11/json11.cpp
 	http_client.cpp osd_ops.cpp pg_states.cpp ../util/timerfd_manager.cpp ../util/str_util.cpp ../util/json_util.cpp ${MSGR_RDMA} ${MSGR_RDMACM}
 )
-target_link_libraries(vitastor_common pthread)
+target_link_libraries(vitastor_common pthread ${OPENSSL_LIBRARIES})
 target_compile_options(vitastor_common PUBLIC -fPIC)
 
 # libvitastor_client.so
@@ -33,6 +33,7 @@ target_link_libraries(vitastor_client
 	${LIBURING_LIBRARIES}
 	${IBVERBS_LIBRARIES}
 	${RDMACM_LIBRARIES}
+	${OPENSSL_LIBRARIES}
 )
 set_target_properties(vitastor_client PROPERTIES VERSION ${VITASTOR_VERSION} SOVERSION 0)
 configure_file(vitastor.pc.in vitastor.pc @ONLY)
@@ -98,9 +99,10 @@ endif (${WITH_QEMU})
 add_executable(test_cluster_client
 	EXCLUDE_FROM_ALL
 	../test/test_cluster_client.cpp
-	pg_states.cpp osd_ops.cpp cluster_client.cpp cluster_client_list.cpp cluster_client_wb.cpp msgr_op.cpp ../test/mock/messenger.cpp msgr_stop.cpp
+	pg_states.cpp osd_ops.cpp cluster_client.cpp cluster_client_list.cpp cluster_client_wb.cpp msgr_op.cpp ../test/mock/messenger.cpp msgr_stop.cpp msgr_encrypt.cpp
 	etcd_state_client.cpp ../util/timerfd_manager.cpp ../util/addr_util.cpp ../util/str_util.cpp ../util/json_util.cpp ../../json11/json11.cpp
 )
+target_link_libraries(test_cluster_client ${OPENSSL_LIBRARIES})
 target_compile_definitions(test_cluster_client PUBLIC -D__MOCK__)
 target_include_directories(test_cluster_client BEFORE PUBLIC ${CMAKE_SOURCE_DIR}/src/test/mock)
 add_dependencies(build_tests test_cluster_client)

src/client/cluster_client.cpp

@@ -958,9 +958,22 @@ bool cluster_client_t::check_rw(cluster_op_t *op)
     {
         op->flags |= OP_IMMEDIATE_COMMIT;
     }
+    auto ino_it = st_cli.inode_config.find(op->inode);
+    if (ino_it != st_cli.inode_config.end() && ino_it->second.enc)
+    {
+        // FIXME: Rework client API by adding open/close and cache inode information in the "FD"
+        op->enc = ino_it->second.enc;
+        if (!op->enc->bitmap_granularity)
+        {
+            op->enc->bitmap_granularity = pool_it->second.bitmap_granularity;
+        }
+    }
+    else
+    {
+        op->enc.reset();
+    }
     if ((op->opcode == OSD_OP_WRITE || op->opcode == OSD_OP_DELETE) && !(op->flags & OSD_OP_IGNORE_READONLY))
     {
-        auto ino_it = st_cli.inode_config.find(op->inode);
         if (ino_it != st_cli.inode_config.end() && ino_it->second.readonly)
         {
             op->retval = -EROFS;
@@ -972,7 +985,6 @@ bool cluster_client_t::check_rw(cluster_op_t *op)
     op->deoptimise_snapshot = false;
     if (enable_writeback && (op->opcode == OSD_OP_READ || op->opcode == OSD_OP_READ_BITMAP || op->opcode == OSD_OP_READ_CHAIN_BITMAP))
     {
-        auto ino_it = st_cli.inode_config.find(op->inode);
         if (ino_it != st_cli.inode_config.end())
         {
             int chain_size = 0;
@@ -1435,6 +1447,7 @@ int cluster_client_t::try_send(cluster_op_t *op, int i, std::function<void(osd_o
                     ? (uint8_t*)op->part_bitmaps + pg_bitmap_size*i : NULL),
                 .bitmap_len = (unsigned)(op->opcode == OSD_OP_READ || op->opcode == OSD_OP_READ_BITMAP || op->opcode == OSD_OP_READ_CHAIN_BITMAP
                     ? pg_bitmap_size : 0),
+                .enc = op->enc,
                 .callback = cb ? cb : [this, part](osd_op_t *op_part)
                 {
                     handle_op_part(part);

src/client/cluster_client.h

@@ -71,6 +71,7 @@ protected:
     cluster_op_t *prev = NULL, *next = NULL;
     int prev_wait = 0;
     uint64_t flush_id = 0;
+    std::shared_ptr<inode_enc_t> enc;
     friend class cluster_client_t;
     friend class writeback_cache_t;
 };

src/client/etcd_state_client.cpp

@@ -22,14 +22,19 @@ etcd_state_client_t::~etcd_state_client_t()
     stop_ws_keepalive();
     if (etcd_watch_ws)
     {
-        http_close(etcd_watch_ws);
+        http_destroy(etcd_watch_ws);
         etcd_watch_ws = NULL;
     }
     if (keepalive_client)
     {
-        http_close(keepalive_client);
+        http_destroy(keepalive_client);
         keepalive_client = NULL;
     }
+    if (http_ctx)
+    {
+        http_context_destroy(http_ctx);
+        http_ctx = NULL;
+    }
 #endif
     if (load_pgs_timer_id >= 0)
     {
@@ -72,10 +77,27 @@ std::vector<std::string> etcd_state_client_t::get_addresses()
     return addrs;
 }
 
+http_context_t *etcd_state_client_t::get_http_ctx()
+{
+    if (!http_ctx)
+    {
+        std::string error;
+        http_ctx = http_context_init(etcd_client_cert, etcd_client_key, etcd_ca, true, error);
+        if (!http_ctx)
+        {
+            fprintf(stderr, "Failed to initialize HTTP context: %s\n", error.c_str());
+            exit(1);
+        }
+    }
+    return http_ctx;
+}
+
 void etcd_state_client_t::etcd_call_oneshot(std::string etcd_address, std::string api, json11::Json payload,
     int timeout, std::function<void(std::string, json11::Json)> callback)
 {
     std::string etcd_api_path;
+    bool ssl = etcd_address.substr(0, 8) == "https://";
+    etcd_address = etcd_address.substr(ssl ? 8 : 7);
     int pos = etcd_address.find('/');
     if (pos >= 0)
     {
@@ -89,16 +111,16 @@ void etcd_state_client_t::etcd_call_oneshot(std::string etcd_address, std::strin
         "Content-Length: "+std::to_string(req.size())+"\r\n"
         "Connection: close\r\n"
         "\r\n"+req;
-    auto http_cli = http_init(tfd);
-    auto cb = [http_cli, callback](const http_response_t *response)
+    auto http_cli = http_init(tfd, get_http_ctx());
+    auto cb = [http_cli, callback](http_message_t *response)
     {
         std::string err;
         json11::Json data;
         response->parse_json_response(err, data);
         callback(err, data);
-        http_close(http_cli);
+        http_destroy(http_cli);
     };
-    http_request(http_cli, etcd_address, req, { .timeout = timeout }, cb);
+    http_request(http_cli, etcd_address, req, { .timeout = timeout, .ssl = ssl }, cb);
 }
 
 void etcd_state_client_t::etcd_call(std::string api, json11::Json payload, int timeout,
@@ -112,6 +134,8 @@ void etcd_state_client_t::etcd_call(std::string api, json11::Json payload, int t
     pick_next_etcd();
     std::string etcd_address = selected_etcd_address;
     std::string etcd_api_path;
+    bool ssl = etcd_address.substr(0, 8) == "https://";
+    etcd_address = etcd_address.substr(ssl ? 8 : 7);
     int pos = etcd_address.find('/');
     if (pos >= 0)
     {
@@ -128,7 +152,7 @@ void etcd_state_client_t::etcd_call(std::string api, json11::Json payload, int t
         "\r\n"+req;
     retries--;
     auto cb = [this, api, payload, timeout, retries, interval, callback,
-        cur_addr = selected_etcd_address](const http_response_t *response)
+        cur_addr = selected_etcd_address](http_message_t *response)
     {
         std::string err;
         json11::Json data;
@@ -164,22 +188,21 @@ void etcd_state_client_t::etcd_call(std::string api, json11::Json payload, int t
             callback(err, data);
     };
     if (!keepalive_client)
-    {
-        keepalive_client = http_init(tfd);
-    }
-    http_request(keepalive_client, etcd_address, req, { .timeout = timeout, .keepalive = true }, cb);
+        keepalive_client = http_init(tfd, get_http_ctx());
+    http_request(keepalive_client, etcd_address, req, { .timeout = timeout, .keepalive = true, .ssl = ssl }, cb);
 }
 
 void etcd_state_client_t::add_etcd_url(std::string addr)
 {
     if (addr.length() > 0)
     {
+        bool ssl = false;
         if (strtolower(addr.substr(0, 7)) == "http://")
             addr = addr.substr(7);
         else if (strtolower(addr.substr(0, 8)) == "https://")
         {
-            fprintf(stderr, "HTTPS is unsupported for etcd. Either use plain HTTP or setup a local proxy for etcd interaction\n");
-            exit(1);
+            addr = addr.substr(8);
+            ssl = true;
         }
         if (!local_ips.size())
             local_ips = getifaddr_list(std::vector<addr_mask_t>(), true);
@@ -194,6 +217,7 @@ void etcd_state_client_t::add_etcd_url(std::string addr)
             check_addr = addr;
         if (pos == std::string::npos)
             addr += "/v3";
+        addr = (ssl ? "https://" : "http://") + addr;
         bool local = false;
         int i;
         for (i = 0; i < local_ips.size(); i++)
@@ -239,6 +263,9 @@ void etcd_state_client_t::parse_config(const json11::Json & config)
             add_etcd_url(ea.string_value());
         }
     }
+    this->etcd_client_cert = config["etcd_client_cert"].string_value();
+    this->etcd_client_key = config["etcd_client_key"].string_value();
+    this->etcd_ca = config["etcd_ca"].string_value();
     this->etcd_prefix = config["etcd_prefix"].string_value();
     if (this->etcd_prefix == "")
     {
@@ -331,6 +358,8 @@ void etcd_state_client_t::start_etcd_watcher()
     pick_next_etcd();
     std::string etcd_address = selected_etcd_address;
     std::string etcd_api_path;
+    bool ssl = etcd_address.substr(0, 8) == "https://";
+    etcd_address = etcd_address.substr(ssl ? 8 : 7);
     int pos = etcd_address.find('/');
     if (pos >= 0)
     {
@@ -339,18 +368,17 @@ void etcd_state_client_t::start_etcd_watcher()
     }
     etcd_watches_initialised = 0;
     ws_alive = 1;
-    if (etcd_watch_ws)
-    {
-        http_close(etcd_watch_ws);
-        etcd_watch_ws = NULL;
-    }
     if (this->log_level > 1)
     {
         fprintf(stderr, "Trying to connect to etcd websocket at %s, watch from revision %ju/%ju/%ju\n", etcd_address.c_str(),
             etcd_watch_revision_config, etcd_watch_revision_osd, etcd_watch_revision_pg);
     }
-    etcd_watch_ws = open_websocket(tfd, etcd_address, etcd_api_path+"/watch", etcd_slow_timeout,
-        [this, cur_addr = selected_etcd_address](const http_response_t *msg)
+    if (!etcd_watch_ws)
+        etcd_watch_ws = http_init(tfd, get_http_ctx());
+    else
+        http_close(etcd_watch_ws);
+    open_websocket(etcd_watch_ws, etcd_address, etcd_api_path+"/watch", { .timeout = etcd_slow_timeout, .ssl = ssl },
+        [this, cur_addr = selected_etcd_address](http_message_t *msg)
     {
         if (msg->body.length())
         {
@@ -393,7 +421,6 @@ void etcd_state_client_t::start_etcd_watcher()
                                 fprintf(stderr, "Revisions before %ju were compacted by etcd, reloading state\n",
                                     data["result"]["compact_revision"].uint64_value());
                                 http_close(etcd_watch_ws);
-                                etcd_watch_ws = NULL;
                                 etcd_watch_revision_config = etcd_watch_revision_osd = etcd_watch_revision_pg = 0;
                                 on_reload_hook();
                             }
@@ -414,10 +441,7 @@ void etcd_state_client_t::start_etcd_watcher()
                 }
                 // Save revision only if it's present in the message - because sometimes etcd sends something without a header, like:
                 // {"error": {"grpc_code": 14, "http_code": 503, "http_status": "Service Unavailable", "message": "error reading from server: EOF"}}
-                // Also don't save revision from the initial created: true messages because they always contain the latest revision
-                if (etcd_watches_initialised == ETCD_TOTAL_WATCHES &&
-                    !data["result"]["header"]["revision"].is_null() &&
-                    !data["result"]["created"].bool_value())
+                if (etcd_watches_initialised == ETCD_TOTAL_WATCHES && !data["result"]["header"]["revision"].is_null())
                 {
                     // Restart watchers from the same revision number as in the last received message,
                     // not from the next one to protect against revision being split into multiple messages,
@@ -470,11 +494,6 @@ void etcd_state_client_t::start_etcd_watcher()
             fprintf(stderr, "Disconnected from etcd %s\n", cur_addr.c_str());
             if (cur_addr == selected_etcd_address)
                 selected_etcd_address = "";
-            if (etcd_watch_ws)
-            {
-                http_close(etcd_watch_ws);
-                etcd_watch_ws = NULL;
-            }
             if (etcd_watches_initialised == 0)
             {
                 // Connection not established, retry in <etcd_quick_timeout>
@@ -551,11 +570,6 @@ void etcd_state_client_t::start_ws_keepalive()
                 {
                     fprintf(stderr, "Websocket ping failed, disconnecting from etcd %s\n", selected_etcd_address.c_str());
                 }
-                if (etcd_watch_ws)
-                {
-                    http_close(etcd_watch_ws);
-                    etcd_watch_ws = NULL;
-                }
                 start_etcd_watcher();
             }
             else
@@ -1282,6 +1296,16 @@ void etcd_state_client_t::parse_state(const etcd_kv_t & kv)
                     else
                         parent_inode_num |= parent_pool_id << (64-POOL_ID_BITS);
                 }
+                std::shared_ptr<inode_enc_t> enc;
+                if (!value["enc_key"].string_value().empty())
+                {
+                    std::vector<uint8_t> k = hexdecode(value["enc_key"].string_value());
+                    if (k.size() == 512/8)
+                    {
+                        enc = std::make_shared<inode_enc_t>();
+                        enc->key = std::move(k);
+                    }
+                }
                 insert_inode_config((inode_config_t){
                     .num = inode_num,
                     .name = value["name"].string_value(),
@@ -1289,6 +1313,7 @@ void etcd_state_client_t::parse_state(const etcd_kv_t & kv)
                     .parent_id = parent_inode_num,
                     .readonly = value["readonly"].bool_value(),
                     .deleted = value["deleted"].bool_value(),
+                    .enc = enc,
                     .meta = value["meta"],
                     .mod_revision = kv.mod_revision,
                 });

src/client/etcd_state_client.h

@@ -4,9 +4,10 @@
 #pragma once
 
 #include <set>
+#include <memory>
 
 #include "json11/json11.hpp"
-#include "object_id.h"
+#include "osd_id.h"
 #include "timerfd_manager.h"
 
 #define ETCD_CONFIG_WATCH_ID 1
@@ -75,6 +76,14 @@ struct pool_config_t
     void *reshard_state = NULL;
 };
 
+struct inode_enc_t
+{
+    int refs = 0;
+    std::vector<uint8_t> key;
+    // FIXME It may also contain snapshot chain and key information
+    uint32_t bitmap_granularity = 0;
+};
+
 struct inode_config_t
 {
     uint64_t num = 0;
@@ -83,6 +92,7 @@ struct inode_config_t
     inode_t parent_id = 0;
     bool readonly = false;
     bool deleted = false;
+    std::shared_ptr<inode_enc_t> enc;
     // Arbitrary metadata
     json11::Json meta;
     // Change revision of the metadata in etcd
@@ -96,6 +106,7 @@ struct inode_watch_t
 };
 
 struct http_co_t;
+struct http_context_t;
 
 struct __attribute__((visibility("default"))) etcd_state_client_t
 {
@@ -125,9 +136,13 @@ public:
     uint32_t global_immediate_commit = IMMEDIATE_NONE;
 
     std::string etcd_prefix;
+    std::string etcd_client_cert;
+    std::string etcd_client_key;
+    std::string etcd_ca;
     int log_level = 0;
     timerfd_manager_t *tfd = NULL;
 
+    http_context_t *http_ctx = NULL;
     http_co_t *etcd_watch_ws = NULL, *keepalive_client = NULL;
     int etcd_watches_initialised = 0;
     uint64_t etcd_watch_revision_config = 0;
@@ -160,6 +175,7 @@ public:
     json11::Json::object serialize_inode_cfg(inode_config_t *cfg);
     etcd_kv_t parse_etcd_kv(const json11::Json & kv_json);
     std::vector<std::string> get_addresses();
+    http_context_t *get_http_ctx();
     void etcd_call_oneshot(std::string etcd_address, std::string api, json11::Json payload, int timeout, std::function<void(std::string, json11::Json)> callback);
     void etcd_call(std::string api, json11::Json payload, int timeout, int retries, int interval, std::function<void(std::string, json11::Json)> callback);
     void etcd_txn(json11::Json txn, int timeout, int retries, int interval, std::function<void(std::string, json11::Json)> callback);

src/client/http_client.cpp

@@ -10,9 +10,17 @@
 #include <unistd.h>
 #include <fcntl.h>
 #include <string.h>
+#include <assert.h>
 
 #include <stdexcept>
 
+#ifdef WITH_OPENSSL
+#include <openssl/bio.h>
+#include <openssl/err.h>
+#include <openssl/pem.h>
+#include <openssl/ssl.h>
+#endif
+
 #include "addr_util.h"
 #include "str_util.h"
 #include "json_util.h"
@@ -24,14 +32,51 @@
 
 static std::string ws_format_frame(int type, uint64_t size);
 static bool ws_parse_frame(std::string & buf, uint8_t & type, std::string & res);
-static void parse_http_headers(std::string & res, http_response_t *parsed);
+static void parse_http_headers(std::string & res, http_message_t *parsed, bool is_request);
+
+struct http_context_t
+{
+    std::string ssl_cert;
+    std::string ssl_key;
+    std::string ssl_ca;
+
+#ifdef WITH_OPENSSL
+    SSL_CTX *ssl_ctx = NULL;
+#endif
+
+    ~http_context_t()
+    {
+#ifdef WITH_OPENSSL
+        if (ssl_ctx)
+        {
+            SSL_CTX_free(ssl_ctx);
+            ssl_ctx = NULL;
+        }
+#endif
+    }
+};
+
+struct http_call_t
+{
+    std::string host;
+    std::string request;
+    http_options_t options;
+    std::function<void(http_message_t *)> cb;
+};
 
 struct http_co_t
 {
+    http_context_t *ctx = NULL;
+#ifdef WITH_OPENSSL
+    SSL *ssl_cli = NULL;
+    BIO *ssl_bio = NULL;
+#endif
+
     timerfd_manager_t *tfd;
-    std::function<void(const http_response_t*)> response_callback;
+    std::function<void(http_message_t*)> response_callback;
 
     int request_timeout = 0;
+    bool ssl = false;
     std::string host;
     std::string request;
     std::string ws_outbox;
@@ -39,7 +84,7 @@ struct http_co_t
     bool want_streaming;
     bool keepalive;
 
-    std::vector<std::function<void()>> keepalive_queue;
+    std::vector<http_call_t> keepalive_queue;
 
     int state = 0;
     std::string connected_host;
@@ -47,10 +92,10 @@ struct http_co_t
     int timeout_id = -1;
     int epoll_events = 0;
     int sent = 0;
-    std::vector<char> rbuf;
+    std::vector<uint8_t> rbuf;
     iovec read_iov, send_iov;
     msghdr read_msg = { 0 }, send_msg = { 0 };
-    http_response_t parsed;
+    http_message_t parsed;
     uint64_t target_response_size = 0;
 
     int onstack = 0;
@@ -70,9 +115,14 @@ struct http_co_t
     void submit_read(bool check_timeout);
     void submit_send();
     bool handle_read();
+#ifdef WITH_OPENSSL
+    bool do_ssl_handshake(bool init_send);
+    void on_ssl_error(int res);
+#endif
     void post_message(uint8_t type, const std::string & msg);
+    void reply(const std::string & msg);
     void send_request(const std::string & host, const std::string & request,
-        const http_options_t & options, std::function<void(const http_response_t *response)> response_callback);
+        const http_options_t & options, std::function<void(http_message_t *response)> response_callback);
 };
 
 #define HTTP_CO_CLOSED 0
@@ -83,20 +133,64 @@ struct http_co_t
 #define HTTP_CO_WEBSOCKET 5
 #define HTTP_CO_CHUNKED 6
 #define HTTP_CO_KEEPALIVE 7
+#define HTTP_CO_SERVER 8
+#define HTTP_CO_REQ_HDR_RECEIVED 9
+#define HTTP_CO_REQUEST_RECEIVED 10
 
 #define DEFAULT_TIMEOUT 5000
 
-http_co_t *http_init(timerfd_manager_t *tfd)
+http_context_t* http_context_init(const std::string & ssl_cert, const std::string & ssl_key,
+    const std::string & ssl_ca, bool verify_peer, std::string & error)
+{
+    http_context_t *ctx = new http_context_t;
+#ifdef WITH_OPENSSL
+    SSL_CTX *ssl_ctx = SSL_CTX_new(TLS_method());
+    ctx->ssl_cert = ssl_cert;
+    ctx->ssl_key = ssl_key;
+    ctx->ssl_ca = ssl_ca;
+    ctx->ssl_ctx = ssl_ctx;
+    if (!ssl_ctx)
+        goto init_err;
+    SSL_CTX_set_verify(ssl_ctx, verify_peer ? SSL_VERIFY_PEER : SSL_VERIFY_NONE, NULL);
+    if (!SSL_CTX_set_min_proto_version(ssl_ctx, TLS1_2_VERSION))
+        goto init_err;
+    if ((ssl_ca != "")
+        ? !SSL_CTX_load_verify_locations(ssl_ctx, ssl_ca.c_str(), NULL)
+        : !SSL_CTX_set_default_verify_paths(ssl_ctx))
+        goto init_err;
+    if (ssl_cert != "" && ssl_key != "" &&
+        (!SSL_CTX_use_certificate_file(ssl_ctx, ssl_cert.c_str(), SSL_FILETYPE_PEM) ||
+        !SSL_CTX_use_PrivateKey_file(ssl_ctx, ssl_key.c_str(), SSL_FILETYPE_PEM)))
+        goto init_err;
+#endif
+    return ctx;
+init_err:
+    error = std::string("openssl initialization failed: ")+ERR_error_string(ERR_get_error(), NULL);
+    delete ctx;
+    return NULL;
+}
+
+void http_context_destroy(http_context_t *ctx)
+{
+    delete ctx;
+}
+
+http_co_t *http_init(timerfd_manager_t *tfd, http_context_t *ctx)
 {
     http_co_t *handler = new http_co_t();
     handler->tfd = tfd;
     handler->state = HTTP_CO_CLOSED;
+    handler->ctx = ctx;
     return handler;
 }
 
-http_co_t* open_websocket(timerfd_manager_t *tfd, const std::string & host, const std::string & path,
-    int timeout, std::function<void(const http_response_t *msg)> response_callback)
+void open_websocket(http_co_t *handler, const std::string & host, const std::string & path,
+    const http_options_t & options, std::function<void(http_message_t *msg)> response_callback)
 {
+    if (handler->state == HTTP_CO_KEEPALIVE && (handler->connected_host != host || handler->ssl != options.ssl))
+        handler->close_connection();
+    if (handler->state != HTTP_CO_KEEPALIVE && handler->state != HTTP_CO_CLOSED)
+        throw std::runtime_error("Attempt to open websocket on a keepalive stream");
     std::string request = "GET "+path+" HTTP/1.1\r\n"
         "Host: "+host+"\r\n"
         "Upgrade: websocket\r\n"
@@ -104,29 +198,54 @@ http_co_t* open_websocket(timerfd_manager_t *tfd, const std::string & host, cons
         "Sec-WebSocket-Key: x3JJHMbDL1EzLkh9GBhXDw==\r\n"
         "Sec-WebSocket-Version: 13\r\n"
         "\r\n";
-    http_co_t *handler = new http_co_t();
-    handler->tfd = tfd;
-    handler->state = HTTP_CO_CLOSED;
     handler->host = host;
-    handler->request_timeout = timeout < 0 ? -1 : (timeout == 0 ? DEFAULT_TIMEOUT : timeout);
+    handler->request_timeout = options.timeout < 0 ? -1 : (options.timeout == 0 ? DEFAULT_TIMEOUT : options.timeout);
     handler->want_streaming = false;
     handler->keepalive = false;
+    handler->ssl = options.ssl;
     handler->request = request;
     handler->response_callback = response_callback;
+    handler->ws_outbox = "";
+    handler->response = "";
+    handler->sent = 0;
+    handler->parsed = {};
     handler->start_ws_connection();
-    return handler;
 }
 
 void http_request(http_co_t *handler, const std::string & host, const std::string & request,
-    const http_options_t & options, std::function<void(const http_response_t *response)> response_callback)
+    const http_options_t & options, std::function<void(http_message_t *response)> response_callback)
 {
     handler->send_request(host, request, options, response_callback);
 }
 
+void http_serve(http_co_t *handler, int peer_fd, const http_options_t & options, std::function<void(http_message_t *msg)> request_callback)
+{
+    if (handler->state != HTTP_CO_SERVER || handler->peer_fd != peer_fd)
+        handler->close_connection();
+    handler->host = "";
+    handler->request_timeout = options.timeout < 0 ? -1 : (options.timeout == 0 ? DEFAULT_TIMEOUT : options.timeout);
+    handler->want_streaming = false;
+    handler->keepalive = false;
+    handler->ssl = options.ssl;
+    handler->request = "";
+    handler->response_callback = request_callback;
+    handler->ws_outbox = "";
+    handler->response = "";
+    handler->sent = 0;
+    handler->parsed = {};
+    handler->peer_fd = peer_fd;
+    handler->state = HTTP_CO_SERVER;
+    handler->tfd->set_fd_handler(peer_fd, false, [handler](int peer_fd, int epoll_events)
+    {
+        handler->epoll_events |= epoll_events;
+        handler->handle_events();
+    });
+}
+
 void http_co_t::run_cb_and_clear()
 {
     parsed.eof = true;
-    std::function<void(const http_response_t*)> cb;
+    std::function<void(http_message_t*)> cb;
     cb.swap(response_callback);
     // Call callback after clearing it because otherwise we may hit reenterability problems
     if (cb != NULL)
@@ -135,7 +254,7 @@ void http_co_t::run_cb_and_clear()
 }
 
 void http_co_t::send_request(const std::string & host, const std::string & request,
-    const http_options_t & options, std::function<void(const http_response_t *response)> response_callback)
+    const http_options_t & options, std::function<void(http_message_t *response)> response_callback)
 {
     stackin();
     if (state == HTTP_CO_WEBSOCKET)
@@ -145,20 +264,18 @@ void http_co_t::send_request(const std::string & host, const std::string & reque
     }
     else if (state != HTTP_CO_KEEPALIVE && state != HTTP_CO_CLOSED)
     {
-        keepalive_queue.push_back([this, host, request, options, response_callback]()
-        {
-            this->send_request(host, request, options, response_callback);
-        });
+        keepalive_queue.emplace_back((http_call_t){ host, request, options, std::move(response_callback) });
         stackout();
         return;
     }
-    if (state == HTTP_CO_KEEPALIVE && connected_host != host)
+    if (state == HTTP_CO_KEEPALIVE && (connected_host != host || ssl != options.ssl))
     {
         close_connection();
     }
     this->request_timeout = options.timeout < 0 ? 0 : (options.timeout == 0 ? DEFAULT_TIMEOUT : options.timeout);
     this->want_streaming = options.want_streaming;
     this->keepalive = options.keepalive;
+    this->ssl = options.ssl;
     this->host = host;
     this->request = request;
     this->response = "";
@@ -190,7 +307,7 @@ void http_co_t::send_request(const std::string & host, const std::string & reque
             else
             {
                 close_connection();
-                parsed = { .error = "HTTP request timed out" };
+                parsed = { .error = "HTTP request timed out", .status_code = ETIMEDOUT };
                 run_cb_and_clear();
             }
             stackout();
@@ -204,6 +321,11 @@ void http_post_message(http_co_t *handler, uint8_t type, const std::string & msg
     handler->post_message(type, msg);
 }
 
+void http_reply(http_co_t *handler, const std::string & reply)
+{
+    handler->reply(reply);
+}
+
 void http_co_t::post_message(uint8_t type, const std::string & msg)
 {
     stackin();
@@ -213,7 +335,8 @@ void http_co_t::post_message(uint8_t type, const std::string & msg)
         request += msg;
         submit_send();
     }
-    else if (state == HTTP_CO_KEEPALIVE || state == HTTP_CO_CHUNKED)
+    else if (state == HTTP_CO_KEEPALIVE || state == HTTP_CO_CHUNKED ||
+        state == HTTP_CO_SERVER || state == HTTP_CO_REQ_HDR_RECEIVED || state == HTTP_CO_REQUEST_RECEIVED)
     {
         throw std::runtime_error("Attempt to send websocket message on a regular HTTP connection");
     }
@@ -225,12 +348,29 @@ void http_co_t::post_message(uint8_t type, const std::string & msg)
     stackout();
 }
 
-void http_close(http_co_t *handler)
+void http_co_t::reply(const std::string & reply)
+{
+    stackin();
+    if (state != HTTP_CO_REQUEST_RECEIVED)
+    {
+        throw std::runtime_error("Attempt to send HTTP response in invalid connection state");
+    }
+    request += reply;
+    submit_send();
+    stackout();
+}
+
+void http_destroy(http_co_t *handler)
 {
     handler->end();
 }
 
-void http_response_t::parse_json_response(std::string & error, json11::Json & r) const
+void http_close(http_co_t *handler)
+{
+    handler->close_connection();
+}
+
+void http_message_t::parse_json_response(std::string & error, json11::Json & r) const
 {
     if (this->error != "")
     {
@@ -277,6 +417,15 @@ void http_co_t::close_connection()
         close(peer_fd);
         peer_fd = -1;
     }
+#ifdef WITH_OPENSSL
+    if (ssl_cli)
+    {
+        // Frees client and bios at once
+        SSL_free(ssl_cli);
+        ssl_cli = NULL;
+    }
+    ssl_bio = NULL;
+#endif
     state = HTTP_CO_CLOSED;
     connected_host = "";
     response = "";
@@ -295,7 +444,7 @@ void http_co_t::start_ws_connection()
             if (state != HTTP_CO_WEBSOCKET)
             {
                 close_connection();
-                parsed = { .error = "Websocket connection timed out" };
+                parsed = { .error = "Websocket connection timed out", .status_code = ETIMEDOUT };
                 run_cb_and_clear();
             }
             stackout();
@@ -311,7 +460,7 @@ void http_co_t::start_connection()
     if (!string_to_addr(host.c_str(), 1, 80, &addr))
     {
         close_connection();
-        parsed = { .error = "Invalid address: "+host };
+        parsed = { .error = "Invalid address: "+host, .status_code = EINVAL };
         run_cb_and_clear();
         stackout();
         return;
@@ -320,19 +469,56 @@ void http_co_t::start_connection()
     if (peer_fd < 0)
     {
         close_connection();
-        parsed = { .error = std::string("socket: ")+strerror(errno) };
+        parsed = { .error = std::string("socket: ")+strerror(errno), .status_code = errno };
         run_cb_and_clear();
         stackout();
         return;
     }
     fcntl(peer_fd, F_SETFL, fcntl(peer_fd, F_GETFL, 0) | O_NONBLOCK);
     epoll_events = 0;
+#ifdef WITH_OPENSSL
+    // https://wiki.openssl.org/index.php/Hostname_validation
+    if (ssl)
+    {
+        if (!ctx)
+            goto init_err;
+        ssl_bio = BIO_new(BIO_s_socket());
+        if (!ssl_bio)
+            goto init_err;
+        if (!BIO_set_fd(ssl_bio, peer_fd, BIO_NOCLOSE))
+            goto init_err;
+        ssl_cli = SSL_new(ctx->ssl_ctx);
+        if (!ssl_cli)
+            goto init_err;
+        SSL_set_bio(ssl_cli, ssl_bio, ssl_bio);
+        if (!SSL_set_tlsext_host_name(ssl_cli, host.c_str()))
+        {
+init_err:
+            if (ssl_cli)
+            {
+                SSL_free(ssl_cli);
+                ssl_cli = NULL;
+            }
+            else if (ssl_bio)
+            {
+                BIO_free(ssl_bio);
+                ssl_bio = NULL;
+            }
+            parsed = { .error = std::string("openssl initialization failed: ")+ERR_error_string(ERR_get_error(), NULL) };
+            response_callback(&parsed);
+            response_callback = NULL;
+            stackout();
+            return;
+        }
+        SSL_set_connect_state(ssl_cli);
+    }
+#endif
     // Finally call connect
     int r = ::connect(peer_fd, (sockaddr*)&addr, sizeof(addr));
     if (r < 0 && errno != EINPROGRESS)
     {
         close_connection();
-        parsed = { .error = std::string("connect: ")+strerror(errno) };
+        parsed = { .error = std::string("connect: ")+strerror(errno), .status_code = errno };
         run_cb_and_clear();
         stackout();
         return;
@@ -367,6 +553,8 @@ void http_co_t::handle_events()
             {
                 if (state == HTTP_CO_HEADERS_RECEIVED)
                     std::swap(parsed.body, response);
+                else if (state == HTTP_CO_SERVER || state == HTTP_CO_REQ_HDR_RECEIVED || state == HTTP_CO_REQUEST_RECEIVED)
+                    parsed = { .error = "client has disconnected normally" };
                 close_connection();
                 run_cb_and_clear();
                 break;
@@ -388,7 +576,7 @@ void http_co_t::handle_connect_result()
     if (result != 0)
     {
         close_connection();
-        parsed = { .error = std::string("connect: ")+strerror(result) };
+        parsed = { .error = std::string("connect: ")+strerror(result), .status_code = result };
         run_cb_and_clear();
         stackout();
         return;
@@ -408,18 +596,44 @@ void http_co_t::handle_connect_result()
 void http_co_t::submit_send()
 {
     stackin();
-    int res;
+    ssize_t res = 0;
 again:
     if (sent < request.size())
     {
-        send_iov = (iovec){ .iov_base = (void*)(request.c_str()+sent), .iov_len = request.size()-sent };
-        send_msg.msg_iov = &send_iov;
-        send_msg.msg_iovlen = 1;
-        res = sendmsg(peer_fd, &send_msg, MSG_NOSIGNAL);
-        if (res < 0)
+        send_iov = (iovec){ .iov_base = (void*)(request.data()+sent), .iov_len = request.size()-sent };
+#ifdef WITH_OPENSSL
+        if (!ssl)
+#endif
         {
-            res = -errno;
+            send_msg.msg_iov = &send_iov;
+            send_msg.msg_iovlen = 1;
+            res = sendmsg(peer_fd, &send_msg, MSG_NOSIGNAL);
+            if (res < 0)
+                res = -errno;
         }
+#ifdef WITH_OPENSSL
+        else
+        {
+            if (!do_ssl_handshake(false))
+                goto out;
+            int ok = SSL_write_ex(ssl_cli, send_iov.iov_base, send_iov.iov_len, (size_t*)&res);
+            if (!ok)
+            {
+                res = SSL_get_error(ssl_cli, ok);
+                if (res == SSL_ERROR_WANT_WRITE || res == 0)
+                    res = 0;
+                else if (res == SSL_ERROR_WANT_READ)
+                    goto out;
+                else if (res == SSL_ERROR_SYSCALL)
+                    res = -errno;
+                else
+                {
+                    on_ssl_error(res);
+                    goto out;
+                }
+            }
+        }
+#endif
         if (res == -EAGAIN || res == -EINTR)
         {
             res = 0;
@@ -427,13 +641,33 @@ again:
         else if (res < 0)
         {
             close_connection();
-            parsed = { .error = std::string("sendmsg: ")+strerror(errno) };
+            parsed = { .error = std::string("sendmsg: ")+strerror(errno), .status_code = errno };
             run_cb_and_clear();
             stackout();
             return;
         }
         sent += res;
-        if (state == HTTP_CO_SENDING_REQUEST)
+        if (state == HTTP_CO_REQUEST_RECEIVED)
+        {
+            if (sent >= request.size())
+            {
+                if (!keepalive)
+                {
+                    close_connection();
+                    parsed = { .error = "connection is not keep-alive" };
+                    run_cb_and_clear();
+                    stackout();
+                    return;
+                }
+                state = HTTP_CO_SERVER;
+                request = "";
+                sent = 0;
+            }
+            else
+                goto again;
+            handle_read();
+        }
+        else if (state == HTTP_CO_SENDING_REQUEST)
         {
             if (sent >= request.size())
                 state = HTTP_CO_REQUEST_SENT;
@@ -447,26 +681,53 @@ again:
             goto again;
         }
     }
+out:
     stackout();
 }
 
 void http_co_t::submit_read(bool check_timeout)
 {
     stackin();
-    int res;
+    ssize_t res = 0;
 again:
     if (rbuf.size() != READ_BUFFER_SIZE)
     {
         rbuf.resize(READ_BUFFER_SIZE);
     }
     read_iov = { .iov_base = rbuf.data(), .iov_len = READ_BUFFER_SIZE };
-    read_msg.msg_iov = &read_iov;
-    read_msg.msg_iovlen = 1;
-    res = recvmsg(peer_fd, &read_msg, 0);
-    if (res < 0)
+#ifdef WITH_OPENSSL
+    if (!ssl)
+#endif
     {
-        res = -errno;
+        read_msg.msg_iov = &read_iov;
+        read_msg.msg_iovlen = 1;
+        res = recvmsg(peer_fd, &read_msg, 0);
+        if (res < 0)
+            res = -errno;
     }
+#ifdef WITH_OPENSSL
+    else
+    {
+        if (!do_ssl_handshake(true))
+            goto out;
+        int ok = SSL_read_ex(ssl_cli, read_iov.iov_base, read_iov.iov_len, (size_t*)&res);
+        if (!ok)
+        {
+            res = SSL_get_error(ssl_cli, ok);
+            if (res == SSL_ERROR_WANT_READ)
+                res = -EAGAIN;
+            else if (res == SSL_ERROR_SYSCALL)
+                res = -errno;
+            else if (res == SSL_ERROR_ZERO_RETURN)
+                res = 0;
+            else
+            {
+                on_ssl_error(res);
+                goto out;
+            }
+        }
+    }
+#endif
     if (res == -EAGAIN || res == -EINTR)
     {
         if (check_timeout)
@@ -477,7 +738,7 @@ again:
             {
                 // Timeout happened and there is no data to read
                 close_connection();
-                parsed = { .error = "HTTP request timed out" };
+                parsed = { .error = "HTTP request timed out", .status_code = ETIMEDOUT };
                 run_cb_and_clear();
             }
         }
@@ -492,23 +753,76 @@ again:
         epoll_events = epoll_events & ~EPOLLIN;
         if (state == HTTP_CO_HEADERS_RECEIVED)
             std::swap(parsed.body, response);
-        close_connection();
         if (res < 0)
-            parsed = { .error = std::string("recvmsg: ")+strerror(-res) };
+            parsed = { .error = std::string("recvmsg: ")+strerror(-res), .status_code = (int)-res };
+        else if (state == HTTP_CO_SERVER || state == HTTP_CO_REQ_HDR_RECEIVED || state == HTTP_CO_REQUEST_RECEIVED)
+            parsed = { .error = "client has disconnected normally" };
+        close_connection();
         run_cb_and_clear();
     }
     else
     {
-        response += std::string(rbuf.data(), res);
+        response += std::string((char*)rbuf.data(), res);
         handle_read();
     }
+out:
     stackout();
 }
 
+#ifdef WITH_OPENSSL
+void http_co_t::on_ssl_error(int res)
+{
+    close_connection();
+    if (res == SSL_ERROR_ZERO_RETURN)
+    {
+        // Client closed the connection
+        parsed = { .error = "peer closed the SSL connection" };
+    }
+    else
+        parsed = { .error = std::string("SSL error: ")+ERR_error_string(ERR_get_error(), NULL), .status_code = EIO };
+    run_cb_and_clear();
+}
+
+bool http_co_t::do_ssl_handshake(bool init_send)
+{
+    if (SSL_is_init_finished(ssl_cli))
+        return true;
+    int r;
+    while (1)
+    {
+        r = SSL_do_handshake(ssl_cli);
+        if (r > 0)
+        {
+            // OK
+            if (init_send)
+                submit_send();
+            return true;
+        }
+        r = SSL_get_error(ssl_cli, r);
+        if (r == SSL_ERROR_WANT_READ)
+        {
+            break;
+        }
+        else
+        {
+            int errcode = ERR_get_error();
+            parsed = { .error = ERR_error_string(errcode, NULL), .status_code = EIO };
+            close_connection();
+            run_cb_and_clear();
+            return false;
+        }
+    }
+    return false;
+}
+#endif
+
 bool http_co_t::handle_read()
 {
     stackin();
-    if (state == HTTP_CO_REQUEST_SENT)
+    if (state == HTTP_CO_REQUEST_RECEIVED)
+    {
+    }
+    else if (state == HTTP_CO_REQUEST_SENT)
     {
         int pos = response.find("\r\n\r\n");
         if (pos >= 0)
@@ -520,7 +834,7 @@ bool http_co_t::handle_read()
                 timeout_id = -1;
             }
             state = HTTP_CO_HEADERS_RECEIVED;
-            parse_http_headers(response, &parsed);
+            parse_http_headers(response, &parsed, false);
             if (parsed.status_code == 101 &&
                 parsed.headers.find("sec-websocket-accept") != parsed.headers.end() &&
                 parsed.headers["upgrade"] == "websocket" &&
@@ -544,7 +858,7 @@ bool http_co_t::handle_read()
                 {
                     // Sorry, unsupported response
                     close_connection();
-                    parsed = { .error = "Response has neither Connection: close, nor Transfer-Encoding: chunked nor Content-Length headers" };
+                    parsed = { .error = "Response has neither Connection: close, nor Transfer-Encoding: chunked nor Content-Length headers", .status_code = EINVAL };
                     run_cb_and_clear();
                     stackout();
                     return false;
@@ -556,14 +870,57 @@ bool http_co_t::handle_read()
             }
         }
     }
-    if (state == HTTP_CO_HEADERS_RECEIVED && target_response_size > 0 && response.size() >= target_response_size)
+    else if (state == HTTP_CO_SERVER)
+    {
+        int pos = response.find("\r\n\r\n");
+        if (pos >= 0)
+        {
+            if (timeout_id >= 0)
+            {
+                // Timeout is cleared when headers are received
+                tfd->clear_timer(timeout_id);
+                timeout_id = -1;
+            }
+            state = HTTP_CO_REQ_HDR_RECEIVED;
+            parse_http_headers(response, &parsed, true);
+            auto conn_it = parsed.headers.find("connection");
+            keepalive = (conn_it != parsed.headers.end() && conn_it->second == "keep-alive");
+            auto enc_it = parsed.headers.find("transfer-encoding");
+            if (enc_it != parsed.headers.end())
+            {
+                // Sorry, unsupported request
+                close_connection();
+                parsed = { .error = "Chunked requests are not supported", .status_code = EINVAL };
+                run_cb_and_clear();
+                stackout();
+                return false;
+            }
+            auto len_it = parsed.headers.find("content-length");
+            target_response_size = stoull_full(len_it != parsed.headers.end() ? len_it->second : "");
+            if (!target_response_size)
+            {
+                state = HTTP_CO_REQUEST_RECEIVED;
+                response_callback(&parsed);
+            }
+        }
+    }
+    if ((state == HTTP_CO_HEADERS_RECEIVED || state == HTTP_CO_REQ_HDR_RECEIVED) &&
+        target_response_size > 0 && response.size() >= target_response_size)
     {
         std::swap(parsed.body, response);
-        if (!keepalive)
-            close_connection();
+        if (state == HTTP_CO_REQ_HDR_RECEIVED)
+        {
+            state = HTTP_CO_REQUEST_RECEIVED;
+            response_callback(&parsed);
+        }
         else
-            state = HTTP_CO_KEEPALIVE;
-        run_cb_and_clear();
+        {
+            if (!keepalive)
+                close_connection();
+            else
+                state = HTTP_CO_KEEPALIVE;
+            run_cb_and_clear();
+        }
     }
     else if (state == HTTP_CO_CHUNKED && response.size() > 0)
     {
@@ -626,26 +983,34 @@ void http_co_t::next_request()
 {
     if (keepalive_queue.size() > 0)
     {
-        auto next = keepalive_queue[0];
-        keepalive_queue.erase(keepalive_queue.begin(), keepalive_queue.begin()+1);
-        next();
+        auto next = std::move(keepalive_queue[0]);
+        keepalive_queue.erase(keepalive_queue.begin());
+        send_request(next.host, next.request, next.options, next.cb);
     }
 }
 
-static void parse_http_headers(std::string & res, http_response_t *parsed)
+static void parse_http_headers(std::string & res, http_message_t *parsed, bool is_request)
 {
     int pos = res.find("\r\n");
     pos = pos < 0 ? res.length() : pos+2;
     std::string status_line = res.substr(0, pos);
     int http_version;
     char *status_text = NULL;
-    sscanf(status_line.c_str(), "HTTP/1.%d %d %ms", &http_version, &parsed->status_code, &status_text);
-    if (status_text)
+    if (!is_request)
     {
-        parsed->status_line = status_text;
-        // %ms = allocate a buffer
-        free(status_text);
-        status_text = NULL;
+        sscanf(status_line.c_str(), "HTTP/1.%d %d %ms", &http_version, &parsed->status_code, &status_text);
+        if (status_text)
+        {
+            parsed->status_line = status_text;
+            // %ms = allocate a buffer
+            free(status_text);
+            status_text = NULL;
+        }
+    }
+    else
+    {
+        // Should be GET/POST / HTTP/1.1
+        parsed->status_line = status_line;
     }
     int prev = pos;
     while ((pos = res.find("\r\n", prev)) >= prev)

src/client/http_client.h

@@ -17,14 +17,19 @@
 
 class timerfd_manager_t;
 
+#pragma GCC visibility push(default)
+
 struct http_options_t
 {
     int timeout;
     bool want_streaming;
     bool keepalive;
+    bool ssl;
 };
 
-struct http_response_t
+struct http_context_t;
+
+struct http_message_t
 {
     std::string error;
 
@@ -41,10 +46,19 @@ struct http_response_t
 // Opened websocket or keepalive HTTP connection
 struct http_co_t;
 
-http_co_t* http_init(timerfd_manager_t *tfd);
-http_co_t* open_websocket(timerfd_manager_t *tfd, const std::string & host, const std::string & path,
-    int timeout, std::function<void(const http_response_t *msg)> on_message);
+http_context_t* http_context_init(const std::string & ssl_cert, const std::string & ssl_key,
+    const std::string & ssl_ca, bool verify_peer, std::string & error);
+void http_context_destroy(http_context_t *ctx);
+http_co_t* http_init(timerfd_manager_t *tfd, http_context_t *ctx = NULL);
+void open_websocket(http_co_t *handler, const std::string & host, const std::string & path,
+    const http_options_t & options, std::function<void(http_message_t *msg)> on_message);
 void http_request(http_co_t *handler, const std::string & host, const std::string & request,
-    const http_options_t & options, std::function<void(const http_response_t *response)> response_callback);
+    const http_options_t & options, std::function<void(http_message_t *response)> response_callback);
 void http_post_message(http_co_t *handler, uint8_t type, const std::string & msg);
+void http_serve(http_co_t *handler, int peer_fd, const http_options_t & options,
+    std::function<void(http_message_t *msg)> request_callback);
+void http_reply(http_co_t *handler, const std::string & reply);
 void http_close(http_co_t *co);
+void http_destroy(http_co_t *co);
+
+#pragma GCC visibility pop

src/client/messenger.cpp

@@ -289,6 +289,16 @@ osd_messenger_t::~osd_messenger_t()
         rdmacm_evch = NULL;
     }
 #endif
+#ifdef WITH_OPENSSL
+    for (auto encrypt_ctx: encrypt_ctx_pool)
+    {
+        destroy_aes_xts_encrypt(encrypt_ctx);
+    }
+    for (auto decrypt_ctx: decrypt_ctx_pool)
+    {
+        destroy_aes_xts_decrypt(decrypt_ctx);
+    }
+#endif
 }
 
 void osd_messenger_t::parse_config(const json11::Json & config)
@@ -323,6 +333,9 @@ void osd_messenger_t::parse_config(const json11::Json & config)
     if (!this->rdma_max_msg || this->rdma_max_msg > 128*1024*1024)
         this->rdma_max_msg = 129*1024;
 #endif
+    this->max_aes_xts_pool_size = config["max_aes_xts_pool_size"].uint64_value();
+    if (!this->max_aes_xts_pool_size)
+        this->max_aes_xts_pool_size = 256;
     if (!osd_num)
         this->iothread_count = (uint32_t)config["client_iothread_count"].uint64_value();
     else
@@ -502,7 +515,7 @@ void osd_messenger_t::try_connect_peer_tcp(osd_num_t peer_osd, const char *peer_
     clients[peer_fd]->peer_state = PEER_CONNECTING;
     clients[peer_fd]->connect_timeout_id = -1;
     clients[peer_fd]->osd_num = peer_osd;
-    clients[peer_fd]->in_buf = malloc_or_die(receive_buffer_size);
+    clients[peer_fd]->in_buf = (uint8_t*)malloc_or_die(receive_buffer_size);
     tfd->set_fd_handler(peer_fd, true, [this](int peer_fd, int epoll_events)
     {
         // Either OUT (connected) or HUP
@@ -739,6 +752,14 @@ void osd_messenger_t::check_peer_config(osd_client_t *cl)
                     fprintf(stderr, "Connected to OSD %ju using RDMA\n", cl->osd_num);
                 }
                 cl->peer_state = PEER_RDMA;
+                tfd->set_fd_handler(cl->peer_fd, false, [this](int peer_fd, int epoll_events)
+                {
+                    // Do not miss the disconnection!
+                    if (epoll_events & EPOLLRDHUP)
+                    {
+                        handle_peer_epoll(peer_fd, epoll_events);
+                    }
+                });
                 // Add the initial receive request
                 init_recv_rdma(cl);
             }
@@ -773,7 +794,7 @@ void osd_messenger_t::accept_connections(int listen_fd)
         cl->peer_port = ntohs(((sockaddr_in*)&addr)->sin_port);
         cl->peer_fd = peer_fd;
         cl->peer_state = PEER_CONNECTED;
-        cl->in_buf = malloc_or_die(receive_buffer_size);
+        cl->in_buf = (uint8_t*)malloc_or_die(receive_buffer_size);
         // Add FD to epoll
         tfd->set_fd_handler(peer_fd, false, [this](int peer_fd, int epoll_events)
         {

src/client/messenger.h

@@ -34,9 +34,6 @@
 
 #define DEFAULT_MIN_ZEROCOPY_SEND_SIZE 32*1024
 
-#define MSGR_SENDP_HDR 1
-#define MSGR_SENDP_FREE 2
-
 struct msgr_sendp_t
 {
     osd_op_t *op;
@@ -48,6 +45,13 @@ struct msgr_rdma_connection_t;
 struct msgr_rdma_context_t;
 #endif
 
+#ifdef WITH_OPENSSL
+struct op_aes_xts_encrypt_t;
+struct op_aes_xts_decrypt_t;
+void destroy_aes_xts_encrypt(op_aes_xts_encrypt_t *encrypt_ctx);
+void destroy_aes_xts_decrypt(op_aes_xts_decrypt_t *decrypt_ctx);
+#endif
+
 struct osd_client_t
 {
     int refs = 0;
@@ -63,20 +67,23 @@ struct osd_client_t
     osd_num_t in_osd_num = 0;
     bool is_incoming = false;
 
-    void *in_buf = NULL;
+    uint8_t *in_buf = NULL;
 
 #ifdef WITH_RDMA
     msgr_rdma_connection_t *rdma_conn = NULL;
 #endif
 
     // Read state
+    op_aes_xts_decrypt_t *decrypt_ctx = NULL;
     int read_ready = 0;
     osd_op_t *read_op = NULL;
+    size_t read_op_size = 0;
+    size_t read_op_pos = 0;
+    size_t read_op_inline_decrypt_pos = 0;
     iovec read_iov = { 0 };
     msghdr read_msg = { 0 };
-    int read_remaining = 0;
-    int read_state = 0;
-    osd_op_buf_list_t recv_list;
+    std::vector<iovec> recv_list;
+    size_t recv_list_size = 0;
     uint64_t read_op_id = 1;
     bool check_sequencing = false;
     bool enable_pg_locks = false;
@@ -92,10 +99,15 @@ struct osd_client_t
     std::set<pool_pg_num_t> dirty_pgs;
 
     // Write state
+    op_aes_xts_encrypt_t *encrypt_ctx = NULL;
+    std::deque<osd_op_t *> write_ops;
+    osd_op_t *write_op = NULL;
+    size_t write_op_pos = 0;
     msghdr write_msg = { 0 };
     int write_state = 0;
-    std::vector<iovec> send_list, next_send_list;
-    std::vector<msgr_sendp_t> outbox, next_outbox;
+    std::vector<iovec> send_list;
+    size_t send_list_size = 0;
+    std::deque<osd_op_t*> send_free_ops;
     std::vector<osd_op_t*> zc_free_list;
 
     ~osd_client_t();
@@ -188,6 +200,7 @@ protected:
     bool use_sync_send_recv = false;
     int min_zerocopy_send_size = DEFAULT_MIN_ZEROCOPY_SEND_SIZE;
     int iothread_count = 0;
+    int max_aes_xts_pool_size = 256;
 
 #ifdef WITH_RDMA
     bool use_rdma = true;
@@ -209,7 +222,12 @@ protected:
     std::vector<int> read_ready_clients;
     std::vector<int> write_ready_clients;
     // We don't use ringloop->set_immediate here because we may have no ringloop in client :)
-    std::deque<osd_op_t*> set_immediate_ops;
+    std::vector<osd_op_t*> set_immediate_ops;
+
+#ifdef WITH_OPENSSL
+    std::vector<op_aes_xts_encrypt_t*> encrypt_ctx_pool;
+    std::vector<op_aes_xts_decrypt_t*> decrypt_ctx_pool;
+#endif
 
 public:
     timerfd_manager_t *tfd = NULL;
@@ -274,9 +292,23 @@ protected:
 
     bool try_send(osd_client_t *cl);
     void handle_send(int result, bool prev, bool more, osd_client_t *cl);
+    bool op_encrypted_copy_data_to(osd_client_t* cl, uint8_t *buf, size_t len, size_t from, size_t & done);
+    size_t op_copy_to(osd_client_t *cl, uint8_t *dst, size_t dst_len);
+    void op_get_write_buffers(osd_client_t *cl, std::vector<iovec> & lst);
+
+    void handle_read(int result, osd_client_t *cl);
+    bool handle_read_buffer(osd_client_t *cl, uint8_t *curbuf, size_t bufsize);
+    bool handle_hdr(osd_client_t *cl);
+    bool allocate_op_buffers(osd_client_t *cl);
+    bool allocate_reply_buffers(osd_client_t *cl, osd_op_t *op);
+    size_t op_copy_from(osd_client_t *cl, uint8_t *src, size_t src_len, size_t & done);
+    bool op_decrypted_copy_data_from(osd_client_t* cl, uint8_t *buf, size_t len, size_t from, size_t & done);
+    void op_decrypt_start(osd_client_t* cl);
+    void op_decrypt_inline(osd_client_t* cl);
+    void op_decrypt_free(osd_client_t* cl);
+    size_t op_get_read_buffers(osd_client_t *cl, std::vector<iovec> & lst);
+    void handle_finished_op(osd_client_t *cl);
 
-    bool handle_read(int result, osd_client_t *cl);
-    bool handle_read_buffer(osd_client_t *cl, void *curbuf, int remain);
     bool handle_finished_read(osd_client_t *cl);
     void handle_op_hdr(osd_client_t *cl);
     bool handle_reply_hdr(osd_client_t *cl);
@@ -286,6 +318,7 @@ protected:
 
 #ifdef WITH_RDMA
     void try_send_rdma(osd_client_t *cl);
+    int try_send_rdma_copy(osd_client_t *cl, uint8_t *dst, int dst_len);
     bool init_recv_rdma(osd_client_t *cl);
     void handle_rdma_events(msgr_rdma_context_t *rdma_context);
     msgr_rdma_context_t* choose_rdma_context(osd_client_t *cl);

src/client/msgr_encrypt.cpp

@@ -0,0 +1,328 @@
+// Copyright (c) Vitaliy Filippov, 2026+
+// License: VNPL-1.1 or GNU GPL-2.0+ (see README.md for details)
+
+#define _XOPEN_SOURCE
+#include <limits.h>
+#include <assert.h>
+
+#include "etcd_state_client.h"
+#include "messenger.h"
+#include "msgr_encrypt.h"
+
+// FIXME Fuck, no streaming...
+
+op_aes_xts_encrypt_t::op_aes_xts_encrypt_t()
+{
+    if (!(ctx = EVP_CIPHER_CTX_new()))
+    {
+        ERR_print_errors_fp(stderr);
+        abort();
+    }
+    EVP_CIPHER_CTX_set_padding(ctx, 0);
+    if (EVP_EncryptInit_ex(ctx, EVP_aes_256_xts(), NULL, NULL, NULL) != 1)
+    {
+        ERR_print_errors_fp(stderr);
+        abort();
+    }
+}
+
+op_aes_xts_encrypt_t::~op_aes_xts_encrypt_t()
+{
+    EVP_CIPHER_CTX_free(ctx);
+}
+
+void op_aes_xts_encrypt_t::start(const uint8_t *key, uint64_t start_offset, size_t block_size)
+{
+    this->start_offset = start_offset;
+    this->key = key;
+    this->block_size = block_size;
+    this->offset = 0;
+}
+
+void op_aes_xts_encrypt_t::update(uint8_t *in, size_t max_in, uint8_t *out, size_t max_out, size_t & done_in, size_t & done_out)
+{
+    if (max_in > block_size - offset%block_size)
+        max_in = block_size - offset%block_size;
+    size_t insize = max_in;
+    size_t outsize = ((offset+insize)/16 - offset/16) * 16;
+    if (outsize > max_out)
+    {
+        // encrypt is used to send data through temporary buffer(s),
+        // so we don't care to support fragmenting output into < 16 b parts
+        insize = (max_out < 16 ? 0 : (max_out & ~15) - offset%16);
+        outsize = ((offset+insize)/16 - offset/16) * 16;
+    }
+    assert(insize <= max_in);
+    assert(outsize <= max_out);
+    if (!(offset % block_size))
+    {
+        uint8_t iv[16] = { 0 };
+        *((uint64_t*)iv) = start_offset + offset;
+        if (EVP_EncryptInit_ex(ctx, NULL, NULL, key, iv) != 1)
+        {
+            ERR_print_errors_fp(stderr);
+            abort();
+        }
+    }
+    int actual_out = 0;
+    if (EVP_EncryptUpdate(ctx, out, &actual_out, in, insize) != 1)
+    {
+        ERR_print_errors_fp(stderr);
+        abort();
+    }
+    assert(actual_out == outsize);
+    done_in += insize;
+    done_out += outsize;
+    offset += insize;
+}
+
+void destroy_aes_xts_encrypt(op_aes_xts_encrypt_t *encrypt_ctx)
+{
+    delete encrypt_ctx;
+}
+
+op_aes_xts_decrypt_t::op_aes_xts_decrypt_t()
+{
+    if (!(ctx = EVP_CIPHER_CTX_new()))
+    {
+        ERR_print_errors_fp(stderr);
+        abort();
+    }
+    EVP_CIPHER_CTX_set_padding(ctx, 0);
+    if (EVP_DecryptInit_ex(ctx, EVP_aes_256_xts(), NULL, NULL, NULL) != 1)
+    {
+        ERR_print_errors_fp(stderr);
+        abort();
+    }
+}
+
+op_aes_xts_decrypt_t::~op_aes_xts_decrypt_t()
+{
+    EVP_CIPHER_CTX_free(ctx);
+}
+
+void op_aes_xts_decrypt_t::start(const uint8_t *key, uint64_t start_offset, size_t block_size)
+{
+    this->start_offset = start_offset;
+    this->key = key;
+    this->block_size = block_size;
+    this->in_offset = 0;
+    this->tmp_pos = 16;
+}
+
+void op_aes_xts_decrypt_t::update(uint8_t *in, size_t max_in, uint8_t *out, size_t max_out, size_t & done_in, size_t & done_out)
+{
+    if (max_in > block_size - in_offset%block_size)
+        max_in = block_size - in_offset%block_size;
+    int actual_out = 0;
+    // Write previously buffered block to support small output buffers
+    if (tmp_pos < 16)
+    {
+        size_t tmp_size = 16-tmp_pos;
+        if (tmp_size > max_out)
+            tmp_size = max_out;
+        memcpy(out, tmp_buf+tmp_pos, tmp_size);
+        tmp_pos += tmp_size;
+        done_out += tmp_size;
+        out += tmp_size;
+        max_out -= tmp_size;
+        if (!max_out)
+            return;
+        assert(tmp_pos == 16);
+    }
+    if (!(in_offset % block_size))
+    {
+        uint8_t iv[16] = { 0 };
+        *((uint64_t*)iv) = start_offset+in_offset;
+        if (EVP_DecryptInit_ex(ctx, NULL, NULL, key, iv) != 1)
+        {
+            ERR_print_errors_fp(stderr);
+            abort();
+        }
+    }
+    size_t insize = max_in;
+    size_t outsize = ((in_offset+insize)/16 - in_offset/16) * 16;
+    if (outsize > max_out)
+    {
+        if (max_out < 16)
+        {
+            // We can only decrypt a partial block
+            insize = 16 - in_offset%16;
+            if (EVP_DecryptUpdate(ctx, tmp_buf, &actual_out, in, insize) != 1)
+            {
+                ERR_print_errors_fp(stderr);
+                abort();
+            }
+            assert(actual_out == 16);
+            in_offset += insize;
+            in += insize;
+            max_in -= insize;
+            tmp_pos = 0;
+            return;
+        }
+        // Otherwise, we can decrypt at least some data into <out> directly
+        insize = (max_out & ~15) - (in_offset % 16);
+        outsize = (max_out & ~15);
+        assert(insize < max_in);
+    }
+    if (EVP_DecryptUpdate(ctx, out, &actual_out, in, insize) != 1)
+    {
+        ERR_print_errors_fp(stderr);
+        abort();
+    }
+    assert(actual_out == outsize);
+    in_offset += insize;
+    done_in += insize;
+    done_out += actual_out;
+}
+
+void destroy_aes_xts_decrypt(op_aes_xts_decrypt_t *decrypt_ctx)
+{
+    delete decrypt_ctx;
+}
+
+bool osd_messenger_t::op_encrypted_copy_data_to(osd_client_t* cl, uint8_t *enc_buf, size_t enc_len, size_t from, size_t & done)
+{
+    auto op = cl->write_op;
+    auto & op_pos = cl->write_op_pos;
+    assert(op->req.hdr.opcode == OSD_OP_WRITE);
+    if (!from)
+    {
+        if (!cl->encrypt_ctx)
+        {
+            if (encrypt_ctx_pool.size())
+            {
+                cl->encrypt_ctx = encrypt_ctx_pool.back();
+                encrypt_ctx_pool.pop_back();
+            }
+            else
+                cl->encrypt_ctx = new op_aes_xts_encrypt_t();
+        }
+        assert(op->enc->key.size() == 512/8);
+        cl->encrypt_ctx->start(op->enc->key.data(), op->req.rw.offset, op->enc->bitmap_granularity);
+    }
+    for (int i = 0; i < op->iov.count; i++)
+    {
+        uint8_t *plain = (uint8_t*)op->iov.buf[i].iov_base;
+        size_t plain_len = op->iov.buf[i].iov_len;
+        while (from < plain_len)
+        {
+            size_t done_in = 0;
+            size_t done_out = 0;
+            cl->encrypt_ctx->update(plain+from, plain_len-from, enc_buf+done, enc_len-done, done_in, done_out);
+            done += done_out;
+            op_pos += done_in;
+            from += done_in;
+            if (!done_in)
+                return false;
+        }
+        from -= plain_len;
+    }
+    if (cl->encrypt_ctx)
+    {
+        if (encrypt_ctx_pool.size() > max_aes_xts_pool_size)
+            delete cl->encrypt_ctx;
+        else
+            encrypt_ctx_pool.push_back(cl->encrypt_ctx);
+        cl->encrypt_ctx = NULL;
+    }
+    return true;
+}
+
+bool osd_messenger_t::op_decrypted_copy_data_from(osd_client_t* cl, uint8_t *enc_buf, size_t enc_len, size_t from, size_t & done)
+{
+    op_decrypt_start(cl);
+    auto op = cl->read_op;
+    auto & op_pos = cl->read_op_pos;
+    assert(op->req.hdr.opcode == OSD_OP_READ);
+    uint64_t offset = from;
+    for (int i = 0; i < op->iov.count; i++)
+    {
+        uint8_t *plain = (uint8_t*)op->iov.buf[i].iov_base;
+        size_t plain_len = op->iov.buf[i].iov_len;
+        while (from < plain_len)
+        {
+            size_t done_in = 0;
+            size_t done_out = 0;
+            cl->decrypt_ctx->update(enc_buf+done, enc_len-done, plain+from, plain_len-from, done_in, done_out);
+            done += done_in;
+            offset += done_in;
+            op_pos += done_out;
+            from += done_out;
+            if (!done_in)
+                return false;
+        }
+        from -= plain_len;
+    }
+    op_decrypt_free(cl);
+    return true;
+}
+
+void osd_messenger_t::op_decrypt_start(osd_client_t* cl)
+{
+    if (!cl->decrypt_ctx)
+    {
+        if (decrypt_ctx_pool.size())
+        {
+            cl->decrypt_ctx = decrypt_ctx_pool.back();
+            decrypt_ctx_pool.pop_back();
+        }
+        else
+            cl->decrypt_ctx = new op_aes_xts_decrypt_t();
+        assert(cl->read_op->enc->key.size() == 512/8);
+        cl->decrypt_ctx->start(cl->read_op->enc->key.data(), cl->read_op->req.rw.offset, cl->read_op->enc->bitmap_granularity);
+    }
+}
+
+void osd_messenger_t::op_decrypt_inline(osd_client_t* cl)
+{
+    op_decrypt_start(cl);
+    osd_op_t *op = cl->read_op;
+    size_t from_in = cl->read_op_inline_decrypt_pos - OSD_PACKET_SIZE - op->reply.rw.bitmap_len;
+    int i = 0;
+    while (i < op->iov.count && from_in >= op->iov.buf[i].iov_len)
+    {
+        from_in -= op->iov.buf[i].iov_len;
+        i++;
+    }
+    size_t from_out = from_in;
+    int j = i;
+    while (i < op->iov.count && j < op->iov.count)
+    {
+        uint8_t *in = (uint8_t*)op->iov.buf[i].iov_base + from_in;
+        size_t in_len = op->iov.buf[i].iov_len - from_in;
+        uint8_t *out = (uint8_t*)op->iov.buf[j].iov_base + from_out;
+        size_t out_len = op->iov.buf[j].iov_len - from_out;
+        size_t done_in = 0;
+        size_t done_out = 0;
+        cl->decrypt_ctx->update(in, in_len, out, out_len, done_in, done_out);
+        if (done_in >= in_len)
+        {
+            i++;
+            from_in = 0;
+        }
+        else
+            from_in += done_in;
+        if (done_out >= out_len)
+        {
+            j++;
+            from_out = 0;
+        }
+        else
+            from_out += done_out;
+    }
+    assert(j >= op->iov.count);
+    op_decrypt_free(cl);
+}
+
+void osd_messenger_t::op_decrypt_free(osd_client_t* cl)
+{
+    if (cl->decrypt_ctx)
+    {
+        if (decrypt_ctx_pool.size() > max_aes_xts_pool_size)
+            delete cl->decrypt_ctx;
+        else
+            decrypt_ctx_pool.push_back(cl->decrypt_ctx);
+        cl->decrypt_ctx = NULL;
+    }
+}

src/client/msgr_encrypt.h

@@ -0,0 +1,44 @@
+// Copyright (c) Vitaliy Filippov, 2026+
+// License: VNPL-1.1 or GNU GPL-2.0+ (see README.md for details)
+
+#include <stdint.h>
+
+#include <openssl/conf.h>
+#include <openssl/evp.h>
+#include <openssl/err.h>
+
+struct op_aes_xts_encrypt_t
+{
+    EVP_CIPHER_CTX *ctx = NULL;
+    uint64_t start_offset = 0;
+    const uint8_t *key = NULL;
+    size_t offset = 0;
+    size_t block_size = 0;
+
+    op_aes_xts_encrypt_t();
+    ~op_aes_xts_encrypt_t();
+
+    void start(const uint8_t *key, uint64_t start_offset, size_t block_size);
+    void update(uint8_t *in, size_t max_in, uint8_t *out, size_t max_out, size_t & done_in, size_t & done_out);
+};
+
+void destroy_aes_xts_encrypt(op_aes_xts_encrypt_t *encrypt_ctx);
+
+struct op_aes_xts_decrypt_t
+{
+    EVP_CIPHER_CTX *ctx = NULL;
+    uint64_t start_offset = 0;
+    const uint8_t *key = NULL;
+    uint8_t tmp_buf[16];
+    size_t tmp_pos = 16;
+    size_t in_offset = 0;
+    size_t block_size = 0;
+
+    op_aes_xts_decrypt_t();
+    ~op_aes_xts_decrypt_t();
+
+    void start(const uint8_t *key, uint64_t start_offset, size_t block_size);
+    void update(uint8_t *in, size_t max_in, uint8_t *out, size_t max_out, size_t & done_in, size_t & done_out);
+};
+
+void destroy_aes_xts_decrypt(op_aes_xts_decrypt_t *decrypt_ctx);

src/client/msgr_op.cpp

@@ -23,6 +23,10 @@ osd_op_t::~osd_op_t()
         // So we don't reuse it, but free it every time
         free(buf);
     }
+    if (enc_buf)
+    {
+        free(enc_buf);
+    }
 }
 
 bool osd_op_t::is_recovery_related()

src/client/msgr_op.h

@@ -3,6 +3,8 @@
 
 #pragma once
 
+#include <memory>
+
 #include <sys/uio.h>
 #include <stdint.h>
 #include <stdio.h>
@@ -152,6 +154,8 @@ struct blockstore_op_t;
 
 struct osd_primary_op_data_t;
 
+struct inode_enc_t;
+
 struct __attribute__((visibility("default"))) osd_op_t
 {
     timespec tv_begin = { 0 }, tv_end = { 0 };
@@ -167,6 +171,8 @@ struct __attribute__((visibility("default"))) osd_op_t
     unsigned bmp_data = 0;
     void *bitmap_buf = NULL;
     void *rmw_buf = NULL;
+    std::shared_ptr<inode_enc_t> enc;
+    uint8_t *enc_buf = NULL;
     osd_primary_op_data_t* op_data = NULL;
     std::function<void(osd_op_t*)> callback;
 

src/client/msgr_rdma.cpp

@@ -555,23 +555,28 @@ static void try_send_rdma_wr(osd_client_t *cl, ibv_sge *sge, int op_sge)
     cl->rdma_conn->cur_send++;
 }
 
-static int try_send_rdma_copy(osd_client_t *cl, uint8_t *dst, int dst_len)
+int osd_messenger_t::try_send_rdma_copy(osd_client_t *cl, uint8_t *dst, int dst_len)
 {
-    auto rc = cl->rdma_conn;
     int total_dst_len = dst_len;
-    while (dst_len > 0 && rc->send_pos < cl->send_list.size())
+    while (dst_len > 0 && cl->write_ops.size())
     {
-        iovec & iov = cl->send_list[rc->send_pos];
-        uint32_t len = (uint32_t)(iov.iov_len-rc->send_buf_pos < dst_len
-            ? iov.iov_len-rc->send_buf_pos : dst_len);
-        memcpy(dst, (uint8_t*)iov.iov_base+rc->send_buf_pos, len);
-        dst += len;
-        dst_len -= len;
-        rc->send_buf_pos += len;
-        if (rc->send_buf_pos >= iov.iov_len)
+        if (!cl->write_op)
         {
-            rc->send_pos++;
-            rc->send_buf_pos = 0;
+            cl->write_op = cl->write_ops.front();
+            cl->write_ops.pop_front();
+        }
+        osd_op_t *op = cl->write_op;
+        size_t copied = op_copy_to(cl, dst, dst_len);
+        if (!copied)
+        {
+            break;
+        }
+        dst += copied;
+        dst_len -= copied;
+        if (!cl->write_op && op->op_type == OSD_OP_IN)
+        {
+            // this is a reply, free the op after sending it
+            cl->send_free_ops.push_back(op);
         }
     }
     return total_dst_len-dst_len;
@@ -617,7 +622,7 @@ void osd_messenger_t::try_send_rdma(osd_client_t *cl)
                 .lkey = rc->send_out.mr->lkey,
             };
             try_send_rdma_wr(cl, &sge, 1);
-            rc->send_sizes.push_back(copied);
+            cl->send_free_ops.push_back(NULL); // end marker
         }
     }
 }
@@ -696,10 +701,6 @@ void osd_messenger_t::handle_rdma_events(msgr_rdma_context_t *rdma_context)
                 continue;
             }
             osd_client_t *cl = cl_it->second;
-            if (cl->peer_state == PEER_STOPPED)
-            {
-                continue;
-            }
             auto rc = cl->rdma_conn;
             if (wc[i].status != IBV_WC_SUCCESS)
             {
@@ -731,51 +732,18 @@ void osd_messenger_t::handle_rdma_events(msgr_rdma_context_t *rdma_context)
             else
             {
                 rc->cur_send--;
-                uint64_t sent_size = rc->send_sizes.at(0);
-                rc->send_sizes.erase(rc->send_sizes.begin(), rc->send_sizes.begin()+1);
+                uint64_t sent_size = wc[i].byte_len;
                 rc->send_done_pos += sent_size;
                 rc->send_out_full = false;
                 if (rc->send_done_pos == rc->send_out_size)
                     rc->send_done_pos = 0;
                 assert(rc->send_done_pos < rc->send_out_size);
-                int send_pos = 0, send_buf_pos = 0;
-                while (sent_size > 0)
+                while (cl->send_free_ops.front())
                 {
-                    if (sent_size >= cl->send_list.at(send_pos).iov_len)
-                    {
-                        sent_size -= cl->send_list[send_pos].iov_len;
-                        send_pos++;
-                    }
-                    else
-                    {
-                        send_buf_pos = sent_size;
-                        sent_size = 0;
-                    }
-                }
-                assert(rc->send_pos >= send_pos);
-                if (rc->send_pos == send_pos)
-                {
-                    rc->send_buf_pos -= send_buf_pos;
-                }
-                rc->send_pos -= send_pos;
-                for (int i = 0; i < send_pos; i++)
-                {
-                    if (cl->outbox[i].flags & MSGR_SENDP_FREE)
-                    {
-                        // Reply fully sent
-                        delete cl->outbox[i].op;
-                    }
-                }
-                if (send_pos > 0)
-                {
-                    cl->send_list.erase(cl->send_list.begin(), cl->send_list.begin()+send_pos);
-                    cl->outbox.erase(cl->outbox.begin(), cl->outbox.begin()+send_pos);
-                }
-                if (send_buf_pos > 0)
-                {
-                    cl->send_list[0].iov_base = (uint8_t*)cl->send_list[0].iov_base + send_buf_pos;
-                    cl->send_list[0].iov_len -= send_buf_pos;
+                    delete cl->send_free_ops.front();
+                    cl->send_free_ops.pop_front();
                 }
+                cl->send_free_ops.pop_front();
                 try_send_rdma(cl);
             }
         }

src/client/msgr_rdma.h

@@ -10,6 +10,8 @@
 #include <vector>
 #include "addr_util.h"
 
+struct osd_op_t;
+
 struct msgr_rdma_address_t
 {
     ibv_gid gid;
@@ -72,9 +74,8 @@ struct msgr_rdma_connection_t
     int cur_send = 0, cur_recv = 0;
     int send_pos = 0, send_buf_pos = 0;
     int next_recv_buf = 0;
-    std::vector<void*> recv_buffers;
+    std::vector<uint8_t*> recv_buffers;
     msgr_rdma_buf_t recv_buf;
-    std::vector<uint64_t> send_sizes;
     msgr_rdma_buf_t send_out;
     int send_out_pos = 0, send_done_pos = 0, send_out_size = 0;
     bool send_out_full = false;

src/client/msgr_rdmacm.cpp

@@ -515,7 +515,7 @@ void osd_messenger_t::rdmacm_established(rdma_cm_event *ev)
     cl->peer_state = PEER_RDMA;
     cl->connect_timeout_id = -1;
     cl->osd_num = peer_osd;
-    cl->in_buf = malloc_or_die(receive_buffer_size);
+    cl->in_buf = (uint8_t*)malloc_or_die(receive_buffer_size);
     cl->rdma_conn = rc;
     clients[conn->peer_fd] = cl;
     if (conn->timeout_id >= 0)

src/client/msgr_receive.cpp

@@ -1,6 +1,8 @@
 // Copyright (c) Vitaliy Filippov, 2019+
 // License: VNPL-1.1 or GNU GPL-2.0+ (see README.md for details)
 
+#define _XOPEN_SOURCE
+#include <limits.h>
 #include "messenger.h"
 
 void osd_messenger_t::read_requests()
@@ -9,13 +11,16 @@ void osd_messenger_t::read_requests()
     {
         int peer_fd = read_ready_clients[i];
         auto cl_it = clients.find(peer_fd);
-        if (cl_it == clients.end() || !cl_it->second || cl_it->second->read_msg.msg_iovlen ||
-            cl_it->second->peer_state == PEER_RDMA || cl_it->second->peer_state == PEER_RDMA_CONNECTING)
+        if (cl_it == clients.end() || !cl_it->second || cl_it->second->read_msg.msg_iovlen)
         {
             continue;
         }
         auto cl = cl_it->second;
-        if (cl->read_remaining < receive_buffer_size)
+        if (cl->read_op && cl->read_op_size-(cl->read_op_pos-OSD_PACKET_SIZE) >= receive_buffer_size)
+        {
+            op_get_read_buffers(cl, cl->recv_list);
+        }
+        if (!cl->recv_list.size())
         {
             cl->read_iov.iov_base = cl->in_buf;
             cl->read_iov.iov_len = receive_buffer_size;
@@ -25,10 +30,11 @@ void osd_messenger_t::read_requests()
         else
         {
             cl->read_iov.iov_base = 0;
-            cl->read_iov.iov_len = cl->read_remaining;
-            cl->read_msg.msg_iov = cl->recv_list.get_iovec();
-            cl->read_msg.msg_iovlen = cl->recv_list.get_size();
+            cl->read_iov.iov_len = 0;
+            cl->read_msg.msg_iov = cl->recv_list.data();
+            cl->read_msg.msg_iovlen = cl->recv_list.size();
         }
+        assert(!cl->read_op || cl->read_op_pos < OSD_PACKET_SIZE || cl->read_op_size >= (cl->read_op_pos-OSD_PACKET_SIZE));
         cl->refs++;
         if (ringloop && !use_sync_send_recv)
         {
@@ -50,7 +56,7 @@ void osd_messenger_t::read_requests()
             }
             ring_data_t* data = ((ring_data_t*)sqe->user_data);
             data->callback = [this, cl](ring_data_t *data) { handle_read(data->res, cl); };
-            io_uring_prep_recvmsg(sqe, peer_fd, &cl->read_msg, 0);
+            io_uring_prep_recvmsg(sqe, peer_fd, &cl->read_msg, cl->recv_list.size() ? MSG_WAITALL : 0);
             if (iothread)
             {
                 iothread->add_sqe(sqe_local);
@@ -70,23 +76,17 @@ void osd_messenger_t::read_requests()
     read_ready_clients.clear();
 }
 
-bool osd_messenger_t::handle_read(int result, osd_client_t *cl)
+void osd_messenger_t::handle_read(int result, osd_client_t *cl)
 {
-    bool ret = false;
     int peer_fd = cl->peer_fd;
-    cl->read_msg.msg_iovlen = 0;
     cl->refs--;
-    if (cl->peer_state == PEER_RDMA)
-    {
-        return true;
-    }
     if (cl->peer_state == PEER_STOPPED)
     {
         if (cl->refs <= 0)
         {
             delete cl;
         }
-        return false;
+        return;
     }
     if (result <= 0 && result != -EAGAIN && result != -EINTR)
     {
@@ -96,9 +96,51 @@ bool osd_messenger_t::handle_read(int result, osd_client_t *cl)
             fprintf(stderr, "Client %d socket read error: %d (%s). Disconnecting client\n", cl->peer_fd, -result, strerror(-result));
         }
         stop_client(cl->peer_fd);
-        return false;
+        return;
     }
-    if (result == -EAGAIN || result == -EINTR || result < cl->read_iov.iov_len)
+    bool full_read = false;
+    if (result > 0)
+    {
+        if (cl->read_iov.iov_base == cl->in_buf)
+        {
+            full_read = result >= cl->read_iov.iov_len;
+            if (!handle_read_buffer(cl, cl->in_buf, result))
+            {
+                clear_immediate_ops(peer_fd);
+                handle_immediate_ops();
+                return;
+            }
+        }
+        else
+        {
+            // Reset OSD ping state
+            cl->ping_time_remaining = 0;
+            cl->idle_time_remaining = osd_idle_timeout;
+            // Long data
+            size_t i = 0;
+            while (i < cl->recv_list.size() && result >= cl->recv_list[i].iov_len)
+            {
+                result -= cl->recv_list[i].iov_len;
+                i++;
+            }
+            if (i < cl->recv_list.size())
+            {
+                cl->recv_list[i].iov_base += result;
+                cl->recv_list[i].iov_len -= result;
+            }
+            else
+            {
+                full_read = true;
+            }
+            cl->recv_list.erase(cl->recv_list.begin(), cl->recv_list.begin()+i);
+            if (!cl->recv_list.size())
+            {
+                handle_finished_op(cl);
+            }
+        }
+    }
+    cl->read_msg.msg_iovlen = 0;
+    if (result == -EAGAIN || result == -EINTR || !full_read)
     {
         cl->read_ready--;
         if (cl->read_ready > 0)
@@ -108,39 +150,7 @@ bool osd_messenger_t::handle_read(int result, osd_client_t *cl)
     {
         read_ready_clients.push_back(cl->peer_fd);
     }
-    if (result > 0)
-    {
-        if (cl->read_iov.iov_base == cl->in_buf)
-        {
-            if (!handle_read_buffer(cl, cl->in_buf, result))
-            {
-                clear_immediate_ops(peer_fd);
-                handle_immediate_ops();
-                return false;
-            }
-        }
-        else
-        {
-            // Long data
-            cl->read_remaining -= result;
-            cl->recv_list.eat(result);
-            if (cl->recv_list.done >= cl->recv_list.count)
-            {
-                if (!handle_finished_read(cl))
-                {
-                    clear_immediate_ops(peer_fd);
-                    handle_immediate_ops();
-                    return false;
-                }
-            }
-        }
-        if (result >= cl->read_iov.iov_len)
-        {
-            ret = true;
-        }
-    }
     handle_immediate_ops();
-    return ret;
 }
 
 void osd_messenger_t::clear_immediate_ops(int peer_fd)
@@ -165,10 +175,8 @@ void osd_messenger_t::clear_immediate_ops(int peer_fd)
 
 void osd_messenger_t::handle_immediate_ops()
 {
-    while (set_immediate_ops.size())
+    for (auto op: set_immediate_ops)
     {
-        auto op = set_immediate_ops.front();
-        set_immediate_ops.pop_front();
         if (op->op_type == OSD_OP_IN)
         {
             exec_op(op);
@@ -179,115 +187,98 @@ void osd_messenger_t::handle_immediate_ops()
             std::function<void(osd_op_t*)>(op->callback)(op);
         }
     }
+    set_immediate_ops.clear();
 }
 
-bool osd_messenger_t::handle_read_buffer(osd_client_t *cl, void *curbuf, int remain)
+bool osd_messenger_t::handle_read_buffer(osd_client_t *cl, uint8_t *curbuf, size_t bufsize)
 {
+    // Reset OSD ping state
+    cl->ping_time_remaining = 0;
+    cl->idle_time_remaining = osd_idle_timeout;
     // Compose operation(s) from the buffer
-    while (remain > 0)
+    size_t done = 0;
+    while (done < bufsize)
     {
         if (!cl->read_op)
         {
             cl->read_op = new osd_op_t;
             cl->read_op->peer_fd = cl->peer_fd;
             cl->read_op->op_type = OSD_OP_IN;
-            cl->recv_list.push_back(cl->read_op->req.buf, OSD_PACKET_SIZE);
-            cl->read_remaining = OSD_PACKET_SIZE;
-            cl->read_state = CL_READ_HDR;
+            cl->read_op_pos = 0;
+            cl->read_op_size = 0;
+            cl->read_op_inline_decrypt_pos = (size_t)-1;
         }
-        while (cl->recv_list.done < cl->recv_list.count && remain > 0)
+        if (cl->read_op_pos < OSD_PACKET_SIZE)
         {
-            iovec* cur = cl->recv_list.get_iovec();
-            if (cur->iov_len > remain)
-            {
-                memcpy(cur->iov_base, curbuf, remain);
-                cl->read_remaining -= remain;
-                cur->iov_len -= remain;
-                cur->iov_base = (uint8_t*)cur->iov_base + remain;
-                remain = 0;
-            }
-            else
-            {
-                memcpy(cur->iov_base, curbuf, cur->iov_len);
-                curbuf = (uint8_t*)curbuf + cur->iov_len;
-                cl->read_remaining -= cur->iov_len;
-                remain -= cur->iov_len;
-                cur->iov_len = 0;
-                cl->recv_list.done++;
-            }
-        }
-        if (cl->recv_list.done >= cl->recv_list.count)
-        {
-            if (!handle_finished_read(cl))
+            int len = OSD_PACKET_SIZE - cl->read_op_pos;
+            if (len > bufsize-done)
+                len = bufsize-done;
+            memcpy(cl->read_op->req.buf + cl->read_op_pos, curbuf+done, len);
+            done += len;
+            cl->read_op_pos += len;
+            if (cl->read_op_pos < OSD_PACKET_SIZE)
+                return true;
+            if (!handle_hdr(cl))
             {
+                stop_client(cl->peer_fd);
                 return false;
             }
         }
+        op_copy_from(cl, curbuf, bufsize, done);
     }
     return true;
 }
 
-bool osd_messenger_t::handle_finished_read(osd_client_t *cl)
+bool osd_messenger_t::handle_hdr(osd_client_t *cl)
 {
-    // Reset OSD ping state
-    cl->ping_time_remaining = 0;
-    cl->idle_time_remaining = osd_idle_timeout;
-    cl->recv_list.reset();
-    if (cl->read_state == CL_READ_HDR)
+    if (cl->read_op->req.hdr.magic == SECONDARY_OSD_REPLY_MAGIC)
     {
-        if (cl->read_op->req.hdr.magic == SECONDARY_OSD_REPLY_MAGIC)
-            return handle_reply_hdr(cl);
-        else if (cl->read_op->req.hdr.magic == SECONDARY_OSD_OP_MAGIC)
+        auto req_it = cl->sent_ops.find(cl->read_op->req.hdr.id);
+        if (req_it == cl->sent_ops.end())
         {
-            if (cl->check_sequencing)
-            {
-                if (cl->read_op->req.hdr.id != cl->read_op_id)
-                {
-                    fprintf(stderr, "Warning: operation sequencing is broken on client %d: expected num %ju, got %ju, stopping client\n", cl->peer_fd, cl->read_op_id, cl->read_op->req.hdr.id);
-                    stop_client(cl->peer_fd);
-                    return false;
-                }
-                cl->read_op_id++;
-            }
-            handle_op_hdr(cl);
+            // Command out of sync. Drop connection
+            fprintf(stderr, "Client %d command out of sync: id %ju\n", cl->peer_fd, cl->read_op->req.hdr.id);
+            return false;
         }
-        else
+        osd_op_t *op = req_it->second;
+        memcpy(op->reply.buf, cl->read_op->req.buf, OSD_PACKET_SIZE);
+        if (!allocate_reply_buffers(cl, op))
+        {
+            return false;
+        }
+        cl->sent_ops.erase(req_it);
+        delete cl->read_op;
+        cl->read_op = op;
+    }
+    else if (cl->read_op->req.hdr.magic == SECONDARY_OSD_OP_MAGIC)
+    {
+        if (cl->check_sequencing)
+        {
+            if (cl->read_op->req.hdr.id != cl->read_op_id)
+            {
+                fprintf(stderr, "Warning: operation sequencing is broken on client %d: expected num %ju, got %ju, stopping client\n", cl->peer_fd, cl->read_op_id, cl->read_op->req.hdr.id);
+                return false;
+            }
+            cl->read_op_id++;
+        }
+        if (!allocate_op_buffers(cl))
         {
-            fprintf(stderr, "Received garbage: magic=%jx id=%ju opcode=%jx from %d\n", cl->read_op->req.hdr.magic, cl->read_op->req.hdr.id, cl->read_op->req.hdr.opcode, cl->peer_fd);
-            stop_client(cl->peer_fd);
             return false;
         }
     }
-    else if (cl->read_state == CL_READ_DATA)
-    {
-        // Operation is ready
-        cl->received_ops.push_back(cl->read_op);
-        set_immediate_ops.push_back(cl->read_op);
-        cl->read_op = NULL;
-        cl->read_state = 0;
-    }
-    else if (cl->read_state == CL_READ_REPLY_DATA)
-    {
-        // Reply is ready
-        handle_reply_ready(cl->read_op);
-        cl->read_op = NULL;
-        cl->read_state = 0;
-    }
     else
     {
-        assert(0);
+        fprintf(stderr, "Received garbage: magic=%jx id=%ju opcode=%jx from %d\n", cl->read_op->req.hdr.magic, cl->read_op->req.hdr.id, cl->read_op->req.hdr.opcode, cl->peer_fd);
+        return false;
     }
     return true;
 }
 
-void osd_messenger_t::handle_op_hdr(osd_client_t *cl)
+bool osd_messenger_t::allocate_op_buffers(osd_client_t *cl)
 {
     osd_op_t *cur_op = cl->read_op;
-    if (cur_op->req.hdr.opcode == OSD_OP_SEC_READ)
-    {
-        cl->read_remaining = 0;
-    }
-    else if (cur_op->req.hdr.opcode == OSD_OP_SEC_WRITE ||
+    cl->read_op_size = 0;
+    if (cur_op->req.hdr.opcode == OSD_OP_SEC_WRITE ||
         cur_op->req.hdr.opcode == OSD_OP_SEC_WRITE_STABLE)
     {
         if (cur_op->req.sec_rw.attr_len > 0)
@@ -296,14 +287,12 @@ void osd_messenger_t::handle_op_hdr(osd_client_t *cl)
                 cur_op->bitmap = cur_op->rmw_buf = malloc_or_die(cur_op->req.sec_rw.attr_len);
             else
                 cur_op->bitmap = &cur_op->bmp_data;
-            cl->recv_list.push_back(cur_op->bitmap, cur_op->req.sec_rw.attr_len);
         }
         if (cur_op->req.sec_rw.len > 0)
         {
             cur_op->buf = memalign_or_die(MEM_ALIGNMENT, cur_op->req.sec_rw.len);
-            cl->recv_list.push_back(cur_op->buf, cur_op->req.sec_rw.len);
         }
-        cl->read_remaining = cur_op->req.sec_rw.len + cur_op->req.sec_rw.attr_len;
+        cl->read_op_size = cur_op->req.sec_rw.len + cur_op->req.sec_rw.attr_len;
     }
     else if (cur_op->req.hdr.opcode == OSD_OP_SEC_STABILIZE ||
         cur_op->req.hdr.opcode == OSD_OP_SEC_ROLLBACK)
@@ -311,27 +300,24 @@ void osd_messenger_t::handle_op_hdr(osd_client_t *cl)
         if (cur_op->req.sec_stab.len > 0)
         {
             cur_op->buf = memalign_or_die(MEM_ALIGNMENT, cur_op->req.sec_stab.len);
-            cl->recv_list.push_back(cur_op->buf, cur_op->req.sec_stab.len);
         }
-        cl->read_remaining = cur_op->req.sec_stab.len;
+        cl->read_op_size = cur_op->req.sec_stab.len;
     }
     else if (cur_op->req.hdr.opcode == OSD_OP_SEC_READ_BMP)
     {
         if (cur_op->req.sec_read_bmp.len > 0)
         {
             cur_op->buf = memalign_or_die(MEM_ALIGNMENT, cur_op->req.sec_read_bmp.len);
-            cl->recv_list.push_back(cur_op->buf, cur_op->req.sec_read_bmp.len);
         }
-        cl->read_remaining = cur_op->req.sec_read_bmp.len;
+        cl->read_op_size = cur_op->req.sec_read_bmp.len;
     }
     else if (cur_op->req.hdr.opcode == OSD_OP_WRITE)
     {
         if (cur_op->req.rw.len > 0)
         {
             cur_op->buf = memalign_or_die(MEM_ALIGNMENT, cur_op->req.rw.len);
-            cl->recv_list.push_back(cur_op->buf, cur_op->req.rw.len);
         }
-        cl->read_remaining = cur_op->req.rw.len;
+        cl->read_op_size = cur_op->req.rw.len;
     }
     else if (cur_op->req.hdr.opcode == OSD_OP_SHOW_CONFIG)
     {
@@ -339,44 +325,15 @@ void osd_messenger_t::handle_op_hdr(osd_client_t *cl)
         {
             cur_op->buf = malloc_or_die(cur_op->req.show_conf.json_len+1);
             ((uint8_t*)cur_op->buf)[cur_op->req.show_conf.json_len] = 0;
-            cl->recv_list.push_back(cur_op->buf, cur_op->req.show_conf.json_len);
         }
-        cl->read_remaining = cur_op->req.show_conf.json_len;
-    }
-    /*else if (cur_op->req.hdr.opcode == OSD_OP_READ ||
-        cur_op->req.hdr.opcode == OSD_OP_SCRUB ||
-        cur_op->req.hdr.opcode == OSD_OP_DESCRIBE)
-    {
-        cl->read_remaining = 0;
-    }*/
-    if (cl->read_remaining > 0)
-    {
-        // Read data
-        cl->read_state = CL_READ_DATA;
-    }
-    else
-    {
-        // Operation is ready
-        cl->received_ops.push_back(cur_op);
-        set_immediate_ops.push_back(cur_op);
-        cl->read_op = NULL;
-        cl->read_state = 0;
+        cl->read_op_size = cur_op->req.show_conf.json_len;
     }
+    return true;
 }
 
-bool osd_messenger_t::handle_reply_hdr(osd_client_t *cl)
+bool osd_messenger_t::allocate_reply_buffers(osd_client_t *cl, osd_op_t *op)
 {
-    auto req_it = cl->sent_ops.find(cl->read_op->req.hdr.id);
-    if (req_it == cl->sent_ops.end())
-    {
-        // Command out of sync. Drop connection
-        fprintf(stderr, "Client %d command out of sync: id %ju\n", cl->peer_fd, cl->read_op->req.hdr.id);
-        stop_client(cl->peer_fd);
-        return false;
-    }
-    osd_op_t *op = req_it->second;
-    memcpy(op->reply.buf, cl->read_op->req.buf, OSD_PACKET_SIZE);
-    cl->sent_ops.erase(req_it);
+    cl->read_op_size = 0;
     if (op->reply.hdr.opcode == OSD_OP_SEC_READ || op->reply.hdr.opcode == OSD_OP_READ)
     {
         // Read data. In this case we assume that the buffer is preallocated by the caller (!)
@@ -387,97 +344,295 @@ bool osd_messenger_t::handle_reply_hdr(osd_client_t *cl)
             // Check reply length to not overflow the buffer
             fprintf(stderr, "Client %d read reply of different length: expected %u+%u, got %jd+%u\n",
                 cl->peer_fd, expected_size, op->bitmap_len, op->reply.hdr.retval, bmp_len);
-            cl->sent_ops[op->req.hdr.id] = op;
-            stop_client(cl->peer_fd);
             return false;
         }
         if (bmp_len > 0)
         {
             assert(op->bitmap);
-            cl->recv_list.push_back(op->bitmap, bmp_len);
-            cl->read_remaining += bmp_len;
+            cl->read_op_size += bmp_len;
         }
         if (op->reply.hdr.retval > 0)
         {
             assert(op->iov.count > 0);
-            cl->recv_list.append(op->iov);
-            cl->read_remaining += op->reply.hdr.retval;
+            cl->read_op_size += op->reply.hdr.retval;
         }
-        if (cl->read_remaining == 0)
-        {
-            goto reuse;
-        }
-        delete cl->read_op;
-        cl->read_op = op;
-        cl->read_state = CL_READ_REPLY_DATA;
     }
     else if (op->reply.hdr.opcode == OSD_OP_SEC_LIST && op->reply.hdr.retval > 0)
     {
         assert(!op->iov.count);
-        delete cl->read_op;
-        cl->read_op = op;
-        cl->read_state = CL_READ_REPLY_DATA;
-        cl->read_remaining = sizeof(obj_ver_id) * op->reply.hdr.retval;
-        op->buf = memalign_or_die(MEM_ALIGNMENT, cl->read_remaining);
-        cl->recv_list.push_back(op->buf, cl->read_remaining);
+        cl->read_op_size = sizeof(obj_ver_id) * op->reply.hdr.retval;
+        op->buf = memalign_or_die(MEM_ALIGNMENT, cl->read_op_size);
     }
     else if (op->reply.hdr.opcode == OSD_OP_SEC_READ_BMP && op->reply.hdr.retval > 0)
     {
         assert(!op->iov.count);
-        delete cl->read_op;
-        cl->read_op = op;
-        cl->read_state = CL_READ_REPLY_DATA;
-        cl->read_remaining = op->reply.hdr.retval;
+        cl->read_op_size = op->reply.hdr.retval;
         free(op->buf);
-        op->buf = memalign_or_die(MEM_ALIGNMENT, cl->read_remaining);
-        cl->recv_list.push_back(op->buf, cl->read_remaining);
+        op->buf = memalign_or_die(MEM_ALIGNMENT, cl->read_op_size);
     }
     else if (op->reply.hdr.opcode == OSD_OP_SHOW_CONFIG && op->reply.hdr.retval > 0)
     {
-        delete cl->read_op;
-        cl->read_op = op;
-        cl->read_state = CL_READ_REPLY_DATA;
-        cl->read_remaining = op->reply.hdr.retval;
+        cl->read_op_size = op->reply.hdr.retval;
         free(op->buf);
         op->buf = malloc_or_die(op->reply.hdr.retval);
-        cl->recv_list.push_back(op->buf, op->reply.hdr.retval);
     }
     else if (op->reply.hdr.opcode == OSD_OP_DESCRIBE && op->reply.describe.result_bytes > 0)
     {
-        delete cl->read_op;
-        cl->read_op = op;
-        cl->read_state = CL_READ_REPLY_DATA;
-        cl->read_remaining = op->reply.describe.result_bytes;
+        cl->read_op_size = op->reply.describe.result_bytes;
         free(op->buf);
         op->buf = malloc_or_die(op->reply.describe.result_bytes);
-        cl->recv_list.push_back(op->buf, op->reply.describe.result_bytes);
-    }
-    else
-    {
-reuse:
-        // It's fine to reuse cl->read_op for the next reply
-        handle_reply_ready(op);
-        cl->recv_list.push_back(cl->read_op->req.buf, OSD_PACKET_SIZE);
-        cl->read_remaining = OSD_PACKET_SIZE;
-        cl->read_state = CL_READ_HDR;
     }
     return true;
 }
 
-void osd_messenger_t::handle_reply_ready(osd_op_t *op)
+size_t osd_messenger_t::op_copy_from(osd_client_t *cl, uint8_t *src, size_t src_len, size_t & done)
 {
-    // Measure subop latency
-    timespec tv_end;
-    clock_gettime(CLOCK_REALTIME, &tv_end);
-    stats.subop_stat_count[op->req.hdr.opcode]++;
-    if (!stats.subop_stat_count[op->req.hdr.opcode])
+    osd_op_t *op = cl->read_op;
+    size_t from = cl->read_op_pos-OSD_PACKET_SIZE;
+    auto op_read_buf = [&](uint8_t *dst, size_t dst_len)
     {
-        stats.subop_stat_count[op->req.hdr.opcode]++;
-        stats.subop_stat_sum[op->req.hdr.opcode] = 0;
+        if (from < dst_len)
+        {
+            size_t n = dst_len-from;
+            if (n > src_len-done)
+                n = src_len-done;
+            memcpy(dst+from, src+done, n);
+            done += n;
+            cl->read_op_pos += n;
+            from += n;
+            if (from < dst_len)
+                return false;
+            from = 0;
+        }
+        else
+            from -= dst_len;
+        return true;
+    };
+    if (op->op_type == OSD_OP_IN)
+    {
+        if (op->req.hdr.opcode == OSD_OP_SEC_WRITE ||
+            op->req.hdr.opcode == OSD_OP_SEC_WRITE_STABLE)
+        {
+            if (!op_read_buf((uint8_t*)op->bitmap, op->req.sec_rw.attr_len))
+                return done;
+            if (!op_read_buf((uint8_t*)op->buf, op->req.sec_rw.len))
+                return done;
+        }
+        else if (op->req.hdr.opcode == OSD_OP_SEC_STABILIZE ||
+            op->req.hdr.opcode == OSD_OP_SEC_ROLLBACK)
+        {
+            if (!op_read_buf((uint8_t*)op->buf, op->req.sec_stab.len))
+                return done;
+        }
+        else if (op->req.hdr.opcode == OSD_OP_SEC_READ_BMP)
+        {
+            if (!op_read_buf((uint8_t*)op->buf, op->req.sec_read_bmp.len))
+                return done;
+        }
+        else if (op->req.hdr.opcode == OSD_OP_WRITE)
+        {
+            if (!op_read_buf((uint8_t*)op->buf, op->req.rw.len))
+                return done;
+        }
+        else if (op->req.hdr.opcode == OSD_OP_SHOW_CONFIG)
+        {
+            if (!op_read_buf((uint8_t*)op->buf, op->req.show_conf.json_len))
+                return done;
+        }
     }
-    stats.subop_stat_sum[op->req.hdr.opcode] += (
-        (tv_end.tv_sec - op->tv_begin.tv_sec)*1000000 +
-        (tv_end.tv_nsec - op->tv_begin.tv_nsec)/1000
-    );
-    set_immediate_ops.push_back(op);
+    else
+    {
+        if (op->reply.hdr.opcode == OSD_OP_SEC_READ)
+        {
+            if (op->reply.sec_rw.attr_len > 0)
+            {
+                if (!op_read_buf((uint8_t*)op->bitmap, op->reply.sec_rw.attr_len))
+                    return done;
+            }
+            if (op->reply.hdr.retval > 0)
+            {
+                for (int i = 0; i < op->iov.count; i++)
+                    if (!op_read_buf((uint8_t*)op->iov.buf[i].iov_base, op->iov.buf[i].iov_len))
+                        return done;
+            }
+        }
+        else if (op->reply.hdr.opcode == OSD_OP_READ)
+        {
+            if (op->reply.rw.bitmap_len > 0)
+            {
+                if (!op_read_buf((uint8_t*)op->bitmap, op->reply.rw.bitmap_len))
+                    return done;
+            }
+            if (op->reply.hdr.retval > 0)
+            {
+                if (op->enc)
+                {
+                    if (!op_decrypted_copy_data_from(cl, src, src_len, from, done))
+                        return done;
+                }
+                else
+                {
+                    for (int i = 0; i < op->iov.count; i++)
+                        if (!op_read_buf((uint8_t*)op->iov.buf[i].iov_base, op->iov.buf[i].iov_len))
+                            return done;
+                }
+            }
+        }
+        else if (op->reply.hdr.opcode == OSD_OP_SEC_LIST && op->reply.hdr.retval > 0)
+        {
+            if (!op_read_buf((uint8_t*)op->buf, sizeof(obj_ver_id) * op->reply.hdr.retval))
+                return done;
+        }
+        else if ((op->reply.hdr.opcode == OSD_OP_SEC_READ_BMP ||
+            op->reply.hdr.opcode == OSD_OP_SHOW_CONFIG) && op->reply.hdr.retval > 0)
+        {
+            if (!op_read_buf((uint8_t*)op->buf, op->reply.hdr.retval))
+                return done;
+        }
+        else if (op->reply.hdr.opcode == OSD_OP_DESCRIBE && op->reply.describe.result_bytes > 0)
+        {
+            if (!op_read_buf((uint8_t*)op->buf, op->reply.describe.result_bytes))
+                return done;
+        }
+    }
+    handle_finished_op(cl);
+    return done;
+}
+
+size_t osd_messenger_t::op_get_read_buffers(osd_client_t *cl, std::vector<iovec> & lst)
+{
+    osd_op_t *op = cl->read_op;
+    size_t from = cl->read_op_pos-OSD_PACKET_SIZE;
+    size_t done = 0;
+    auto op_read_buf = [&](uint8_t *dst, size_t dst_len)
+    {
+        if (lst.size() >= IOV_MAX)
+            return false;
+        if (from < dst_len)
+        {
+            lst.push_back((iovec){ .iov_base = dst+from, .iov_len = dst_len-from });
+            cl->read_op_pos += dst_len-from;
+            done += dst_len-from;
+            from = 0;
+        }
+        else
+            from -= dst_len;
+        return true;
+    };
+    if (op->op_type == OSD_OP_IN)
+    {
+        if (op->req.hdr.opcode == OSD_OP_SEC_WRITE ||
+            op->req.hdr.opcode == OSD_OP_SEC_WRITE_STABLE)
+        {
+            if (!op_read_buf((uint8_t*)op->bitmap, op->req.sec_rw.attr_len))
+                return done;
+            if (!op_read_buf((uint8_t*)op->buf, op->req.sec_rw.len))
+                return done;
+        }
+        else if (op->req.hdr.opcode == OSD_OP_SEC_STABILIZE ||
+            op->req.hdr.opcode == OSD_OP_SEC_ROLLBACK)
+        {
+            if (!op_read_buf((uint8_t*)op->buf, op->req.sec_stab.len))
+                return done;
+        }
+        else if (op->req.hdr.opcode == OSD_OP_SEC_READ_BMP)
+        {
+            if (!op_read_buf((uint8_t*)op->buf, op->req.sec_read_bmp.len))
+                return done;
+        }
+        else if (op->req.hdr.opcode == OSD_OP_WRITE)
+        {
+            if (!op_read_buf((uint8_t*)op->buf, op->req.rw.len))
+                return done;
+        }
+        else if (op->req.hdr.opcode == OSD_OP_SHOW_CONFIG)
+        {
+            if (!op_read_buf((uint8_t*)op->buf, op->req.show_conf.json_len))
+                return done;
+        }
+    }
+    else
+    {
+        if (op->reply.hdr.opcode == OSD_OP_SEC_READ)
+        {
+            if (op->reply.sec_rw.attr_len > 0)
+            {
+                if (!op_read_buf((uint8_t*)op->bitmap, op->reply.sec_rw.attr_len))
+                    return done;
+            }
+            if (op->reply.hdr.retval > 0)
+            {
+                for (int i = 0; i < op->iov.count; i++)
+                    if (!op_read_buf((uint8_t*)op->iov.buf[i].iov_base, op->iov.buf[i].iov_len))
+                        return done;
+            }
+        }
+        else if (op->reply.hdr.opcode == OSD_OP_READ)
+        {
+            if (op->reply.rw.bitmap_len > 0)
+            {
+                if (!op_read_buf((uint8_t*)op->bitmap, op->reply.rw.bitmap_len))
+                    return done;
+            }
+            if (op->reply.hdr.retval > 0)
+            {
+                if (op->enc)
+                    cl->read_op_inline_decrypt_pos = cl->read_op_pos;
+                for (int i = 0; i < op->iov.count; i++)
+                    if (!op_read_buf((uint8_t*)op->iov.buf[i].iov_base, op->iov.buf[i].iov_len))
+                        return done;
+            }
+        }
+        else if (op->reply.hdr.opcode == OSD_OP_SEC_LIST && op->reply.hdr.retval > 0)
+        {
+            if (!op_read_buf((uint8_t*)op->buf, sizeof(obj_ver_id) * op->reply.hdr.retval))
+                return done;
+        }
+        else if ((op->reply.hdr.opcode == OSD_OP_SEC_READ_BMP ||
+            op->reply.hdr.opcode == OSD_OP_SHOW_CONFIG) && op->reply.hdr.retval > 0)
+        {
+            if (!op_read_buf((uint8_t*)op->buf, op->reply.hdr.retval))
+                return done;
+        }
+        else if (op->reply.hdr.opcode == OSD_OP_DESCRIBE && op->reply.describe.result_bytes > 0)
+        {
+            if (!op_read_buf((uint8_t*)op->buf, op->reply.describe.result_bytes))
+                return done;
+        }
+    }
+    return done;
+}
+
+void osd_messenger_t::handle_finished_op(osd_client_t *cl)
+{
+    osd_op_t *op = cl->read_op;
+    if (op->op_type == OSD_OP_IN)
+    {
+        // Operation is ready
+        cl->received_ops.push_back(op);
+    }
+    else
+    {
+        // Inline decryption
+        if (cl->read_op_inline_decrypt_pos != (size_t)-1)
+        {
+            op_decrypt_inline(cl);
+            cl->read_op_inline_decrypt_pos = (size_t)-1;
+        }
+        // Measure subop (outbound op) latency
+        timespec tv_end;
+        clock_gettime(CLOCK_REALTIME, &tv_end);
+        stats.subop_stat_count[op->req.hdr.opcode]++;
+        if (!stats.subop_stat_count[op->req.hdr.opcode])
+        {
+            stats.subop_stat_count[op->req.hdr.opcode]++;
+            stats.subop_stat_sum[op->req.hdr.opcode] = 0;
+        }
+        stats.subop_stat_sum[op->req.hdr.opcode] += (
+            (tv_end.tv_sec - op->tv_begin.tv_sec)*1000000 +
+            (tv_end.tv_nsec - op->tv_begin.tv_nsec)/1000
+        );
+    }
+    set_immediate_ops.push_back(op);
+    cl->read_op = NULL;
 }

src/client/msgr_send.cpp

@@ -15,6 +15,7 @@ void osd_messenger_t::outbox_push(osd_op_t *cur_op)
     {
         clock_gettime(CLOCK_REALTIME, &cur_op->tv_begin);
         cur_op->req.hdr.id = ++cl->send_op_id;
+        cl->sent_ops[cur_op->req.hdr.id] = cur_op;
     }
     else
     {
@@ -35,77 +36,9 @@ void osd_messenger_t::outbox_push(osd_op_t *cur_op)
             delete cur_op;
             return;
         }
-    }
-    auto & to_send_list = cl->write_msg.msg_iovlen ? cl->next_send_list : cl->send_list;
-    auto & to_outbox = cl->write_msg.msg_iovlen ? cl->next_outbox : cl->outbox;
-    if (cur_op->op_type == OSD_OP_IN)
-    {
         measure_exec(cur_op);
-        to_send_list.push_back((iovec){ .iov_base = cur_op->reply.buf, .iov_len = OSD_PACKET_SIZE });
-    }
-    else
-    {
-        to_send_list.push_back((iovec){ .iov_base = cur_op->req.buf, .iov_len = OSD_PACKET_SIZE });
-        cl->sent_ops[cur_op->req.hdr.id] = cur_op;
-    }
-    to_outbox.push_back((msgr_sendp_t){ .op = cur_op, .flags = MSGR_SENDP_HDR });
-    // Bitmap
-    if (cur_op->op_type == OSD_OP_IN &&
-        cur_op->req.hdr.opcode == OSD_OP_SEC_READ &&
-        cur_op->reply.sec_rw.attr_len > 0)
-    {
-        to_send_list.push_back((iovec){
-            .iov_base = cur_op->bitmap,
-            .iov_len = cur_op->reply.sec_rw.attr_len,
-        });
-        to_outbox.push_back((msgr_sendp_t){ .op = cur_op, .flags = 0 });
-    }
-    else if (cur_op->op_type == OSD_OP_OUT &&
-        (cur_op->req.hdr.opcode == OSD_OP_SEC_WRITE || cur_op->req.hdr.opcode == OSD_OP_SEC_WRITE_STABLE) &&
-        cur_op->req.sec_rw.attr_len > 0)
-    {
-        to_send_list.push_back((iovec){
-            .iov_base = cur_op->bitmap,
-            .iov_len = cur_op->req.sec_rw.attr_len,
-        });
-        to_outbox.push_back((msgr_sendp_t){ .op = cur_op, .flags = 0 });
-    }
-    // Operation data
-    if ((cur_op->op_type == OSD_OP_IN
-        ? (cur_op->req.hdr.opcode == OSD_OP_READ ||
-        cur_op->req.hdr.opcode == OSD_OP_SEC_READ ||
-        cur_op->req.hdr.opcode == OSD_OP_SEC_LIST ||
-        cur_op->req.hdr.opcode == OSD_OP_SHOW_CONFIG ||
-        cur_op->req.hdr.opcode == OSD_OP_DESCRIBE)
-        : (cur_op->req.hdr.opcode == OSD_OP_WRITE ||
-        cur_op->req.hdr.opcode == OSD_OP_SEC_WRITE ||
-        cur_op->req.hdr.opcode == OSD_OP_SEC_WRITE_STABLE ||
-        cur_op->req.hdr.opcode == OSD_OP_SEC_STABILIZE ||
-        cur_op->req.hdr.opcode == OSD_OP_SEC_ROLLBACK ||
-        cur_op->req.hdr.opcode == OSD_OP_SHOW_CONFIG)) && cur_op->iov.count > 0)
-    {
-        for (int i = 0; i < cur_op->iov.count; i++)
-        {
-            if (cur_op->iov.buf[i].iov_len > 0)
-            {
-                assert(cur_op->iov.buf[i].iov_base);
-                to_send_list.push_back(cur_op->iov.buf[i]);
-                to_outbox.push_back((msgr_sendp_t){ .op = cur_op, .flags = 0 });
-            }
-        }
-    }
-    if (cur_op->req.hdr.opcode == OSD_OP_SEC_READ_BMP)
-    {
-        if (cur_op->op_type == OSD_OP_IN && cur_op->reply.hdr.retval > 0)
-            to_send_list.push_back((iovec){ .iov_base = cur_op->buf, .iov_len = (size_t)cur_op->reply.hdr.retval });
-        else if (cur_op->op_type == OSD_OP_OUT && cur_op->req.sec_read_bmp.len > 0)
-            to_send_list.push_back((iovec){ .iov_base = cur_op->buf, .iov_len = (size_t)cur_op->req.sec_read_bmp.len });
-        to_outbox.push_back((msgr_sendp_t){ .op = cur_op, .flags = 0 });
-    }
-    if (cur_op->op_type == OSD_OP_IN)
-    {
-        to_outbox[to_outbox.size()-1].flags |= MSGR_SENDP_FREE;
     }
+    cl->write_ops.push_back(cur_op);
 #ifdef WITH_RDMA
     if (cl->peer_state == PEER_RDMA)
     {
@@ -116,7 +49,7 @@ void osd_messenger_t::outbox_push(osd_op_t *cur_op)
     if (!ringloop)
     {
         // FIXME: It's worse because it doesn't allow batching
-        while (cl->outbox.size())
+        while (cl->write_ops.size())
         {
             try_send(cl);
         }
@@ -184,11 +117,25 @@ void osd_messenger_t::measure_exec(osd_op_t *cur_op)
 bool osd_messenger_t::try_send(osd_client_t *cl)
 {
     int peer_fd = cl->peer_fd;
-    if (!cl->send_list.size() || cl->write_msg.msg_iovlen > 0)
+    if (!cl->write_op && !cl->write_ops.size() || cl->write_msg.msg_iovlen > 0)
     {
         return true;
     }
     assert(cl->peer_state != PEER_RDMA);
+    while ((cl->write_op || cl->write_ops.size()) && cl->send_list.size() < IOV_MAX)
+    {
+        if (!cl->write_op)
+        {
+            cl->write_op = cl->write_ops.front();
+            cl->write_ops.pop_front();
+        }
+        osd_op_t *op = cl->write_op;
+        op_get_write_buffers(cl, cl->send_list);
+        if (!cl->write_op && op->op_type == OSD_OP_IN)
+        {
+            cl->send_free_ops.push_back(op);
+        }
+    }
     if (ringloop && !use_sync_send_recv)
     {
         auto iothread = iothreads.size() ? iothreads[peer_fd % iothreads.size()] : NULL;
@@ -201,20 +148,24 @@ bool osd_messenger_t::try_send(osd_client_t *cl)
             data_local = {};
         }
         if (!sqe)
+        {
             return false;
+        }
+        cl->send_list_size = 0;
+        for (auto & iov: cl->send_list)
+        {
+            cl->send_list_size += iov.iov_len;
+        }
         cl->write_msg.msg_iov = cl->send_list.data();
         cl->write_msg.msg_iovlen = cl->send_list.size() < IOV_MAX ? cl->send_list.size() : IOV_MAX;
         cl->refs++;
         ring_data_t* data = ((ring_data_t*)sqe->user_data);
         data->callback = [this, cl](ring_data_t *data) { handle_send(data->res, data->prev, data->more, cl); };
         bool use_zc = has_sendmsg_zc && min_zerocopy_send_size >= 0;
-        if (use_zc && min_zerocopy_send_size > 0)
+        if (use_zc && min_zerocopy_send_size > 0 &&
+            cl->send_list_size/cl->write_msg.msg_iovlen < min_zerocopy_send_size)
         {
-            size_t avg_size = 0;
-            for (size_t i = 0; i < cl->write_msg.msg_iovlen; i++)
-                avg_size += cl->write_msg.msg_iov[i].iov_len;
-            if (avg_size/cl->write_msg.msg_iovlen < min_zerocopy_send_size)
-                use_zc = false;
+            use_zc = false;
         }
         if (use_zc)
         {
@@ -251,7 +202,7 @@ void osd_messenger_t::send_replies()
     {
         int peer_fd = write_ready_clients[i];
         auto cl_it = clients.find(peer_fd);
-        if (cl_it != clients.end() && cl_it->second->peer_state != PEER_RDMA && !try_send(cl_it->second))
+        if (cl_it != clients.end() && !try_send(cl_it->second))
         {
             write_ready_clients.erase(write_ready_clients.begin(), write_ready_clients.begin() + i);
             return;
@@ -265,6 +216,7 @@ void osd_messenger_t::handle_send(int result, bool prev, bool more, osd_client_t
     if (!prev)
     {
         cl->write_msg.msg_iovlen = 0;
+        cl->send_list.clear();
     }
     if (!more)
     {
@@ -297,64 +249,42 @@ void osd_messenger_t::handle_send(int result, bool prev, bool more, osd_client_t
                 cl->zc_free_list.erase(cl->zc_free_list.begin(), cl->zc_free_list.begin()+i+1);
             return;
         }
-        int done = 0;
-        while (result > 0 && done < cl->send_list.size())
+        if (cl->send_list_size > result)
         {
-            iovec & iov = cl->send_list[done];
-            if (iov.iov_len <= result)
-            {
-                if (cl->outbox[done].flags & MSGR_SENDP_FREE)
-                {
-                    // Reply fully sent
-                    if (more)
-                        cl->zc_free_list.push_back(cl->outbox[done].op);
-                    else
-                        delete cl->outbox[done].op;
-                }
-                result -= iov.iov_len;
-                done++;
-            }
+            fprintf(stderr, "Client %d socket write error: expected to send "
+                "%zu bytes with MSG_WAITALL but sent %u. Disconnecting client\n", cl->peer_fd, cl->send_list_size, result);
+            stop_client(cl->peer_fd);
+            return;
+        }
+        for (auto op: cl->send_free_ops)
+        {
+            if (more)
+                cl->zc_free_list.push_back(op);
             else
-            {
-                iov.iov_len -= result;
-                iov.iov_base = (uint8_t*)iov.iov_base + result;
-                break;
-            }
+                delete op;
         }
         if (more)
-        {
-            int expected = cl->send_list.size() < IOV_MAX ? cl->send_list.size() : IOV_MAX;
-            if (done != expected)
-            {
-                fprintf(stderr, "Client %d socket write error: expected to send "
-                    "%d iovecs with MSG_WAITALL but sent %d. Disconnecting client\n", cl->peer_fd, expected, done);
-                stop_client(cl->peer_fd);
-                return;
-            }
             cl->zc_free_list.push_back(NULL); // end marker
-        }
-        if (done > 0)
-        {
-            cl->send_list.erase(cl->send_list.begin(), cl->send_list.begin()+done);
-            cl->outbox.erase(cl->outbox.begin(), cl->outbox.begin()+done);
-        }
-        if (cl->next_send_list.size())
-        {
-            cl->send_list.insert(cl->send_list.end(), cl->next_send_list.begin(), cl->next_send_list.end());
-            cl->outbox.insert(cl->outbox.end(), cl->next_outbox.begin(), cl->next_outbox.end());
-            cl->next_send_list.clear();
-            cl->next_outbox.clear();
-        }
-        cl->write_state = cl->outbox.size() > 0 ? CL_WRITE_READY : 0;
+        cl->send_free_ops.clear();
+        cl->write_state = cl->write_op || cl->write_ops.size() ? CL_WRITE_READY : 0;
 #ifdef WITH_RDMA
-        if (cl->rdma_conn && !cl->outbox.size() && cl->peer_state == PEER_RDMA_CONNECTING)
+        if (cl->rdma_conn && !cl->write_op && !cl->write_ops.size() && cl->peer_state == PEER_RDMA_CONNECTING)
         {
+            // FIXME: Do something better than just forgetting the FD
             // FIXME: Ignore pings during RDMA state transition
             if (log_level > 0)
             {
                 fprintf(stderr, "Successfully connected with client %d using RDMA\n", cl->peer_fd);
             }
             cl->peer_state = PEER_RDMA;
+            tfd->set_fd_handler(cl->peer_fd, false, [this](int peer_fd, int epoll_events)
+            {
+                // Do not miss the disconnection!
+                if (epoll_events & EPOLLRDHUP)
+                {
+                    handle_peer_epoll(peer_fd, epoll_events);
+                }
+            });
             // Add the initial receive request
             init_recv_rdma(cl);
         }
@@ -365,3 +295,158 @@ void osd_messenger_t::handle_send(int result, bool prev, bool more, osd_client_t
         write_ready_clients.push_back(cl->peer_fd);
     }
 }
+
+static inline bool op_write_headers(osd_op_t *op, std::function<bool(uint8_t*, size_t)> op_write_buf)
+{
+    // Header
+    if (!op_write_buf((op->op_type == OSD_OP_IN ? op->reply.buf : op->req.buf), OSD_PACKET_SIZE))
+        return false;
+    // Bitmap
+    if (op->op_type == OSD_OP_IN &&
+        op->req.hdr.opcode == OSD_OP_SEC_READ &&
+        op->reply.sec_rw.attr_len > 0)
+    {
+        if (!op_write_buf((uint8_t*)op->bitmap, op->reply.sec_rw.attr_len))
+            return false;
+    }
+    else if (op->op_type == OSD_OP_OUT &&
+        (op->req.hdr.opcode == OSD_OP_SEC_WRITE || op->req.hdr.opcode == OSD_OP_SEC_WRITE_STABLE) &&
+        op->req.sec_rw.attr_len > 0)
+    {
+        if (!op_write_buf((uint8_t*)op->bitmap, op->req.sec_rw.attr_len))
+            return false;
+    }
+    if (op->req.hdr.opcode == OSD_OP_SEC_READ_BMP)
+    {
+        if (op->op_type == OSD_OP_IN && op->reply.hdr.retval > 0)
+        {
+            if (!op_write_buf((uint8_t*)op->buf, (size_t)op->reply.hdr.retval))
+                return false;
+        }
+        else if (op->op_type == OSD_OP_OUT && op->req.sec_read_bmp.len > 0)
+        {
+            if (!op_write_buf((uint8_t*)op->buf, (size_t)op->req.sec_read_bmp.len))
+                return false;
+        }
+    }
+    return true;
+}
+
+static inline bool op_has_data(osd_op_t *op)
+{
+    return (op->op_type == OSD_OP_IN
+        ? (op->req.hdr.opcode == OSD_OP_READ ||
+        op->req.hdr.opcode == OSD_OP_SEC_READ ||
+        op->req.hdr.opcode == OSD_OP_SEC_LIST ||
+        op->req.hdr.opcode == OSD_OP_SHOW_CONFIG ||
+        op->req.hdr.opcode == OSD_OP_DESCRIBE)
+        : (op->req.hdr.opcode == OSD_OP_WRITE ||
+        op->req.hdr.opcode == OSD_OP_SEC_WRITE ||
+        op->req.hdr.opcode == OSD_OP_SEC_WRITE_STABLE ||
+        op->req.hdr.opcode == OSD_OP_SEC_STABILIZE ||
+        op->req.hdr.opcode == OSD_OP_SEC_ROLLBACK ||
+        op->req.hdr.opcode == OSD_OP_SHOW_CONFIG)) && op->iov.count > 0;
+}
+
+size_t osd_messenger_t::op_copy_to(osd_client_t *cl, uint8_t *dst, size_t dst_len)
+{
+    size_t done = 0;
+    size_t from = cl->write_op_pos;
+    auto op_write_buf = [&](uint8_t *src, size_t src_len)
+    {
+        if (from < src_len)
+        {
+            size_t n = src_len-from;
+            if (n > dst_len-done)
+                n = dst_len-done;
+            memcpy(dst+done, src+from, n);
+            done += n;
+            cl->write_op_pos += n;
+            from += n;
+            if (from < src_len)
+                return false;
+            from = 0;
+        }
+        else
+            from -= src_len;
+        return true;
+    };
+    if (!op_write_headers(cl->write_op, op_write_buf))
+    {
+        return done;
+    }
+    // Operation data
+    if (op_has_data(cl->write_op))
+    {
+        if (cl->write_op->enc)
+        {
+            if (!op_encrypted_copy_data_to(cl, dst, dst_len, from, done))
+            {
+                return done;
+            }
+        }
+        else
+        {
+            for (int i = 0; i < cl->write_op->iov.count; i++)
+            {
+                if (!op_write_buf((uint8_t*)cl->write_op->iov.buf[i].iov_base, cl->write_op->iov.buf[i].iov_len))
+                    return done;
+            }
+        }
+    }
+    cl->write_op = NULL;
+    cl->write_op_pos = 0;
+    return done;
+}
+
+void osd_messenger_t::op_get_write_buffers(osd_client_t *cl, std::vector<iovec> & lst)
+{
+    size_t from = cl->write_op_pos;
+    auto op_write_buf = [&](uint8_t *src, size_t src_len)
+    {
+        if (lst.size() >= IOV_MAX)
+            return false;
+        if (from < src_len)
+        {
+            lst.push_back((iovec){ .iov_base = src+from, .iov_len = src_len-from });
+            cl->write_op_pos += src_len-from;
+            from = 0;
+        }
+        else
+            from -= src_len;
+        return true;
+    };
+    if (!op_write_headers(cl->write_op, op_write_buf))
+    {
+        return;
+    }
+    // Operation data
+    if (op_has_data(cl->write_op))
+    {
+        if (cl->write_op->enc)
+        {
+            if (lst.size() >= IOV_MAX)
+                return;
+            // No way except to allocate a temporary buffer and encrypt data to it
+            assert(cl->write_op->req.hdr.opcode == OSD_OP_WRITE);
+            size_t remsize = cl->write_op->req.rw.len - from + (from % 16);
+            assert(remsize > 0);
+            assert(!cl->write_op->enc_buf);
+            cl->write_op->enc_buf = (uint8_t*)malloc_or_die(remsize);
+            size_t done = 0;
+            bool end = op_encrypted_copy_data_to(cl, cl->write_op->enc_buf, remsize, from, done);
+            assert(end);
+            lst.push_back((iovec){ .iov_base = cl->write_op->enc_buf, .iov_len = remsize });
+        }
+        else
+        {
+            for (int i = 0; i < cl->write_op->iov.count; i++)
+            {
+                if (!op_write_buf((uint8_t*)cl->write_op->iov.buf[i].iov_base, cl->write_op->iov.buf[i].iov_len))
+                    return;
+            }
+        }
+    }
+    cl->write_op = NULL;
+    cl->write_op_pos = 0;
+}

src/client/msgr_stop.cpp

@@ -57,7 +57,6 @@ void osd_messenger_t::stop_client(int peer_fd, bool force, bool force_delete)
     {
         return;
     }
-    clear_immediate_ops(peer_fd);
     if (log_level > 0)
     {
         if (cl->osd_num)
@@ -73,6 +72,22 @@ void osd_messenger_t::stop_client(int peer_fd, bool force, bool force_delete)
             fprintf(stderr, "[OSD %ju] Stopping client %d (regular client)\n", osd_num, peer_fd);
         }
     }
+    if (cl->encrypt_ctx)
+    {
+        if (encrypt_ctx_pool.size() > max_aes_xts_pool_size)
+            destroy_aes_xts_encrypt(cl->encrypt_ctx);
+        else
+            encrypt_ctx_pool.push_back(cl->encrypt_ctx);
+        cl->encrypt_ctx = NULL;
+    }
+    if (cl->decrypt_ctx)
+    {
+        if (decrypt_ctx_pool.size() > max_aes_xts_pool_size)
+            destroy_aes_xts_decrypt(cl->decrypt_ctx);
+        else
+            decrypt_ctx_pool.push_back(cl->decrypt_ctx);
+        cl->decrypt_ctx = NULL;
+    }
     // First set state to STOPPED so another stop_client() call doesn't try to free it again
     cl->refs++;
     int prev_state = cl->peer_state;
@@ -169,6 +184,13 @@ osd_client_t::~osd_client_t()
     }
     // Cancel outbound ops
     cancel_ops();
+    for (osd_op_t *op: send_free_ops)
+    {
+        if (op)
+        {
+            delete op;
+        }
+    }
     for (osd_op_t *op: zc_free_list)
     {
         if (op)

src/client/object_id.h

@@ -20,15 +20,6 @@ typedef uint64_t inode_t;
 // Pool ID is 16 bits long
 typedef uint32_t pool_id_t;
 
-typedef uint64_t osd_num_t;
-typedef uint32_t pg_num_t;
-
-struct pool_pg_num_t
-{
-    pool_id_t pool_id;
-    pg_num_t pg_num;
-};
-
 // 16 bytes per object/stripe id
 // stripe = (start of the parity stripe + peer role)
 // i.e. for example (256KB + one of 0,1,2)
@@ -70,21 +61,6 @@ inline bool operator < (const obj_ver_id & a, const obj_ver_id & b)
     return a.oid < b.oid || a.oid == b.oid && a.version < b.version;
 }
 
-inline bool operator < (const pool_pg_num_t & a, const pool_pg_num_t & b)
-{
-    return a.pool_id < b.pool_id || a.pool_id == b.pool_id && a.pg_num < b.pg_num;
-}
-
-inline bool operator == (const pool_pg_num_t & a, const pool_pg_num_t & b)
-{
-    return a.pool_id == b.pool_id && a.pg_num == b.pg_num;
-}
-
-inline bool operator != (const pool_pg_num_t & a, const pool_pg_num_t & b)
-{
-    return a.pool_id != b.pool_id || a.pg_num != b.pg_num;
-}
-
 namespace std
 {
     template<> struct hash<object_id>

src/client/osd_ops.h

@@ -4,6 +4,7 @@
 #pragma once
 
 #include "object_id.h"
+#include "osd_id.h"
 
 // Magic numbers
 #define SECONDARY_OSD_OP_MAGIC      0x2bd7b10325434553l

src/client/vitastor.pc.in

@@ -6,7 +6,7 @@ includedir=${prefix}/@CMAKE_INSTALL_INCLUDEDIR@
 
 Name: Vitastor
 Description: Vitastor client library
-Version: 3.0.6
+Version: 3.0.3
 Libs: -L${libdir} -lvitastor_client
 Cflags: -I${includedir}
 

src/cmd/CMakeLists.txt

@@ -2,6 +2,15 @@ cmake_minimum_required(VERSION 2.8.12)
 
 project(vitastor)
 
+set(OPENAPI_JSON_H "${CMAKE_CURRENT_BINARY_DIR}/openapi.json.h")
+add_custom_command(
+	OUTPUT ${OPENAPI_JSON_H}
+	COMMAND ${CMAKE_COMMAND} -E echo const char* openapi_description = R\\\"json\\\( > ${OPENAPI_JSON_H}
+	COMMAND ${CMAKE_COMMAND} -E cat ${CMAKE_CURRENT_SOURCE_DIR}/openapi.json >> ${OPENAPI_JSON_H}
+	COMMAND ${CMAKE_COMMAND} -E echo "\\)json\\\"\\;" >> ${OPENAPI_JSON_H}
+	DEPENDS openapi.json
+)
+
 # libvitastor_cli.a
 add_library(vitastor_cli STATIC
 	cli_common.cpp
@@ -27,8 +36,11 @@ add_library(vitastor_cli STATIC
 	cli_pool_ls.cpp
 	cli_pool_modify.cpp
 	cli_pool_rm.cpp
+	cli_serve.cpp
+	${OPENAPI_JSON_H}
 )
 target_compile_options(vitastor_cli PUBLIC -fPIC)
+target_include_directories(vitastor_cli PRIVATE ${CMAKE_CURRENT_BINARY_DIR})
 
 # vitastor-cli
 add_executable(vitastor-cli

src/cmd/cli.cpp

@@ -37,6 +37,7 @@ static const char* help_text =
     "  --sort FIELD    Sort by specified field (name, size, used_size, <read|write|delete>_<iops|bps|lat|queue>)\n"
     "  -r|--reverse    Sort in descending order\n"
     "  -n|--count N    Only list first N items\n"
+    "  --ids ID1,ID2   Only list images with specified full IDs\n"
     "  --tree          Show image snapshot/clone tree\n"
     "\n"
     "vitastor-cli create -s|--size <size> [-p|--pool <id|name>] [--parent <parent_name>[@<snapshot>]] <name>\n"
@@ -217,7 +218,7 @@ static const char* help_text =
     "vitastor-cli rm-pool|pool-rm [--force] <id|name>\n"
     "  Remove a pool. Refuses to remove pools with images without --force.\n"
     "\n"
-    "vitastor-cli ls-pools|pool-ls|ls-pool|pools [-l] [--detail] [--sort FIELD] [-r] [-n N] [--stats] [<glob> ...]\n"
+    "vitastor-cli ls-pools|pool-ls|ls-pool|pools [-l] [--detail] [--sort FIELD] [-r] [-n N] [<glob> ...]\n"
     "  List pools (only matching <glob> patterns if passed).\n"
     "  -l|--long       Also report I/O statistics\n"
     "  --detail        Use list format (not table), show all details\n"
@@ -225,6 +226,14 @@ static const char* help_text =
     "  -r|--reverse    Sort in descending order\n"
     "  -n|--count N    Only list first N items\n"
     "\n"
+    "vitastor-cli serve\n"
+    "  Start HTTP server able to handle CLI commands over a REST API. Options:\n"
+    "  --bind_address ADDR  Specify server IP address or addresses, separated by space. Default is 127.0.0.1.\n"
+    "  --port 8080          Specify server port.\n"
+    "  --ssl_cert FILE      Path to server SSL certificate file (PEM format).\n"
+    "  --ssl_key FILE       Path to server SSL private key file.\n"
+    "  --ssl_ca FILE        Path to file with SSL CA certificates used to validate client connections.\n"
+    "\n"
     "Use vitastor-cli --help <command> for command details or vitastor-cli --help --all for all details.\n"
     "\n"
     "GLOBAL OPTIONS:\n"
@@ -319,27 +328,24 @@ static json11::Json::object parse_args(int narg, const char *args[])
     return cfg;
 }
 
-static int run(cli_tool_t *p, json11::Json::object cfg)
+std::function<bool(cli_result_t &)> cli_tool_t::start(json11::Json::object cfg, cli_result_t & result)
 {
-    cli_result_t result = {};
-    p->is_command_line = true;
-    p->parse_config(cfg);
     json11::Json::array cmd = cfg["command"].array_items();
     cfg.erase("command");
     std::function<bool(cli_result_t &)> action_cb;
     if (!cmd.size())
     {
-        result = { .err = EINVAL, .text = "command is missing" };
+        result = { .err = EOPNOTSUPP, .text = "command is missing" };
     }
     else if (cmd[0] == "status")
     {
         // Show cluster status
-        action_cb = p->start_status(cfg);
+        action_cb = start_status(cfg);
     }
     else if (cmd[0] == "df")
     {
         // Show pool space stats
-        action_cb = p->start_pool_ls(cfg);
+        action_cb = start_pool_ls(cfg);
     }
     else if (cmd[0] == "ls")
     {
@@ -349,7 +355,7 @@ static int run(cli_tool_t *p, json11::Json::object cfg)
             cmd.erase(cmd.begin(), cmd.begin()+1);
             cfg["names"] = cmd;
         }
-        action_cb = p->start_ls(cfg);
+        action_cb = start_ls(cfg);
     }
     else if (cmd[0] == "snap-create")
     {
@@ -364,7 +370,7 @@ static int run(cli_tool_t *p, json11::Json::object cfg)
         {
             cfg["image"] = name.substr(0, pos);
             cfg["snapshot"] = name.substr(pos + 1);
-            action_cb = p->start_create(cfg);
+            action_cb = start_create(cfg);
         }
     }
     else if (cmd[0] == "create")
@@ -374,7 +380,7 @@ static int run(cli_tool_t *p, json11::Json::object cfg)
         {
             cfg["image"] = cmd[1];
         }
-        action_cb = p->start_create(cfg);
+        action_cb = start_create(cfg);
     }
     else if (cmd[0] == "modify")
     {
@@ -383,12 +389,12 @@ static int run(cli_tool_t *p, json11::Json::object cfg)
         {
             cfg["image"] = cmd[1];
         }
-        action_cb = p->start_modify(cfg);
+        action_cb = start_modify(cfg);
     }
     else if (cmd[0] == "rm-data")
     {
         // Delete inode data
-        action_cb = p->start_rm_data(cfg);
+        action_cb = start_rm_data(cfg);
     }
     else if (cmd[0] == "rm-osd")
     {
@@ -398,7 +404,7 @@ static int run(cli_tool_t *p, json11::Json::object cfg)
             cmd.erase(cmd.begin(), cmd.begin()+1);
             cfg["osd_id"] = cmd;
         }
-        action_cb = p->start_rm_osd(cfg);
+        action_cb = start_rm_osd(cfg);
     }
     else if (cmd[0] == "merge-data")
     {
@@ -409,7 +415,7 @@ static int run(cli_tool_t *p, json11::Json::object cfg)
             if (cmd.size() > 2)
                 cfg["to"] = cmd[2];
         }
-        action_cb = p->start_merge(cfg);
+        action_cb = start_merge(cfg);
     }
     else if (cmd[0] == "flatten")
     {
@@ -418,7 +424,7 @@ static int run(cli_tool_t *p, json11::Json::object cfg)
         {
             cfg["image"] = cmd[1];
         }
-        action_cb = p->start_flatten(cfg);
+        action_cb = start_flatten(cfg);
     }
     else if (cmd[0] == "dd")
     {
@@ -432,16 +438,31 @@ static int run(cli_tool_t *p, json11::Json::object cfg)
                 cfg[arg.substr(0, p)] = arg.substr(p+1);
             }
         }
-        action_cb = p->start_dd(cfg);
+        action_cb = start_dd(cfg);
     }
     else if (cmd[0] == "rm")
     {
         // Remove multiple snapshots and rebase their children
-        if (cfg["exact"].bool_value() || cfg["matching"].bool_value())
+        if (cfg["names"].is_array())
+        {
+            cfg["globs"] = cfg["names"];
+            cfg.erase("names");
+            cfg["exact"] = true;
+            cfg["matching"] = false;
+            action_cb = start_rm_wildcard(cfg);
+        }
+        else if (cfg["matching"].is_array())
+        {
+            cfg["globs"] = cfg["matching"];
+            cfg["exact"] = false;
+            cfg["matching"] = true;
+            action_cb = start_rm_wildcard(cfg);
+        }
+        else if (cfg["exact"].bool_value() || cfg["matching"].bool_value())
         {
             cmd.erase(cmd.begin(), cmd.begin()+1);
             cfg["globs"] = cmd;
-            action_cb = p->start_rm_wildcard(cfg);
+            action_cb = start_rm_wildcard(cfg);
         }
         else
         {
@@ -451,41 +472,41 @@ static int run(cli_tool_t *p, json11::Json::object cfg)
                 if (cmd.size() > 2)
                     cfg["to"] = cmd[2];
             }
-            action_cb = p->start_rm(cfg);
+            action_cb = start_rm(cfg);
         }
     }
     else if (cmd[0] == "describe")
     {
         // Describe unclean objects
-        action_cb = p->start_describe(cfg);
+        action_cb = start_describe(cfg);
     }
     else if (cmd[0] == "fix")
     {
         // Fix inconsistent objects (by deleting some copies)
-        action_cb = p->start_fix(cfg);
+        action_cb = start_fix(cfg);
     }
     else if (cmd[0] == "alloc-osd")
     {
         // Allocate a new OSD number
-        action_cb = p->start_alloc_osd(cfg);
+        action_cb = start_alloc_osd(cfg);
     }
     else if (cmd[0] == "osd-tree")
     {
         // Print OSD tree
-        action_cb = p->start_osd_tree(cfg);
+        cfg["as_tree"] = true;
+        action_cb = start_osd_tree(cfg);
     }
     else if (cmd[0] == "osds" || cmd[0] == "ls-osds" || cmd[0] == "ls-osd" || cmd[0] == "osd-ls")
     {
         // Print OSD list
-        cfg["flat"] = true;
-        action_cb = p->start_osd_tree(cfg);
+        action_cb = start_osd_tree(cfg);
     }
     else if (cmd[0] == "modify-osd")
     {
         // Modify OSD configuration
         if (cmd.size() > 1)
             cfg["osd_num"] = cmd[1];
-        action_cb = p->start_modify_osd(cfg);
+        action_cb = start_modify_osd(cfg);
     }
     else if (cmd[0] == "pg-list" || cmd[0] == "pg-ls" || cmd[0] == "list-pg" || cmd[0] == "ls-pg" || cmd[0] == "ls-pgs" || cmd[0] == "pgs")
     {
@@ -495,7 +516,7 @@ static int run(cli_tool_t *p, json11::Json::object cfg)
             cmd.erase(cmd.begin(), cmd.begin()+1);
             cfg["pg_state"] = cmd;
         }
-        action_cb = p->start_pg_list(cfg);
+        action_cb = start_pg_list(cfg);
     }
     else if (cmd[0] == "create-pool" || cmd[0] == "pool-create")
     {
@@ -504,16 +525,16 @@ static int run(cli_tool_t *p, json11::Json::object cfg)
         {
             cfg["name"] = cmd[1];
         }
-        action_cb = p->start_pool_create(cfg);
+        action_cb = start_pool_create(cfg);
     }
     else if (cmd[0] == "modify-pool" || cmd[0] == "pool-modify")
     {
         // Modify existing pool
         if (cmd.size() > 1)
         {
-            cfg["old_name"] = cmd[1];
+            cfg["pool"] = cmd[1];
         }
-        action_cb = p->start_pool_modify(cfg);
+        action_cb = start_pool_modify(cfg);
     }
     else if (cmd[0] == "rm-pool" || cmd[0] == "pool-rm")
     {
@@ -522,7 +543,7 @@ static int run(cli_tool_t *p, json11::Json::object cfg)
         {
             cfg["pool"] = cmd[1];
         }
-        action_cb = p->start_pool_rm(cfg);
+        action_cb = start_pool_rm(cfg);
     }
     else if (cmd[0] == "ls-pool" || cmd[0] == "pool-ls" || cmd[0] == "ls-pools" || cmd[0] == "pools")
     {
@@ -533,12 +554,26 @@ static int run(cli_tool_t *p, json11::Json::object cfg)
             cmd.erase(cmd.begin(), cmd.begin()+1);
             cfg["names"] = cmd;
         }
-        action_cb = p->start_pool_ls(cfg);
+        action_cb = start_pool_ls(cfg);
+    }
+    else if (cmd[0] == "serve")
+    {
+        // Start HTTP server
+        action_cb = start_serve(cfg);
     }
     else
     {
-        result = { .err = EINVAL, .text = "unknown command: "+cmd[0].string_value() };
+        result = { .err = EOPNOTSUPP, .text = "unknown command: "+cmd[0].string_value() };
     }
+    return action_cb;
+}
+
+static int run(cli_tool_t *p, json11::Json::object cfg)
+{
+    cli_result_t result = {};
+    p->is_command_line = true;
+    p->parse_config(cfg);
+    auto action_cb = p->start(cfg, result);
     if (action_cb != NULL)
     {
         // Create client
@@ -550,6 +585,7 @@ static int run(cli_tool_t *p, json11::Json::object cfg)
         {
             result = r;
             action_cb = NULL;
+            p->ringloop->submit();
         });
         // Loop until it completes
         while (action_cb != NULL)

src/cmd/cli.h

@@ -7,6 +7,7 @@
 
 #include "json11/json11.hpp"
 #include "object_id.h"
+#include "osd_id.h"
 #include "ringloop.h"
 #include <functional>
 
@@ -46,6 +47,7 @@ public:
     json11::Json etcd_result;
 
     void parse_config(json11::Json::object & cfg);
+    void parse_api_opts(json11::Json::object & cfg);
     json11::Json parse_tags(std::string tags);
 
     void change_parent(inode_t cur, inode_t new_parent, cli_result_t *result);
@@ -56,8 +58,10 @@ public:
     friend struct snap_flattener_t;
     friend struct snap_remover_t;
 
+    std::function<bool(cli_result_t &)> start(json11::Json::object cfg, cli_result_t & result);
     std::function<bool(cli_result_t &)> start_alloc_osd(json11::Json);
     std::function<bool(cli_result_t &)> start_create(json11::Json);
+    std::function<bool(cli_result_t &)> start_dd(json11::Json);
     std::function<bool(cli_result_t &)> start_describe(json11::Json);
     std::function<bool(cli_result_t &)> start_fix(json11::Json);
     std::function<bool(cli_result_t &)> start_flatten(json11::Json);
@@ -75,8 +79,8 @@ public:
     std::function<bool(cli_result_t &)> start_rm_data(json11::Json);
     std::function<bool(cli_result_t &)> start_rm_osd(json11::Json);
     std::function<bool(cli_result_t &)> start_rm_wildcard(json11::Json);
+    std::function<bool(cli_result_t &)> start_serve(json11::Json);
     std::function<bool(cli_result_t &)> start_status(json11::Json);
-    std::function<bool(cli_result_t &)> start_dd(json11::Json);
 
     // Should be called like loop_and_wait(start_status(), <completion callback>)
     void loop_and_wait(std::function<bool(cli_result_t &)> loop_cb, std::function<void(const cli_result_t &)> complete_cb);

src/cmd/cli_common.cpp

@@ -101,6 +101,16 @@ inode_config_t* cli_tool_t::get_inode_cfg(const std::string & name)
     return NULL;
 }
 
+void cli_tool_t::parse_api_opts(json11::Json::object & cfg)
+{
+    iodepth = cfg["iodepth"].uint64_value();
+    if (!iodepth)
+        iodepth = 32;
+    parallel_osds = cfg["parallel_osds"].uint64_value();
+    if (!parallel_osds)
+        parallel_osds = 4;
+}
+
 void cli_tool_t::parse_config(json11::Json::object & cfg)
 {
     for (auto kv_it = cfg.begin(); kv_it != cfg.end();)
@@ -121,15 +131,10 @@ void cli_tool_t::parse_config(json11::Json::object & cfg)
     else
         color = isatty(1);
     json_output = cfg["json"].bool_value();
-    iodepth = cfg["iodepth"].uint64_value();
-    if (!iodepth)
-        iodepth = 32;
-    parallel_osds = cfg["parallel_osds"].uint64_value();
-    if (!parallel_osds)
-        parallel_osds = 4;
     log_level = cfg["log_level"].int64_value();
     progress = cfg["progress"].uint64_value() ? true : false;
     list_first = cfg["wait_list"].uint64_value() ? true : false;
+    parse_api_opts(cfg);
 }
 
 struct cli_result_looper_t
@@ -153,7 +158,6 @@ void cli_tool_t::loop_and_wait(std::function<bool(cli_result_t &)> loop_cb, std:
             ringloop->unregister_consumer(&looper->consumer);
             looper->loop_cb = NULL;
             looper->complete_cb(looper->result);
-            ringloop->submit();
             delete looper;
             return;
         }

src/cmd/cli_create.cpp

@@ -64,7 +64,8 @@ struct image_creator_t
         }
         if (new_pool_id)
         {
-            if (pools.find(new_pool_id) == pools.end())
+            auto pool_it = pools.find(new_pool_id);
+            if (pool_it == pools.end())
             {
                 result = (cli_result_t){ .err = ENOENT, .text = "Pool "+std::to_string(new_pool_id)+" does not exist" };
                 state = 100;
@@ -194,15 +195,31 @@ resume_3:
         // Save into inode_config for library users to be able to take it from there immediately
         new_cfg.mod_revision = parent->etcd_result["header"]["revision"].uint64_value();
         parent->cli->st_cli.insert_inode_config(new_cfg);
+        auto img = json11::Json::object {
+            { "inode_id", INODE_WITH_POOL(new_pool_id, new_id) },
+            { "inode_num", new_id },
+            { "name", image_name },
+            { "pool_id", (uint64_t)new_pool_id },
+            { "size", size },
+        };
+        {
+            auto new_pool_it = parent->cli->st_cli.pool_config.find(new_pool_id);
+            if (new_pool_it != parent->cli->st_cli.pool_config.end())
+            {
+                img["pool_name"] = new_pool_it->second.name;
+            }
+        }
+        if (new_parent_id)
+        {
+            img["parent_name"] = new_parent;
+            img["parent_inode_id"] = new_parent_id;
+            img["parent_inode_num"] = INODE_NO_POOL(new_parent_id);
+            img["parent_pool_id"] = (uint64_t)INODE_POOL(new_parent_id);
+        }
         result = (cli_result_t){
             .err = 0,
             .text = "Image "+image_name+" created",
-            .data = json11::Json::object {
-                { "name", image_name },
-                { "pool", new_pool_name },
-                { "parent", new_parent },
-                { "size", size },
-            }
+            .data = img,
         };
         state = 100;
     }
@@ -272,13 +289,23 @@ resume_4:
         // Save into inode_config for library users to be able to take it from there immediately
         new_cfg.mod_revision = parent->etcd_result["header"]["revision"].uint64_value();
         parent->cli->st_cli.insert_inode_config(new_cfg);
+        {
+            auto new_pool_it = parent->cli->st_cli.pool_config.find(new_pool_id);
+            new_pool_name = new_pool_it != parent->cli->st_cli.pool_config.end() ? new_pool_it->second.name : "";
+        }
         result = (cli_result_t){
             .err = 0,
             .text = "Snapshot "+image_name+"@"+new_snap+" created",
             .data = json11::Json::object {
-                { "name", image_name+"@"+new_snap },
-                { "pool", (uint64_t)new_pool_id },
-                { "parent", new_parent },
+                { "inode_id", INODE_WITH_POOL(new_pool_id, new_id) },
+                { "inode_num", new_id },
+                { "name", image_name },
+                { "pool_id", (uint64_t)new_pool_id },
+                { "pool_name", new_pool_name },
+                { "parent_name", image_name+"@"+new_snap },
+                { "parent_inode_id", INODE_WITH_POOL(old_pool_id, old_id) },
+                { "parent_inode_num", old_id },
+                { "parent_pool_id", (uint64_t)old_pool_id },
                 { "size", size },
             }
         };
@@ -554,8 +581,16 @@ std::function<bool(cli_result_t &)> cli_tool_t::start_create(json11::Json cfg)
     auto image_creator = new image_creator_t();
     image_creator->parent = this;
     image_creator->image_name = cfg["image"].string_value();
-    image_creator->new_pool_id = cfg["pool"].uint64_value();
-    image_creator->new_pool_name = cfg["pool"].string_value();
+    if (!cfg["pool"].is_null())
+    {
+        image_creator->new_pool_id = cfg["pool"].uint64_value();
+        image_creator->new_pool_name = cfg["pool"].string_value();
+    }
+    else
+    {
+        image_creator->new_pool_id = cfg["pool_id"].uint64_value();
+        image_creator->new_pool_name = cfg["pool_name"].string_value();
+    }
     image_creator->force = cfg["force"].bool_value();
     image_creator->force_size = cfg["force_size"].bool_value();
     if (cfg["image_meta"].is_object())

src/cmd/cli_dd.cpp

@@ -442,7 +442,7 @@ struct cli_dd_t
                 }
                 delete cur_read;
             }
-            else if (!is_zero(read_op->bitmap_buf, (read_op->len/iinfo.in_granularity+7)/8))
+            else if (!is_zero(read_op->bitmap_buf, read_op->len/iinfo.in_granularity/8))
             {
                 vitastor_read(cur_read);
             }
@@ -864,7 +864,7 @@ resume_2:
         // Copy data
         if (iinfo.in_seekable && iseek >= iinfo.in_size)
         {
-            result = (cli_result_t){ .err = -EINVAL, .text = "Input seek position is beyond end of input" };
+            result = (cli_result_t){ .err = EINVAL, .text = "Input seek position is beyond end of input" };
             goto close_end;
         }
         if (!iinfo.iwatch && !iinfo.in_seekable && iseek)

src/cmd/cli_describe.cpp

@@ -57,12 +57,24 @@ struct cli_describe_t
 
     void parse_options(json11::Json cfg)
     {
-        only_pool = cfg["pool"].uint64_value();
-        if (!only_pool && cfg["pool"].is_string())
+        uint64_t pool_id;
+        std::string pool_name;
+        if (!cfg["pool"].is_null())
+        {
+            pool_id = cfg["pool"].uint64_value();
+            pool_name = pool_id ? "" : cfg["pool"].string_value();
+        }
+        else
+        {
+            pool_id = cfg["pool_id"].uint64_value();
+            pool_name = pool_id ? "" : cfg["pool_name"].string_value();
+        }
+        only_pool = pool_id;
+        if (!only_pool && pool_name != "")
         {
             for (auto & pp: parent->cli->st_cli.pool_config)
             {
-                if (pp.second.name == cfg["pool"].string_value())
+                if (pp.second.name == pool_name)
                 {
                     only_pool = pp.first;
                     break;
@@ -106,6 +118,22 @@ struct cli_describe_t
             if (cfg["object_state"].string_value().find("misplaced") != std::string::npos)
                 object_state |= OBJ_MISPLACED;
         }
+        else if (!object_state && cfg["object_state"].is_array())
+        {
+            for (auto & st: cfg["object_state"].array_items())
+            {
+                if (st == "inconsistent")
+                    object_state |= OBJ_INCONSISTENT;
+                else if (st == "corrupted")
+                    object_state |= OBJ_CORRUPTED;
+                else if (st == "incomplete")
+                    object_state |= OBJ_INCOMPLETE;
+                else if (st == "degraded")
+                    object_state |= OBJ_DEGRADED;
+                else if (st == "misplaced")
+                    object_state |= OBJ_MISPLACED;
+            }
+        }
     }
 
     void loop()

src/cmd/cli_ls.cpp

@@ -17,6 +17,7 @@ struct image_lister_t
     std::string list_pool_name;
     std::string sort_field;
     std::set<std::string> only_names;
+    std::vector<uint64_t> only_ids;
     bool reverse = false;
     bool exact = false;
     bool tree = false;
@@ -63,7 +64,6 @@ struct image_lister_t
             auto item = json11::Json::object {
                 { "name", ic.second.name },
                 { "size", ic.second.size },
-                { "used_size", 0 },
                 { "readonly", ic.second.readonly },
                 { "pool_id", (uint64_t)INODE_POOL(ic.second.num) },
                 { "pool_name", good_pool ? pool_it->second.name : "? (ID:"+std::to_string(INODE_POOL(ic.second.num))+")" },
@@ -202,20 +202,33 @@ resume_1:
     json11::Json::array to_list()
     {
         json11::Json::array list;
-        for (auto & kv: stats)
+        if (only_ids.size())
         {
-            if (!only_names.size())
+            for (auto & id: only_ids)
             {
-                list.push_back(kv.second);
-            }
-            else
-            {
-                for (auto & glob: only_names)
+                if (stats.find(id) != stats.end())
                 {
-                    if (exact ? (kv.second["name"].string_value() == glob) : stupid_glob(kv.second["name"].string_value(), glob))
+                    list.push_back(stats[id]);
+                }
+            }
+        }
+        else
+        {
+            for (auto & kv: stats)
+            {
+                if (!only_names.size())
+                {
+                    list.push_back(kv.second);
+                }
+                else
+                {
+                    for (auto & glob: only_names)
                     {
-                        list.push_back(kv.second);
-                        break;
+                        if (exact ? (kv.second["name"].string_value() == glob) : stupid_glob(kv.second["name"].string_value(), glob))
+                        {
+                            list.push_back(kv.second);
+                            break;
+                        }
                     }
                 }
             }
@@ -570,17 +583,30 @@ std::function<bool(cli_result_t &)> cli_tool_t::start_ls(json11::Json cfg)
     lister->parent = this;
     lister->exact = cfg["exact"].bool_value();
     lister->tree = cfg["tree"].bool_value();
-    lister->list_pool_id = cfg["pool"].uint64_value();
-    lister->list_pool_name = lister->list_pool_id ? "" : cfg["pool"].as_string();
+    if (!cfg["pool"].is_null())
+    {
+        lister->list_pool_id = cfg["pool"].uint64_value();
+        lister->list_pool_name = lister->list_pool_id ? "" : cfg["pool"].as_string();
+    }
+    else
+    {
+        lister->list_pool_id = cfg["pool_id"].uint64_value();
+        lister->list_pool_name = lister->list_pool_id ? "" : cfg["pool_name"].string_value();
+    }
     lister->show_stats = cfg["long"].bool_value();
     lister->show_delete = cfg["del"].bool_value();
     lister->sort_field = cfg["sort"].string_value() != "" ? cfg["sort"].string_value() : "name";
     lister->reverse = cfg["reverse"].bool_value();
     lister->max_count = cfg["count"].uint64_value();
+    if (cfg["names"].is_string())
+        lister->only_names.insert(cfg["names"].string_value());
     for (auto & item: cfg["names"].array_items())
-    {
         lister->only_names.insert(item.string_value());
-    }
+    if (cfg["ids"].is_string())
+        for (auto & item: explode(",", cfg["ids"].string_value(), true))
+            lister->only_ids.push_back(stoull_full(item));
+    for (auto & item: cfg["ids"].array_items())
+        lister->only_ids.push_back(item.uint64_value());
     return [lister](cli_result_t & result)
     {
         lister->loop();

src/cmd/cli_merge.cpp

@@ -374,7 +374,7 @@ struct snap_merger_t
         result = (cli_result_t){ .text = "Done, layers from "+from_name+" to "+to_name+" merged into "+target_name, .data = json11::Json::object {
             { "from", from_name },
             { "to", to_name },
-            { "into", target_name },
+            { "target", target_name },
         }};
         state = 100;
     resume_100:

src/cmd/cli_modify.cpp

@@ -88,10 +88,7 @@ struct image_changer_t
             (!new_size && !force_size || cfg.size == new_size || cfg.size >= new_size && inc_size) &&
             (new_name == "" || new_name == image_name))
         {
-            result = (cli_result_t){ .err = 0, .text = "No change", .data = json11::Json::object {
-                { "error_code", 0 },
-                { "error_text", "No change" },
-            }};
+            result = (cli_result_t){ .err = 0, .text = "No change", .data = fill_img(cfg) };
             state = 100;
             return;
         }
@@ -235,15 +232,42 @@ resume_2:
         result = (cli_result_t){
             .err = 0,
             .text = "Image "+image_name+" modified",
-            .data = json11::Json::object {
-                { "name", image_name },
-                { "inode", INODE_NO_POOL(inode_num) },
-                { "pool", (uint64_t)INODE_POOL(inode_num) },
-                { "size", new_size },
-            }
+            .data = fill_img(cfg)
         };
         state = 100;
     }
+
+    json11::Json fill_img(inode_config_t & cfg)
+    {
+        auto img = json11::Json::object {
+            { "inode_id", inode_num },
+            { "inode_num", INODE_NO_POOL(inode_num) },
+            { "name", cfg.name },
+            { "pool_id", (uint64_t)INODE_POOL(inode_num) },
+            { "size", cfg.size },
+            { "readonly", cfg.readonly },
+            { "deleted", cfg.deleted },
+        };
+        {
+            auto pool_it = parent->cli->st_cli.pool_config.find(INODE_POOL(inode_num));
+            if (pool_it != parent->cli->st_cli.pool_config.end())
+            {
+                img["pool_name"] = pool_it->second.name;
+            }
+        }
+        if (cfg.parent_id)
+        {
+            auto parent_it = parent->cli->st_cli.inode_config.find(cfg.parent_id);
+            if (parent_it != parent->cli->st_cli.inode_config.end())
+            {
+                img["parent_name"] = parent_it->second.name;
+            }
+            img["parent_inode_id"] = cfg.parent_id;
+            img["parent_inode_num"] = INODE_NO_POOL(cfg.parent_id);
+            img["parent_pool_id"] = (uint64_t)INODE_POOL(cfg.parent_id);
+        }
+        return img;
+    }
 };
 
 std::function<bool(cli_result_t &)> cli_tool_t::start_modify(json11::Json cfg)

src/cmd/cli_osd_tree.cpp

@@ -41,7 +41,7 @@ struct osd_tree_printer_t
 {
     cli_tool_t *parent;
     json11::Json cfg;
-    bool flat = false;
+    bool as_tree = false;
     bool show_stats = false;
 
     int state = 0;
@@ -209,11 +209,14 @@ resume_1:
             for (int i = 1; i < node_seq.size(); i++)
             {
                 auto & node = placement_tree->nodes.at(node_seq[i]);
-                fmt_items.push_back(json11::Json::object{
-                    { "type", node.level },
-                    { "name", node.name },
-                    { "parent", node.parent },
-                });
+                if (as_tree)
+                {
+                    fmt_items.push_back(json11::Json::object{
+                        { "type", node.level },
+                        { "name", node.name },
+                        { "parent", node.parent },
+                    });
+                }
                 for (uint64_t osd_num: node.child_osds)
                 {
                     auto & osd = placement_tree->osds.at(osd_num);
@@ -221,17 +224,22 @@ resume_1:
                         { "type", "osd" },
                         { "name", osd.num },
                         { "parent", node.name },
-                        { "up", osd.up ? "up" : "down" },
+                        { "up", osd.up },
                         { "size", osd.size },
                         { "free", osd.free },
                         { "reweight", osd.reweight },
                         { "noout", osd.noout },
                         { "tags", osd.tags },
-                        { "block", (uint64_t)osd.block_size },
-                        { "bitmap", (uint64_t)osd.bitmap_granularity },
-                        { "commit", osd.immediate_commit == IMMEDIATE_NONE ? "none" : (osd.immediate_commit == IMMEDIATE_ALL ? "all" : "small") },
-                        { "op_stats", osd_stats[osd_num]["op_stats"] },
+                        { "data_block_size", (uint64_t)osd.block_size },
+                        { "bitmap_granularity", (uint64_t)osd.bitmap_granularity },
+                        { "immediate_commit", osd.immediate_commit == IMMEDIATE_NONE ? "none" : (osd.immediate_commit == IMMEDIATE_ALL ? "all" : "small") },
                     };
+                    if (show_stats)
+                    {
+                        json_osd["op_stats"] = osd_stats[osd_num]["op_stats"];
+                        json_osd["subop_stats"] = osd_stats[osd_num]["subop_stats"];
+                        json_osd["recovery_stats"] = osd_stats[osd_num]["recovery_stats"];
+                    }
                     if (osd_stats[osd_num]["slow_ops_primary"].uint64_value() > 0)
                     {
                         json_osd["slow_ops_primary"] = osd_stats[osd_num]["slow_ops_primary"];
@@ -249,7 +257,7 @@ resume_1:
         for (int i = 1; i < node_seq.size(); i++)
         {
             auto & node = placement_tree->nodes.at(node_seq[i]);
-            if (!flat)
+            if (as_tree)
             {
                 fmt_items.push_back(json11::Json::object{
                     { "type", str_repeat("  ", indents[i]) + node.level },
@@ -257,7 +265,7 @@ resume_1:
                 });
             }
             std::string parent = node.name;
-            if (flat)
+            if (!as_tree)
             {
                 auto cur = &placement_tree->nodes.at(node.name);
                 while (cur->parent != "" && cur->parent != node.name)
@@ -270,7 +278,7 @@ resume_1:
             {
                 auto & osd = placement_tree->osds.at(osd_num);
                 auto fmt = json11::Json::object{
-                    { "type", (flat ? "osd" : str_repeat("  ", indents[i]+1) + "osd") },
+                    { "type", (!as_tree ? "osd" : str_repeat("  ", indents[i]+1) + "osd") },
                     { "name", osd.num },
                     { "parent", parent },
                     { "up", osd.up ? "up" : "down" },
@@ -300,7 +308,7 @@ resume_1:
             }
         }
         json11::Json::array cols;
-        if (!flat)
+        if (as_tree)
         {
             cols.push_back(json11::Json::object{
                 { "key", "type" },
@@ -309,9 +317,9 @@ resume_1:
         }
         cols.push_back(json11::Json::object{
             { "key", "name" },
-            { "title", flat ? "OSD" : "NAME" },
+            { "title", !as_tree ? "OSD" : "NAME" },
         });
-        if (flat)
+        if (!as_tree)
         {
             cols.push_back(json11::Json::object{
                 { "key", "parent" },
@@ -414,7 +422,7 @@ std::function<bool(cli_result_t &)> cli_tool_t::start_osd_tree(json11::Json cfg)
     auto osd_tree_printer = new osd_tree_printer_t();
     osd_tree_printer->parent = this;
     osd_tree_printer->cfg = cfg;
-    osd_tree_printer->flat = cfg["flat"].bool_value();
+    osd_tree_printer->as_tree = cfg["as_tree"].bool_value();
     osd_tree_printer->show_stats = cfg["long"].bool_value();
     return [osd_tree_printer](cli_result_t & result)
     {

src/cmd/cli_pg_ls.cpp

@@ -282,10 +282,16 @@ std::function<bool(cli_result_t &)> cli_tool_t::start_pg_list(json11::Json cfg)
 {
     auto pg_lister = new pg_lister_t();
     pg_lister->parent = this;
-    if (cfg["pool"].uint64_value())
+    if (!cfg["pool"].is_null())
+    {
         pg_lister->pool_id = cfg["pool"].uint64_value();
+        pg_lister->pool_name = pg_lister->pool_id ? "" : cfg["pool"].string_value();
+    }
     else
-        pg_lister->pool_name = cfg["pool"].string_value();
+    {
+        pg_lister->pool_id = cfg["pool_id"].uint64_value();
+        pg_lister->pool_name = pg_lister->pool_id ? "" : cfg["pool_name"].string_value();
+    }
     for (auto & st: cfg["pg_state"].array_items())
         pg_lister->pg_state.push_back(st.string_value());
     if (cfg["pg_state"].is_string())

src/cmd/cli_pool_create.cpp

@@ -213,7 +213,7 @@ resume_3:
                     if (failure_domain != "osd")
                         pool_err += "\n- different parent '"+failure_domain+"' nodes";
                     result = (cli_result_t){
-                        .err = EINVAL,
+                        .err = EBUSY,
                         .text = pool_err,
                     };
                     state = 100;

src/cmd/cli_pool_ls.cpp

@@ -206,7 +206,7 @@ resume_1:
                 { "space_efficiency", pool_stats[pool_cfg.id]["space_efficiency"].number_value() },
                 { "pg_real_size", pool_stats[pool_cfg.id]["pg_real_size"].uint64_value() },
                 { "osd_count", (uint64_t)pg_per_osd.size() },
-                { "backfillfull", pool_cfg.backfillfull },
+                { "backfillfull", !!pool_cfg.backfillfull },
             };
         }
         // Include full pool config
@@ -546,6 +546,10 @@ resume_3:
                 { "write_fmt", "Write" },
                 { "delete_fmt", "Delete" },
             };
+            if (sort_field == "osd_tags" || sort_field == "primary_affinity_tags")
+            {
+                sort_field += "_fmt";
+            }
             auto list = to_list();
             size_t title_len = 0;
             for (auto & item: list)
@@ -666,15 +670,12 @@ std::function<bool(cli_result_t &)> cli_tool_t::start_pool_ls(json11::Json cfg)
     lister->show_stats = cfg["long"].bool_value();
     lister->detailed = cfg["detail"].bool_value();
     lister->sort_field = cfg["sort"].string_value();
-    if ((lister->sort_field == "osd_tags") ||
-        (lister->sort_field == "primary_affinity_tags" ))
-        lister->sort_field = lister->sort_field + "_fmt";
     lister->reverse = cfg["reverse"].bool_value();
     lister->max_count = cfg["count"].uint64_value();
+    if (cfg["names"].is_string())
+        lister->only_names.insert(cfg["names"].string_value());
     for (auto & item: cfg["names"].array_items())
-    {
         lister->only_names.insert(item.string_value());
-    }
     return [lister](cli_result_t & result)
     {
         lister->loop();

src/cmd/cli_pool_modify.cpp

@@ -39,16 +39,21 @@ struct pool_changer_t
             goto resume_1;
         else if (state == 2)
             goto resume_2;
-        pool_id = stoull_full(cfg["old_name"].string_value());
-        if (!pool_id)
+        if (!cfg["pool"].is_null())
         {
-            pool_name = cfg["old_name"].string_value();
-            if (pool_name == "")
-            {
-                result = (cli_result_t){ .err = ENOENT, .text = "Pool ID or name is required to modify it" };
-                state = 100;
-                return;
-            }
+            pool_id = cfg["pool"].uint64_value();
+            pool_name = pool_id ? "" : cfg["pool"].string_value();
+        }
+        else
+        {
+            pool_id = cfg["id"].uint64_value();
+            pool_name = pool_id ? "" : cfg["old_name"].string_value();
+        }
+        if (!pool_id && pool_name == "")
+        {
+            result = (cli_result_t){ .err = ENOENT, .text = "Pool ID or name is required to modify it" };
+            state = 100;
+            return;
         }
 resume_0:
         // Get pools from etcd
@@ -179,7 +184,7 @@ resume_2:
         result = (cli_result_t){
             .err = 0,
             .text = "Pool "+pool_name+" updated",
-            .data = new_pools,
+            .data = new_pools[std::to_string(pool_id)],
         };
         state = 100;
     }

src/cmd/cli_pool_rm.cpp

@@ -196,7 +196,9 @@ resume_3:
         result = (cli_result_t){
             .err = 0,
             .text = "Pool "+pool_name+" deleted",
-            .data = new_pools
+            .data = json11::Json::object{
+                {"ok", true},
+            },
         };
         state = 100;
     }
@@ -207,8 +209,16 @@ std::function<bool(cli_result_t &)> cli_tool_t::start_pool_rm(json11::Json cfg)
     auto pool_remover = new pool_remover_t();
     pool_remover->parent = this;
 
-    pool_remover->pool_id = cfg["pool"].uint64_value();
-    pool_remover->pool_name = pool_remover->pool_id ? "" : cfg["pool"].as_string();
+    if (!cfg["pool"].is_null())
+    {
+        pool_remover->pool_id = cfg["pool"].uint64_value();
+        pool_remover->pool_name = pool_remover->pool_id ? "" : cfg["pool"].as_string();
+    }
+    else
+    {
+        pool_remover->pool_id = cfg["pool_id"].uint64_value();
+        pool_remover->pool_name = pool_remover->pool_id ? "" : cfg["pool_name"].as_string();
+    }
 
     pool_remover->force = !cfg["force"].is_null();
 

src/cmd/cli_rm_data.cpp

@@ -264,7 +264,7 @@ struct rm_inode_t
             {
                 // Error
                 result = (cli_result_t){
-                    .err = EIO,
+                    .err = EBUSY,
                     .text = "Failed: some blocks were not removed",
                     .data = data,
                 };

src/cmd/cli_rm_osd.cpp

@@ -181,6 +181,7 @@ struct rm_osd_t
             if (dry_run || is_dataloss && !force_dataloss || is_warning && !force_warning)
             {
                 result.err = is_dataloss && !force_dataloss || is_warning && !force_warning ? EBUSY : 0;
+                result.data = json11::Json::object{ { "pool_effects", pool_effects } };
                 state = 100;
                 return;
             }
@@ -332,6 +333,9 @@ struct rm_osd_t
         ids = (osd_ids.size() > 1 ? "OSDs " : "OSD ")+ids+(osd_ids.size() > 1 ? " are" : " is")+" removed from etcd";
         state = 100;
         result.text = (result.text != "" ? ids+"\n"+result.text : ids);
+        result.data = json11::Json::object {
+            {"deleted_osds", osd_ids},
+        };
         result.err = 0;
     }
 

src/cmd/cli_rm_wildcard.cpp

@@ -195,6 +195,8 @@ std::function<bool(cli_result_t &)> cli_tool_t::start_rm_wildcard(json11::Json c
     auto wildcard_remover = new wildcard_remover_t();
     wildcard_remover->parent = this;
     wildcard_remover->cfg = cfg;
+    if (cfg["globs"].is_string())
+        wildcard_remover->globs.push_back(cfg["globs"].string_value());
     for (auto & glob: cfg["globs"].array_items())
         wildcard_remover->globs.push_back(glob.string_value());
     wildcard_remover->exact = cfg["exact"].bool_value();

src/cmd/cli_serve.cpp

@@ -0,0 +1,439 @@
+// Copyright (c) Vitaliy Filippov, 2019+
+// License: VNPL-1.1 (see README.md for details)
+
+#include <signal.h>
+#include <sys/socket.h>
+#include <netinet/tcp.h>
+#include <stdexcept>
+#include "cli.h"
+#include "cluster_client.h"
+#include "epoll_manager.h"
+#include "http_client.h"
+#include "str_util.h"
+#include "json_util.h"
+#include "addr_util.h"
+#include "openapi.json.h"
+
+struct cli_serve_conn_t
+{
+    int peer_fd = 0;
+    std::string peer_addr;
+    http_co_t *co = NULL;
+    cli_tool_t *p = NULL;
+    cli_result_t result;
+    bool keepalive = false;
+    bool closed = false;
+    timespec request_time;
+    std::string request_method;
+    std::string request_path;
+    std::string request_body;
+    std::string response_type;
+    std::function<bool(cli_result_t &)> action_cb;
+};
+
+struct cli_serve_path_t
+{
+    std::string cmd;
+    bool allow_get;
+};
+
+// Serve vitastor-cli commands over HTTP in JSON format
+struct cli_serve_t
+{
+    std::map<std::string, cli_serve_path_t> cmd_paths = {
+        {"data/delete",   {"rm-data", false}},
+        {"data/describe", {"describe", true}},
+        {"data/fix",      {"fix", false}},
+        {"data/merge",    {"merge-data", false}},
+        {"image/create",  {"create", false}},
+        {"image/delete",  {"rm", false}},
+        {"image/flatten", {"flatten", false}},
+        {"image/list",    {"ls", true}},
+        {"image/modify",  {"modify", false}},
+        {"osd/alloc",     {"alloc-osd", false}},
+        {"osd/delete",    {"rm-osd", false}},
+        {"osd/list",      {"ls-osd", true}},
+        {"osd/modify",    {"modify-osd", false}},
+        {"pg/list",       {"ls-pgs", true}},
+        {"pool/create",   {"create-pool", false}},
+        {"pool/delete",   {"rm-pool", false}},
+        {"pool/list",     {"pools", true}},
+        {"pool/modify",   {"modify-pool", false}},
+        {"status",        {"status", true}},
+    };
+
+    cli_tool_t *parent = NULL;
+    json11::Json options;
+    cli_result_t result;
+
+    bool log_body = false;
+    bool stop = false;
+    std::vector<std::string> bind_addresses;
+    int port = 0;
+    int listen_backlog = 0;
+    bool ssl = false;
+    std::vector<int> listen_fds;
+    http_context_t *http_ctx = NULL;
+    std::set<cli_serve_conn_t*> connections;
+
+    int state = 0;
+
+    bool is_done()
+    {
+        return state == 100;
+    }
+
+    void loop()
+    {
+        if (state == 1)
+            goto resume_1;
+        else if (state == 2)
+            goto resume_2;
+        else if (state == 100)
+            return;
+        if (options["bind_address"].is_string())
+            bind_addresses = explode(" ", options["bind_address"].string_value(), true);
+        else
+            bind_addresses.push_back("127.0.0.1");
+        port = options["port"].uint64_value();
+        if (!port)
+            port = 8080;
+        else if (port < 0 || port > 65535)
+        {
+            result = (cli_result_t){ .err = EINVAL, .text = "HTTP port can't be larger than 65536" };
+            state = 100;
+            return;
+        }
+        listen_backlog = options["listen_backlog"].uint64_value();
+        if (!listen_backlog)
+            listen_backlog = 128;
+        ssl = json_is_true(options["ssl"]);
+        if (ssl)
+        {
+            std::string ssl_cert = options["ssl_cert"].string_value();
+            std::string ssl_key = options["ssl_key"].string_value();
+            std::string ssl_ca = options["ssl_ca"].string_value();
+            std::string error;
+            http_ctx = http_context_init(ssl_cert, ssl_key, ssl_ca, ssl_ca != "", error);
+            if (error != "")
+            {
+                result = (cli_result_t){ .err = EINVAL, .text = error };
+                state = 100;
+                return;
+            }
+        }
+        for (auto & bind_address: bind_addresses)
+        {
+            int listen_fd = create_and_bind_socket(bind_address, port, listen_backlog, NULL);
+            fcntl(listen_fd, F_SETFL, fcntl(listen_fd, F_GETFL, 0) | O_NONBLOCK);
+            parent->epmgr->set_fd_handler(listen_fd, false, [this](int fd, int events)
+            {
+                accept_connections(fd);
+            });
+            listen_fds.push_back(listen_fd);
+        }
+    resume_1:
+        if (!stop)
+        {
+            state = 1;
+            return;
+        }
+        for (auto conn: connections)
+        {
+            http_close(conn->co);
+        }
+    resume_2:
+        // Wait for all connections to finish
+        if (connections.size() > 0)
+        {
+            state = 2;
+            return;
+        }
+        if (http_ctx)
+        {
+            http_context_destroy(http_ctx);
+            http_ctx = NULL;
+        }
+        state = 100;
+    }
+
+    void accept_connections(int listen_fd)
+    {
+        sockaddr_storage addr;
+        socklen_t peer_addr_size = sizeof(addr);
+        int peer_fd;
+        while ((peer_fd = accept(listen_fd, (sockaddr*)&addr, &peer_addr_size)) >= 0)
+        {
+            auto peer_addr_str = addr_to_string(addr);
+            assert(peer_fd != 0);
+            timespec ts;
+            clock_gettime(CLOCK_REALTIME, &ts);
+            printf("[%s.%03ju] New connection %d from %s\n", format_datetime(ts.tv_sec).c_str(), (uint64_t)ts.tv_nsec/1000000,
+                peer_fd, peer_addr_str.c_str());
+            fcntl(peer_fd, F_SETFL, fcntl(peer_fd, F_GETFL, 0) | O_NONBLOCK);
+            int one = 1;
+            setsockopt(peer_fd, SOL_TCP, TCP_NODELAY, &one, sizeof(one));
+            cli_serve_conn_t *conn = new cli_serve_conn_t;
+            conn->peer_fd = peer_fd;
+            conn->peer_addr = peer_addr_str;
+            conn->co = http_init(parent->epmgr->tfd, http_ctx);
+            http_serve(conn->co, peer_fd, (http_options_t){ .ssl = ssl }, [this, conn](http_message_t *msg)
+            {
+                process_request(conn, msg);
+            });
+            connections.insert(conn);
+            // Try to accept next connection
+            peer_addr_size = sizeof(addr);
+        }
+        if (peer_fd == -1 && errno != EAGAIN)
+        {
+            throw std::runtime_error(std::string("accept: ") + strerror(errno));
+        }
+    }
+
+    int map_to_http(int err, std::string *text)
+    {
+        int code = 0;
+        if (err == EINVAL)
+        {
+            code = 400;
+            if (text)
+                *text = "Bad Request";
+        }
+        else if (err == EOPNOTSUPP)
+        {
+            code = 404;
+            if (text)
+                *text = "Not Found";
+        }
+        else if (err == ENOSYS)
+        {
+            code = 405;
+            if (text)
+                *text = "Method Not Allowed";
+        }
+        else if (err == EAGAIN)
+        {
+            code = 409;
+            if (text)
+                *text = "Update Conflict";
+        }
+        else if (err == ENOTEMPTY || err == EEXIST || err == ENOENT || err == EBUSY)
+        {
+            code = 412;
+            if (text)
+                *text = "Precondition Failed";
+        }
+        else /*if (err == EIO || err == EBADF)*/
+        {
+            code = 500;
+            if (text)
+                *text = "Internal Server Error";
+        }
+        return code;
+    }
+
+    std::string cli_http_response(cli_serve_conn_t *conn)
+    {
+        timespec now;
+        clock_gettime(CLOCK_REALTIME, &now);
+        int code = 200;
+        std::string response;
+        if (conn->result.err)
+        {
+            std::string status_line;
+            code = map_to_http(conn->result.err, &status_line);
+            response = "HTTP/1.1 "+std::to_string(code)+" "+status_line+"\r\n";
+        }
+        else
+        {
+            response = "HTTP/1.1 200 OK\r\n";
+        }
+        response += (conn->keepalive
+            ? "Connection: keep-alive\r\n"
+            : "Connection: close\r\n");
+        std::string body;
+        if (!conn->result.data.is_null())
+        {
+            response += "Content-Type: application/json\r\n";
+            body = conn->result.data.dump();
+        }
+        else
+        {
+            if (!conn->response_type.empty())
+                response += "Content-Type: "+conn->response_type+"\r\n";
+            else
+                response += "Content-Type: text/plain; charset=utf-8\r\n";
+            body = conn->result.text;
+        }
+        response += "Content-Length: "+std::to_string(body.size())+"\r\n\r\n";
+        response += body;
+        if (conn->request_method.find("\n") != std::string::npos)
+            conn->request_method = str_replace(conn->request_method, "\n", "%0a");
+        if (conn->request_method.find(" ") != std::string::npos)
+            conn->request_method = str_replace(conn->request_method, " ", "%20");
+        if (conn->request_path.find("\n") != std::string::npos)
+            conn->request_path = str_replace(conn->request_path, "\n", "%0a");
+        if (conn->request_path.find(" ") != std::string::npos)
+            conn->request_path = str_replace(conn->request_path, " ", "%20");
+        uint64_t response_time = (now.tv_sec-conn->request_time.tv_sec)*1000 + (now.tv_nsec-conn->request_time.tv_nsec)/1000000;
+        printf("[%s.%03ju] %s %s %s %d %.03f sec\n",
+            format_datetime(now.tv_sec).c_str(), (uint64_t)now.tv_nsec/1000000,
+            conn->peer_addr.c_str(), conn->request_method.c_str(), conn->request_path.c_str(), code,
+            response_time/1000.0);
+        if (log_body)
+        {
+            if (conn->request_body.find("\n") != std::string::npos)
+                conn->request_body = str_replace(conn->request_body, "\n", " ");
+            if (conn->request_body.size())
+                printf("  %s\n", conn->request_body.c_str());
+            printf("  %s\n", body.c_str());
+        }
+        return response;
+    }
+
+    void process_request(cli_serve_conn_t *conn, http_message_t *msg)
+    {
+        timespec ts;
+        clock_gettime(CLOCK_REALTIME, &ts);
+        if (!msg->error.empty())
+        {
+            // connection is closed
+            fprintf(stderr, "[%s.%03ju] Connection %d closed: %s\n", format_datetime(ts.tv_sec).c_str(),
+                (uint64_t)ts.tv_nsec/1000000, conn->peer_fd, msg->error.c_str());
+            if (conn->p)
+                conn->closed = true;
+            else
+            {
+                connections.erase(conn);
+                http_destroy(conn->co);
+                delete conn;
+            }
+            return;
+        }
+        conn->keepalive = msg->headers.find("connection") != msg->headers.end() &&
+            msg->headers.at("connection") == "keep-alive";
+        conn->p = new cli_tool_t;
+        conn->p->iodepth = parent->iodepth;
+        conn->p->parallel_osds = parent->parallel_osds;
+        conn->p->json_output = true;
+        conn->p->ringloop = parent->ringloop;
+        conn->p->epmgr = parent->epmgr;
+        conn->p->cli = parent->cli;
+        conn->p->is_command_line = false;
+        // Parse request
+        auto req_line = explode(" ", msg->status_line, true);
+        if (req_line.size() < 2)
+        {
+            if (req_line[0] == "")
+                req_line[0] = "-";
+            req_line.push_back("-");
+        }
+        conn->request_time = ts;
+        conn->request_method = std::move(req_line[0]);
+        conn->request_path = std::move(req_line[1]);
+        conn->request_body = std::move(msg->body);
+        conn->response_type = "";
+        auto ctype = msg->headers["content-type"];
+        if (conn->request_method != "GET" && conn->request_method != "POST")
+        {
+            conn->result = { .err = ENOSYS, .text = "Unsupported request method "+conn->request_method };
+        }
+        else if (ctype != (conn->request_method == "GET" ? "" : "application/json"))
+        {
+            conn->result = { .err = EINVAL, .text = "Unsupported Content-Type: "+ctype+" for "+conn->request_method+" requests" };
+        }
+        else
+        {
+            auto uri = explode("?", conn->request_path, true);
+            uri[0] = trim(uri[0], "/");
+            auto cmd_it = cmd_paths.find(uri[0]);
+            if (uri[0] == "")
+            {
+                std::string text = "Supported APIs:\n\n- GET /openapi\n";
+                for (auto & pp: cmd_paths)
+                {
+                    text += (pp.second.allow_get ? "- GET" : "- POST") + (" /" + pp.first) + "\n";
+                }
+                conn->result = { .text = text };
+            }
+            else if (uri[0] == "openapi")
+            {
+                conn->response_type = "application/json";
+                conn->result = { .text = openapi_description };
+            }
+            else if (cmd_it == cmd_paths.end())
+            {
+                conn->result = { .err = EOPNOTSUPP, .text = "unknown command: "+uri[0] };
+            }
+            else if (conn->request_method == "GET" && !cmd_it->second.allow_get)
+            {
+                conn->result = { .err = ENOSYS, .text = "method /"+uri[0]+" only allows POST requests" };
+            }
+            else
+            {
+                std::string error;
+                json11::Json::object cfg;
+                if (conn->request_method == "POST")
+                {
+                    cfg = json11::Json::parse(conn->request_body, error).object_items();
+                }
+                else
+                {
+                    // Parse URI
+                    cfg = parse_uri_params(uri[1]);
+                }
+                if (error != "")
+                {
+                    conn->result = { .err = EINVAL, .text = "Invalid JSON in body: "+error };
+                }
+                else
+                {
+                    cfg["command"] = json11::Json::array{cmd_it->second.cmd};
+                    conn->p->parse_api_opts(cfg);
+                    conn->action_cb = conn->p->start(cfg, conn->result);
+                }
+            }
+        }
+        if (!conn->action_cb)
+        {
+            http_reply(conn->co, cli_http_response(conn));
+            delete conn->p;
+            conn->p = NULL;
+            return;
+        }
+        conn->p->loop_and_wait(conn->action_cb, [this, conn](const cli_result_t & r)
+        {
+            conn->result = r;
+            conn->action_cb = NULL;
+            delete conn->p;
+            conn->p = NULL;
+            if (!conn->closed)
+                http_reply(conn->co, cli_http_response(conn));
+            else
+            {
+                connections.erase(conn);
+                http_destroy(conn->co);
+                delete conn;
+            }
+        });
+    }
+};
+
+std::function<bool(cli_result_t &)> cli_tool_t::start_serve(json11::Json cfg)
+{
+    auto server = new cli_serve_t();
+    server->parent = this;
+    server->options = cfg;
+    return [server](cli_result_t & result)
+    {
+        server->loop();
+        if (server->is_done())
+        {
+            result = server->result;
+            delete server;
+            return true;
+        }
+        return false;
+    };
+}

src/cmd/cli_status.cpp

@@ -18,6 +18,7 @@ struct status_printer_t
 {
     cli_tool_t *parent;
 
+    cli_result_t result;
     int state = 0;
     json11::Json::array mon_members;
     json11::Json agg_stats;
@@ -89,7 +90,7 @@ resume_2:
             return;
         if (parent->etcd_err.err)
         {
-            fprintf(stderr, "%s\n", parent->etcd_err.text.c_str());
+            result = parent->etcd_err;
             state = 100;
             return;
         }
@@ -107,7 +108,8 @@ resume_2:
             if (etcd_states[i]["error"].is_null())
             {
                 etcd_alive++;
-                etcd_db_size = etcd_states[i]["dbSize"].uint64_value();
+                uint64_t db_size = etcd_states[i]["dbSize"].uint64_value();
+                etcd_db_size = db_size > etcd_db_size ? db_size : etcd_db_size;
             }
         }
         int mon_count = 0;
@@ -265,7 +267,7 @@ resume_2:
                     obj_n = agg_stats["object_counts"][str].uint64_value() * parent->cli->st_cli.global_block_size;
                 json_status[str+"_data"] = obj_n;
             }
-            printf("%s\n", json11::Json(json_status).dump().c_str());
+            result.data = json11::Json(json_status);
             state = 100;
             return;
         }
@@ -406,7 +408,7 @@ std::function<bool(cli_result_t &)> cli_tool_t::start_status(json11::Json cfg)
         printer->loop();
         if (printer->is_done())
         {
-            result = { .err = 0 };
+            result = printer->result;
             delete printer;
             return true;
         }

src/cmd/openapi.json

@@ -0,0 +1,937 @@
+{
+    "openapi": "3.0.0",
+    "info": {
+        "description": "Vitastor control API",
+        "version": "3.0.2",
+        "license": {
+            "name": "VNPL 1.1"
+        }
+    },
+    "paths": {
+        "/status": { "get": {
+            "summary": "Show server status",
+            "operationId": "status",
+            "produces": [ "application/json" ],
+            "parameters": [],
+            "responses": {
+                "200": { "content": { "application/json": { "schema": {
+                    "type": "object",
+                    "properties": {
+                        "pool_count": { "type": "integer", "format": "uint64", "description": "Total number of pools" },
+                        "active_pool_count": { "type": "integer", "format": "uint64", "description": "Number of active pools" },
+                        "backfillfull_pools": { "type": "array", "items": { "type": "integer", "format": "uint64" }, "description": "ID(s) of backfillfull pools" },
+                        "clean_data": { "type": "integer", "format": "uint64", "description": "Clean data - user size in bytes" },
+                        "misplaced_data": { "type": "integer", "format": "uint64", "description": "Misplaced data - user size in bytes" },
+                        "degraded_data": { "type": "integer", "format": "uint64", "description": "Degraded data - user size in bytes" },
+                        "incomplete_data": { "type": "integer", "format": "uint64", "description": "Incomplete data - user size in bytes" },
+                        "osd_count": { "type": "integer", "format": "uint64", "description": "Total number of OSDs" },
+                        "osd_up": { "type": "integer", "format": "uint64", "description": "Number of active OSDs" },
+                        "osds_full": { "type": "integer", "format": "uint64", "description": "Number of full OSDs" },
+                        "osds_nearfull": { "type": "integer", "format": "uint64", "description": "Number of nearfull OSDs" },
+                        "osds_primary_slow_ops": { "type": "array", "items": { "type": "integer", "format": "uint64" }, "description": "Numbers of OSDs with slow/hung client operations" },
+                        "osds_secondary_slow_ops": { "type": "array", "items": { "type": "integer", "format": "uint64" }, "description": "Numbers of OSDs with slow/hung storage operations" },
+                        "total_raw": { "type": "integer", "format": "uint64", "description": "Total size of all OSDs in bytes" },
+                        "free_raw": { "type": "integer", "format": "uint64", "description": "Total free space on OSDs in bytes" },
+                        "down_raw": { "type": "integer", "format": "uint64", "description": "Total size of stopped OSDs in bytes" },
+                        "free_down_raw": { "type": "integer", "format": "uint64", "description": "Total free space on stopped OSDs in bytes" },
+                        "etcd_count": { "type": "integer", "format": "uint64", "description": "Total number of etcd endpoints in the config" },
+                        "etcd_alive": { "type": "integer", "format": "uint64", "description": "Number of healthy etcd endpoints" },
+                        "etcd_db_size": { "type": "integer", "format": "uint64", "description": "Estimated maximum size of etcd databases (dbSize)" },
+                        "mon_count": { "type": "integer", "format": "uint64", "description": "Number of active monitors" },
+                        "mon_master": { "type": "string", "description": "Name of the host with active monitor" },
+                        "readonly": { "type": "boolean", "description": "readonly flag" },
+                        "no_rebalance": { "type": "boolean", "description": "no_rebalance flag" },
+                        "no_recovery": { "type": "boolean", "description": "no_recovery flag" },
+                        "no_scrub": { "type": "boolean", "description": "no_scrub flag" },
+                        "object_counts": { "type": "object", "properties": {
+                            "object": { "type": "integer", "format": "uint64", "description": "Total number of objects" },
+                            "clean": { "type": "integer", "format": "uint64", "description": "Number of clean objects" },
+                            "degraded": { "type": "integer", "format": "uint64", "description": "Number of degraded objects" },
+                            "incomplete": { "type": "integer", "format": "uint64", "description": "Number of incomplete objects" },
+                            "misplaced": { "type": "integer", "format": "uint64", "description": "Number of misplaced objects" }
+                        } },
+                        "pg_states": { "type": "object", "additionalProperties": { "type": "string" }, "description": "Number of PGs by state" },
+                        "op_stats": { "$ref": "#/components/schemas/OpStats" },
+                        "recovery_stats": { "$ref": "#/components/schemas/RecoveryStats" }
+                    }
+                } } } }
+            }
+        } },
+        "/image/list": { "get": {
+            "summary": "List images",
+            "operationId": "imageListGet",
+            "produces": [ "application/json" ],
+            "parameters": [ {
+                "name": "params",
+                "in": "query",
+                "schema": { "$ref": "#/components/schemas/ImageListParams" },
+                "style": "form",
+                "explode": true
+            } ],
+            "responses": {
+                "200": { "content": { "application/json": { "schema": { "$ref": "#/components/schemas/ImageList" } } } }
+            }
+        }, "post": {
+            "summary": "List images",
+            "operationId": "imageListPost",
+            "consumes": [ "application/json" ],
+            "produces": [ "application/json" ],
+            "requestBody": { "content": { "application/json": { "schema": { "$ref": "#/components/schemas/ImageListParams" } } } },
+            "responses": {
+                "200": { "content": { "application/json": { "schema": { "$ref": "#/components/schemas/ImageList" } } } }
+            }
+        } },
+        "/image/create": { "post": {
+            "summary": "Create an image, snapshot, or a clone",
+            "operationId": "imageCreate",
+            "consumes": [ "application/json" ],
+            "produces": [ "application/json" ],
+            "requestBody": { "content": { "application/json": { "schema": {
+                "type": "object",
+                "required": [ "image" ],
+                "properties": {
+                    "image": { "type": "string", "description": "Image name" },
+                    "snapshot": { "type": "string", "description": "Snapshot name" },
+                    "size": {
+                        "oneOf": [
+                            { "type": "integer", "format": "uint64" },
+                            { "type": "string", "pattern": "^\\d+[KMGT]$" }
+                        ],
+                        "description": "Image size in bytes or with K/M/G/T suffix (not for snapshots)"
+                    },
+                    "pool_id": { "type": "integer", "format": "uint64", "description": "Pool ID for the new image/snapshot" },
+                    "pool_name": { "type": "string", "description": "Pool name for the new image/snapshot" },
+                    "parent": { "type": "string", "description": "Create a clone with this parent image name" }
+                }
+            } } } },
+            "responses": {
+                "200": { "content": { "application/json": { "schema": { "$ref": "#/components/schemas/Image" } } } },
+                "400": { "$ref": "#/components/responses/Invalid" },
+                "409": { "$ref": "#/components/responses/UpdateConflict" },
+                "412": { "$ref": "#/components/responses/PreconditionFailed" }
+            }
+        } },
+        "/image/modify": { "post": {
+            "summary": "Modify an image",
+            "operationId": "imageModify",
+            "consumes": [ "application/json" ],
+            "produces": [ "application/json" ],
+            "requestBody": { "content": { "application/json": { "schema": {
+                "type": "object",
+                "required": [ "image" ],
+                "properties": {
+                    "image": { "type": "string", "description": "Old full name of the image" },
+                    "rename": { "type": "string", "description": "New full image name" },
+                    "resize": {
+                        "oneOf": [
+                            { "type": "integer", "format": "uint64" },
+                            { "type": "string", "pattern": "^\\d+[KMGT]$" }
+                        ],
+                        "description": "New image size in bytes or with K/M/G/T suffix"
+                    },
+                    "readonly": { "type": "boolean", "description": "Make the image read-only" },
+                    "readwrite": { "type": "boolean", "description": "Make the image read-write" },
+                    "deleted": { "type": "boolean", "description": "Set or clear the 'deleted' flag" },
+                    "force": { "type": "boolean", "description": "Proceed with shrinking or setting readwrite flag even if the image has children" },
+                    "down_ok": { "type": "boolean", "description": "Proceed with shrinking even if some data will be left on unavailable OSDs" }
+                }
+            } } } },
+            "responses": {
+                "200": { "content": { "application/json": { "schema": { "$ref": "#/components/schemas/Image" } } } },
+                "400": { "$ref": "#/components/responses/Invalid" },
+                "409": { "$ref": "#/components/responses/UpdateConflict" },
+                "412": { "$ref": "#/components/responses/PreconditionFailed" }
+            }
+        } },
+        "/image/delete": { "post": {
+            "summary": "Delete image(s)",
+            "operationId": "imageDelete",
+            "consumes": [ "application/json" ],
+            "produces": [ "application/json" ],
+            "requestBody": { "content": { "application/json": { "schema": {
+                "oneOf": [ {
+                    "type": "object",
+                    "required": [ "names" ],
+                    "properties": {
+                        "names": { "type": "array", "items": { "type": "string" }, "description": "Exact image name list to delete" },
+                        "writers_stopped": { "type": "boolean", "description": "Allow 'inverse' snapshot deletion optimisation" },
+                        "down_ok": { "type": "boolean", "description": "Continue deletion/merging even if some data will be left on unavailable OSDs" }
+                    }
+                }, {
+                    "type": "object",
+                    "required": [ "matching" ],
+                    "properties": {
+                        "matching": { "type": "array", "items": { "type": "string" }, "description": "Glob image name patterns to delete" },
+                        "writers_stopped": { "type": "boolean", "description": "Allow 'inverse' snapshot deletion optimisation" },
+                        "down_ok": { "type": "boolean", "description": "Continue deletion/merging even if some data will be left on unavailable OSDs" }
+                    }
+                }, {
+                    "type": "object",
+                    "required": [ "from" ],
+                    "properties": {
+                        "from": { "type": "string", "description": "Parent image in a parent-child sequence to delete" },
+                        "to": { "type": "string", "description": "Child image in a parent-child sequence to delete" },
+                        "writers_stopped": { "type": "boolean", "description": "Allow 'inverse' snapshot deletion optimisation" },
+                        "down_ok": { "type": "boolean", "description": "Continue deletion/merging even if some data will be left on unavailable OSDs" }
+                    }
+                } ]
+            } } } },
+            "responses": {
+                "200": { "content": { "application/json": { "schema": {
+                    "type": "object",
+                    "properties": {
+                        "deleted_ids": { "type": "array", "items": { "type": "integer", "format": "uint64" }, "description": "Full inode ID(s) of deleted images" },
+                        "deleted_images": { "type": "array", "items": { "type": "string" }, "description": "Deleted image name(s)" },
+                        "rebased_images": { "type": "array", "items": { "type": "string" }, "description": "Rebased child image name(s)" },
+                        "renamed_from": { "type": "string", "description": "Name of the renamed snapshot when 'inverse' deletion is enabled" },
+                        "renamed_to": { "type": "string", "description": "Name of the rename destination when 'inverse' deletion is enabled" }
+                    }
+                } } } },
+                "400": { "$ref": "#/components/responses/Invalid" },
+                "409": { "$ref": "#/components/responses/UpdateConflict" },
+                "412": { "$ref": "#/components/responses/PreconditionFailed" }
+            }
+        } },
+        "/image/flatten": { "post": {
+            "summary": "Flatten an image",
+            "operationId": "imageFlatten",
+            "consumes": [ "application/json" ],
+            "produces": [ "application/json" ],
+            "requestBody": { "content": { "application/json": { "schema": {
+                "type": "object",
+                "required": [ "names" ],
+                "properties": {
+                    "image": { "type": "string", "description": "Image name to flatten" }
+                }
+            } } } },
+            "responses": {
+                "200": { "content": { "text/plain": { "schema": { "type": "string", "description": "Empty response" } } } },
+                "400": { "$ref": "#/components/responses/Invalid" },
+                "409": { "$ref": "#/components/responses/UpdateConflict" },
+                "412": { "$ref": "#/components/responses/PreconditionFailed" }
+            }
+        } },
+        "/osd/list": { "get": {
+            "summary": "List OSDs",
+            "operationId": "osdListGet",
+            "produces": [ "application/json" ],
+            "parameters": [ {
+                "name": "params",
+                "in": "query",
+                "schema": { "$ref": "#/components/schemas/OsdListParams" },
+                "style": "form",
+                "explode": true
+            } ],
+            "responses": {
+                "200": { "content": { "application/json": { "schema": { "$ref": "#/components/schemas/OsdList" } } } }
+            }
+        }, "post": {
+            "summary": "List OSDs",
+            "operationId": "osdListPost",
+            "consumes": [ "application/json" ],
+            "produces": [ "application/json" ],
+            "requestBody": { "content": { "application/json": { "schema": { "$ref": "#/components/schemas/OsdListParams" } } } },
+            "responses": {
+                "200": { "content": { "application/json": { "schema": { "$ref": "#/components/schemas/OsdList" } } } }
+            }
+        } },
+        "/osd/alloc": { "post": {
+            "summary": "Allocate a new OSD number",
+            "operationId": "osdAlloc",
+            "produces": [ "application/json" ],
+            "responses": {
+                "200": { "content": { "application/json": { "schema": {
+                    "type": "integer",
+                    "format": "uint64",
+                    "description": "New OSD number"
+                } } } },
+                "400": { "$ref": "#/components/responses/Invalid" },
+                "412": { "$ref": "#/components/responses/PreconditionFailed" }
+            }
+        } },
+        "/osd/delete": { "post": {
+            "summary": "Delete stopped OSD(s) from etcd",
+            "operationId": "osdDelete",
+            "consumes": [ "application/json" ],
+            "produces": [ "application/json" ],
+            "requestBody": { "content": { "application/json": { "schema": {
+                "type": "object",
+                "properties": {
+                    "force": { "type": "boolean", "description": "Allow to remove non-empty OSDs" },
+                    "allow_data_loss": { "type": "boolean", "description": "Allow to remove non-empty OSDs even if it will lead to data loss" },
+                    "dry_run": { "type": "boolean", "description": "Only check if the removal is possible but do not delete" }
+                }
+            } } } },
+            "responses": {
+                "200": { "content": { "application/json": { "schema": {
+                    "type": "object",
+                    "properties": {
+                        "deleted_osds": {
+                            "type": "array",
+                            "items": { "type": "integer", "format": "uint64" },
+                            "description": "Deleted OSD numbers"
+                        }
+                    },
+                    "format": "uint64",
+                    "description": "New OSD number"
+                } } } },
+                "400": { "$ref": "#/components/responses/Invalid" },
+                "409": { "$ref": "#/components/responses/UpdateConflict" },
+                "412": { "content": { "application/json": { "schema": {
+                    "type": "object",
+                    "properties": {
+                        "pool_effects": {
+                            "type": "array",
+                            "items": {
+                                "type": "object",
+                                "properties": {
+                                    "pool_id": { "type": "integer", "format": "uint64", "description": "Affected pool ID" },
+                                    "pool_name": { "type": "string", "description": "Affected pool name" },
+                                    "effect": {
+                                        "type": "string",
+                                        "enum": [ "incomplete", "has_incomplete", "offline", "degraded" ],
+                                        "description": "Possible pool PG state after removing OSD(s)"
+                                    }
+                                }
+                            }
+                        }
+                    }
+                } } } }
+            }
+        } },
+        "/osd/modify": { "post": {
+            "summary": "Modify OSD reweight, tags or noout flag",
+            "operationId": "osdDelete",
+            "consumes": [ "application/json" ],
+            "produces": [ "application/json" ],
+            "requestBody": { "content": { "application/json": { "schema": {
+                "type": "object",
+                "properties": {
+                    "osd_num": { "type": "integer", "format": "uint64", "description": "OSD number to modify" },
+                    "reweight": { "type": "number", "minimum": 0, "maximum": 1, "description": "New OSD weight" },
+                    "tags": { "type": "array", "items": { "type": "string" }, "description": "New OSD tags" },
+                    "noout": { "type": "boolean", "description": "New value of the noout flag" }
+                }
+            } } } },
+            "responses": {
+                "200": { "content": { "application/json": { "schema": {
+                    "type": "object",
+                    "properties": {
+                        "reweight": { "type": "number", "minimum": 0, "maximum": 1, "description": "OSD weight" },
+                        "tags": { "type": "array", "items": { "type": "string" }, "description": "OSD tags" },
+                        "noout": { "type": "boolean", "description": "noout flag" }
+                    },
+                    "format": "uint64",
+                    "description": "New OSD number"
+                } } } },
+                "400": { "$ref": "#/components/responses/Invalid" },
+                "409": { "$ref": "#/components/responses/UpdateConflict" },
+                "412": { "$ref": "#/components/responses/PreconditionFailed" }
+            }
+        } },
+        "/pool/list": { "get": {
+            "summary": "List pools",
+            "operationId": "poolListGet",
+            "produces": [ "application/json" ],
+            "parameters": [ {
+                "name": "params",
+                "in": "query",
+                "schema": { "$ref": "#/components/schemas/PoolListParams" },
+                "style": "form",
+                "explode": true
+            } ],
+            "responses": {
+                "200": { "content": { "application/json": { "schema": { "$ref": "#/components/schemas/PoolList" } } } }
+            }
+        }, "post": {
+            "summary": "List pools",
+            "operationId": "poolListPost",
+            "consumes": [ "application/json" ],
+            "produces": [ "application/json" ],
+            "requestBody": { "content": { "application/json": { "schema": { "$ref": "#/components/schemas/PoolListParams" } } } },
+            "responses": {
+                "200": { "content": { "application/json": { "schema": { "$ref": "#/components/schemas/PoolList" } } } }
+            }
+        } },
+        "/pool/create": { "post": {
+            "summary": "Create a pool",
+            "operationId": "poolCreate",
+            "consumes": [ "application/json" ],
+            "produces": [ "application/json" ],
+            "requestBody": { "content": { "application/json": { "schema": {
+                "allOf": [ {
+                    "$ref": "#/components/schemas/PoolConfig"
+                }, { "type": "object", "properties": {
+                    "wait": { "type": "boolean", "description": "Wait for the new pool to come online" },
+                    "force": { "type": "boolean", "description": "Do not check that cluster has enough OSDs to create the pool" }
+                } } ]
+            } } } },
+            "responses": {
+                "200": { "content": { "application/json": { "schema": {
+                    "$ref": "#/components/schemas/PoolConfig"
+                } } } },
+                "400": { "$ref": "#/components/responses/Invalid" },
+                "409": { "$ref": "#/components/responses/UpdateConflict" },
+                "412": { "$ref": "#/components/responses/PreconditionFailed" }
+            }
+        } },
+        "/pool/modify": { "post": {
+            "summary": "Modify a pool",
+            "operationId": "poolModify",
+            "consumes": [ "application/json" ],
+            "produces": [ "application/json" ],
+            "requestBody": { "content": { "application/json": { "schema": {
+                "allOf": [ {
+                    "$ref": "#/components/schemas/PoolConfig"
+                }, { "type": "object", "properties": {
+                    "id": { "type": "integer", "format": "uint64", "description": "Pool ID to modify" },
+                    "old_name": { "type": "string", "description": "Pool name to modify" },
+                    "force": { "type": "boolean", "description": "Allow modifications which may lead to data loss" }
+                } } ]
+            } } } },
+            "responses": {
+                "200": { "content": { "application/json": { "schema": {
+                    "$ref": "#/components/schemas/PoolConfig"
+                } } } },
+                "400": { "$ref": "#/components/responses/Invalid" },
+                "409": { "$ref": "#/components/responses/UpdateConflict" },
+                "412": { "$ref": "#/components/responses/PreconditionFailed" }
+            }
+        } },
+        "/pool/delete": { "post": {
+            "summary": "Delete an pool",
+            "operationId": "poolDelete",
+            "consumes": [ "application/json" ],
+            "produces": [ "application/json" ],
+            "requestBody": { "content": { "application/json": { "schema": {
+                "type": "object",
+                "properties": {
+                    "pool_id": { "type": "integer", "format": "uint64", "description": "Pool ID to delete" },
+                    "pool_name": { "type": "string", "description": "Pool name to delete" },
+                    "force": { "type": "boolean", "description": "Allow to delete non-empty pools" }
+                }
+            } } } },
+            "responses": {
+                "200": { "content": { "application/json": { "schema": {
+                    "type": "object",
+                    "properties": { "ok": { "type": "boolean" } }
+                } } } },
+                "400": { "$ref": "#/components/responses/Invalid" },
+                "409": { "$ref": "#/components/responses/UpdateConflict" },
+                "412": { "$ref": "#/components/responses/PreconditionFailed" }
+            }
+        } },
+        "/pg/list": { "get": {
+            "summary": "List PGs",
+            "operationId": "pgListGet",
+            "produces": [ "application/json" ],
+            "parameters": [ {
+                "name": "params",
+                "in": "query",
+                "schema": { "$ref": "#/components/schemas/PgListParams" },
+                "style": "form",
+                "explode": true
+            } ],
+            "responses": {
+                "200": { "content": { "application/json": { "schema": { "$ref": "#/components/schemas/PgList" } } } }
+            }
+        }, "post": {
+            "summary": "List PGs",
+            "operationId": "pgListPost",
+            "consumes": [ "application/json" ],
+            "produces": [ "application/json" ],
+            "requestBody": { "content": { "application/json": { "schema": { "$ref": "#/components/schemas/PgListParams" } } } },
+            "responses": {
+                "200": { "content": { "application/json": { "schema": { "$ref": "#/components/schemas/PgList" } } } }
+            }
+        } },
+        "/data/delete": { "post": {
+            "summary": "Remove inode data without changing metadata",
+            "operationId": "dataDelete",
+            "consumes": [ "application/json" ],
+            "produces": [ "application/json" ],
+            "requestBody": { "content": { "application/json": { "schema": {
+                "type": "object",
+                "properties": {
+                    "pool": { "type": "integer", "format": "uint64", "description": "Pool ID" },
+                    "inode": { "type": "integer", "format": "uint64", "description": "Inode number" },
+                    "min_offset": { "type": "integer", "format": "uint64", "description": "Purge only data starting with specified offset" },
+                    "max_offset": { "type": "integer", "format": "uint64", "description": "Purge only data before specified offset" }
+                }
+            } } } },
+            "responses": {
+                "200": { "content": { "application/json": { "schema": {
+                    "type": "object",
+                    "properties": {
+                        "removed_objects": { "type": "integer", "format": "uint64", "description": "Removed object count" }
+                    }
+                } } } },
+                "400": { "$ref": "#/components/responses/Invalid" },
+                "412": { "content": { "application/json": { "schema": {
+                    "type": "object",
+                    "properties": {
+                        "total_objects": { "type": "integer", "format": "uint64", "description": "Total object count" },
+                        "removed_objects": { "type": "integer", "format": "uint64", "description": "Removed object count" },
+                        "error_count": { "type": "integer", "format": "uint64", "description": "Error count" },
+                        "inactive_osds": {
+                            "type": "array",
+                            "items": { "type": "integer", "format": "uint64" },
+                            "description": "Inactive OSDs possibly with some non-removed data left"
+                        },
+                        "inactive_pgs": {
+                            "type": "array",
+                            "items": { "type": "integer", "format": "uint64" },
+                            "description": "Inactive PGs possibly with some non-removed data left"
+                        }
+                    }
+                } } } }
+            }
+        } },
+        "/data/describe": { "get": {
+            "summary": "Describe unclean object locations in the cluster",
+            "operationId": "dataDescribeGet",
+            "produces": [ "application/json" ],
+            "parameters": [ {
+                "name": "params",
+                "in": "query",
+                "schema": { "$ref": "#/components/schemas/DataDescribeParams" },
+                "style": "form",
+                "explode": true
+            } ],
+            "responses": {
+                "200": { "content": { "application/json": { "schema": { "$ref": "#/components/schemas/DataDescribe" } } } }
+            }
+        }, "post": {
+            "summary": "Describe unclean object locations in the cluster",
+            "operationId": "dataDescribePost",
+            "consumes": [ "application/json" ],
+            "produces": [ "application/json" ],
+            "requestBody": { "content": { "application/json": { "schema": { "$ref": "#/components/schemas/DataDescribeParams" } } } },
+            "responses": {
+                "200": { "content": { "application/json": { "schema": { "$ref": "#/components/schemas/DataDescribe" } } } }
+            }
+        } },
+        "/data/fix": { "post": {
+            "summary": "Fix inconsistent objects in the cluster by deleting some copies",
+            "operationId": "dataDelete",
+            "consumes": [ "application/json" ],
+            "produces": [ "application/json" ],
+            "requestBody": { "content": { "application/json": { "schema": {
+                "type": "object",
+                "properties": {
+                    "objects": {
+                        "type": "array",
+                        "items": {
+                            "type": "object",
+                            "properties": {
+                                "inode": { "type": "integer", "format": "uint64", "description": "Full inode ID with pool" },
+                                "stripe": { "type": "integer", "format": "uint64", "description": "Object offset within inode" }
+                            }
+                        }
+                    },
+                    "bad_osds": {
+                        "type": "array",
+                        "items": { "type": "integer", "format": "uint64" },
+                        "description": "Remove inconsistent copies/parts of objects from these OSDs"
+                    },
+                    "part": { "type": "integer", "minimum": 0, "maximum": 255, "description": "Only remove EC object part with this number" },
+                    "check": { "type": "boolean", "default": true, "description": "Do safety checks before removing parts" }
+                }
+            } } } },
+            "responses": {
+                "200": { "content": { "application/json": { "schema": {
+                    "type": "object",
+                    "properties": {
+                        "inode": { "type": "integer", "format": "uint64", "description": "Full inode ID with pool" },
+                        "stripe": { "type": "integer", "format": "uint64", "description": "Object offset within inode" },
+                        "part": { "type": "integer", "minimum": 0, "maximum": 255, "description": "Object part number" },
+                        "osd_num": { "type": "integer", "format": "uint64", "description": "OSD number with this part" }
+                    }
+                } } } },
+                "400": { "$ref": "#/components/responses/Invalid" }
+            }
+        } },
+        "/data/merge": { "post": {
+            "summary": "Merge layer data without changing metadata",
+            "operationId": "dataMerge",
+            "consumes": [ "application/json" ],
+            "produces": [ "application/json" ],
+            "requestBody": { "content": { "application/json": { "schema": {
+                "type": "object",
+                "properties": {
+                    "from": { "type": "string", "description": "First parent layer to merge" },
+                    "to": { "type": "string", "description": "Last child layer to merge" },
+                    "target": { "type": "string", "description": "Merge target (must be between from and to)" },
+                    "delete_source": { "type": "boolean", "description": "Delete source data after merging" },
+                    "fsync_interval": { "type": "integer", "format": "uint64", "description": "Fsync every this number of writes during merge" },
+                    "cas": { "type": "boolean", "description": "Use CAS writes to allow online merging" }
+                }
+            } } } },
+            "responses": {
+                "200": { "content": { "application/json": { "schema": {
+                    "type": "object",
+                    "properties": {
+                        "from": { "type": "string", "description": "First merged parent layer" },
+                        "to": { "type": "string", "description": "Last merged child layer" },
+                        "target": { "type": "string", "description": "Merge target" }
+                    }
+                } } } },
+                "400": { "$ref": "#/components/responses/Invalid" },
+                "412": { "$ref": "#/components/responses/PreconditionFailed" }
+            }
+        } }
+    },
+    "components": {
+        "responses": {
+            "Invalid": {
+                "description": "Invalid Arguments (EINVAL)",
+                "content": {
+                    "text/plain": {
+                        "schema": { "type": "string", "description": "Error text" }
+                    }
+                }
+            },
+            "PreconditionFailed": {
+                "description": "Precondition Failed (ENOTEMPTY, EEXIST or ENOENT)",
+                "content": {
+                    "text/plain": {
+                        "schema": { "type": "string", "description": "Error text" }
+                    }
+                }
+            },
+            "UpdateConflict": {
+                "description": "Update Conflict",
+                "content": {
+                    "text/plain": {
+                        "schema": { "type": "string", "description": "Error text" }
+                    }
+                }
+            }
+        },
+        "schemas": {
+            "OpStat": {
+                "type": "object",
+                "properties": {
+                    "bytes": { "type": "integer", "format": "uint64", "description": "Total bytes processed" },
+                    "count": { "type": "integer", "format": "uint64", "description": "Number of processed operations" },
+                    "usec": { "type": "integer", "format": "uint64", "description": "Total operation processing time in microseconds" },
+                    "bps": { "type": "integer", "format": "uint64", "description": "Bytes per second (average since previous report)" },
+                    "iops": { "type": "integer", "format": "uint64", "description": "Operations per second (average since previous report)" },
+                    "lat": { "type": "integer", "format": "uint64", "description": "Operation latency (average since previous report)" }
+                }
+            },
+            "SubOpStat": {
+                "type": "object",
+                "properties": {
+                    "count": { "type": "integer", "format": "uint64", "description": "Number of processed operations" },
+                    "usec": { "type": "integer", "format": "uint64", "description": "Total operation processing time in microseconds" },
+                    "iops": { "type": "integer", "format": "uint64", "description": "Operations per second (average since previous report)" },
+                    "lat": { "type": "integer", "format": "uint64", "description": "Operation latency (average since previous report)" }
+                }
+            },
+            "RecoveryStat": {
+                "type": "object",
+                "properties": {
+                    "bytes": { "type": "integer", "format": "uint64", "description": "Total bytes processed" },
+                    "count": { "type": "integer", "format": "uint64", "description": "Number of processed operations" },
+                    "bps": { "type": "integer", "format": "uint64", "description": "Bytes per second (average since previous report)" },
+                    "iops": { "type": "integer", "format": "uint64", "description": "Operations per second (average since previous report)" }
+                }
+            },
+            "RecoveryStats": {
+                "description": "Recovery statistics",
+                "type": "object",
+                "properties": {
+                    "misplaced": { "$ref": "#/components/schemas/RecoveryStat" },
+                    "degraded": { "$ref": "#/components/schemas/RecoveryStat" }
+                }
+            },
+            "OpStats": {
+                "description": "Operation statistics (processing time without client RTT)",
+                "type": "object",
+                "properties": {
+                    "read": { "$ref": "#/components/schemas/OpStat" },
+                    "write": { "$ref": "#/components/schemas/OpStat" },
+                    "write_stable": { "$ref": "#/components/schemas/OpStat" },
+                    "sync": { "$ref": "#/components/schemas/OpStat" },
+                    "stabilize": { "$ref": "#/components/schemas/OpStat" },
+                    "rollback": { "$ref": "#/components/schemas/OpStat" },
+                    "delete": { "$ref": "#/components/schemas/OpStat" },
+                    "list": { "$ref": "#/components/schemas/OpStat" },
+                    "show_config": { "$ref": "#/components/schemas/OpStat" },
+                    "primary_read": { "$ref": "#/components/schemas/OpStat" },
+                    "primary_write": { "$ref": "#/components/schemas/OpStat" },
+                    "primary_sync": { "$ref": "#/components/schemas/OpStat" },
+                    "primary_delete": { "$ref": "#/components/schemas/OpStat" },
+                    "ping": { "$ref": "#/components/schemas/OpStat" },
+                    "sec_read_bmp": { "$ref": "#/components/schemas/OpStat" },
+                    "scrub": { "$ref": "#/components/schemas/OpStat" },
+                    "describe": { "$ref": "#/components/schemas/OpStat" },
+                    "sec_lock": { "$ref": "#/components/schemas/OpStat" }
+                }
+            },
+            "SubOpStats": {
+                "description": "Suboperation statistics (processing+RTT to secondary as primary sees it)",
+                "type": "object",
+                "properties": {
+                    "read": { "$ref": "#/components/schemas/SubOpStat" },
+                    "write": { "$ref": "#/components/schemas/SubOpStat" },
+                    "write_stable": { "$ref": "#/components/schemas/SubOpStat" },
+                    "sync": { "$ref": "#/components/schemas/SubOpStat" },
+                    "stabilize": { "$ref": "#/components/schemas/SubOpStat" },
+                    "rollback": { "$ref": "#/components/schemas/SubOpStat" },
+                    "delete": { "$ref": "#/components/schemas/SubOpStat" },
+                    "list": { "$ref": "#/components/schemas/SubOpStat" },
+                    "show_config": { "$ref": "#/components/schemas/SubOpStat" },
+                    "ping": { "$ref": "#/components/schemas/SubOpStat" },
+                    "sec_read_bmp": { "$ref": "#/components/schemas/SubOpStat" },
+                    "sec_lock": { "$ref": "#/components/schemas/SubOpStat" }
+                }
+            },
+            "ImageListParams": {
+                "type": "object",
+                "properties": {
+                    "globs": { "type": "array", "items": { "type": "string" }, "description": "Glob patterns or exact name matches" },
+                    "exact": { "type": "boolean", "description": "Treat globs as exact name matches, not as patterns" },
+                    "pool_id": { "type": "integer", "format": "uint64", "description": "Pool ID to list" },
+                    "pool_name": { "type": "string", "description": "Pool name to list" },
+                    "long": { "type": "boolean", "description": "Include allocated size and I/O statistics" },
+                    "sort": {
+                        "type": "string",
+                        "description": "Sort by specified field",
+                        "enum": [
+                            "name", "size", "used_size",
+                            "read_iops", "read_bps", "read_lat", "read_queue",
+                            "write_iops", "write_bps", "write_lat", "write_queue",
+                            "delete_iops", "delete_bps", "delete_lat", "delete_queue"
+                        ]
+                    },
+                    "reverse": { "type": "boolean", "description": "Sort in descending order" },
+                    "count": { "type": "integer", "format": "uint64", "description": "Only list first N items" }
+                }
+            },
+            "Image": {
+                "type": "object",
+                "properties": {
+                    "inode_id": { "type": "integer", "format": "uint64", "description": "Full inode ID with pool" },
+                    "inode_num": { "type": "integer", "format": "uint64", "description": "Inode number without pool" },
+                    "name": { "type": "string", "description": "Image name" },
+                    "pool_id": { "type": "integer", "format": "uint64", "description": "Pool ID" },
+                    "pool_name": { "type": "string", "description": "Pool name" },
+                    "parent_name": { "type": "string", "description": "Name of the parent image (layer)" },
+                    "parent_inode_id": { "type": "integer", "format": "uint64", "description": "Full ID of the parent inode" },
+                    "parent_inode_num": { "type": "integer", "format": "uint64", "description": "Parent inode number without pool" },
+                    "parent_pool_id": { "type": "integer", "format": "uint64", "description": "Pool ID of the parent inode" },
+                    "size": { "type": "integer", "format": "uint64", "description": "Image size in bytes" },
+                    "used_size": { "type": "integer", "format": "uint64", "description": "Used space in bytes" },
+                    "readonly": { "type": "boolean", "description": "Readonly flag" },
+                    "deleted": { "type": "boolean", "description": "Deleted flag" }
+                }
+            },
+            "ImageList": {
+                "type": "array",
+                "items": { "$ref": "#/components/schemas/Image" }
+            },
+            "OsdListParams": {
+                "type": "object",
+                "properties": {
+                    "as_tree": { "type": "boolean", "description": "Include all placement tree nodes, not just OSDs" },
+                    "long": { "type": "boolean", "description": "Include per-OSD I/O statistics" }
+                }
+            },
+            "OsdList": {
+                "type": "array",
+                "items": {
+                    "type": "object",
+                    "properties": {
+                        "name": { "type": "string", "description": "OSD number" },
+                        "type": { "type": "string", "description": "Placement tree node type (osd for OSDs)" },
+                        "up": { "type": "boolean", "description": "Whether the OSD is active" },
+                        "parent": { "type": "string", "description": "Parent placement tree node name (host name, etc)" },
+                        "size": { "type": "integer", "format": "uint64", "description": "OSD size in bytes" },
+                        "free": { "type": "integer", "format": "uint64", "description": "Free space on the OSD in bytes" },
+                        "reweight": { "type": "number", "minimum": 0, "maximum": 1, "description": "OSD weight in data distribution" },
+                        "tags": { "type": "array", "items": { "type": "string" }, "description": "OSD tags" },
+                        "data_block_size": { "type": "integer", "format": "uint64" },
+                        "bitmap_granularity": { "type": "integer", "format": "uint64" },
+                        "immediate_commit": { "type": "string" },
+                        "noout": { "type": "boolean", "description": "noout flag" },
+                        "slow_ops_primary": { "type": "integer", "format": "uint64", "description": "Number of slow/hung client operations on this OSD" },
+                        "slow_ops_secondary": { "type": "integer", "format": "uint64", "description": "Number of slow/hung storage operations on this OSD" },
+                        "op_stats": { "$ref": "#/components/schemas/OpStats" },
+                        "subop_stats": { "$ref": "#/components/schemas/SubOpStats" },
+                        "recovery_stats": { "$ref": "#/components/schemas/RecoveryStats" }
+                    }
+                }
+            },
+            "PoolListParams": {
+                "type": "object",
+                "properties": {
+                    "globs": { "type": "array", "items": { "type": "string" }, "description": "Glob patterns to restrict listing" },
+                    "long": { "type": "boolean", "description": "Include per-pool I/O statistics" },
+                    "sort": { "type": "string", "description": "Sort by specified field" },
+                    "reverse": { "type": "boolean", "description": "Sort in descending order" },
+                    "count": { "type": "integer", "format": "uint64", "description": "Only list first N items" }
+                }
+            },
+            "PoolConfig": {
+                "type": "object",
+                "properties": {
+                    "name": { "type": "string", "description": "Pool name" },
+                    "used_for_app": { "type": "string", "description": "Application name pool is used for" },
+                    "scheme": { "type": "string", "enum": [ "replicated", "ec", "xor" ], "description": "Redundancy scheme" },
+                    "pg_count": { "type": "integer", "format": "uint64", "description": "Configured PG count" },
+                    "pg_size": { "type": "integer", "minimum": 1, "maximum": 255, "description": "Number of replicas or total data and parity EC chunks" },
+                    "pg_minsize": { "type": "integer", "minimum": 1, "maximum": 255, "description": "Minimum active OSD count in a PG to allow activity" },
+                    "parity_chunks": { "type": "integer", "minimum": 1, "maximum": 255, "description": "Parity chunk count for EC" },
+                    "local_reads": { "type": "string", "enum": [ "primary", "nearest", "random" ], "description": "Local read policy" },
+                    "root_node": { "type": "string", "description": "Root placement tree node for pool OSDs" },
+                    "pg_stripe_size": { "type": "integer", "format": "uint64", "description": "Stripe size to map to one PG in bytes" },
+                    "failure_domain": { "type": "string", "description": "Failure domain" },
+                    "osd_tags": { "type": "array", "items": { "type": "string" }, "description": "Put pool only on OSDs tagged with all specified tags" },
+                    "primary_affinity_tags": { "type": "array", "items": { "type": "string" }, "description": "Tags to prefer for primary OSDs" },
+                    "block_size": { "type": "integer", "format": "uint64", "description": "Put pool only on OSDs with this data block size" },
+                    "bitmap_granularity": { "type": "integer", "format": "uint64", "description": "Put pool only on OSDs with this logical sector size" },
+                    "immediate_commit": { "type": "string", "enum": [ "none", "small", "all" ], "description": "Put pool only on OSDs with this or larger immediate_commit" },
+                    "scrub_interval": { "type": "string", "pattern": "^\\d+[smhdMy]$", "description": "Automatic scrub interval" },
+                    "level_placement": { "type": "string", "description": "Additional failure domain rules" },
+                    "raw_placement": { "type": "string", "description": "Raw PG generation rules" },
+                    "max_osd_combinations": { "type": "integer", "format": "uint64", "description": "Maximum number of random combinations during PG generation" }
+                }
+            },
+            "PoolList": {
+                "type": "array",
+                "items": {
+                    "type": "object",
+                    "allOf": [ {
+                        "$ref": "#/components/schemas/PoolConfig"
+                    }, { "type": "object", "properties": {
+                        "id": { "type": "integer", "format": "uint64", "description": "Pool ID" },
+                        "status": {
+                            "type": "string",
+                            "enum": [ "active", "has_misplaced", "has_degraded", "degraded", "has_incomplete", "incomplete", "inactive" ],
+                            "description": "Pool status (worst status of pool PGs)"
+                        },
+                        "real_pg_count": { "type": "integer", "format": "uint64", "description": "Currently active PG count" },
+                        "total_raw": { "type": "integer", "format": "uint64", "description": "Total raw space on all pool OSDs" },
+                        "used_raw": { "type": "integer", "format": "uint64", "description": "Raw OSD space used by the pool data" },
+                        "max_available": { "type": "integer", "format": "uint64", "description": "Maximum available user space in the pool" },
+                        "raw_to_usable": { "type": "number", "minimum": 1, "description": "Raw to user space ratio" },
+                        "space_efficiency": { "type": "number", "minimum": 0, "maximum": 1, "description": "Space efficiency (1 - wasted due to unbalanced PGs)" },
+                        "pg_real_size": { "type": "integer", "format": "uint64", "description": "Minimum of PG size and available failure domain count" },
+                        "osd_count": { "type": "integer", "format": "uint64", "description": "Number of OSDs used for the pool" },
+                        "backfillfull": { "type": "boolean", "description": "Backfillfull (is rebalance stopped due to low space)?" },
+                        "misplaced_count": { "type": "integer", "format": "uint64", "description": "Misplaced object count" },
+                        "degraded_count": { "type": "integer", "format": "uint64", "description": "Degraded object count" },
+                        "incomplete_count": { "type": "integer", "format": "uint64", "description": "Incomplete object count" },
+                        "object_count": { "type": "integer", "format": "uint64", "description": "Total object count" },
+                        "read_bps": { "type": "integer", "format": "uint64", "description": "Read bytes per second" },
+                        "read_iops": { "type": "integer", "format": "uint64", "description": "Read operations per second" },
+                        "read_lat": { "type": "integer", "format": "uint64", "description": "Average read latency" },
+                        "write_bps": { "type": "integer", "format": "uint64", "description": "Write bytes per second" },
+                        "write_iops": { "type": "integer", "format": "uint64", "description": "Write operations per second" },
+                        "write_lat": { "type": "integer", "format": "uint64", "description": "Average write latency" },
+                        "delete_bps": { "type": "integer", "format": "uint64", "description": "Delete bytes per second" },
+                        "delete_iops": { "type": "integer", "format": "uint64", "description": "Delete operations per second" },
+                        "delete_lat": { "type": "integer", "format": "uint64", "description": "Average delete latency" }
+                    } } ]
+                }
+            },
+            "PgListParams": {
+                "type": "object",
+                "properties": {
+                    "pool_id": { "type": "integer", "format": "uint64", "description": "Pool ID" },
+                    "pool_name": { "type": "string", "description": "Pool name" },
+                    "pg_state": {
+                        "type": "array",
+                        "items": { "type": "string" },
+                        "description": "PG state combinations (OR) separated by + (AND) and possibly with ! for negation"
+                    },
+                    "min": { "type": "integer", "format": "uint64", "description": "First PG number to show" },
+                    "max": { "type": "integer", "format": "uint64", "description": "Last PG number to show" },
+                    "osd": {
+                        "type": "array",
+                        "items": { "type": "integer", "format": "uint64" },
+                        "description": "Only return PGs with data on one of these OSDs"
+                    }
+                }
+            },
+            "PgList": {
+                "type": "array",
+                "items": {
+                    "type": "object",
+                    "properties": {
+                        "pool_id": { "type": "integer", "format": "uint64", "description": "Pool ID" },
+                        "pool_name": { "type": "string", "description": "Pool name" },
+                        "pg_num": { "type": "integer", "format": "uint64", "description": "PG number" },
+                        "target_set": {
+                            "type": "array",
+                            "items": { "type": "integer", "format": "uint64" },
+                            "description": "Target OSD set for this PG"
+                        },
+                        "target_primary": { "type": "integer", "format": "uint64", "description": "Target primary OSD of this PG (0 = no primary)" },
+                        "epoch": { "type": "integer", "format": "uint64", "description": "PG epoch number" },
+                        "next_scrub": { "type": "integer", "format": "uint64", "description": "UNIX timestamp of the next planned scrub" },
+                        "target_history": {
+                            "type": "array",
+                            "items": { "type": "array", "items": { "type": "integer", "format": "uint64" } },
+                            "description": "Other OSD sets of this PG which possibly still have some data left"
+                        },
+                        "pause": { "type": "boolean", "description": "Pause PG control flag from monitor" },
+                        "cur_primary": { "type": "integer", "format": "uint64", "description": "Current primary OSD of this PG (0 = no primary)" },
+                        "state": {
+                            "type": "array",
+                            "items": { "type": "string", "enum": [
+                                "starting", "peering", "incomplete", "active", "repeering", "stopping", "offline",
+                                "degraded", "has_inconsistent", "has_corrupted", "has_incomplete", "has_degraded",
+                                "has_misplaced", "has_unclean", "has_invalid", "left_on_dead", "scrubbing"
+                            ] },
+                            "description": "PG state"
+                        },
+                        "all_peers": {
+                            "type": "array",
+                            "items": { "type": "array", "items": { "type": "integer", "format": "uint64" } },
+                            "description": "Other OSDs of this PG which possibly still have some data left"
+                        }
+                    }
+                }
+            },
+            "DataDescribeParams": {
+                "type": "object",
+                "properties": {
+                    "pool_id": { "type": "integer", "format": "uint64", "description": "Only list objects from the given pool (by ID)" },
+                    "pool_name": { "type": "string", "description": "Only list objects from the given pool (by name)" },
+                    "pg": { "type": "integer", "format": "uint64", "description": "Only list objects in the given PG of the pool" },
+                    "inode": { "type": "integer", "format": "uint64", "description": "Single inode number to list" },
+                    "min_inode": { "type": "integer", "format": "uint64", "description": "Minimum inode number to list" },
+                    "max_inode": { "type": "integer", "format": "uint64", "description": "Maximum inode number to list" },
+                    "object_state": {
+                        "type": "array",
+                        "items": { "type": "string", "enum": [ "degraded", "misplaced", "incomplete", "corrupted", "inconsistent" ] },
+                        "description": "Only list objects in given state(s)"
+                    },
+                    "osd": {
+                        "type": "array",
+                        "items": { "type": "integer", "format": "uint64" },
+                        "description": "Only list objects from these primary OSD(s)"
+                    }
+                }
+            },
+            "DataDescribe": {
+                "type": "array",
+                "items": {
+                    "type": "object",
+                    "properties": {
+                        "inode": { "type": "integer", "format": "uint64", "description": "Full inode ID with pool" },
+                        "stripe": { "type": "integer", "format": "uint64", "description": "Object offset within inode" },
+                        "part": { "type": "integer", "minimum": 0, "maximum": 255, "description": "Object part number" },
+                        "osd_num": { "type": "integer", "format": "uint64", "description": "OSD number with this part" },
+                        "corrupted": { "type": "boolean", "description": "Part is corrupted flag" },
+                        "inconsistent": { "type": "boolean", "description": "Inconsistency flag (parts do not match)" },
+                        "outdated": { "type": "boolean", "description": "Part on this OSD is outdated" }
+                    }
+                }
+            }
+        }
+    }
+}

src/disk_tool/CMakeLists.txt

@@ -7,7 +7,7 @@ add_executable(vitastor-disk
 	disk_tool.cpp disk_simple_offsets.cpp
 	disk_tool_discard.cpp disk_tool_journal.cpp disk_tool_meta.cpp disk_tool_prepare.cpp disk_tool_resize.cpp
 	disk_tool_resize_auto.cpp disk_tool_udev.cpp disk_tool_utils.cpp disk_tool_upgrade.cpp
-	../util/crc32c.c ../util/str_util.cpp ../util/json_util.cpp ../../json11/json11.cpp ../util/rw_blocking.cpp ../util/allocator.cpp ../util/ringloop.cpp
+	../util/crc32c.c ../util/xxhash.c ../util/str_util.cpp ../util/json_util.cpp ../../json11/json11.cpp ../util/rw_blocking.cpp ../util/allocator.cpp ../util/ringloop.cpp
 	../blockstore/blockstore_disk.cpp ../blockstore/blockstore_heap.cpp ../blockstore/multilist.cpp
 )
 target_link_libraries(vitastor-disk

src/disk_tool/disk_tool.h

@@ -136,8 +136,8 @@ struct disk_tool_t
     void choose_journal_block(uint32_t je_size);
     int resize_rebuild_journal();
     int resize_write_new_journal();
-    void remap_big_write(heap_entry_t *wr);
-    void remap_small_write(heap_entry_t *wr);
+    void remap_big_write(blockstore_heap_t *heap, heap_entry_t *wr);
+    void remap_small_write(blockstore_heap_t *heap, heap_entry_t *wr);
     void fill_old_clean_entry(blockstore_heap_t *heap, heap_entry_t *big_wr);
     void fill_old_journal_entry(blockstore_heap_t *heap, heap_entry_t *wr);
     int resize_rebuild_meta();
@@ -175,7 +175,6 @@ struct disk_tool_t
 void disk_tool_simple_offsets(json11::Json cfg, bool json_output);
 
 uint64_t sscanf_json(const char *fmt, const json11::Json & str);
-void fromhexstr(const std::string & from, int bytes, uint8_t *to);
 int disable_cache(const std::string & dev);
 uint64_t get_atomic_write_size(const std::string & dev);
 uint64_t get_device_size(const std::string & dev, bool should_exist = false);

src/disk_tool/disk_tool_journal.cpp

@@ -4,6 +4,7 @@
 #include <unistd.h>
 
 #include "disk_tool.h"
+#include "str_util.h"
 #include "malloc_or_die.h"
 
 int disk_tool_t::dump_journal()

src/disk_tool/disk_tool_meta.cpp

@@ -3,7 +3,9 @@
 
 #include "disk_tool.h"
 #include "rw_blocking.h"
+#include "osd_id.h"
 #include "json_util.h"
+#include "str_util.h"
 #include "malloc_or_die.h"
 
 int disk_tool_t::process_meta(std::function<void(blockstore_meta_header_v3_t *)> hdr_fn,
@@ -331,7 +333,7 @@ void disk_tool_t::dump_meta_header(blockstore_meta_header_v3_t *hdr)
     }
     else
     {
-        printf("{\"version\":\"0.5\",\"meta_block_size\":%ju,\"entries\":[\n", dsk.meta_block_size);
+        printf("{\"version\":\"0.5\",\"meta_block_size\":%u,\"entries\":[\n", dsk.meta_block_size);
     }
     first_entry = true;
 }
@@ -751,7 +753,7 @@ close_err0:
                 {
                     *wr->get_checksum(&heap) = sscanf_json("%jx", write_entry["data_crc32c"]);
                 }
-                wr->crc32c = wr->calc_crc32c();
+                wr->checksum = wr->calc_checksum(&heap);
                 assert((uint8_t*)wr + wr->size == new_meta_buf + meta_offset + used_space);
             }
         }
@@ -794,7 +796,7 @@ close_err:
             fromhexstr(meta_entry["ext_bitmap"].string_value(), new_clean_entry_bitmap_size, wr->get_ext_bitmap(&heap));
             if (new_meta_hdr->data_csum_type != 0)
                 fromhexstr(meta_entry["data_csum"].string_value(), new_data_csum_size, wr->get_checksums(&heap));
-            wr->crc32c = wr->calc_crc32c();
+            wr->checksum = wr->calc_checksum(&heap);
             assert((uint8_t*)wr + wr->size == new_meta_buf + meta_offset + used_space);
             auto j_it = journal_by_object.find(oid);
             if (j_it != journal_by_object.end())
@@ -860,7 +862,7 @@ close_err:
                         assert(0);
                     }
                     wr->size = wr->get_size(&heap);
-                    wr->crc32c = wr->calc_crc32c();
+                    wr->checksum = wr->calc_checksum(&heap);
                     assert((uint8_t*)wr + wr->size == new_meta_buf + meta_offset + used_space);
                 }
             }

src/disk_tool/disk_tool_prepare.cpp

@@ -4,6 +4,7 @@
 #include "disk_tool.h"
 #include "str_util.h"
 #include "json_util.h"
+#include "osd_id.h"
 
 void disk_tool_t::parse_meta_reserve()
 {
@@ -149,12 +150,9 @@ int disk_tool_t::prepare_one(std::map<std::string, std::string> options, int is_
     {
         if (options["block_size"] == "")
             options["block_size"] = "1M";
-        if (is_hybrid && options["atomic_write_size"] == "")
-            options["atomic_write_size"] = "0";
         if (is_hybrid && options["throttle_small_writes"] == "")
             options["throttle_small_writes"] = "1";
-        if (!is_hybrid && options.find("data_csum_type") != options.end() && options.at("data_csum_type") != "" &&
-            options["csum_block_size"] == "")
+        if (!is_hybrid && options.find("data_csum_type") != options.end() && options.at("data_csum_type") != "")
             options["csum_block_size"] = "32k";
     }
     else if (!json_is_true(options["disable_data_fsync"]))

src/disk_tool/disk_tool_resize.cpp

@@ -212,11 +212,11 @@ void disk_tool_t::resize_init(blockstore_meta_header_v3_t *hdr)
     dsk.calc_lengths();
     if (((new_data_offset-dsk.data_offset) % dsk.data_block_size))
     {
-        fprintf(stderr, "Data alignment mismatch: old data offset is 0x%jx, new is 0x%jx, but alignment on %jx should be equal\n",
+        fprintf(stderr, "Data alignment mismatch: old data offset is 0x%jx, new is 0x%jx, but alignment on %x should be equal\n",
             dsk.data_offset, new_data_offset, dsk.data_block_size);
         exit(1);
     }
-    data_idx_diff = ((int64_t)(dsk.data_offset-new_data_offset))/((int64_t)dsk.data_block_size);
+    data_idx_diff = ((int64_t)(dsk.data_offset-new_data_offset)) / dsk.data_block_size;
     free_first = new_data_offset > dsk.data_offset ? (new_data_offset-dsk.data_offset) / dsk.data_block_size : 0;
     free_last = (new_data_offset+new_data_len < dsk.data_offset+dsk.data_len)
         ? (dsk.data_offset+dsk.data_len-new_data_offset-new_data_len) / dsk.data_block_size
@@ -351,7 +351,7 @@ int disk_tool_t::resize_copy_data()
                         if (data->res != dsk.data_block_size)
                         {
                             fprintf(
-                                stderr, "Failed to read %ju bytes at %ju from %s: %s\n", dsk.data_block_size,
+                                stderr, "Failed to read %u bytes at %ju from %s: %s\n", dsk.data_block_size,
                                 dsk.data_offset + moving_blocks[i].old_loc*dsk.data_block_size, dsk.data_device.c_str(),
                                 data->res < 0 ? strerror(-data->res) : "short read"
                             );
@@ -376,7 +376,7 @@ int disk_tool_t::resize_copy_data()
                         if (data->res != dsk.data_block_size)
                         {
                             fprintf(
-                                stderr, "Failed to write %ju bytes at %ju to %s: %s\n", dsk.data_block_size,
+                                stderr, "Failed to write %u bytes at %ju to %s: %s\n", dsk.data_block_size,
                                 dsk.data_offset + moving_blocks[i].new_loc*dsk.data_block_size, dsk.data_device.c_str(),
                                 data->res < 0 ? strerror(-data->res) : "short write"
                             );
@@ -531,7 +531,7 @@ int disk_tool_t::resize_write_new_journal()
     return 0;
 }
 
-void disk_tool_t::remap_big_write(heap_entry_t *wr)
+void disk_tool_t::remap_big_write(blockstore_heap_t *heap, heap_entry_t *wr)
 {
     uint64_t block_num = wr->big().block_num;
     auto remap_it = data_remap.find(block_num);
@@ -544,10 +544,10 @@ void disk_tool_t::remap_big_write(heap_entry_t *wr)
     }
     block_num += data_idx_diff;
     wr->big().block_num = block_num;
-    wr->crc32c = wr->calc_crc32c();
+    wr->checksum = wr->calc_checksum(heap);
 }
 
-void disk_tool_t::remap_small_write(heap_entry_t *wr)
+void disk_tool_t::remap_small_write(blockstore_heap_t *heap, heap_entry_t *wr)
 {
     if (new_meta_format == BLOCKSTORE_META_FORMAT_HEAP && wr->small().len > 0)
     {
@@ -559,7 +559,7 @@ void disk_tool_t::remap_small_write(heap_entry_t *wr)
         memcpy(new_journal_ptr, buffer_area+wr->small().location, wr->small().len);
         wr->small().location = new_journal_ptr-new_journal_buf;
         new_journal_ptr += wr->small().len;
-        wr->crc32c = wr->calc_crc32c();
+        wr->checksum = wr->calc_checksum(heap);
     }
 }
 
@@ -673,11 +673,11 @@ int disk_tool_t::resize_rebuild_meta()
             {
                 if (wr->type() == BS_HEAP_BIG_WRITE || wr->type() == BS_HEAP_BIG_INTENT)
                 {
-                    remap_big_write(wr);
+                    remap_big_write(heap, wr);
                 }
                 else if (wr->type() == BS_HEAP_SMALL_WRITE)
                 {
-                    remap_small_write(wr);
+                    remap_small_write(heap, wr);
                 }
                 else if (wr->type() != BS_HEAP_DELETE && new_meta_format != BLOCKSTORE_META_FORMAT_HEAP)
                 {
@@ -786,7 +786,7 @@ int disk_tool_t::resize_rebuild_meta()
                     memcpy(((uint8_t*)wr) + sizeof(heap_big_write_t) + new_clean_entry_bitmap_size, bitmap+new_clean_entry_bitmap_size, new_clean_entry_bitmap_size);
                     memcpy(((uint8_t*)wr) + sizeof(heap_big_write_t) + 2*new_clean_entry_bitmap_size, bitmap+2*new_clean_entry_bitmap_size, new_data_csum_size);
                 }
-                wr->crc32c = wr->calc_crc32c();
+                wr->checksum = wr->calc_checksum(&dsk);
                 new_meta_pos += wr->size;
             }
             else

src/disk_tool/disk_tool_udev.cpp

@@ -140,10 +140,7 @@ uint32_t disk_tool_t::write_osd_superblock(std::string device, json11::Json para
     }
     close(fd);
     free(buf);
-    if (!test_mode)
-    {
-        shell_exec({ "udevadm", "trigger", "--settle", device }, "", NULL, NULL);
-    }
+    shell_exec({ "udevadm", "trigger", "--settle", device }, "", NULL, NULL);
     return sb_size;
 }