WIP vitastor-cli create-user

Seems slightly overcomplicated...

.gitea/workflows/buildenv.Dockerfile

@@ -1,28 +1,29 @@
-FROM node:16-bullseye
+FROM node:16-bookworm
 
 WORKDIR /root
 
-ADD ./docker/vitastor.gpg /etc/apt/trusted.gpg.d
+ADD ./docker/etc/apt/trusted.gpg.d /etc/apt/trusted.gpg.d
 
-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; \
+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; \
     echo >> /etc/apt/preferences; \
     echo 'Package: *' >> /etc/apt/preferences; \
-    echo 'Pin: release a=bullseye-backports' >> /etc/apt/preferences; \
+    echo 'Pin: release n=bookworm-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 libasan5 \
+RUN apt-get -y install etcd qemu-system-x86 qemu-block-extra qemu-utils fio libasan8 \
     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
+RUN apt-get update && apt-get -y install jq lp-solve sudo nfs-common fdisk parted libc-ares-dev udev
 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
@@ -648,6 +702,24 @@ jobs:
           echo ""
         done
 
+  test_snapshot_chain_encrypted:
+    runs-on: ubuntu-latest
+    needs: build
+    container: ${{env.TEST_IMAGE}}:${{github.sha}}
+    steps:
+    - name: Run test
+      id: test
+      timeout-minutes: 3
+      run: ENCRYPTED=1 /root/vitastor/tests/test_snapshot_chain.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_snapshot_chain:
     runs-on: ubuntu-latest
     needs: build
@@ -1206,6 +1278,96 @@ jobs:
           echo ""
         done
 
+  test_checksum:
+    runs-on: ubuntu-latest
+    needs: build
+    container: ${{env.TEST_IMAGE}}:${{github.sha}}
+    steps:
+    - name: Run test
+      id: test
+      timeout-minutes: 3
+      run: /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_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
+    container: ${{env.TEST_IMAGE}}:${{github.sha}}
+    steps:
+    - name: Run test
+      id: test
+      timeout-minutes: 3
+      run: OLD=1 /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_corrupt_all:
+    runs-on: ubuntu-latest
+    needs: build
+    container: ${{env.TEST_IMAGE}}:${{github.sha}}
+    steps:
+    - name: Run test
+      id: test
+      timeout-minutes: 3
+      run: /root/vitastor/tests/test_corrupt_all.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_corrupt_all:
+    runs-on: ubuntu-latest
+    needs: build
+    container: ${{env.TEST_IMAGE}}:${{github.sha}}
+    steps:
+    - name: Run test
+      id: test
+      timeout-minutes: 3
+      run: OLD=1 /root/vitastor/tests/test_corrupt_all.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_reweight_half:
     runs-on: ubuntu-latest
     needs: build
@@ -1980,3 +2142,39 @@ jobs:
           echo ""
         done
 
+  test_write_encrypted:
+    runs-on: ubuntu-latest
+    needs: build
+    container: ${{env.TEST_IMAGE}}:${{github.sha}}
+    steps:
+    - name: Run test
+      id: test
+      timeout-minutes: 3
+      run: /root/vitastor/tests/test_write_encrypted.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_write_encrypted_ec:
+    runs-on: ubuntu-latest
+    needs: build
+    container: ${{env.TEST_IMAGE}}:${{github.sha}}
+    steps:
+    - name: Run test
+      id: test
+      timeout-minutes: 3
+      run: SCHEME=ec /root/vitastor/tests/test_write_encrypted.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
+

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

@@ -38,6 +38,14 @@ for my $line (<>)
             {
                 $test_name .= '_antietcd';
             }
+            elsif ($1 eq 'ETCD_SCHEME' && $2 eq 'https')
+            {
+                $test_name .= '_https';
+            }
+            elsif ($1 eq 'ENCRYPTED')
+            {
+                $test_name .= '_encrypted';
+            }
             elsif ($1 eq 'OLD')
             {
                 $test_name =~ s/^test_/test_old_/s;

.gitignore

@@ -3,3 +3,4 @@
 package-lock.json
 fio
 qemu
+node_modules

CMakeLists.txt

@@ -2,7 +2,7 @@ cmake_minimum_required(VERSION 2.8.12)
 
 project(vitastor)
 
-set(VITASTOR_VERSION "3.0.2")
+set(VITASTOR_VERSION "3.0.5")
 
 include(CTest)
 

README-ru.md

@@ -62,6 +62,7 @@ Vitastor поддерживает QEMU-драйвер, протоколы UBLK,
     - [Дисковые параметры OSD](docs/config/layout-osd.ru.md)
     - [Прочие параметры OSD](docs/config/osd.ru.md)
     - [Параметры мониторов](docs/config/monitor.ru.md)
+    - [Безопасность](docs/config/security.ru.md)
   - [Настройки пулов](docs/config/pool.ru.md)
   - [Метаданные образов в etcd](docs/config/inode.ru.md)
 - Использование

README.md

@@ -62,6 +62,7 @@ Read more details in the documentation. You can start from here: [Quick Start](d
     - [OSD Disk Layout](docs/config/layout-osd.en.md)
     - [OSD Runtime Parameters](docs/config/osd.en.md)
     - [Monitor](docs/config/monitor.en.md)
+    - [Security](docs/config/security.en.md)
   - [Pool configuration](docs/config/pool.en.md)
   - [Image metadata in etcd](docs/config/inode.en.md)
 - Usage

csi/Dockerfile

@@ -1,5 +1,5 @@
 # Compile stage
-FROM golang:bookworm AS build
+FROM golang:trixie 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:bookworm
+FROM debian:trixie
 
 LABEL maintainers="Vitaliy Filippov <vitalif@yourcmc.ru>"
 LABEL description="Vitastor CSI Driver"
@@ -25,20 +25,20 @@ RUN apt-get update && \
         # NFS mount dependencies
         nfs-common netbase \
         # dependencies of qemu-storage-daemon
-        libnuma1 liburing2 libglib2.0-0 libfuse3-3 libaio1 libzstd1 libnettle8 \
-        libgmp10 libhogweed6 libp11-kit0 libidn2-0 libunistring2 libtasn1-6 libpcre2-8-0 libffi8 && \
+        libaio1t64 libc6 libfuse3-4 libglib2.0-0t64 libgmp10 libgnutls30t64 \
+        libhogweed6t64 libnettle8t64 libnuma1 libselinux1 liburing2 libzstd1 zlib1g && \
     apt-get clean && \
     (echo options nbd nbds_max=128 > /etc/modprobe.d/nbd.conf)
 
 COPY --from=build /app/vitastor-csi /bin/
 
-RUN (echo deb http://vitastor.io/debian bookworm main > /etc/apt/sources.list.d/vitastor.list) && \
+RUN (echo deb http://vitastor.io/debian trixie main > /etc/apt/sources.list.d/vitastor.list) && \
     ((echo 'Package: *'; echo 'Pin: origin "vitastor.io"'; echo 'Pin-Priority: 1000') > /etc/apt/preferences.d/vitastor.pref) && \
     wget -q -O /etc/apt/trusted.gpg.d/vitastor.gpg https://vitastor.io/debian/pubkey.gpg && \
     apt-get update && \
     apt-get install -y vitastor-client ibverbs-providers && \
-    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 && \
+    wget https://vitastor.io/archive/qemu/qemu-trixie-10.0.2%2Bds-2%2Bvitastor1/qemu-utils_10.0.2%2Bds-2%2Bvitastor1_amd64.deb && \
+    wget https://vitastor.io/archive/qemu/qemu-trixie-10.0.2%2Bds-2%2Bvitastor1/qemu-block-extra_10.0.2%2Bds-2%2Bvitastor1_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/Dockerfile.local

@@ -1,5 +1,5 @@
 # Compile stage
-FROM golang:bookworm AS build
+FROM golang:trixie 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:bookworm
+FROM debian:trixie
 
 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-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 && \
+    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 && \
     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 ?= v3.0.2
+VITASTOR_VERSION ?= v3.0.5
 
 all: build push
 

csi/deploy/004-csi-nodeplugin.yaml

@@ -49,7 +49,7 @@ spec:
             capabilities:
               add: ["SYS_ADMIN"]
             allowPrivilegeEscalation: true
-          image: vitalif/vitastor-csi:v3.0.2
+          image: vitalif/vitastor-csi:v3.0.5
           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:v3.0.2
+          image: vitalif/vitastor-csi:v3.0.5
           args:
             - "--node=$(NODE_ID)"
             - "--endpoint=$(CSI_ENDPOINT)"

csi/src/config.go

@@ -5,7 +5,7 @@ package vitastor
 
 const (
     vitastorCSIDriverName    = "csi.vitastor.io"
-    vitastorCSIDriverVersion = "3.0.2"
+    vitastorCSIDriverVersion = "3.0.5"
 )
 
 // Config struct fills the parameters of request or user input

debian/build-vitastor-ubuntu-resolute.sh

@@ -0,0 +1,5 @@
+#!/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 (3.0.2-1) unstable; urgency=medium
+vitastor (3.0.5-1) unstable; urgency=medium
 
   * Bugfixes
 

debian/control

@@ -3,7 +3,7 @@ Section: admin
 Priority: optional
 Maintainer: Vitaliy Filippov <vitalif@yourcmc.ru>
 Build-Depends: debhelper, g++ (>= 8), libstdc++6 (>= 8),
-  linux-libc-dev, libgoogle-perftools-dev, libjerasure-dev, libgf-complete-dev,
+  linux-libc-dev, libgoogle-perftools-dev, libjerasure-dev, libgf-complete-dev, libc-ares-dev,
   libibverbs-dev, librdmacm-dev, libisal-dev, cmake, pkg-config, libnl-3-dev, libnl-genl-3-dev,
   node-bindings <!nocheck>, node-gyp, node-nan
 Standards-Version: 4.5.0

debian/vitastor-buildenv.Dockerfile

@@ -25,7 +25,7 @@ RUN set -e -x; \
     echo 'APT::Install-Suggests false;' >> /etc/apt/apt.conf
 
 RUN apt-get update && \
-    apt-get -y install fio libgoogle-perftools-dev devscripts libjerasure-dev cmake \
+    apt-get -y install fio libgoogle-perftools-dev devscripts libjerasure-dev cmake libc-ares-dev \
         libibverbs-dev librdmacm-dev libisal-dev libnl-3-dev libnl-genl-3-dev curl nodejs npm node-nan node-bindings && \
     apt-get -y build-dep fio && \
     apt-get --download-only source fio

docker/Dockerfile

@@ -1,6 +1,6 @@
 # Build Docker image with Vitastor packages
 
-FROM debian:bookworm
+FROM debian:trixie
 
 ADD etc/apt /etc/apt/
 RUN apt-get update && apt-get -y install vitastor ibverbs-providers udev systemd qemu-system-x86 qemu-system-common qemu-block-extra qemu-utils jq nfs-common && apt-get clean

docker/Makefile

@@ -1,4 +1,4 @@
-VITASTOR_VERSION ?= v3.0.2
+VITASTOR_VERSION ?= v3.0.5
 
 all: build push
 

docker/etc/apt/preferences

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

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

@@ -1,2 +1,2 @@
-deb http://vitastor.io/debian bookworm main
-deb http://http.debian.net/debian/ bookworm-backports main
+deb http://vitastor.io/debian trixie main
+#deb http://http.debian.net/debian/ trixie-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 \
+ExecStart=bash -c 'docker run --rm -i -v /etc/vitastor:/etc/vitastor -v /dev:/dev -v /run:/run -e SYSTEMD_IN_CHROOT=0 \
     --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=v3.0.2
+VITASTOR_VERSION=v3.0.5
 
 # Additional arguments for all containers
 # For example, you may want to specify a custom logging driver here

docker/install.sh

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

docs/config.en.md

@@ -38,3 +38,4 @@ In the future, additional configuration methods may be added:
 - [OSD Disk Layout](config/layout-osd.en.md)
 - [OSD Runtime Parameters](config/osd.en.md)
 - [Monitor](config/monitor.en.md)
+- [Security Parameters](config/security.en.md)

docs/config.ru.md

@@ -41,3 +41,4 @@
 - [Дисковые параметры OSD](config/layout-osd.ru.md)
 - [Прочие параметры OSD](config/osd.ru.md)
 - [Параметры мониторов](config/monitor.ru.md)
+- [Параметры безопасности](config/security.ru.md)

docs/config/layout-osd.en.md

@@ -198,8 +198,14 @@ put a modified value into etcd key /vitastor/config/global.
 - Type: string
 - Default: none
 
-Data checksum type to use. May be "crc32c" or "none". Set to "crc32c" to
-enable data checksums.
+Data and metadata checksum type to use. May be "crc32c", "xxh3_32" or "none".
+Select crc32c or xxh3_32 and set csum_block_size to enable data checksums.
+
+Both crc32c and xxh3_32 are almost equally fast, xxh3_32 is safer. xxh3_32 is
+the xxhash3 algorithm truncated from 64 to 32 bits (which is still a good hash).
+
+Note that enabled data checksums either increase memory usage or reduce
+performance. Check details in [csum_block_size](#csum_block_size) description.
 
 ## csum_block_size
 

docs/config/layout-osd.ru.md

@@ -209,8 +209,12 @@ journal_block_size и meta_block_size. Однако на данный момен
 - Тип: строка
 - Значение по умолчанию: none
 
-Тип используемых OSD контрольных сумм данных. Может быть "crc32c" или "none".
-Установите в "crc32c", чтобы включить расчёт и проверку контрольных сумм данных.
+Тип используемых OSD контрольных сумм данных и метаданных. Может быть "crc32c",
+"xxh3_32" или "none". Выберите crc32c или xxh3_32 и установите csum_block_size,
+чтобы включить контрольные суммы данных.
+
+И crc32c, и xxh3_32 примерно одинаково быстры, xxh3_32 надёжней. xxh3_32 - это
+алгоритм xxhash3, обрезанный с 64 до 32 бит (это всё равно хороший хеш).
 
 Следует понимать, что контрольные суммы в зависимости от размера блока их
 расчёта либо увеличивают потребление памяти, либо снижают производительность.

docs/config/network.en.md

@@ -22,7 +22,6 @@ between clients, OSDs and etcd.
 - [rdma_max_msg](#rdma_max_msg)
 - [rdma_max_recv](#rdma_max_recv)
 - [rdma_max_send](#rdma_max_send)
-- [rdma_odp](#rdma_odp)
 - [peer_connect_interval](#peer_connect_interval)
 - [peer_connect_timeout](#peer_connect_timeout)
 - [osd_idle_timeout](#osd_idle_timeout)
@@ -102,11 +101,6 @@ found or if `osd_network` is not specified. Auto-selection is also
 unsupported with old libibverbs < v32, like in Debian 10 Buster or
 CentOS 7.
 
-Vitastor supports all adapters, even ones without ODP support, like
-Mellanox ConnectX-3 and non-Mellanox cards. Versions up to Vitastor
-1.2.0 required ODP which is only present in Mellanox ConnectX >= 4.
-See also [rdma_odp](#rdma_odp).
-
 Run `ibv_devinfo -v` as root to list available RDMA devices and their
 features.
 
@@ -116,6 +110,23 @@ the manual of your network vendor for details about setting up the switch
 for RoCEv2 correctly. Usually it means setting up Lossless Ethernet with
 PFC (Priority Flow Control) and ECN (Explicit Congestion Notification).
 
+Vitastor supports all adapters, even ones without ODP (On-Demand Paging)
+support, like Mellanox ConnectX-3 and non-Mellanox cards. ODP is only present
+in Mellanox ConnectX >= 4 adapters and allows to skip memory registration
+for RDMA and thus, in theory, avoid memory copying.
+
+Versions up to Vitastor 1.2.0 required ODP, then it was disabled by default,
+but it was still supported up to 3.0.3. Now ODP support is removed because it
+actually only hurts performance: an example 3-node cluster with 8 NVMe in each
+node and 2*25 GBit/s ConnectX-6 RDMA network pushed 3950000 read iops without
+ODP, but only 239000 iops with ODP.
+
+This happens because Mellanox ODP implementation seems to be based on
+message retransmissions when the adapter doesn't know about the buffer yet -
+it likely uses standard "RNR retransmissions" (RNR = receiver not ready)
+which is generally slow in RDMA/RoCE networks. Here's a presentation about
+it from ISPASS-2021 conference: https://tkygtr6.github.io/pub/ISPASS21_slides.pdf
+
 ## rdma_port_num
 
 - Type: integer
@@ -187,28 +198,6 @@ less than `rdma_max_recv` so the receiving side doesn't run out of buffers.
 Doesn't affect memory usage - additional memory isn't allocated for send
 operations.
 
-## rdma_odp
-
-- Type: boolean
-- Default: false
-
-Use RDMA with On-Demand Paging. ODP is currently only available on Mellanox
-ConnectX-4 and newer adapters. ODP allows to not register memory explicitly
-for RDMA adapter to be able to use it. This, in turn, allows to skip memory
-copying during sending. One would think this should improve performance, but
-**in reality** RDMA performance with ODP is **drastically** worse. Example
-3-node cluster with 8 NVMe in each node and 2*25 GBit/s ConnectX-6 RDMA network
-without ODP pushes 3950000 read iops, but only 239000 iops with ODP...
-
-This happens because Mellanox ODP implementation seems to be based on
-message retransmissions when the adapter doesn't know about the buffer yet -
-it likely uses standard "RNR retransmissions" (RNR = receiver not ready)
-which is generally slow in RDMA/RoCE networks. Here's a presentation about
-it from ISPASS-2021 conference: https://tkygtr6.github.io/pub/ISPASS21_slides.pdf
-
-ODP support is retained in the code just in case a good ODP implementation
-appears one day.
-
 ## peer_connect_interval
 
 - Type: seconds

docs/config/network.ru.md

@@ -22,7 +22,6 @@
 - [rdma_max_msg](#rdma_max_msg)
 - [rdma_max_recv](#rdma_max_recv)
 - [rdma_max_send](#rdma_max_send)
-- [rdma_odp](#rdma_odp)
 - [peer_connect_interval](#peer_connect_interval)
 - [peer_connect_timeout](#peer_connect_timeout)
 - [osd_idle_timeout](#osd_idle_timeout)
@@ -101,12 +100,6 @@ RoCEv1/RoCEv2, и даже позволяет полностью отключи
 не задана. Также автовыбор не поддерживается со старыми версиями библиотеки
 libibverbs < v32, например в Debian 10 Buster или CentOS 7.
 
-Vitastor поддерживает все модели адаптеров, включая те, у которых
-нет поддержки ODP, то есть вы можете использовать RDMA с ConnectX-3 и
-картами производства не Mellanox. Версии Vitastor до 1.2.0 включительно
-требовали ODP, который есть только на Mellanox ConnectX 4 и более новых.
-См. также [rdma_odp](#rdma_odp).
-
 Запустите `ibv_devinfo -v` от имени суперпользователя, чтобы посмотреть
 список доступных RDMA-устройств, их параметры и возможности.
 
@@ -117,6 +110,24 @@ Vitastor поддерживает все модели адаптеров, вкл
 подразумевает настройку сети без потерь на основе PFC (Priority Flow
 Control) и ECN (Explicit Congestion Notification).
 
+Vitastor поддерживает все модели адаптеров, включая те, у которых нет
+поддержки ODP (On-Demand Paging), например, ConnectX-3 и карты производства
+не Mellanox. Функция ODP доступна только на адаптерах Mellanox ConnectX-4 и
+более новых и позволяет не регистрировать память для её использования RDMA-картой,
+благодаря чему в теории можно избежать лишних копирований памяти.
+
+Версии Vitastor до 1.2.0 включительно требовали ODP, потом функция был отключена
+по умолчанию, но поддерживалась вплоть до версии 3.0.3. Сейчас поддержка ODP
+полностью удалена, так как на самом деле она только портит производительность:
+например, на 3-узловом кластере с 8 NVMe в каждом узле и сетью 2*25 Гбит/с на
+чтение с RDMA без ODP удаётся снять 3950000 iops, а с ODP - всего 239000 iops.
+
+Это происходит из-за того, что реализация ODP у Mellanox неоптимальная и
+основана на повторной передаче сообщений, когда карте не известен буфер -
+вероятно, на стандартных "RNR retransmission" (RNR = receiver not ready).
+А данные повторные передачи в RDMA/RoCE - всегда очень медленная штука.
+Презентация на эту тему с конференции ISPASS-2021: https://tkygtr6.github.io/pub/ISPASS21_slides.pdf
+
 ## rdma_port_num
 
 - Тип: целое число
@@ -192,29 +203,6 @@ OSD в любом случае согласовывают реальное зн
 Не влияет на потребление памяти - дополнительная память на операции отправки
 не выделяется.
 
-## rdma_odp
-
-- Тип: булево (да/нет)
-- Значение по умолчанию: false
-
-Использовать RDMA с On-Demand Paging. ODP - функция, доступная пока что
-исключительно на адаптерах Mellanox ConnectX-4 и более новых. ODP позволяет
-не регистрировать память для её использования RDMA-картой. Благодаря этому
-можно не копировать данные при отправке их в сеть и, казалось бы, это должно
-улучшать производительность - но **по факту** получается так, что
-производительность только ухудшается, причём сильно. Пример - на 3-узловом
-кластере с 8 NVMe в каждом узле и сетью 2*25 Гбит/с на чтение с RDMA без ODP
-удаётся снять 3950000 iops, а с ODP - всего 239000 iops...
-
-Это происходит из-за того, что реализация ODP у Mellanox неоптимальная и
-основана на повторной передаче сообщений, когда карте не известен буфер -
-вероятно, на стандартных "RNR retransmission" (RNR = receiver not ready).
-А данные повторные передачи в RDMA/RoCE - всегда очень медленная штука.
-Презентация на эту тему с конференции ISPASS-2021: https://tkygtr6.github.io/pub/ISPASS21_slides.pdf
-
-Возможность использования ODP сохранена в коде на случай, если вдруг в один
-прекрасный день появится хорошая реализация ODP.
-
 ## peer_connect_interval
 
 - Тип: секунды

docs/config/osd.en.md

@@ -38,6 +38,7 @@ 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)
@@ -279,13 +280,19 @@ Maximum number of journal flushers (see above min_flusher_count).
 - Type: boolean
 - Default: true
 
-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.
+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.
 
 ## inmemory_journal
 
@@ -364,6 +371,8 @@ 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
@@ -377,6 +386,17 @@ 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

@@ -39,6 +39,7 @@
 - [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)
@@ -287,13 +288,19 @@ Flusher - это микро-поток (корутина), которая коп
 - Тип: булево (да/нет)
 - Значение по умолчанию: true
 
-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.
+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.
 
 ## inmemory_journal
 
@@ -376,6 +383,8 @@ 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)
@@ -391,6 +400,18 @@ 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/security.en.md

@@ -0,0 +1,150 @@
+[Documentation](../../README.md#documentation) → [Configuration](../config.en.md) → Security Parameters
+
+-----
+
+[Читать на русском](security.ru.md)
+
+# Security Parameters
+
+These parameters affect your Vitastor installation security and apply to OSDs, monitors and clients.
+
+Most of them can be set in /etc/vitastor/vitastor.conf and in etcd, but don't support online modification.
+
+- [etcd_client_cert](#etcd_client_cert)
+- [etcd_client_key](#etcd_client_key)
+- [etcd_ca](#etcd_ca)
+- [osd_etcd_client_cert](#osd_etcd_client_cert)
+- [osd_etcd_client_key](#osd_etcd_client_key)
+- [mon_etcd_client_cert](#mon_etcd_client_cert)
+- [mon_etcd_client_key](#mon_etcd_client_key)
+- [vault_url](#vault_url)
+- [vault_secret_api_path](#vault_secret_api_path)
+- [vault_client_cert](#vault_client_cert)
+- [vault_client_key](#vault_client_key)
+- [vault_ca](#vault_ca)
+- [vault_timeout_ms](#vault_timeout_ms)
+- [vault_error_timeout_sec](#vault_error_timeout_sec)
+- [vault_refresh_leeway_sec](#vault_refresh_leeway_sec)
+- [max_aes_xts_pool_size](#max_aes_xts_pool_size)
+
+## etcd_client_cert
+
+- Type: string
+
+Client TLS certificate to use for Vitastor client (not OSD and not monitor)
+etcd https connections. May be path to a file or just a PEM string with certificate.
+In the latter case, string must begin with "-----BEGIN CERTIFICATE-----".
+
+## etcd_client_key
+
+- Type: string
+
+Private key for etcd_client_cert (also a file or a PEM string).
+
+## etcd_ca
+
+- Type: string
+
+Trusted TLS CA to verify etcd server certificate. May be path to a file,
+directory or just a PEM string with certificate.
+
+## osd_etcd_client_cert
+
+- Type: string
+
+Same as [etcd_client_cert](#etcd_client_cert), but only for OSDs.
+OSDs, clients and monitors should have different permissions, so they should
+use different certificates.
+
+## osd_etcd_client_key
+
+- Type: string
+
+Same as [etcd_client_key](#etcd_client_key), but only for OSDs.
+
+## mon_etcd_client_cert
+
+- Type: string
+
+Same as [etcd_client_cert](#etcd_client_cert), but only for Vitastor monitors.
+
+## mon_etcd_client_key
+
+- Type: string
+
+Same as [etcd_client_key](#etcd_client_key), but only for Vitastor monitors.
+
+## vault_url
+
+- Type: string
+
+Vault base URL.
+
+Vitastor clients support AES-256-XTS image data encryption with different per-image keys.
+Encryption is performed by the client, OSDs don't have access to decrypted data.
+
+Encryption keys may be stored in etcd or, for the increased security level, in an external
+[HashiCorp Vault](https://developer.hashicorp.com/vault/) or [OpenBao](https://openbao.org/)
+instance.
+
+Vitastor clients use [v1 k/v secrets engine](https://openbao.org/api-docs/secret/kv/kv-v1/)
+and [TLS authentication engine](https://openbao.org/api-docs/auth/cert/) in Vault.
+
+In that case, only key IDs are stored in etcd.
+
+## vault_secret_api_path
+
+- Type: string
+- Default: /v1/secret/
+
+Vault v1 secret API mount path to use.
+
+## vault_client_cert
+
+- Type: string
+
+Client TLS certificate to use for Vault connections. Just like [etcd_client_cert](#etcd_client_cert),
+may be path to a file or just a certificate in PEM string.
+
+## vault_client_key
+
+- Type: string
+
+Private key for vault_client_cert (also a file or a PEM string).
+
+## vault_ca
+
+- Type: string
+
+Trusted TLS CA to verify Vault server certificate. May be path to a file,
+directory or just a PEM string with certificate.
+
+## vault_timeout_ms
+
+- Type: integer
+- Default: 5000
+
+Timeout for Vault requests in milliseconds.
+
+## vault_error_timeout_sec
+
+- Type: integer
+- Default: 60
+
+Time (in seconds) to wait before retrying after receiving an error from Vault.
+
+## vault_refresh_leeway_sec
+
+- Type: integer
+- Default: 60
+
+Extra time (in seconds) before real Vault token lease_timeout to refresh it, just
+in case of system clock drift.
+
+## max_aes_xts_pool_size
+
+- Type: integer
+- Default: 256
+
+Maximum number of OpenSSL encryption contexts cached in OSD memory. Probably
+doesn't require modification.

docs/config/security.ru.md

@@ -0,0 +1,154 @@
+[Документация](../../README-ru.md#документация) → [Конфигурация](../config.ru.md) → Параметры безопасности
+
+-----
+
+[Read in English](security.en.md)
+
+# Параметры безопасности
+
+Данные параметры затрагивают безопасность инсталляций Vitastor и используются
+OSD, мониторами и клиентами.
+
+Большая их часть может задаваться в /etc/vitastor/vitastor.conf и в etcd, но не
+поддерживает онлайн-изменение.
+
+- [etcd_client_cert](#etcd_client_cert)
+- [etcd_client_key](#etcd_client_key)
+- [etcd_ca](#etcd_ca)
+- [osd_etcd_client_cert](#osd_etcd_client_cert)
+- [osd_etcd_client_key](#osd_etcd_client_key)
+- [mon_etcd_client_cert](#mon_etcd_client_cert)
+- [mon_etcd_client_key](#mon_etcd_client_key)
+- [vault_url](#vault_url)
+- [vault_secret_api_path](#vault_secret_api_path)
+- [vault_client_cert](#vault_client_cert)
+- [vault_client_key](#vault_client_key)
+- [vault_ca](#vault_ca)
+- [vault_timeout_ms](#vault_timeout_ms)
+- [vault_error_timeout_sec](#vault_error_timeout_sec)
+- [vault_refresh_leeway_sec](#vault_refresh_leeway_sec)
+- [max_aes_xts_pool_size](#max_aes_xts_pool_size)
+
+## etcd_client_cert
+
+- Тип: строка
+
+Клиентский TLS сертификат для https-подключений к etcd для клиентов Vitastor
+(не OSD и не мониторов). Может быть путём к файлу или просто строкой с
+сертификатом в формате PEM. В последнем случае строка должна начинаться с
+"-----BEGIN CERTIFICATE-----".
+
+## etcd_client_key
+
+- Тип: строка
+
+Закрытый ключ для сертификата etcd_client_cert (также путь к файлу или PEM строка).
+
+## etcd_ca
+
+- Тип: строка
+
+Доверенный корневой TLS-сертификат для проверки сертификата сервера etcd.
+Может быть путём к файлу, директории или просто строкой с сертификатом в
+формате PEM.
+
+## osd_etcd_client_cert
+
+- Тип: строка
+
+Аналогично [etcd_client_cert](#etcd_client_cert), но только для OSD.
+OSD, клиенты и мониторы должны иметь разные привилегии, поэтому они должны
+использовать разные сертификаты.
+
+## osd_etcd_client_key
+
+- Тип: строка
+
+Аналогично [etcd_client_key](#etcd_client_key), но только для OSD.
+
+## mon_etcd_client_cert
+
+- Тип: строка
+
+Аналогично [etcd_client_cert](#etcd_client_cert), но только для мониторов Vitastor.
+
+## mon_etcd_client_key
+
+- Тип: строка
+
+Аналогично [etcd_client_key](#etcd_client_key), но только для мониторов Vitastor.
+
+## vault_url
+
+- Тип: строка
+
+Базовый адрес Vault.
+
+Клиенты Vitastor поддерживают AES-256-XTS шифрование данных образов с отдельными ключами на
+каждый образ. Данные шифруются клиентами, OSD не имеют доступа к незашифрованным данным.
+
+Ключи шифрования могут храниться в etcd или, для повышенного уровня безопасности, во внешнем
+[HashiCorp Vault](https://developer.hashicorp.com/vault/) или [OpenBao](https://openbao.org/).
+
+Клиенты Vitastor используют [движок секретов v1](https://openbao.org/api-docs/secret/kv/kv-v1/)
+и [TLS-аутентификацию](https://openbao.org/api-docs/auth/cert/) в Vault.
+
+В этом случае, только ID ключей хранятся в etcd.
+
+## vault_secret_api_path
+
+- Тип: строка
+- Значение по умолчанию: /v1/secret/
+
+Путь к API секретов v1 для использования клиентами.
+
+## vault_client_cert
+
+- Тип: строка
+
+Клиентский TLS сертификат для подключений к Vault. Как и [etcd_client_cert](#etcd_client_cert),
+может быть путём к файлу или просто PEM-строкой с сертификатом.
+
+## vault_client_key
+
+- Тип: строка
+
+Закрытый ключ для сертификата vault_client_cert (также путь к файлу или PEM строка).
+
+## vault_ca
+
+- Тип: строка
+
+Доверенный корневой TLS-сертификат для проверки сертификата сервера Vault.
+Может быть путём к файлу, директории или просто строкой с сертификатом в
+формате PEM.
+
+## vault_timeout_ms
+
+- Тип: целое число
+- Значение по умолчанию: 5000
+
+Максимально время выполнения Vault-запросов в миллисекундах.
+
+## vault_error_timeout_sec
+
+- Тип: целое число
+- Значение по умолчанию: 60
+
+Время (в секундах) для ожидания перед повторной попыткой при получении ошибки от Vault.
+
+## vault_refresh_leeway_sec
+
+- Тип: целое число
+- Значение по умолчанию: 60
+
+Зазор времени (в секундах), чтобы обновлять токены Vault чуть раньше их реального
+lease_timeout, на случай "ухода" системных часов.
+
+## max_aes_xts_pool_size
+
+- Тип: целое число
+- Значение по умолчанию: 256
+
+Максимальное количество кэшируемых в памяти OSD контекстов шифрования OpenSSL.
+Вряд ли требует изменения.

docs/config/src/included.en.md

@@ -44,6 +44,8 @@
 
 {{../../config/monitor.en.md|indent=2}}
 
+{{../../config/security.en.md|indent=2}}
+
 {{../../config/pool.en.md|indent=2}}
 
 {{../../config/inode.en.md|indent=2}}

docs/config/src/included.ru.md

@@ -44,6 +44,8 @@
 
 {{../../config/monitor.ru.md|indent=2}}
 
+{{../../config/security.ru.md|indent=2}}
+
 {{../../config/pool.ru.md|indent=2}}
 
 {{../../config/inode.ru.md|indent=2}}

docs/config/src/layout-osd.yml

@@ -233,11 +233,21 @@
   type: string
   default: none
   info: |
-    Data checksum type to use. May be "crc32c" or "none". Set to "crc32c" to
-    enable data checksums.
+    Data and metadata checksum type to use. May be "crc32c", "xxh3_32" or "none".
+    Select crc32c or xxh3_32 and set csum_block_size to enable data checksums.
+
+    Both crc32c and xxh3_32 are almost equally fast, xxh3_32 is safer. xxh3_32 is
+    the xxhash3 algorithm truncated from 64 to 32 bits (which is still a good hash).
+
+    Note that enabled data checksums either increase memory usage or reduce
+    performance. Check details in [csum_block_size](#csum_block_size) description.
   info_ru: |
-    Тип используемых OSD контрольных сумм данных. Может быть "crc32c" или "none".
-    Установите в "crc32c", чтобы включить расчёт и проверку контрольных сумм данных.
+    Тип используемых OSD контрольных сумм данных и метаданных. Может быть "crc32c",
+    "xxh3_32" или "none". Выберите crc32c или xxh3_32 и установите csum_block_size,
+    чтобы включить контрольные суммы данных.
+
+    И crc32c, и xxh3_32 примерно одинаково быстры, xxh3_32 надёжней. xxh3_32 - это
+    алгоритм xxhash3, обрезанный с 64 до 32 бит (это всё равно хороший хеш).
 
     Следует понимать, что контрольные суммы в зависимости от размера блока их
     расчёта либо увеличивают потребление памяти, либо снижают производительность.

docs/config/src/network.yml

@@ -84,11 +84,6 @@
     unsupported with old libibverbs < v32, like in Debian 10 Buster or
     CentOS 7.
 
-    Vitastor supports all adapters, even ones without ODP support, like
-    Mellanox ConnectX-3 and non-Mellanox cards. Versions up to Vitastor
-    1.2.0 required ODP which is only present in Mellanox ConnectX >= 4.
-    See also [rdma_odp](#rdma_odp).
-
     Run `ibv_devinfo -v` as root to list available RDMA devices and their
     features.
 
@@ -97,6 +92,23 @@
     the manual of your network vendor for details about setting up the switch
     for RoCEv2 correctly. Usually it means setting up Lossless Ethernet with
     PFC (Priority Flow Control) and ECN (Explicit Congestion Notification).
+
+    Vitastor supports all adapters, even ones without ODP (On-Demand Paging)
+    support, like Mellanox ConnectX-3 and non-Mellanox cards. ODP is only present
+    in Mellanox ConnectX >= 4 adapters and allows to skip memory registration
+    for RDMA and thus, in theory, avoid memory copying.
+
+    Versions up to Vitastor 1.2.0 required ODP, then it was disabled by default,
+    but it was still supported up to 3.0.3. Now ODP support is removed because it
+    actually only hurts performance: an example 3-node cluster with 8 NVMe in each
+    node and 2*25 GBit/s ConnectX-6 RDMA network pushed 3950000 read iops without
+    ODP, but only 239000 iops with ODP.
+
+    This happens because Mellanox ODP implementation seems to be based on
+    message retransmissions when the adapter doesn't know about the buffer yet -
+    it likely uses standard "RNR retransmissions" (RNR = receiver not ready)
+    which is generally slow in RDMA/RoCE networks. Here's a presentation about
+    it from ISPASS-2021 conference: https://tkygtr6.github.io/pub/ISPASS21_slides.pdf
   info_ru: |
     Название RDMA-устройства для связи с Vitastor OSD (например, "rocep5s0f0").
     Если не указано, Vitastor попробует найти RoCE-устройство, соответствующее
@@ -105,12 +117,6 @@
     не задана. Также автовыбор не поддерживается со старыми версиями библиотеки
     libibverbs < v32, например в Debian 10 Buster или CentOS 7.
 
-    Vitastor поддерживает все модели адаптеров, включая те, у которых
-    нет поддержки ODP, то есть вы можете использовать RDMA с ConnectX-3 и
-    картами производства не Mellanox. Версии Vitastor до 1.2.0 включительно
-    требовали ODP, который есть только на Mellanox ConnectX 4 и более новых.
-    См. также [rdma_odp](#rdma_odp).
-
     Запустите `ibv_devinfo -v` от имени суперпользователя, чтобы посмотреть
     список доступных RDMA-устройств, их параметры и возможности.
 
@@ -120,6 +126,24 @@
     коммутатора для RoCEv2 ищите в документации производителя. Обычно это
     подразумевает настройку сети без потерь на основе PFC (Priority Flow
     Control) и ECN (Explicit Congestion Notification).
+
+    Vitastor поддерживает все модели адаптеров, включая те, у которых нет
+    поддержки ODP (On-Demand Paging), например, ConnectX-3 и карты производства
+    не Mellanox. Функция ODP доступна только на адаптерах Mellanox ConnectX-4 и
+    более новых и позволяет не регистрировать память для её использования RDMA-картой,
+    благодаря чему в теории можно избежать лишних копирований памяти.
+
+    Версии Vitastor до 1.2.0 включительно требовали ODP, потом функция был отключена
+    по умолчанию, но поддерживалась вплоть до версии 3.0.3. Сейчас поддержка ODP
+    полностью удалена, так как на самом деле она только портит производительность:
+    например, на 3-узловом кластере с 8 NVMe в каждом узле и сетью 2*25 Гбит/с на
+    чтение с RDMA без ODP удаётся снять 3950000 iops, а с ODP - всего 239000 iops.
+
+    Это происходит из-за того, что реализация ODP у Mellanox неоптимальная и
+    основана на повторной передаче сообщений, когда карте не известен буфер -
+    вероятно, на стандартных "RNR retransmission" (RNR = receiver not ready).
+    А данные повторные передачи в RDMA/RoCE - всегда очень медленная штука.
+    Презентация на эту тему с конференции ISPASS-2021: https://tkygtr6.github.io/pub/ISPASS21_slides.pdf
 - name: rdma_port_num
   type: int
   info: |
@@ -218,45 +242,6 @@
     у принимающей стороны в процессе работы не заканчивались буферы на приём.
     Не влияет на потребление памяти - дополнительная память на операции отправки
     не выделяется.
-- name: rdma_odp
-  type: bool
-  default: false
-  online: false
-  info: |
-    Use RDMA with On-Demand Paging. ODP is currently only available on Mellanox
-    ConnectX-4 and newer adapters. ODP allows to not register memory explicitly
-    for RDMA adapter to be able to use it. This, in turn, allows to skip memory
-    copying during sending. One would think this should improve performance, but
-    **in reality** RDMA performance with ODP is **drastically** worse. Example
-    3-node cluster with 8 NVMe in each node and 2*25 GBit/s ConnectX-6 RDMA network
-    without ODP pushes 3950000 read iops, but only 239000 iops with ODP...
-
-    This happens because Mellanox ODP implementation seems to be based on
-    message retransmissions when the adapter doesn't know about the buffer yet -
-    it likely uses standard "RNR retransmissions" (RNR = receiver not ready)
-    which is generally slow in RDMA/RoCE networks. Here's a presentation about
-    it from ISPASS-2021 conference: https://tkygtr6.github.io/pub/ISPASS21_slides.pdf
-
-    ODP support is retained in the code just in case a good ODP implementation
-    appears one day.
-  info_ru: |
-    Использовать RDMA с On-Demand Paging. ODP - функция, доступная пока что
-    исключительно на адаптерах Mellanox ConnectX-4 и более новых. ODP позволяет
-    не регистрировать память для её использования RDMA-картой. Благодаря этому
-    можно не копировать данные при отправке их в сеть и, казалось бы, это должно
-    улучшать производительность - но **по факту** получается так, что
-    производительность только ухудшается, причём сильно. Пример - на 3-узловом
-    кластере с 8 NVMe в каждом узле и сетью 2*25 Гбит/с на чтение с RDMA без ODP
-    удаётся снять 3950000 iops, а с ODP - всего 239000 iops...
-
-    Это происходит из-за того, что реализация ODP у Mellanox неоптимальная и
-    основана на повторной передаче сообщений, когда карте не известен буфер -
-    вероятно, на стандартных "RNR retransmission" (RNR = receiver not ready).
-    А данные повторные передачи в RDMA/RoCE - всегда очень медленная штука.
-    Презентация на эту тему с конференции ISPASS-2021: https://tkygtr6.github.io/pub/ISPASS21_slides.pdf
-
-    Возможность использования ODP сохранена в коде на случай, если вдруг в один
-    прекрасный день появится хорошая реализация ODP.
 - name: peer_connect_interval
   type: sec
   min: 1

docs/config/src/osd.yml

@@ -253,21 +253,33 @@
   type: bool
   default: true
   info: |
-    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.
+    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.
   info_ru: |
-    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.
+    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.
 - name: inmemory_journal
   type: bool
   default: true
@@ -386,11 +398,15 @@
     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
@@ -402,6 +418,8 @@
     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
@@ -412,6 +430,20 @@
     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/config/src/package.json

@@ -0,0 +1,5 @@
+{
+  "dependencies": {
+    "yaml": "^2.8.2"
+  }
+}

docs/config/src/security.en.md

@@ -0,0 +1,5 @@
+# Security Parameters
+
+These parameters affect your Vitastor installation security and apply to OSDs, monitors and clients.
+
+Most of them can be set in /etc/vitastor/vitastor.conf and in etcd, but don't support online modification.

docs/config/src/security.ru.md

@@ -0,0 +1,7 @@
+# Параметры безопасности
+
+Данные параметры затрагивают безопасность инсталляций Vitastor и используются
+OSD, мониторами и клиентами.
+
+Большая их часть может задаваться в /etc/vitastor/vitastor.conf и в etcd, но не
+поддерживает онлайн-изменение.

docs/config/src/security.yml

@@ -0,0 +1,131 @@
+- name: etcd_client_cert
+  type: string
+  info: |
+    Client TLS certificate to use for Vitastor client (not OSD and not monitor)
+    etcd https connections. May be path to a file or just a PEM string with certificate.
+    In the latter case, string must begin with "-----BEGIN CERTIFICATE-----".
+  info_ru: |
+    Клиентский TLS сертификат для https-подключений к etcd для клиентов Vitastor
+    (не OSD и не мониторов). Может быть путём к файлу или просто строкой с
+    сертификатом в формате PEM. В последнем случае строка должна начинаться с
+    "-----BEGIN CERTIFICATE-----".
+- name: etcd_client_key
+  type: string
+  info: Private key for etcd_client_cert (also a file or a PEM string).
+  info_ru: Закрытый ключ для сертификата etcd_client_cert (также путь к файлу или PEM строка).
+- name: etcd_ca
+  type: string
+  info: |
+    Trusted TLS CA to verify etcd server certificate. May be path to a file,
+    directory or just a PEM string with certificate.
+  info_ru: |
+    Доверенный корневой TLS-сертификат для проверки сертификата сервера etcd.
+    Может быть путём к файлу, директории или просто строкой с сертификатом в
+    формате PEM.
+- name: osd_etcd_client_cert
+  type: string
+  info: |
+    Same as [etcd_client_cert](#etcd_client_cert), but only for OSDs.
+    OSDs, clients and monitors should have different permissions, so they should
+    use different certificates.
+  info_ru: |
+    Аналогично [etcd_client_cert](#etcd_client_cert), но только для OSD.
+    OSD, клиенты и мониторы должны иметь разные привилегии, поэтому они должны
+    использовать разные сертификаты.
+- name: osd_etcd_client_key
+  type: string
+  info: Same as [etcd_client_key](#etcd_client_key), but only for OSDs.
+  info_ru: Аналогично [etcd_client_key](#etcd_client_key), но только для OSD.
+- name: mon_etcd_client_cert
+  type: string
+  info: Same as [etcd_client_cert](#etcd_client_cert), but only for Vitastor monitors.
+  info_ru: Аналогично [etcd_client_cert](#etcd_client_cert), но только для мониторов Vitastor.
+- name: mon_etcd_client_key
+  type: string
+  info: Same as [etcd_client_key](#etcd_client_key), but only for Vitastor monitors.
+  info_ru: Аналогично [etcd_client_key](#etcd_client_key), но только для мониторов Vitastor.
+- name: vault_url
+  type: string
+  info: |
+    Vault base URL.
+
+    Vitastor clients support AES-256-XTS image data encryption with different per-image keys.
+    Encryption is performed by the client, OSDs don't have access to decrypted data.
+
+    Encryption keys may be stored in etcd or, for the increased security level, in an external
+    [HashiCorp Vault](https://developer.hashicorp.com/vault/) or [OpenBao](https://openbao.org/)
+    instance.
+
+    Vitastor clients use [v1 k/v secrets engine](https://openbao.org/api-docs/secret/kv/kv-v1/)
+    and [TLS authentication engine](https://openbao.org/api-docs/auth/cert/) in Vault.
+
+    In that case, only key IDs are stored in etcd.
+  info_ru: |
+    Базовый адрес Vault.
+
+    Клиенты Vitastor поддерживают AES-256-XTS шифрование данных образов с отдельными ключами на
+    каждый образ. Данные шифруются клиентами, OSD не имеют доступа к незашифрованным данным.
+
+    Ключи шифрования могут храниться в etcd или, для повышенного уровня безопасности, во внешнем
+    [HashiCorp Vault](https://developer.hashicorp.com/vault/) или [OpenBao](https://openbao.org/).
+
+    Клиенты Vitastor используют [движок секретов v1](https://openbao.org/api-docs/secret/kv/kv-v1/)
+    и [TLS-аутентификацию](https://openbao.org/api-docs/auth/cert/) в Vault.
+
+    В этом случае, только ID ключей хранятся в etcd.
+- name: vault_secret_api_path
+  type: string
+  default: /v1/secret/
+  info: Vault v1 secret API mount path to use.
+  info_ru: Путь к API секретов v1 для использования клиентами.
+- name: vault_client_cert
+  type: string
+  info: |
+    Client TLS certificate to use for Vault connections. Just like [etcd_client_cert](#etcd_client_cert),
+    may be path to a file or just a certificate in PEM string.
+  info_ru: |
+    Клиентский TLS сертификат для подключений к Vault. Как и [etcd_client_cert](#etcd_client_cert),
+    может быть путём к файлу или просто PEM-строкой с сертификатом.
+- name: vault_client_key
+  type: string
+  info: Private key for vault_client_cert (also a file or a PEM string).
+  info_ru: Закрытый ключ для сертификата vault_client_cert (также путь к файлу или PEM строка).
+- name: vault_ca
+  type: string
+  info: |
+    Trusted TLS CA to verify Vault server certificate. May be path to a file,
+    directory or just a PEM string with certificate.
+  info_ru: |
+    Доверенный корневой TLS-сертификат для проверки сертификата сервера Vault.
+    Может быть путём к файлу, директории или просто строкой с сертификатом в
+    формате PEM.
+- name: vault_timeout_ms
+  type: int
+  default: 5000
+  info: Timeout for Vault requests in milliseconds.
+  info_ru: Максимально время выполнения Vault-запросов в миллисекундах.
+- name: vault_error_timeout_sec
+  type: int
+  default: 60
+  info: |
+    Time (in seconds) to wait before retrying after receiving an error from Vault.
+  info_ru: |
+    Время (в секундах) для ожидания перед повторной попыткой при получении ошибки от Vault.
+- name: vault_refresh_leeway_sec
+  type: int
+  default: 60
+  info: |
+    Extra time (in seconds) before real Vault token lease_timeout to refresh it, just
+    in case of system clock drift.
+  info_ru: |
+    Зазор времени (в секундах), чтобы обновлять токены Vault чуть раньше их реального
+    lease_timeout, на случай "ухода" системных часов.
+- name: max_aes_xts_pool_size
+  type: int
+  default: 256
+  info: |
+    Maximum number of OpenSSL encryption contexts cached in OSD memory. Probably
+    doesn't require modification.
+  info_ru: |
+    Максимальное количество кэшируемых в памяти OSD контекстов шифрования OpenSSL.
+    Вряд ли требует изменения.

docs/installation/docker.en.md

@@ -26,13 +26,37 @@ 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:v3.0.2`
+   `docker pull vitalif/vitastor:v3.0.5`
 2. Install scripts to the host system: \
-   `docker run --rm -it -v /etc:/host-etc -v /usr/bin:/host-bin vitalif/vitastor:v3.0.2 install.sh`
+   `docker run --rm -it -v /etc:/host-etc -v /usr/bin:/host-bin vitalif/vitastor:v3.0.5 install.sh`
 3. Reload udev rules: \
    `udevadm control --reload-rules`
+4. Enable the vitastor-host service: \
+   `systemctl enable --now vitastor-host`
 
-And you can return to [Quick Start](../intro/quickstart.en.md).
+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.
 
 ## Upgrading Containers
 

docs/installation/docker.ru.md

@@ -25,14 +25,39 @@ Vitastor можно установить в Docker/Podman. При этом etcd,
 Инструкция по установке максимально простая.
 
 1. Скачайте Docker-образ желаемой версии: \
-   `docker pull vitalif/vitastor:v3.0.2`
+   `docker pull vitalif/vitastor:v3.0.5`
 2. Установите скрипты в хост-систему командой: \
-   `docker run --rm -it -v /etc:/host-etc -v /usr/bin:/host-bin vitalif/vitastor:v3.0.2 install.sh`
+   `docker run --rm -it -v /etc:/host-etc -v /usr/bin:/host-bin vitalif/vitastor:v3.0.5 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

@@ -33,15 +33,17 @@
   - 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 clones: not required
+  - RHEL 9/10 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

@@ -33,15 +33,17 @@
   - 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: не нужно
+  - Клоны RHEL 9/10: 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

docs/installation/source.en.md

@@ -15,8 +15,8 @@
 - gcc and g++ 8 or newer, clang 10 or newer, or other compiler with C++11 plus
   designated initializers support from C++20
 - CMake
-- jerasure headers and libraries
-- ISA-L, libibverbs and librdmacm headers and libraries (optional)
+- jerasure, c-ares headers and libraries
+- ISA-L, libibverbs, librdmacm, libnl3 headers and libraries (optional)
 - tcmalloc (google-perftools-dev)
 
 ## Basic instructions

docs/installation/source.ru.md

@@ -15,8 +15,8 @@
 - gcc и g++ >= 8, либо clang >= 10, либо другой компилятор с поддержкой C++11 плюс
   назначенных инициализаторов (designated initializers) из C++20
 - CMake
-- Заголовки и библиотеки jerasure
-- Опционально - заголовки и библиотеки ISA-L, libibverbs, librdmacm
+- Заголовки и библиотеки jerasure, c-ares
+- Опционально - заголовки и библиотеки ISA-L, libibverbs, librdmacm, libnl3
 - tcmalloc (google-perftools-dev)
 
 ## Базовая инструкция

docs/intro/features.en.md

@@ -41,6 +41,8 @@
 - [Built-in Prometheus metric exporter](../config/monitor.en.md#enable_prometheus)
 - [NFS RDMA support](../usage/nfs.en.md#rdma) (probably also usable for GPUDirect)
 - [S3](../installation/s3.en.md)
+- [TLS support for etcd connections](../config/security.en.md)
+- [AES-256-XTS image encryption](../usage/cli.en.md#create) and [Vault support](../config/security.en.md#vault_url) for key storage
 
 ## Plugins and tools
 

docs/intro/features.ru.md

@@ -43,6 +43,8 @@
 - [Встроенный Prometheus-экспортер метрик](../config/monitor.ru.md#enable_prometheus)
 - [Поддержка NFS RDMA](../usage/nfs.ru.md#rdma) (вероятно, также подходящая для GPUDirect)
 - [S3](../installation/s3.ru.md)
+- [Поддержка TLS-соединений с etcd](../config/security.ru.md)
+- [AES-256-XTS шифрование данных](../usage/cli.ru.md#create) и [поддержка Vault](../config/security.ru.md#vault_url) для хранения ключей
 
 ## Драйверы и инструменты
 

docs/usage/cli.en.md

@@ -125,18 +125,31 @@ bench-kaveri         kaveri    10 G  10 G    0 B/s  0     0      0 us  0 B/s  0
 
 ## create
 
-`vitastor-cli create -s|--size <size> [-p|--pool <id|name>] [--parent <parent_name>[@<snapshot>]] <name>`
+`vitastor-cli create -s|--size SIZE [OPTIONS] <name>`
 
-Create an image. You may use K/M/G/T suffixes for `<size>`. If `--parent` is specified,
-a copy-on-write image clone is created. Parent must be a snapshot (readonly image).
-Pool must be specified if there is more than one pool.
+Create an image. Options:
+
+* `-s|--size SIZE` - New image size in bytes or with a K/M/G/T unit suffix.
+* `-p|--pool POOL` - Specify pool for the new image (may be omitted if there is only 1 pool).
+* `--parent PARENT` - Create a copy-on-write image clone based on PARENT (or PARENT@SNAPSHOT).
+  If parent is not a snapshot, it must be a read-only image.
+* `--enc-key random` - Generate a new random AES-256-XTS encryption key for the new image.
+* `--enc-key HEX` - Set a specified AES-256-XTS key (64 bytes in hex) for the new image.
+* `--enc-key vault:ID` - Use an encryption key from an external Vault secret with specified ID.
 
 ```
-vitastor-cli create --snapshot <snapshot> [-p|--pool <id|name>] <image>
-vitastor-cli snap-create [-p|--pool <id|name>] <image>@<snapshot>
+vitastor-cli create --snapshot <snapshot> [OPTIONS] <image>
+vitastor-cli snap-create [OPTIONS] <image>@<snapshot>
 ```
 
-Create a snapshot of image `<name>` (either form can be used). May be used live if only a single writer is active.
+Create a snapshot of image `<image>`. May be used live if only a single writer is active.
+
+Options:
+
+* `-p|--pool POOL` - Move image to pool POOL, leaving the snapshot in the old pool.
+* `--enc-key random` - Change image encryption key to a new random AES-256-XTS key.
+* `--enc-key KEY` - Change image encryption key to a specified key, Vault key or to an empty key.
+  By default, the image retains its old encryption key when taking a snapshot.
 
 See also about [how to export snapshots](qemu.en.md#exporting-snapshots).
 
@@ -151,6 +164,7 @@ You should resize file system in the image, if present, before shrinking it.
 * `--deleted 1|0` - Set/clear 'deleted image' flag (set automatically during unfinished deletes).
 * `-f|--force` - Proceed with shrinking or setting readwrite flag even if the image has children.
 * `--down-ok` - Proceed with shrinking even if some data will be left on unavailable OSDs.
+* `--enc-key HEX` - Change image encryption key (allowed only with `--force`).
 
 ## dd
 

docs/usage/cli.ru.md

@@ -127,19 +127,32 @@ bench-kaveri         kaveri    10 G  10 G    0 B/s  0     0      0 us  0 B/s  0
 
 ## create
 
-`vitastor-cli create -s|--size <size> [-p|--pool <id|name>] [--parent <parent_name>[@<snapshot>]] <name>`
+`vitastor-cli create -s|--size SIZE [ОПЦИИ] <name>`
 
-Создать образ. Для размера `<size>` можно использовать суффиксы K/M/G/T (килобайт-мегабайт-гигабайт-терабайт).
-Если указана опция `--parent`, создаётся клон образа. Родитель `<parent_name>[@<snapshot>]` должен быть
-снимком (или просто немодифицируемым образом). Пул обязательно указывать, если в кластере больше одного пула.
+Создать образ. Опции:
+
+* `-s|--size SIZE` - Размер нового образа в байтах или с суффиксом K/M/G/T (кило/мега/гига/терабайт).
+* `-p|--pool POOL` - Создать образ в заданном пуле (можно не указывать, если пул всего один).
+* `--parent PARENT` - Создать легковесный клон на основе образа `PARENT` или снимка `PARENT@SNAP`.
+  Если `PARENT` - не снимок, он должен быть помечен как образ только для чтения.
+* `--enc-key random` - Сгенерировать случайный ключ шифрования AES-256-XTS для нового образа.
+* `--enc-key HEX` - Установить заданный ключ AES-256-XTS (64 байта в hex) для нового образа.
+* `--enc-key vault:ID` - Использовать ключ из внешнего секрета с заданным ID из Vault.
 
 ```
-vitastor-cli create --snapshot <snapshot> [-p|--pool <id|name>] <image>
-vitastor-cli snap-create [-p|--pool <id|name>] <image>@<snapshot>
+vitastor-cli create --snapshot <snapshot> [ОПЦИИ] <image>
+vitastor-cli snap-create [ОПЦИИ] <image>@<snapshot>
 ```
 
-Создать снимок образа `<name>` (можно использовать любую форму команды). Снимок можно создавать без остановки
-клиентов, если пишущий клиент максимум 1.
+Создать снимок образа `<image>` (можно использовать любую форму команды).
+Снимок можно создавать без остановки клиентов, если пишущих клиентов не больше одного.
+
+Опции:
+
+* `-p|--pool POOL` - Переместить образ в пул POOL, оставив снимок в старом пуле.
+* `--enc-key random` - Изменить ключ шифрования образа на новый случайный ключ AES-256-XTS.
+* `--enc-key KEY` - Изменить ключ шифрования образа на заданный ключ, ключ из Vault или пустой ключ.
+  По умолчанию шифрованные образы сохраняют старый ключ при снятии снимка.
 
 Смотрите также информацию о том, [как экспортировать снимки](qemu.ru.md#экспорт-снимков).
 
@@ -156,6 +169,7 @@ vitastor-cli snap-create [-p|--pool <id|name>] <image>@<snapshot>
 * `--deleted 1|0` - Установить/снять флаг "образ удалён" (устанавливается при незавершённом удалении).
 * `-f|--force` - Разрешить уменьшение или перевод в чтение-запись образа, у которого есть клоны.
 * `--down-ok` - Разрешить уменьшение, даже если часть данных останется неудалённой на недоступных OSD.
+* `--enc-key HEX` - Изменить ключ шифрования образа (разрешено только с `--force`).
 
 ## dd
 

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 +17,30 @@ 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.mon_etcd_client_cert || config.etcd_client_cert)
+        {
+            this.opts.cert = config.mon_etcd_client_cert || config.etcd_client_cert;
+            if (this.opts.cert.substr(0, 5) != '-----')
+                this.opts.cert = fs.readFileSync(this.opts.cert, { encoding: 'utf-8' });
+        }
+        if (config.mon_etcd_client_key || config.etcd_client_key)
+        {
+            this.opts.key = config.mon_etcd_client_key || config.etcd_client_key;
+            if (this.opts.key.substr(0, 5) != '-----')
+                this.opts.key = fs.readFileSync(this.opts.key, { encoding: 'utf-8' });
+        }
+        if (config.etcd_ca)
+        {
+            this.opts.ca = config.etcd_ca;
+            if (this.opts.ca.substr(0, 5) != '-----')
+                this.opts.ca = fs.readFileSync(this.opts.ca, { encoding: 'utf-8' });
+        }
     }
 
     parse_etcd_addresses(addrs)
@@ -39,7 +60,7 @@ 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 +151,7 @@ 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 +293,7 @@ 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 +319,7 @@ class EtcdAdapter
     }
 }
 
-function POST(url, body, timeout)
+function POST(url, body, timeout, opts)
 {
     return new Promise(ok =>
     {
@@ -310,10 +331,10 @@ 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 +45,14 @@ 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: "",
+            osd_etcd_client_cert: "",
+            osd_etcd_client_key: "",
+            mon_etcd_client_cert: "",
+            mon_etcd_client_key: "",
+            etcd_ca: "",
             // mon
             etcd_mon_ttl: 5, // min: 1
             etcd_mon_timeout: 1000, // ms. min: 0
@@ -217,6 +224,8 @@ const etcd_tree = {
                     parent_id?: <inode_t>,
                     readonly?: boolean,
                     deleted?: boolean,
+                    enc_key?: string,
+                    meta?: any,
                 }
             }
         }, */
@@ -383,6 +392,7 @@ const etcd_tree = {
             /* <name>: {
                 id: uint64_t,
                 pool_id: uint64_t,
+                // ...plus a copy of everything from config/inode/x/y
             }, */
         },
         maxid: {

mon/http_server.js

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

mon/lp_optimizer/lp_optimizer.js

@@ -10,16 +10,19 @@ const NO_OSD = 'Z';
 async function lp_solve(text)
 {
     const cp = child_process.spawn('lp_solve');
-    let stdout = '', stderr = '', finish_cb;
+    let stdout = '', stderr = '', finish_cb, finished = 0;
     cp.stdout.on('data', buf => stdout += buf.toString());
     cp.stderr.on('data', buf => stderr += buf.toString());
-    cp.on('exit', () => finish_cb && finish_cb());
+    cp.stdout.on('end', () => finish_cb());
+    cp.stderr.on('end', () => finish_cb());
     cp.stdin.write(text);
     cp.stdin.end();
-    if (cp.exitCode == null)
+    await new Promise(ok => (finish_cb = () =>
     {
-        await new Promise(ok => finish_cb = ok);
-    }
+        finished++;
+        if (finished == 2)
+            ok();
+    }));
     if (!stdout.trim())
     {
         return null;

mon/package.json

@@ -1,6 +1,6 @@
 {
   "name": "vitastor-mon",
-  "version": "3.0.2",
+  "version": "3.0.5",
   "description": "Vitastor SDS monitor service",
   "main": "mon-main.js",
   "scripts": {
@@ -9,7 +9,7 @@
   "author": "Vitaliy Filippov",
   "license": "UNLICENSED",
   "dependencies": {
-    "antietcd": "^1.2.2",
+    "antietcd": "^1.2.4",
     "sprintf-js": "^1.1.2",
     "ws": "^7.2.5"
   },

mon/pg_gen.js

@@ -195,7 +195,7 @@ async function generate_pool_pgs(state, global_config, pool_id, osd_tree, levels
     const folded_tree = make_hier_tree(global_config, folded.nodes.reduce((a, c) => { a[c.id] = c; return a; }, {}));
     const old_pg_count = prev_pgs.length;
     const optimize_cfg = {
-        osd_weights: folded.nodes.reduce((a, c) => { if (Number(c.id)) { a[c.id] = c.size; } return a; }, {}),
+        osd_weights: folded.nodes.reduce((a, c) => { if (/^\d+$/.exec(c.id) && c.size != null) { a[c.id] = c.size||0; } return a; }, {}),
         combinator: use_rules
             // new algorithm:
             ? new RuleCombinator(folded_tree, rules, pool_cfg.max_osd_combinations)

node-binding/package.json

@@ -1,6 +1,6 @@
 {
   "name": "vitastor",
-  "version": "3.0.2",
+  "version": "3.0.5",
   "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 = '3.0.2'
+VITASTOR_VERSION = '3.0.5'
 
 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.
@@ -291,6 +291,14 @@ 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."""
 
@@ -302,7 +310,7 @@ class VitastorDriver(driver.CloneableImageVD,
 
         LOG.debug("creating volume '%s'", vol_name)
 
-        self._create_image(vol_name, { 'size': size })
+        self._cli('create volume', 'create', vol_name, '--size', size)
 
         if volume.encryption_key_id:
             self._create_encrypted_volume(volume, volume.obj_context)
@@ -346,7 +354,7 @@ 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._create_snapshot(vol_name, vol_name+'@'+snap_name)
+        self._cli('create snapshot', 'snap-create', vol_name+'@'+snap_name)
 
     def snapshot_revert_use_temp_snapshot(self):
         """Disable the use of a temporary snapshot on revert."""
@@ -359,21 +367,8 @@ class VitastorDriver(driver.CloneableImageVD,
         snap_name = utils.convert_str(snapshot.name)
 
         # Delete the image and recreate it from the snapshot
-        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)
+        self._cli('delete image', 'rm', vol_name)
+        self._cli('recreate image', 'create', '--parent', vol_name+'@'+snap_name, vol_name)
 
     def delete_snapshot(self, snapshot):
         """Deletes a snapshot."""
@@ -381,15 +376,7 @@ class VitastorDriver(driver.CloneableImageVD,
         vol_name = utils.convert_str(snapshot.volume_name)
         snap_name = utils.convert_str(snapshot.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)
+        self._cli('remove snapshot', 'rm', vol_name+'@'+snap_name)
 
     def _child_count(self, parents):
         children = 0
@@ -427,13 +414,7 @@ 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
-            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'],
-            })
+            self._cli('create clone', 'create', '--parent', src_name, '--size', size, dest_name)
             return {}
 
         clone_snap = "%s@%s.clone_snap" % (src_name, dest_name)
@@ -446,15 +427,12 @@ class VitastorDriver(driver.CloneableImageVD,
             clone_snap = dest_name
             make_img = False
 
-        LOG.debug("creating layer '%s' under '%s'", clone_snap, src_name)
-        new_cfg = self._create_snapshot(src_name, clone_snap, True)
+        LOG.debug("creating snapshot '%s'", clone_snap)
+        self._cli('create base snapshot', 'snap-create', '--allow-existing', '1', clone_snap)
+
         if make_img:
             # Then create a clone from it
-            new_cfg = self._create_image(dest_name, {
-                'size': size,
-                'parent_id': new_cfg['parent_id'],
-                'parent_pool_id': new_cfg['parent_pool_id'],
-            })
+            self._cli('create clone', 'create', '--parent', clone_snap, '--size', size, dest_name)
 
         return {}
 
@@ -464,7 +442,8 @@ class VitastorDriver(driver.CloneableImageVD,
         vol_name = utils.convert_str(volume.name)
         snap_name = utils.convert_str(snapshot.name)
 
-        snap = self._get_image('volume-'+snapshot.volume_id+'@'+snap_name)
+        src_snap = 'volume-'+snapshot.volume_id+'@'+snap_name
+        snap = self._get_image(src_snap)
         if not snap:
             raise exception.SnapshotNotFound(snapshot_id = snap_name)
         snap_inode_id = int(resp['responses'][0]['kvs'][0]['value']['id'])
@@ -473,12 +452,8 @@ 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):
@@ -505,49 +480,7 @@ class VitastorDriver(driver.CloneableImageVD,
         """Deletes a logical volume."""
 
         vol_name = utils.convert_str(volume.name)
-
-        # 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 })
+        self._cli('delete volume', 'rm', '--matching', vol_name, vol_name+'@*', '--progress', '0')
 
     def retype(self, context, volume, new_type, diff, host):
         """Change extra type specifications for a volume."""
@@ -567,98 +500,6 @@ 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',
@@ -697,13 +538,9 @@ class VitastorDriver(driver.CloneableImageVD,
                     size = int(volume.size) * units.Gi
                     dest_name = utils.convert_str(volume.name)
                     # Find or create the base snapshot
-                    snap_cfg = self._create_snapshot(base_vol.name, base_vol.name+'@.clone_snap', True)
+                    self._cli('create base snapshot', 'create', '--allow-existing', '1', base_vol.name+'@.clone_snap')
                     # Then create a clone from it
-                    new_cfg = self._create_image(dest_name, {
-                        'size': size,
-                        'parent_id': snap_cfg['parent_id'],
-                        'parent_pool_id': snap_cfg['parent_pool_id'],
-                    })
+                    self._cli('create clone', 'create', dest_name, '--size', size, '--parent', base_vol.name+'@.clone_snap')
                     return ({}, True)
         return ({}, False)
 
@@ -770,26 +607,8 @@ class VitastorDriver(driver.CloneableImageVD,
     def extend_volume(self, volume, new_size):
         """Extend an existing volume."""
         vol_name = utils.convert_str(volume.name)
-        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
+        size = int(new_size) * units.Gi
+        self._cli('extend volume', 'modify', vol_name, '--resize', new_size)
         LOG.debug(
             "Extend volume from %(old_size)s GB to %(new_size)s GB.",
             {'old_size': volume.size, 'new_size': new_size}
@@ -862,28 +681,7 @@ class VitastorDriver(driver.CloneableImageVD,
         """
         from_name = self._get_existing_name(existing_ref)
         to_name = utils.convert_str(volume.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
+        self._cli('rename', 'modify', from_name, '--rename', to_name)
 
     def unmanage(self, volume):
         pass
@@ -956,7 +754,7 @@ 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._rename(from_name, to_name)
+        self._cli('rename', 'modify', from_name, '--rename', to_name)
 
     def unmanage_snapshot(self, snapshot):
         """Removes the specified snapshot from Cinder management."""

rpm/build-tarball.sh

@@ -21,7 +21,7 @@ rpmbuild -bp fio.spec
 cd $VITASTOR
 VER=$(grep ^Version: rpm/vitastor-$REL.spec | awk '{print $2}')
 rm -rf fio
-ln -s ~/rpmbuild/BUILD/fio*/ fio
+ln -s $(ls -d ~/rpmbuild/BUILD/fio*/ | grep -v SPECPARTS) fio
 sh copy-fio-includes.sh
 rm fio
 mv fio-copy fio

rpm/vitastor-el10.Dockerfile

@@ -0,0 +1,17 @@
+# Build packages for AlmaLinux 10 inside a container
+# cd ..
+# docker pull --platform=linux/amd64/v2 quay.io/almalinuxorg/almalinux:10
+# docker build -t vitastor-buildenv:el10 -f rpm/vitastor-el10.Dockerfile .
+# docker run -i --rm -v ./:/root/vitastor vitastor-buildenv:el10 /root/vitastor/rpm/vitastor-build.sh
+
+FROM quay.io/almalinuxorg/almalinux:10
+
+WORKDIR /root
+
+RUN sed -i 's/enabled=0/enabled=1/' /etc/yum.repos.d/*.repo
+RUN dnf -y install epel-release dnf-plugins-core
+RUN dnf -y install https://vitastor.io/rpms/centos/10/vitastor-release-1.0-1.el10.noarch.rpm
+RUN dnf -y install gcc-c++ gperftools-devel fio nodejs rpm-build jerasure-devel isa-l-devel gf-complete-devel rdma-core-devel cmake libnl3-devel c-ares-devel
+RUN dnf download --source fio
+RUN rpm --nomd5 -i fio*.src.rpm
+RUN cd ~/rpmbuild/SPECS && dnf builddep -y --spec fio.spec

rpm/vitastor-el10.spec

@@ -0,0 +1,199 @@
+Name:           vitastor
+Version:        3.0.5
+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-3.0.5.el10.tar.gz
+
+BuildRequires:  gperftools-devel
+BuildRequires:  gcc-c++
+BuildRequires:  nodejs >= 10
+BuildRequires:  jerasure-devel
+BuildRequires:  isa-l-devel
+BuildRequires:  gf-complete-devel
+BuildRequires:  rdma-core-devel
+BuildRequires:  cmake
+BuildRequires:  libnl3-devel
+BuildRequires:  c-ares-devel
+Requires:       vitastor-osd = %{version}-%{release}
+Requires:       vitastor-mon = %{version}-%{release}
+Requires:       vitastor-client = %{version}-%{release}
+Requires:       vitastor-client-devel = %{version}-%{release}
+Requires:       vitastor-fio = %{version}-%{release}
+
+%description
+Vitastor is a small, simple and fast clustered block storage (storage for VM drives),
+architecturally similar to Ceph which means strong consistency, primary-replication,
+symmetric clustering and automatic data distribution over any number of drives of any
+size with configurable redundancy (replication or erasure codes/XOR).
+
+
+%package -n vitastor-osd
+Summary:        Vitastor - OSD
+Requires:       vitastor-client = %{version}-%{release}
+Requires:       util-linux
+Requires:       parted
+
+
+%description -n vitastor-osd
+Vitastor object storage daemon, i.e. server program that stores data.
+
+
+%package -n vitastor-mon
+Summary:        Vitastor - monitor
+Requires:       nodejs >= 10
+Requires:       lpsolve
+
+
+%description -n vitastor-mon
+Vitastor monitor, i.e. server program responsible for watching cluster state and
+scheduling cluster-level operations.
+
+
+%package -n vitastor-client
+Summary:        Vitastor - client
+
+
+%description -n vitastor-client
+Vitastor client library and command-line interface.
+
+
+%package -n vitastor-client-devel
+Summary:        Vitastor - development files
+Group:          Development/Libraries
+Requires:       vitastor-client = %{version}-%{release}
+
+
+%description -n vitastor-client-devel
+Vitastor library headers for development.
+
+
+%package -n vitastor-fio
+Summary:        Vitastor - fio drivers
+Group:          Development/Libraries
+Requires:       vitastor-client = %{version}-%{release}
+Requires:       fio = 3.36-5.el10
+
+
+%description -n vitastor-fio
+Vitastor fio drivers for benchmarking.
+
+
+%package -n vitastor-opennebula
+Summary:        Vitastor for OpenNebula
+Group:          Development/Libraries
+Requires:       vitastor-client
+Requires:       jq
+Requires:       python3-lxml
+Requires:       patch
+Requires:       qemu-kvm-block-vitastor
+
+
+%description -n vitastor-opennebula
+Vitastor storage plugin for OpenNebula.
+
+
+%prep
+%setup -q
+
+
+%build
+%cmake
+%cmake_build
+
+
+%install
+rm -rf $RPM_BUILD_ROOT
+%cmake_install
+cd mon
+npm install --production
+cd ..
+mkdir -p %buildroot/usr/lib/vitastor
+cp -r mon %buildroot/usr/lib/vitastor
+mv %buildroot/usr/lib/vitastor/mon/scripts/make-etcd %buildroot/usr/lib/vitastor/mon/
+mkdir -p %buildroot/lib/systemd/system
+cp mon/scripts/vitastor.target mon/scripts/vitastor-mon.service mon/scripts/vitastor-osd@.service %buildroot/lib/systemd/system
+mkdir -p %buildroot/lib/udev/rules.d
+cp mon/scripts/90-vitastor.rules %buildroot/lib/udev/rules.d
+mkdir -p %buildroot/var/lib/one
+cp -r opennebula/remotes %buildroot/var/lib/one
+cp opennebula/install.sh %buildroot/var/lib/one/remotes/datastore/vitastor/
+mkdir -p %buildroot/etc/
+cp -r opennebula/sudoers.d %buildroot/etc/
+
+
+%files
+%doc GPL-2.0.txt VNPL-1.1.txt README.md README-ru.md
+
+
+%files -n vitastor-osd
+%_bindir/vitastor-osd
+%_bindir/vitastor-disk
+%_bindir/vitastor-dump-journal
+/lib/systemd/system/vitastor-osd@.service
+/lib/systemd/system/vitastor.target
+/lib/udev/rules.d/90-vitastor.rules
+
+
+%pre -n vitastor-osd
+groupadd -r -f vitastor 2>/dev/null ||:
+useradd -r -g vitastor -s /sbin/nologin -c "Vitastor daemons" -M -d /nonexistent vitastor 2>/dev/null ||:
+install -o vitastor -g vitastor -d /var/log/vitastor
+mkdir -p /etc/vitastor
+
+
+%files -n vitastor-mon
+/usr/lib/vitastor/mon
+/lib/systemd/system/vitastor-mon.service
+
+
+%pre -n vitastor-mon
+groupadd -r -f vitastor 2>/dev/null ||:
+useradd -r -g vitastor -s /sbin/nologin -c "Vitastor daemons" -M -d /nonexistent vitastor 2>/dev/null ||:
+mkdir -p /etc/vitastor
+mkdir -p /var/lib/vitastor
+chown vitastor:vitastor /var/lib/vitastor
+
+
+%files -n vitastor-client
+%_bindir/vitastor-nbd
+%_bindir/vitastor-ublk
+%_bindir/vitastor-nfs
+%_bindir/vitastor-cli
+%_bindir/vitastor-rm
+%_bindir/vitastor-kv
+%_bindir/vitastor-kv-stress
+%_bindir/vita
+%_libdir/libvitastor_client.so*
+%_libdir/libvitastor_kv.so*
+
+
+%files -n vitastor-client-devel
+%_includedir/vitastor_c.h
+%_includedir/vitastor_kv.h
+%_libdir/pkgconfig
+
+
+%files -n vitastor-fio
+%_libdir/libfio_vitastor.so
+%_libdir/libfio_vitastor_blk.so
+%_libdir/libfio_vitastor_sec.so
+
+
+%files -n vitastor-opennebula
+/var/lib/one
+/etc/sudoers.d/opennebula-vitastor
+
+
+%triggerin -n vitastor-opennebula -- opennebula
+[ $2 = 0 ] || exit 0
+/var/lib/one/remotes/datastore/vitastor/install.sh
+
+
+# Turn off the brp-python-bytecompile script
+%global __os_install_post %(echo '%{__os_install_post}' | sed -e 's!/usr/lib[^[:space:]]*/brp-python-bytecompile[[:space:]].*$!!g')
+
+
+%changelog

rpm/vitastor-el7.Dockerfile

@@ -15,7 +15,7 @@ RUN yum -y --enablerepo=extras install centos-release-scl epel-release yum-utils
 RUN perl -i -pe 's!mirrorlist=!#mirrorlist=!s; s!#\s*baseurl=http://mirror.centos.org!baseurl=http://vault.centos.org!' /etc/yum.repos.d/CentOS-SCLo-scl*.repo
 RUN yum -y install https://vitastor.io/rpms/centos/7/vitastor-release-1.0-1.el7.noarch.rpm
 RUN yum -y install devtoolset-9-gcc-c++ devtoolset-9-libatomic-devel gcc make cmake gperftools-devel \
-    fio rh-nodejs12 jerasure-devel libisa-l-devel gf-complete-devel rdma-core-devel libnl3-devel
+    fio rh-nodejs12 jerasure-devel libisa-l-devel gf-complete-devel rdma-core-devel libnl3-devel c-ares-devel
 RUN yumdownloader --disablerepo=centos-sclo-rh --source fio
 RUN rpm --nomd5 -i fio*.src.rpm
 RUN rm -f /etc/yum.repos.d/CentOS-Media.repo

rpm/vitastor-el7.spec

@@ -1,11 +1,11 @@
 Name:           vitastor
-Version:        3.0.2
+Version:        3.0.5
 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-3.0.2.el7.tar.gz
+Source0:        vitastor-3.0.5.el7.tar.gz
 
 BuildRequires:  gperftools-devel
 BuildRequires:  devtoolset-9-gcc-c++
@@ -17,6 +17,7 @@ BuildRequires:  gf-complete-devel
 BuildRequires:  rdma-core-devel
 BuildRequires:  cmake3
 BuildRequires:  libnl3-devel
+BuildRequires:  c-ares-devel
 Requires:       vitastor-osd = %{version}-%{release}
 Requires:       vitastor-mon = %{version}-%{release}
 Requires:       vitastor-client = %{version}-%{release}

rpm/vitastor-el8.Dockerfile

@@ -13,7 +13,7 @@ RUN dnf -y install centos-release-advanced-virtualization epel-release dnf-plugi
 RUN sed -i 's/^mirrorlist=/#mirrorlist=/; s!#baseurl=.*!baseurl=http://vault.centos.org/centos/8.4.2105/virt/$basearch/$avdir/!; s!^baseurl=.*Source/.*!baseurl=http://vault.centos.org/centos/8.4.2105/virt/Source/advanced-virtualization/!' /etc/yum.repos.d/CentOS-Advanced-Virtualization.repo
 RUN yum -y install https://vitastor.io/rpms/centos/8/vitastor-release-1.0-1.el8.noarch.rpm
 RUN dnf -y install gcc-toolset-9 gcc-toolset-9-gcc-c++ gperftools-devel \
-    fio nodejs rpm-build jerasure-devel libisa-l-devel gf-complete-devel libibverbs-devel libarchive cmake libnl3-devel
+    fio nodejs rpm-build jerasure-devel libisa-l-devel gf-complete-devel libibverbs-devel libarchive cmake libnl3-devel c-ares-devel
 RUN dnf download --source fio
 RUN rpm --nomd5 -i fio*.src.rpm
 RUN cd ~/rpmbuild/SPECS && dnf builddep -y --enablerepo=powertools --spec fio.spec

rpm/vitastor-el8.spec

@@ -1,11 +1,11 @@
 Name:           vitastor
-Version:        3.0.2
+Version:        3.0.5
 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-3.0.2.el8.tar.gz
+Source0:        vitastor-3.0.5.el8.tar.gz
 
 BuildRequires:  gperftools-devel
 BuildRequires:  gcc-toolset-9-gcc-c++
@@ -16,6 +16,7 @@ BuildRequires:  gf-complete-devel
 BuildRequires:  rdma-core-devel
 BuildRequires:  cmake
 BuildRequires:  libnl3-devel
+BuildRequires:  c-ares-devel
 Requires:       vitastor-osd = %{version}-%{release}
 Requires:       vitastor-mon = %{version}-%{release}
 Requires:       vitastor-client = %{version}-%{release}

rpm/vitastor-el9.Dockerfile

@@ -10,7 +10,7 @@ WORKDIR /root
 RUN sed -i 's/enabled=0/enabled=1/' /etc/yum.repos.d/*.repo
 RUN dnf -y install epel-release dnf-plugins-core
 RUN dnf -y install https://vitastor.io/rpms/centos/9/vitastor-release-1.0-1.el9.noarch.rpm
-RUN dnf -y install gcc-c++ gperftools-devel fio nodejs rpm-build jerasure-devel libisa-l-devel gf-complete-devel rdma-core-devel libarchive cmake libnl3-devel
+RUN dnf -y install gcc-c++ gperftools-devel fio nodejs rpm-build jerasure-devel libisa-l-devel gf-complete-devel rdma-core-devel libarchive cmake libnl3-devel c-ares-devel
 RUN dnf download --source fio
 RUN rpm --nomd5 -i fio*.src.rpm
 RUN cd ~/rpmbuild/SPECS && dnf builddep -y --spec fio.spec

rpm/vitastor-el9.spec

@@ -1,11 +1,11 @@
 Name:           vitastor
-Version:        3.0.2
+Version:        3.0.5
 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-3.0.2.el9.tar.gz
+Source0:        vitastor-3.0.5.el9.tar.gz
 
 BuildRequires:  gperftools-devel
 BuildRequires:  gcc-c++
@@ -16,6 +16,7 @@ BuildRequires:  gf-complete-devel
 BuildRequires:  rdma-core-devel
 BuildRequires:  cmake
 BuildRequires:  libnl3-devel
+BuildRequires:  c-ares-devel
 Requires:       vitastor-osd = %{version}-%{release}
 Requires:       vitastor-mon = %{version}-%{release}
 Requires:       vitastor-client = %{version}-%{release}

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="3.0.2")
+add_definitions(-DVITASTOR_VERSION="3.0.5")
 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 +75,14 @@ if (RDMACM_LIBRARIES)
 	add_definitions(-DWITH_RDMACM)
 endif (RDMACM_LIBRARIES)
 
+find_package(OpenSSL REQUIRED)
+if (OPENSSL_FOUND)
+	add_definitions(-DWITH_OPENSSL)
+endif (OPENSSL_FOUND)
+
+pkg_check_modules(CARES REQUIRED libcares)
+include_directories(${CARES_INCLUDE_DIRS})
+
 if (${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.h

@@ -187,7 +187,6 @@ public:
     // MUST be called only when nobody makes any modifications to the DB for this pool
     virtual void* reshard_start(pool_id_t pool, uint32_t pg_count, uint32_t pg_stripe_size, uint64_t chunk_limit) = 0;
     virtual bool reshard_continue(void *reshard_state, uint64_t chunk_limit) = 0;
-    virtual void reshard_abort(void *reshard_state) = 0;
 
     // Event loop
     virtual void loop() = 0;

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() &&
@@ -173,9 +177,7 @@ void blockstore_disk_t::parse_config(std::map<std::string, std::string> & config
     }
     if (data_block_size / 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);
+        throw std::runtime_error("Data block size must be at least bitmap_granularity*8");
     }
     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;
-    uint32_t data_block_size;
+    uint64_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
-    uint32_t journal_block_size = 4096;
+    uint64_t journal_block_size = 4096;
     // Metadata block size - minimum_io_size of the metadata device is the best choice
-    uint32_t meta_block_size = 4096;
+    uint64_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;
@@ -301,12 +311,13 @@ 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 %u bytes). ",
+                fprintf(stderr, "Error: entry is too large in metadata block %u at %u (%u > max %ju 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
@@ -331,7 +342,19 @@ corrupted_block:
                     block_num, block_offset, wr->size, sizeof(heap_entry_t));
                 goto corrupted_block;
             }
-            wr->entry_type &= ~BS_HEAP_GARBAGE;
+            if (wr->is_garbage())
+            {
+                // Garbage collection is only performed when writing new entries into the block
+                // because it needs a fake LSN and modified blocks require consecutive modified LSNs
+                // That's why garbage entries may persist on disk
+                if (log_level > 5)
+                {
+                    fprintf(stderr, "Notice: skipping garbage entry %jx:%jx v%ju l%ju in metadata block %u at %u\n",
+                        wr->inode, wr->stripe, wr->version, wr->lsn, block_num, block_offset);
+                }
+                block_offset += wr->size;
+                continue;
+            }
             if ((wr->entry_type & BS_HEAP_TYPE) < BS_HEAP_BIG_WRITE ||
                 (wr->entry_type & BS_HEAP_TYPE) > BS_HEAP_ROLLBACK ||
                 (wr->entry_type & ~(BS_HEAP_TYPE|BS_HEAP_STABLE)) ||
@@ -345,6 +368,7 @@ corrupted_object:
                 if (allow_corrupted)
                 {
                     fprintf(stderr, "Entry is corrupted, skipping\n");
+                    recheck_modified_blocks.insert(block_num);
                     block_offset += wr->size;
                     continue;
                 }
@@ -360,7 +384,7 @@ corrupted_object:
             {
                 // Small writes require accessing offset & len to calculate correct length,
                 // so require at least sizeof(heap_small_write_t) for them
-                fprintf(stderr, "Error: entry %jx:%jx v%ju has invalid size in metadata block %u at %u (%u < min %zu bytes). Metadata is corrupted, aborting\n",
+                fprintf(stderr, "Error: entry %jx:%jx v%ju has invalid size in metadata block %u at %u (%u < min %zu bytes)\n",
                     wr->inode, wr->stripe, wr->version, block_num, block_offset, wr->size, sizeof(heap_small_write_t));
                 goto corrupted_object;
             }
@@ -371,12 +395,12 @@ 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 in metadata block %u at %u is corrupt (crc32c mismatch: expected %08x, got %08x). Metadata is corrupted, aborting\n",
-                    wr->inode, wr->stripe, wr->version,
-                    block_num, block_offset, expected_crc32c, wr->crc32c);
+                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_checksum, wr->checksum);
                 goto corrupted_object;
             }
             // Verify offset & len
@@ -385,7 +409,7 @@ corrupted_object:
                 wr->small().offset % dsk->bitmap_granularity ||
                 wr->small().len % dsk->bitmap_granularity))
             {
-                fprintf(stderr, "Error: %s entry %jx:%jx v%ju has invalid offset/length: %u/%u. Metadata is incompatible with current parameters, aborting\n",
+                fprintf(stderr, "Error: %s entry %jx:%jx v%ju has invalid offset/length: %u/%u. Metadata is incompatible with current parameters. ",
                     wr->type() == BS_HEAP_SMALL_WRITE ? "small_write" : "intent_write",
                     wr->inode, wr->stripe, wr->version, wr->small().offset, wr->small().len);
                 goto corrupted_object;
@@ -395,7 +419,7 @@ corrupted_object:
                 wr->big_intent().offset % dsk->bitmap_granularity ||
                 wr->big_intent().len % dsk->bitmap_granularity))
             {
-                fprintf(stderr, "Error: big_intent entry %jx:%jx v%ju has invalid offset/length: %u/%u. Metadata is incompatible with current parameters, aborting\n",
+                fprintf(stderr, "Error: big_intent entry %jx:%jx v%ju has invalid offset/length: %u/%u. Metadata is incompatible with current parameters. ",
                     wr->inode, wr->stripe, wr->version, wr->big_intent().offset, wr->big_intent().len);
                 goto corrupted_object;
             }
@@ -422,7 +446,7 @@ int blockstore_heap_t::load_blocks(uint64_t disk_offset, uint64_t size, uint8_t
             next_lsn = wr->lsn;
         }
         entries_loaded++;
-        insert_list_item(li);
+        loaded_list_items.push_back(li);
         modify_alloc(block_num, [&](heap_block_info_t & inf)
         {
             if (!inf.entries.size())
@@ -458,19 +482,22 @@ bool blockstore_heap_t::validate_object(heap_entry_t *obj)
         next_wr = wr;
         if (wr->type() == BS_HEAP_ROLLBACK)
         {
-            if (commit_wr && wr->version > commit_wr->version)
-            {
-                // rollback may not come before commit with a smaller version
-                fprintf(stderr, "Error: rollback entry %jx:%jx v%ju l%ju comes before a commit entry v%ju l%ju\n",
-                    wr->inode, wr->stripe, wr->version, wr->lsn, commit_wr->version, commit_wr->lsn);
-                return false;
-            }
             rollback_wr = wr;
             continue;
         }
         if (wr->type() == BS_HEAP_COMMIT)
         {
-            commit_wr = wr;
+            if (commit_wr && wr->version > commit_wr->version)
+            {
+                // commit may not come before commit with a smaller version
+                fprintf(stderr, "Error: commit entry %jx:%jx v%ju l%ju comes before a commit entry v%ju l%ju\n",
+                    wr->inode, wr->stripe, wr->version, wr->lsn, commit_wr->version, commit_wr->lsn);
+                return false;
+            }
+            if (!commit_wr)
+            {
+                commit_wr = wr;
+            }
             continue;
         }
         if (wr->entry_type & BS_HEAP_STABLE)
@@ -523,6 +550,23 @@ bool blockstore_heap_t::validate_object(heap_entry_t *obj)
     return true;
 }
 
+void blockstore_heap_t::finish_load()
+{
+    if (loaded_list_items.size())
+    {
+        // Sort everything and load in correct order
+        std::sort(loaded_list_items.begin(), loaded_list_items.end(), [this](const heap_list_item_t* a, const heap_list_item_t* b)
+        {
+            return a->entry.lsn < b->entry.lsn;
+        });
+        for (auto & li: loaded_list_items)
+        {
+            insert_list_item(li);
+        }
+        loaded_list_items.clear();
+    }
+}
+
 void blockstore_heap_t::fill_recheck_queue()
 {
     for (auto & pgp: block_index)
@@ -709,6 +753,7 @@ bool blockstore_heap_t::recheck_small_writes(std::function<void(bool is_data, ui
     }
     if (!recheck_queue_filled)
     {
+        finish_load();
         fill_recheck_queue();
         recheck_queue_filled = true;
     }
@@ -800,7 +845,7 @@ std::vector<uint32_t> blockstore_heap_t::get_recheck_modified_blocks()
     return modified;
 }
 
-int blockstore_heap_t::finish_load(bool allow_corrupted)
+int blockstore_heap_t::finish_recheck()
 {
     if (!marked_used_blocks)
     {
@@ -837,14 +882,17 @@ bool blockstore_heap_t::calc_checksums(heap_entry_t *wr, uint8_t *data, bool set
         {
             return true;
         }
-        uint32_t len = 0;
         if (wr->type() == BS_HEAP_SMALL_WRITE || wr->type() == BS_HEAP_INTENT_WRITE)
             len = wr->small().len;
         else if (wr->type() == BS_HEAP_BIG_INTENT)
             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;
@@ -854,13 +902,14 @@ bool blockstore_heap_t::calc_checksums(heap_entry_t *wr, uint8_t *data, bool set
     }
     if (wr->type() == BS_HEAP_BIG_WRITE)
     {
+        assert(offset != UINT32_MAX && len != UINT32_MAX);
         return calc_block_checksums((uint32_t*)(wr->get_checksums(this) + offset/dsk->csum_block_size * (dsk->data_csum_type & 0xFF)),
             data, wr->get_int_bitmap(this), offset, offset+len, set, NULL);
     }
     if (wr->type() == BS_HEAP_BIG_INTENT)
     {
         auto & bi = wr->big_intent();
-        return calc_block_checksums((uint32_t*)(wr->get_checksums(this) + offset/dsk->csum_block_size * (dsk->data_csum_type & 0xFF)),
+        return calc_block_checksums((uint32_t*)(wr->get_checksums(this) + bi.offset/dsk->csum_block_size * (dsk->data_csum_type & 0xFF)),
             data, wr->get_int_bitmap(this), bi.offset, bi.offset+bi.len, set, NULL);
     }
     assert(wr->type() == BS_HEAP_SMALL_WRITE || wr->type() == BS_HEAP_INTENT_WRITE);
@@ -893,11 +942,26 @@ 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;
@@ -914,42 +978,89 @@ 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;
 }
 
@@ -1077,16 +1188,6 @@ bool blockstore_heap_t::reshard_check(pool_id_t pool, uint32_t pg_count, uint32_
         set_it->second.pg_stripe_size == pg_stripe_size);
 }
 
-void blockstore_heap_t::reshard_abort(void* reshard_state)
-{
-    heap_reshard_state_t *st = (heap_reshard_state_t*)reshard_state;
-    for (auto sh_it = st->old_shards.begin(); sh_it != st->old_shards.end(); sh_it++)
-    {
-        block_index[sh_it->first] = std::move(sh_it->second);
-    }
-    delete st;
-}
-
 heap_entry_t *blockstore_heap_t::lock_and_read_entry(object_id oid)
 {
     auto obj = read_entry(oid);
@@ -1099,27 +1200,6 @@ heap_entry_t *blockstore_heap_t::lock_and_read_entry(object_id oid)
     return obj;
 }
 
-heap_entry_t *blockstore_heap_t::read_locked_entry(object_id oid, uint64_t lsn)
-{
-    auto obj = read_entry(oid);
-    assert(obj);
-    for (auto wr = obj; wr; wr = prev(wr))
-    {
-        if (wr->is_overwrite())
-        {
-            if (lsn == wr->lsn)
-            {
-                return obj;
-            }
-            else
-            {
-                obj = prev(wr);
-            }
-        }
-    }
-    return NULL;
-}
-
 bool blockstore_heap_t::unlock_entry(object_id oid)
 {
     auto mvcc_it = object_mvcc.find(oid);
@@ -1156,6 +1236,35 @@ heap_entry_t *blockstore_heap_t::read_entry(object_id oid)
     return &li->entry;
 }
 
+void blockstore_heap_t::gc_block(heap_block_info_t & inf)
+{
+    if (inf.has_garbage)
+    {
+        size_t i = 0, j = 0;
+        for (; i < inf.entries.size(); i++)
+        {
+            if (inf.entries[i]->entry.is_garbage())
+            {
+                // old entry invalidated by a newer one, mark it as freeable on block write
+                // assign a 'virtual' LSN to track GC completion
+                assert(!inf.mod_lsn_to || inf.mod_lsn_to == next_lsn);
+                uint64_t gc_lsn = ++next_lsn;
+                inf.mod_lsn = inf.mod_lsn ? inf.mod_lsn : gc_lsn;
+                inf.mod_lsn_to = gc_lsn;
+                push_inflight_lsn(gc_lsn, &inf.entries[i]->entry, HEAP_INFLIGHT_GC);
+            }
+            else
+            {
+                if (j != i)
+                    inf.entries[j] = inf.entries[i];
+                j++;
+            }
+        }
+        inf.entries.resize(j);
+        inf.has_garbage = false;
+    }
+}
+
 int blockstore_heap_t::allocate_entry(uint32_t entry_size, uint32_t *block_num, bool allow_last_free)
 {
     if (last_allocated_block != UINT32_MAX)
@@ -1222,31 +1331,7 @@ int blockstore_heap_t::allocate_entry(uint32_t entry_size, uint32_t *block_num,
     }
     // Write into the same block
     auto & inf = block_info.at(last_allocated_block);
-    if (inf.has_garbage)
-    {
-        size_t i = 0, j = 0;
-        for (; i < inf.entries.size(); i++)
-        {
-            if (inf.entries[i]->entry.is_garbage())
-            {
-                // old entry invalidated by a newer one, mark it as freeable on block write
-                // assign a 'virtual' LSN to track GC completion
-                assert(!inf.mod_lsn_to || inf.mod_lsn_to == next_lsn);
-                uint64_t gc_lsn = ++next_lsn;
-                inf.mod_lsn = inf.mod_lsn ? inf.mod_lsn : gc_lsn;
-                inf.mod_lsn_to = gc_lsn;
-                push_inflight_lsn(gc_lsn, &inf.entries[i]->entry, HEAP_INFLIGHT_GC);
-            }
-            else
-            {
-                if (j != i)
-                    inf.entries[j] = inf.entries[i];
-                j++;
-            }
-        }
-        inf.entries.resize(j);
-        inf.has_garbage = false;
-    }
+    gc_block(inf);
     *block_num = last_allocated_block;
     modify_alloc(last_allocated_block, [&](heap_block_info_t & inf)
     {
@@ -1327,7 +1412,7 @@ 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;
 }
 
@@ -1347,7 +1432,6 @@ int blockstore_heap_t::add_small_write(object_id oid, heap_entry_t **obj_ptr, ui
     // Small writes are written in parallel with buffered data so they require explicit_complete
     return add_entry(wr_size, modified_block, false, true, [&](heap_entry_t *wr)
     {
-        printf("add_small_write t%u %lx:%lx l%lu v%lu %u +%u loc:%lx\n", type, oid.inode, oid.stripe, wr->lsn, version, offset, len, location);
         wr->entry_type = type;
         wr->inode = oid.inode;
         wr->stripe = oid.stripe;
@@ -1377,7 +1461,6 @@ int blockstore_heap_t::add_big_write(object_id oid, heap_entry_t *old_head, bool
     // Big writes are written after writing data so they don't require explicit_complete
     return add_entry(wr_size, modified_block, false, false, [&](heap_entry_t *wr)
     {
-        printf("add_big_write %lx:%lx l%lu v%lu loc:%lx\n", oid.inode, oid.stripe, wr->lsn, version, location);
         wr->entry_type = BS_HEAP_BIG_WRITE | (stable ? BS_HEAP_STABLE : 0);
         wr->inode = oid.inode;
         wr->stripe = oid.stripe;
@@ -1389,7 +1472,7 @@ 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);
@@ -1404,7 +1487,6 @@ int blockstore_heap_t::add_redirect_intent(object_id oid, heap_entry_t **obj_ptr
     // Big-redirect intents, just like regular big writes, are written after writing data so they don't require explicit_complete
     return add_entry(wr_size, modified_block, false, false, [&](heap_entry_t *wr)
     {
-        printf("add_redir_intent %lx:%lx l%lu v%lu %u +%u loc:%lx\n", oid.inode, oid.stripe, wr->lsn, version, offset, len, location);
         wr->entry_type = BS_HEAP_BIG_INTENT|BS_HEAP_STABLE;
         wr->inode = oid.inode;
         wr->stripe = oid.stripe;
@@ -1419,9 +1501,9 @@ 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, offset, len);
+        calc_checksums(wr, (uint8_t*)data, true);
         *obj_ptr = wr;
     });
 }
@@ -1441,7 +1523,6 @@ int blockstore_heap_t::add_big_intent(object_id oid, heap_entry_t **obj_ptr, uin
     // Big intents are written before writing data so they require explicit_complete
     return add_entry(wr_size, modified_block, false, true, [&](heap_entry_t *wr)
     {
-        printf("add_big_intent %lx:%lx l%lu v%lu %u +%u loc:%lx\n", oid.inode, oid.stripe, wr->lsn, version, offset, len, obj->big_location(this));
         wr->entry_type = BS_HEAP_BIG_INTENT | BS_HEAP_STABLE;
         wr->inode = oid.inode;
         wr->stripe = oid.stripe;
@@ -1458,14 +1539,14 @@ 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, dsk->clean_entry_bitmap_size);
+                memcpy(wr->get_checksums(this), checksums, get_csum_size(wr));
             else
             {
-                memcpy(wr->get_checksums(this), obj->get_checksums(this), dsk->clean_entry_bitmap_size);
-                calc_checksums(wr, (uint8_t*)data, true, offset, len);
+                memcpy(wr->get_checksums(this), obj->get_checksums(this), get_csum_size(wr));
+                calc_checksums(wr, (uint8_t*)data, true);
             }
         }
         else
@@ -1497,7 +1578,6 @@ int blockstore_heap_t::add_compact(heap_entry_t *obj, uint64_t compact_version,
     // Compaction entry is added after copying data so it doesn't require explicit_complete
     return add_entry(wr_size, modified_block, true, false, [&](heap_entry_t *new_wr)
     {
-        printf("add_compact %lx:%lx l%lu v%lu loc:%lx\n", obj->inode, obj->stripe, compact_lsn, compact_version, compact_location);
         new_wr->entry_type = BS_HEAP_BIG_WRITE|BS_HEAP_STABLE;
         new_wr->inode = obj->inode;
         new_wr->stripe = obj->stripe;
@@ -1506,7 +1586,7 @@ 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));
     });
 }
@@ -1587,7 +1667,7 @@ int blockstore_heap_t::add_commit(heap_entry_t *obj, uint64_t version, uint32_t
     }
     if (!uncommitted)
     {
-        return EBUSY;
+        return 0;
     }
     return add_simple(obj, version, modified_block, BS_HEAP_COMMIT);
 }
@@ -1597,23 +1677,32 @@ int blockstore_heap_t::add_rollback(heap_entry_t *obj, uint64_t version, uint32_
     heap_entry_t *wr = obj;
     bool found_uncommitted = false;
     uint64_t commit_version = 0;
-    while (wr && !wr->is_overwrite())
+    uint64_t rollback_version = UINT64_MAX;
+    while (wr)
     {
         if (wr->type() == BS_HEAP_ROLLBACK)
         {
-            auto rollback_version = wr->version;
-            wr = prev(wr);
-            while (wr->version > rollback_version)
+            if (wr->version <= version)
             {
-                assert(!(wr->entry_type & BS_HEAP_STABLE));
-                wr = prev(wr);
+                // All previous writes are already rolled back, stop
+                break;
             }
+            rollback_version = wr->version;
+            wr = prev(wr);
             continue;
         }
         if (wr->type() == BS_HEAP_COMMIT)
         {
             if (commit_version < wr->version)
+            {
                 commit_version = wr->version;
+            }
+            wr = prev(wr);
+            continue;
+        }
+        if (wr->version > rollback_version)
+        {
+            // Already rolled back, skip
             wr = prev(wr);
             continue;
         }
@@ -1624,14 +1713,10 @@ int blockstore_heap_t::add_rollback(heap_entry_t *obj, uint64_t version, uint32_
             {
                 return EBUSY;
             }
-            else if (wr->version == version)
+            else
             {
                 break;
             }
-            else if (wr->version < version)
-            {
-                return ENOENT;
-            }
         }
         else if (wr->version > version)
         {
@@ -1812,9 +1897,9 @@ void blockstore_heap_t::iterate_with_stable(heap_entry_t *obj, uint64_t max_lsn,
         }
         else
         {
-            // 1) 1 2 3 ROLLBACK(2) COMMIT(3) -> impossible
+            // 1) 1 2 3 ROLLBACK(2) COMMIT(3) -> 3 is unstable
             // 2) 1 2 3 4 ROLLBACK(3) COMMIT(2) -> OK
-            // 3) 1 2 3 ROLLBACK(2) 3 COMMIT(3) -> first 3 shouldn't be treated as stable
+            // 3) 1 2 3 ROLLBACK(2) 3 COMMIT(3) -> first 3 is unstable
             // 4) 1 2 3 COMMIT(3) ROLLBACK(2) -> impossible
             //    I.e. a rollback always has version >= previous commit
             // 5) 1 2 3 4 5 ROLLBACK(4) 5 ROLLBACK(3)
@@ -1971,24 +2056,21 @@ int blockstore_heap_t::list_objects(uint32_t pg_num, object_id min_oid, object_i
                 return;
             }
             uint64_t stable_version = 0;
-            auto first_wr = obj;
-            for (auto wr = first_wr; wr; wr = prev(wr))
+            iterate_with_stable(obj, UINT64_MAX, [&](heap_entry_t* wr, bool stable)
             {
-                if ((wr->entry_type & BS_HEAP_STABLE) || wr->type() == BS_HEAP_COMMIT || wr->type() == BS_HEAP_ROLLBACK)
+                if (stable)
                 {
                     stable_version = wr->version;
-                    break;
+                    return false;
                 }
-                else
+                if (unstable_size >= 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_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 };
+                return true;
+            });
             if (stable_version)
             {
                 if (res_size >= res_alloc)
@@ -2050,7 +2132,13 @@ void blockstore_heap_t::free_data(inode_t inode, uint64_t location)
         inode = (INODE_POOL(inode) << POOL_ID_BITS);
     assert(data_alloc->get(location / dsk->data_block_size));
     data_alloc->set(location / dsk->data_block_size, false);
-    inode_space_stats[inode] -= dsk->data_block_size;
+    auto sp_it = inode_space_stats.find(inode);
+    if (sp_it != inode_space_stats.end())
+    {
+        sp_it->second -= dsk->data_block_size;
+        if (sp_it->second == 0)
+            inode_space_stats.erase(sp_it);
+    }
     data_used_space -= dsk->data_block_size;
 }
 
@@ -2279,7 +2367,9 @@ 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;
@@ -2333,7 +2423,7 @@ void blockstore_heap_t::set_no_inode_stats(const std::vector<uint64_t> & pool_id
     {
         // Recalculate if changed
         if (ps.second.no_inode_stats == 2 || ps.second.no_inode_stats == 1)
-            recalc_inode_space_stats(ps.first, ps.second.no_inode_stats == 1);
+            recalc_inode_space_stats(ps.first, ps.second.no_inode_stats == 2);
         ps.second.no_inode_stats &= 1;
     }
 }
@@ -2344,8 +2434,8 @@ void blockstore_heap_t::recalc_inode_space_stats(uint64_t pool_id, bool per_inod
     auto sp_begin = inode_space_stats.lower_bound((pool_id << (64-POOL_ID_BITS)));
     auto sp_end = inode_space_stats.lower_bound(((pool_id+1) << (64-POOL_ID_BITS)));
     inode_space_stats.erase(sp_begin, sp_end);
-    uint32_t pg_count = ps.pg_count ? ps.pg_count : 1;
-    for (uint32_t pg_num = 1; pg_num <= pg_count; pg_num++)
+    uint32_t pg_count = ps.pg_count;
+    for (uint32_t pg_num = pg_count ? 1 : 0; pg_num <= pg_count; pg_num++)
     {
         auto & pg_idx = block_index[(pool_id << (64-POOL_ID_BITS)) | pg_num];
         for (auto & ip: pg_idx)
@@ -2383,7 +2473,10 @@ 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
@@ -200,6 +201,7 @@ class blockstore_heap_t
 
     bool marked_used_blocks = false;
     bool recheck_queue_filled = false;
+    std::vector<heap_list_item_t*> loaded_list_items;
     std::set<uint32_t> recheck_modified_blocks;
     std::deque<heap_entry_t*> recheck_queue;
     int recheck_in_progress = 0;
@@ -215,6 +217,7 @@ class blockstore_heap_t
     void defragment_block(uint32_t block_num);
     void reshard_add(heap_reshard_state_t *st, heap_list_item_t *li);
 
+    void gc_block(heap_block_info_t & inf);
     int allocate_entry(uint32_t entry_size, uint32_t *block_num, bool allow_last_free);
     void insert_list_item(heap_list_item_t *li);
     int add_entry(uint32_t wr_size, uint32_t *modified_block, bool allow_last_free,
@@ -237,31 +240,29 @@ public:
         std::function<void(uint32_t, uint32_t, uint8_t*)> handle_block);
     int load_blocks(uint64_t disk_offset, uint64_t size, uint8_t *buf,
         bool allow_corrupted, uint64_t &entries_loaded);
-    // finish loading
-    int finish_load(bool allow_corrupted = false);
+    // finish loading - should be called after load_blocks
+    void finish_load();
     // get blocks which are modified during loading and should be written to the disk
     // before finishing initialization if not R/O
     std::vector<uint32_t> get_recheck_modified_blocks();
     // recheck small write data after reading the database from disk
     bool recheck_small_writes(std::function<void(bool is_data, uint64_t offset, uint64_t len, uint8_t* buf, std::function<void()>)> read_buffer, int queue_depth);
+    int finish_recheck();
     // reshard database according to the pool's PG count
     void* reshard_start(pool_id_t pool, uint32_t pg_count, uint32_t pg_stripe_size, uint64_t chunk_limit);
     bool reshard_continue(void* reshard_state, uint64_t chunk_limit);
     bool reshard_check(pool_id_t pool, uint32_t pg_count, uint32_t pg_stripe_size);
-    void reshard_abort(void* reshard_state);
     void set_no_inode_stats(const std::vector<uint64_t> & pool_ids);
     void recalc_inode_space_stats(uint64_t pool_id, bool per_inode);
     // read an object entry and lock it against removal
     // in the future, may become asynchronous
     heap_entry_t *lock_and_read_entry(object_id oid);
-    // re-read a locked object entry with the given lsn (pointer may be invalidated)
-    heap_entry_t *read_locked_entry(object_id oid, uint64_t lsn);
     // read an object entry without locking it
     heap_entry_t *read_entry(object_id oid);
     // unlock an entry
     bool unlock_entry(object_id oid);
     // set or verify checksums in a write request
-    bool calc_checksums(heap_entry_t *wr, uint8_t *data, bool set, uint32_t offset = 0, uint32_t len = 0);
+    bool calc_checksums(heap_entry_t *wr, uint8_t *data, bool set, uint32_t offset = UINT32_MAX, uint32_t len = UINT32_MAX);
     // set or verify raw block checksums
     bool calc_block_checksums(uint32_t *block_csums, uint8_t *data, uint8_t *bitmap, uint32_t start, uint32_t end,
         bool set, std::function<void(uint32_t, uint32_t, uint32_t)> bad_block_cb);

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;
@@ -406,8 +406,3 @@ bool blockstore_impl_t::reshard_continue(void *reshard_state, uint64_t chunk_lim
 {
     return heap->reshard_continue(reshard_state, chunk_limit);
 }
-
-void blockstore_impl_t::reshard_abort(void *reshard_state)
-{
-    return heap->reshard_abort(reshard_state);
-}

src/blockstore/blockstore_impl.h

@@ -78,6 +78,7 @@ 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;
@@ -191,7 +192,6 @@ public:
 
     void* reshard_start(pool_id_t pool, uint32_t pg_count, uint32_t pg_stripe_size, uint64_t chunk_limit);
     bool reshard_continue(void *reshard_state, uint64_t chunk_limit);
-    void reshard_abort(void *reshard_state);
 
     // Event loop
     void loop();

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 (%u/%u/%u, %u/%u, %ju).\n",
+                " differs from OSD configuration (%ju/%ju/%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, false, loaded);
+            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);
             if (r != 0)
                 exit(1);
             entries_loaded += loaded;
@@ -239,6 +239,7 @@ resume_4:
         return 1;
     }
     // metadata read finished
+    bs->heap->finish_load();
     printf("Metadata entries loaded: %ju, used blocks: %ju / %ju\n", entries_loaded, bs->heap->get_data_used_space() / bs->dsk.data_block_size, bs->dsk.block_count);
     if (zero_on_init && !bs->dsk.disable_meta_fsync)
     {
@@ -284,7 +285,7 @@ resume_6:
     }, bs->meta_write_recheck_parallelism);
     return 1;
 resume_7:
-    if (bs->heap->finish_load() != 0)
+    if (bs->heap->finish_recheck() != 0)
     {
         exit(1);
     }

src/blockstore/blockstore_open.cpp

@@ -28,6 +28,7 @@ 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_stable.cpp

@@ -57,9 +57,9 @@ int blockstore_impl_t::dequeue_stable(blockstore_op_t *op)
             }
             assert(res == 0);
         }
+resume_1:
         if (priv->modified_block != UINT32_MAX && priv->modified_block2 != priv->modified_block)
         {
-resume_1:
             BS_SUBMIT_CHECK_SQES(1);
             prepare_meta_block_write(priv->modified_block);
 resume_2:

src/blockstore/blockstore_write.cpp

@@ -13,38 +13,36 @@ bool blockstore_impl_t::enqueue_write(blockstore_op_t *op)
     return true;
 }
 
-bool blockstore_impl_t::prepare_meta_block_write(uint32_t modified_block)
+void blockstore_impl_t::prepare_meta_block_write(uint32_t modified_block)
 {
-    auto mod_it = modified_blocks.find(modified_block);
-    if (mod_it != modified_blocks.end())
-    {
-        return !mod_it->second.sent;
-    }
+    if (modified_blocks.find(modified_block) != modified_blocks.end())
+        return;
     io_uring_sqe *sqe = get_sqe();
     assert(sqe != NULL);
     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, buf](ring_data_t *data)
+    data->callback = [this, modified_block](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);
         }
-        modified_blocks.erase(modified_block);
+        auto it = modified_blocks.find(modified_block);
+        assert(it != modified_blocks.end());
+        free(it->second.buf);
+        modified_blocks.erase(it);
         heap->complete_block_write(modified_block);
         ringloop->wakeup();
     };
     io_uring_prep_writev(
-        sqe, dsk.meta_fd, &data->iov, 1, dsk.meta_offset + (modified_block+1)*dsk.meta_block_size
+        sqe, dsk.meta_fd, &data->iov, 1, dsk.meta_offset + ((uint64_t)modified_block+1)*dsk.meta_block_size
     );
     unsynced_meta_write_count++;
     pending_modified_blocks.push_back(modified_block);
     modified_blocks[modified_block] = { .sent = false, .buf = buf };
-    return true;
 }
 
 bool blockstore_impl_t::meta_block_is_pending(uint32_t modified_block)
@@ -125,7 +123,6 @@ int blockstore_impl_t::dequeue_write(blockstore_op_t *op)
     heap_entry_t *obj = heap->read_entry(op->oid);
     if (op->opcode == BS_OP_DELETE)
     {
-        return continue_delete(op, 0);
         // Delete
         if (!obj || obj->type() == BS_HEAP_DELETE)
         {
@@ -138,38 +135,17 @@ int blockstore_impl_t::dequeue_write(blockstore_op_t *op)
         BS_SUBMIT_CHECK_SQES(1);
         int res = heap->add_delete(obj, &PRIV(op)->modified_block);
         if (res == ENOSPC)
-        {
             goto enospc;
-        }
         assert(res == 0);
+        prepare_meta_block_write(PRIV(op)->modified_block);
+        PRIV(op)->pending_ops++;
+        PRIV(op)->op_state = 5;
         write_iodepth++;
-resume_1:
-        while (!prepare_meta_block_write(PRIV(op)->modified_block))
-        {
-            PRIV(op)->op_state = 1;
-            return 1;
-        }
-rseume_2:
-        while (meta_block_is_pending(PRIV(op)->modified_block))
-        {
-            PRIV(op)->op_state = 2;
-            return 1;
-        }
-resume_3:
-resume_4:
-        if (!throttle_write(op, 3))
-        {
-            return 1;
-        }
-        write_iodepth--;
-        ack_write(op);
-        return 2;
     }
     // FIXME: Allow to do initial writes as buffered, not redirected
     // FIXME: Allow to do direct writes over holes
     else if (!obj || obj->type() == BS_HEAP_DELETE || op->offset == 0 && op->len == dsk.data_block_size)
     {
-        return continue_big_write(op, 10);
         // Big (redirect) write
         PRIV(op)->write_type = dsk.disable_data_fsync || op->opcode != BS_OP_WRITE_STABLE ? BS_HEAP_BIG_WRITE : _REDIRECT_INTENT;
         BS_SUBMIT_CHECK_SQES(1);
@@ -189,7 +165,6 @@ enospc:
             flusher->request_trim();
             return 0;
         }
-        write_iodepth++;
         uint64_t loc = PRIV(op)->location;
 #ifdef BLOCKSTORE_DEBUG
         printf(
@@ -203,72 +178,18 @@ enospc:
         data->iov = (struct iovec){ op->buf, op->len };
         data->callback = [this, op](ring_data_t *data) { handle_write_event(data, op); };
         io_uring_prep_writev(sqe, dsk.data_fd, &data->iov, 1, dsk.data_offset + loc + op->offset);
-        if (PRIV(op)->write_type == BS_HEAP_BIG_WRITE)
-            inflight_big++;
         PRIV(op)->pending_ops++;
-resume_10:
-        if (PRIV(op)->pending_ops > 0)
-        {
-            PRIV(op)->op_state = 10;
-            return 1;
-        }
+        write_iodepth++;
         if (PRIV(op)->write_type == BS_HEAP_BIG_WRITE)
         {
-            inflight_big--;
-resume_11:
-resume_12:
-resume_13:
-            if (!fsync_big_write(op, 11))
-                return 1;
-        }
-        heap_entry_t *obj = heap->read_entry(op->oid);
-        int res = 0;
-        if (PRIV(op)->write_type == _REDIRECT_INTENT)
-        {
-            res = heap->add_redirect_intent(op->oid, &obj, op->version, op->offset, op->len,
-                PRIV(op)->location, op->bitmap, (uint8_t*)op->buf, &PRIV(op)->modified_block);
+            PRIV(op)->op_state = 1;
+            inflight_big++;
         }
         else
-        {
-            res = heap->add_big_write(op->oid, obj, op->opcode == BS_OP_WRITE_STABLE,
-                op->version, op->offset, op->len, PRIV(op)->location, op->bitmap, (uint8_t*)op->buf, &PRIV(op)->modified_block);
-        }
-        if (res == ENOSPC)
-        {
-            if (!heap->get_to_compact_count())
-            {
-                // no space
-                heap->free_data(op->oid.inode, PRIV(op)->location);
-                write_iodepth--;
-                op->retval = -ENOSPC;
-                FINISH_OP(op);
-                return 2;
-            }
-            PRIV(op)->wait_for = WAIT_COMPACTION;
-            PRIV(op)->wait_detail = heap->get_compacted_count();
-            flusher->request_trim();
-            return 0;
-        }
-        assert(res == 0);
-resume_14:
-        while (!prepare_meta_block_write(PRIV(op)->modified_block))
-        {
-            PRIV(op)->op_state = 14;
-            return 1;
-        }
-resume_15:
-        while (meta_block_is_pending(PRIV(op)->modified_block))
-        {
-            PRIV(op)->op_state = 15;
-            return 1;
-        }
-        write_iodepth--;
-        ack_write(op);
-        return 2;
+            PRIV(op)->op_state = 3;
     }
     else if (intent_write_allowed(op, obj))
     {
-        return continue_intent_write(op, 20);
         // Direct intent-write
         BS_SUBMIT_CHECK_SQES(1);
         int res = 0;
@@ -304,41 +225,13 @@ resume_15:
             assert(res == 0);
             PRIV(op)->lsn = obj->lsn;
         }
-        write_iodepth++;
-resume_20:
-        while (!prepare_meta_block_write(PRIV(op)->modified_block))
-        {
-            PRIV(op)->op_state = 20;
-            return 1;
-        }
-resume_21:
-        while (meta_block_is_pending(PRIV(op)->modified_block))
-        {
-            PRIV(op)->op_state = 21;
-            return 1;
-        }
-        // Direct intent-write
-        // LSN is not marked as completed so big_write won't be freed
-        BS_SUBMIT_GET_SQE(sqe, data);
-        data->iov = (struct iovec){ op->buf, op->len };
-        data->callback = [this, op](ring_data_t *data) { handle_write_event(data, op); };
-        io_uring_prep_writev(sqe, dsk.data_fd, &data->iov, 1, dsk.data_offset + PRIV(op)->location + op->offset);
-        if (dsk.use_atomic_flag)
-            sqe->rw_flags = RWF_ATOMIC;
+        prepare_meta_block_write(PRIV(op)->modified_block);
         PRIV(op)->pending_ops++;
-resume_22:
-        if (PRIV(op)->pending_ops > 0)
-        {
-            PRIV(op)->op_state = 22;
-            return 1;
-        }
-        write_iodepth--;
-        ack_write(op);
-        return 2;
+        PRIV(op)->op_state = 9;
+        write_iodepth++;
     }
     else
     {
-        return continue_small_write(op, 30);
         // Small (buffered) overwrite
         // First check if there is free buffer space
         PRIV(op)->write_type = BS_HEAP_SMALL_WRITE;
@@ -351,20 +244,19 @@ resume_22:
             return 0;
         }
         // There is sufficient space. Check SQE(s)
-        BS_SUBMIT_CHECK_SQES(1 + (op->len > 0 ? 1 : 0)); ---> refactor too
+        BS_SUBMIT_CHECK_SQES(1 + (op->len > 0 ? 1 : 0));
         int res = heap->add_small_write(op->oid, &obj, (BS_HEAP_SMALL_WRITE | (op->opcode == BS_OP_WRITE_STABLE ? BS_HEAP_STABLE : 0)),
             op->version, op->offset, op->len, loc, op->bitmap, (uint8_t*)op->buf, &PRIV(op)->modified_block);
         if (res == 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);
@@ -420,46 +312,27 @@ again:
         goto again;
     }
 resume_2:
-
-
-resume_4:
-resume_6:
-
-resume_8:
-    ack
-    return 2;
-resume_10:
-    return 1;
-}
-
-bool blockstore_impl_t::fsync_big_write(blockstore_op_t *op, int base_state)
-{
-    if (PRIV(op)->state == base_state)
-        goto resume_0;
-    else if (PRIV(op)->state == base_state+1)
-        goto resume_1;
-    else if (PRIV(op)->state == base_state+2)
-        goto resume_2;
     // We must fsync all big writes to avoid complex write workflows
     // It's OK for all HDDs and for server SSDs, but slightly worse for desktop SSDs
+    inflight_big--;
     if (!dsk.disable_data_fsync)
     {
         // fsync data in a batch
-resume_0:
+resume_11:
         if (inflight_big > 0)
         {
-            PRIV(op)->op_state = base_state;
-            return false;
+            PRIV(op)->op_state = 11;
+            return 1;
         }
         if (fsyncing_data)
         {
-resume_1:
+resume_12:
             if (fsyncing_data)
             {
-                PRIV(op)->op_state = base_state+1;
-                return false;
+                PRIV(op)->op_state = 12;
+                return 1;
             }
-            return true;
+            goto resume_4;
         }
         fsyncing_data = true;
         BS_SUBMIT_GET_SQE(sqe, data);
@@ -471,23 +344,48 @@ resume_1:
             handle_write_event(data, op);
         };
         PRIV(op)->pending_ops++;
-resume_2:
-        if (PRIV(op)->pending_ops > 0)
-        {
-            PRIV(op)->op_state = base_state+2;
-            return false;
-        }
+        PRIV(op)->op_state = 3;
+        return 1;
     }
-    return true;
-}
-
-bool blockstore_impl_t::throttle_write(blockstore_op_t *op, int base_state)
-{
-    // Apply throttling to not fill the journal too quickly for the SSD+HDD case
-    if (PRIV(op)->op_state >= base_state+1)
+resume_4:
     {
-        return true;
+        BS_SUBMIT_CHECK_SQES(1);
+        auto obj = heap->read_entry(op->oid);
+        int res = 0;
+        if (PRIV(op)->write_type == _REDIRECT_INTENT)
+        {
+            res = heap->add_redirect_intent(op->oid, &obj, op->version, op->offset, op->len,
+                PRIV(op)->location, op->bitmap, (uint8_t*)op->buf, &PRIV(op)->modified_block);
+        }
+        else
+        {
+            res = heap->add_big_write(op->oid, obj, op->opcode == BS_OP_WRITE_STABLE,
+                op->version, op->offset, op->len, PRIV(op)->location, op->bitmap, (uint8_t*)op->buf, &PRIV(op)->modified_block);
+        }
+        if (res == ENOSPC)
+        {
+            if (!heap->get_to_compact_count())
+            {
+                // no space
+                heap->free_data(op->oid.inode, PRIV(op)->location);
+                write_iodepth--;
+                op->retval = -ENOSPC;
+                FINISH_OP(op);
+                return 2;
+            }
+            PRIV(op)->wait_for = WAIT_COMPACTION;
+            PRIV(op)->wait_detail = heap->get_compacted_count();
+            flusher->request_trim();
+            return 0;
+        }
+        assert(res == 0);
+        prepare_meta_block_write(PRIV(op)->modified_block);
+        PRIV(op)->pending_ops++;
+        PRIV(op)->op_state = 5;
+        return 1;
     }
+resume_6:
+    // Apply throttling to not fill the journal too quickly for the SSD+HDD case
     if (PRIV(op)->write_type == BS_HEAP_SMALL_WRITE && throttle_small_writes)
     {
         // Apply throttling
@@ -508,21 +406,18 @@ bool blockstore_impl_t::throttle_write(blockstore_op_t *op, int base_state)
         if (ref_us > exec_us + throttle_threshold_us)
         {
             // Pause reply
-            PRIV(op)->op_state = base_state;
+            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)->op_state++;
+                PRIV(op)->pending_ops--;
                 ringloop->wakeup();
             });
-            return false;
+            return 1;
         }
     }
-    return true;
-}
-
-void blockstore_impl_t::ack_write(blockstore_op_t *op)
-{
+resume_8:
     // Acknowledge write
 #ifdef BLOCKSTORE_DEBUG
     printf("Ack write %jx:%jx v%ju\n", op->oid.inode, op->oid.stripe, op->version);
@@ -549,7 +444,21 @@ void blockstore_impl_t::ack_write(blockstore_op_t *op)
         unsynced_data_write_count++;
         intent_write_counter++;
     }
+    write_iodepth--;
     FINISH_OP(op);
+    return 2;
+resume_10:
+    // Direct intent-write
+    // LSN is not marked as completed so big_write won't be freed
+    BS_SUBMIT_GET_SQE(sqe, data);
+    data->iov = (struct iovec){ op->buf, op->len };
+    data->callback = [this, op](ring_data_t *data) { handle_write_event(data, op); };
+    io_uring_prep_writev(sqe, dsk.data_fd, &data->iov, 1, dsk.data_offset + PRIV(op)->location + op->offset);
+    if (dsk.use_atomic_flag)
+        sqe->rw_flags = RWF_ATOMIC;
+    PRIV(op)->pending_ops++;
+    PRIV(op)->op_state = 7;
+    return 1;
 }
 
 void blockstore_impl_t::handle_write_event(ring_data_t *data, blockstore_op_t *op)

src/blockstore/v1/impl.cpp

@@ -480,17 +480,6 @@ resume_1:
     return true;
 }
 
-void blockstore_impl_t::reshard_abort(void *reshard_state)
-{
-    bs_reshard_state_t *st = (bs_reshard_state_t*)reshard_state;
-    for (auto sh_it = st->old_shards.begin(); sh_it != st->old_shards.end(); sh_it++)
-    {
-        auto & to = clean_db_shards[sh_it->first];
-        to.swap(sh_it->second);
-    }
-    delete st;
-}
-
 void blockstore_impl_t::process_list(blockstore_op_t *op)
 {
     uint32_t list_pg = op->pg_number+1;

src/blockstore/v1/impl.h

@@ -290,7 +290,6 @@ public:
     // Reshard database for a pool
     void* reshard_start(pool_id_t pool, uint32_t pg_count, uint32_t pg_stripe_size, uint64_t chunk_limit);
     bool reshard_continue(void *reshard_state, uint64_t chunk_limit);
-    void reshard_abort(void *reshard_state);
 
     // Event loop
     void loop();

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 (%u/%u/%u, %u/%u).\n",
+                " differs from OSD configuration (%ju/%ju/%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

@@ -620,8 +620,7 @@ 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 +
-            item_start/dsk.csum_block_size*(dsk.data_csum_type & 0xFF));
+        uint8_t *csum = !dsk.csum_block_size ? 0 : (clean_entry_bitmap + dsk.clean_entry_bitmap_size);
         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} ${CARES_LIBRARIES})
 target_compile_options(vitastor_common PUBLIC -fPIC)
 
 # libvitastor_client.so
@@ -24,6 +24,7 @@ add_library(vitastor_client SHARED
 	cluster_client.cpp
 	cluster_client_list.cpp
 	cluster_client_wb.cpp
+	cluster_client_icache.cpp
 	vitastor_c.cpp
 )
 set_target_properties(vitastor_client PROPERTIES PUBLIC_HEADER "client/vitastor_c.h")
@@ -33,6 +34,7 @@ target_link_libraries(vitastor_client
 	${LIBURING_LIBRARIES}
 	${IBVERBS_LIBRARIES}
 	${RDMACM_LIBRARIES}
+	${OPENSSL_LIBRARIES}
 )
 set_target_properties(vitastor_client PROPERTIES VERSION ${VITASTOR_VERSION} SOVERSION 0)
 configure_file(vitastor.pc.in vitastor.pc @ONLY)
@@ -98,9 +100,10 @@ endif (${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 cluster_client_icache.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

@@ -62,6 +62,7 @@ cluster_client_t::cluster_client_t(ring_loop_t *ringloop, timerfd_manager_t *tfd
     st_cli.on_change_node_placement_hook = [this]() { on_change_node_placement_hook(); };
     st_cli.on_load_pgs_hook = [this](bool success) { on_load_pgs_hook(success); };
     st_cli.on_reload_hook = [this]() { st_cli.load_global_config(); };
+    st_cli.on_inode_change_hook = [this](uint64_t inode, bool removed) { on_change_inode_hook(inode, removed); };
 
     st_cli.parse_config(config);
     st_cli.infinite_start = false;
@@ -77,6 +78,7 @@ cluster_client_t::cluster_client_t(ring_loop_t *ringloop, timerfd_manager_t *tfd
 
 cluster_client_t::~cluster_client_t()
 {
+    vault_destroy();
     if (retry_timeout_id >= 0)
     {
         tfd->clear_timer(retry_timeout_id);
@@ -481,6 +483,8 @@ void cluster_client_t::on_load_config_hook(json11::Json::object & etcd_global_co
         self_tree_metrics.clear();
         client_hostname = new_hostname;
     }
+    // vault
+    vault_parse_config();
     msgr.parse_config(config);
     st_cli.parse_config(config);
     st_cli.load_pgs();
@@ -607,6 +611,9 @@ void cluster_client_t::on_change_pool_config_hook()
             pg_counts[pool_item.first] = pool_item.second.real_pg_count;
         }
     }
+    inode_cache.clear();
+    inode_cache_children.clear();
+    vault_keys.clear();
     continue_ops();
 }
 
@@ -673,6 +680,10 @@ bool cluster_client_t::flush()
 {
     if (!ringloop)
     {
+        if (vault_loading)
+        {
+            return false;
+        }
         if (wb->writeback_queue.size())
         {
             wb->start_writebacks(this, 0);
@@ -695,7 +706,7 @@ bool cluster_client_t::flush()
         sync_done = true;
     };
     execute(sync);
-    while (!sync_done)
+    while (!sync_done || vault_loading)
     {
         ringloop->loop();
         if (!sync_done)
@@ -958,10 +969,40 @@ bool cluster_client_t::check_rw(cluster_op_t *op)
     {
         op->flags |= OP_IMMEDIATE_COMMIT;
     }
+    bool searched = false;
+    std::shared_ptr<inode_cache_t> icache;
+    if (op->opcode == OSD_OP_READ || op->opcode == OSD_OP_WRITE)
+    {
+        if (!searched)
+        {
+            icache = inode_cache_get(op->inode);
+            searched = true;
+        }
+        if (icache && icache->has_parent_loop && op->opcode == OSD_OP_READ)
+        {
+            op->retval = -EINVAL;
+            auto cb = std::move(op->callback);
+            cb(op);
+            return false;
+        }
+        if (icache && icache->op_enc)
+        {
+            // Use shared_ptr aliasing to attach op_enc to the inode cache entry
+            op->enc = std::shared_ptr<osd_op_enc_t>(icache, icache->op_enc);
+        }
+        else
+            op->enc.reset();
+    }
+    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)
+        if (!searched)
+        {
+            icache = inode_cache_get(op->inode);
+            searched = true;
+        }
+        if (icache && icache->readonly)
         {
             op->retval = -EROFS;
             auto cb = std::move(op->callback);
@@ -972,33 +1013,39 @@ 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())
+        if (!searched)
         {
-            int chain_size = 0;
-            while (ino_it != st_cli.inode_config.end() && ino_it->second.parent_id)
+            icache = inode_cache_get(op->inode);
+            searched = true;
+        }
+        if (icache)
+        {
+            for (auto & parent: icache->chain)
             {
-                // Check for loops - FIXME check it in etcd_state_client
-                if (ino_it->second.parent_id == op->inode ||
-                    chain_size > st_cli.inode_config.size())
-                {
-                    op->retval = -EINVAL;
-                    auto cb = std::move(op->callback);
-                    cb(op);
-                    return false;
-                }
-                if (INODE_POOL(ino_it->second.parent_id) == INODE_POOL(ino_it->first) &&
-                    wb->has_inode(ino_it->second.parent_id))
+                if (INODE_POOL(parent) == INODE_POOL(op->inode) && wb->has_inode(parent))
                 {
                     // Deoptimise reads - we have dirty data for one of the parent layer(s).
                     op->deoptimise_snapshot = true;
                     break;
                 }
-                chain_size++;
-                ino_it = st_cli.inode_config.find(ino_it->second.parent_id);
             }
         }
     }
+    if (icache && icache->err_code)
+    {
+        if (icache->err_code == EPERM)
+        {
+            op->retval = -EPERM;
+            auto cb = std::move(op->callback);
+            cb(op);
+            return false;
+        }
+        else if (icache->err_code == EAGAIN)
+        {
+            key_wait_ops.push_back(op);
+            return false;
+        }
+    }
     return true;
 }
 
@@ -1121,31 +1168,33 @@ resume_2:
         // because if some operations were invalid for the new PG count we'd get errors
         if (op->opcode == OSD_OP_READ || op->opcode == OSD_OP_READ_CHAIN_BITMAP)
         {
-            // Check parent inode
-            auto ino_it = st_cli.inode_config.find(op->cur_inode);
-            // Skip parents from the same pool
-            int skipped = 0;
-            while (!op->deoptimise_snapshot &&
-                ino_it != st_cli.inode_config.end() && ino_it->second.parent_id &&
-                INODE_POOL(ino_it->second.parent_id) == INODE_POOL(op->cur_inode))
+            uint64_t next_inode = 0;
+            auto icache = inode_cache_get(op->cur_inode);
+            if (icache)
             {
-                // Check for loops - FIXME check it in etcd_state_client
-                if (ino_it->second.parent_id == op->inode ||
-                    skipped > st_cli.inode_config.size())
+                if (icache->has_parent_loop)
                 {
                     op->retval = -EINVAL;
                     erase_op(op);
                     return 1;
                 }
-                skipped++;
-                ino_it = st_cli.inode_config.find(ino_it->second.parent_id);
+                if (op->deoptimise_snapshot)
+                {
+                    if (icache->chain.size() > 1)
+                        next_inode = icache->chain[1];
+                }
+                else
+                {
+                    if (icache->other_pool_parent_id)
+                        next_inode = icache->other_pool_parent_id;
+                }
             }
-            if (ino_it != st_cli.inode_config.end() &&
-                ino_it->second.parent_id &&
-                ino_it->second.parent_id != op->inode)
+            if (next_inode)
             {
                 // Continue reading from the parent inode
-                op->cur_inode = ino_it->second.parent_id;
+                icache = inode_cache_get(next_inode);
+                op->cur_inode = next_inode;
+                op->enc = (icache && icache->op_enc ? std::shared_ptr<osd_op_enc_t>(icache, icache->op_enc) : nullptr);
                 op->parts.clear();
                 op->done_count = 0;
                 goto resume_0;
@@ -1252,7 +1301,11 @@ void cluster_client_t::slice_rw(cluster_op_t *op)
         // Allocate memory for the bitmap
         unsigned object_bitmap_size = ((op->len / pool_cfg.bitmap_granularity + 7) / 8);
         object_bitmap_size = (object_bitmap_size < 8 ? 8 : object_bitmap_size);
-        unsigned bitmap_mem = object_bitmap_size + (pool_cfg.data_block_size / pool_cfg.bitmap_granularity / 8 * pg_data_size) * op->parts.size();
+        unsigned bitmap_mem = object_bitmap_size +
+            op->parts.size() * pg_data_size *
+            (pool_cfg.data_block_size / pool_cfg.bitmap_granularity / 8
+            // read chain info - 1 byte per block
+            + (op->enc ? op->len/pool_cfg.bitmap_granularity : 0));
         if (!op->bitmap_buf || op->bitmap_buf_size < bitmap_mem)
         {
             op->bitmap_buf = realloc_or_die(op->bitmap_buf, bitmap_mem);
@@ -1407,9 +1460,9 @@ int cluster_client_t::try_send(cluster_op_t *op, int i, std::function<void(osd_o
             int peer_fd = peer_it->second;
             part->flags |= PART_SENT|PART_VALID;
             op->inflight_count++;
-            uint64_t pg_bitmap_size = (pool_cfg.data_block_size / pool_cfg.bitmap_granularity / 8) * (
-                pool_cfg.scheme == POOL_SCHEME_REPLICATED ? 1 : pool_cfg.pg_size-pool_cfg.parity_chunks
-            );
+            uint32_t pg_data_size = (pool_cfg.scheme == POOL_SCHEME_REPLICATED ? 1 : pool_cfg.pg_size-pool_cfg.parity_chunks);
+            uint64_t pg_bitmap_size = pg_data_size * (pool_cfg.data_block_size / pool_cfg.bitmap_granularity / 8
+                + (op->opcode == OSD_OP_READ && op->enc ? pool_cfg.data_block_size/pool_cfg.bitmap_granularity : 0));
             uint64_t meta_rev = 0;
             if (op->opcode != OSD_OP_READ_BITMAP && op->opcode != OSD_OP_DELETE && !op->deoptimise_snapshot)
             {
@@ -1428,6 +1481,7 @@ int cluster_client_t::try_send(cluster_op_t *op, int i, std::function<void(osd_o
                     .inode = op->cur_inode,
                     .offset = part->offset,
                     .len = part->len,
+                    .flags = op->opcode == OSD_OP_READ && op->enc && !op->deoptimise_snapshot ? OSD_OP_RETURN_CHAIN : 0,
                     .meta_revision = meta_rev,
                     .version = op->opcode == OSD_OP_WRITE || op->opcode == OSD_OP_DELETE ? op->version : 0,
                 } },
@@ -1435,6 +1489,7 @@ 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

@@ -5,6 +5,7 @@
 
 #include "messenger.h"
 #include "etcd_state_client.h"
+#include "../util/robin_hood.h"
 
 #define DEFAULT_CLIENT_MAX_DIRTY_BYTES 32*1024*1024
 #define DEFAULT_CLIENT_MAX_DIRTY_OPS 1024
@@ -71,6 +72,7 @@ protected:
     cluster_op_t *prev = NULL, *next = NULL;
     int prev_wait = 0;
     uint64_t flush_id = 0;
+    std::shared_ptr<osd_op_enc_t> enc;
     friend class cluster_client_t;
     friend class writeback_cache_t;
 };
@@ -80,6 +82,25 @@ struct inode_list_osd_t;
 struct inode_list_pg_t;
 class writeback_cache_t;
 
+struct inode_cache_t
+{
+    std::vector<inode_t> chain;
+    uint8_t *key_data = NULL;
+    osd_op_enc_t *op_enc = NULL;
+    bool readonly = false;
+    bool has_parent_loop = false;
+    inode_t other_pool_parent_id = 0;
+    int err_code = 0;
+
+    ~inode_cache_t();
+};
+
+struct vault_load_key_t
+{
+    int key_state = 0;
+    std::string key;
+};
+
 // FIXME: Split into public and private interfaces
 class __attribute__((visibility("default"))) cluster_client_t
 {
@@ -89,8 +110,8 @@ public:
     timerfd_manager_t *tfd = NULL;
     ring_loop_t *ringloop = NULL;
 
-    std::map<pool_id_t, uint64_t> pg_counts;
-    std::map<pool_pg_num_t, osd_num_t> pg_primary;
+    // config:
+
     // client_max_dirty_* is actually "max unsynced", for the case when immediate_commit is off
     uint64_t client_max_dirty_bytes = 0;
     uint64_t client_max_dirty_ops = 0;
@@ -102,12 +123,23 @@ public:
     uint64_t client_max_writeback_iodepth = 0;
     std::string conf_hostname;
 
+    std::string vault_url;
+    std::string vault_client_cert;
+    std::string vault_client_key;
+    std::string vault_ca;
+    std::string vault_secret_api_path;
+    uint64_t vault_timeout_ms = 0;
+    uint64_t vault_error_timeout_sec = 0;
+    uint64_t vault_refresh_leeway_sec = 0;
+
     int log_level = 0;
     int client_retry_interval = 50; // ms
     int client_eio_retry_interval = 1000; // ms
     bool client_retry_enospc = true;
     int client_wait_up_timeout = 16; // sec (for listings)
 
+    // state:
+
     std::string client_hostname;
     std::map<std::string, int> self_tree_metrics;
     std::map<osd_num_t, int> osd_tree_metrics;
@@ -115,6 +147,7 @@ public:
     int retry_timeout_id = -1;
     int retry_timeout_duration = 0;
     std::vector<cluster_op_t*> offline_ops;
+    std::vector<cluster_op_t*> key_wait_ops;
     cluster_op_t *op_queue_head = NULL, *op_queue_tail = NULL;
     writeback_cache_t *wb = NULL;
     std::set<osd_num_t> dirty_osds;
@@ -123,7 +156,22 @@ public:
     void *scrap_buffer = NULL;
     unsigned scrap_buffer_size = 0;
 
+    // inodes require some extra state for read/write, it's stored here.
+    // moreover, robin_hood access is slightly faster than std::map :)
+    robin_hood::unordered_flat_map<inode_t, std::shared_ptr<inode_cache_t>> inode_cache;
+    std::set<std::pair<inode_t, inode_t>> inode_cache_children;
+
+    http_context_t *vault_http_ctx = NULL;
+    http_co_t *vault_http_cli = NULL;
+    bool vault_loading = false;
+    std::string vault_token;
+    bool vault_auth_error = false;
+    timespec vault_token_expire = {};
+    std::vector<std::string> vault_key_load_queue;
+    std::map<std::string, vault_load_key_t> vault_keys;
+
     bool pgs_loaded = false;
+    std::map<pool_id_t, uint64_t> pg_counts;
     ring_consumer_t consumer;
     std::vector<std::function<void(void)>> on_ready_hooks;
     int list_retry_timeout_id = -1;
@@ -163,6 +211,13 @@ protected:
 #endif
     void continue_ops(int time_passed = 0);
 
+    std::shared_ptr<inode_cache_t> inode_cache_get(inode_t ino);
+    void vault_parse_config();
+    bool vault_check_token();
+    void vault_load_keys();
+    void vault_destroy();
+    void vault_parse_secret(const std::string & key_id, const std::string & err, json11::Json data);
+
 protected:
     bool affects_osd(uint64_t inode, uint64_t offset, uint64_t len, osd_num_t osd);
     bool affects_pg(uint64_t inode, uint64_t offset, uint64_t len, pool_id_t pool_id, pg_num_t pg_num);
@@ -173,6 +228,7 @@ protected:
     void on_change_pg_state_hook(pool_id_t pool_id, pg_num_t pg_num, osd_num_t prev_primary);
     void on_change_osd_state_hook(uint64_t peer_osd);
     void on_change_node_placement_hook();
+    void on_change_inode_hook(uint64_t inode, bool removed);
 
     void execute_internal(cluster_op_t *op);
     void execute_cas(cluster_op_t *op);
@@ -189,6 +245,7 @@ protected:
     void erase_op(cluster_op_t *op);
     void calc_wait(cluster_op_t *op);
     void inc_wait(uint64_t opcode, uint64_t flags, cluster_op_t *next, int inc);
+
     void continue_lists();
     bool continue_listing(inode_list_t *lst);
     bool restart_listing(inode_list_t* lst);

src/client/cluster_client_icache.cpp

@@ -0,0 +1,367 @@
+// Copyright (c) Vitaliy Filippov, 2019+
+// License: VNPL-1.1 or GNU GPL-2.0+ (see README.md for details)
+
+#include <stdexcept>
+#include <assert.h>
+#include "cluster_client_impl.h"
+#include "http_client.h"
+#include "str_util.h"
+
+#define VAULT_KEY_NOT_LOADED 0
+#define VAULT_KEY_LOADING 1
+#define VAULT_KEY_LOADED 2
+#define VAULT_KEY_ERROR 3
+
+inode_cache_t::~inode_cache_t()
+{
+    if (key_data)
+    {
+        free(key_data);
+        key_data = NULL;
+        op_enc = NULL;
+    }
+}
+
+void cluster_client_t::vault_destroy()
+{
+    if (vault_http_ctx)
+    {
+#ifndef __MOCK__
+        http_destroy(vault_http_cli);
+        http_context_destroy(vault_http_ctx);
+        vault_http_cli = NULL;
+        vault_http_ctx = NULL;
+#endif
+    }
+}
+
+void cluster_client_t::vault_parse_config()
+{
+    vault_url = config["vault_url"].string_value();
+    vault_client_cert = config["vault_client_cert"].string_value();
+    vault_client_key = config["vault_client_key"].string_value();
+    vault_ca = config["vault_ca"].string_value();
+    vault_secret_api_path = "/v1/secret/";
+    if (config["vault_secret_api_path"].is_string())
+        vault_secret_api_path = config["vault_secret_api_path"].string_value();
+    vault_timeout_ms = config["vault_timeout_ms"].uint64_value();
+    if (!vault_timeout_ms)
+        vault_timeout_ms = 5000;
+    vault_error_timeout_sec = config["vault_error_timeout_sec"].uint64_value();
+    if (!vault_error_timeout_sec)
+        vault_error_timeout_sec = 60;
+    vault_refresh_leeway_sec = config["vault_refresh_leeway_sec"].uint64_value();
+    if (!vault_refresh_leeway_sec)
+        vault_refresh_leeway_sec = 60;
+}
+
+// FIXME: Rework client API by adding open/close and cache inode information in the "FD" (maybe)
+void cluster_client_t::on_change_inode_hook(uint64_t inode, bool removed)
+{
+    std::vector<inode_t> children = { inode };
+    for (size_t i = 0; i < children.size(); i++)
+    {
+        auto it = inode_cache_children.lower_bound(std::make_pair(children[i], (inode_t)0));
+        while (it != inode_cache_children.end() && it->first == children[i])
+        {
+            children.push_back(it->second);
+            it++;
+        }
+    }
+    for (auto & inode: children)
+    {
+        auto it = inode_cache.find(inode);
+        if (it != inode_cache.end())
+        {
+            auto icache = it->second;
+            for (auto & parent: icache->chain)
+            {
+                inode_cache_children.erase(std::make_pair(parent, inode));
+            }
+            inode_cache.erase(it);
+        }
+    }
+}
+
+std::shared_ptr<inode_cache_t> cluster_client_t::inode_cache_get(inode_t ino)
+{
+    auto icache_it = inode_cache.find(ino);
+    if (icache_it != inode_cache.end())
+    {
+        return icache_it->second;
+    }
+    // Fill inode cache
+    auto ino_it = st_cli.inode_config.find(ino);
+    if (ino_it == st_cli.inode_config.end())
+    {
+        inode_cache[ino] = NULL;
+        return NULL;
+    }
+    auto pool_it = st_cli.pool_config.find(INODE_POOL(ino));
+    if (pool_it == st_cli.pool_config.end())
+    {
+        inode_cache[ino] = NULL;
+        return NULL;
+    }
+    auto & inode_cfg = ino_it->second;
+    auto & pool_cfg = pool_it->second;
+    std::shared_ptr<inode_cache_t> icache = std::make_shared<inode_cache_t>();
+    icache->readonly = inode_cfg.readonly;
+    icache->chain.push_back(ino);
+    std::vector<inode_config_t*> chain_cfg;
+    // FIXME: Allow unencrypted read & write when all chain is encrypted with the same key
+    int enc_key_count = !inode_cfg.enc_key.empty() ? 1 : 0;
+    if (inode_cfg.parent_id)
+    {
+        // Check for loops and cache the chain
+        robin_hood::unordered_flat_set<inode_t> seen;
+        seen.insert(ino);
+        uint64_t parent_id = inode_cfg.parent_id;
+        while (parent_id)
+        {
+            if (seen.find(parent_id) != seen.end())
+            {
+                icache->has_parent_loop = true;
+                break;
+            }
+            seen.insert(parent_id);
+            ino_it = st_cli.inode_config.find(parent_id);
+            if (INODE_POOL(parent_id) == INODE_POOL(ino))
+            {
+                icache->chain.push_back(parent_id);
+                if (ino_it == st_cli.inode_config.end())
+                    chain_cfg.push_back(NULL);
+                else
+                {
+                    chain_cfg.push_back(&ino_it->second);
+                    if (!ino_it->second.enc_key.empty())
+                        enc_key_count++;
+                }
+            }
+            else if (!icache->other_pool_parent_id)
+                icache->other_pool_parent_id = parent_id;
+            if (ino_it == st_cli.inode_config.end())
+                break;
+            parent_id = ino_it->second.parent_id;
+        }
+    }
+    // Check external keys and wait for loading, if required
+    if (enc_key_count)
+    {
+        for (size_t i = 0; i <= chain_cfg.size(); i++)
+        {
+            inode_config_t *cfg = !i ? &inode_cfg : chain_cfg[i-1];
+            if (cfg && cfg->enc_key.substr(0, strlen(VAULT_KEY_PREFIX)) == VAULT_KEY_PREFIX)
+            {
+                auto & ik = vault_keys[inode_cfg.enc_key];
+                if (ik.key_state == VAULT_KEY_ERROR || vault_url.empty())
+                {
+                    icache->err_code = EPERM;
+                    enc_key_count = 0;
+                }
+                else if (ik.key_state == VAULT_KEY_NOT_LOADED)
+                {
+                    ik.key_state = VAULT_KEY_LOADING;
+                    vault_key_load_queue.push_back(inode_cfg.enc_key);
+                    vault_load_keys();
+                    icache->err_code = EAGAIN;
+                    enc_key_count = 0;
+                }
+                else if (ik.key_state == VAULT_KEY_LOADING)
+                {
+                    icache->err_code = EAGAIN;
+                    enc_key_count = 0;
+                }
+                else
+                {
+                    assert(ik.key_state == VAULT_KEY_LOADED);
+                }
+            }
+        }
+    }
+    // Generate encryption key chain, if applicable
+    if (enc_key_count)
+    {
+        uint8_t *key_data = (uint8_t*)malloc_or_die(
+            AES_256_XTS_KEY_SIZE * enc_key_count +
+            sizeof(uint8_t*) * icache->chain.size() +
+            sizeof(osd_op_enc_t)
+        );
+        uint8_t **keys = (uint8_t**)(key_data + AES_256_XTS_KEY_SIZE * enc_key_count);
+        osd_op_enc_t *enc = (osd_op_enc_t*)((uint8_t*)keys + sizeof(uint8_t*)*icache->chain.size());
+        size_t key_pos = 0;
+        for (size_t i = 0; i <= chain_cfg.size(); i++)
+        {
+            inode_config_t *cfg = !i ? &inode_cfg : chain_cfg[i-1];
+            if (cfg && !cfg->enc_key.empty())
+            {
+                const auto & key = cfg->enc_key.substr(0, strlen(VAULT_KEY_PREFIX)) == VAULT_KEY_PREFIX
+                    ? vault_keys.at(cfg->enc_key).key
+                    : cfg->enc_key;
+                assert(key_pos < AES_256_XTS_KEY_SIZE * enc_key_count);
+                assert(key.size() == 2*AES_256_XTS_KEY_SIZE);
+                keys[i] = key_data + key_pos;
+                fromhexstr(key, AES_256_XTS_KEY_SIZE, key_data + key_pos);
+                key_pos += AES_256_XTS_KEY_SIZE;
+            }
+            else
+                keys[i] = NULL;
+        }
+        enc->key_chain = keys;
+        enc->chain_size = icache->chain.size();
+        enc->read_chain_bitmap_pos = pool_cfg.data_block_size/pool_cfg.bitmap_granularity/8;
+        enc->bitmap_granularity = pool_cfg.bitmap_granularity;
+        icache->key_data = key_data;
+        icache->op_enc = enc;
+    }
+    inode_cache[ino] = icache;
+    for (auto & parent: icache->chain)
+    {
+        if (parent != ino)
+            inode_cache_children.insert(std::make_pair(parent, ino));
+    }
+    return icache;
+}
+
+#ifndef __MOCK__
+bool cluster_client_t::vault_check_token()
+{
+    timespec now;
+    clock_gettime(CLOCK_REALTIME, &now);
+    if (!vault_token_expire.tv_sec || vault_token_expire.tv_sec < now.tv_sec)
+    {
+        vault_loading = true;
+        http_json_post(
+            vault_http_cli, vault_url+"/v1/auth/cert/login", json11::Json::object{}, "",
+            (http_options_t){ .timeout = (int)vault_timeout_ms, .keepalive = true },
+            [this](http_message_t *response)
+            {
+                clock_gettime(CLOCK_REALTIME, &vault_token_expire);
+                vault_loading = false;
+                std::string err;
+                json11::Json data;
+                response->parse_json_response(err, data);
+                if (err != "")
+                {
+                    vault_token_expire.tv_sec += vault_error_timeout_sec;
+                    fprintf(stderr, "Vault request failed: %s\n", err.c_str());
+                }
+                else
+                {
+                    uint64_t ttl = data["auth"]["lease_duration"].uint64_value();
+                    vault_token = data["auth"]["client_token"].string_value();
+                    if (vault_token.empty() || !ttl)
+                    {
+                        vault_token_expire.tv_sec += vault_error_timeout_sec;
+                        fprintf(stderr, "No token or lease_duration in Vault response: %s\n", data.dump().c_str());
+                    }
+                    else
+                    {
+                        if (ttl < vault_refresh_leeway_sec)
+                            vault_token_expire.tv_sec += ttl/2;
+                        else
+                            vault_token_expire.tv_sec += ttl - vault_refresh_leeway_sec;
+                    }
+                }
+                vault_load_keys();
+            }
+        );
+        return false;
+    }
+    if (vault_token.empty())
+    {
+        // Auth error happened, mark all loads as failed
+        for (auto & key_id: vault_key_load_queue)
+        {
+            auto & k = vault_keys[key_id];
+            k.key_state = VAULT_KEY_ERROR;
+        }
+        vault_key_load_queue.clear();
+        auto ops = std::move(key_wait_ops);
+        for (cluster_op_t *op: ops)
+            inode_cache.erase(op->inode);
+        for (cluster_op_t *op: ops)
+            execute_internal(op);
+        return false;
+    }
+    return true;
+}
+#endif
+
+void cluster_client_t::vault_load_keys()
+{
+    if (vault_loading || !vault_key_load_queue.size())
+    {
+        return;
+    }
+#ifdef __MOCK__
+    vault_loading = true;
+#else
+    if (!vault_http_ctx)
+    {
+        std::string error;
+        vault_http_ctx = http_context_init(tfd, vault_client_cert, vault_client_key, vault_ca, true, error);
+        if (!vault_http_ctx)
+        {
+            fprintf(stderr, "Failed to initialize HTTP context for Vault: %s\n", error.c_str());
+            exit(1);
+        }
+        vault_http_cli = http_init(vault_http_ctx);
+    }
+    if (!vault_check_token())
+    {
+        return;
+    }
+    std::string key_id = vault_key_load_queue[0];
+    vault_key_load_queue.erase(vault_key_load_queue.begin());
+    vault_loading = true;
+    http_get(
+        vault_http_cli, vault_url+vault_secret_api_path+key_id.substr(strlen(VAULT_KEY_PREFIX)), "X-Vault-Token: "+vault_token+"\r\n",
+        (http_options_t){ .timeout = (int)vault_timeout_ms, .keepalive = true },
+        [this, key_id](http_message_t *response)
+        {
+            vault_loading = false;
+            std::string err;
+            json11::Json data;
+            response->parse_json_response(err, data);
+            vault_parse_secret(key_id, err, data);
+        }
+    );
+#endif
+}
+
+void cluster_client_t::vault_parse_secret(const std::string & key_id, const std::string & err, json11::Json data)
+{
+    vault_loading = false;
+    auto & k = vault_keys[key_id];
+    if (err != "")
+    {
+        k.key_state = VAULT_KEY_ERROR;
+        fprintf(stderr, "Vault %s%s%s request failed: %s\n", vault_url.c_str(),
+            vault_secret_api_path.c_str(), key_id.c_str()+strlen(VAULT_KEY_PREFIX), err.c_str());
+    }
+    else
+    {
+        auto hexkey = data["data"]["key"].string_value();
+        if (hexkey.empty() || !ishexstr(hexkey) || hexkey.size() != 2*AES_256_XTS_KEY_SIZE)
+        {
+            k.key_state = VAULT_KEY_ERROR;
+            fprintf(stderr, "Vault /v1/secret/%s request failed: 'key' is empty or has invalid format\n", key_id.c_str());
+        }
+        else
+        {
+            k.key_state = VAULT_KEY_LOADED;
+            k.key = hexkey;
+        }
+    }
+    if (vault_key_load_queue.empty())
+    {
+        auto ops = std::move(key_wait_ops);
+        for (cluster_op_t *op: ops)
+            inode_cache.erase(op->inode);
+        for (cluster_op_t *op: ops)
+            execute_internal(op);
+    }
+    else
+        vault_load_keys();
+}

src/client/etcd_state_client.cpp

@@ -1,7 +1,10 @@
 // Copyright (c) Vitaliy Filippov, 2019+
 // License: VNPL-1.1 or GNU GPL-2.0+ (see README.md for details)
 
+#include <assert.h>
+#include "malloc_or_die.h"
 #include "osd_ops.h"
+#include "msgr_op.h"
 #include "pg_states.h"
 #include "etcd_state_client.h"
 #ifndef __MOCK__
@@ -22,14 +25,19 @@ 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)
     {
@@ -72,55 +80,51 @@ std::vector<std::string> etcd_state_client_t::get_addresses()
     return addrs;
 }
 
-void etcd_state_client_t::etcd_call_oneshot(std::string etcd_address, std::string api, json11::Json payload,
+http_context_t *etcd_state_client_t::get_http_ctx()
+{
+    if (!http_ctx)
+    {
+        std::string error;
+        http_ctx = http_context_init(tfd, 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(const std::string & etcd_url, const std::string & api, json11::Json payload,
     int timeout, std::function<void(std::string, json11::Json)> callback)
 {
-    std::string etcd_api_path;
-    int pos = etcd_address.find('/');
-    if (pos >= 0)
-    {
-        etcd_api_path = etcd_address.substr(pos);
-        etcd_address = etcd_address.substr(0, pos);
-    }
-    std::string req = payload.dump();
-    req = "POST "+etcd_api_path+api+" HTTP/1.1\r\n"
-        "Host: "+etcd_address+"\r\n"
-        "Content-Type: application/json\r\n"
-        "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(get_http_ctx());
+    http_json_post(http_cli, etcd_url+api, payload, "", { .timeout = timeout }, [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_request(http_cli, etcd_address, req, { .timeout = timeout }, cb);
+        http_destroy(http_cli);
+    });
 }
 
-void etcd_state_client_t::etcd_call(std::string api, json11::Json payload, int timeout,
+void etcd_state_client_t::etcd_call(const std::string & api, json11::Json payload, int timeout,
     int retries, int interval, std::function<void(std::string, json11::Json)> callback)
 {
-    if (!etcd_addresses.size() && !etcd_local.size())
+    pick_next_etcd([=]()
     {
-        fprintf(stderr, "etcd_address is missing in Vitastor configuration\n");
-        exit(1);
-    }
-    pick_next_etcd();
-    std::string etcd_address = selected_etcd_address;
-    std::string etcd_api_path;
-    int pos = etcd_address.find('/');
-    if (pos >= 0)
-    {
-        etcd_api_path = etcd_address.substr(pos);
-        etcd_address = etcd_address.substr(0, pos);
-    }
+        etcd_call_selected(api, payload, timeout, retries, interval, callback);
+    });
+}
+
+void etcd_state_client_t::etcd_call_selected(const std::string & api, json11::Json payload, int timeout,
+    int retries, int interval, std::function<void(std::string, json11::Json)> callback)
+{
+    const auto & url = selected_etcd_url;
     std::string req = payload.dump();
-    req = "POST "+etcd_api_path+api+" HTTP/1.1\r\n"
-        "Host: "+etcd_address+"\r\n"
+    req = "POST "+url.path+api+" HTTP/1.1\r\n"
+        "Host: "+url.hostname+"\r\n"
         "Content-Type: application/json\r\n"
         "Content-Length: "+std::to_string(req.size())+"\r\n"
         "Connection: keep-alive\r\n"
@@ -128,15 +132,15 @@ 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 = url.addr](http_message_t *response)
     {
         std::string err;
         json11::Json data;
         response->parse_json_response(err, data);
         if (err != "")
         {
-            if (cur_addr == selected_etcd_address)
-                selected_etcd_address = "";
+            if (cur_addr == selected_etcd_url.addr)
+                selected_etcd_url = (http_url_t){};
             if (retries > 0)
             {
                 if (this->log_level > 0)
@@ -164,54 +168,58 @@ 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(get_http_ctx());
+    http_request(keepalive_client, url.addr, req, { .timeout = timeout, .keepalive = true, .ssl = url.ssl }, cb);
 }
 
-void etcd_state_client_t::add_etcd_url(std::string addr)
+
+void etcd_state_client_t::add_etcd_url(std::string etcd_address)
 {
-    if (addr.length() > 0)
+    if (etcd_address.size() > 0)
     {
-        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);
-        }
         if (!local_ips.size())
-            local_ips = getifaddr_list(std::vector<addr_mask_t>(), true);
-        std::string check_addr;
-        int pos = addr.find('/');
-        int pos2 = addr.find(':');
-        if (pos2 >= 0)
-            check_addr = addr.substr(0, pos2);
-        else if (pos >= 0)
-            check_addr = addr.substr(0, pos);
-        else
-            check_addr = addr;
-        if (pos == std::string::npos)
-            addr += "/v3";
-        bool local = false;
-        int i;
-        for (i = 0; i < local_ips.size(); i++)
         {
-            if (local_ips[i] == check_addr)
-            {
-                local = true;
-                break;
-            }
+            // Fill local_ips
+            for (auto & ip: getifaddr_list(std::vector<addr_mask_t>(), true))
+                local_ips.insert(ip);
         }
-        auto & to = local ? this->etcd_local : this->etcd_addresses;
+        std::string etcd_api_path;
+        bool ssl = false;
+        if (etcd_address.substr(0, 8) == "https://")
+        {
+            ssl = true;
+            etcd_address = etcd_address.substr(8);
+        }
+        else if (etcd_address.substr(0, 7) == "http://")
+            etcd_address = etcd_address.substr(7);
+        auto pos = etcd_address.find('/');
+        if (pos != std::string::npos)
+        {
+            etcd_api_path = etcd_address.substr(pos);
+            etcd_address = etcd_address.substr(0, pos);
+        }
+        else
+            etcd_api_path = "/v3";
+        pos = etcd_address.find(':');
+        auto check_addr = (pos != std::string::npos ? etcd_address.substr(0, pos) : etcd_address);
+        bool is_local = local_ips.find(check_addr) != local_ips.end();
+        auto & to = (is_local ? etcd_local : etcd_addresses);
+        check_addr = (ssl ? "https://" : "http://") + etcd_address + etcd_api_path;
+        size_t i;
         for (i = 0; i < to.size(); i++)
         {
-            if (to[i] == addr)
+            if (to[i] == check_addr)
                 break;
         }
         if (i >= to.size())
-            to.push_back(addr);
+        {
+            to.push_back(check_addr);
+            // Check if it's a domain name
+            sockaddr_storage ss;
+            bool is_name = !is_local && !string_to_addr(etcd_address, true, 0, &ss);
+            auto & to_addr = (is_local ? etcd_local_addr_urls : (is_name ? etcd_name_urls : etcd_nonlocal_addr_urls));
+            to_addr.push_back((http_url_t){ .ssl = ssl, .addr = etcd_address, .hostname = etcd_address, .path = etcd_api_path });
+        }
     }
 }
 
@@ -219,6 +227,9 @@ void etcd_state_client_t::parse_config(const json11::Json & config)
 {
     this->etcd_local.clear();
     this->etcd_addresses.clear();
+    this->etcd_local_addr_urls.clear();
+    this->etcd_nonlocal_addr_urls.clear();
+    this->etcd_name_urls.clear();
     if (config["etcd_address"].is_string())
     {
         std::string ea = config["etcd_address"].string_value();
@@ -239,6 +250,17 @@ void etcd_state_client_t::parse_config(const json11::Json & config)
             add_etcd_url(ea.string_value());
         }
     }
+    if (this->osd_num)
+    {
+        this->etcd_client_cert = config["osd_etcd_client_cert"].string_value();
+        this->etcd_client_key = config["osd_etcd_client_key"].string_value();
+    }
+    else
+    {
+        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 == "")
     {
@@ -282,6 +304,7 @@ void etcd_state_client_t::parse_config(const json11::Json & config)
     {
         this->etcd_min_reload_interval = 50;
     }
+    this->use_image_index_for_auth = config["use_image_index_for_auth"].bool_value();
     if (this->etcd_ws_keepalive_interval != old_etcd_ws_keepalive_interval && ws_keepalive_timer >= 0)
     {
 #ifndef __MOCK__
@@ -291,66 +314,131 @@ void etcd_state_client_t::parse_config(const json11::Json & config)
     }
 }
 
-void etcd_state_client_t::pick_next_etcd()
-{
-    if (selected_etcd_address != "")
-        return;
-    if (addresses_to_try.size() == 0)
-    {
-        // Prefer local etcd, if any
-        for (int i = 0; i < etcd_local.size(); i++)
-            addresses_to_try.push_back(etcd_local[i]);
-        std::vector<int> ns;
-        for (int i = 0; i < etcd_addresses.size(); i++)
-            ns.push_back(i);
-        if (!rand_initialized)
-        {
-            timespec tv;
-            clock_gettime(CLOCK_REALTIME, &tv);
-            srand48(tv.tv_sec*1000000000 + tv.tv_nsec);
-            rand_initialized = true;
-        }
-        while (ns.size())
-        {
-            int i = lrand48() % ns.size();
-            addresses_to_try.push_back(etcd_addresses[ns[i]]);
-            ns.erase(ns.begin()+i, ns.begin()+i+1);
-        }
-    }
-    selected_etcd_address = addresses_to_try[0];
-    addresses_to_try.erase(addresses_to_try.begin(), addresses_to_try.begin()+1);
-}
-
-void etcd_state_client_t::start_etcd_watcher()
+void etcd_state_client_t::pick_next_etcd(std::function<void()> cb)
 {
     if (!etcd_addresses.size() && !etcd_local.size())
     {
         fprintf(stderr, "etcd_address is missing in Vitastor configuration\n");
         exit(1);
     }
-    pick_next_etcd();
-    std::string etcd_address = selected_etcd_address;
-    std::string etcd_api_path;
-    int pos = etcd_address.find('/');
-    if (pos >= 0)
+    if (selected_etcd_url.addr != "")
     {
-        etcd_api_path = etcd_address.substr(pos);
-        etcd_address = etcd_address.substr(0, pos);
+        cb();
+        return;
     }
+    if (etcd_urls_to_try.size() != 0)
+    {
+        selected_etcd_url = std::move(etcd_urls_to_try[0]);
+        etcd_urls_to_try.erase(etcd_urls_to_try.begin());
+        cb();
+        return;
+    }
+    on_resolve_queue.push_back(std::move(cb));
+    if (on_resolve_queue.size() > 1)
+    {
+        // Already resolving
+        return;
+    }
+    assert(!resolve_count);
+    local_to_try = 0;
+    for (auto & url: etcd_local_addr_urls)
+    {
+        // Prefer local IPs, if any
+        etcd_urls_to_try.push_back(url);
+        local_to_try++;
+    }
+    for (auto & url: etcd_nonlocal_addr_urls)
+    {
+        etcd_urls_to_try.push_back(url);
+    }
+    resolve_count++;
+    for (auto & url: etcd_name_urls)
+    {
+        resolve_count++;
+        http_resolve(get_http_ctx(), url.ssl, url.addr, [this, url](const std::string & error, const std::vector<std::string>& addresses)
+        {
+            if (error != "")
+                fprintf(stderr, "Error resolving %s: %s\n", url.addr.c_str(), error.c_str());
+            for (auto & addr: addresses)
+            {
+                auto url_copy = url;
+                url_copy.addr = addr;
+                if (local_ips.find(addr) != local_ips.end())
+                {
+                    etcd_urls_to_try.insert(etcd_urls_to_try.begin(), std::move(url_copy));
+                    local_to_try++;
+                }
+                else
+                    etcd_urls_to_try.push_back(std::move(url_copy));
+            }
+            resolve_count--;
+            if (!resolve_count)
+                pick_next_etcd_on_resolve();
+        });
+    }
+    resolve_count--;
+    if (!resolve_count)
+    {
+        pick_next_etcd_on_resolve();
+    }
+}
+
+void etcd_state_client_t::pick_next_etcd_on_resolve()
+{
+    if (!etcd_urls_to_try.size())
+    {
+        fprintf(stderr, "None of etcd_address could be resolved\n");
+        exit(1);
+    }
+    if (!rand_initialized)
+    {
+        timespec tv;
+        clock_gettime(CLOCK_REALTIME, &tv);
+        srand48(tv.tv_sec*1000000000 + tv.tv_nsec);
+        rand_initialized = true;
+    }
+    // Shuffle addresses
+    for (size_t i = etcd_urls_to_try.size()-1; i > local_to_try; i--)
+    {
+        size_t j = local_to_try + lrand48() % (i - local_to_try);
+        if (j != i)
+            std::swap(etcd_urls_to_try[i], etcd_urls_to_try[j]);
+    }
+    selected_etcd_url = std::move(etcd_urls_to_try[0]);
+    etcd_urls_to_try.erase(etcd_urls_to_try.begin());
+    auto cbs = std::move(on_resolve_queue);
+    for (auto cb: cbs)
+    {
+        cb();
+    }
+}
+
+void etcd_state_client_t::start_etcd_watcher()
+{
+    pick_next_etcd([this]()
+    {
+        start_etcd_watcher_selected();
+    });
+}
+
+void etcd_state_client_t::start_etcd_watcher_selected()
+{
+    const auto & url = selected_etcd_url;
     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(),
+        fprintf(stderr, "Trying to connect to etcd websocket at %s%s%s (hostname %s), watch from revision %ju/%ju/%ju\n",
+            url.ssl ? "https://" : "http://", url.addr.c_str(), url.path.c_str(), url.hostname.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(get_http_ctx());
+    else
+        http_close(etcd_watch_ws);
+    etcd_total_watches = (use_image_index_for_auth ? 4 : 3);
+    open_websocket(etcd_watch_ws, url.addr, url.hostname, url.path+"/watch", { .timeout = etcd_slow_timeout, .ssl = url.ssl },
+        [this, cur_addr = url.addr](http_message_t *msg)
     {
         if (msg->body.length())
         {
@@ -368,11 +456,12 @@ void etcd_state_client_t::start_etcd_watcher()
                 {
                     if (watch_id == ETCD_CONFIG_WATCH_ID ||
                         watch_id == ETCD_PG_STATE_WATCH_ID ||
-                        watch_id == ETCD_OSD_STATE_WATCH_ID)
+                        watch_id == ETCD_OSD_STATE_WATCH_ID ||
+                        watch_id == ETCD_INDEX_WATCH_ID)
                     {
                         etcd_watches_initialised++;
                     }
-                    if (etcd_watches_initialised == ETCD_TOTAL_WATCHES && this->log_level > 0)
+                    if (etcd_watches_initialised == etcd_total_watches && this->log_level > 0)
                     {
                         fprintf(stderr, "Successfully subscribed to etcd at %s, revision %ju/%ju/%ju\n", cur_addr.c_str(),
                             etcd_watch_revision_config, etcd_watch_revision_osd, etcd_watch_revision_pg);
@@ -393,8 +482,7 @@ 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;
+                                etcd_watch_revision_config = etcd_watch_revision_osd = etcd_watch_revision_pg = etcd_watch_revision_index = 0;
                                 on_reload_hook();
                             }
                             return;
@@ -414,7 +502,10 @@ 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"}}
-                if (etcd_watches_initialised == ETCD_TOTAL_WATCHES && !data["result"]["header"]["revision"].is_null())
+                // 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())
                 {
                     // 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,
@@ -435,7 +526,9 @@ void etcd_state_client_t::start_etcd_watcher()
                         etcd_watch_revision_pg = watch_rev;
                     else if (watch_id == ETCD_OSD_STATE_WATCH_ID)
                         etcd_watch_revision_osd = watch_rev;
-                    addresses_to_try.clear();
+                    else if (watch_id == ETCD_INDEX_WATCH_ID)
+                        etcd_watch_revision_index = watch_rev;
+                    etcd_urls_to_try.clear();
                 }
                 // First gather all changes into a hash to remove multiple overwrites
                 std::map<std::string, etcd_kv_t> changes;
@@ -465,13 +558,8 @@ void etcd_state_client_t::start_etcd_watcher()
         if (msg->eof)
         {
             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 (cur_addr == selected_etcd_url.addr)
+                selected_etcd_url = (http_url_t){};
             if (etcd_watches_initialised == 0)
             {
                 // Connection not established, retry in <etcd_quick_timeout>
@@ -515,6 +603,18 @@ void etcd_state_client_t::start_etcd_watcher()
             { "progress_notify", true },
         } }
     }).dump());
+    if (use_image_index_for_auth)
+    {
+        http_post_message(etcd_watch_ws, WS_TEXT, json11::Json(json11::Json::object {
+            { "create_request", json11::Json::object {
+                { "key", base64_encode(etcd_prefix+"/index/image/") },
+                { "range_end", base64_encode(etcd_prefix+"/index/image0") },
+                { "start_revision", etcd_watch_revision_index },
+                { "watch_id", ETCD_INDEX_WATCH_ID },
+                { "progress_notify", true },
+            } }
+        }).dump());
+    }
     // FIXME: Do not watch /pg/history/ at all in client code (not in OSD)
     if (on_start_watcher_hook)
     {
@@ -538,7 +638,7 @@ void etcd_state_client_t::start_ws_keepalive()
     {
         ws_keepalive_timer = tfd->set_timer(etcd_ws_keepalive_interval*1000, true, [this](int)
         {
-            if (!etcd_watch_ws || etcd_watches_initialised < ETCD_TOTAL_WATCHES)
+            if (!etcd_watch_ws || etcd_watches_initialised < etcd_total_watches)
             {
                 // Do nothing
             }
@@ -546,12 +646,7 @@ void etcd_state_client_t::start_ws_keepalive()
             {
                 if (this->log_level > 0)
                 {
-                    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;
+                    fprintf(stderr, "Websocket ping failed, disconnecting from etcd %s\n", selected_etcd_url.addr.c_str());
                 }
                 start_etcd_watcher();
             }
@@ -689,6 +784,15 @@ void etcd_state_client_t::load_pgs()
             } }
         },
     };
+    if (use_image_index_for_auth)
+    {
+        txn.push_back(json11::Json::object {
+            { "request_range", json11::Json::object {
+                { "key", base64_encode(etcd_prefix+"/index/image/") },
+                { "range_end", base64_encode(etcd_prefix+"/index/image0") },
+            } }
+        });
+    }
     json11::Json::object req = { { "success", txn } };
     json11::Json checks = load_pgs_checks_hook != NULL ? load_pgs_checks_hook() : json11::Json();
     if (checks.array_items().size() > 0)
@@ -1262,33 +1366,40 @@ void etcd_state_client_t::parse_state(const etcd_kv_t & kv)
             }
             else
             {
-                inode_t parent_inode_num = value["parent_id"].uint64_value();
-                if (parent_inode_num && !(parent_inode_num >> (64-POOL_ID_BITS)))
+                insert_inode_config(deserialize_inode_cfg(inode_num, kv.value, kv.mod_revision));
+            }
+        }
+    }
+    else if (key.substr(0, etcd_prefix.length()+13) == etcd_prefix+"/index/image/")
+    {
+        // <etcd_prefix>/index/image/%s
+        // Used for environments based on etcd authentication to grant client image access
+        // by names - granting by ID is inconvenient because of snapshots
+        std::string inode_name = key.substr(etcd_prefix.length()+13);
+        if (!value.is_object())
+        {
+            auto n_it = this->inode_by_name.find(inode_name);
+            if (n_it != this->inode_by_name.end())
+            {
+                uint64_t inode_num = n_it->second;
+                if (on_inode_change_hook != NULL)
                 {
-                    uint64_t parent_pool_id = value["parent_pool"].uint64_value();
-                    if (!parent_pool_id)
-                        parent_inode_num |= pool_id << (64-POOL_ID_BITS);
-                    else if (parent_pool_id >= POOL_ID_MAX)
-                    {
-                        fprintf(
-                            stderr, "Inode %ju/%ju parent_pool value is invalid, ignoring parent setting\n",
-                            inode_num >> (64-POOL_ID_BITS), inode_num & (((uint64_t)1 << (64-POOL_ID_BITS)) - 1)
-                        );
-                        parent_inode_num = 0;
-                    }
-                    else
-                        parent_inode_num |= parent_pool_id << (64-POOL_ID_BITS);
+                    on_inode_change_hook(inode_num, true);
                 }
-                insert_inode_config((inode_config_t){
-                    .num = inode_num,
-                    .name = value["name"].string_value(),
-                    .size = value["size"].uint64_value(),
-                    .parent_id = parent_inode_num,
-                    .readonly = value["readonly"].bool_value(),
-                    .deleted = value["deleted"].bool_value(),
-                    .meta = value["meta"],
-                    .mod_revision = kv.mod_revision,
-                });
+                this->inode_config.erase(inode_num);
+                this->inode_by_name.erase(n_it);
+            }
+        }
+        else if (value["name"] == inode_name)
+        {
+            uint64_t inode_num = INODE_WITH_POOL(value["pool_id"].uint64_value(), value["id"].uint64_value());
+            if (!INODE_POOL(inode_num) || !INODE_NO_POOL(inode_num))
+            {
+                fprintf(stderr, "Invalid inode ID in key %s: %ju\n", key.c_str(), inode_num);
+            }
+            else
+            {
+                insert_inode_config(deserialize_inode_cfg(inode_num, value, kv.mod_revision));
             }
         }
     }
@@ -1377,6 +1488,10 @@ json11::Json::object etcd_state_client_t::serialize_inode_cfg(inode_config_t *cf
             new_cfg["parent_pool"] = (uint64_t)INODE_POOL(cfg->parent_id);
         new_cfg["parent_id"] = (uint64_t)INODE_NO_POOL(cfg->parent_id);
     }
+    if (!cfg->enc_key.empty())
+    {
+        new_cfg["enc_key"] = cfg->enc_key;
+    }
     if (cfg->readonly)
     {
         new_cfg["readonly"] = true;
@@ -1392,6 +1507,50 @@ json11::Json::object etcd_state_client_t::serialize_inode_cfg(inode_config_t *cf
     return new_cfg;
 }
 
+inode_config_t etcd_state_client_t::deserialize_inode_cfg(uint64_t inode_num, json11::Json value, uint64_t mod_revision)
+{
+    inode_t parent_inode_num = value["parent_id"].uint64_value();
+    if (parent_inode_num && !INODE_POOL(parent_inode_num))
+    {
+        uint64_t parent_pool_id = value["parent_pool"].uint64_value();
+        if (!parent_pool_id)
+            parent_inode_num = INODE_WITH_POOL(INODE_POOL(inode_num), parent_inode_num);
+        else if (parent_pool_id >= POOL_ID_MAX)
+        {
+            fprintf(
+                stderr, "Inode %u/%ju parent_pool value is invalid, ignoring parent setting\n",
+                INODE_POOL(inode_num), INODE_NO_POOL(inode_num)
+            );
+            parent_inode_num = 0;
+        }
+        else
+            parent_inode_num |= parent_pool_id << (64-POOL_ID_BITS);
+    }
+    std::string enc_key;
+    if (!value["enc_key"].is_null())
+    {
+        enc_key = value["enc_key"].string_value();
+        if (enc_key.substr(0, strlen(VAULT_KEY_PREFIX)) != VAULT_KEY_PREFIX &&
+            (enc_key.size() != 2*AES_256_XTS_KEY_SIZE || !ishexstr(enc_key)))
+        {
+            enc_key = "";
+            fprintf(stderr, "Inode %u/%ju has invalid enc_key, should be %u bit hex string or Vault key reference\n",
+                INODE_POOL(inode_num), INODE_NO_POOL(inode_num), AES_256_XTS_KEY_SIZE);
+        }
+    }
+    return (inode_config_t){
+        .num = inode_num,
+        .name = value["name"].string_value(),
+        .size = value["size"].uint64_value(),
+        .parent_id = parent_inode_num,
+        .readonly = value["readonly"].bool_value(),
+        .deleted = value["deleted"].bool_value(),
+        .enc_key = std::move(enc_key),
+        .meta = value["meta"],
+        .mod_revision = mod_revision,
+    };
+}
+
 int etcd_state_client_t::address_count()
 {
     return etcd_addresses.size() + etcd_local.size();

src/client/etcd_state_client.h

@@ -4,21 +4,24 @@
 #pragma once
 
 #include <set>
+#include <memory>
 
 #include "json11/json11.hpp"
-#include "osd_id.h"
+#include "object_id.h"
 #include "timerfd_manager.h"
 
 #define ETCD_CONFIG_WATCH_ID 1
 #define ETCD_OSD_STATE_WATCH_ID 2
 #define ETCD_PG_STATE_WATCH_ID 3
-#define ETCD_TOTAL_WATCHES 3
+#define ETCD_INDEX_WATCH_ID 4
 
 #define DEFAULT_BLOCK_SIZE 128*1024
 #define MIN_DATA_BLOCK_SIZE 4*1024
 #define MAX_DATA_BLOCK_SIZE 128*1024*1024
 #define DEFAULT_BITMAP_GRANULARITY 4096
 
+#define VAULT_KEY_PREFIX "vault:"
+
 #ifndef IMMEDIATE_NONE
 #define IMMEDIATE_NONE 0
 #define IMMEDIATE_SMALL 1
@@ -83,6 +86,7 @@ struct inode_config_t
     inode_t parent_id = 0;
     bool readonly = false;
     bool deleted = false;
+    std::string enc_key;
     // Arbitrary metadata
     json11::Json meta;
     // Change revision of the metadata in etcd
@@ -95,23 +99,42 @@ struct inode_watch_t
     inode_config_t cfg = {};
 };
 
+struct http_url_t
+{
+    bool ssl;
+    std::string addr;
+    std::string hostname;
+    std::string path;
+};
+
 struct http_co_t;
+struct http_context_t;
 
 struct __attribute__((visibility("default"))) etcd_state_client_t
 {
 protected:
-    std::vector<std::string> local_ips;
-    std::vector<std::string> etcd_addresses;
+    std::set<std::string> local_ips;
     std::vector<std::string> etcd_local;
-    std::string selected_etcd_address;
-    std::vector<std::string> addresses_to_try;
+    std::vector<std::string> etcd_addresses;
+    std::vector<http_url_t> etcd_local_addr_urls;
+    std::vector<http_url_t> etcd_nonlocal_addr_urls;
+    std::vector<http_url_t> etcd_name_urls;
+    size_t local_to_try = 0;
+    std::vector<http_url_t> etcd_urls_to_try;
+    http_url_t selected_etcd_url;
+    size_t resolve_count = 0;
     std::vector<inode_watch_t*> watches;
+    std::vector<std::function<void()>> on_resolve_queue;
     bool new_pg_config = false;
+    bool use_image_index_for_auth = false;
     int ws_keepalive_timer = -1;
     int ws_alive = 0;
     bool rand_initialized = false;
     void add_etcd_url(std::string);
-    void pick_next_etcd();
+    void pick_next_etcd(std::function<void()> cb);
+    void pick_next_etcd_on_resolve();
+    void etcd_call_selected(const std::string & api, json11::Json payload, int timeout, int retries, int interval, std::function<void(std::string, json11::Json)> callback);
+    void start_etcd_watcher_selected();
 public:
     int etcd_keepalive_timeout = 30;
     int etcd_ws_keepalive_interval = 5;
@@ -124,15 +147,22 @@ public:
     uint32_t global_bitmap_granularity = DEFAULT_BITMAP_GRANULARITY;
     uint32_t global_immediate_commit = IMMEDIATE_NONE;
 
+    uint64_t osd_num = 0;
     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;
+    int etcd_total_watches = 0;
     uint64_t etcd_watch_revision_config = 0;
     uint64_t etcd_watch_revision_osd = 0;
     uint64_t etcd_watch_revision_pg = 0;
+    uint64_t etcd_watch_revision_index = 0;
     timespec etcd_last_reload = {};
     int load_pgs_timer_id = -1;
     std::map<pool_id_t, pool_config_t> pool_config;
@@ -158,10 +188,12 @@ public:
     std::function<void(http_co_t *)> on_start_watcher_hook;
 
     json11::Json::object serialize_inode_cfg(inode_config_t *cfg);
+    inode_config_t deserialize_inode_cfg(uint64_t inode_num, json11::Json value, uint64_t mod_revision);
     etcd_kv_t parse_etcd_kv(const json11::Json & kv_json);
     std::vector<std::string> get_addresses();
-    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);
+    http_context_t *get_http_ctx();
+    void etcd_call_oneshot(const std::string & etcd_address, const std::string & api, json11::Json payload, int timeout, std::function<void(std::string, json11::Json)> callback);
+    void etcd_call(const 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);
     void etcd_txn_slow(json11::Json txn, std::function<void(std::string, json11::Json)> callback);
     void start_etcd_watcher();

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 +46,25 @@ 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(timerfd_manager_t *tfd, const std::string & ssl_cert, const std::string & ssl_key,
+    const std::string & ssl_ca, bool verify_peer, std::string & error);
+void http_resolve(http_context_t *ctx, bool ssl, std::string host,
+    std::function<void(const std::string & error, const std::vector<std::string> & addrs)> cb);
+void http_context_destroy(http_context_t *ctx);
+http_co_t* http_init(http_context_t *ctx = NULL);
+void open_websocket(http_co_t *handler, const std::string & addr, const std::string & hostname, 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_get(http_co_t *handler, const std::string & url, const std::string & headers,
+    const http_options_t & options, std::function<void(http_message_t *response)> response_callback);
+void http_json_post(http_co_t *handler, const std::string & url, json11::Json body, const std::string & headers,
+    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

@@ -145,7 +145,7 @@ void osd_messenger_t::init()
         rdma_contexts = msgr_rdma_context_t::create_all(
             osd_num && osd_cluster_network_masks.size() ? osd_cluster_network_masks : osd_network_masks,
             rdma_device != "" ? rdma_device.c_str() : NULL,
-            rdma_port_num, rdma_gid_index, rdma_mtu, rdma_odp, log_level
+            rdma_port_num, rdma_gid_index, rdma_mtu, log_level
         );
         if (!rdma_contexts.size())
         {
@@ -289,6 +289,14 @@ osd_messenger_t::~osd_messenger_t()
         rdmacm_evch = NULL;
     }
 #endif
+    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);
+    }
 }
 
 void osd_messenger_t::parse_config(const json11::Json & config)
@@ -322,8 +330,10 @@ void osd_messenger_t::parse_config(const json11::Json & config)
     this->rdma_max_msg = config["rdma_max_msg"].uint64_value();
     if (!this->rdma_max_msg || this->rdma_max_msg > 128*1024*1024)
         this->rdma_max_msg = 129*1024;
-    this->rdma_odp = config["rdma_odp"].bool_value();
 #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
@@ -503,7 +513,7 @@ 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
@@ -740,14 +750,6 @@ 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);
             }
@@ -782,7 +784,7 @@ 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 +45,11 @@ struct msgr_rdma_connection_t;
 struct msgr_rdma_context_t;
 #endif
 
+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);
+
 struct osd_client_t
 {
     int refs = 0;
@@ -63,7 +65,7 @@ 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;
@@ -72,14 +74,18 @@ struct osd_client_t
     // Read state
     int read_ready = 0;
     osd_op_t *read_op = NULL;
+    size_t read_op_size = 0;
+    size_t read_op_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;
+    op_aes_xts_decrypt_t *decrypt_ctx = NULL;
+    size_t read_op_inline_decrypt_pos = 0;
+    size_t read_op_inline_decrypt_in = 0;
 
     // Incoming operations
     std::vector<osd_op_t*> received_ops;
@@ -92,11 +98,16 @@ struct osd_client_t
     std::set<pool_pg_num_t> dirty_pgs;
 
     // Write state
+    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;
+    op_aes_xts_encrypt_t *encrypt_ctx = NULL;
 
     ~osd_client_t();
     void cancel_ops();
@@ -188,6 +199,7 @@ 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;
@@ -200,7 +212,6 @@ protected:
     std::vector<msgr_rdma_context_t *> rdma_contexts;
     uint64_t rdma_max_sge = 0, rdma_max_send = 0, rdma_max_recv = 0;
     uint64_t rdma_max_msg = 0;
-    bool rdma_odp = false;
     rdma_event_channel *rdmacm_evch = NULL;
     std::map<rdma_cm_id*, osd_client_t*> rdmacm_connections;
     std::map<rdma_cm_id*, rdmacm_connecting_t*> rdmacm_connecting;
@@ -210,7 +221,10 @@ 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::vector<osd_op_t*> set_immediate_ops;
+    std::deque<osd_op_t*> set_immediate_ops;
+
+    std::vector<op_aes_xts_encrypt_t*> encrypt_ctx_pool;
+    std::vector<op_aes_xts_decrypt_t*> decrypt_ctx_pool;
 
 public:
     timerfd_manager_t *tfd = NULL;
@@ -275,9 +289,24 @@ protected:
 
     bool try_send(osd_client_t *cl);
     void handle_send(int result, bool prev, bool more, osd_client_t *cl);
+    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);
+    size_t op_get_read_buffers(osd_client_t *cl, std::vector<iovec> & lst);
+    void handle_finished_op(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);
+    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);
 
-    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);
@@ -287,8 +316,7 @@ protected:
 
 #ifdef WITH_RDMA
     void try_send_rdma(osd_client_t *cl);
-    void try_send_rdma_odp(osd_client_t *cl);
-    void try_send_rdma_nodp(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 +1,474 @@
+// Copyright (c) Vitaliy Filippov, 2026+
+// License: VNPL-1.1 or GNU GPL-2.0+ (see README.md for details)
+
+#include <assert.h>
+
+#include "etcd_state_client.h"
+#include "messenger.h"
+#include "msgr_encrypt.h"
+
+op_aes_xts_encrypt_t::op_aes_xts_encrypt_t()
+{
+#ifdef WITH_OPENSSL
+    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();
+    }
+#else
+    fprintf(stderr, "Error: Vitastor is built without encryption support\n");
+    abort();
+#endif
+}
+
+op_aes_xts_encrypt_t::~op_aes_xts_encrypt_t()
+{
+    assert(!encrypted);
+#ifdef WITH_OPENSSL
+    EVP_CIPHER_CTX_free(ctx);
+#endif
+    if (tmp)
+        free(tmp);
+}
+
+void op_aes_xts_encrypt_t::start(uint8_t *key, uint64_t start_offset, size_t block_size)
+{
+    assert(!encrypted);
+    this->start_offset = start_offset;
+    this->key = key;
+    this->block_size = block_size;
+    this->offset = 0;
+    this->encrypted = false;
+    this->tmp_pos = 0;
+    if (tmp && tmp_size != block_size)
+    {
+        free(tmp);
+        tmp = NULL;
+        tmp_size = 0;
+    }
+#ifdef WITH_OPENSSL
+    if (EVP_EncryptInit_ex(ctx, NULL, NULL, key, NULL) != 1)
+    {
+        ERR_print_errors_fp(stderr);
+        abort();
+    }
+#endif
+}
+
+void op_aes_xts_encrypt_t::encrypt_block(uint8_t *in, uint8_t *out)
+{
+#ifdef WITH_OPENSSL
+    uint8_t iv[16] = { 0 };
+    *((uint64_t*)iv) = start_offset + offset - offset%block_size;
+    if (EVP_EncryptInit_ex(ctx, NULL, NULL, NULL, iv) != 1)
+    {
+        ERR_print_errors_fp(stderr);
+        abort();
+    }
+    int actual_out = 0;
+    if (EVP_EncryptUpdate(ctx, out, &actual_out, in, block_size) != 1)
+    {
+        ERR_print_errors_fp(stderr);
+        abort();
+    }
+    assert(actual_out == block_size);
+#endif
+}
+
+// FIXME: Copy-paste
+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)
+{
+    // Fucking AES-XTS implementations (all of them) don't have streaming support,
+    // crafting IV to resume encryption is slow, so we have to accumulate a full block
+    // and encrypt it at once :-(
+    // And then we have to support consuming it in parts because it's simpler for the
+    // higher layers.
+    if (encrypted)
+    {
+        // Copy accumulated and encrypted output
+        assert(tmp);
+        if (max_out > block_size - tmp_pos)
+            max_out = block_size - tmp_pos;
+        memcpy(out, tmp + tmp_pos, max_out);
+        done_out += max_out;
+        tmp_pos += max_out;
+        if (tmp_pos >= block_size)
+            encrypted = false;
+    }
+    else if (max_in < block_size - offset%block_size)
+    {
+        // Just accumulate input
+        if (!tmp)
+        {
+            tmp = (uint8_t*)malloc_or_die(block_size);
+            tmp_size = block_size;
+        }
+        memcpy(tmp + offset%block_size, in, max_in);
+        done_in += max_in;
+        offset += max_in;
+    }
+    else if (max_out < block_size)
+    {
+        // Accumulate and encrypt input in <tmp>, then copy part of it to <out>
+        if (!tmp)
+        {
+            tmp = (uint8_t*)malloc_or_die(block_size);
+            tmp_size = block_size;
+        }
+        max_in = block_size - offset%block_size;
+        memcpy(tmp + offset%block_size, in, max_in);
+        encrypt_block(tmp, tmp);
+        encrypted = true;
+        memcpy(out, tmp, max_out);
+        tmp_pos = max_out;
+        done_in += max_in;
+        offset += max_in;
+        done_out += max_out;
+    }
+    else if (!(offset%block_size))
+    {
+        // Full block - simplest case
+        encrypt_block(in, out);
+        done_in += block_size;
+        offset += block_size;
+        done_out += block_size;
+    }
+    else
+    {
+        // Accumulate input and encrypt directly to <output>
+        assert(tmp);
+        max_in = block_size - offset%block_size;
+        memcpy(tmp + offset%block_size, in, max_in);
+        encrypt_block(tmp, out);
+        done_in += max_in;
+        offset += max_in;
+        done_out += block_size;
+    }
+}
+
+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()
+{
+#ifdef WITH_OPENSSL
+    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();
+    }
+#else
+    fprintf(stderr, "Error: Vitastor is built without encryption support\n");
+    abort();
+#endif
+}
+
+op_aes_xts_decrypt_t::~op_aes_xts_decrypt_t()
+{
+    assert(!decrypted);
+#ifdef WITH_OPENSSL
+    EVP_CIPHER_CTX_free(ctx);
+#endif
+    if (tmp)
+        free(tmp);
+}
+
+void op_aes_xts_decrypt_t::start(uint8_t **key_chain, size_t chain_size, uint8_t *key_indexes, uint64_t start_offset, size_t block_size)
+{
+    assert(!decrypted);
+    this->start_offset = start_offset;
+    this->key_chain = chain_size > 1 ? key_chain : 0;
+    this->chain_size = chain_size > 1 ? chain_size : 0;
+    this->key_indexes = chain_size > 1 ? key_indexes : NULL;
+    assert(chain_size <= 1 || key_indexes != NULL);
+    this->block_size = block_size;
+    this->offset = 0;
+    this->tmp_pos = 0;
+    if (tmp && tmp_size != block_size)
+    {
+        free(tmp);
+        tmp = NULL;
+        tmp_size = 0;
+    }
+#ifdef WITH_OPENSSL
+    if (chain_size == 1 && key_chain[0] && EVP_DecryptInit_ex(ctx, NULL, NULL, key_chain[0], NULL) != 1)
+    {
+        ERR_print_errors_fp(stderr);
+        abort();
+    }
+#endif
+}
+
+void op_aes_xts_decrypt_t::decrypt_block(uint8_t *in, uint8_t *out)
+{
+    uint8_t *key = NULL;
+    if (chain_size)
+    {
+        assert(key_indexes[offset/block_size] < chain_size);
+        key = key_chain[key_indexes[offset/block_size]];
+        if (!key)
+        {
+            if (in != out)
+                memcpy(out, in, block_size);
+            return;
+        }
+    }
+#ifdef WITH_OPENSSL
+    uint8_t iv[16] = { 0 };
+    *((uint64_t*)iv) = start_offset + offset - offset%block_size;
+    if (EVP_DecryptInit_ex(ctx, NULL, NULL, key, iv) != 1)
+    {
+        ERR_print_errors_fp(stderr);
+        abort();
+    }
+    int actual_out = 0;
+    if (EVP_DecryptUpdate(ctx, out, &actual_out, in, block_size) != 1)
+    {
+        ERR_print_errors_fp(stderr);
+        abort();
+    }
+    assert(actual_out == block_size);
+#endif
+}
+
+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)
+{
+    // Fucking AES-XTS implementations (all of them) don't have streaming support,
+    // crafting IV to resume decryption is slow, so we have to accumulate a full block
+    // and decrypt it at once :-(
+    // And then we have to support consuming it in parts because clients sometimes need
+    // fragmented output.
+    if (decrypted)
+    {
+        // Copy accumulated and decrypted output
+        assert(tmp);
+        if (max_out > block_size - tmp_pos)
+            max_out = block_size - tmp_pos;
+        memcpy(out, tmp + tmp_pos, max_out);
+        done_out += max_out;
+        tmp_pos += max_out;
+        if (tmp_pos >= block_size)
+            decrypted = false;
+    }
+    else if (max_in < block_size - offset%block_size)
+    {
+        // Just accumulate input
+        if (!tmp)
+        {
+            tmp = (uint8_t*)malloc_or_die(block_size);
+            tmp_size = block_size;
+        }
+        memcpy(tmp + offset%block_size, in, max_in);
+        done_in += max_in;
+        offset += max_in;
+    }
+    else if (max_out < block_size)
+    {
+        // Accumulate and decrypt input in <tmp>, then copy part of it to <out>
+        if (!tmp)
+        {
+            tmp = (uint8_t*)malloc_or_die(block_size);
+            tmp_size = block_size;
+        }
+        max_in = block_size - offset%block_size;
+        memcpy(tmp + offset%block_size, in, max_in);
+        decrypt_block(tmp, tmp);
+        decrypted = true;
+        memcpy(out, tmp, max_out);
+        tmp_pos = max_out;
+        done_in += max_in;
+        offset += max_in;
+        done_out += max_out;
+    }
+    else if (!(offset%block_size))
+    {
+        // Full block - simplest case
+        decrypt_block(in, out);
+        done_in += block_size;
+        offset += block_size;
+        done_out += block_size;
+    }
+    else
+    {
+        // Accumulate input and decrypt directly to <output>
+        assert(tmp);
+        max_in = block_size - offset%block_size;
+        memcpy(tmp + offset%block_size, in, max_in);
+        decrypt_block(tmp, out);
+        done_in += max_in;
+        offset += max_in;
+        done_out += block_size;
+    }
+}
+
+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_chain[0]);
+        cl->encrypt_ctx->start(op->enc->key_chain[0], 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 || cl->encrypt_ctx->has_buffered())
+        {
+            if (done >= enc_len)
+                return false;
+            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;
+        }
+        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;
+    assert(op->req.hdr.opcode == OSD_OP_READ);
+    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)
+        {
+            if (done >= enc_len)
+                return false;
+            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;
+            cl->read_op_pos += done_out;
+            cl->read_op_inline_decrypt_in += done_in;
+            from += done_out;
+            if (!done_out)
+                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();
+        auto & enc = cl->read_op->enc;
+        cl->decrypt_ctx->start(enc->key_chain, enc->chain_size,
+            (cl->read_op->req.rw.flags & OSD_OP_RETURN_CHAIN) ? (uint8_t*)cl->read_op->bitmap + enc->read_chain_bitmap_pos : 0,
+            cl->read_op->req.rw.offset, 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_in;
+    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 = cl->read_op_inline_decrypt_pos - OSD_PACKET_SIZE - op->reply.rw.bitmap_len;
+    int j = 0;
+    while (j < op->iov.count && from_out >= op->iov.buf[j].iov_len)
+    {
+        from_out -= op->iov.buf[j].iov_len;
+        j++;
+    }
+    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 +1,67 @@
+// Copyright (c) Vitaliy Filippov, 2026+
+// License: VNPL-1.1 or GNU GPL-2.0+ (see README.md for details)
+
+#include <stdint.h>
+
+// WITH_OPENSSL is left to possibly support other crypto libraries
+#ifdef WITH_OPENSSL
+#include <openssl/conf.h>
+#include <openssl/evp.h>
+#include <openssl/err.h>
+#endif
+
+class op_aes_xts_encrypt_t
+{
+#ifdef WITH_OPENSSL
+    EVP_CIPHER_CTX *ctx = NULL;
+#endif
+    uint64_t start_offset = 0;
+    uint8_t *key = NULL;
+    size_t offset = 0;
+    size_t block_size = 0;
+    uint8_t *tmp = NULL;
+    size_t tmp_size = 0;
+    size_t tmp_pos = 0;
+    bool encrypted = false;
+
+    void encrypt_block(uint8_t *in, uint8_t *out);
+
+public:
+    op_aes_xts_encrypt_t();
+    ~op_aes_xts_encrypt_t();
+
+    inline bool has_buffered() { return encrypted; };
+    void start(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);
+
+class op_aes_xts_decrypt_t
+{
+#ifdef WITH_OPENSSL
+    EVP_CIPHER_CTX *ctx = NULL;
+#endif
+    uint64_t start_offset = 0;
+    uint8_t **key_chain = NULL;
+    size_t chain_size = 0;
+    uint8_t *key_indexes = NULL;
+    size_t offset = 0;
+    size_t block_size = 0;
+    uint8_t *tmp = NULL;
+    size_t tmp_size = 0;
+    size_t tmp_pos = 0;
+    bool decrypted = false;
+
+    void decrypt_block(uint8_t *in, uint8_t *out);
+
+public:
+    op_aes_xts_decrypt_t();
+    ~op_aes_xts_decrypt_t();
+
+    inline bool has_buffered() { return decrypted; };
+    void start(uint8_t **key_chain, size_t chain_size, uint8_t *key_indexes, 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);