Support RDMA devices without Implicit ODP using mlockall()

UPD: Seems it won't work because ibv_reg_mr() takes a permissions argument
and doesn't allow more permissions than allowed by the kernel for memory
mappings. So the only way to register all memory is probably to iterate
over /proc/PID/maps... :)

Mellanox docs mention that older MLNX_OFED emulated ODP, so maybe it's
still possible to use it, but it's not confirmed.

CMakeLists.txt

@@ -2,6 +2,6 @@ cmake_minimum_required(VERSION 2.8)
 
 project(vitastor)
 
-set(VERSION "0.6.13")
+set(VERSION "0.6.12")
 
 add_subdirectory(src)

csi/Makefile

@@ -1,4 +1,4 @@
-VERSION ?= v0.6.13
+VERSION ?= v0.6.12
 
 all: build push
 

csi/deploy/004-csi-nodeplugin.yaml

@@ -49,7 +49,7 @@ spec:
             capabilities:
               add: ["SYS_ADMIN"]
             allowPrivilegeEscalation: true
-          image: vitalif/vitastor-csi:v0.6.13
+          image: vitalif/vitastor-csi:v0.6.12
           args:
             - "--node=$(NODE_ID)"
             - "--endpoint=$(CSI_ENDPOINT)"

csi/deploy/007-csi-provisioner.yaml

@@ -116,7 +116,7 @@ spec:
             privileged: true
             capabilities:
               add: ["SYS_ADMIN"]
-          image: vitalif/vitastor-csi:v0.6.13
+          image: vitalif/vitastor-csi:v0.6.12
           args:
             - "--node=$(NODE_ID)"
             - "--endpoint=$(CSI_ENDPOINT)"

csi/src/config.go

@@ -5,7 +5,7 @@ package vitastor
 
 const (
     vitastorCSIDriverName    = "csi.vitastor.io"
-    vitastorCSIDriverVersion = "0.6.13"
+    vitastorCSIDriverVersion = "0.6.12"
 )
 
 // Config struct fills the parameters of request or user input

debian/changelog

@@ -1,4 +1,4 @@
-vitastor (0.6.13-1) unstable; urgency=medium
+vitastor (0.6.12-1) unstable; urgency=medium
 
   * RDMA support
   * Bugfixes

debian/vitastor.Dockerfile

@@ -33,8 +33,8 @@ RUN set -e -x; \
     mkdir -p /root/packages/vitastor-$REL; \
     rm -rf /root/packages/vitastor-$REL/*; \
     cd /root/packages/vitastor-$REL; \
-    cp -r /root/vitastor vitastor-0.6.13; \
+    cp -r /root/vitastor vitastor-0.6.12; \
-    cd vitastor-0.6.13; \
+    cd vitastor-0.6.12; \
     ln -s /root/fio-build/fio-*/ ./fio; \
     FIO=$(head -n1 fio/debian/changelog | perl -pe 's/^.*\((.*?)\).*$/$1/'); \
     ls /usr/include/linux/raw.h || cp ./debian/raw.h /usr/include/linux/raw.h; \
@@ -47,8 +47,8 @@ RUN set -e -x; \
     rm -rf a b; \
     echo "dep:fio=$FIO" > debian/fio_version; \
     cd /root/packages/vitastor-$REL; \
-    tar --sort=name --mtime='2020-01-01' --owner=0 --group=0 --exclude=debian -cJf vitastor_0.6.13.orig.tar.xz vitastor-0.6.13; \
+    tar --sort=name --mtime='2020-01-01' --owner=0 --group=0 --exclude=debian -cJf vitastor_0.6.12.orig.tar.xz vitastor-0.6.12; \
-    cd vitastor-0.6.13; \
+    cd vitastor-0.6.12; \
     V=$(head -n1 debian/changelog | perl -pe 's/^.*\((.*?)\).*$/$1/'); \
     DEBFULLNAME="Vitaliy Filippov <vitalif@yourcmc.ru>" dch -D $REL -v "$V""$REL" "Rebuild for $REL"; \
     DEB_BUILD_OPTIONS=nocheck dpkg-buildpackage --jobs=auto -sa; \

docs/params/network.yml

@@ -48,19 +48,28 @@
   type: string
   info: |
     RDMA device name to use for Vitastor OSD communications (for example,
-    "rocep5s0f0"). Please note that Vitastor RDMA requires Implicit On-Demand
+    "rocep5s0f0"). Please note that if your RDMA device doesn't support
-    Paging (Implicit ODP) and Scatter/Gather (SG) support from the RDMA device
+    Implicit ODP (Implicit On-Demand Paging) then all Vitastor OSDs and clients
-    to work. For example, Mellanox ConnectX-3 and older adapters don't have
+    will have to use mlockall() to lock all application memory to use RDMA.
-    Implicit ODP, so they're unsupported by Vitastor. Run `ibv_devinfo -v` as
+    In case of the native Vitastor QEMU driver with RDMA, all virtual machine
-    root to list available RDMA devices and their features.
+    memory will be locked if your RDMA device doesn't support Implicit ODP.
+
+    Notably, Mellanox ConnectX-3 and older adapters don't support Implicit ODP,
+    while ConnectX-4 and newer do. Run `ibv_devinfo -v` as root to list
+    available RDMA devices and their features.
   info_ru: |
     Название RDMA-устройства для связи с Vitastor OSD (например, "rocep5s0f0").
-    Имейте в виду, что поддержка RDMA в Vitastor требует функций устройства
+    Имейте в виду, что если ваше устройство не поддерживает Implicit ODP
-    Implicit On-Demand Paging (Implicit ODP) и Scatter/Gather (SG). Например,
+    (Implicit On-Demand Paging), то все OSD и клиенты Vitastor будут вынуждены
-    адаптеры Mellanox ConnectX-3 и более старые не поддерживают Implicit ODP и
+    блокировать всю память приложения с помощью mlockall(), чтобы задействовать
-    потому не поддерживаются в Vitastor. Запустите `ibv_devinfo -v` от имени
+    RDMA. В случае нативного QEMU-драйвера это будет означать, что при
-    суперпользователя, чтобы посмотреть список доступных RDMA-устройств, их
+    использовании RDMA на устройстве без поддержки Implicit ODP блокироваться
-    параметры и возможности.
+    от выгрузки будет вся память виртуальных машин.
+
+    В случае с адаптерами Mellanox Implicit ODP поддерживается начиная с
+    ConnectX-4. ConnectX-3 и более старые адаптеры не поддерживают Implicit ODP.
+    Чтобы посмотреть список своих RDMA-устройств и их возможностей, запустите
+    `ibv_devinfo -v` от имени суперпользователя.
 - name: rdma_port_num
   type: int
   default: 1

patches/cinder-vitastor.py

@@ -50,7 +50,7 @@ from cinder.volume import configuration
 from cinder.volume import driver
 from cinder.volume import volume_utils
 
-VERSION = '0.6.13'
+VERSION = '0.6.12'
 
 LOG = logging.getLogger(__name__)
 

rpm/build-tarball.sh

@@ -25,4 +25,4 @@ rm fio
 mv fio-copy fio
 FIO=`rpm -qi fio | perl -e 'while(<>) { /^Epoch[\s:]+(\S+)/ && print "$1:"; /^Version[\s:]+(\S+)/ && print $1; /^Release[\s:]+(\S+)/ && print "-$1"; }'`
 perl -i -pe 's/(Requires:\s*fio)([^\n]+)?/$1 = '$FIO'/' $VITASTOR/rpm/vitastor-el$EL.spec
-tar --transform 's#^#vitastor-0.6.13/#' --exclude 'rpm/*.rpm' -czf $VITASTOR/../vitastor-0.6.13$(rpm --eval '%dist').tar.gz *
+tar --transform 's#^#vitastor-0.6.12/#' --exclude 'rpm/*.rpm' -czf $VITASTOR/../vitastor-0.6.12$(rpm --eval '%dist').tar.gz *

rpm/vitastor-el7.Dockerfile

@@ -34,7 +34,7 @@ ADD . /root/vitastor
 RUN set -e; \
     cd /root/vitastor/rpm; \
     sh build-tarball.sh; \
-    cp /root/vitastor-0.6.13.el7.tar.gz ~/rpmbuild/SOURCES; \
+    cp /root/vitastor-0.6.12.el7.tar.gz ~/rpmbuild/SOURCES; \
     cp vitastor-el7.spec ~/rpmbuild/SPECS/vitastor.spec; \
     cd ~/rpmbuild/SPECS/; \
     rpmbuild -ba vitastor.spec; \

rpm/vitastor-el7.spec

@@ -1,11 +1,11 @@
 Name:           vitastor
-Version:        0.6.13
+Version:        0.6.12
 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-0.6.13.el7.tar.gz
+Source0:        vitastor-0.6.12.el7.tar.gz
 
 BuildRequires:  liburing-devel >= 0.6
 BuildRequires:  gperftools-devel

rpm/vitastor-el8.Dockerfile

@@ -33,7 +33,7 @@ ADD . /root/vitastor
 RUN set -e; \
     cd /root/vitastor/rpm; \
     sh build-tarball.sh; \
-    cp /root/vitastor-0.6.13.el8.tar.gz ~/rpmbuild/SOURCES; \
+    cp /root/vitastor-0.6.12.el8.tar.gz ~/rpmbuild/SOURCES; \
     cp vitastor-el8.spec ~/rpmbuild/SPECS/vitastor.spec; \
     cd ~/rpmbuild/SPECS/; \
     rpmbuild -ba vitastor.spec; \

rpm/vitastor-el8.spec

@@ -1,11 +1,11 @@
 Name:           vitastor
-Version:        0.6.13
+Version:        0.6.12
 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-0.6.13.el8.tar.gz
+Source0:        vitastor-0.6.12.el8.tar.gz
 
 BuildRequires:  liburing-devel >= 0.6
 BuildRequires:  gperftools-devel

src/CMakeLists.txt

@@ -15,7 +15,7 @@ if("${CMAKE_INSTALL_PREFIX}" MATCHES "^/usr/local/?$")
 	set(CMAKE_INSTALL_RPATH "${CMAKE_INSTALL_PREFIX}/${CMAKE_INSTALL_LIBDIR}")
 endif()
 
-add_definitions(-DVERSION="0.6.13")
+add_definitions(-DVERSION="0.6.12")
 add_definitions(-Wall -Wno-sign-compare -Wno-comment -Wno-parentheses -Wno-pointer-arith -fdiagnostics-color=always -I ${CMAKE_SOURCE_DIR}/src)
 if (${WITH_ASAN})
 	add_definitions(-fsanitize=address -fno-omit-frame-pointer)

src/fio_cluster.cpp

@@ -351,7 +351,7 @@ static enum fio_q_status sec_queue(struct thread_data *td, struct io_u *io)
         }
         else
         {
-            printf("+++ %s 0x%lx 0x%llx+%lx\n",
+            printf("+++ %s 0x%lx 0x%llx+%llx\n",
                 io->ddir == DDIR_READ ? "READ" : "WRITE",
                 (uint64_t)io, io->offset, io->xfer_buflen);
         }

src/fio_sec_osd.cpp

@@ -355,7 +355,7 @@ static int sec_getevents(struct thread_data *td, unsigned int min, unsigned int
         {
             if (reply.hdr.retval != io->xfer_buflen)
             {
-                fprintf(stderr, "Short read: retval = %ld instead of %lu\n", reply.hdr.retval, io->xfer_buflen);
+                fprintf(stderr, "Short read: retval = %ld instead of %llu\n", reply.hdr.retval, io->xfer_buflen);
                 exit(1);
             }
             // Support bitmap
@@ -380,7 +380,7 @@ static int sec_getevents(struct thread_data *td, unsigned int min, unsigned int
         {
             if (reply.hdr.retval != io->xfer_buflen)
             {
-                fprintf(stderr, "Short write: retval = %ld instead of %lu\n", reply.hdr.retval, io->xfer_buflen);
+                fprintf(stderr, "Short write: retval = %ld instead of %llu\n", reply.hdr.retval, io->xfer_buflen);
                 exit(1);
             }
         }

src/msgr_rdma.cpp

@@ -3,6 +3,7 @@
 
 #include <stdio.h>
 #include <stdlib.h>
+#include <sys/mman.h>
 #include "msgr_rdma.h"
 #include "messenger.h"
 
@@ -54,6 +55,7 @@ msgr_rdma_connection_t::~msgr_rdma_connection_t()
 msgr_rdma_context_t *msgr_rdma_context_t::create(const char *ib_devname, uint8_t ib_port, uint8_t gid_index, uint32_t mtu, int log_level)
 {
     int res;
+    bool odp = true;
     ibv_device **dev_list = NULL;
     msgr_rdma_context_t *ctx = new msgr_rdma_context_t();
     ctx->mtu = mtu;
@@ -117,9 +119,9 @@ msgr_rdma_context_t *msgr_rdma_context_t::create(const char *ib_devname, uint8_t
         fprintf(stderr, "RDMA device %s must have local LID because it's not Ethernet, but LID is zero\n", ibv_get_device_name(ctx->dev));
         goto cleanup;
     }
-    if (ibv_query_gid(ctx->context, ib_port, gid_index, &ctx->my_gid))
+    if ((res = ibv_query_gid(ctx->context, ib_port, gid_index, &ctx->my_gid)) != 0)
     {
-        fprintf(stderr, "Couldn't read RDMA device %s GID index %d\n", ibv_get_device_name(ctx->dev), gid_index);
+        fprintf(stderr, "Couldn't read RDMA device %s GID index %d: %s\n", ibv_get_device_name(ctx->dev), gid_index, strerror(res));
         goto cleanup;
     }
 
@@ -131,9 +133,9 @@ msgr_rdma_context_t *msgr_rdma_context_t::create(const char *ib_devname, uint8_t
     }
 
     {
-        if (ibv_query_device_ex(ctx->context, NULL, &ctx->attrx))
+        if ((res = ibv_query_device_ex(ctx->context, NULL, &ctx->attrx)) != 0)
         {
-            fprintf(stderr, "Couldn't query RDMA device for its features\n");
+            fprintf(stderr, "Couldn't query RDMA device %s for its features: %s\n", ibv_get_device_name(ctx->dev), strerror(res));
             goto cleanup;
         }
         if (!(ctx->attrx.odp_caps.general_caps & IBV_ODP_SUPPORT) ||
@@ -141,15 +143,20 @@ msgr_rdma_context_t *msgr_rdma_context_t::create(const char *ib_devname, uint8_t
             !(ctx->attrx.odp_caps.per_transport_caps.rc_odp_caps & IBV_ODP_SUPPORT_SEND) ||
             !(ctx->attrx.odp_caps.per_transport_caps.rc_odp_caps & IBV_ODP_SUPPORT_RECV))
         {
-            fprintf(stderr, "The RDMA device isn't implicit ODP (On-Demand Paging) capable or does not support RC send and receive with ODP\n");
+            fprintf(stderr, "Warning: RDMA device isn't implicit ODP (On-Demand Paging) capable, trying to lock all application memory\n");
-            goto cleanup;
+            if (mlockall(MCL_CURRENT|MCL_FUTURE|MCL_ONFAULT) != 0)
+            {
+                fprintf(stderr, "mlockall() failed: %s\n", strerror(errno));
+                goto cleanup;
+            }
+            odp = false;
         }
     }
 
-    ctx->mr = ibv_reg_mr(ctx->pd, NULL, SIZE_MAX, IBV_ACCESS_LOCAL_WRITE | IBV_ACCESS_ON_DEMAND);
+    ctx->mr = ibv_reg_mr(ctx->pd, NULL, SIZE_MAX, IBV_ACCESS_LOCAL_WRITE | (odp ? IBV_ACCESS_ON_DEMAND : 0));
     if (!ctx->mr)
     {
-        fprintf(stderr, "Couldn't register RDMA memory region\n");
+        fprintf(stderr, "Couldn't register RDMA memory region: %s\n", strerror(errno));
         goto cleanup;
     }
 

src/nbd_proxy.cpp

@@ -363,8 +363,7 @@ public:
         setsid();
         if (fork())
             exit(0);
-        if (chdir("/") != 0)
+        chdir("/");
-            fprintf(stderr, "Warning: Failed to chdir into /\n");
         close(0);
         close(1);
         close(2);
@@ -526,11 +525,7 @@ protected:
         {
             goto end_unmap;
         }
-        r = write(qd_fd, "32768", 5);
+        write(qd_fd, "32768", 5);
-        if (r != 5)
-        {
-            fprintf(stderr, "Warning: Failed to configure max_sectors_kb\n");
-        }
         close(qd_fd);
         if (!fork())
         {

src/osd.cpp

@@ -57,11 +57,7 @@ osd_t::osd_t(const json11::Json & config, ring_loop_t *ringloop)
     if (this->config["osd_memlock"] == "true" || this->config["osd_memlock"] == "1" || this->config["osd_memlock"] == "yes")
     {
         // Lock all OSD memory if requested
-        if (mlockall(MCL_CURRENT|MCL_FUTURE
+        if (mlockall(MCL_CURRENT|MCL_FUTURE|MCL_ONFAULT) != 0)
-#ifdef MCL_ONFAULT
-            | MCL_ONFAULT
-#endif
-            ) != 0)
         {
             fprintf(stderr, "osd_memlock is set to true, but mlockall() failed: %s\n", strerror(errno));
             exit(-1);

src/vitastor.pc.in

@@ -6,7 +6,7 @@ includedir=${prefix}/@CMAKE_INSTALL_INCLUDEDIR@
 
 Name: Vitastor
 Description: Vitastor client library
-Version: 0.6.13
+Version: 0.6.12
 Libs: -L${libdir} -lvitastor_client
 Cflags: -I${includedir}