Skip to contentSkip to conversation

block/rbd: migrate to coroutines and add write zeroes support

PATCH Superseded Peter Lieven 2 participants 24 messages Dec 27, 2020 – Jan 19, 2021 Block layer

v1 of 4 Superseded Latest is v4 →

lore
Share & tools
lore
#0 lore raw

cover block/rbd: migrate to coroutines and add write zeroes support

this series migrates the qemu rbd driver from the old aio emulation to native coroutines and adds write zeroes support which is important for block operations.

To archive this we first bump the librbd requirement to the already outdated luminous release of ceph to get rid of some wrappers and ifdef'ry in the code.

Peter Lieven (7):
block/rbd: bump librbd requirement to luminous release block/rbd: store object_size in BDRVRBDState block/rbd: use stored image_size in qemu_rbd_getlength block/rbd: add bdrv_{attach,detach}_aio_context block/rbd: migrate from aio to coroutines block/rbd: add write zeroes support block/rbd: change request alignment to 1 byte

 block/rbd.c | 421 +++++++++++++++++-----------------------------------
 configure   |   7 +-
 2 files changed, 139 insertions(+), 289 deletions(-)
Signature · 1 line
2.17.1
#1 lore raw

PATCH 1/7 block/rbd: bump librbd requirement to luminous release

even luminous (version 12.2) is unmaintained for over 3 years now. Bump the requirement to get rid of the ifdef'ry in the code.

---
 block/rbd.c | 120 ++++------------------------------------------------
 configure   |   7 +--
 2 files changed, 12 insertions(+), 115 deletions(-)
block/rbd.c +8 -112
index 9bd2bce716..650e27c351 100644
@@ -55,24 +55,10 @@
  * leading "\".
  */
 
-/* rbd_aio_discard added in 0.1.2 */
-#if LIBRBD_VERSION_CODE >= LIBRBD_VERSION(0, 1, 2)
-#define LIBRBD_SUPPORTS_DISCARD
-#else
-#undef LIBRBD_SUPPORTS_DISCARD
-#endif
-
 #define OBJ_MAX_SIZE (1UL << OBJ_DEFAULT_OBJ_ORDER)
 
 #define RBD_MAX_SNAPS 100
 
-/* The LIBRBD_SUPPORTS_IOVEC is defined in librbd.h */
-#ifdef LIBRBD_SUPPORTS_IOVEC
-#define LIBRBD_USE_IOVEC 1
-#else
-#define LIBRBD_USE_IOVEC 0
-#endif
-
 typedef enum {
     RBD_AIO_READ,
     RBD_AIO_WRITE,
@@ -84,7 +70,6 @@ typedef struct RBDAIOCB {
     BlockAIOCB common;
     int64_t ret;
     QEMUIOVector *qiov;
-    char *bounce;
     RBDAIOCmd cmd;
     int error;
     struct BDRVRBDState *s;
@@ -94,7 +79,6 @@ typedef struct RADOSCB {
     RBDAIOCB *acb;
     struct BDRVRBDState *s;
     int64_t size;
-    char *buf;
     int64_t ret;
 } RADOSCB;
 
@@ -332,13 +316,9 @@ static int qemu_rbd_set_keypairs(rados_t cluster, const char *keypairs_json,
 
 static void qemu_rbd_memset(RADOSCB *rcb, int64_t offs)
 {
-    if (LIBRBD_USE_IOVEC) {
-        RBDAIOCB *acb = rcb->acb;
-        iov_memset(acb->qiov->iov, acb->qiov->niov, offs, 0,
-                   acb->qiov->size - offs);
-    } else {
-        memset(rcb->buf + offs, 0, rcb->size - offs);
-    }
+    RBDAIOCB *acb = rcb->acb;
+    iov_memset(acb->qiov->iov, acb->qiov->niov, offs, 0,
+               acb->qiov->size - offs);
 }
 
 /* FIXME Deprecate and remove keypairs or make it available in QMP. */
@@ -493,13 +473,6 @@ static void qemu_rbd_complete_aio(RADOSCB *rcb)
 
     g_free(rcb);
 
-    if (!LIBRBD_USE_IOVEC) {
-        if (acb->cmd == RBD_AIO_READ) {
-            qemu_iovec_from_buf(acb->qiov, 0, acb->bounce, acb->qiov->size);
-        }
-        qemu_vfree(acb->bounce);
-    }
-
     acb->common.cb(acb->common.opaque, (acb->ret > 0 ? 0 : acb->ret));
 
     qemu_aio_unref(acb);
@@ -866,28 +839,6 @@ static void rbd_finish_aiocb(rbd_completion_t c, RADOSCB *rcb)
                                      rbd_finish_bh, rcb);
 }
 
-static int rbd_aio_discard_wrapper(rbd_image_t image,
-                                   uint64_t off,
-                                   uint64_t len,
-                                   rbd_completion_t comp)
-{
-#ifdef LIBRBD_SUPPORTS_DISCARD
-    return rbd_aio_discard(image, off, len, comp);
-#else
-    return -ENOTSUP;
-#endif
-}
-
-static int rbd_aio_flush_wrapper(rbd_image_t image,
-                                 rbd_completion_t comp)
-{
-#ifdef LIBRBD_SUPPORTS_AIO_FLUSH
-    return rbd_aio_flush(image, comp);
-#else
-    return -ENOTSUP;
-#endif
-}
-
 static BlockAIOCB *rbd_start_aio(BlockDriverState *bs,
                                  int64_t off,
                                  QEMUIOVector *qiov,
@@ -910,21 +861,6 @@ static BlockAIOCB *rbd_start_aio(BlockDriverState *bs,
 
     rcb = g_new(RADOSCB, 1);
 
-    if (!LIBRBD_USE_IOVEC) {
-        if (cmd == RBD_AIO_DISCARD || cmd == RBD_AIO_FLUSH) {
-            acb->bounce = NULL;
-        } else {
-            acb->bounce = qemu_try_blockalign(bs, qiov->size);
-            if (acb->bounce == NULL) {
-                goto failed;
-            }
-        }
-        if (cmd == RBD_AIO_WRITE) {
-            qemu_iovec_to_buf(acb->qiov, 0, acb->bounce, qiov->size);
-        }
-        rcb->buf = acb->bounce;
-    }
-
     acb->ret = 0;
     acb->error = 0;
     acb->s = s;
@@ -938,7 +874,7 @@ static BlockAIOCB *rbd_start_aio(BlockDriverState *bs,
     }
 
     switch (cmd) {
-    case RBD_AIO_WRITE: {
+    case RBD_AIO_WRITE:
         /*
          * RBD APIs don't allow us to write more than actual size, so in order
          * to support growing images, we resize the image before write
@@ -950,25 +886,16 @@ static BlockAIOCB *rbd_start_aio(BlockDriverState *bs,
                 goto failed_completion;
             }
         }
-#ifdef LIBRBD_SUPPORTS_IOVEC
-            r = rbd_aio_writev(s->image, qiov->iov, qiov->niov, off, c);
-#else
-            r = rbd_aio_write(s->image, off, size, rcb->buf, c);
-#endif
+        r = rbd_aio_writev(s->image, qiov->iov, qiov->niov, off, c);
         break;
-    }
     case RBD_AIO_READ:
-#ifdef LIBRBD_SUPPORTS_IOVEC
-            r = rbd_aio_readv(s->image, qiov->iov, qiov->niov, off, c);
-#else
-            r = rbd_aio_read(s->image, off, size, rcb->buf, c);
-#endif
+        r = rbd_aio_readv(s->image, qiov->iov, qiov->niov, off, c);
         break;
     case RBD_AIO_DISCARD:
-        r = rbd_aio_discard_wrapper(s->image, off, size, c);
+        r = rbd_aio_discard(s->image, off, size, c);
         break;
     case RBD_AIO_FLUSH:
-        r = rbd_aio_flush_wrapper(s->image, c);
+        r = rbd_aio_flush(s->image, c);
         break;
     default:
         r = -EINVAL;
@@ -983,9 +910,6 @@ failed_completion:
     rbd_aio_release(c);
 failed:
     g_free(rcb);
-    if (!LIBRBD_USE_IOVEC) {
-        qemu_vfree(acb->bounce);
-    }
 
     qemu_aio_unref(acb);
     return NULL;
@@ -1011,7 +935,6 @@ static BlockAIOCB *qemu_rbd_aio_pwritev(BlockDriverState *bs,
                          RBD_AIO_WRITE);
 }
 
-#ifdef LIBRBD_SUPPORTS_AIO_FLUSH
 static BlockAIOCB *qemu_rbd_aio_flush(BlockDriverState *bs,
                                       BlockCompletionFunc *cb,
                                       void *opaque)
@@ -1019,20 +942,6 @@ static BlockAIOCB *qemu_rbd_aio_flush(BlockDriverState *bs,
     return rbd_start_aio(bs, 0, NULL, 0, cb, opaque, RBD_AIO_FLUSH);
 }
 
-#else
-
-static int qemu_rbd_co_flush(BlockDriverState *bs)
-{
-#if LIBRBD_VERSION_CODE >= LIBRBD_VERSION(0, 1, 1)
-    /* rbd_flush added in 0.1.1 */
-    BDRVRBDState *s = bs->opaque;
-    return rbd_flush(s->image);
-#else
-    return 0;
-#endif
-}
-#endif
-
 static int qemu_rbd_getinfo(BlockDriverState *bs, BlockDriverInfo *bdi)
 {
     BDRVRBDState *s = bs->opaque;
@@ -1198,7 +1107,6 @@ static int qemu_rbd_snap_list(BlockDriverState *bs,
     return snap_count;
 }
 
-#ifdef LIBRBD_SUPPORTS_DISCARD
 static BlockAIOCB *qemu_rbd_aio_pdiscard(BlockDriverState *bs,
                                          int64_t offset,
                                          int bytes,
@@ -1208,9 +1116,7 @@ static BlockAIOCB *qemu_rbd_aio_pdiscard(BlockDriverState *bs,
     return rbd_start_aio(bs, offset, NULL, bytes, cb, opaque,
                          RBD_AIO_DISCARD);
 }
-#endif
 
-#ifdef LIBRBD_SUPPORTS_INVALIDATE
 static void coroutine_fn qemu_rbd_co_invalidate_cache(BlockDriverState *bs,
                                                       Error **errp)
 {
@@ -1220,7 +1126,6 @@ static void coroutine_fn qemu_rbd_co_invalidate_cache(BlockDriverState *bs,
         error_setg_errno(errp, -r, "Failed to invalidate the cache");
     }
 }
-#endif
 
 static QemuOptsList qemu_rbd_create_opts = {
     .name = "rbd-create-opts",
@@ -1278,23 +1183,14 @@ static BlockDriver bdrv_rbd = {
     .bdrv_aio_preadv        = qemu_rbd_aio_preadv,
     .bdrv_aio_pwritev       = qemu_rbd_aio_pwritev,
 
-#ifdef LIBRBD_SUPPORTS_AIO_FLUSH
     .bdrv_aio_flush         = qemu_rbd_aio_flush,
-#else
-    .bdrv_co_flush_to_disk  = qemu_rbd_co_flush,
-#endif
-
-#ifdef LIBRBD_SUPPORTS_DISCARD
     .bdrv_aio_pdiscard      = qemu_rbd_aio_pdiscard,
-#endif
 
     .bdrv_snapshot_create   = qemu_rbd_snap_create,
     .bdrv_snapshot_delete   = qemu_rbd_snap_remove,
     .bdrv_snapshot_list     = qemu_rbd_snap_list,
     .bdrv_snapshot_goto     = qemu_rbd_snap_rollback,
-#ifdef LIBRBD_SUPPORTS_INVALIDATE
     .bdrv_co_invalidate_cache = qemu_rbd_co_invalidate_cache,
-#endif
 
     .strong_runtime_opts    = qemu_rbd_strong_runtime_opts,
 };
configure +4 -3
index c228f7c21e..e07f80cab4 100755
@@ -3663,8 +3663,9 @@ if test "$rbd" != "no" ; then
 #include <stdio.h>
 #include <rbd/librbd.h>
 int main(void) {
-    rados_t cluster;
-    rados_create(&cluster, NULL);
+#if LIBRBD_VERSION_CODE < LIBRBD_VERSION(1, 12, 0)
+#error
+#endif
     return 0;
 }
 EOF
@@ -3673,7 +3674,7 @@ EOF
     rbd=yes
   else
     if test "$rbd" = "yes" ; then
-      feature_not_found "rados block device" "Install librbd/ceph devel"
+      feature_not_found "rados block device" "Install librbd/ceph devel (>= 12.0.0)"
     fi
     rbd=no
   fi
Signature · 1 line
2.17.1
#2 lore raw

PATCH 2/7 block/rbd: store object_size in BDRVRBDState

In reply to Peter Lieven
---
 block/rbd.c | 18 +++++++-----------
 1 file changed, 7 insertions(+), 11 deletions(-)
block/rbd.c +7 -11
index 650e27c351..bc8cf8af9b 100644
@@ -90,6 +90,7 @@ typedef struct BDRVRBDState {
     char *snap;
     char *namespace;
     uint64_t image_size;
+    uint64_t object_size;
 } BDRVRBDState;
 
 static int qemu_rbd_connect(rados_t *cluster, rados_ioctx_t *io_ctx,
@@ -663,6 +664,7 @@ static int qemu_rbd_open(BlockDriverState *bs, QDict *options, int flags,
     const QDictEntry *e;
     Error *local_err = NULL;
     char *keypairs, *secretid;
+    rbd_image_info_t info;
     int r;
 
     keypairs = g_strdup(qdict_get_try_str(options, "=keyvalue-pairs"));
@@ -727,13 +729,15 @@ static int qemu_rbd_open(BlockDriverState *bs, QDict *options, int flags,
         goto failed_open;
     }
 
-    r = rbd_get_size(s->image, &s->image_size);
+    r = rbd_stat(s->image, &info, sizeof(info));
     if (r < 0) {
-        error_setg_errno(errp, -r, "error getting image size from %s",
+        error_setg_errno(errp, -r, "error getting image info from %s",
                          s->image_name);
         rbd_close(s->image);
         goto failed_open;
     }
+    s->image_size = info.size;
+    s->object_size = info.obj_size;
 
     /* If we are using an rbd snapshot, we must be r/o, otherwise
      * leave as-is */
@@ -945,15 +949,7 @@ static BlockAIOCB *qemu_rbd_aio_flush(BlockDriverState *bs,
 static int qemu_rbd_getinfo(BlockDriverState *bs, BlockDriverInfo *bdi)
 {
     BDRVRBDState *s = bs->opaque;
-    rbd_image_info_t info;
-    int r;
-
-    r = rbd_stat(s->image, &info, sizeof(info));
-    if (r < 0) {
-        return r;
-    }
-
-    bdi->cluster_size = info.obj_size;
+    bdi->cluster_size = s->object_size;
     return 0;
 }
 
Signature · 1 line
2.17.1
#3 lore raw

PATCH 3/7 block/rbd: use stored image_size in qemu_rbd_getlength

In reply to Peter Lieven
---
 block/rbd.c | 10 +---------
 1 file changed, 1 insertion(+), 9 deletions(-)
block/rbd.c +1 -9
index bc8cf8af9b..a2da70e37f 100644
@@ -956,15 +956,7 @@ static int qemu_rbd_getinfo(BlockDriverState *bs, BlockDriverInfo *bdi)
 static int64_t qemu_rbd_getlength(BlockDriverState *bs)
 {
     BDRVRBDState *s = bs->opaque;
-    rbd_image_info_t info;
-    int r;
-
-    r = rbd_stat(s->image, &info, sizeof(info));
-    if (r < 0) {
-        return r;
-    }
-
-    return info.size;
+    return s->image_size;
 }
 
 static int coroutine_fn qemu_rbd_co_truncate(BlockDriverState *bs,
Signature · 1 line
2.17.1
#4 lore raw

Re: [PATCH 3/7] block/rbd: use stored image_size in qemu_rbd_getlength

Quoted 28 lines · Peter Lieven
---
 block/rbd.c | 10 +---------
 1 file changed, 1 insertion(+), 9 deletions(-)
block/rbd.c +1 -10
index bc8cf8af9b..a2da70e37f 100644
@@ -956,15 +956,7 @@ static int qemu_rbd_getinfo(BlockDriverState *bs, BlockDriverInfo *bdi)
 static int64_t qemu_rbd_getlength(BlockDriverState *bs)
 {
     BDRVRBDState *s = bs->opaque;
-    rbd_image_info_t info;
-    int r;
-
-    r = rbd_stat(s->image, &info, sizeof(info));
-    if (r < 0) {
-        return r;
-    }
-
-    return info.size;
+    return s->image_size;
 }
 
 static int coroutine_fn qemu_rbd_co_truncate(BlockDriverState *bs,
--
Quoted 1 line

2.17.1

An RBD image can technically change size dynamically while in-use. The original code would provide the most up-to-date length but this version will always return the size of the image when it was opened.

--
Jason

#5 lore raw

Re: [PATCH 3/7] block/rbd: use stored image_size in qemu_rbd_getlength

Am 14.01.21 um 20:18 schrieb Jason Dillaman:

Quoted 26 lines · Peter Lieven
Quoted 25 lines
---
 block/rbd.c | 10 +---------
 1 file changed, 1 insertion(+), 9 deletions(-)
block/rbd.c +1 -9
index bc8cf8af9b..a2da70e37f 100644
@@ -956,15 +956,7 @@ static int qemu_rbd_getinfo(BlockDriverState *bs, BlockDriverInfo *bdi)
 static int64_t qemu_rbd_getlength(BlockDriverState *bs)
 {
     BDRVRBDState *s = bs->opaque;
-    rbd_image_info_t info;
-    int r;
-
-    r = rbd_stat(s->image, &info, sizeof(info));
-    if (r < 0) {
-        return r;
-    }
-
-    return info.size;
+    return s->image_size;
 }
 
Quoted 6 lines
Quoted 3 lines
block/rbd.c +0 -1
 static int coroutine_fn qemu_rbd_co_truncate(BlockDriverState *bs,
--

2.17.1

An RBD image can technically change size dynamically while in-use. The original code would provide the most up-to-date length but this version will always return the size of the image when it was opened.

Agreed, but Qemu won't propagate this info to the guest unless Qemu truncate is called with length 0.

Anyway, if we want support this we should adjust s->image_size if the size has changed.

Peter

#6 lore raw

PATCH 4/7 block/rbd: add bdrv_{attach,detach}_aio_context

In reply to Peter Lieven
---
 block/rbd.c | 21 +++++++++++++++++++--
 1 file changed, 19 insertions(+), 2 deletions(-)
block/rbd.c +19 -2
index a2da70e37f..27b232f4d8 100644
@@ -91,6 +91,7 @@ typedef struct BDRVRBDState {
     char *namespace;
     uint64_t image_size;
     uint64_t object_size;
+    AioContext *aio_context;
 } BDRVRBDState;
 
 static int qemu_rbd_connect(rados_t *cluster, rados_ioctx_t *io_ctx,
@@ -749,6 +750,8 @@ static int qemu_rbd_open(BlockDriverState *bs, QDict *options, int flags,
         }
     }
 
+    s->aio_context = bdrv_get_aio_context(bs);
+
     /* When extending regular files, we get zeros from the OS */
     bs->supported_truncate_flags = BDRV_REQ_ZERO_WRITE;
 
@@ -839,8 +842,7 @@ static void rbd_finish_aiocb(rbd_completion_t c, RADOSCB *rcb)
     rcb->ret = rbd_aio_get_return_value(c);
     rbd_aio_release(c);
 
-    replay_bh_schedule_oneshot_event(bdrv_get_aio_context(acb->common.bs),
-                                     rbd_finish_bh, rcb);
+    replay_bh_schedule_oneshot_event(acb->s->aio_context, rbd_finish_bh, rcb);
 }
 
 static BlockAIOCB *rbd_start_aio(BlockDriverState *bs,
@@ -1151,6 +1153,18 @@ static const char *const qemu_rbd_strong_runtime_opts[] = {
     NULL
 };
 
+static void qemu_rbd_attach_aio_context(BlockDriverState *bs,
+                                       AioContext *new_context)
+{
+    BDRVRBDState *s = bs->opaque;
+    s->aio_context = new_context;
+}
+
+static void qemu_rbd_detach_aio_context(BlockDriverState *bs)
+{
+
+}
+
 static BlockDriver bdrv_rbd = {
     .format_name            = "rbd",
     .instance_size          = sizeof(BDRVRBDState),
@@ -1180,6 +1194,9 @@ static BlockDriver bdrv_rbd = {
     .bdrv_snapshot_goto     = qemu_rbd_snap_rollback,
     .bdrv_co_invalidate_cache = qemu_rbd_co_invalidate_cache,
 
+    .bdrv_attach_aio_context  = qemu_rbd_attach_aio_context,
+    .bdrv_detach_aio_context  = qemu_rbd_detach_aio_context,
+
     .strong_runtime_opts    = qemu_rbd_strong_runtime_opts,
 };
 
Signature · 1 line
2.17.1
#7 lore raw

Re: [PATCH 4/7] block/rbd: add bdrv_{attach,detach}_aio_context

Quoted 40 lines · Peter Lieven
---
 block/rbd.c | 21 +++++++++++++++++++--
 1 file changed, 19 insertions(+), 2 deletions(-)
block/rbd.c +4 -2
index a2da70e37f..27b232f4d8 100644
@@ -91,6 +91,7 @@ typedef struct BDRVRBDState {
     char *namespace;
     uint64_t image_size;
     uint64_t object_size;
+    AioContext *aio_context;
 } BDRVRBDState;
 
 static int qemu_rbd_connect(rados_t *cluster, rados_ioctx_t *io_ctx,
@@ -749,6 +750,8 @@ static int qemu_rbd_open(BlockDriverState *bs, QDict *options, int flags,
         }
     }
 
+    s->aio_context = bdrv_get_aio_context(bs);
+
     /* When extending regular files, we get zeros from the OS */
     bs->supported_truncate_flags = BDRV_REQ_ZERO_WRITE;
 
@@ -839,8 +842,7 @@ static void rbd_finish_aiocb(rbd_completion_t c, RADOSCB *rcb)
     rcb->ret = rbd_aio_get_return_value(c);
     rbd_aio_release(c);
 
-    replay_bh_schedule_oneshot_event(bdrv_get_aio_context(acb->common.bs),
-                                     rbd_finish_bh, rcb);
+    replay_bh_schedule_oneshot_event(acb->s->aio_context, rbd_finish_bh, rcb);
 }
 
 static BlockAIOCB *rbd_start_aio(BlockDriverState *bs,
@@ -1151,6 +1153,18 @@ static const char *const qemu_rbd_strong_runtime_opts[] = {
     NULL
 };

block/rbd.c static const char *const qemu_rbd_strong_runtime_opts[] = { · line 1164

Quoted 10 lines
block/rbd.c +9 -0
 
+static void qemu_rbd_attach_aio_context(BlockDriverState *bs,
+                                       AioContext *new_context)
+{
+    BDRVRBDState *s = bs->opaque;
+    s->aio_context = new_context;
+}
+
+static void qemu_rbd_detach_aio_context(BlockDriverState *bs)
+{

I don't know enough about the internals of QEMU, but this seems suspicious to be a no-op.

Quoted 16 lines
block/rbd.c +5 -1
+}
+
 static BlockDriver bdrv_rbd = {
     .format_name            = "rbd",
     .instance_size          = sizeof(BDRVRBDState),
@@ -1180,6 +1194,9 @@ static BlockDriver bdrv_rbd = {
     .bdrv_snapshot_goto     = qemu_rbd_snap_rollback,
     .bdrv_co_invalidate_cache = qemu_rbd_co_invalidate_cache,
 
+    .bdrv_attach_aio_context  = qemu_rbd_attach_aio_context,
+    .bdrv_detach_aio_context  = qemu_rbd_detach_aio_context,
+
     .strong_runtime_opts    = qemu_rbd_strong_runtime_opts,
 };
 
--
Quoted 3 lines

2.17.1

--
Jason

#8 lore raw

Re: [PATCH 4/7] block/rbd: add bdrv_{attach,detach}_aio_context

Am 14.01.21 um 20:18 schrieb Jason Dillaman:

Quoted 41 lines · Peter Lieven
Quoted 40 lines
---
 block/rbd.c | 21 +++++++++++++++++++--
 1 file changed, 19 insertions(+), 2 deletions(-)
block/rbd.c +4 -2
index a2da70e37f..27b232f4d8 100644
@@ -91,6 +91,7 @@ typedef struct BDRVRBDState {
     char *namespace;
     uint64_t image_size;
     uint64_t object_size;
+    AioContext *aio_context;
 } BDRVRBDState;
 
 static int qemu_rbd_connect(rados_t *cluster, rados_ioctx_t *io_ctx,
@@ -749,6 +750,8 @@ static int qemu_rbd_open(BlockDriverState *bs, QDict *options, int flags,
         }
     }
 
+    s->aio_context = bdrv_get_aio_context(bs);
+
     /* When extending regular files, we get zeros from the OS */
     bs->supported_truncate_flags = BDRV_REQ_ZERO_WRITE;
 
@@ -839,8 +842,7 @@ static void rbd_finish_aiocb(rbd_completion_t c, RADOSCB *rcb)
     rcb->ret = rbd_aio_get_return_value(c);
     rbd_aio_release(c);
 
-    replay_bh_schedule_oneshot_event(bdrv_get_aio_context(acb->common.bs),
-                                     rbd_finish_bh, rcb);
+    replay_bh_schedule_oneshot_event(acb->s->aio_context, rbd_finish_bh, rcb);
 }
 
 static BlockAIOCB *rbd_start_aio(BlockDriverState *bs,
@@ -1151,6 +1153,18 @@ static const char *const qemu_rbd_strong_runtime_opts[] = {
     NULL
 };
 
Quoted 11 lines

block/rbd.c static const char *const qemu_rbd_strong_runtime_opts[] = { · line 1164

Quoted 9 lines
block/rbd.c +9 -0
+static void qemu_rbd_attach_aio_context(BlockDriverState *bs,
+                                       AioContext *new_context)
+{
+    BDRVRBDState *s = bs->opaque;
+    s->aio_context = new_context;
+}
+
+static void qemu_rbd_detach_aio_context(BlockDriverState *bs)
+{

I don't know enough about the internals of QEMU, but this seems suspicious to be a no-op.

You are right, I was believing attach and detach aio_context functions always needs to be implemented both at the same time.

Normally this is the point where internal timers will be deleted or polling an fd will be stopped.

We can leave it completely out or set s->aio_context = NULL if we don't want to omit it.

Peter

#9 lore raw

PATCH 5/7 block/rbd: migrate from aio to coroutines

In reply to Peter Lieven
---
 block/rbd.c | 247 ++++++++++++++++++----------------------------------
 1 file changed, 84 insertions(+), 163 deletions(-)
block/rbd.c +84 -163
index 27b232f4d8..2d77d0007f 100644
@@ -66,22 +66,6 @@ typedef enum {
     RBD_AIO_FLUSH
 } RBDAIOCmd;
 
-typedef struct RBDAIOCB {
-    BlockAIOCB common;
-    int64_t ret;
-    QEMUIOVector *qiov;
-    RBDAIOCmd cmd;
-    int error;
-    struct BDRVRBDState *s;
-} RBDAIOCB;
-
-typedef struct RADOSCB {
-    RBDAIOCB *acb;
-    struct BDRVRBDState *s;
-    int64_t size;
-    int64_t ret;
-} RADOSCB;
-
 typedef struct BDRVRBDState {
     rados_t cluster;
     rados_ioctx_t io_ctx;
@@ -94,6 +78,13 @@ typedef struct BDRVRBDState {
     AioContext *aio_context;
 } BDRVRBDState;
 
+typedef struct RBDTask {
+    BDRVRBDState *s;
+    Coroutine *co;
+    bool complete;
+    int64_t ret;
+} RBDTask;
+
 static int qemu_rbd_connect(rados_t *cluster, rados_ioctx_t *io_ctx,
                             BlockdevOptionsRbd *opts, bool cache,
                             const char *keypairs, const char *secretid,
@@ -316,13 +307,6 @@ static int qemu_rbd_set_keypairs(rados_t cluster, const char *keypairs_json,
     return ret;
 }
 
-static void qemu_rbd_memset(RADOSCB *rcb, int64_t offs)
-{
-    RBDAIOCB *acb = rcb->acb;
-    iov_memset(acb->qiov->iov, acb->qiov->niov, offs, 0,
-               acb->qiov->size - offs);
-}
-
 /* FIXME Deprecate and remove keypairs or make it available in QMP. */
 static int qemu_rbd_do_create(BlockdevCreateOptions *options,
                               const char *keypairs, const char *password_secret,
@@ -440,46 +424,6 @@ exit:
     return ret;
 }
 
-/*
- * This aio completion is being called from rbd_finish_bh() and runs in qemu
- * BH context.
- */
-static void qemu_rbd_complete_aio(RADOSCB *rcb)
-{
-    RBDAIOCB *acb = rcb->acb;
-    int64_t r;
-
-    r = rcb->ret;
-
-    if (acb->cmd != RBD_AIO_READ) {
-        if (r < 0) {
-            acb->ret = r;
-            acb->error = 1;
-        } else if (!acb->error) {
-            acb->ret = rcb->size;
-        }
-    } else {
-        if (r < 0) {
-            qemu_rbd_memset(rcb, 0);
-            acb->ret = r;
-            acb->error = 1;
-        } else if (r < rcb->size) {
-            qemu_rbd_memset(rcb, r);
-            if (!acb->error) {
-                acb->ret = rcb->size;
-            }
-        } else if (!acb->error) {
-            acb->ret = r;
-        }
-    }
-
-    g_free(rcb);
-
-    acb->common.cb(acb->common.opaque, (acb->ret > 0 ? 0 : acb->ret));
-
-    qemu_aio_unref(acb);
-}
-
 static char *qemu_rbd_mon_host(BlockdevOptionsRbd *opts, Error **errp)
 {
     const char **vals;
@@ -817,88 +761,49 @@ static int qemu_rbd_resize(BlockDriverState *bs, uint64_t size)
     return 0;
 }
 
-static const AIOCBInfo rbd_aiocb_info = {
-    .aiocb_size = sizeof(RBDAIOCB),
-};
-
-static void rbd_finish_bh(void *opaque)
+static void qemu_rbd_finish_bh(void *opaque)
 {
-    RADOSCB *rcb = opaque;
-    qemu_rbd_complete_aio(rcb);
+    RBDTask *task = opaque;
+    task->complete = 1;
+    aio_co_wake(task->co);
 }
 
-/*
- * This is the callback function for rbd_aio_read and _write
- *
- * Note: this function is being called from a non qemu thread so
- * we need to be careful about what we do here. Generally we only
- * schedule a BH, and do the rest of the io completion handling
- * from rbd_finish_bh() which runs in a qemu context.
- */
-static void rbd_finish_aiocb(rbd_completion_t c, RADOSCB *rcb)
+static void qemu_rbd_completion_cb(rbd_completion_t c, RBDTask *task)
 {
-    RBDAIOCB *acb = rcb->acb;
-
-    rcb->ret = rbd_aio_get_return_value(c);
+    task->ret = rbd_aio_get_return_value(c);
     rbd_aio_release(c);
-
-    replay_bh_schedule_oneshot_event(acb->s->aio_context, rbd_finish_bh, rcb);
+    aio_bh_schedule_oneshot(task->s->aio_context, qemu_rbd_finish_bh, task);
 }
 
-static BlockAIOCB *rbd_start_aio(BlockDriverState *bs,
-                                 int64_t off,
-                                 QEMUIOVector *qiov,
-                                 int64_t size,
-                                 BlockCompletionFunc *cb,
-                                 void *opaque,
-                                 RBDAIOCmd cmd)
+static int coroutine_fn qemu_rbd_start_co(BlockDriverState *bs,
+                                          uint64_t offset,
+                                          uint64_t bytes,
+                                          QEMUIOVector *qiov,
+                                          int flags,
+                                          RBDAIOCmd cmd)
 {
-    RBDAIOCB *acb;
-    RADOSCB *rcb = NULL;
-    rbd_completion_t c;
-    int r;
-
     BDRVRBDState *s = bs->opaque;
+    RBDTask task = { .s = s, .co = qemu_coroutine_self() };
+    rbd_completion_t c;
+    int r, ret = -EIO;
 
-    acb = qemu_aio_get(&rbd_aiocb_info, bs, cb, opaque);
-    acb->cmd = cmd;
-    acb->qiov = qiov;
-    assert(!qiov || qiov->size == size);
-
-    rcb = g_new(RADOSCB, 1);
-
-    acb->ret = 0;
-    acb->error = 0;
-    acb->s = s;
+    assert(!qiov || qiov->size == bytes);
 
-    rcb->acb = acb;
-    rcb->s = acb->s;
-    rcb->size = size;
-    r = rbd_aio_create_completion(rcb, (rbd_callback_t) rbd_finish_aiocb, &c);
+    r = rbd_aio_create_completion(&task,
+                                  (rbd_callback_t) qemu_rbd_completion_cb, &c);
     if (r < 0) {
         goto failed;
     }
 
     switch (cmd) {
-    case RBD_AIO_WRITE:
-        /*
-         * RBD APIs don't allow us to write more than actual size, so in order
-         * to support growing images, we resize the image before write
-         * operations that exceed the current size.
-         */
-        if (off + size > s->image_size) {
-            r = qemu_rbd_resize(bs, off + size);
-            if (r < 0) {
-                goto failed_completion;
-            }
-        }
-        r = rbd_aio_writev(s->image, qiov->iov, qiov->niov, off, c);
-        break;
     case RBD_AIO_READ:
-        r = rbd_aio_readv(s->image, qiov->iov, qiov->niov, off, c);
+        r = rbd_aio_readv(s->image, qiov->iov, qiov->niov, offset, c);
+        break;
+    case RBD_AIO_WRITE:
+        r = rbd_aio_writev(s->image, qiov->iov, qiov->niov, offset, c);
         break;
     case RBD_AIO_DISCARD:
-        r = rbd_aio_discard(s->image, off, size, c);
+        r = rbd_aio_discard(s->image, offset, bytes, c);
         break;
     case RBD_AIO_FLUSH:
         r = rbd_aio_flush(s->image, c);
@@ -908,44 +813,71 @@ static BlockAIOCB *rbd_start_aio(BlockDriverState *bs,
     }
 
     if (r < 0) {
+        task.complete = 1;
+        task.ret = r;
+    }
+
+    while (!task.complete) {
+        qemu_coroutine_yield();
+    }
+
+    if (task.ret < 0) {
+        error_report("rbd request failed: cmd %d offset %" PRIu64 " bytes %"
+                     PRIu64 " flags %d task.ret %" PRIi64 " (%s)", cmd, offset,
+                     bytes, flags, task.ret, strerror(-task.ret));
         goto failed_completion;
     }
-    return &acb->common;
+
+    /* zero pad short reads */
+    if (cmd == RBD_AIO_READ && task.ret < qiov->size) {
+        qemu_iovec_memset(qiov, task.ret, 0, qiov->size - task.ret);
+    }
+
+    return 0;
 
 failed_completion:
     rbd_aio_release(c);
 failed:
-    g_free(rcb);
+    return ret;
+}
 
-    qemu_aio_unref(acb);
-    return NULL;
+static int
+coroutine_fn qemu_rbd_co_preadv(BlockDriverState *bs, uint64_t offset,
+                               uint64_t bytes, QEMUIOVector *qiov,
+                               int flags)
+{
+    return qemu_rbd_start_co(bs, offset, bytes, qiov, flags, RBD_AIO_READ);
 }
 
-static BlockAIOCB *qemu_rbd_aio_preadv(BlockDriverState *bs,
-                                       uint64_t offset, uint64_t bytes,
-                                       QEMUIOVector *qiov, int flags,
-                                       BlockCompletionFunc *cb,
-                                       void *opaque)
+static int
+coroutine_fn qemu_rbd_co_pwritev(BlockDriverState *bs, uint64_t offset,
+                                 uint64_t bytes, QEMUIOVector *qiov,
+                                 int flags)
 {
-    return rbd_start_aio(bs, offset, qiov, bytes, cb, opaque,
-                         RBD_AIO_READ);
+    BDRVRBDState *s = bs->opaque;
+    /*
+     * RBD APIs don't allow us to write more than actual size, so in order
+     * to support growing images, we resize the image before write
+     * operations that exceed the current size.
+     */
+    if (offset + bytes > s->image_size) {
+        int r = qemu_rbd_resize(bs, offset + bytes);
+        if (r < 0) {
+            return r;
+        }
+    }
+    return qemu_rbd_start_co(bs, offset, bytes, qiov, flags, RBD_AIO_WRITE);
 }
 
-static BlockAIOCB *qemu_rbd_aio_pwritev(BlockDriverState *bs,
-                                        uint64_t offset, uint64_t bytes,
-                                        QEMUIOVector *qiov, int flags,
-                                        BlockCompletionFunc *cb,
-                                        void *opaque)
+static int coroutine_fn qemu_rbd_co_flush(BlockDriverState *bs)
 {
-    return rbd_start_aio(bs, offset, qiov, bytes, cb, opaque,
-                         RBD_AIO_WRITE);
+    return qemu_rbd_start_co(bs, 0, 0, NULL, 0, RBD_AIO_FLUSH);
 }
 
-static BlockAIOCB *qemu_rbd_aio_flush(BlockDriverState *bs,
-                                      BlockCompletionFunc *cb,
-                                      void *opaque)
+static int coroutine_fn qemu_rbd_co_pdiscard(BlockDriverState *bs,
+                                             int64_t offset, int count)
 {
-    return rbd_start_aio(bs, 0, NULL, 0, cb, opaque, RBD_AIO_FLUSH);
+    return qemu_rbd_start_co(bs, offset, count, NULL, 0, RBD_AIO_DISCARD);
 }
 
 static int qemu_rbd_getinfo(BlockDriverState *bs, BlockDriverInfo *bdi)
@@ -1097,16 +1029,6 @@ static int qemu_rbd_snap_list(BlockDriverState *bs,
     return snap_count;
 }
 
-static BlockAIOCB *qemu_rbd_aio_pdiscard(BlockDriverState *bs,
-                                         int64_t offset,
-                                         int bytes,
-                                         BlockCompletionFunc *cb,
-                                         void *opaque)
-{
-    return rbd_start_aio(bs, offset, NULL, bytes, cb, opaque,
-                         RBD_AIO_DISCARD);
-}
-
 static void coroutine_fn qemu_rbd_co_invalidate_cache(BlockDriverState *bs,
                                                       Error **errp)
 {
@@ -1182,11 +1104,10 @@ static BlockDriver bdrv_rbd = {
     .bdrv_co_truncate       = qemu_rbd_co_truncate,
     .protocol_name          = "rbd",
 
-    .bdrv_aio_preadv        = qemu_rbd_aio_preadv,
-    .bdrv_aio_pwritev       = qemu_rbd_aio_pwritev,
-
-    .bdrv_aio_flush         = qemu_rbd_aio_flush,
-    .bdrv_aio_pdiscard      = qemu_rbd_aio_pdiscard,
+    .bdrv_co_preadv         = qemu_rbd_co_preadv,
+    .bdrv_co_pwritev        = qemu_rbd_co_pwritev,
+    .bdrv_co_flush_to_disk  = qemu_rbd_co_flush,
+    .bdrv_co_pdiscard       = qemu_rbd_co_pdiscard,
 
     .bdrv_snapshot_create   = qemu_rbd_snap_create,
     .bdrv_snapshot_delete   = qemu_rbd_snap_remove,
Signature · 1 line
2.17.1
#10 lore raw

Re: [PATCH 5/7] block/rbd: migrate from aio to coroutines

Quoted 192 lines · Peter Lieven
---
 block/rbd.c | 247 ++++++++++++++++++----------------------------------
 1 file changed, 84 insertions(+), 163 deletions(-)
block/rbd.c +26 -110
index 27b232f4d8..2d77d0007f 100644
@@ -66,22 +66,6 @@ typedef enum {
     RBD_AIO_FLUSH
 } RBDAIOCmd;
 
-typedef struct RBDAIOCB {
-    BlockAIOCB common;
-    int64_t ret;
-    QEMUIOVector *qiov;
-    RBDAIOCmd cmd;
-    int error;
-    struct BDRVRBDState *s;
-} RBDAIOCB;
-
-typedef struct RADOSCB {
-    RBDAIOCB *acb;
-    struct BDRVRBDState *s;
-    int64_t size;
-    int64_t ret;
-} RADOSCB;
-
 typedef struct BDRVRBDState {
     rados_t cluster;
     rados_ioctx_t io_ctx;
@@ -94,6 +78,13 @@ typedef struct BDRVRBDState {
     AioContext *aio_context;
 } BDRVRBDState;
 
+typedef struct RBDTask {
+    BDRVRBDState *s;
+    Coroutine *co;
+    bool complete;
+    int64_t ret;
+} RBDTask;
+
 static int qemu_rbd_connect(rados_t *cluster, rados_ioctx_t *io_ctx,
                             BlockdevOptionsRbd *opts, bool cache,
                             const char *keypairs, const char *secretid,
@@ -316,13 +307,6 @@ static int qemu_rbd_set_keypairs(rados_t cluster, const char *keypairs_json,
     return ret;
 }
 
-static void qemu_rbd_memset(RADOSCB *rcb, int64_t offs)
-{
-    RBDAIOCB *acb = rcb->acb;
-    iov_memset(acb->qiov->iov, acb->qiov->niov, offs, 0,
-               acb->qiov->size - offs);
-}
-
 /* FIXME Deprecate and remove keypairs or make it available in QMP. */
 static int qemu_rbd_do_create(BlockdevCreateOptions *options,
                               const char *keypairs, const char *password_secret,
@@ -440,46 +424,6 @@ exit:
     return ret;
 }
 
-/*
- * This aio completion is being called from rbd_finish_bh() and runs in qemu
- * BH context.
- */
-static void qemu_rbd_complete_aio(RADOSCB *rcb)
-{
-    RBDAIOCB *acb = rcb->acb;
-    int64_t r;
-
-    r = rcb->ret;
-
-    if (acb->cmd != RBD_AIO_READ) {
-        if (r < 0) {
-            acb->ret = r;
-            acb->error = 1;
-        } else if (!acb->error) {
-            acb->ret = rcb->size;
-        }
-    } else {
-        if (r < 0) {
-            qemu_rbd_memset(rcb, 0);
-            acb->ret = r;
-            acb->error = 1;
-        } else if (r < rcb->size) {
-            qemu_rbd_memset(rcb, r);
-            if (!acb->error) {
-                acb->ret = rcb->size;
-            }
-        } else if (!acb->error) {
-            acb->ret = r;
-        }
-    }
-
-    g_free(rcb);
-
-    acb->common.cb(acb->common.opaque, (acb->ret > 0 ? 0 : acb->ret));
-
-    qemu_aio_unref(acb);
-}
-
 static char *qemu_rbd_mon_host(BlockdevOptionsRbd *opts, Error **errp)
 {
     const char **vals;
@@ -817,88 +761,49 @@ static int qemu_rbd_resize(BlockDriverState *bs, uint64_t size)
     return 0;
 }
 
-static const AIOCBInfo rbd_aiocb_info = {
-    .aiocb_size = sizeof(RBDAIOCB),
-};
-
-static void rbd_finish_bh(void *opaque)
+static void qemu_rbd_finish_bh(void *opaque)
 {
-    RADOSCB *rcb = opaque;
-    qemu_rbd_complete_aio(rcb);
+    RBDTask *task = opaque;
+    task->complete = 1;
+    aio_co_wake(task->co);
 }
 
-/*
- * This is the callback function for rbd_aio_read and _write
- *
- * Note: this function is being called from a non qemu thread so
- * we need to be careful about what we do here. Generally we only
- * schedule a BH, and do the rest of the io completion handling
- * from rbd_finish_bh() which runs in a qemu context.
- */
-static void rbd_finish_aiocb(rbd_completion_t c, RADOSCB *rcb)
+static void qemu_rbd_completion_cb(rbd_completion_t c, RBDTask *task)
 {
-    RBDAIOCB *acb = rcb->acb;
-
-    rcb->ret = rbd_aio_get_return_value(c);
+    task->ret = rbd_aio_get_return_value(c);
     rbd_aio_release(c);
-
-    replay_bh_schedule_oneshot_event(acb->s->aio_context, rbd_finish_bh, rcb);
+    aio_bh_schedule_oneshot(task->s->aio_context, qemu_rbd_finish_bh, task);
 }
 
-static BlockAIOCB *rbd_start_aio(BlockDriverState *bs,
-                                 int64_t off,
-                                 QEMUIOVector *qiov,
-                                 int64_t size,
-                                 BlockCompletionFunc *cb,
-                                 void *opaque,
-                                 RBDAIOCmd cmd)
+static int coroutine_fn qemu_rbd_start_co(BlockDriverState *bs,
+                                          uint64_t offset,
+                                          uint64_t bytes,
+                                          QEMUIOVector *qiov,
+                                          int flags,
+                                          RBDAIOCmd cmd)
 {
-    RBDAIOCB *acb;
-    RADOSCB *rcb = NULL;
-    rbd_completion_t c;
-    int r;
-
     BDRVRBDState *s = bs->opaque;
+    RBDTask task = { .s = s, .co = qemu_coroutine_self() };
+    rbd_completion_t c;
+    int r, ret = -EIO;
 
-    acb = qemu_aio_get(&rbd_aiocb_info, bs, cb, opaque);
-    acb->cmd = cmd;
-    acb->qiov = qiov;
-    assert(!qiov || qiov->size == size);
-
-    rcb = g_new(RADOSCB, 1);
-
-    acb->ret = 0;
-    acb->error = 0;
-    acb->s = s;
+    assert(!qiov || qiov->size == bytes);
 
-    rcb->acb = acb;
-    rcb->s = acb->s;
-    rcb->size = size;
-    r = rbd_aio_create_completion(rcb, (rbd_callback_t) rbd_finish_aiocb, &c);
+    r = rbd_aio_create_completion(&task,
+                                  (rbd_callback_t) qemu_rbd_completion_cb, &c);
     if (r < 0) {
         goto failed;
     }

block/rbd.c static int qemu_rbd_resize(BlockDriverState *bs, uint64_t size) · line 890

Quoted 10 lines
block/rbd.c +0 -8
 
     switch (cmd) {
-    case RBD_AIO_WRITE:
-        /*
-         * RBD APIs don't allow us to write more than actual size, so in order
-         * to support growing images, we resize the image before write
-         * operations that exceed the current size.
-         */
-        if (off + size > s->image_size) {
-            r = qemu_rbd_resize(bs, off + size);

I realize this is inherited code that is being refactored, but is it really possible for QEMU to issue IOs outside the block extents? This also has the same issue as the previous commit comment where the "image_size" is only initialized at image open time and therefore this path will keep getting hit if IOs are issued beyond the original block device extents.

Quoted 38 lines
block/rbd.c +17 -9
-            if (r < 0) {
-                goto failed_completion;
-            }
-        }
-        r = rbd_aio_writev(s->image, qiov->iov, qiov->niov, off, c);
-        break;
     case RBD_AIO_READ:
-        r = rbd_aio_readv(s->image, qiov->iov, qiov->niov, off, c);
+        r = rbd_aio_readv(s->image, qiov->iov, qiov->niov, offset, c);
+        break;
+    case RBD_AIO_WRITE:
+        r = rbd_aio_writev(s->image, qiov->iov, qiov->niov, offset, c);
         break;
     case RBD_AIO_DISCARD:
-        r = rbd_aio_discard(s->image, off, size, c);
+        r = rbd_aio_discard(s->image, offset, bytes, c);
         break;
     case RBD_AIO_FLUSH:
         r = rbd_aio_flush(s->image, c);
@@ -908,44 +813,71 @@ static BlockAIOCB *rbd_start_aio(BlockDriverState *bs,
     }
 
     if (r < 0) {
+        task.complete = 1;
+        task.ret = r;
+    }
+
+    while (!task.complete) {
+        qemu_coroutine_yield();
+    }
+
+    if (task.ret < 0) {
+        error_report("rbd request failed: cmd %d offset %" PRIu64 " bytes %"
+                     PRIu64 " flags %d task.ret %" PRIi64 " (%s)", cmd, offset,
+                     bytes, flags, task.ret, strerror(-task.ret));
         goto failed_completion;
     }
-    return &acb->common;

block/rbd.c static BlockAIOCB *rbd_start_aio(BlockDriverState *bs, · line 839

Quoted 10 lines
block/rbd.c +7 -0
+
+    /* zero pad short reads */
+    if (cmd == RBD_AIO_READ && task.ret < qiov->size) {
+        qemu_iovec_memset(qiov, task.ret, 0, qiov->size - task.ret);
+    }
+
+    return 0;
 
 failed_completion:
     rbd_aio_release(c);

Wasn't this already released in "qemu_rbd_completion_cb"?

Quoted 99 lines
block/rbd.c +34 -37
 failed:
-    g_free(rcb);
+    return ret;
+}
 
-    qemu_aio_unref(acb);
-    return NULL;
+static int
+coroutine_fn qemu_rbd_co_preadv(BlockDriverState *bs, uint64_t offset,
+                               uint64_t bytes, QEMUIOVector *qiov,
+                               int flags)
+{
+    return qemu_rbd_start_co(bs, offset, bytes, qiov, flags, RBD_AIO_READ);
 }
 
-static BlockAIOCB *qemu_rbd_aio_preadv(BlockDriverState *bs,
-                                       uint64_t offset, uint64_t bytes,
-                                       QEMUIOVector *qiov, int flags,
-                                       BlockCompletionFunc *cb,
-                                       void *opaque)
+static int
+coroutine_fn qemu_rbd_co_pwritev(BlockDriverState *bs, uint64_t offset,
+                                 uint64_t bytes, QEMUIOVector *qiov,
+                                 int flags)
 {
-    return rbd_start_aio(bs, offset, qiov, bytes, cb, opaque,
-                         RBD_AIO_READ);
+    BDRVRBDState *s = bs->opaque;
+    /*
+     * RBD APIs don't allow us to write more than actual size, so in order
+     * to support growing images, we resize the image before write
+     * operations that exceed the current size.
+     */
+    if (offset + bytes > s->image_size) {
+        int r = qemu_rbd_resize(bs, offset + bytes);
+        if (r < 0) {
+            return r;
+        }
+    }
+    return qemu_rbd_start_co(bs, offset, bytes, qiov, flags, RBD_AIO_WRITE);
 }
 
-static BlockAIOCB *qemu_rbd_aio_pwritev(BlockDriverState *bs,
-                                        uint64_t offset, uint64_t bytes,
-                                        QEMUIOVector *qiov, int flags,
-                                        BlockCompletionFunc *cb,
-                                        void *opaque)
+static int coroutine_fn qemu_rbd_co_flush(BlockDriverState *bs)
 {
-    return rbd_start_aio(bs, offset, qiov, bytes, cb, opaque,
-                         RBD_AIO_WRITE);
+    return qemu_rbd_start_co(bs, 0, 0, NULL, 0, RBD_AIO_FLUSH);
 }
 
-static BlockAIOCB *qemu_rbd_aio_flush(BlockDriverState *bs,
-                                      BlockCompletionFunc *cb,
-                                      void *opaque)
+static int coroutine_fn qemu_rbd_co_pdiscard(BlockDriverState *bs,
+                                             int64_t offset, int count)
 {
-    return rbd_start_aio(bs, 0, NULL, 0, cb, opaque, RBD_AIO_FLUSH);
+    return qemu_rbd_start_co(bs, offset, count, NULL, 0, RBD_AIO_DISCARD);
 }
 
 static int qemu_rbd_getinfo(BlockDriverState *bs, BlockDriverInfo *bdi)
@@ -1097,16 +1029,6 @@ static int qemu_rbd_snap_list(BlockDriverState *bs,
     return snap_count;
 }
 
-static BlockAIOCB *qemu_rbd_aio_pdiscard(BlockDriverState *bs,
-                                         int64_t offset,
-                                         int bytes,
-                                         BlockCompletionFunc *cb,
-                                         void *opaque)
-{
-    return rbd_start_aio(bs, offset, NULL, bytes, cb, opaque,
-                         RBD_AIO_DISCARD);
-}
-
 static void coroutine_fn qemu_rbd_co_invalidate_cache(BlockDriverState *bs,
                                                       Error **errp)
 {
@@ -1182,11 +1104,10 @@ static BlockDriver bdrv_rbd = {
     .bdrv_co_truncate       = qemu_rbd_co_truncate,
     .protocol_name          = "rbd",
 
-    .bdrv_aio_preadv        = qemu_rbd_aio_preadv,
-    .bdrv_aio_pwritev       = qemu_rbd_aio_pwritev,
-
-    .bdrv_aio_flush         = qemu_rbd_aio_flush,
-    .bdrv_aio_pdiscard      = qemu_rbd_aio_pdiscard,
+    .bdrv_co_preadv         = qemu_rbd_co_preadv,
+    .bdrv_co_pwritev        = qemu_rbd_co_pwritev,
+    .bdrv_co_flush_to_disk  = qemu_rbd_co_flush,
+    .bdrv_co_pdiscard       = qemu_rbd_co_pdiscard,
 
     .bdrv_snapshot_create   = qemu_rbd_snap_create,
     .bdrv_snapshot_delete   = qemu_rbd_snap_remove,
--
Quoted 3 lines

2.17.1

--
Jason

#11 lore raw

Re: [PATCH 5/7] block/rbd: migrate from aio to coroutines

Am 14.01.21 um 20:19 schrieb Jason Dillaman:

Quoted 193 lines · Peter Lieven
Quoted 192 lines
---
 block/rbd.c | 247 ++++++++++++++++++----------------------------------
 1 file changed, 84 insertions(+), 163 deletions(-)
block/rbd.c +26 -110
index 27b232f4d8..2d77d0007f 100644
@@ -66,22 +66,6 @@ typedef enum {
     RBD_AIO_FLUSH
 } RBDAIOCmd;
 
-typedef struct RBDAIOCB {
-    BlockAIOCB common;
-    int64_t ret;
-    QEMUIOVector *qiov;
-    RBDAIOCmd cmd;
-    int error;
-    struct BDRVRBDState *s;
-} RBDAIOCB;
-
-typedef struct RADOSCB {
-    RBDAIOCB *acb;
-    struct BDRVRBDState *s;
-    int64_t size;
-    int64_t ret;
-} RADOSCB;
-
 typedef struct BDRVRBDState {
     rados_t cluster;
     rados_ioctx_t io_ctx;
@@ -94,6 +78,13 @@ typedef struct BDRVRBDState {
     AioContext *aio_context;
 } BDRVRBDState;
 
+typedef struct RBDTask {
+    BDRVRBDState *s;
+    Coroutine *co;
+    bool complete;
+    int64_t ret;
+} RBDTask;
+
 static int qemu_rbd_connect(rados_t *cluster, rados_ioctx_t *io_ctx,
                             BlockdevOptionsRbd *opts, bool cache,
                             const char *keypairs, const char *secretid,
@@ -316,13 +307,6 @@ static int qemu_rbd_set_keypairs(rados_t cluster, const char *keypairs_json,
     return ret;
 }
 
-static void qemu_rbd_memset(RADOSCB *rcb, int64_t offs)
-{
-    RBDAIOCB *acb = rcb->acb;
-    iov_memset(acb->qiov->iov, acb->qiov->niov, offs, 0,
-               acb->qiov->size - offs);
-}
-
 /* FIXME Deprecate and remove keypairs or make it available in QMP. */
 static int qemu_rbd_do_create(BlockdevCreateOptions *options,
                               const char *keypairs, const char *password_secret,
@@ -440,46 +424,6 @@ exit:
     return ret;
 }
 
-/*
- * This aio completion is being called from rbd_finish_bh() and runs in qemu
- * BH context.
- */
-static void qemu_rbd_complete_aio(RADOSCB *rcb)
-{
-    RBDAIOCB *acb = rcb->acb;
-    int64_t r;
-
-    r = rcb->ret;
-
-    if (acb->cmd != RBD_AIO_READ) {
-        if (r < 0) {
-            acb->ret = r;
-            acb->error = 1;
-        } else if (!acb->error) {
-            acb->ret = rcb->size;
-        }
-    } else {
-        if (r < 0) {
-            qemu_rbd_memset(rcb, 0);
-            acb->ret = r;
-            acb->error = 1;
-        } else if (r < rcb->size) {
-            qemu_rbd_memset(rcb, r);
-            if (!acb->error) {
-                acb->ret = rcb->size;
-            }
-        } else if (!acb->error) {
-            acb->ret = r;
-        }
-    }
-
-    g_free(rcb);
-
-    acb->common.cb(acb->common.opaque, (acb->ret > 0 ? 0 : acb->ret));
-
-    qemu_aio_unref(acb);
-}
-
 static char *qemu_rbd_mon_host(BlockdevOptionsRbd *opts, Error **errp)
 {
     const char **vals;
@@ -817,88 +761,49 @@ static int qemu_rbd_resize(BlockDriverState *bs, uint64_t size)
     return 0;
 }
 
-static const AIOCBInfo rbd_aiocb_info = {
-    .aiocb_size = sizeof(RBDAIOCB),
-};
-
-static void rbd_finish_bh(void *opaque)
+static void qemu_rbd_finish_bh(void *opaque)
 {
-    RADOSCB *rcb = opaque;
-    qemu_rbd_complete_aio(rcb);
+    RBDTask *task = opaque;
+    task->complete = 1;
+    aio_co_wake(task->co);
 }
 
-/*
- * This is the callback function for rbd_aio_read and _write
- *
- * Note: this function is being called from a non qemu thread so
- * we need to be careful about what we do here. Generally we only
- * schedule a BH, and do the rest of the io completion handling
- * from rbd_finish_bh() which runs in a qemu context.
- */
-static void rbd_finish_aiocb(rbd_completion_t c, RADOSCB *rcb)
+static void qemu_rbd_completion_cb(rbd_completion_t c, RBDTask *task)
 {
-    RBDAIOCB *acb = rcb->acb;
-
-    rcb->ret = rbd_aio_get_return_value(c);
+    task->ret = rbd_aio_get_return_value(c);
     rbd_aio_release(c);
-
-    replay_bh_schedule_oneshot_event(acb->s->aio_context, rbd_finish_bh, rcb);
+    aio_bh_schedule_oneshot(task->s->aio_context, qemu_rbd_finish_bh, task);
 }
 
-static BlockAIOCB *rbd_start_aio(BlockDriverState *bs,
-                                 int64_t off,
-                                 QEMUIOVector *qiov,
-                                 int64_t size,
-                                 BlockCompletionFunc *cb,
-                                 void *opaque,
-                                 RBDAIOCmd cmd)
+static int coroutine_fn qemu_rbd_start_co(BlockDriverState *bs,
+                                          uint64_t offset,
+                                          uint64_t bytes,
+                                          QEMUIOVector *qiov,
+                                          int flags,
+                                          RBDAIOCmd cmd)
 {
-    RBDAIOCB *acb;
-    RADOSCB *rcb = NULL;
-    rbd_completion_t c;
-    int r;
-
     BDRVRBDState *s = bs->opaque;
+    RBDTask task = { .s = s, .co = qemu_coroutine_self() };
+    rbd_completion_t c;
+    int r, ret = -EIO;
 
-    acb = qemu_aio_get(&rbd_aiocb_info, bs, cb, opaque);
-    acb->cmd = cmd;
-    acb->qiov = qiov;
-    assert(!qiov || qiov->size == size);
-
-    rcb = g_new(RADOSCB, 1);
-
-    acb->ret = 0;
-    acb->error = 0;
-    acb->s = s;
+    assert(!qiov || qiov->size == bytes);
 
-    rcb->acb = acb;
-    rcb->s = acb->s;
-    rcb->size = size;
-    r = rbd_aio_create_completion(rcb, (rbd_callback_t) rbd_finish_aiocb, &c);
+    r = rbd_aio_create_completion(&task,
+                                  (rbd_callback_t) qemu_rbd_completion_cb, &c);
     if (r < 0) {
         goto failed;
     }
 
Quoted 15 lines

block/rbd.c static int qemu_rbd_resize(BlockDriverState *bs, uint64_t size) · line 890

Quoted 9 lines
block/rbd.c +0 -8
     switch (cmd) {
-    case RBD_AIO_WRITE:
-        /*
-         * RBD APIs don't allow us to write more than actual size, so in order
-         * to support growing images, we resize the image before write
-         * operations that exceed the current size.
-         */
-        if (off + size > s->image_size) {
-            r = qemu_rbd_resize(bs, off + size);

I realize this is inherited code that is being refactored, but is it really possible for QEMU to issue IOs outside the block extents? This also has the same issue as the previous commit comment where the "image_size" is only initialized at image open time and therefore this path will keep getting hit if IOs are issued beyond the original block device extents.

This is for the case that you came to the idea that putting e.g. a qcow2 file on rbd which actually can grow during runtime.

If you put a raw image on an rbd image the only way to grow is with block_resize (which invoces truncate) or with qemu-img resize...

I personally was unsure if it makes sense at all to supprt anything except raw on rbd, but it was supported before so I left this in.

Quoted 47 lines
Quoted 47 lines
block/rbd.c +24 -9
-            if (r < 0) {
-                goto failed_completion;
-            }
-        }
-        r = rbd_aio_writev(s->image, qiov->iov, qiov->niov, off, c);
-        break;
     case RBD_AIO_READ:
-        r = rbd_aio_readv(s->image, qiov->iov, qiov->niov, off, c);
+        r = rbd_aio_readv(s->image, qiov->iov, qiov->niov, offset, c);
+        break;
+    case RBD_AIO_WRITE:
+        r = rbd_aio_writev(s->image, qiov->iov, qiov->niov, offset, c);
         break;
     case RBD_AIO_DISCARD:
-        r = rbd_aio_discard(s->image, off, size, c);
+        r = rbd_aio_discard(s->image, offset, bytes, c);
         break;
     case RBD_AIO_FLUSH:
         r = rbd_aio_flush(s->image, c);
@@ -908,44 +813,71 @@ static BlockAIOCB *rbd_start_aio(BlockDriverState *bs,
     }
 
     if (r < 0) {
+        task.complete = 1;
+        task.ret = r;
+    }
+
+    while (!task.complete) {
+        qemu_coroutine_yield();
+    }
+
+    if (task.ret < 0) {
+        error_report("rbd request failed: cmd %d offset %" PRIu64 " bytes %"
+                     PRIu64 " flags %d task.ret %" PRIi64 " (%s)", cmd, offset,
+                     bytes, flags, task.ret, strerror(-task.ret));
         goto failed_completion;
     }
-    return &acb->common;
+
+    /* zero pad short reads */
+    if (cmd == RBD_AIO_READ && task.ret < qiov->size) {
+        qemu_iovec_memset(qiov, task.ret, 0, qiov->size - task.ret);
+    }
+
+    return 0;
 
Quoted 3 lines

block/rbd.c static BlockAIOCB *rbd_start_aio(BlockDriverState *bs, · line 839

Quoted 2 lines
block/rbd.c +0 -0
 failed_completion:
     rbd_aio_release(c);

Wasn't this already released in "qemu_rbd_completion_cb"?

You are right, in case of task.ret < 0 we need to go to "failed" and not "failed_completion".

If the I/O is successful we leave the coroutine with "return 0" just above the label.

Peter

#12 lore raw

PATCH 6/7 block/rbd: add write zeroes support

In reply to Peter Lieven
---
 block/rbd.c | 31 ++++++++++++++++++++++++++++++-
 1 file changed, 30 insertions(+), 1 deletion(-)
block/rbd.c +30 -1
index 2d77d0007f..27b4404adf 100644
@@ -63,7 +63,8 @@ typedef enum {
     RBD_AIO_READ,
     RBD_AIO_WRITE,
     RBD_AIO_DISCARD,
-    RBD_AIO_FLUSH
+    RBD_AIO_FLUSH,
+    RBD_AIO_WRITE_ZEROES
 } RBDAIOCmd;
 
 typedef struct BDRVRBDState {
@@ -221,8 +222,12 @@ done:
 
 static void qemu_rbd_refresh_limits(BlockDriverState *bs, Error **errp)
 {
+    BDRVRBDState *s = bs->opaque;
     /* XXX Does RBD support AIO on less than 512-byte alignment? */
     bs->bl.request_alignment = 512;
+#ifdef LIBRBD_SUPPORTS_WRITE_ZEROES
+    bs->bl.pwrite_zeroes_alignment = s->object_size;
+#endif
 }
 
 
@@ -695,6 +700,9 @@ static int qemu_rbd_open(BlockDriverState *bs, QDict *options, int flags,
     }
 
     s->aio_context = bdrv_get_aio_context(bs);
+#ifdef LIBRBD_SUPPORTS_WRITE_ZEROES
+    bs->supported_zero_flags = BDRV_REQ_MAY_UNMAP;
+#endif
 
     /* When extending regular files, we get zeros from the OS */
     bs->supported_truncate_flags = BDRV_REQ_ZERO_WRITE;
@@ -808,6 +816,11 @@ static int coroutine_fn qemu_rbd_start_co(BlockDriverState *bs,
     case RBD_AIO_FLUSH:
         r = rbd_aio_flush(s->image, c);
         break;
+#ifdef LIBRBD_SUPPORTS_WRITE_ZEROES
+    case RBD_AIO_WRITE_ZEROES:
+        r = rbd_aio_write_zeroes(s->image, offset, bytes, c, 0, 0);
+        break;
+#endif
     default:
         r = -EINVAL;
     }
@@ -880,6 +893,19 @@ static int coroutine_fn qemu_rbd_co_pdiscard(BlockDriverState *bs,
     return qemu_rbd_start_co(bs, offset, count, NULL, 0, RBD_AIO_DISCARD);
 }
 
+#ifdef LIBRBD_SUPPORTS_WRITE_ZEROES
+static int
+coroutine_fn qemu_rbd_co_pwrite_zeroes(BlockDriverState *bs, int64_t offset,
+                                      int count, BdrvRequestFlags flags)
+{
+    if (!(flags & BDRV_REQ_MAY_UNMAP)) {
+        return -ENOTSUP;
+    }
+    return qemu_rbd_start_co(bs, offset, count, NULL, flags,
+                             RBD_AIO_WRITE_ZEROES);
+}
+#endif
+
 static int qemu_rbd_getinfo(BlockDriverState *bs, BlockDriverInfo *bdi)
 {
     BDRVRBDState *s = bs->opaque;
@@ -1108,6 +1134,9 @@ static BlockDriver bdrv_rbd = {
     .bdrv_co_pwritev        = qemu_rbd_co_pwritev,
     .bdrv_co_flush_to_disk  = qemu_rbd_co_flush,
     .bdrv_co_pdiscard       = qemu_rbd_co_pdiscard,
+#ifdef LIBRBD_SUPPORTS_WRITE_ZEROES
+    .bdrv_co_pwrite_zeroes  = qemu_rbd_co_pwrite_zeroes,
+#endif
 
     .bdrv_snapshot_create   = qemu_rbd_snap_create,
     .bdrv_snapshot_delete   = qemu_rbd_snap_remove,
Signature · 1 line
2.17.1
#13 lore raw

Re: [PATCH 6/7] block/rbd: add write zeroes support

Quoted 21 lines · Peter Lieven
---
 block/rbd.c | 31 ++++++++++++++++++++++++++++++-
 1 file changed, 30 insertions(+), 1 deletion(-)
block/rbd.c +2 -1
index 2d77d0007f..27b4404adf 100644
@@ -63,7 +63,8 @@ typedef enum {
     RBD_AIO_READ,
     RBD_AIO_WRITE,
     RBD_AIO_DISCARD,
-    RBD_AIO_FLUSH
+    RBD_AIO_FLUSH,
+    RBD_AIO_WRITE_ZEROES
 } RBDAIOCmd;
 
 typedef struct BDRVRBDState {

block/rbd.c done: · line 229

Quoted 8 lines
block/rbd.c +3 -0
@@ -221,8 +222,12 @@ done:
 
 static void qemu_rbd_refresh_limits(BlockDriverState *bs, Error **errp)
 {
+    BDRVRBDState *s = bs->opaque;
     /* XXX Does RBD support AIO on less than 512-byte alignment? */
     bs->bl.request_alignment = 512;
+#ifdef LIBRBD_SUPPORTS_WRITE_ZEROES
+    bs->bl.pwrite_zeroes_alignment = s->object_size;

The optimal alignment is 512 bytes -- but it can safely work just fine down to 1 byte alignment since it will pad the request with additional writes if needed.

Quoted 28 lines
block/rbd.c +9 -0
+#endif
 }
 
 
@@ -695,6 +700,9 @@ static int qemu_rbd_open(BlockDriverState *bs, QDict *options, int flags,
     }
 
     s->aio_context = bdrv_get_aio_context(bs);
+#ifdef LIBRBD_SUPPORTS_WRITE_ZEROES
+    bs->supported_zero_flags = BDRV_REQ_MAY_UNMAP;
+#endif
 
     /* When extending regular files, we get zeros from the OS */
     bs->supported_truncate_flags = BDRV_REQ_ZERO_WRITE;
@@ -808,6 +816,11 @@ static int coroutine_fn qemu_rbd_start_co(BlockDriverState *bs,
     case RBD_AIO_FLUSH:
         r = rbd_aio_flush(s->image, c);
         break;
+#ifdef LIBRBD_SUPPORTS_WRITE_ZEROES
+    case RBD_AIO_WRITE_ZEROES:
+        r = rbd_aio_write_zeroes(s->image, offset, bytes, c, 0, 0);
+        break;
+#endif
     default:
         r = -EINVAL;
     }
@@ -880,6 +893,19 @@ static int coroutine_fn qemu_rbd_co_pdiscard(BlockDriverState *bs,
     return qemu_rbd_start_co(bs, offset, count, NULL, 0, RBD_AIO_DISCARD);

block/rbd.c static int coroutine_fn qemu_rbd_co_pdiscard(BlockDriverState *bs, · line 903

Quoted 10 lines
block/rbd.c +8 -0
 }
 
+#ifdef LIBRBD_SUPPORTS_WRITE_ZEROES
+static int
+coroutine_fn qemu_rbd_co_pwrite_zeroes(BlockDriverState *bs, int64_t offset,
+                                      int count, BdrvRequestFlags flags)
+{
+    if (!(flags & BDRV_REQ_MAY_UNMAP)) {
+        return -ENOTSUP;
+    }

There is a "RBD_WRITE_ZEROES_FLAG_THICK_PROVISION" flag that you can use to optionally disable unmap.

Quoted 19 lines
block/rbd.c +8 -1
+    return qemu_rbd_start_co(bs, offset, count, NULL, flags,
+                             RBD_AIO_WRITE_ZEROES);
+}
+#endif
+
 static int qemu_rbd_getinfo(BlockDriverState *bs, BlockDriverInfo *bdi)
 {
     BDRVRBDState *s = bs->opaque;
@@ -1108,6 +1134,9 @@ static BlockDriver bdrv_rbd = {
     .bdrv_co_pwritev        = qemu_rbd_co_pwritev,
     .bdrv_co_flush_to_disk  = qemu_rbd_co_flush,
     .bdrv_co_pdiscard       = qemu_rbd_co_pdiscard,
+#ifdef LIBRBD_SUPPORTS_WRITE_ZEROES
+    .bdrv_co_pwrite_zeroes  = qemu_rbd_co_pwrite_zeroes,
+#endif
 
     .bdrv_snapshot_create   = qemu_rbd_snap_create,
     .bdrv_snapshot_delete   = qemu_rbd_snap_remove,
--
Quoted 3 lines

2.17.1

--
Jason

#14 lore raw

Re: [PATCH 6/7] block/rbd: add write zeroes support

Am 14.01.21 um 20:19 schrieb Jason Dillaman:

Quoted 22 lines · Peter Lieven
Quoted 21 lines
---
 block/rbd.c | 31 ++++++++++++++++++++++++++++++-
 1 file changed, 30 insertions(+), 1 deletion(-)
block/rbd.c +2 -1
index 2d77d0007f..27b4404adf 100644
@@ -63,7 +63,8 @@ typedef enum {
     RBD_AIO_READ,
     RBD_AIO_WRITE,
     RBD_AIO_DISCARD,
-    RBD_AIO_FLUSH
+    RBD_AIO_FLUSH,
+    RBD_AIO_WRITE_ZEROES
 } RBDAIOCmd;
 
 typedef struct BDRVRBDState {
@@ -221,8 +222,12 @@ done:
 
Quoted 10 lines

block/rbd.c done: · line 229

Quoted 7 lines
block/rbd.c +3 -0
 static void qemu_rbd_refresh_limits(BlockDriverState *bs, Error **errp)
 {
+    BDRVRBDState *s = bs->opaque;
     /* XXX Does RBD support AIO on less than 512-byte alignment? */
     bs->bl.request_alignment = 512;
+#ifdef LIBRBD_SUPPORTS_WRITE_ZEROES
+    bs->bl.pwrite_zeroes_alignment = s->object_size;

The optimal alignment is 512 bytes -- but it can safely work just fine down to 1 byte alignment since it will pad the request with additional writes if needed.

Okay and this will likely be faster than having Qemu doing that request split, right?

If we drop the alignment hint Qemu will pass the original request.

Quoted 31 lines
Quoted 31 lines
block/rbd.c +9 -0
+#endif
 }
 
 
@@ -695,6 +700,9 @@ static int qemu_rbd_open(BlockDriverState *bs, QDict *options, int flags,
     }
 
     s->aio_context = bdrv_get_aio_context(bs);
+#ifdef LIBRBD_SUPPORTS_WRITE_ZEROES
+    bs->supported_zero_flags = BDRV_REQ_MAY_UNMAP;
+#endif
 
     /* When extending regular files, we get zeros from the OS */
     bs->supported_truncate_flags = BDRV_REQ_ZERO_WRITE;
@@ -808,6 +816,11 @@ static int coroutine_fn qemu_rbd_start_co(BlockDriverState *bs,
     case RBD_AIO_FLUSH:
         r = rbd_aio_flush(s->image, c);
         break;
+#ifdef LIBRBD_SUPPORTS_WRITE_ZEROES
+    case RBD_AIO_WRITE_ZEROES:
+        r = rbd_aio_write_zeroes(s->image, offset, bytes, c, 0, 0);
+        break;
+#endif
     default:
         r = -EINVAL;
     }
@@ -880,6 +893,19 @@ static int coroutine_fn qemu_rbd_co_pdiscard(BlockDriverState *bs,
     return qemu_rbd_start_co(bs, offset, count, NULL, 0, RBD_AIO_DISCARD);
 }
 
Quoted 10 lines

block/rbd.c static int coroutine_fn qemu_rbd_co_pdiscard(BlockDriverState *bs, · line 903

Quoted 8 lines
block/rbd.c +8 -0
+#ifdef LIBRBD_SUPPORTS_WRITE_ZEROES
+static int
+coroutine_fn qemu_rbd_co_pwrite_zeroes(BlockDriverState *bs, int64_t offset,
+                                      int count, BdrvRequestFlags flags)
+{
+    if (!(flags & BDRV_REQ_MAY_UNMAP)) {
+        return -ENOTSUP;
+    }

There is a "RBD_WRITE_ZEROES_FLAG_THICK_PROVISION" flag that you can use to optionally disable unmap.

I have seen that. If you want I can support for this, too. But afaik this

is only available since Octopus release?

Peter

#15 lore raw

Re: [PATCH 6/7] block/rbd: add write zeroes support

Quoted 84 lines · Peter Lieven

Am 14.01.21 um 20:19 schrieb Jason Dillaman:

Quoted 22 lines · Peter Lieven
Quoted 21 lines
---
 block/rbd.c | 31 ++++++++++++++++++++++++++++++-
 1 file changed, 30 insertions(+), 1 deletion(-)
block/rbd.c +2 -1
index 2d77d0007f..27b4404adf 100644
@@ -63,7 +63,8 @@ typedef enum {
     RBD_AIO_READ,
     RBD_AIO_WRITE,
     RBD_AIO_DISCARD,
-    RBD_AIO_FLUSH
+    RBD_AIO_FLUSH,
+    RBD_AIO_WRITE_ZEROES
 } RBDAIOCmd;
 
 typedef struct BDRVRBDState {
@@ -221,8 +222,12 @@ done:
 
Quoted 10 lines

block/rbd.c done: · line 229

Quoted 7 lines
block/rbd.c +3 -0
 static void qemu_rbd_refresh_limits(BlockDriverState *bs, Error **errp)
 {
+    BDRVRBDState *s = bs->opaque;
     /* XXX Does RBD support AIO on less than 512-byte alignment? */
     bs->bl.request_alignment = 512;
+#ifdef LIBRBD_SUPPORTS_WRITE_ZEROES
+    bs->bl.pwrite_zeroes_alignment = s->object_size;

The optimal alignment is 512 bytes -- but it can safely work just fine down to 1 byte alignment since it will pad the request with additional writes if needed.

Okay and this will likely be faster than having Qemu doing that request split, right?

If we drop the alignment hint Qemu will pass the original request.

Quoted 31 lines
Quoted 31 lines
block/rbd.c +9 -0
+#endif
 }
 
 
@@ -695,6 +700,9 @@ static int qemu_rbd_open(BlockDriverState *bs, QDict *options, int flags,
     }
 
     s->aio_context = bdrv_get_aio_context(bs);
+#ifdef LIBRBD_SUPPORTS_WRITE_ZEROES
+    bs->supported_zero_flags = BDRV_REQ_MAY_UNMAP;
+#endif
 
     /* When extending regular files, we get zeros from the OS */
     bs->supported_truncate_flags = BDRV_REQ_ZERO_WRITE;
@@ -808,6 +816,11 @@ static int coroutine_fn qemu_rbd_start_co(BlockDriverState *bs,
     case RBD_AIO_FLUSH:
         r = rbd_aio_flush(s->image, c);
         break;
+#ifdef LIBRBD_SUPPORTS_WRITE_ZEROES
+    case RBD_AIO_WRITE_ZEROES:
+        r = rbd_aio_write_zeroes(s->image, offset, bytes, c, 0, 0);
+        break;
+#endif
     default:
         r = -EINVAL;
     }
@@ -880,6 +893,19 @@ static int coroutine_fn qemu_rbd_co_pdiscard(BlockDriverState *bs,
     return qemu_rbd_start_co(bs, offset, count, NULL, 0, RBD_AIO_DISCARD);
 }
 
Quoted 10 lines

block/rbd.c static int coroutine_fn qemu_rbd_co_pdiscard(BlockDriverState *bs, · line 903

Quoted 8 lines
block/rbd.c +8 -0
+#ifdef LIBRBD_SUPPORTS_WRITE_ZEROES
+static int
+coroutine_fn qemu_rbd_co_pwrite_zeroes(BlockDriverState *bs, int64_t offset,
+                                      int count, BdrvRequestFlags flags)
+{
+    if (!(flags & BDRV_REQ_MAY_UNMAP)) {
+        return -ENOTSUP;
+    }

There is a "RBD_WRITE_ZEROES_FLAG_THICK_PROVISION" flag that you can use to optionally disable unmap.

Quoted 3 lines

I have seen that. If you want I can support for this, too. But afaik this

is only available since Octopus release?

True -- I didn't backport that support to Nautilus since it was a new feature vs the bug-fix that the write-zeroes API was introduced to fix.

Quoted 4 lines

Peter

Signature · 1 line
Jason
#16 lore raw

PATCH 7/7 block/rbd: change request alignment to 1 byte

In reply to Peter Lieven

since we implement byte interfaces and librbd supports aio on byte granularity we can lift the 512 byte alignment.

---
 block/rbd.c | 2 --
 1 file changed, 2 deletions(-)
block/rbd.c +0 -2
index 27b4404adf..8673e8f553 100644
@@ -223,8 +223,6 @@ done:
 static void qemu_rbd_refresh_limits(BlockDriverState *bs, Error **errp)
 {
     BDRVRBDState *s = bs->opaque;
-    /* XXX Does RBD support AIO on less than 512-byte alignment? */
-    bs->bl.request_alignment = 512;
 #ifdef LIBRBD_SUPPORTS_WRITE_ZEROES
     bs->bl.pwrite_zeroes_alignment = s->object_size;
 #endif
Signature · 1 line
2.17.1
#17 lore raw

Re: [PATCH 7/7] block/rbd: change request alignment to 1 byte

Quoted 14 lines · Peter Lieven

since we implement byte interfaces and librbd supports aio on byte granularity we can lift the 512 byte alignment.

---
 block/rbd.c | 2 --
 1 file changed, 2 deletions(-)
block/rbd.c +0 -0
index 27b4404adf..8673e8f553 100644

block/rbd.c done: · line 227

Quoted 5 lines
block/rbd.c +0 -2
@@ -223,8 +223,6 @@ done:
 static void qemu_rbd_refresh_limits(BlockDriverState *bs, Error **errp)
 {
     BDRVRBDState *s = bs->opaque;
-    /* XXX Does RBD support AIO on less than 512-byte alignment? */
-    bs->bl.request_alignment = 512;

Just a suggestion, but perhaps improve discard alignment, max discard, optimal alignment (if that's something QEMU handles internally) if not overridden by the user.

Quoted 7 lines
block/rbd.c +0 -1
 #ifdef LIBRBD_SUPPORTS_WRITE_ZEROES
     bs->bl.pwrite_zeroes_alignment = s->object_size;
 #endif
--

2.17.1

--
Jason

#18 lore raw

Re: [PATCH 7/7] block/rbd: change request alignment to 1 byte

Am 14.01.21 um 20:19 schrieb Jason Dillaman:

Quoted 9 lines · Peter Lieven
Quoted 8 lines

since we implement byte interfaces and librbd supports aio on byte granularity we can lift the 512 byte alignment.

---
 block/rbd.c | 2 --
 1 file changed, 2 deletions(-)
Quoted 13 lines

block/rbd.c done: · line 227

Quoted 10 lines
block/rbd.c +0 -2
index 27b4404adf..8673e8f553 100644
@@ -223,8 +223,6 @@ done:
 static void qemu_rbd_refresh_limits(BlockDriverState *bs, Error **errp)
 {
     BDRVRBDState *s = bs->opaque;
-    /* XXX Does RBD support AIO on less than 512-byte alignment? */
-    bs->bl.request_alignment = 512;

Just a suggestion, but perhaps improve discard alignment, max discard, optimal alignment (if that's something QEMU handles internally) if not overridden by the user.

Qemu supports max_discard and discard_alignment. Is there a call to get these limits

from librbd?

What do you mean by optimal_alignment? The object size?

Peter

#19 lore raw

Re: [PATCH 7/7] block/rbd: change request alignment to 1 byte

Quoted 26 lines · Peter Lieven

Am 14.01.21 um 20:19 schrieb Jason Dillaman:

Quoted 9 lines · Peter Lieven
Quoted 8 lines

since we implement byte interfaces and librbd supports aio on byte granularity we can lift the 512 byte alignment.

---
 block/rbd.c | 2 --
 1 file changed, 2 deletions(-)
Quoted 13 lines

block/rbd.c done: · line 227

Quoted 10 lines
block/rbd.c +0 -2
index 27b4404adf..8673e8f553 100644
@@ -223,8 +223,6 @@ done:
 static void qemu_rbd_refresh_limits(BlockDriverState *bs, Error **errp)
 {
     BDRVRBDState *s = bs->opaque;
-    /* XXX Does RBD support AIO on less than 512-byte alignment? */
-    bs->bl.request_alignment = 512;

Just a suggestion, but perhaps improve discard alignment, max discard, optimal alignment (if that's something QEMU handles internally) if not overridden by the user.

Quoted 6 lines

Qemu supports max_discard and discard_alignment. Is there a call to get these limits

from librbd?

What do you mean by optimal_alignment? The object size?

krbd does a good job of initializing defaults [1] where optimal and discard alignment is 64KiB (can actually be 4KiB now), max IO size for writes, discards, and write-zeroes is the object size * the stripe count.

Quoted 4 lines

Peter

[1] https://github.com/torvalds/linux/blob/master/drivers/block/rbd.c#L4981

Signature · 1 line
Jason
#20 lore raw

Re: [PATCH 7/7] block/rbd: change request alignment to 1 byte

Am 15.01.21 um 16:27 schrieb Jason Dillaman:

Quoted 27 lines · Peter Lieven
Quoted 26 lines

Am 14.01.21 um 20:19 schrieb Jason Dillaman:

Quoted 9 lines · Peter Lieven
Quoted 8 lines

since we implement byte interfaces and librbd supports aio on byte granularity we can lift the 512 byte alignment.

---
 block/rbd.c | 2 --
 1 file changed, 2 deletions(-)
Quoted 13 lines

block/rbd.c done: · line 227

Quoted 10 lines
block/rbd.c +0 -2
index 27b4404adf..8673e8f553 100644
@@ -223,8 +223,6 @@ done:
 static void qemu_rbd_refresh_limits(BlockDriverState *bs, Error **errp)
 {
     BDRVRBDState *s = bs->opaque;
-    /* XXX Does RBD support AIO on less than 512-byte alignment? */
-    bs->bl.request_alignment = 512;

Just a suggestion, but perhaps improve discard alignment, max discard, optimal alignment (if that's something QEMU handles internally) if not overridden by the user.

Qemu supports max_discard and discard_alignment. Is there a call to get these limits

Quoted 8 lines
Quoted 4 lines

from librbd?

What do you mean by optimal_alignment? The object size?

krbd does a good job of initializing defaults [1] where optimal and discard alignment is 64KiB (can actually be 4KiB now), max IO size for writes, discards, and write-zeroes is the object size * the stripe count.

Okay, I will have a look at it. If qemu issues a write, discard, write_zero greater than

obj_size  * stripe count will librbd split it internally or will the request fail?

Regarding the alignment it seems that rbd_dev->opts->alloc_size is something that comes from the device

configuration and not from rbd? I don't have that information inside the Qemu RBD driver.

Peter

#21 lore raw

Re: [PATCH 7/7] block/rbd: change request alignment to 1 byte

Quoted 39 lines · Peter Lieven

Am 15.01.21 um 16:27 schrieb Jason Dillaman:

Quoted 27 lines · Peter Lieven
Quoted 26 lines

Am 14.01.21 um 20:19 schrieb Jason Dillaman:

Quoted 9 lines · Peter Lieven
Quoted 8 lines

since we implement byte interfaces and librbd supports aio on byte granularity we can lift the 512 byte alignment.

---
 block/rbd.c | 2 --
 1 file changed, 2 deletions(-)
Quoted 13 lines

block/rbd.c done: · line 227

Quoted 10 lines
block/rbd.c +0 -2
index 27b4404adf..8673e8f553 100644
@@ -223,8 +223,6 @@ done:
 static void qemu_rbd_refresh_limits(BlockDriverState *bs, Error **errp)
 {
     BDRVRBDState *s = bs->opaque;
-    /* XXX Does RBD support AIO on less than 512-byte alignment? */
-    bs->bl.request_alignment = 512;

Just a suggestion, but perhaps improve discard alignment, max discard, optimal alignment (if that's something QEMU handles internally) if not overridden by the user.

Qemu supports max_discard and discard_alignment. Is there a call to get these limits

Quoted 8 lines
Quoted 4 lines

from librbd?

What do you mean by optimal_alignment? The object size?

krbd does a good job of initializing defaults [1] where optimal and discard alignment is 64KiB (can actually be 4KiB now), max IO size for writes, discards, and write-zeroes is the object size * the stripe count.

Quoted 3 lines

Okay, I will have a look at it. If qemu issues a write, discard, write_zero greater than

obj_size * stripe count will librbd split it internally or will the request fail?

librbd will handle it as needed. My goal is really just to get the hints down the guest OS.

Quoted 3 lines

Regarding the alignment it seems that rbd_dev->opts->alloc_size is something that comes from the device

configuration and not from rbd? I don't have that information inside the Qemu RBD driver.

librbd doesn't really have the information either. The 64KiB guess that krbd uses was a compromise since that was the default OSD allocation size for HDDs since Luminous. Starting with Pacific that default is going down to 4KiB.

Quoted 4 lines

Peter

Signature · 1 line
Jason
#22 lore raw

Re: [PATCH 7/7] block/rbd: change request alignment to 1 byte

Am 18.01.21 um 23:33 schrieb Jason Dillaman:

Quoted 43 lines · Peter Lieven
Quoted 31 lines

Am 15.01.21 um 16:27 schrieb Jason Dillaman:

Quoted 30 lines · Peter Lieven
Quoted 29 lines

Am 14.01.21 um 20:19 schrieb Jason Dillaman:

Quoted 9 lines · Peter Lieven
Quoted 8 lines

since we implement byte interfaces and librbd supports aio on byte granularity we can lift the 512 byte alignment.

---
 block/rbd.c | 2 --
 1 file changed, 2 deletions(-)
Quoted 13 lines

block/rbd.c done: · line 227

Quoted 10 lines
block/rbd.c +0 -2
index 27b4404adf..8673e8f553 100644
@@ -223,8 +223,6 @@ done:
 static void qemu_rbd_refresh_limits(BlockDriverState *bs, Error **errp)
 {
     BDRVRBDState *s = bs->opaque;
-    /* XXX Does RBD support AIO on less than 512-byte alignment? */
-    bs->bl.request_alignment = 512;

Just a suggestion, but perhaps improve discard alignment, max discard, optimal alignment (if that's something QEMU handles internally) if not overridden by the user.

Qemu supports max_discard and discard_alignment. Is there a call to get these limits

from librbd?

What do you mean by optimal_alignment? The object size?

Quoted 8 lines
Quoted 4 lines

krbd does a good job of initializing defaults [1] where optimal and discard alignment is 64KiB (can actually be 4KiB now), max IO size for writes, discards, and write-zeroes is the object size * the stripe count.

Okay, I will have a look at it. If qemu issues a write, discard, write_zero greater than

obj_size * stripe count will librbd split it internally or will the request fail?

librbd will handle it as needed. My goal is really just to get the hints down the guest OS.

Quoted 7 lines
Quoted 3 lines

Regarding the alignment it seems that rbd_dev->opts->alloc_size is something that comes from the device

configuration and not from rbd? I don't have that information inside the Qemu RBD driver.

librbd doesn't really have the information either. The 64KiB guess that krbd uses was a compromise since that was the default OSD allocation size for HDDs since Luminous. Starting with Pacific that default is going down to 4KiB.

I will try to adjust these values as far as it is possible and makes sense.

Is there a way to check the minimum supported OSD release in the backend from librbd / librados?

Anyway, I want to sent a V2 by the end of this week.

Peter

#23 lore raw

Re: [PATCH 7/7] block/rbd: change request alignment to 1 byte

Quoted 54 lines · Peter Lieven

Am 18.01.21 um 23:33 schrieb Jason Dillaman:

Quoted 43 lines · Peter Lieven
Quoted 31 lines

Am 15.01.21 um 16:27 schrieb Jason Dillaman:

Quoted 30 lines · Peter Lieven
Quoted 29 lines

Am 14.01.21 um 20:19 schrieb Jason Dillaman:

Quoted 9 lines · Peter Lieven
Quoted 8 lines

since we implement byte interfaces and librbd supports aio on byte granularity we can lift the 512 byte alignment.

---
 block/rbd.c | 2 --
 1 file changed, 2 deletions(-)
Quoted 13 lines

block/rbd.c done: · line 227

Quoted 10 lines
block/rbd.c +0 -2
index 27b4404adf..8673e8f553 100644
@@ -223,8 +223,6 @@ done:
 static void qemu_rbd_refresh_limits(BlockDriverState *bs, Error **errp)
 {
     BDRVRBDState *s = bs->opaque;
-    /* XXX Does RBD support AIO on less than 512-byte alignment? */
-    bs->bl.request_alignment = 512;

Just a suggestion, but perhaps improve discard alignment, max discard, optimal alignment (if that's something QEMU handles internally) if not overridden by the user.

Qemu supports max_discard and discard_alignment. Is there a call to get these limits

from librbd?

What do you mean by optimal_alignment? The object size?

Quoted 8 lines
Quoted 4 lines

krbd does a good job of initializing defaults [1] where optimal and discard alignment is 64KiB (can actually be 4KiB now), max IO size for writes, discards, and write-zeroes is the object size * the stripe count.

Okay, I will have a look at it. If qemu issues a write, discard, write_zero greater than

obj_size * stripe count will librbd split it internally or will the request fail?

librbd will handle it as needed. My goal is really just to get the hints down the guest OS.

Quoted 7 lines
Quoted 3 lines

Regarding the alignment it seems that rbd_dev->opts->alloc_size is something that comes from the device

configuration and not from rbd? I don't have that information inside the Qemu RBD driver.

librbd doesn't really have the information either. The 64KiB guess that krbd uses was a compromise since that was the default OSD allocation size for HDDs since Luminous. Starting with Pacific that default is going down to 4KiB.

Quoted 4 lines

I will try to adjust these values as far as it is possible and makes sense.

Is there a way to check the minimum supported OSD release in the backend from librbd / librados?

It's not a minimum -- RADOS will gladly access 1 byte writes as well. It's really just the optimal (performance and space-wise). Sadly, there is no realistic way to query this data from the backend.

Quoted 7 lines

Anyway, I want to sent a V2 by the end of this week.

Peter

Signature · 1 line
Jason