Support streaming to an intermediate layer

The current way to obtain the list of existing block jobs is to iterate over all root nodes and check which ones own a job.

Since we want to be able to support block jobs in other nodes as well, this patch keeps a list of jobs that is updated everytime one is created or destroyed.

This also updates qmp_query_block_jobs() to use this new list.

Signed-off-by Alberto Garcia <[email protected]>

---
  blockdev.c               | 19 ++++++++-----------
  blockjob.c               | 13 +++++++++++++
  include/block/blockjob.h | 14 ++++++++++++++
  3 files changed, 35 insertions(+), 11 deletions(-)

Currently, block jobs can only be owned by root nodes. This patch allows block jobs to be in any arbitrary node, by making the following changes:

- Block jobs can now be identified by the node name of their BlockDriverState in addition to the device name. Since both device and node names live in the same namespace there's no ambiguity.

- The "device" parameter used by all commands that operate on block jobs can also be a node name now.

- The node name is used as a fallback to fill in the BlockJobInfo structure and all BLOCK_JOB_* events if there is no device name for that job.

Signed-off-by Alberto Garcia <[email protected]>

---
  block/mirror.c            |  5 +++--
  blockdev.c                | 14 +++++++-------
  blockjob.c                | 17 +++++++++--------
  docs/qmp/qmp-events.txt   |  8 ++++----
  include/qapi/qmp/qerror.h |  3 ---
  qapi/block-core.json      | 18 +++++++++---------
  6 files changed, 32 insertions(+), 33 deletions(-)

block/mirror.c +0 -0

index 4056164..189e8f8 100644

block/mirror.c +0 -0

          return;
      }
 

blockdev.c +4 -2

index 567d5e3..dc5d931 100644
@@ -2643,18 +2643,18 @@ out:
      aio_context_release(aio_context);
  }
 
-/* Get the block job for a given device name and acquire its AioContext */
-static BlockJob *find_block_job(const char *device, AioContext **aio_context,
+/* Get the block job for a given device or node name
+ * and acquire its AioContext */
+static BlockJob *find_block_job(const char *device_or_node,
+                                AioContext **aio_context,

blockdev.c +1 -2

      }
-    bs = blk_bs(blk);
 
      *aio_context = bdrv_get_aio_context(bs);
      aio_context_acquire(*aio_context);

@@ -2668,7 +2668,7 @@ static BlockJob *find_block_job(const char *device, AioContext **aio_context,
 
  notfound:
      error_set(errp, ERROR_CLASS_DEVICE_NOT_ACTIVE,
-              "No active block job on device '%s'", device);
+              "No active block job on node '%s'", device_or_node);
      *aio_context = NULL;
      return NULL;
  }

blockjob.c +0 -0

index 562362b..b2b6cc9 100644

blockjob.c +8 -7

          return NULL;
      }
      bdrv_ref(bs);

@@ -121,8 +121,9 @@ void block_job_set_speed(BlockJob *job, int64_t speed, Error **errp)
  void block_job_complete(BlockJob *job, Error **errp)
  {
      if (job->paused || job->cancelled || !job->driver->complete) {
-        error_set(errp, QERR_BLOCK_JOB_NOT_READY,
-                  bdrv_get_device_name(job->bs));
+        error_setg(errp,
+                   "The active block job for node '%s' cannot be completed",
+                   bdrv_get_device_or_node_name(job->bs));
          return;
      }
 

@@ -268,7 +269,7 @@ BlockJobInfo *block_job_query(BlockJob *job)
  {
      BlockJobInfo *info = g_new0(BlockJobInfo, 1);
      info->type      = g_strdup(BlockJobType_lookup[job->driver->job_type]);
-    info->device    = g_strdup(bdrv_get_device_name(job->bs));
+    info->device    = g_strdup(bdrv_get_device_or_node_name(job->bs));
      info->len       = job->len;
      info->busy      = job->busy;
      info->paused    = job->paused;

@@ -290,7 +291,7 @@ static void block_job_iostatus_set_err(BlockJob *job, int error)
  void block_job_event_cancelled(BlockJob *job)
  {
      qapi_event_send_block_job_cancelled(job->driver->job_type,
-                                        bdrv_get_device_name(job->bs),
+                                        bdrv_get_device_or_node_name(job->bs),
                                          job->len,
                                          job->offset,
                                          job->speed,

@@ -300,7 +301,7 @@ void block_job_event_cancelled(BlockJob *job)
  void block_job_event_completed(BlockJob *job, const char *msg)
  {
      qapi_event_send_block_job_completed(job->driver->job_type,
-                                        bdrv_get_device_name(job->bs),
+                                        bdrv_get_device_or_node_name(job->bs),
                                          job->len,
                                          job->offset,
                                          job->speed,

@@ -314,7 +315,7 @@ void block_job_event_ready(BlockJob *job)
      job->ready = true;
 
      qapi_event_send_block_job_ready(job->driver->job_type,
-                                    bdrv_get_device_name(job->bs),
+                                    bdrv_get_device_or_node_name(job->bs),
                                      job->len,
                                      job->offset,
                                      job->speed, &error_abort);

@@ -343,7 +344,7 @@ BlockErrorAction block_job_error_action(BlockJob *job, BlockDriverState *bs,
      default:
          abort();
      }
-    qapi_event_send_block_job_error(bdrv_get_device_name(job->bs),
+    qapi_event_send_block_job_error(bdrv_get_device_or_node_name(job->bs),
                                      is_read ? IO_OPERATION_TYPE_READ :
                                      IO_OPERATION_TYPE_WRITE,
                                      action, &error_abort);

docs/qmp/qmp-events.txt +4 -4

index b19e490..db90c61 100644
@@ -89,7 +89,7 @@ Data:
 
  - "type":     Job type (json-string; "stream" for image streaming
                                       "commit" for block commit)
-- "device":   Device name (json-string)
+- "device":   Device name, or node name if not present (json-string)
  - "len":      Maximum progress value (json-int)
  - "offset":   Current progress value (json-int)
                On success this is equal to len.

@@ -113,7 +113,7 @@ Data:
 
  - "type":     Job type (json-string; "stream" for image streaming
                                       "commit" for block commit)
-- "device":   Device name (json-string)
+- "device":   Device name, or node name if not present (json-string)
  - "len":      Maximum progress value (json-int)
  - "offset":   Current progress value (json-int)
                On success this is equal to len.

@@ -140,7 +140,7 @@ Emitted when a block job encounters an error.
 
  Data:
 
-- "device": device name (json-string)
+- "device": device name, or node name if not present (json-string)
  - "operation": I/O operation (json-string, "read" or "write")
  - "action": action that has been taken, it's one of the following (json-string):
      "ignore": error has been ignored, the job may fail later

@@ -164,7 +164,7 @@ Data:
 
  - "type":     Job type (json-string; "stream" for image streaming
                                       "commit" for block commit)
-- "device":   Device name (json-string)
+- "device":   Device name, or node name if not present (json-string)
  - "len":      Maximum progress value (json-int)
  - "offset":   Current progress value (json-int)
                On success this is equal to len.

include/qapi/qmp/qerror.h +0 -3

index e567339..4ba442e 100644
@@ -37,9 +37,6 @@ void qerror_report_err(Error *err);
  #define QERR_BASE_NOT_FOUND \
      ERROR_CLASS_GENERIC_ERROR, "Base '%s' not found"
 
-#define QERR_BLOCK_JOB_NOT_READY \
-    ERROR_CLASS_GENERIC_ERROR, "The active block job for device '%s' cannot be completed"
-
  #define QERR_BUS_NO_HOTPLUG \
      ERROR_CLASS_GENERIC_ERROR, "Bus '%s' does not support hotplugging"
 

qapi/block-core.json +8 -8

index 21e6cb5..3639454 100644
@@ -1065,7 +1065,7 @@
  #
  # Throttling can be disabled by setting the speed to 0.
  #
-# @device: the device name
+# @device: the device or node name of the owner of the block job.
  #
  # @speed:  the maximum speed, in bytes per second, or 0 for unlimited.
  #          Defaults to 0.

@@ -1096,7 +1096,7 @@
  # operation can be started at a later time to finish copying all data from the
  # backing file.
  #
-# @device: the device name
+# @device: the device or node name of the owner of the block job.
  #
  # @force: #optional whether to allow cancellation of a paused job (default
  #         false).  Since 1.3.

@@ -1122,7 +1122,7 @@
  # the operation is actually paused.  Cancelling a paused job automatically
  # resumes it.
  #
-# @device: the device name
+# @device: the device or node name of the owner of the block job.
  #
  # Returns: Nothing on success
  #          If no background operation is active on this device, DeviceNotActive

@@ -1142,7 +1142,7 @@
  #
  # This command also clears the error status of the job.
  #
-# @device: the device name
+# @device: the device or node name of the owner of the block job.
  #
  # Returns: Nothing on success
  #          If no background operation is active on this device, DeviceNotActive

@@ -1168,7 +1168,7 @@
  #
  # A cancelled or paused job cannot be completed.
  #
-# @device: the device name
+# @device: the device or node name of the owner of the block job.
  #
  # Returns: Nothing on success
  #          If no background operation is active on this device, DeviceNotActive

@@ -1821,7 +1821,7 @@
  #
  # @type: job type
  #
-# @device: device name
+# @device: device name, or node name if not present
  #
  # @len: maximum progress value
  #

@@ -1852,7 +1852,7 @@
  #
  # @type: job type
  #
-# @device: device name
+# @device: device name, or node name if not present
  #
  # @len: maximum progress value
  #

@@ -1875,7 +1875,7 @@
  #
  # Emitted when a block job encounters an error
  #
-# @device: device name
+# @device: device name, or node name if not present
  #
  # @operation: I/O operation
  #

We need to call stream_complete() in order to do all the necessary clean-ups, even if there's an early failure. At the moment it's only useful to make sure that s->backing_file_str is not leaked,

but it
will become more important as we introduce support for streaming to any intermediate node.

Signed-off-by Alberto Garcia <[email protected]>

---
  block/stream.c | 11 ++++++-----
  1 file changed, 6 insertions(+), 5 deletions(-)

This makes sure that the image we are steaming into is open in read-write mode during the operation.

Operation blockers are also set in all intermediate nodes, since they will be removed from the chain afterwards.

Finally, this also unblocks the stream operation in backing files.

Signed-off-by Alberto Garcia <[email protected]>

---
  block.c        |  4 +++-
  block/stream.c | 36 ++++++++++++++++++++++++++++++++++++
  2 files changed, 39 insertions(+), 1 deletion(-)

block.c +3 -1

index 25289ef..e892cb4 100644
@@ -1240,9 +1240,11 @@ void bdrv_set_backing_hd(BlockDriverState *bs, BlockDriverState *backing_hd)
              backing_hd->drv ? backing_hd->drv->format_name : "");
 
      bdrv_op_block_all(bs->backing_hd, bs->backing_blocker);
-    /* Otherwise we won't be able to commit due to check in bdrv_commit */
+    /* Otherwise we won't be able to commit or stream */
      bdrv_op_unblock(bs->backing_hd, BLOCK_OP_TYPE_COMMIT_TARGET,
                      bs->backing_blocker);
+    bdrv_op_unblock(bs->backing_hd, BLOCK_OP_TYPE_STREAM,
+                    bs->backing_blocker);
  out:
      bdrv_refresh_limits(bs, NULL);
  }

block/stream.c +2 -0

index 37bfd8b..327d964 100644
@@ -33,6 +33,8 @@ typedef struct StreamBlockJob {
      BlockDriverState *base;
      BlockdevOnError on_error;
      char *backing_file_str;
+    int bs_flags;
+    Error *blocker;
  } StreamBlockJob;
 
  static int coroutine_fn stream_populate(BlockDriverState *bs,

block/stream.c +11 -0

      BlockDriverState *base = s->base;
 
+    /* Remove all blockers set in stream_start() */
+    for (i = job->bs->backing_hd; i && i != s->base; i = i->backing_hd) {
+        bdrv_op_unblock_all(i, s->blocker);
+    }
+    error_free(s->blocker);
+
      if (!block_job_is_cancelled(&s->common) && data->reached_end &&
          data->ret == 0) {
          const char *base_id = NULL, *base_fmt = NULL;

@@ -103,6 +112,11 @@ static void stream_complete(BlockJob *job, void *opaque)
          close_unused_images(job->bs, base, base_id);
      }
 
+    /* Reopen the image back in read-only mode if necessary */
+    if (s->bs_flags != bdrv_get_flags(job->bs)) {
+        bdrv_reopen(job->bs, s->bs_flags, NULL);
+    }
+
      g_free(s->backing_file_str);
      block_job_completed(&s->common, data->ret);
      g_free(data);

This patch makes the 'device' paramater of the 'block-stream' command allow a node name as well as a device name.

In addition to that, operation blockers will be checked in all intermediate nodes between the top and the base node.

Signed-off-by Alberto Garcia <[email protected]>

---
  blockdev.c           | 20 ++++++++++----------
  qapi/block-core.json | 10 +++++++---
  2 files changed, 17 insertions(+), 13 deletions(-)

blockdev.c +0 -0

index dc5d931..f24cf2d 100644

blockdev.c +9 -8

      BlockDriverState *base_bs = NULL;
      AioContext *aio_context;
      Error *local_err = NULL;

@@ -2115,20 +2114,14 @@ void qmp_block_stream(const char *device,
          on_error = BLOCKDEV_ON_ERROR_REPORT;
      }
 
-    blk = blk_by_name(device);
-    if (!blk) {
-        error_set(errp, QERR_DEVICE_NOT_FOUND, device);
+    bs = bdrv_lookup_bs(device, device, errp);
+    if (!bs) {
          return;
      }
-    bs = blk_bs(blk);
 
      aio_context = bdrv_get_aio_context(bs);
      aio_context_acquire(aio_context);
 
-    if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_STREAM, errp)) {
-        goto out;
-    }
-
      if (has_base) {
          base_bs = bdrv_find_backing_image(bs, base);
          if (base_bs == NULL) {

@@ -2139,6 +2132,13 @@ void qmp_block_stream(const char *device,
          base_name = base;
      }
 
+    /* Check for op blockers in the whole chain between bs and base */
+    for (i = bs; i != NULL && i != base_bs; i = i->backing_hd) {
+        if (bdrv_op_is_blocked(i, BLOCK_OP_TYPE_STREAM, errp)) {
+            goto out;
+        }
+    }
+
      /* if we are streaming the entire chain, the result will have no backing
       * file, and specifying one is therefore an error */
      if (base_bs == NULL && has_backing_file) {

qapi/block-core.json +7 -3

index 3639454..60b9664 100644
@@ -1014,6 +1014,10 @@
  # with query-block-jobs.  The operation can be stopped before it has completed
  # using the block-job-cancel command.
  #
+# The node that receives the data is called the top image, can be located
+# in any part of the whole chain and can be specified using its device
+# or node name.
+#
  # If a base file is specified then sectors are not copied from that base file and
  # its backing chain.  When streaming completes the image file will have the base
  # file as its backing file.  This can be used to stream a subset of the backing

@@ -1022,12 +1026,12 @@
  # On successful completion the image file is updated to drop the backing file
  # and the BLOCK_JOB_COMPLETED event is emitted.
  #
-# @device: the device name
+# @device: the device or node name of the top image
  #
  # @base:   #optional the common backing file name
  #
-# @backing-file: #optional The backing file string to write into the active
-#                          layer. This filename is not validated.
+# @backing-file: #optional The backing file string to write into the top
+#                          image. This filename is not validated.
  #
  #                          If a pathname string is such that it cannot be
  #                          resolved by QEMU, that means that subsequent QMP or

Signed-off-by Alberto Garcia <[email protected]>

---
  docs/live-block-ops.txt | 32 ++++++++++++++++++++------------
  1 file changed, 20 insertions(+), 12 deletions(-)

docs/live-block-ops.txt +2 -2

index a257087..5e969fd 100644
@@ -10,9 +10,9 @@ Snapshot live merge
  Given a snapshot chain, described in this document in the following
  format:
 
-[A] -> [B] -> [C] -> [D]
+[A] -> [B] -> [C] -> [D] -> [E]
 
-Where the rightmost object ([D] in the example) described is the current
+Where the rightmost object ([E] in the example) described is the current
  image which the guest OS has write access to. To the left of it is its base
  image, and so on accordingly until the leftmost image, which has no
  base.

@@ -21,28 +21,36 @@ The snapshot live merge operation transforms such a chain into a
  smaller one with fewer elements, such as this transformation relative
  to the first example:

docs/live-block-ops.txt +0 -0

  Live block copy
  ===============

There's only one change in this series: there's now a list of block jobs that is updated everytime a new one is created or destroyed.

This way we can iterate the list directly, rather than searching all BlockDriverStates for the ones that own a block job.

Otherwise please check the summary I wrote for the previous series, because the rest of the changes are detailed there:

    https://lists.gnu.org/archive/html/qemu-devel/2015-03/msg04798.html

v3:
- Keep a list of block jobs and make qmp_query_block_jobs() iterate over it.

v2: https://lists.gnu.org/archive/html/qemu-devel/2015-03/msg04798.html
- The 'block-stream' command does not have a 'node-name' parameter anymore and reuses 'device' for that purpose.
- Block jobs can now be owned by any intermediate node, and not just by the ones at the root. query-block-jobs is updated to reflect that change.
- The 'device' parameter of all 'block-job-*' commands can now take a node name.
- The BlockJobInfo type and all BLOCK_JOB_* events report the node name in the 'device' field if the node does not have a device name.
- All intermediate nodes are blocked (and checked for blockers) during the streaming operation.

Regards,

Berto

Alberto Garcia (6):
block: keep a list of block jobs
block: allow block jobs in any arbitrary node
block: never cancel a streaming job without running stream_complete()
block: Support streaming to an intermediate layer
block: Add QMP support for streaming to an intermediate layer
docs: Document how to stream to an intermediate layer

  block.c                   |  4 +++-
  block/mirror.c            |  5 +++--
  block/stream.c            | 47 ++++++++++++++++++++++++++++++++++++-----
  blockdev.c                | 53 ++++++++++++++++++++++-------------------------
  blockjob.c                | 30 ++++++++++++++++++++-------
  docs/live-block-ops.txt   | 32 +++++++++++++++++-----------
  docs/qmp/qmp-events.txt   |  8 +++----
  include/block/blockjob.h  | 14 +++++++++++++
  include/qapi/qmp/qerror.h |  3 ---
  qapi/block-core.json      | 28 ++++++++++++++-----------
  10 files changed, 149 insertions(+), 75 deletions(-)

Quoted 10 lines

block/mirror.c +0 -2

index 4056164..189e8f8 100644
@@ -607,8 +607,9 @@ static void mirror_complete(BlockJob *job, Error **errp)
          return;
      }
      if (!s->synced) {
-        error_set(errp, QERR_BLOCK_JOB_NOT_READY,
-                  bdrv_get_device_name(job->bs));

block/mirror.c static void mirror_complete(BlockJob *job, Error **errp) · line 612

Quoted 3 lines

block/mirror.c +3 -0

+        error_setg(errp,
+                   "The active block job for device '%s' cannot be completed",
+                   bdrv_get_device_name(job->bs));

I know there is no way right now that bdrv_get_device_name() returns an empty string, but somehow I'd feel more consistent for this to be bdrv_get_device_or_node_name() and the string saying "node" instead of "device". Your choice, though, since it's completely correct for now.

Quoted 9 lines

block/mirror.c +4 -1

+/* Get the block job for a given device or node name
+ * and acquire its AioContext */
+static BlockJob *find_block_job(const char *device_or_node,
+                                AioContext **aio_context,
                                  Error **errp)
  {
-    BlockBackend *blk;
      BlockDriverState *bs;

block/mirror.c

Quoted 5 lines

block/mirror.c +2 -2

-    blk = blk_by_name(device);
-    if (!blk) {
+    bs = bdrv_lookup_bs(device_or_node, device_or_node, NULL);
+    if (!bs) {
          goto notfound;

It would be possible to pass errp to bdrv_lookup_bs() and move the error_set() under notfound to the if (!bs->job) block (I'd find the error message confusing if there just is no node with that name; but on the other hand, it wouldn't be a regression).

Quoted 7 lines

blockjob.c +0 -0

index 562362b..b2b6cc9 100644
@@ -52,7 +52,7 @@ void *block_job_create(const BlockJobDriver *driver, BlockDriverState *bs,
      BlockJob *job;
 

blockjob.c void *block_job_create(const BlockJobDriver *driver, BlockDriverState *bs, · line 55

Quoted 3 lines

blockjob.c +1 -1

      if (bs->job) {
-        error_set(errp, QERR_DEVICE_IN_USE, bdrv_get_device_name(bs));
+        error_set(errp, QERR_DEVICE_IN_USE, bdrv_get_device_or_node_name(bs));

Hm, the error message only mentions "device" now... Not too nice.

Quoted 4 lines

blockjob.c +0 -0

@@ -1895,7 +1895,7 @@
  #
  # @type: job type
  #

blockjob.c line 1901

Quoted 5 lines

blockjob.c +1 -1

-# @device: device name
+# @device: device name, or node name if not present
  #
  # @len: maximum progress value
  #

I think you need to apply the same treatment for the comment of BlockJobInfo, too.

block/stream.c void stream_start(BlockDriverState *bs, BlockDriverState *base, · line 276

Quoted 2 lines

block/stream.c +2 -0

+    orig_bs_flags = bdrv_get_flags(bs);
+    if (!(orig_bs_flags & BDRV_O_RDWR)) {

I feel like we don't want to do this if we're not streaming to an intermediate layer but to the top layer (because that means there is some reason for the BDS to be read-only beyond it just being a backing BDS).

block/stream.c void stream_start(BlockDriverState *bs, BlockDriverState *base, · line 291

Quoted 2 lines

block/stream.c +2 -0

+    /* Block all intermediate nodes between bs and base, because they
+     * will disappear from the chain after this operation */

Hm, do we really need to? There's a difference between "it doesn't make sense, but it works if you want to" and "it will break". Shouldn't it be enough that the intermediate nodes are all read-only anyway (hopefully)?

But then again, it probably won't hurt and I don't really want to think about the implications of trying to run a block-commit or a separate block-stream on the chain...

Quoted 3 lines · Max Reitz

Quoted 2 lines

block/stream.c +2 -0

+    orig_bs_flags = bdrv_get_flags(bs);
+    if (!(orig_bs_flags & BDRV_O_RDWR)) {

Quoted 4 lines

I feel like we don't want to do this if we're not streaming to an intermediate layer but to the top layer (because that means there is some reason for the BDS to be read-only beyond it just being a backing BDS).

I didn't think about this... that I can fix easily, but I wonder what's the scenario where the top layer is read-only.

Quoted 4 lines

But then again, it probably won't hurt and I don't really want to think about the implications of trying to run a block-commit or a separate block-stream on the chain...

I actually tried with scenarios such as A>B>C>D>E, streaming from B to E and from A to C simultaneously, and it seems to work (at least I didn't see any obvious problems), but I don't think we want to support that because a) I don't see the use case and b) we're likely opening a can of worms.

What we can do is stream from A to C and from D to E at the same time, that would be allowed with these patches and I don't see any obvious reason why it would fail.

Quoted 8 lines

Quoted 7 lines

block/stream.c void stream_start(BlockDriverState *bs, BlockDriverState *base, · line 276

Quoted 2 lines

block/stream.c +2 -0

+    orig_bs_flags = bdrv_get_flags(bs);
+    if (!(orig_bs_flags & BDRV_O_RDWR)) {

I feel like we don't want to do this if we're not streaming to an intermediate layer but to the top layer (because that means there is some reason for the BDS to be read-only beyond it just being a backing BDS).

Quoted 2 lines

I didn't think about this... that I can fix easily, but I wonder what's the scenario where the top layer is read-only.

Suppose I tell qemu to open the chain 'base <- mid <- top' read-only, and then later decide to stream base into mid or top. It should be fine from the guest's perspective to keep the guest in read-only mode, while still using streaming to reduce the chain.

Quoted 5 lines

I actually tried with scenarios such as A>B>C>D>E, streaming from B to E and from A to C simultaneously, and it seems to work (at least I didn't see any obvious problems), but I don't think we want to support that because a) I don't see the use case and b) we're likely opening a can of worms.

Streaming doesn't corrupt any part of the chain, so I think you are safe doing overlapping streams like that. However, I also agree that it is not worth supporting, and that we are safer not allowing overlapping streams.

block/stream.c void stream_start(BlockDriverState *bs, BlockDriverState *base, · line 276

Quoted 2 lines

block/stream.c +2 -0

+    orig_bs_flags = bdrv_get_flags(bs);
+    if (!(orig_bs_flags & BDRV_O_RDWR)) {

I feel like we don't want to do this if we're not streaming to an intermediate layer but to the top layer (because that means there is some reason for the BDS to be read-only beyond it just being a backing BDS).

Quoted 9 lines

Quoted 9 lines

Quoted 8 lines

block/stream.c void stream_start(BlockDriverState *bs, BlockDriverState *base, · line 276

Quoted 3 lines

block/stream.c +2 -0

+    orig_bs_flags = bdrv_get_flags(bs);
+    if (!(orig_bs_flags & BDRV_O_RDWR)) {

I feel like we don't want to do this if we're not streaming to an intermediate layer but to the top layer (because that means there is some reason for the BDS to be read-only beyond it just being a backing BDS).

Quoted 7 lines

Quoted 2 lines

I didn't think about this... that I can fix easily, but I wonder what's the scenario where the top layer is read-only.

Suppose I tell qemu to open the chain 'base <- mid <- top' read-only, and then later decide to stream base into mid or top. It should be fine from the guest's perspective to keep the guest in read-only mode, while still using streaming to reduce the chain.

Ok, I'll add the check for the top image then.

Quoted 10 lines

Quoted 5 lines

I actually tried with scenarios such as A>B>C>D>E, streaming from B to E and from A to C simultaneously, and it seems to work (at least I didn't see any obvious problems), but I don't think we want to support that because a) I don't see the use case and b) we're likely opening a can of worms.

Streaming doesn't corrupt any part of the chain, so I think you are safe doing overlapping streams like that. However, I also agree that it is not worth supporting, and that we are safer not allowing overlapping streams.

I'm thinking about cases where streaming from B to E finishes (and all the intermediate nodes are removed from the chain) while A to C is still ongoing. At the very least it doesn't seem to make sense to allow that kind of things.

Quoted 6 lines · Max Reitz

block/stream.c void stream_start(BlockDriverState *bs, BlockDriverState *base, · line 276

Quoted 2 lines

block/stream.c +2 -0

+    orig_bs_flags = bdrv_get_flags(bs);
+    if (!(orig_bs_flags & BDRV_O_RDWR)) {

I feel like we don't want to do this if we're not streaming to an intermediate layer but to the top layer (because that means there is some reason for the BDS to be read-only beyond it just being a backing BDS).

Looks like we don't actually need to do anything, bdrv_reopen_prepare() already takes care of checking the BDRV_O_ALLOW_RDWR flag, which is not set if the image was opened in read-only mode.