A race condition is currently present between the clean up attempt of
the QEMU process and the execution of qemu-img. The actual (bad)
output is:
-Warning: Image size mismatch!
-Images are identical.
+qemu-img: Could not open '<build_dir>/tests/qemu-iotests/scratch/t.raw': Failed to get "consistent read" lock
+Is another process using the image?
A KILL signal is sent to the QEMU process, but qemu-img may begin to
run before the QEMU process is really gone. qemu-img will then
attempt to open the TEST_IMG file before it can secure a lock on it.
This attempts a more graceful shutdown, and waits for the QEMU process
to exit.
Signed-off-by: Cleber Rosa <crosa@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Jeff Cody <jcody@redhat.com>
Reviewed-by: John Snow <jsnow@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Mirror calculates job len from current I/O progress:
s->common.len = s->common.offset +
(cnt + s->sectors_in_flight) * BDRV_SECTOR_SIZE;
The final "len" of a failed mirror job in iotests 109 depends on the
subtle timing of the completion of read and write issued in the first
mirror iteration. The second iteration may or may not have run when the
I/O error happens, resulting in non-deterministic output of the
BLOCK_JOB_COMPLETED event text.
Similar to what was done in a752e4786, filter out the field to make the
test robust.
Signed-off-by: Fam Zheng <famz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Tested-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
BDRV_POLL_WHILE waits for the started I/O by releasing bs's ctx then polling
the main context, which relies on the yielded coroutine continuing on bs->ctx
before notifying qemu_aio_context with bdrv_wakeup().
Thus, using qemu_coroutine_enter to start I/O is wrong because if the coroutine
is entered from main loop, co->ctx will be qemu_aio_context, as a result of the
"release, poll, acquire" loop of BDRV_POLL_WHILE, race conditions happen when
both main thread and the iothread access the same BDS:
main loop iothread
-----------------------------------------------------------------------
blockdev_snapshot
aio_context_acquire(bs->ctx)
virtio_scsi_data_plane_handle_cmd
bdrv_drained_begin(bs->ctx)
bdrv_flush(bs)
bdrv_co_flush(bs) aio_context_acquire(bs->ctx).enter
...
qemu_coroutine_yield(co)
BDRV_POLL_WHILE()
aio_context_release(bs->ctx)
aio_context_acquire(bs->ctx).return
...
aio_co_wake(co)
aio_poll(qemu_aio_context) ...
co_schedule_bh_cb() ...
qemu_coroutine_enter(co) ...
/* (A) bdrv_co_flush(bs) /* (B) I/O on bs */
continues... */
aio_context_release(bs->ctx)
aio_context_acquire(bs->ctx)
Note that in above case, bdrv_drained_begin() doesn't do the "release,
poll, acquire" in BDRV_POLL_WHILE, because bs->in_flight == 0.
Fix this by using bdrv_coroutine_enter and enter coroutine in the right
context.
iotests 109 output is updated because the coroutine reenter flow during
mirror job complete is different (now through co_queue_wakeup, instead
of the unconditional qemu_coroutine_switch before), making the end job
len different.
Signed-off-by: Fam Zheng <famz@redhat.com>
Acked-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
109 iotest is broken for raw after 0965a41e99
[mirror: double performance of the bulk stage if the disc is full]
The problem is with finishing block-job with error: before specified
patch mirror was not very async and it created one big request at disk
start, this request finished with error and qemu produced
BLOCK_JOB_COMPLETED with zero progress.
After 0965a41, mirror starts several smaller requests in parallel, when
BLOCK_JOB_COMPLETED emited we have some successful non-zero progress.
This patch solves the issue by filtering out progress from 109 test
output.
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
For some time now, qemu-img compare has been able to compare
unaligned images. So we no longer need test 109's hack of
resizing to sector boundaries before invoking compare.
Signed-off-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-id: 1463416983-28318-3-git-send-email-eblake@redhat.com
Signed-off-by: Max Reitz <mreitz@redhat.com>
The last sub-chunk is rounded up to the copy granularity in the target
image, resulting in a larger size than the source.
Add a function to clip the copied sectors to the end.
This undoes the "wrong" changes to tests/qemu-iotests/109.out in
e5b43573e2. The remaining two offset changes are okay.
[ kwolf: Use DIV_ROUND_UP to calculate nb_chunks now ]
Reported-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Fam Zheng <famz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Jeff Cody <jcody@redhat.com>
The "pnum < nb_sectors" condition in deciding whether to actually copy
data is unnecessarily strict, and the qiov initialization is
unnecessarily for bdrv_aio_write_zeroes and bdrv_aio_discard.
Rewrite mirror_iteration to fix both flaws.
The output of iotests 109 is updated because we now report the offset
and len slightly differently in mirroring progress.
Signed-off-by: Fam Zheng <famz@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-id: 1454637630-10585-2-git-send-email-famz@redhat.com
Signed-off-by: Jeff Cody <jcody@redhat.com>
This is forbidden if the raw driver was probed.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-id: 1416497234-29880-10-git-send-email-kwolf@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
