This patch rewrites the ctx->dispatching optimization, which was the cause
of some mysterious hangs that could be reproduced on aarch64 KVM only.
The hangs were indirectly caused by aio_poll() and in particular by
flash memory updates's call to blk_write(), which invokes aio_poll().
Fun stuff: they had an extremely short race window, so much that
adding all kind of tracing to either the kernel or QEMU made it
go away (a single printf made it half as reproducible).
On the plus side, the failure mode (a hang until the next keypress)
made it very easy to examine the state of the process with a debugger.
And there was a very nice reproducer from Laszlo, which failed pretty
often (more than half of the time) on any version of QEMU with a non-debug
kernel; it also failed fast, while still in the firmware. So, it could
have been worse.
For some unknown reason they happened only with virtio-scsi, but
that's not important. It's more interesting that they disappeared with
io=native, making thread-pool.c a likely suspect for where the bug arose.
thread-pool.c is also one of the few places which use bottom halves
across threads, by the way.
I hope that no other similar bugs exist, but just in case :) I am
going to describe how the successful debugging went... Since the
likely culprit was the ctx->dispatching optimization, which mostly
affects bottom halves, the first observation was that there are two
qemu_bh_schedule() invocations in the thread pool: the one in the aio
worker and the one in thread_pool_completion_bh. The latter always
causes the optimization to trigger, the former may or may not. In
order to restrict the possibilities, I introduced new functions
qemu_bh_schedule_slow() and qemu_bh_schedule_fast():
/* qemu_bh_schedule_slow: */
ctx = bh->ctx;
bh->idle = 0;
if (atomic_xchg(&bh->scheduled, 1) == 0) {
event_notifier_set(&ctx->notifier);
}
/* qemu_bh_schedule_fast: */
ctx = bh->ctx;
bh->idle = 0;
assert(ctx->dispatching);
atomic_xchg(&bh->scheduled, 1);
Notice how the atomic_xchg is still in qemu_bh_schedule_slow(). This
was already debated a few months ago, so I assumed it to be correct.
In retrospect this was a very good idea, as you'll see later.
Changing thread_pool_completion_bh() to qemu_bh_schedule_fast() didn't
trigger the assertion (as expected). Changing the worker's invocation
to qemu_bh_schedule_slow() didn't hide the bug (another assumption
which luckily held). This already limited heavily the amount of
interaction between the threads, hinting that the problematic events
must have triggered around thread_pool_completion_bh().
As mentioned early, invoking a debugger to examine the state of a
hung process was pretty easy; the iothread was always waiting on a
poll(..., -1) system call. Infinite timeouts are much rarer on x86,
and this could be the reason why the bug was never observed there.
With the buggy sequence more or less resolved to an interaction between
thread_pool_completion_bh() and poll(..., -1), my "tracing" strategy was
to just add a few qemu_clock_get_ns(QEMU_CLOCK_REALTIME) calls, hoping
that the ordering of aio_ctx_prepare(), aio_ctx_dispatch, poll() and
qemu_bh_schedule_fast() would provide some hint. The output was:
(gdb) p last_prepare
$3 = 103885451
(gdb) p last_dispatch
$4 = 103876492
(gdb) p last_poll
$5 = 115909333
(gdb) p last_schedule
$6 = 115925212
Notice how the last call to qemu_poll_ns() came after aio_ctx_dispatch().
This makes little sense unless there is an aio_poll() call involved,
and indeed with a slightly different instrumentation you can see that
there is one:
(gdb) p last_prepare
$3 = 107569679
(gdb) p last_dispatch
$4 = 107561600
(gdb) p last_aio_poll
$5 = 110671400
(gdb) p last_schedule
$6 = 110698917
So the scenario becomes clearer:
iothread VCPU thread
--------------------------------------------------------------------------
aio_ctx_prepare
aio_ctx_check
qemu_poll_ns(timeout=-1)
aio_poll
aio_dispatch
thread_pool_completion_bh
qemu_bh_schedule()
At this point bh->scheduled = 1 and the iothread has not been woken up.
The solution must be close, but this alone should not be a problem,
because the bottom half is only rescheduled to account for rare situations
(see commit 3c80ca1, thread-pool: avoid deadlock in nested aio_poll()
calls, 2014-07-15).
Introducing a third thread---a thread pool worker thread, which
also does qemu_bh_schedule()---does bring out the problematic case.
The third thread must be awakened *after* the callback is complete and
thread_pool_completion_bh has redone the whole loop, explaining the
short race window. And then this is what happens:
thread pool worker
--------------------------------------------------------------------------
<I/O completes>
qemu_bh_schedule()
Tada, bh->scheduled is already 1, so qemu_bh_schedule() does nothing
and the iothread is never woken up. This is where the bh->scheduled
optimization comes into play---it is correct, but removing it would
have masked the bug.
So, what is the bug?
Well, the question asked by the ctx->dispatching optimization ("is any
active aio_poll dispatching?") was wrong. The right question to ask
instead is "is any active aio_poll *not* dispatching", i.e. in the prepare
or poll phases? In that case, the aio_poll is sleeping or might go to
sleep anytime soon, and the EventNotifier must be invoked to wake
it up.
In any other case (including if there is *no* active aio_poll at all!)
we can just wait for the next prepare phase to pick up the event (e.g. a
bottom half); the prepare phase will avoid the blocking and service the
bottom half.
Expressing the invariant with a logic formula, the broken one looked like:
!(exists(thread): in_dispatching(thread)) => !optimize
or equivalently:
!(exists(thread):
in_aio_poll(thread) && in_dispatching(thread)) => !optimize
In the correct one, the negation is in a slightly different place:
(exists(thread):
in_aio_poll(thread) && !in_dispatching(thread)) => !optimize
or equivalently:
(exists(thread): in_prepare_or_poll(thread)) => !optimize
Even if the difference boils down to moving an exclamation mark :)
the implementation is quite different. However, I think the new
one is simpler to understand.
In the old implementation, the "exists" was implemented with a boolean
value. This didn't really support well the case of multiple concurrent
event loops, but I thought that this was okay: aio_poll holds the
AioContext lock so there cannot be concurrent aio_poll invocations, and
I was just considering nested event loops. However, aio_poll _could_
indeed be concurrent with the GSource. This is why I came up with the
wrong invariant.
In the new implementation, "exists" is computed simply by counting how many
threads are in the prepare or poll phases. There are some interesting
points to consider, but the gist of the idea remains:
1) AioContext can be used through GSource as well; as mentioned in the
patch, bit 0 of the counter is reserved for the GSource.
2) the counter need not be updated for a non-blocking aio_poll, because
it won't sleep forever anyway. This is just a matter of checking
the "blocking" variable. This requires some changes to the win32
implementation, but is otherwise not too complicated.
3) as mentioned above, the new implementation will not call aio_notify
when there is *no* active aio_poll at all. The tests have to be
adjusted for this change. The calls to aio_notify in async.c are fine;
they only want to kick aio_poll out of a blocking wait, but need not
do anything if aio_poll is not running.
4) nested aio_poll: these just work with the new implementation; when
a nested event loop is invoked, the outer event loop is never in the
prepare or poll phases. The outer event loop thus has already decremented
the counter.
Reported-by: Richard W. M. Jones <rjones@redhat.com>
Reported-by: Laszlo Ersek <lersek@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Fam Zheng <famz@redhat.com>
Tested-by: Richard W.M. Jones <rjones@redhat.com>
Message-id: 1437487673-23740-5-git-send-email-pbonzini@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
This patch moves bdrv_attach_child() from the individual places that add
a backing file to a BDS to bdrv_set_backing_hd(), which is called by all
of them. It also adds bdrv_detach_child() there.
For normal operation (starting with one backing file chain and not
changing it until the topmost image is closed) and live snapshots, this
constitutes no change in behaviour.
For all other cases, this is a fix for the bug that the old backing file
was still referenced as a child, and the new one wasn't referenced.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
This is the counterpart for bdrv_open_child(). It decreases the
reference count of the child BDS and removes it from the list of
children of the given parent BDS.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
It is the same as bdrv_open_image(), except that it doesn't only return
success or failure, but the newly created BdrvChild object for the new
child node.
As the BdrvChild object already contains a BlockDriverState pointer (and
this is supposed to become the only pointer so that bdrv_append() and
friends can just change a single pointer in BdrvChild), the pbs
parameter is removed for bdrv_open_child().
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
There is job resource leak in function mirror_start_job,
although bdrv_create_dirty_bitmap is unlikely failed.
Add block_job_release for each release when needed.
Signed-off-by: Ting Wang <kathy.wangting@huawei.com>
Reviewed-by: John Snow <jsnow@redhat.com>
Message-id: 1435311455-56048-1-git-send-email-kathy.wangting@huawei.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Using this function would always be wrong because a dirty bitmap must
have a specific owner that consumes the dirty bits and calls
bdrv_reset_dirty_bitmap().
Remove the unused function to avoid future misuse.
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Fam Zheng <famz@redhat.com>
Reviewed-by: John Snow <jsnow@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
If specified as "true", it allows discarding on target sectors where source is
not allocated.
Signed-off-by: Fam Zheng <famz@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Like bdrv_is_allocated_above, this function follows the backing chain until seeing
BDRV_BLOCK_ALLOCATED. Base is not included.
Reimplement bdrv_is_allocated on top.
[Initialized bdrv_co_get_block_status_above() ret to 0 to silence
mingw64 compiler warning about the unitialized variable. assert(bs !=
base) prevents that case but I suppose the program could be compiled
with -DNDEBUG.
--Stefan]
Signed-off-by: Fam Zheng <famz@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
If we wish to make differential backups a feature that's easy to access,
it might be pertinent to rename the "dirty-bitmap" mode to "incremental"
to make it clear what /type/ of backup the dirty-bitmap is helping us
perform.
This is an API breaking change, but 2.4 has not yet gone live,
so we have this flexibility.
Signed-off-by: John Snow <jsnow@redhat.com>
Message-id: 1433463642-21840-2-git-send-email-jsnow@redhat.com
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
In particular, don't include it into headers.
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Luiz Capitulino <lcapitulino@redhat.com>
In particular, don't include it into headers.
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Luiz Capitulino <lcapitulino@redhat.com>
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.22 (GNU/Linux)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=QgYk
-----END PGP SIGNATURE-----
Merge remote-tracking branch 'remotes/kevin/tags/for-upstream' into staging
Block layer core and image format patches
# gpg: Signature made Fri Jun 12 16:08:53 2015 BST using RSA key ID C88F2FD6
# gpg: Good signature from "Kevin Wolf <kwolf@redhat.com>"
* remotes/kevin/tags/for-upstream: (25 commits)
block: Fix reopen flag inheritance
block: Add BlockDriverState.inherits_from
block: Add list of children to BlockDriverState
queue.h: Add QLIST_FIX_HEAD_PTR()
block: Drain requests before swapping nodes in bdrv_swap()
block: Move flag inheritance to bdrv_open_inherit()
block: Use QemuOpts in bdrv_open_common()
block: Use macro for cache option names
vmdk: Use bdrv_open_image()
quorum: Use bdrv_open_image()
check-qdict: Test cases for new functions
qdict: Add qdict_{set,copy}_default()
qdict: Add qdict_array_entries()
iotests: Add tests for overriding BDRV_O_PROTOCOL
block: driver should override flags in bdrv_open()
block: Change bitmap truncate conditional to assertion
block: record new size in bdrv_dirty_bitmap_truncate
raw-posix: Fix .bdrv_co_get_block_status() for unaligned image size
vmdk: Use vmdk_find_index_in_cluster everywhere
vmdk: Fix index_in_cluster calculation in vmdk_co_get_block_status
...
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Version: GnuPG v1
iQEcBAABAgAGBQJVevNrAAoJEJykq7OBq3PIn3wH/iTkoxK+8c6vHY4gAyW/3Vel
JAWyDULliLg/mJiyDyz+9hbyb3FmDsQ7Gve5laFCqSES8IChe/fAoR2dXAapHbJm
VroBii+C4MIakAjB1eSGVTMeJVgyq0SxCbEgo6MSgCNRg8pH/ne4kGMn0elsyMQm
VrxqsoW1qda0gLbybu7uRKmdZlb5Eklk0zQmBlzd25j3xgMXmo7ZcXW0d7dCi/xd
ot01aPNHWKy2FyYsI+N1SMagCedjn0FFCHMMc08W34olUOAschV+Croxi6gupbTl
RdlYNnB6BBYLgT7mxYuKKd20r4Q8Si3E5YjIsYaXIskbCOxDuryKboyheMXWqdA=
=mx5x
-----END PGP SIGNATURE-----
Merge remote-tracking branch 'remotes/stefanha/tags/block-pull-request' into staging
# gpg: Signature made Fri Jun 12 15:57:47 2015 BST using RSA key ID 81AB73C8
# gpg: Good signature from "Stefan Hajnoczi <stefanha@redhat.com>"
# gpg: aka "Stefan Hajnoczi <stefanha@gmail.com>"
* remotes/stefanha/tags/block-pull-request:
qemu-iotests: expand test 093 to support group throttling
throttle: Update throttle infrastructure copyright
throttle: add the name of the ThrottleGroup to BlockDeviceInfo
throttle: acquire the ThrottleGroup lock in bdrv_swap()
throttle: Add throttle group support
throttle: Add throttle group infrastructure tests
throttle: Add throttle group infrastructure
throttle: Extract timers from ThrottleState into a separate structure
raw-posix: Fix .bdrv_co_get_block_status() for unaligned image size
Revert "iothread: release iothread around aio_poll"
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Currently, the block layer assumes that any block node can have only one
parent, and if it has a parent, that it inherits some options/flags from
this parent.
This is not true any more: With references used in block device
creation, a single node can be used by multiple parents, or it can be
created separately and not inherit flags from any parent.
To handle reopens correctly, a node must know from which parent it
inherited options. This patch adds the information to BlockDriverState.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
This allows iterating over all children of a given BDS, not only
including bs->file and bs->backing_hd, but also driver-specific
ones like VMDK extents or Quorum children.
For bdrv_swap(), the list of children of the swapped BDS stays at that
BDS (because that's where the pointers stay as well). The list head
moves and pointers to it must be fixed up therefore.
The list of children in the parent of the swapped BDS is not affected by
the swap. The contents of the BDS objects is swapped, so the existing
pointer in the parent automatically points to the newly swapped in BDS.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Instead of letting every caller of bdrv_open() determine the right flags
for its child node manually and pass them to the function, pass the
parent node and the role of the newly opened child (like backing file,
protocol layer, etc.).
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Jeff Cody <jcody@redhat.com>
Reviewed-by: Alberto Garcia <berto@igalia.com>
bdrv_swap() touches the fields of a BlockDriverState that are
protected by the ThrottleGroup lock. Although those fields end up in
their original place, they are temporarily swapped in the process,
so there's a chance that an operation on a member of the same group
happening on a different thread can try to use them.
Signed-off-by: Alberto Garcia <berto@igalia.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-id: d92dc40d7c4f1fc5cda5cbbf4ffb7a4670b79d17.1433779731.git.berto@igalia.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
The throttle group support use a cooperative round robin scheduling
algorithm.
The principles of the algorithm are simple:
- Each BDS of the group is used as a token in a circular way.
- The active BDS computes if a wait must be done and arms the right
timer.
- If a wait must be done the token timer will be armed so the token
will become the next active BDS.
Signed-off-by: Alberto Garcia <berto@igalia.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-id: f0082a86f3ac01c46170f7eafe2101a92e8fde39.1433779731.git.berto@igalia.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Group throttling will share ThrottleState between multiple bs.
As a consequence the ThrottleState will be accessed by multiple aio
context.
Timers are tied to their aio context so they must go out of the
ThrottleState structure.
This commit paves the way for each bs of a common ThrottleState to
have its own timer.
Signed-off-by: Benoit Canet <benoit.canet@nodalink.com>
Signed-off-by: Alberto Garcia <berto@igalia.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-id: 6cf9ea96d8b32ae2f8769cead38f68a6a0c8c909.1433779731.git.berto@igalia.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
All users are converted to qemu_set_fd_handler now, drop
qemu_set_fd_handler2 and IOHandlerRecord.fd_read_poll.
Signed-off-by: Fam Zheng <famz@redhat.com>
Message-id: 1433400324-7358-9-git-send-email-famz@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Bitmaps can be in a handful of different states with potentially
more to come as we tool around with migration and persistence patches.
Management applications may need to know why certain bitmaps are
unavailable for various commands, e.g. busy in another operation,
busy being migrated, etc.
Right now, all we offer is BlockDirtyInfo's boolean member 'frozen'.
Instead of adding more booleans, replace it by an enumeration member
'status' with values 'active' and 'frozen'. Then add new value
'disabled'.
Incompatible change. Fine because the changed part hasn't been
released so far.
Suggested-by: Eric Blake <eblake@redhat.com>
Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
[Commit message tweaked]
Signed-off-by: Markus Armbruster <armbru@redhat.com>
The patch introduces new concept: minimal memory alignment for bounce
buffers. Original so called "optimal" value is actually minimal required
value for aligment. It should be used for validation that the IOVec
is properly aligned and bounce buffer is not required.
Though, from the performance point of view, it would be better if
bounce buffer or IOVec allocated by QEMU will be aligned stricter.
The patch does not change any alignment value yet.
Signed-off-by: Denis V. Lunev <den@openvz.org>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Message-id: 1431441056-26198-2-git-send-email-den@openvz.org
CC: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
CC: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Move the code to install coroutine and aio emulation function pointers
in a BlockDriver to its own function.
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
The dirty bitmap functions are called from the block I/O processing
code. Make them visible to block_int.h users so they can be used
outside block.c.
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
We often don't need the BlockDriverState for functions
that operate on bitmaps. Remove it.
Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-id: 1429314609-29776-15-git-send-email-jsnow@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Add bdrv_clear_dirty_bitmap and a matching QMP command,
qmp_block_dirty_bitmap_clear that enables a user to reset
the bitmap attached to a drive.
This allows us to reset a bitmap in the event of a full
drive backup.
Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-id: 1429314609-29776-12-git-send-email-jsnow@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
For "dirty-bitmap" sync mode, the block job will iterate through the
given dirty bitmap to decide if a sector needs backup (backup all the
dirty clusters and skip clean ones), just as allocation conditions of
"top" sync mode.
Signed-off-by: Fam Zheng <famz@redhat.com>
Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-id: 1429314609-29776-11-git-send-email-jsnow@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
A bitmap successor is an anonymous BdrvDirtyBitmap that is intended to
be created just prior to a sensitive operation (e.g. Incremental Backup)
that can either succeed or fail, but during the course of which we still
want a bitmap tracking writes.
On creating a successor, we "freeze" the parent bitmap which prevents
its deletion, enabling, anonymization, or creating a bitmap with the
same name.
On success, the parent bitmap can "abdicate" responsibility to the
successor, which will inherit its name. The successor will have been
tracking writes during the course of the backup operation. The parent
will be safely deleted.
On failure, we can "reclaim" the successor from the parent, unifying
them such that the resulting bitmap describes all writes occurring since
the last successful backup, for instance. Reclamation will thaw the
parent, but not explicitly re-enable it.
BdrvDirtyBitmap operations that target a single bitmap are protected
by assertions that the bitmap is not frozen and/or disabled.
BdrvDirtyBitmap operations that target a group of bitmaps, such as
bdrv_{set,reset}_dirty will ignore frozen/disabled drives with a
conditional instead.
Internal functions that enable/disable dirty bitmaps have assertions
added to them to prevent modifying frozen bitmaps.
Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-id: 1429314609-29776-10-git-send-email-jsnow@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Add a status indicating the enabled/disabled state of the bitmap.
A bitmap is by default enabled, but you can lock the bitmap into
a read-only state by setting disabled = true.
A previous version of this patch added a QMP interface for changing
the state of the bitmap, but it has since been removed for now until
a use case emerges where this state must be revealed to the user.
The disabled state WILL be used internally for bitmap migration and
bitmap persistence.
Signed-off-by: Fam Zheng <famz@redhat.com>
Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-id: 1429314609-29776-9-git-send-email-jsnow@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
This returns the granularity (in bytes) of dirty bitmap,
which matches the QMP interface and the existing query
interface.
Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-id: 1429314609-29776-6-git-send-email-jsnow@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
The new command pair is added to manage a user created dirty bitmap. The
dirty bitmap's name is mandatory and must be unique for the same device,
but different devices can have bitmaps with the same names.
The granularity is an optional field. If it is not specified, we will
choose a default granularity based on the cluster size if available,
clamped to between 4K and 64K to mirror how the 'mirror' code was
already choosing granularity. If we do not have cluster size info
available, we choose 64K. This code has been factored out into a helper
shared with block/mirror.
This patch also introduces the 'block_dirty_bitmap_lookup' helper,
which takes a device name and a dirty bitmap name and validates the
lookup, returning NULL and setting errp if there is a problem with
either field. This helper will be re-used in future patches in this
series.
The types added to block-core.json will be re-used in future patches
in this series, see:
'qapi: Add transaction support to block-dirty-bitmap-{add, enable, disable}'
Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-id: 1429314609-29776-5-git-send-email-jsnow@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
We treat this field with a variety of different types everywhere
in the code. Now it's just uint32_t.
Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-id: 1429314609-29776-4-git-send-email-jsnow@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
This field will be set for user created dirty bitmap. Also pass in an
error pointer to bdrv_create_dirty_bitmap, so when a name is already
taken on this BDS, it can report an error message. This is not global
check, two BDSes can have dirty bitmap with a common name.
Implemented bdrv_find_dirty_bitmap to find a dirty bitmap by name, will
be used later when other QMP commands want to reference dirty bitmap by
name.
Add bdrv_dirty_bitmap_make_anon. This unsets the name of dirty bitmap.
Signed-off-by: Fam Zheng <famz@redhat.com>
Signed-off-by: John Snow <jsnow@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-id: 1429314609-29776-3-git-send-email-jsnow@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
The image field in BlockDeviceInfo is supposed to contain an ImageInfo
object. However that is being filled in by bdrv_query_info(), not by
bdrv_block_device_info(), which is where BlockDeviceInfo is actually
created.
Anyone calling bdrv_block_device_info() directly will get a null image
field. As a consequence of this, the HMP command 'info block -n -v'
crashes QEMU.
This patch moves the code that fills in that field from
bdrv_query_info() to bdrv_block_device_info().
Signed-off-by: Alberto Garcia <berto@igalia.com>
Message-id: 1429271563-3765-1-git-send-email-berto@igalia.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
This function gets the device name associated with a BlockDriverState,
or its node name if the device name is empty.
Signed-off-by: Alberto Garcia <berto@igalia.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-id: 4fa30aa8d61d9052ce266fd5429a59a14e941255.1428485266.git.berto@igalia.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
This patch changes block_job_pause to increase the pause counter and
block_job_resume to decrease it.
The counter will allow calling block_job_pause/block_job_resume
unconditionally on a job when we need to suspend the IO temporarily.
From now on, each block_job_resume must be paired with a block_job_pause
to keep the counter balanced.
The user pause from QMP or HMP will only trigger block_job_pause once
until it's resumed, this is achieved by adding a user_paused flag in
BlockJob.
One occurrence of block_job_resume in mirror_complete is replaced with
block_job_enter which does what is necessary.
In block_job_cancel, the cancel flag is good enough to instruct
coroutines to quit loop, so use block_job_enter to replace the unpaired
block_job_resume.
Upon block job IO error, user is notified about the entering to the
pause state, so this pause belongs to user pause, set the flag
accordingly and expect a matching QMP resume.
[Extended doc comments as suggested by Paolo Bonzini
<pbonzini@redhat.com>.
--Stefan]
Signed-off-by: Fam Zheng <famz@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Alberto Garcia <berto@igalia.com>
Message-id: 1428069921-2957-2-git-send-email-famz@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
This is the first step in pushing down acquire/release, and will let
rfifolock drop the contention callback feature.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-id: 1424449612-18215-3-git-send-email-pbonzini@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
By using thread-local storage, aio_poll can stop using global data during
g_poll_ns. This will make it possible to drop callbacks from rfifolock.
[Moved npfd = 0 assignment to end of walking_handlers region as
suggested by Paolo. This resolves the assert(npfd == 0) assertion
failure in pollfds_cleanup().
--Stefan]
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-id: 1424449612-18215-2-git-send-email-pbonzini@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Max Reitz <mreitz@redhat.com>
Message-Id: <1424887718-10800-13-git-send-email-mreitz@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Max Reitz <mreitz@redhat.com>
Message-Id: <1424887718-10800-11-git-send-email-mreitz@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Max Reitz <mreitz@redhat.com>
Message-Id: <1424887718-10800-9-git-send-email-mreitz@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
All callers are converted, so drop it.
Signed-off-by: Fam Zheng <famz@redhat.com>
Message-id: 1425296209-1476-5-git-send-email-famz@redhat.com
Reviewed-by: Max Reitz <mreitz@redhat.com>
Signed-off-by: Max Reitz <mreitz@redhat.com>
Add driver functions for geometry and blocksize detection
Signed-off-by: Ekaterina Tumanova <tumanova@linux.vnet.ibm.com>
Reviewed-by: Thomas Huth <thuth@linux.vnet.ibm.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-id: 1424087278-49393-2-git-send-email-tumanova@linux.vnet.ibm.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Add a creation option to qcow2 for setting the refcount order of images
to be created, and respect that option's value.
This breaks some test outputs, fix them.
Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Previously, qemu block driver of sheepdog used hard-coded VDI object size.
This patch enables users to handle VDI object size.
When you start qemu, you don't need to specify additional command option.
But when you create the VDI which doesn't have default object size
with qemu-img command, you specify object_size option.
If you want to create a VDI of 8MB object size,
you need to specify following command option.
# qemu-img create -o object_size=8M sheepdog:test1 100M
In addition, when you don't specify qemu-img command option,
a default value of sheepdog cluster is used for creating VDI.
# qemu-img create sheepdog:test2 100M
Signed-off-by: Teruaki Ishizaki <ishizaki.teruaki@lab.ntt.co.jp>
Acked-by: Hitoshi Mitake <mitake.hitoshi@lab.ntt.co.jp>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
qemu_coroutine_enter() is now the only user of coroutine_swap(). Both
functions are short, so inline it.
Also, using COROUTINE_YIELD is now even more confusing because this code
is never called during qemu_coroutine_yield() any more. In fact, this
value is never read back, so we can just introduce a new COROUTINE_ENTER
which documents the purpose of the task switch better.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Even if the caller has both the old and the new AioContext's, there can
be a deadlock, due to the leading bdrv_drain_all.
Suppose there are four io threads (A, B, A0, B0) with A and B owning a
BDS for each (bs_a, bs_b); Now A wants to move bs_a to iothread A0, and
B wants to move bs_b to B0, at the same time:
iothread A iothread B
--------------------------------------------------------------------------
aio_context_acquire(A0) /* OK */ aio_context_acquire(B0) /* OK */
bdrv_set_aio_context(bs_a, A0) bdrv_set_aio_context(bs_b, B0)
-> bdrv_drain_all() -> bdrv_drain_all()
-> acquire A /* OK */ -> acquire A /* blocked */
-> acquire B /* blocked */ -> acquire B
... ...
Deadlock happens because A is waiting for B, and B is waiting for A.
Signed-off-by: Fam Zheng <famz@redhat.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Message-Id: <1423969591-23646-2-git-send-email-famz@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Now that request clamping is done in the BlockBackend, the "growable"
field can be removed from the BlockDriverState. All BDSs are now treated
as being "growable" (that is, they are allowed to grow; they are not
necessarily actually able to).
Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-id: 1423162705-32065-16-git-send-email-mreitz@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
