ratelimit_calculate_delay() previously reset the accounting every time
slice, no matter how much data had been processed before. This had (at
least) two consequences:
1. The minimum speed is rather large, e.g. 5 MiB/s for commit and stream.
Not sure if there are real-world use cases where this would be a
problem. Mirroring and backup over a slow link (e.g. DSL) would
come to mind, though.
2. Tests for block job operations (e.g. cancel) were rather racy
All block jobs currently use a time slice of 100ms. That's a
reasonable value to get smooth output during regular
operation. However this also meant that the state of block jobs
changed every 100ms, no matter how low the configured limit was. On
busy hosts, qemu often transferred additional chunks until the test
case had a chance to cancel the job.
Fix the block job rate limit code to delay for more than one time
slice to address the above issues. To make it easier to handle
oversized chunks we switch the semantics from returning a delay
_before_ the current request to a delay _after_ the current
request. If necessary, this delay consists of multiple time slice
units.
Since the mirror job sends multiple chunks in one go even if the rate
limit was exceeded in between, we need to keep track of the start of
the current time slice so we can correctly re-compute the delay for
the updated amount of data.
The minimum bandwidth now is 1 data unit per time slice. The block
jobs are currently passing the amount of data transferred in sectors
and using 100ms time slices, so this translates to 5120
bytes/second. With chunk sizes usually being O(512KiB), tests have
plenty of time (O(100s)) to operate on block jobs. The chance of a
race condition now is fairly remote, except possibly on insanely
loaded systems.
Signed-off-by: Sascha Silbe <silbe@linux.vnet.ibm.com>
Message-id: 1467127721-9564-2-git-send-email-silbe@linux.vnet.ibm.com
Reviewed-by: Max Reitz <mreitz@redhat.com>
Signed-off-by: Max Reitz <mreitz@redhat.com>
In practice the entry argument is always known at creation time, and
it is confusing that sometimes qemu_coroutine_enter is used with a
non-NULL argument to re-enter a coroutine (this happens in
block/sheepdog.c and tests/test-coroutine.c). So pass the opaque value
at creation time, for consistency with e.g. aio_bh_new.
Mostly done with the following semantic patch:
@ entry1 @
expression entry, arg, co;
@@
- co = qemu_coroutine_create(entry);
+ co = qemu_coroutine_create(entry, arg);
...
- qemu_coroutine_enter(co, arg);
+ qemu_coroutine_enter(co);
@ entry2 @
expression entry, arg;
identifier co;
@@
- Coroutine *co = qemu_coroutine_create(entry);
+ Coroutine *co = qemu_coroutine_create(entry, arg);
...
- qemu_coroutine_enter(co, arg);
+ qemu_coroutine_enter(co);
@ entry3 @
expression entry, arg;
@@
- qemu_coroutine_enter(qemu_coroutine_create(entry), arg);
+ qemu_coroutine_enter(qemu_coroutine_create(entry, arg));
@ reentry @
expression co;
@@
- qemu_coroutine_enter(co, NULL);
+ qemu_coroutine_enter(co);
except for the aforementioned few places where the semantic patch
stumbled (as expected) and for test_co_queue, which would otherwise
produce an uninitialized variable warning.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Fam Zheng <famz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
This patch adds a new optional 'job-id' parameter to 'block-stream',
allowing the user to specify the ID of the block job to be created.
The HMP 'block_stream' command remains unchanged.
Signed-off-by: Alberto Garcia <berto@igalia.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
When a new job is created, the job ID is taken from the device name of
the BDS. This patch adds a new 'job_id' parameter to let the caller
provide one instead.
This patch also verifies that the ID is always unique and well-formed.
This causes problems in a couple of places where no ID is being set,
because the BDS does not have a device name.
In the case of test_block_job_start() (from test-blockjob-txn.c) we
can simply use this new 'job_id' parameter to set the missing ID.
In the case of img_commit() (from qemu-img.c) we still don't have the
API to make commit_active_start() set the job ID, so we solve it by
setting a default value. We'll get rid of this as soon as we extend
the API.
Signed-off-by: Alberto Garcia <berto@igalia.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
This changes the streaming block job to use the job's BlockBackend for
performing the COR reads. job->bs isn't used by the streaming code any
more afterwards.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Alberto Garcia <berto@igalia.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Block jobs don't actually make use of the iostatus for their BDSes, but
they manage a separate block job iostatus. Still, they require that it
is enabled for the source BDS and they enable it automatically for the
target and set the error handling mode - which ends up never being used
by the job.
This patch removes all of the BDS iostatus handling from the block job,
which removes another few bs->blk accesses.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
When block job errors were introduced, we assigned the iostatus of the
target BDS "just in case". The field has never been accessible for the
user because the target isn't listed in query-block.
Before we can allow the user to have a second BlockBackend on the
target, we need to clean this up. If anything, we would want to set the
iostatus for the internal BB of the job (which we can always do later),
but certainly not for a separate BB which the job doesn't even use.
As a nice side effect, this gets us rid of another bs->blk use.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
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 if we introduce support for streaming to
any intermediate node.
Signed-off-by: Alberto Garcia <berto@igalia.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Fam Zheng <famz@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-id: 2abedf2debc65c250560237f31a8e6756883c8fc.1458566441.git.berto@igalia.com
Signed-off-by: Jeff Cody <jcody@redhat.com>
Commit 57cb38b included qapi/error.h into qemu/osdep.h to get the
Error typedef. Since then, we've moved to include qemu/osdep.h
everywhere. Its file comment explains: "To avoid getting into
possible circular include dependencies, this file should not include
any other QEMU headers, with the exceptions of config-host.h,
compiler.h, os-posix.h and os-win32.h, all of which are doing a
similar job to this file and are under similar constraints."
qapi/error.h doesn't do a similar job, and it doesn't adhere to
similar constraints: it includes qapi-types.h. That's in excess of
100KiB of crap most .c files don't actually need.
Add the typedef to qemu/typedefs.h, and include that instead of
qapi/error.h. Include qapi/error.h in .c files that need it and don't
get it now. Include qapi-types.h in qom/object.h for uint16List.
Update scripts/clean-includes accordingly. Update it further to match
reality: replace config.h by config-target.h, add sysemu/os-posix.h,
sysemu/os-win32.h. Update the list of includes in the qemu/osdep.h
comment quoted above similarly.
This reduces the number of objects depending on qapi/error.h from "all
of them" to less than a third. Unfortunately, the number depending on
qapi-types.h shrinks only a little. More work is needed for that one.
Signed-off-by: Markus Armbruster <armbru@redhat.com>
[Fix compilation without the spice devel packages. - Paolo]
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Clean up includes so that osdep.h is included first and headers
which it implies are not included manually.
This commit was created with scripts/clean-includes.
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
These options are only relevant for the user of a whole BDS tree (like a
guest device or a block job) and should thus be moved into the
BlockBackend.
Signed-off-by: Max Reitz <mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
This simplifies the code somewhat, especially when dropping whole
backing file subchains.
The exception is the mirroring code that does adventurous things with
bdrv_swap() and in order to keep it working, I had to duplicate most of
bdrv_set_backing_hd() locally. We'll get rid again of this ugliness
shortly.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Alberto Garcia <berto@igalia.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
This is the final step in converting all of the BlockDriverState
pointers that block drivers use to BdrvChild.
After this patch, bs->children contains the full list of child nodes
that are referenced by a given BDS, and these children are only
referenced through BdrvChild, so that updating the pointer in there is
enough for changing edges in the graph.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Alberto Garcia <berto@igalia.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Reviewed-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>
These macros expand into error class enumeration constant, comma,
string. Unclean. Has been that way since commit 13f59ae.
The error class is always ERROR_CLASS_GENERIC_ERROR since the previous
commit.
Clean up as follows:
* Prepend every use of a QERR_ macro by ERROR_CLASS_GENERIC_ERROR, and
delete it from the QERR_ macro. No change after preprocessing.
* Rewrite error_set(ERROR_CLASS_GENERIC_ERROR, ...) into
error_setg(...). Again, no change after preprocessing.
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>
The stream block job must run in the BlockDriverState AioContext so that
it works with dataplane.
The basics of acquiring the AioContext are easy in blockdev.c.
The tricky part is the completion code which drops part of the backing
file chain. This must be done in the main loop where bdrv_unref() and
bdrv_close() are safe to call. Use block_job_defer_to_main_loop() to
achieve that.
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-id: 1413889440-32577-9-git-send-email-stefanha@redhat.com
I'll use it with block backends shortly, and the name is going to fit
badly there. It's a block layer thing anyway, not just a block driver
thing.
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
On some image chains, QEMU may not always be able to resolve the
filenames properly, when updating the backing file of an image
after a block job.
For instance, certain relative pathnames may fail, or drives may
have been specified originally by file descriptor (e.g. /dev/fd/???),
or a relative protocol pathname may have been used.
In these instances, QEMU may lack the information to be able to make
the correct choice, but the user or management layer most likely does
have that knowledge.
With this extension to the block-stream api, the user is able to change
the backing file of the active layer as part of the block-stream
operation.
This allows the change to be 'safe', in the sense that if the attempt
to write the active image metadata fails, then the block-stream
operation returns failure, without disrupting the guest.
If a backing file string is not specified in the command, the backing
file string to use is determined in the same manner as it was
previously.
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Jeff Cody <jcody@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
In order to let event defines use existing types later, instead of
redefine new ones, some old type defines for spice and vnc are changed,
and BlockErrorAction is moved from block.h to qapi schema. Note that
BlockErrorAction is not merged with BlockdevOnError.
At this point, VncInfo is not made a child of VncBasicInfo, because
VncBasicInfo has mandatory fields where VncInfo makes them optional.
Signed-off-by: Wenchao Xia <wenchaoqemu@gmail.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Luiz Capitulino <lcapitulino@redhat.com>
We need to handle the coming backing_blocker properly, so don't open
code the assignment, instead, call bdrv_set_backing_hd to change
backing_hd.
Signed-off-by: Fam Zheng <famz@redhat.com>
Reviewed-by: Jeff Cody <jcody@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
When reopening with different flags, or when backing files disappear
from the chain, the limits may change. Make sure they get updated in
these cases.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Wenchao Xia <xiawenc@linux.vnet.ibm.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Benoît Canet <benoit@irqsave.net>
If a block device is unbacked, a streaming blockjob should immediately
finish instead of beginning to try to stream, then noticing the backing
file does not contain even the first sector (since it does not exist)
and then finishing normally.
Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Wenchao Xia <xiawenc@linux.vnet.ibm.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Switch the string to enum type BlockJobType in BlockJobDriver.
Signed-off-by: Fam Zheng <famz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
We will use BlockJobType as the enum type name of block jobs in QAPI,
rename current BlockJobType to BlockJobDriver, which will eventually
become a set of operations, similar to block drivers.
Signed-off-by: Fam Zheng <famz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
The patch fixes a warning from gcc (Debian 4.6.3-14+rpi1) 4.6.3:
block/stream.c:141:22: error:
‘copy’ may be used uninitialized in this function [-Werror=uninitialized]
This is not a real bug - a better compiler would not complain.
Now 'copy' has always a defined value, so the check for ret >= 0
can be removed.
Signed-off-by: Stefan Weil <sw@weilnetz.de>
Acked-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Some bdrv_is_allocated callers do not expect errors, but the fallback
in qcow2.c might make other callers trip on assertion failures or
infinite loops.
Fix the callers to always look for errors.
Cc: qemu-stable@nongnu.org
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Now that bdrv_is_allocated detects coroutine context, the two can
use the same code.
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
bdrv_is_allocated can detect coroutine context and go through a fast
path, similar to other block layer functions.
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Manage BlockDriverState lifecycle with refcnt, so bdrv_delete() is no
longer public and should be called by bdrv_unref() if refcnt is
decreased to 0.
This is an identical change because effectively, there's no multiple
reference of BDS now: no caller of bdrv_ref() yet, only bdrv_new() sets
bs->refcnt to 1, so all bdrv_unref() now actually delete the BDS.
Signed-off-by: Fam Zheng <famz@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Convert block_job_sleep_ns and co_sleep_ns to use the new timer
API.
Signed-off-by: Alex Bligh <alex@alex.org.uk>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
If a block driver has no file descriptors to monitor but there are still
active requests, it can return 1 from .io_flush(). This is used to spin
during synchronous I/O.
Stop relying on .io_flush() and instead check
QLIST_EMPTY(&bs->tracked_requests) to decide whether there are active
requests.
This is the first step in removing .io_flush() so that event loops no
longer need to have the concept of synchronous I/O. Eventually we may
be able to kill synchronous I/O completely by running everything in a
coroutine, but that is future work.
Note this patch moves bs->throttled_reqs initialization to bdrv_new() so
that bdrv_requests_pending(bs) can safely access it. In practice bs is
g_malloc0() so the memory is already zeroed but it's safer to initialize
the queue properly.
We also need to fix up block/stream.c:close_unused_images() to prevent
traversing a dangling pointer while it rearranges the backing file
chain. This is necessary since the new bdrv_drain_all() traverses the
backing file chain.
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
This patch adds support for error management to streaming.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
It is possible to create an image that is larger than its backing file.
Reading beyond the end of the backing file produces zeroes if no writes
have been made to those sectors in the image file.
This patch finishes streaming early when the end of the backing file is
reached. Without this patch the block job hangs and continually tries
to stream the first sectors beyond the end of the backing file.
To reproduce the hung block job bug:
$ qemu-img create -f qcow2 backing.qcow2 128M
$ qemu-img create -f qcow2 -o backing_file=backing.qcow2 image.qcow2 6G
$ qemu -drive if=virtio,cache=none,file=image.qcow2
(qemu) block_stream virtio0
(qemu) info block-jobs
The qemu-iotests 030 streaming test still passes.
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
is_allocated_base has complex semantics that are not really usable
outside streaming. Split the check in two parts, where the allocated
state for the top bs is moved to the caller. The resulting function
is more generally useful.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Unallocated sectors should really never be accessed by the guest,
so there's no need to copy them during the streaming process.
If they are read by the guest during streaming, guest-initiated
copy-on-read will copy them (we're in the base == NULL case, which
enables copy on read). If they are read after we disconnect the
image from the base, they will read as zeroes anyway.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
This fixes inability to make progress in streaming if the quota is set
to less than the amount of data that an I/O operation has to write.
In this case, limit->dispatched + n will always be above the quota and,
due to the "goto retry" to recheck cancellation and allocation, streaming
will livelock.
This can be reproduced with "block_job_set_speed ide0-hd0 1b". Of course,
with this patch the requested limit will not be obeyed. That could be
done with another patch that caps is_allocated's n argument by the slice
quota.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
When an image is modified to point to the new backing file, the backing
file format is set to NULL, which means auto-probe. This is wrong, in
fact it is a small security problem.
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
The limitation on not having I/O after cancellation cannot really be
kept. Even streaming has a very small race window where you could
cancel a job and have it report completion. If this window is hit,
bdrv_change_backing_file() will yield and possibly cause accesses to
dangling pointers etc.
So, let's just assume that we cannot know exactly what will happen
after the coroutine has set busy to false. We can set a very lax
condition:
- if we cancel the job, the coroutine won't set it to false again
(and hence will not call co_sleep_ns again).
- block_job_cancel_sync will wait for the coroutine to exit, which
pretty much ensures no race.
Instead, we track the coroutine that executes the job and put very
strict conditions on what to do while it is quiescent (busy = false).
First of all, the coroutine must never set busy = false while the job
has been cancelled. Second, the coroutine can be reentered arbitrarily
while it is quiescent, so you cannot really do anything but co_sleep_ns at
that time. This condition is obeyed by the block_job_sleep_ns function.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
This function abstracts the pretty complex semantics of the "busy"
member of BlockJob.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
These are needed to print "info block" output correctly. QCOW2 does this
because it needs it to write the header, but QED does not, and common code
is the right place to do it.
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
