Directly pass the QEMUIOVector on instead of linearising it.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
The wait_time variable is in seconds. Reflect this in a comment and use
NANOSECONDS_PER_SECOND instead of BLOCK_IO_SLICE_TIME * 10 (which
happens to have the right value).
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Tested-By: Benoit Canet <benoit@irqsave.net>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
The current slice is extended when an I/O request exceeds the limit.
There is no need to extend the slice every time we check a request.
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Tested-By: Benoit Canet <benoit@irqsave.net>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
It is not necessary to adjust the slice time at runtime. We already
extend the current slice in order to carry over accounting into the next
slice. Changing the actual slice time value introduces oscillations.
The guest may experience large changes in throughput or IOPS from one
moment to the next when slice times are adjusted.
Reported-by: Benoît Canet <benoit@irqsave.net>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Tested-By: Benoit Canet <benoit@irqsave.net>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
I/O throttling relies on bdrv_acct_done() which is called when a request
completes. This leaves a blind spot since we only charge for completed
requests, not submitted requests.
For example, if there is 1 operation remaining in this time slice the
guest could submit 3 operations and they will all be submitted
successfully since they don't actually get accounted for until they
complete.
Originally we probably thought this is okay since the requests will be
accounted when the time slice is extended. In practice it causes
fluctuations since the guest can exceed its I/O limit and it will be
punished for this later on.
Account for I/O upon submission so that I/O limits are enforced
properly.
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Tested-By: Benoit Canet <benoit@irqsave.net>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
bdrv_open_common() implements direct use of protocols by copying the
pre-opened BlockDriverStates to bs using bdrv_swap(). It did however
first set some fields in bs, which end up in file after the swap. When
bdrv_open() destroys file, it appears to be open, and because it isn't,
qemu could segfault while trying to close it.
Reorder the operations to return immediately in such cases so that file
is correctly detected as closed.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
After this patch, using -drive with an empty file name continues to open
the file if driver-specific options are used. If no driver-specific
options are specified, the semantics stay as it was: It defines a drive
without an inserted medium.
In order to achieve this, bdrv_open() must be made safe to work with a
NULL filename parameter. The assumption that is made is that only block
drivers which implement bdrv_parse_filename() support using driver
specific options and could therefore work without a filename. These
drivers must make sure to cope with NULL in their implementation of
.bdrv_open() (this is only NBD for now). For all other drivers, the
block layer code will make sure to error out before calling into their
code - they can't possibly work without a filename.
Now an NBD connection can be opened like this:
qemu-system-x86_64 -drive file.driver=nbd,file.port=1234,file.host=::1
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
bdrv_open() uses two different variables called options. Rename one of
them to avoid confusion and to allow the outer one to be accessed
everywhere.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
If a driver needs structured data and not just a string, it can provide
a .bdrv_parse_filename callback now that parses the command line string
into separate options. Keeping this separate from .bdrv_open_filename
ensures that the preferred way of directly specifying the options always
works as well if parsing the string works.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Specify -drive file.option=... on the command line to pass the option to
the protocol instead of the format driver.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
brdv_truncate() is also called from readv/writev commands on self-
growing file based storage. this will result in requests waiting
for theirselves to complete.
This reverts commit 9a665b2b86.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
realpath(3) is used to get an absolute path to the image file when
creating a -drive snapshot=on temporary qcow2. This does not work for
protocols since their filenames ("proto:foo:...") do not correspond to
file system paths.
Commit 7c96d46ec2 ("Let snapshot work with
protocols") skipped realpath(3) for protocols. Later on the "raw"
format was introduced and broke the check.
Use path_has_protocol(filename) to decide if this image uses a protocol
or a filename.
Reported-by: Richard Jones <rjones@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
For now bdrv_get_aio_context() is just a stub that calls
qemu_aio_get_context() since the block layer is currently tied to the
main loop AioContext.
Add the stub now so that the block layer can begin accessing its
AioContext.
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
The options are passed down to the block drivers, which are supposed to
remove all options they have processed. Anything that is left over in
the end is an unknown option and results in an error.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
It doesn't do anything yet except storing the options QDict in the
BlockDriverState.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
During a commit of 'all' using the HMP non-live commit, the operation
is aborted and returns error on the first error enountered. When
non-COW drives are in use (e.g. ejected floppy, cdrom, or drives without
a backing parent), that means a commit all will return an error of either
-ENOMEDIUM or -ENOTSUP. This is not desirable, so for the 'all' commit
case, only attempt the commit if both bs->drv and bs->backing_hd are
present.
More succinctly: 'commit all' now means a commit on all COW drives.
This means an individual commit to a specific non-COW drive will still
return the appropriate error (-ENOMEDIUM if eject / not present, -ENOTSUP
if no backing file).
Reported-by: Jan Kiszka <jan.kiszka@web.de>
Signed-off-by: Jeff Cody <jcody@redhat.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
It is better to present homogeneous hardware independent of the storage
technology that is chosen on the host, hence we make discard a host
parameter; the user can choose whether to pass it down to the image
format and protocol, or to ignore it.
Using DISCARD with filesystems can cause very severe fragmentation, so it
is left default-off for now. This can change later when we implement the
"anchor" operation for efficient management of preallocated files.
There is still one choice to make: whether DISCARD has an effect on the
dirty bitmap or not. I chose yes, though there is a disadvantage: if
the guest is buggy and issues discards for data that is in use, there
will be no way to migrate storage for that guest without downgrading
the machine type to an older one.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
bdrv_truncate() invalidates the bdrv_check_request() result for
in-flight requests, so there should better be none.
Cc: qemu-stable@nongnu.org
Signed-off-by: Peter Lieven <pl@kamp.de>
Reported-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
There can be a need to turn output to stdout off. This patch adds a -q option
that enable "Quiet mode". In Quiet mode, only errors are printed out.
Signed-off-by: Miroslav Rezanina <mrezanin@redhat.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
There's no synchronous wrapper for bdrv_co_is_allocated_above function
so it's not possible to check for sector allocation in an image with
a backing file.
Signed-off-by: Miroslav Rezanina <mrezanin@redhat.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
In an image chain, if the base image is smaller than the current
image, we need to make sure to use the current images count of
unallocated blocks once we get to the end of the base image. Without
this change the code will return 0 blocks when it gets to the end
of the base image and mirror_run will fail its assertion.
Signed-off-by: Vishvananda Ishaya <vishvananda@gmail.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
This is needed in the following patch.
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
This actually uses the dirty bitmap in the block layer, and converts
mirroring to use an HBitmapIter.
Reviewed-by: Laszlo Ersek <lersek@redhat.com> (except block/mirror.c parts)
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Note that resetting bits in the dirty bitmap is done _before_ actually
processing the request. Writes, instead, set bits after the request
is completed.
This way, when there are concurrent write and discard requests, the
outcome will always be that the blocks are marked dirty. This scenario
should never happen, but it is safer to do it this way.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
bdrv_io_limits_enable() starts a new slice, but does not set io_base
correctly for that slice.
Here is how io_base is used:
bytes_base = bs->nr_bytes[is_write] - bs->io_base.bytes[is_write];
bytes_res = (unsigned) nb_sectors * BDRV_SECTOR_SIZE;
if (bytes_base + bytes_res <= bytes_limit) {
/* no wait */
} else {
/* operation needs to be throttled */
}
As a result, any I/O operations that are triggered between now and
bs->slice_end are incorrectly limited. If 10 MB of data has been
written since the VM was started, QEMU thinks that 10 MB of data has
been written in this slice. This leads to a I/O lockup in the guest.
We fix this by delaying the start of a new slice to the next
call of bdrv_exceed_io_limits().
Signed-off-by: Peter Lieven <pl@kamp.de>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
The qiov_is_aligned() function checks whether a QEMUIOVector meets a
BlockDriverState's alignment requirements. This is needed by
virtio-blk-data-plane so:
1. Move the function from block/raw-posix.c to block/block.c.
2. Make it public in block/block.h.
3. Rename to bdrv_qiov_is_aligned().
4. Change return type from int to bool.
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
A blank CD or DVD is visible as a zero-sized disks. Probing such
disks will lead to an EIO and a failure to start the VM. Treating
them as raw is a better solution.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
This allows removing of MinGW specific code and improves
reentrancy for POSIX hosts.
[Removed unused ret variable in qemu_get_timedate() to fix warning:
vl.c: In function ‘qemu_get_timedate’:
vl.c:451:16: error: variable ‘ret’ set but not used [-Werror=unused-but-set-variable]
-- Stefan Hajnoczi]
Signed-off-by: Stefan Weil <sw@weilnetz.de>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
This makes the blkdebug suspend/resume functionality available in
qemu-io. Use it like this:
$ ./qemu-io blkdebug::/tmp/test.qcow2
qemu-io> break write_aio req_a
qemu-io> aio_write 0 4k
qemu-io> blkdebug: Suspended request 'req_a'
qemu-io> resume req_a
blkdebug: Resuming request 'req_a'
qemu-io> wrote 4096/4096 bytes at offset 0
4 KiB, 1 ops; 0:00:30.71 (133.359788 bytes/sec and 0.0326 ops/sec)
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
This commit adds an Error ** argument to bdrv_img_create() and set it
appropriately on error.
Callers of bdrv_img_create() pass NULL for the new argument and still
rely on bdrv_img_create()'s return value. Next commits will change
callers to use the Error object instead.
Signed-off-by: Luiz Capitulino <lcapitulino@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
This fixes problems that are caused by the additional open/close cycle
of the existing format probing, for example related to qemu-nbd without
-t option or file descriptor passing.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Instead of waiting for all requests to complete, wait just for the
specific request that should be cancelled.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
The local string tmp_filename is passed to function get_tmp_filename
which expects a string with minimum size MAX_PATH for w32 hosts.
MAX_PATH is 260 and PATH_MAX is 259, so tmp_filename was too short.
Commit eba25057b9 introduced this
regression.
Signed-off-by: Stefan Weil <sw@weilnetz.de>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
Now that AIOPool no longer keeps a freelist, it isn't really a "pool"
anymore. Rename it to AIOCBInfo and make it const since it no longer
needs to be modified.
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
AIO control blocks are frequently acquired and released because each aio
request involves at least one AIOCB. Therefore, we pool them to avoid
heap allocation overhead.
The problem with the freelist approach in AIOPool is thread-safety. If
we want BlockDriverStates to associate with AioContexts that execute in
multiple threads, then a global freelist becomes a problem.
This patch drops the freelist and instead uses g_slice_alloc() which is
tuned for per-thread fixed-size object pools. qemu_aio_get() and
qemu_aio_release() are now thread-safe.
Note that the change from g_malloc0() to g_slice_alloc() should be safe
since the freelist reuse case doesn't zero the AIOCB either.
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
* kwolf/for-anthony: (32 commits)
osdep: Less restrictive F_SEFL in qemu_dup_flags()
qemu-iotests: add testcases for mirroring on-source-error/on-target-error
qmp: add pull_event function
mirror: add support for on-source-error/on-target-error
iostatus: forward block_job_iostatus_reset to block job
qemu-iotests: add mirroring test case
mirror: implement completion
qmp: add drive-mirror command
mirror: introduce mirror job
block: introduce BLOCK_JOB_READY event
block: add block-job-complete
block: rename block_job_complete to block_job_completed
block: export dirty bitmap information in query-block
block: introduce new dirty bitmap functionality
block: add bdrv_open_backing_file
block: add bdrv_query_stats
block: add bdrv_query_info
qemu-config: Add new -add-fd command line option
monitor: Prevent removing fd from set during init
monitor: Enable adding an inherited fd to an fd set
...
Conflicts:
vl.c
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
