The handling for the XFS_IOC_DIOINFO ioctl is currently quite excessive:
This is not a "real" feature like the other features that we provide with
the "--enable-xxx" and "--disable-xxx" switches for the configure script,
since this does not influence lots of code (it's only about one call to
xfsctl() in file-posix.c), so people don't gain much with the ability to
disable this with "--disable-xfsctl".
It's also unfortunate that the ioctl will be disabled on Linux in case
the user did not install the right xfsprogs-devel package before running
configure. Thus let's simplify this by providing the ioctl definition
on our own, so we can completely get rid of the header dependency and
thus the related code in the configure script.
Suggested-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Thomas Huth <thuth@redhat.com>
Message-Id: <20211215125824.250091-1-thuth@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
At the end of a reopen, we already call bdrv_refresh_limits(), which
should update bs->request_alignment according to the new file
descriptor. However, raw_probe_alignment() relies on s->needs_alignment
and just uses 1 if it isn't set. We neglected to update this field, so
starting with cache=writeback and then reopening with cache=none means
that we get an incorrect bs->request_alignment == 1 and unaligned
requests fail instead of being automatically aligned.
Fix this by recalculating s->needs_alignment in raw_refresh_limits()
before calling raw_probe_alignment().
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20211104113109.56336-1-kwolf@redhat.com>
Reviewed-by: Hanna Reitz <hreitz@redhat.com>
Reviewed-by: Stefano Garzarella <sgarzare@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20211115145409.176785-13-kwolf@redhat.com>
[hreitz: Fix iotest 142 for block sizes greater than 512 by operating on
a file with a size of 1 MB]
Signed-off-by: Hanna Reitz <hreitz@redhat.com>
Message-Id: <20211116101431.105252-1-hreitz@redhat.com>
Between the submission of a request and the unplug, other devices
with larger limits may have been queued new requests without flushing
the batch.
Using the new `dev_max_batch` parameter, laio_io_unplug() can check
if the batch exceeds the device limit to flush the current batch.
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Stefano Garzarella <sgarzare@redhat.com>
Message-Id: <20211026162346.253081-4-sgarzare@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
This new parameter can be used by block devices to limit the
Linux AIO batch size more than the limit set by the AIO context.
file-posix backend supports this, passing its `aio-max-batch` option
previously added.
Add an helper function to calculate the maximum batch size.
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Stefano Garzarella <sgarzare@redhat.com>
Message-Id: <20211026162346.253081-3-sgarzare@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Commit d7ddd0a161 ("linux-aio: limit the batch size using
`aio-max-batch` parameter") added a way to limit the batch size
of Linux AIO backend for the entire AIO context.
The same AIO context can be shared by multiple devices, so
latency-sensitive devices may want to limit the batch size even
more to avoid increasing latency.
For this reason we add the `aio-max-batch` option to the file
backend, which will be used by the next commits to limit the size of
batches including requests generated by this device.
Suggested-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Stefano Garzarella <sgarzare@redhat.com>
Message-Id: <20211026162346.253081-2-sgarzare@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
AIO discards regressed as a result of the following commit:
0dfc7af2 block/file-posix: Optimize for macOS
When trying to run blkdiscard within a Linux guest, the request would
fail, with some errors in dmesg:
---- [ snip ] ----
[ 4.010070] sd 2:0:0:0: [sda] tag#0 FAILED Result: hostbyte=DID_OK
driverbyte=DRIVER_SENSE
[ 4.011061] sd 2:0:0:0: [sda] tag#0 Sense Key : Aborted Command
[current]
[ 4.011061] sd 2:0:0:0: [sda] tag#0 Add. Sense: I/O process
terminated
[ 4.011061] sd 2:0:0:0: [sda] tag#0 CDB: Unmap/Read sub-channel 42
00 00 00 00 00 00 00 18 00
[ 4.011061] blk_update_request: I/O error, dev sda, sector 0
---- [ snip ] ----
This turns out to be a result of a flaw in changes to the error value
translation logic in handle_aiocb_discard(). The default return value
may be left untranslated in some configurations, and the wrong variable
is used in one translation.
Fix both issues.
Fixes: 0dfc7af2b2 ("block/file-posix: Optimize for macOS")
Cc: qemu-stable@nongnu.org
Signed-off-by: Ari Sundholm <ari@tuxera.com>
Signed-off-by: Emil Karlson <jkarlson@tuxera.com>
Reviewed-by: Akihiko Odaki <akihiko.odaki@gmail.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-Id: <20211019110954.4170931-1-ari@tuxera.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Linux limits the size of iovecs to 1024 (UIO_MAXIOV in the kernel
sources, IOV_MAX in POSIX). Because of this, on some host adapters
requests with many iovecs are rejected with -EINVAL by the
io_submit() or readv()/writev() system calls.
In fact, the same limit applies to SG_IO as well. To fix both the
EINVAL and the possible performance issues from using fewer iovecs
than allowed by Linux (some HBAs have max_segments as low as 128),
introduce a separate entry in BlockLimits to hold the max_segments
value from sysfs. This new limit is used only for SG_IO and clamped
to bs->bl.max_iov anyway, just like max_hw_transfer is clamped to
bs->bl.max_transfer.
Reported-by: Halil Pasic <pasic@linux.ibm.com>
Cc: Hanna Reitz <hreitz@redhat.com>
Cc: Kevin Wolf <kwolf@redhat.com>
Cc: qemu-block@nongnu.org
Cc: qemu-stable@nongnu.org
Fixes: 18473467d5 ("file-posix: try BLKSECTGET on block devices too, do not round to power of 2", 2021-06-25)
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-Id: <20210923130436.1187591-1-pbonzini@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
We are generally moving to int64_t for both offset and bytes parameters
on all io paths.
Main motivation is realization of 64-bit write_zeroes operation for
fast zeroing large disk chunks, up to the whole disk.
We chose signed type, to be consistent with off_t (which is signed) and
with possibility for signed return type (where negative value means
error).
So, convert driver discard handlers bytes parameter to int64_t.
The only caller of all updated function is bdrv_co_pdiscard in
block/io.c. It is already prepared to work with 64bit requests, but
pass at most max(bs->bl.max_pdiscard, INT_MAX) to the driver.
Let's look at all updated functions:
blkdebug: all calculations are still OK, thanks to
bdrv_check_qiov_request().
both rule_check and bdrv_co_pdiscard are 64bit
blklogwrites: pass to blk_loc_writes_co_log which is 64bit
blkreplay, copy-on-read, filter-compress: pass to bdrv_co_pdiscard, OK
copy-before-write: pass to bdrv_co_pdiscard which is 64bit and to
cbw_do_copy_before_write which is 64bit
file-posix: one handler calls raw_account_discard() is 64bit and both
handlers calls raw_do_pdiscard(). Update raw_do_pdiscard, which pass
to RawPosixAIOData::aio_nbytes, which is 64bit (and calls
raw_account_discard())
gluster: somehow, third argument of glfs_discard_async is size_t.
Let's set max_pdiscard accordingly.
iscsi: iscsi_allocmap_set_invalid is 64bit,
!is_byte_request_lun_aligned is 64bit.
list.num is uint32_t. Let's clarify max_pdiscard and
pdiscard_alignment.
mirror_top: pass to bdrv_mirror_top_do_write() which is
64bit
nbd: protocol limitation. max_pdiscard is alredy set strict enough,
keep it as is for now.
nvme: buf.nlb is uint32_t and we do shift. So, add corresponding limits
to nvme_refresh_limits().
preallocate: pass to bdrv_co_pdiscard() which is 64bit.
rbd: pass to qemu_rbd_start_co() which is 64bit.
qcow2: calculations are still OK, thanks to bdrv_check_qiov_request(),
qcow2_cluster_discard() is 64bit.
raw-format: raw_adjust_offset() is 64bit, bdrv_co_pdiscard too.
throttle: pass to bdrv_co_pdiscard() which is 64bit and to
throttle_group_co_io_limits_intercept() which is 64bit as well.
test-block-iothread: bytes argument is unused
Great! Now all drivers are prepared to handle 64bit discard requests,
or else have explicit max_pdiscard limits.
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20210903102807.27127-11-vsementsov@virtuozzo.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Eric Blake <eblake@redhat.com>
We are generally moving to int64_t for both offset and bytes parameters
on all io paths.
Main motivation is realization of 64-bit write_zeroes operation for
fast zeroing large disk chunks, up to the whole disk.
We chose signed type, to be consistent with off_t (which is signed) and
with possibility for signed return type (where negative value means
error).
So, convert driver write_zeroes handlers bytes parameter to int64_t.
The only caller of all updated function is bdrv_co_do_pwrite_zeroes().
bdrv_co_do_pwrite_zeroes() itself is of course OK with widening of
callee parameter type. Also, bdrv_co_do_pwrite_zeroes()'s
max_write_zeroes is limited to INT_MAX. So, updated functions all are
safe, they will not get "bytes" larger than before.
Still, let's look through all updated functions, and add assertions to
the ones which are actually unprepared to values larger than INT_MAX.
For these drivers also set explicit max_pwrite_zeroes limit.
Let's go:
blkdebug: calculations can't overflow, thanks to
bdrv_check_qiov_request() in generic layer. rule_check() and
bdrv_co_pwrite_zeroes() both have 64bit argument.
blklogwrites: pass to blk_log_writes_co_log() with 64bit argument.
blkreplay, copy-on-read, filter-compress: pass to
bdrv_co_pwrite_zeroes() which is OK
copy-before-write: Calls cbw_do_copy_before_write() and
bdrv_co_pwrite_zeroes, both have 64bit argument.
file-posix: both handler calls raw_do_pwrite_zeroes, which is updated.
In raw_do_pwrite_zeroes() calculations are OK due to
bdrv_check_qiov_request(), bytes go to RawPosixAIOData::aio_nbytes
which is uint64_t.
Check also where that uint64_t gets handed:
handle_aiocb_write_zeroes_block() passes a uint64_t[2] to
ioctl(BLKZEROOUT), handle_aiocb_write_zeroes() calls do_fallocate()
which takes off_t (and we compile to always have 64-bit off_t), as
does handle_aiocb_write_zeroes_unmap. All look safe.
gluster: bytes go to GlusterAIOCB::size which is int64_t and to
glfs_zerofill_async works with off_t.
iscsi: Aha, here we deal with iscsi_writesame16_task() that has
uint32_t num_blocks argument and iscsi_writesame16_task() has
uint16_t argument. Make comments, add assertions and clarify
max_pwrite_zeroes calculation.
iscsi_allocmap_() functions already has int64_t argument
is_byte_request_lun_aligned is simple to update, do it.
mirror_top: pass to bdrv_mirror_top_do_write which has uint64_t
argument
nbd: Aha, here we have protocol limitation, and NBDRequest::len is
uint32_t. max_pwrite_zeroes is cleanly set to 32bit value, so we are
OK for now.
nvme: Again, protocol limitation. And no inherent limit for
write-zeroes at all. But from code that calculates cdw12 it's obvious
that we do have limit and alignment. Let's clarify it. Also,
obviously the code is not prepared to handle bytes=0. Let's handle
this case too.
trace events already 64bit
preallocate: pass to handle_write() and bdrv_co_pwrite_zeroes(), both
64bit.
rbd: pass to qemu_rbd_start_co() which is 64bit.
qcow2: offset + bytes and alignment still works good (thanks to
bdrv_check_qiov_request()), so tail calculation is OK
qcow2_subcluster_zeroize() has 64bit argument, should be OK
trace events updated
qed: qed_co_request wants int nb_sectors. Also in code we have size_t
used for request length which may be 32bit. So, let's just keep
INT_MAX as a limit (aligning it down to pwrite_zeroes_alignment) and
don't care.
raw-format: Is OK. raw_adjust_offset and bdrv_co_pwrite_zeroes are both
64bit.
throttle: Both throttle_group_co_io_limits_intercept() and
bdrv_co_pwrite_zeroes() are 64bit.
vmdk: pass to vmdk_pwritev which is 64bit
quorum: pass to quorum_co_pwritev() which is 64bit
Hooray!
At this point all block drivers are prepared to support 64bit
write-zero requests, or have explicitly set max_pwrite_zeroes.
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20210903102807.27127-8-vsementsov@virtuozzo.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
[eblake: use <= rather than < in assertions relying on max_pwrite_zeroes]
Signed-off-by: Eric Blake <eblake@redhat.com>
We are generally moving to int64_t for both offset and bytes parameters
on all io paths.
Main motivation is realization of 64-bit write_zeroes operation for
fast zeroing large disk chunks, up to the whole disk.
We chose signed type, to be consistent with off_t (which is signed) and
with possibility for signed return type (where negative value means
error).
So, convert driver copy_range handlers parameters which are already
64bit to signed type.
Now let's consider all callers. Simple
git grep '\->bdrv_co_copy_range'
shows the only caller:
bdrv_co_copy_range_internal(), which does bdrv_check_request32(),
so everything is OK.
Still, the functions may be called directly, not only by drv->...
Let's check:
git grep '\.bdrv_co_copy_range_\(from\|to\)\s*=' | \
awk '{print $4}' | sed 's/,//' | sed 's/&//' | sort | uniq | \
while read func; do git grep "$func(" | \
grep -v "$func(BlockDriverState"; done
shows no more callers. So, we are done.
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-Id: <20210903102807.27127-6-vsementsov@virtuozzo.com>
Signed-off-by: Eric Blake <eblake@redhat.com>
We are generally moving to int64_t for both offset and bytes parameters
on all io paths.
Main motivation is realization of 64-bit write_zeroes operation for
fast zeroing large disk chunks, up to the whole disk.
We chose signed type, to be consistent with off_t (which is signed) and
with possibility for signed return type (where negative value means
error).
So, convert driver write handlers parameters which are already 64bit to
signed type.
While being here, convert also flags parameter to be BdrvRequestFlags.
Now let's consider all callers. Simple
git grep '\->bdrv_\(aio\|co\)_pwritev\(_part\)\?'
shows that's there three callers of driver function:
bdrv_driver_pwritev() and bdrv_driver_pwritev_compressed() in
block/io.c, both pass int64_t, checked by bdrv_check_qiov_request() to
be non-negative.
qcow2_save_vmstate() does bdrv_check_qiov_request().
Still, the functions may be called directly, not only by drv->...
Let's check:
git grep '\.bdrv_\(aio\|co\)_pwritev\(_part\)\?\s*=' | \
awk '{print $4}' | sed 's/,//' | sed 's/&//' | sort | uniq | \
while read func; do git grep "$func(" | \
grep -v "$func(BlockDriverState"; done
shows several callers:
qcow2:
qcow2_co_truncate() write at most up to @offset, which is checked in
generic qcow2_co_truncate() by bdrv_check_request().
qcow2_co_pwritev_compressed_task() pass the request (or part of the
request) that already went through normal write path, so it should
be OK
qcow:
qcow_co_pwritev_compressed() pass int64_t, it's updated by this patch
quorum:
quorum_co_pwrite_zeroes() pass int64_t and int - OK
throttle:
throttle_co_pwritev_compressed() pass int64_t, it's updated by this
patch
vmdk:
vmdk_co_pwritev_compressed() pass int64_t, it's updated by this
patch
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20210903102807.27127-5-vsementsov@virtuozzo.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Eric Blake <eblake@redhat.com>
We are generally moving to int64_t for both offset and bytes parameters
on all io paths.
Main motivation is realization of 64-bit write_zeroes operation for
fast zeroing large disk chunks, up to the whole disk.
We chose signed type, to be consistent with off_t (which is signed) and
with possibility for signed return type (where negative value means
error).
So, convert driver read handlers parameters which are already 64bit to
signed type.
While being here, convert also flags parameter to be BdrvRequestFlags.
Now let's consider all callers. Simple
git grep '\->bdrv_\(aio\|co\)_preadv\(_part\)\?'
shows that's there three callers of driver function:
bdrv_driver_preadv() in block/io.c, passes int64_t, checked by
bdrv_check_qiov_request() to be non-negative.
qcow2_load_vmstate() does bdrv_check_qiov_request().
do_perform_cow_read() has uint64_t argument. And a lot of things in
qcow2 driver are uint64_t, so converting it is big job. But we must
not work with requests that don't satisfy bdrv_check_qiov_request(),
so let's just assert it here.
Still, the functions may be called directly, not only by drv->...
Let's check:
git grep '\.bdrv_\(aio\|co\)_preadv\(_part\)\?\s*=' | \
awk '{print $4}' | sed 's/,//' | sed 's/&//' | sort | uniq | \
while read func; do git grep "$func(" | \
grep -v "$func(BlockDriverState"; done
The only one such caller:
QEMUIOVector qiov = QEMU_IOVEC_INIT_BUF(qiov, &data, 1);
...
ret = bdrv_replace_test_co_preadv(bs, 0, 1, &qiov, 0);
in tests/unit/test-bdrv-drain.c, and it's OK obviously.
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20210903102807.27127-4-vsementsov@virtuozzo.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
[eblake: fix typos]
Signed-off-by: Eric Blake <eblake@redhat.com>
bdrv_co_block_status() does it for us, we do not need to do it here.
The advantage of not capping *pnum is that bdrv_co_block_status() can
cache larger data regions than requested by its caller.
Signed-off-by: Hanna Reitz <hreitz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20210812084148.14458-5-hreitz@redhat.com>
Similar to other handle_aiocb_* functions, handle_aiocb_ioctl needs to cater
for the possibility that ioctl is interrupted by a signal. Otherwise, the
I/O is incorrectly reported as a failure to the guest.
Reported-by: Gordon Watson <gwatson@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
iOS hosts do not have these defined so we fallback to the
default behaviour.
Co-authored-by: Warner Losh <imp@bsdimp.com>
Signed-off-by: Joelle van Dyne <j@getutm.app>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Try all the possible ioctls for disk size as long as they are
supported, to keep the #if ladder simple.
Extracted and cleaned up from a patch by Joelle van Dyne and
Warner Losh.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
On Darwin (iOS), there are no system level APIs for directly accessing
host block devices. We detect this at configure time.
Signed-off-by: Joelle van Dyne <j@getutm.app>
Message-Id: <20210315180341.31638-2-j@getutm.app>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
bs->sg is only true for character devices, but block devices can also
be used with scsi-block and scsi-generic. Unfortunately BLKSECTGET
returns bytes in an int for /dev/sgN devices, and sectors in a short
for block devices, so account for that in the code.
The maximum transfer also need not be a power of 2 (for example I have
seen disks with 1280 KiB maximum transfer) so there's no need to pass
the result through pow2floor.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
For block host devices, I/O can happen through either the kernel file
descriptor I/O system calls (preadv/pwritev, io_submit, io_uring)
or the SCSI passthrough ioctl SG_IO.
In the latter case, the size of each transfer can be limited by the
HBA, while for file descriptor I/O the kernel is able to split and
merge I/O in smaller pieces as needed. Applying the HBA limits to
file descriptor I/O results in more system calls and suboptimal
performance, so this patch splits the max_transfer limit in two:
max_transfer remains valid and is used in general, while max_hw_transfer
is limited to the maximum hardware size. max_hw_transfer can then be
included by the scsi-generic driver in the block limits page, to ensure
that the stricter hardware limit is used.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
I/O to a disk via read/write is not limited by the number of segments allowed
by the host adapter; the kernel can split requests if needed, and the limit
imposed by the host adapter can be very low (256k or so) to avoid that SG_IO
returns EINVAL if memory is heavily fragmented.
Since this value is only interesting for SG_IO-based I/O, do not include
it in the max_transfer and only take it into account when patching the
block limits VPD page in the scsi-generic device.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Even though it was only called for devices that have bs->sg set (which
must be character devices), sg_get_max_segments looked at /sys/dev/block
which only works for block devices.
On Linux the sg driver has its own way to provide the maximum number of
iovecs in a scatter/gather list, so add support for it. The block device
path is kept because it will be reinstated in the next patches.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Connor Kuehl <ckuehl@redhat.com>
Reviewed-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
A flush failure is a critical failure scenario for some operations.
For example, it will prevent migration from completing, as it will
make vm_stop() report an error. Thus it is important to have a
trace point present for debugging.
Reviewed-by: Connor Kuehl <ckuehl@redhat.com>
Reviewed-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
When fdatasync() fails on a file backend we set a flag that
short-circuits any future attempts to call fdatasync(). The
first failure returns the true errno, but the later short-
circuited calls return a generic EIO. The latter is unhelpful
because fdatasync() can return a variety of errnos, including
EACCESS.
Reviewed-by: Connor Kuehl <ckuehl@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
If fallocate(... FALLOC_FL_ZERO_RANGE ...) returns EINVAL, it's likely
an indication that the file system is buggy and does not implement
unaligned accesses right. We still might be lucky with the other
fallback fallocate() calls later in this function, though, so we should
not return immediately and try the others first.
Since FALLOC_FL_ZERO_RANGE could also return EINVAL if the file descriptor
is not a regular file, we ignore this filesystem bug silently, without
printing an error message for the user.
Signed-off-by: Thomas Huth <thuth@redhat.com>
Message-Id: <20210527172020.847617-3-thuth@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
A customer reported that running
qemu-img convert -t none -O qcow2 -f qcow2 input.qcow2 output.qcow2
fails for them with the following error message when the images are
stored on a GPFS file system :
qemu-img: error while writing sector 0: Invalid argument
After analyzing the strace output, it seems like the problem is in
handle_aiocb_write_zeroes(): The call to fallocate(FALLOC_FL_PUNCH_HOLE)
returns EINVAL, which can apparently happen if the file system has
a different idea of the granularity of the operation. It's arguably
a bug in GPFS, since the PUNCH_HOLE mode should not result in EINVAL
according to the man-page of fallocate(), but the file system is out
there in production and so we have to deal with it. In commit 294682cc3a
("block: workaround for unaligned byte range in fallocate()") we also
already applied the a work-around for the same problem to the earlier
fallocate(FALLOC_FL_ZERO_RANGE) call, so do it now similar with the
PUNCH_HOLE call. But instead of silently catching and returning
-ENOTSUP (which causes the caller to fall back to writing zeroes),
let's rather inform the user once about the buggy file system and
try the other fallback instead.
Signed-off-by: Thomas Huth <thuth@redhat.com>
Message-Id: <20210527172020.847617-2-thuth@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Move bdrv_reopen_multiple to new paradigm of permission update:
first update graph relations, then do refresh the permissions.
We have to modify reopen process in file-posix driver: with new scheme
we don't have prepared permissions in raw_reopen_prepare(), so we
should reconfigure fd in raw_check_perm(). Still this seems more native
and simple anyway.
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20210428151804.439460-31-vsementsov@virtuozzo.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
The 'host_device' and 'host_cdrom' drivers must be used instead.
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
It's better to pass &error_abort than just assert that result is 0: on
crash, we'll immediately see the reason in the backtrace.
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20201211183934.169161-2-vsementsov@virtuozzo.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
[eblake: fix iotest 206 fallout]
Signed-off-by: Eric Blake <eblake@redhat.com>
When a call to fcntl(2) for the purpose of adding file locks fails
with an error other than EAGAIN or EACCES, report the error returned
by fcntl.
EAGAIN or EACCES are elided as they are considered to be common
failures, indicating that a conflicting lock is held by another
process.
No errors are elided when removing file locks.
Signed-off-by: David Edmondson <david.edmondson@oracle.com>
Message-Id: <20210113164447.2545785-1-david.edmondson@oracle.com>
Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Signed-off-by: Max Reitz <mreitz@redhat.com>
We'll need a separate function, which will only "mark" request
serialising with specified align but not wait for conflicting
requests. So, it will be like old bdrv_mark_request_serialising(),
before merging bdrv_wait_serialising_requests_locked() into it.
To reduce the possible mess, let's do the following:
Public function that does both marking and waiting will be called
bdrv_make_request_serialising, and private function which will only
"mark" will be called tracked_request_set_serialising().
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-Id: <20201021145859.11201-6-vsementsov@virtuozzo.com>
Signed-off-by: Max Reitz <mreitz@redhat.com>
local_err is not initialized to NULL, it will cause a assert error as below:
qemu/util/error.c:59: error_setv: Assertion `*errp == NULL' failed.
Fixes: c644751069
Reported-by: Euler Robot <euler.robot@huawei.com>
Signed-off-by: Pan Nengyuan <pannengyuan@huawei.com>
Reviewed-by: Stefano Garzarella <sgarzare@redhat.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Li Qiang <liq3ea@gmail.com>
Signed-off-by: Chen Qun <kuhn.chenqun@huawei.com>
Message-Id: <20201023061218.2080844-8-kuhn.chenqun@huawei.com>
Signed-off-by: Laurent Vivier <laurent@vivier.eu>
We are going to modify block layer to work with 64bit requests. And
first step is moving to int64_t type for both offset and bytes
arguments in all block request related functions.
It's mostly safe (when widening signed or unsigned int to int64_t), but
switching from uint64_t is questionable.
So, let's first establish the set of requests we want to work with.
First signed int64_t should be enough, as off_t is signed anyway. Then,
obviously offset + bytes should not overflow.
And most interesting: (offset + bytes) being aligned up should not
overflow as well. Aligned to what alignment? First thing that comes in
mind is bs->bl.request_alignment, as we align up request to this
alignment. But there is another thing: look at
bdrv_mark_request_serialising(). It aligns request up to some given
alignment. And this parameter may be bdrv_get_cluster_size(), which is
often a lot greater than bs->bl.request_alignment.
Note also, that bdrv_mark_request_serialising() uses signed int64_t for
calculations. So, actually, we already depend on some restrictions.
Happily, bdrv_get_cluster_size() returns int and
bs->bl.request_alignment has 32bit unsigned type, but defined to be a
power of 2 less than INT_MAX. So, we may establish, that INT_MAX is
absolute maximum for any kind of alignment that may occur with the
request.
Note, that bdrv_get_cluster_size() is not documented to return power
of 2, still bdrv_mark_request_serialising() behaves like it is.
Also, backup uses bdi.cluster_size and is not prepared to it not being
power of 2.
So, let's establish that Qemu supports only power-of-2 clusters and
alignments.
So, alignment can't be greater than 2^30.
Finally to be safe with calculations, to not calculate different
maximums for different nodes (depending on cluster size and
request_alignment), let's simply set QEMU_ALIGN_DOWN(INT64_MAX, 2^30)
as absolute maximum bytes length for Qemu. Actually, it's not much less
than INT64_MAX.
OK, then, let's apply it to block/io.
Let's consider all block/io entry points of offset/bytes:
4 bytes/offset interface functions: bdrv_co_preadv_part(),
bdrv_co_pwritev_part(), bdrv_co_copy_range_internal() and
bdrv_co_pdiscard() and we check them all with bdrv_check_request().
We also have one entry point with only offset: bdrv_co_truncate().
Check the offset.
And one public structure: BdrvTrackedRequest. Happily, it has only
three external users:
file-posix.c: adopted by this patch
write-threshold.c: only read fields
test-write-threshold.c: sets obviously small constant values
Better is to make the structure private and add corresponding
interfaces.. Still it's not obvious what kind of interface is needed
for file-posix.c. Let's keep it public but add corresponding
assertions.
After this patch we'll convert functions in block/io.c to int64_t bytes
and offset parameters. We can assume that offset/bytes pair always
satisfy new restrictions, and make
corresponding assertions where needed. If we reach some offset/bytes
point in block/io.c missing bdrv_check_request() it is considered a
bug. As well, if block/io.c modifies a offset/bytes request, expanding
it more then aligning up to request_alignment, it's a bug too.
For all io requests except for discard we keep for now old restriction
of 32bit request length.
iotest 206 output error message changed, as now test disk size is
larger than new limit. Add one more test case with new maximum disk
size to cover too-big-L1 case.
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20201203222713.13507-5-vsementsov@virtuozzo.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
We should not set overlap_bytes:
1. Don't worry: it is calculated by bdrv_mark_request_serialising() and
will be equal to or greater than bytes anyway.
2. If the request was already aligned up to some greater alignment,
than we may break things: we reduce overlap_bytes, and further
bdrv_mark_request_serialising() may not help, as it will not restore
old bigger alignment.
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20201203222713.13507-2-vsementsov@virtuozzo.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
The scenario is that when accessing a volume on an NFS filesystem
without supporting the file lock, Qemu will complain "Failed to lock
byte 100", even when setting the file.locking = off.
We should do file lock related operations only when the file.locking is
enabled, otherwise, the syscall of 'fcntl' will return non-zero.
Signed-off-by: Li Feng <fengli@smartx.com>
Message-Id: <1607341446-85506-1-git-send-email-fengli@smartx.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
On Linux, fallocate(fd, FALLOC_FL_PUNCH_HOLE) when it is used on a block device,
without O_DIRECT can return -EBUSY if it races with another write to the same page.
Since this is rare and discard is not a critical operation, ignore this error
Signed-off-by: Maxim Levitsky <mlevitsk@redhat.com>
Message-Id: <20201111153913.41840-2-mlevitsk@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Currently at startup if using cache=none on a filesystem lacking
O_DIRECT such as tmpfs, at startup QEMU prints
qemu-system-x86_64: -drive file=/tmp/foo.img,cache=none: file system may not support O_DIRECT
qemu-system-x86_64: -drive file=/tmp/foo.img,cache=none: Could not open '/tmp/foo.img': Invalid argument
while at QMP level the hint is missing, so QEMU reports just
"error": {
"class": "GenericError",
"desc": "Could not open '/tmp/foo.img': Invalid argument"
}
which is close to useless for the end user trying to figure out what
they did wrong.
With this change at startup QEMU prints
qemu-system-x86_64: -drive file=/tmp/foo.img,cache=none: Unable to open '/tmp/foo.img': filesystem does not support O_DIRECT
while at the QMP level QEMU reports a massively more informative
"error": {
"class": "GenericError",
"desc": "Unable to open '/tmp/foo.img': filesystem does not support O_DIRECT"
}
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
We want to introduce a new version of qemu_open() that uses an Error
object for reporting problems and make this it the preferred interface.
Rename the existing method to release the namespace for the new impl.
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
The `detect-zeroes=unmap` option may issue unaligned
`FALLOC_FL_PUNCH_HOLE` requests, raw block devices can (and will) return
`EINVAL`, qemu should then write the zeroes to the blockdev instead of
issuing an `IO_ERROR`.
The problem can be reprodced like this:
$ qemu-io -c 'write -P 0 42 1234' --image-opts driver=host_device,filename=/dev/loop0,detect-zeroes=unmap
write failed: Invalid argument
Signed-off-by: Antoine Damhet <antoine.damhet@blade-group.com>
Message-Id: <20200717135603.51180-1-antoine.damhet@blade-group.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20200717105426.51134-4-kwolf@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
For Linux block devices, being able to open the device read-write
doesn't necessarily mean that the device is actually writable (one
example is a read-only LV, as you get with lvchange -pr <device>). We
have check_hdev_writable() to check this condition and fail opening the
image read-write if it's not actually writable.
However, this check doesn't take auto-read-only into account, but
results in a hard failure instead of downgrading to read-only where
possible.
Fix this and do the writable check not based on BDRV_O_RDWR, but only
when this actually results in opening the file read-write. A second
check is inserted in raw_reconfigure_getfd() to have the same check when
dynamic auto-read-only upgrades an image file from read-only to
read-write.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20200717105426.51134-3-kwolf@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
We'll need to call it in raw_open_common(), so move the function to
avoid a forward declaration.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20200717105426.51134-2-kwolf@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Since commit a6b257a08e ('file-posix: Handle undetectable alignment'),
we assume that if we open a file with O_DIRECT and alignment probing
returns 1, we just couldn't find out the real alignment requirement
because some filesystems make the requirement only for allocated blocks.
In this case, a safe default of 4k is used.
This is too strict for NFS, which does actually allow byte-aligned
requests even with O_DIRECT. Because we can't distinguish both cases
with generic code, let's just look at the file system magic and disable
s->needs_alignment for NFS. This way, O_DIRECT can still be used on NFS
for images that are not aligned to 4k.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-Id: <20200716142601.111237-3-kwolf@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Especially when O_DIRECT is used with image files so that the page cache
indirection can't cause a merge of allocating requests, the file will
fragment on the file system layer, with a potentially very small
fragment size (this depends on the requests the guest sent).
On Linux, fragmentation can be reduced by setting an extent size hint
when creating the file (at least on XFS, it can't be set any more after
the first extent has been allocated), basically giving raw files a
"cluster size" for allocation.
This adds a create option to set the extent size hint, and changes the
default from not setting a hint to setting it to 1 MB. The main reason
why qcow2 defaults to smaller cluster sizes is that COW becomes more
expensive, which is not an issue with raw files, so we can choose a
larger size. The tradeoff here is only potentially wasted disk space.
For qcow2 (or other image formats) over file-posix, the advantage should
even be greater because they grow sequentially without leaving holes, so
there won't be wasted space. Setting even larger extent size hints for
such images may make sense. This can be done with the new option, but
let's keep the default conservative for now.
The effect is very visible with a test that intentionally creates a
badly fragmented file with qemu-img bench (the time difference while
creating the file is already remarkable) and then looks at the number of
extents and the time a simple "qemu-img map" takes.
Without an extent size hint:
$ ./qemu-img create -f raw -o extent_size_hint=0 ~/tmp/test.raw 10G
Formatting '/home/kwolf/tmp/test.raw', fmt=raw size=10737418240 extent_size_hint=0
$ ./qemu-img bench -f raw -t none -n -w ~/tmp/test.raw -c 1000000 -S 8192 -o 0
Sending 1000000 write requests, 4096 bytes each, 64 in parallel (starting at offset 0, step size 8192)
Run completed in 25.848 seconds.
$ ./qemu-img bench -f raw -t none -n -w ~/tmp/test.raw -c 1000000 -S 8192 -o 4096
Sending 1000000 write requests, 4096 bytes each, 64 in parallel (starting at offset 4096, step size 8192)
Run completed in 19.616 seconds.
$ filefrag ~/tmp/test.raw
/home/kwolf/tmp/test.raw: 2000000 extents found
$ time ./qemu-img map ~/tmp/test.raw
Offset Length Mapped to File
0 0x1e8480000 0 /home/kwolf/tmp/test.raw
real 0m1,279s
user 0m0,043s
sys 0m1,226s
With the new default extent size hint of 1 MB:
$ ./qemu-img create -f raw -o extent_size_hint=1M ~/tmp/test.raw 10G
Formatting '/home/kwolf/tmp/test.raw', fmt=raw size=10737418240 extent_size_hint=1048576
$ ./qemu-img bench -f raw -t none -n -w ~/tmp/test.raw -c 1000000 -S 8192 -o 0
Sending 1000000 write requests, 4096 bytes each, 64 in parallel (starting at offset 0, step size 8192)
Run completed in 11.833 seconds.
$ ./qemu-img bench -f raw -t none -n -w ~/tmp/test.raw -c 1000000 -S 8192 -o 4096
Sending 1000000 write requests, 4096 bytes each, 64 in parallel (starting at offset 4096, step size 8192)
Run completed in 10.155 seconds.
$ filefrag ~/tmp/test.raw
/home/kwolf/tmp/test.raw: 178 extents found
$ time ./qemu-img map ~/tmp/test.raw
Offset Length Mapped to File
0 0x1e8480000 0 /home/kwolf/tmp/test.raw
real 0m0,061s
user 0m0,040s
sys 0m0,014s
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20200707142329.48303-1-kwolf@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
When all we do with an Error we receive into a local variable is
propagating to somewhere else, we can just as well receive it there
right away. The previous commit did that with a Coccinelle script I
consider fairly trustworthy. This commit uses the same script with
the matching of return taken out, i.e. we convert
if (!foo(..., &err)) {
...
error_propagate(errp, err);
...
}
to
if (!foo(..., errp)) {
...
...
}
This is unsound: @err could still be read between afterwards. I don't
know how to express "no read of @err without an intervening write" in
Coccinelle. Instead, I manually double-checked for uses of @err.
Suboptimal line breaks tweaked manually. qdev_realize() simplified
further to placate scripts/checkpatch.pl.
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-Id: <20200707160613.848843-36-armbru@redhat.com>
When all we do with an Error we receive into a local variable is
propagating to somewhere else, we can just as well receive it there
right away. Convert
if (!foo(..., &err)) {
...
error_propagate(errp, err);
...
return ...
}
to
if (!foo(..., errp)) {
...
...
return ...
}
where nothing else needs @err. Coccinelle script:
@rule1 forall@
identifier fun, err, errp, lbl;
expression list args, args2;
binary operator op;
constant c1, c2;
symbol false;
@@
if (
(
- fun(args, &err, args2)
+ fun(args, errp, args2)
|
- !fun(args, &err, args2)
+ !fun(args, errp, args2)
|
- fun(args, &err, args2) op c1
+ fun(args, errp, args2) op c1
)
)
{
... when != err
when != lbl:
when strict
- error_propagate(errp, err);
... when != err
(
return;
|
return c2;
|
return false;
)
}
@rule2 forall@
identifier fun, err, errp, lbl;
expression list args, args2;
expression var;
binary operator op;
constant c1, c2;
symbol false;
@@
- var = fun(args, &err, args2);
+ var = fun(args, errp, args2);
... when != err
if (
(
var
|
!var
|
var op c1
)
)
{
... when != err
when != lbl:
when strict
- error_propagate(errp, err);
... when != err
(
return;
|
return c2;
|
return false;
|
return var;
)
}
@depends on rule1 || rule2@
identifier err;
@@
- Error *err = NULL;
... when != err
Not exactly elegant, I'm afraid.
The "when != lbl:" is necessary to avoid transforming
if (fun(args, &err)) {
goto out
}
...
out:
error_propagate(errp, err);
even though other paths to label out still need the error_propagate().
For an actual example, see sclp_realize().
Without the "when strict", Coccinelle transforms vfio_msix_setup(),
incorrectly. I don't know what exactly "when strict" does, only that
it helps here.
The match of return is narrower than what I want, but I can't figure
out how to express "return where the operand doesn't use @err". For
an example where it's too narrow, see vfio_intx_enable().
Silently fails to convert hw/arm/armsse.c, because Coccinelle gets
confused by ARMSSE being used both as typedef and function-like macro
there. Converted manually.
Line breaks tidied up manually. One nested declaration of @local_err
deleted manually. Preexisting unwanted blank line dropped in
hw/riscv/sifive_e.c.
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-Id: <20200707160613.848843-35-armbru@redhat.com>
