As we have attempted before
(https://lists.gnu.org/archive/html/qemu-devel/2019-01/msg06451.html,
"file-posix: Cache lseek result for data regions";
https://lists.nongnu.org/archive/html/qemu-block/2021-02/msg00934.html,
"file-posix: Cache next hole"), this patch seeks to reduce the number of
SEEK_DATA/HOLE operations the file-posix driver has to perform. The
main difference is that this time it is implemented as part of the
general block layer code.
The problem we face is that on some filesystems or in some
circumstances, SEEK_DATA/HOLE is unreasonably slow. Given the
implementation is outside of qemu, there is little we can do about its
performance.
We have already introduced the want_zero parameter to
bdrv_co_block_status() to reduce the number of SEEK_DATA/HOLE calls
unless we really want zero information; but sometimes we do want that
information, because for files that consist largely of zero areas,
special-casing those areas can give large performance boosts. So the
real problem is with files that consist largely of data, so that
inquiring the block status does not gain us much performance, but where
such an inquiry itself takes a lot of time.
To address this, we want to cache data regions. Most of the time, when
bad performance is reported, it is in places where the image is iterated
over from start to end (qemu-img convert or the mirror job), so a simple
yet effective solution is to cache only the current data region.
(Note that only caching data regions but not zero regions means that
returning false information from the cache is not catastrophic: Treating
zeroes as data is fine. While we try to invalidate the cache on zero
writes and discards, such incongruences may still occur when there are
other processes writing to the image.)
We only use the cache for nodes without children (i.e. protocol nodes),
because that is where the problem is: Drivers that rely on block-status
implementations outside of qemu (e.g. SEEK_DATA/HOLE).
Resolves: https://gitlab.com/qemu-project/qemu/-/issues/307
Signed-off-by: Hanna Reitz <hreitz@redhat.com>
Message-Id: <20210812084148.14458-3-hreitz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
[hreitz: Added `local_file == bs` assertion, as suggested by Vladimir]
Signed-off-by: Hanna Reitz <hreitz@redhat.com>
When bdrv_pad_request() fails in bdrv_co_preadv_part(), bs->in_flight
has been increased, but is never decreased again. This leads to a hang
when trying to drain the block node.
This bug was observed with Windows guests which issue a request that
fully uses IOV_MAX during installation, so that when padding is
necessary (O_DIRECT with a 4k sector size block device on the host),
adding another entry causes failure.
Call bdrv_dec_in_flight() to fix this. There is a larger problem to
solve here because this request shouldn't even fail, but Windows doesn't
seem to care and with this minimal fix the installation succeeds. So
given that we're already in freeze, let's take this minimal fix for 6.1.
Fixes: 98ca45494f
Fixes: https://bugzilla.redhat.com/show_bug.cgi?id=1972079
Reported-by: Qing Wang <qinwang@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20210727154923.91067-1-kwolf@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
No need to start a tracked request that will always fail. The choice
to check read-only after bdrv_inc_in_flight() predates 1bc5f09f2e
(block: Use tracked request for truncate), but waiting for serializing
requests can make the effect more noticeable.
Signed-off-by: Eric Blake <eblake@redhat.com>
Message-Id: <20210609163034.997943-1-eblake@redhat.com>
Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Signed-off-by: Kevin Wolf <kwolf@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>
It's better to use accessor function instead of bs->read_only directly.
In some places use bdrv_is_writable() instead of
checking both BDRV_O_RDWR set and BDRV_O_INACTIVE not set.
In bdrv_open_common() it's a bit strange to add one more variable, but
we are going to drop bs->read_only in the next patch, so new ro local
variable substitutes it here.
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20210527154056.70294-2-vsementsov@virtuozzo.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
They are unused now.
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-Id: <20210506090621.11848-3-vsementsov@virtuozzo.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Max Reitz <mreitz@redhat.com>
write-notifiers are used only for write-threshold. New code for such
purpose should create filters.
Let's better special-case write-threshold and drop write notifiers at
all. (Actually, write-threshold is special-cased anyway, as the only
user of write-notifiers)
So, create a new direct interface for bdrv_co_write_req_prepare() and
drop all write-notifier related logic from write-threshold.c.
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-Id: <20210506090621.11848-2-vsementsov@virtuozzo.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
[mreitz: Adjusted comment as per Eric's suggestion]
Signed-off-by: Max Reitz <mreitz@redhat.com>
To be used in further commit.
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20210428151804.439460-28-vsementsov@virtuozzo.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
It will be used to stop retrying NBD requests on mirror cancel.
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-Id: <20210205163720.887197-2-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 now copy_range parameters which are already 64bit to signed
type.
It's safe as we don't work with requests overflowing BDRV_MAX_LENGTH
(which is less than INT64_MAX), and do check the requests in
bdrv_co_copy_range_internal() (by bdrv_check_request32(), which calls
bdrv_check_request()).
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20201211183934.169161-17-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).
Now, since bdrv_co_preadv_part() and bdrv_co_pwritev_part() have been
updated, update all their wrappers.
For all of them type of 'bytes' is widening, so callers are safe. We
have update request_fn in blkverify.c simultaneously. Still it's just a
pointer to one of bdrv_co_pwritev() or bdrv_co_preadv(), and type is
widening for callers of the request_fn anyway.
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20201211183934.169161-16-vsementsov@virtuozzo.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
[eblake: grammar tweak]
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, prepare bdrv_co_preadv_part() and bdrv_co_pwritev_part() and their
remaining dependencies now.
bdrv_pad_request() is updated simultaneously, as pointer to bytes passed
to it both from bdrv_co_pwritev_part() and bdrv_co_preadv_part().
So, all callers of bdrv_pad_request() are updated to pass 64bit bytes.
bdrv_pad_request() is already good for 64bit requests, add
corresponding assertion.
Look at bdrv_co_preadv_part() and bdrv_co_pwritev_part().
Type is widening, so callers are safe. Let's look inside the functions.
In bdrv_co_preadv_part() and bdrv_aligned_pwritev() we only pass bytes
to other already int64_t interfaces (and some obviously safe
calculations), it's OK.
In bdrv_co_do_zero_pwritev() aligned_bytes may become large now, still
it's passed to bdrv_aligned_pwritev which supports int64_t bytes.
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20201211183934.169161-15-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, prepare bdrv_aligned_preadv() now.
Make the bytes variable in bdrv_padding_rmw_read() int64_t, as it is
only used for pass-through to bdrv_aligned_preadv().
All bdrv_aligned_preadv() callers are safe as type is widening. Let's
look inside:
- add a new-style assertion that request is good.
- callees bdrv_is_allocated(), bdrv_co_do_copy_on_readv() supports
int64_t bytes
- conversion of bytes_remaining is OK, as we never have requests
overflowing BDRV_MAX_LENGTH
- looping through bytes_remaining is ok, num is updated to int64_t
- for bdrv_driver_preadv we have same limit of max_transfer
- qemu_iovec_memset is OK, as bytes+qiov_offset should not overflow
qiov->size anyway (thanks to bdrv_check_qiov_request())
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20201211183934.169161-14-vsementsov@virtuozzo.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
[eblake: grammar tweak]
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, prepare bdrv_co_do_copy_on_readv() now.
'bytes' type widening, so callers are safe. Look at the function
itself:
bytes, skip_bytes and progress become int64_t.
bdrv_round_to_clusters() is OK, cluster_bytes now may be large.
trace_bdrv_co_do_copy_on_readv() is OK
looping through cluster_bytes is still OK.
pnum is still capped to max_transfer, and to MAX_BOUNCE_BUFFER when we
are going to do COR operation. Therefor calculations in
qemu_iovec_from_buf() and bdrv_driver_preadv() should not change.
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20201211183934.169161-13-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, prepare bdrv_aligned_pwritev() now and convert the dependencies:
bdrv_co_write_req_prepare() and bdrv_co_write_req_finish() to signed
type bytes.
Conversion of bdrv_co_write_req_prepare() and
bdrv_co_write_req_finish() is definitely safe, as all requests in
block/io must not overflow BDRV_MAX_LENGTH. Still add assertions.
For bdrv_aligned_pwritev() 'bytes' type is widened, so callers are
safe. Let's check usage of the parameter inside the function.
Passing to bdrv_co_write_req_prepare() and bdrv_co_write_req_finish()
is OK.
Passing to qemu_iovec_* is OK after new assertion. All other callees
are already updated to int64_t.
Checking alignment is not changed, offset + bytes and qiov_offset +
bytes calculations are safe (thanks to new assertions).
max_transfer is kept to be int for now. It has a default of INT_MAX
here, and some drivers may rely on it. It's to be refactored later.
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20201211183934.169161-12-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, prepare bdrv_co_do_pwrite_zeroes() now.
Callers are safe, as converting int to int64_t is safe. Concentrate on
'bytes' usage in the function (thx to Eric Blake):
compute 'int tail' via % 'int alignment' - safe
fragmentation loop 'int num' - still fragments with a cap on
max_transfer
use of 'num' within the loop
MIN(bytes, max_transfer) as well as %alignment - still works, so
calculations in if (head) {} are safe
clamp size by 'int max_write_zeroes' - safe
drv->bdrv_co_pwrite_zeroes(int) - safe because of clamping
clamp size by 'int max_transfer' - safe
buf allocation is still clamped to max_transfer
qemu_iovec_init_buf(size_t) - safe because of clamping
bdrv_driver_pwritev(uint64_t) - safe
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20201211183934.169161-11-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 wrappers parameters which are already 64bit to
signed type.
Requests in block/io.c must never exceed BDRV_MAX_LENGTH (which is less
than INT64_MAX), which makes the conversion to signed 64bit type safe.
Add corresponding assertions.
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20201211183934.169161-10-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).
All requests in block/io must not overflow BDRV_MAX_LENGTH, all
external users of BdrvTrackedRequest already have corresponding
assertions, so we are safe. Add some assertions still.
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20201211183934.169161-9-vsementsov@virtuozzo.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Eric Blake <eblake@redhat.com>
Operations with qiov add more restrictions on bytes, let's cover it.
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20201211183934.169161-8-vsementsov@virtuozzo.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Eric Blake <eblake@redhat.com>
Make bdrv_pad_request() honest: return error if
qemu_iovec_init_extended() failed.
Update also bdrv_padding_destroy() to clean the structure for safety.
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20201211183934.169161-6-vsementsov@virtuozzo.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Eric Blake <eblake@redhat.com>
Prepare for the following patch when bdrv_pad_request() will be able to
fail. Update the comments.
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20201211183934.169161-5-vsementsov@virtuozzo.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
[eblake: grammar tweak]
Signed-off-by: Eric Blake <eblake@redhat.com>
Calculation of sum may theoretically overflow, so use 64bit type and
add some good assertions.
Use int64_t constantly.
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20201211183934.169161-4-vsementsov@virtuozzo.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
[eblake: tweak assertion order]
Signed-off-by: Eric Blake <eblake@redhat.com>
Actually, we can't extend the io vector in all cases. Handle possible
MAX_IOV and size_t overflows.
For now add assertion to callers (actually they rely on success anyway)
and fix them in the following patch.
Add also some additional good assertions to qemu_iovec_init_slice()
while being here.
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20201211183934.169161-3-vsementsov@virtuozzo.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Eric Blake <eblake@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>
Add the new member supported_read_flags to the BlockDriverState
structure. It will control the flags set for copy-on-read operations.
Make the block generic layer evaluate supported read flags before they
go to a block driver.
Suggested-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Signed-off-by: Andrey Shinkevich <andrey.shinkevich@virtuozzo.com>
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
[vsementsov: use assert instead of abort]
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-Id: <20201216061703.70908-8-vsementsov@virtuozzo.com>
Signed-off-by: Max Reitz <mreitz@redhat.com>
Add flag to make serialising request no wait: if there are conflicting
requests, just return error immediately. It's will be used in upcoming
preallocate filter.
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-Id: <20201021145859.11201-7-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>
bs is linked in req, so no needs to pass it separately. Most of
tracked-requests API doesn't have bs argument. Actually, after this
patch only tracked_request_begin has it, but it's for purpose.
While being here, also add a comment about what "_locked" is.
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-Id: <20201021145859.11201-5-vsementsov@virtuozzo.com>
Signed-off-by: Max Reitz <mreitz@redhat.com>
To be reused in separate.
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-Id: <20201021145859.11201-4-vsementsov@virtuozzo.com>
Signed-off-by: Max Reitz <mreitz@redhat.com>
The comments states, that on misaligned request we should have already
been waiting. But for bdrv_padding_rmw_read, we called
bdrv_mark_request_serialising with align = request_alignment, and now
we serialise with align = cluster_size. So we may have to wait again
with larger alignment.
Note, that the only user of BDRV_REQ_SERIALISING is backup which issues
cluster-aligned requests, so seems the assertion should not fire for
now. But it's wrong anyway.
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Message-Id: <20201021145859.11201-3-vsementsov@virtuozzo.com>
Signed-off-by: Max Reitz <mreitz@redhat.com>
If bdrv_co_yield_to_drain() is called for draining a block node that
runs in a different AioContext, it keeps that AioContext locked while it
yields and schedules a BH in the AioContext to do the actual drain.
As long as executing the BH is the very next thing that the event loop
of the node's AioContext does, this actually happens to work, but when
it tries to execute something else that wants to take the AioContext
lock, it will deadlock. (In the bug report, this other thing is a
virtio-scsi device running virtio_scsi_data_plane_handle_cmd().)
Instead, always drop the AioContext lock across the yield and reacquire
it only when the coroutine is reentered. The BH needs to unconditionally
take the lock for itself now.
This fixes the 'block_resize' QMP command on a block node that runs in
an iothread.
Cc: qemu-stable@nongnu.org
Fixes: eb94b81a94
Fixes: https://bugzilla.redhat.com/show_bug.cgi?id=1903511
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20201203172311.68232-4-kwolf@redhat.com>
Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
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>
Move bdrv_is_inserted() calls into callers.
We are going to make bdrv_check_byte_request() a clean thing.
bdrv_is_inserted() is not about checking the request, it's about
checking the bs. So, it should be separate.
With this patch we probably change error path for some failure
scenarios. But depending on the fact that querying too big request on
empty cdrom (or corrupted qcow2 node with no drv) will result in EIO
and not ENOMEDIUM would be very strange. More over, we are going to
move to 64bit requests, so larger requests will be allowed anyway.
More over, keeping in mind that cdrom is the only driver that has
.bdrv_is_inserted() handler it's strange that we should care so much
about it in generic block layer, intuitively we should just do read and
write, and cdrom driver should return correct errors if it is not
inserted. But it's a work for another series.
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20201203222713.13507-4-vsementsov@virtuozzo.com>
Reviewed-by: Alberto Garcia <berto@igalia.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
This simplifies following commit.
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20201203222713.13507-3-vsementsov@virtuozzo.com>
Reviewed-by: Alberto Garcia <berto@igalia.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
When checking for allocation across a chain, it's already easy to
count the depth within the chain at which the allocation is found.
Instead of throwing that information away, return it to the caller.
Existing callers only cared about allocated/non-allocated, but having
a depth available will be used by NBD in the next patch.
Signed-off-by: Eric Blake <eblake@redhat.com>
Message-Id: <20201027050556.269064-9-eblake@redhat.com>
Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
[eblake: rebase to master]
Signed-off-by: Eric Blake <eblake@redhat.com>
If a BDS gets deleted during blk_drain_all(), it might miss a
call to bdrv_do_drained_end(). This means missing a call to
aio_enable_external() and the AIO context remains disabled for
ever. This can cause a device to become irresponsive and to
disrupt the guest execution, ie. hang, loop forever or worse.
This scenario is quite easy to encounter with virtio-scsi
on POWER when punching multiple blockdev-create QMP commands
while the guest is booting and it is still running the SLOF
firmware. This happens because SLOF disables/re-enables PCI
devices multiple times via IO/MEM/MASTER bits of PCI_COMMAND
register after the initial probe/feature negotiation, as it
tends to work with a single device at a time at various stages
like probing and running block/network bootloaders without
doing a full reset in-between. This naturally generates many
dataplane stops and starts, and thus many drain sections that
can race with blockdev_create_run(). In the end, SLOF bails
out.
It is somehow reproducible on x86 but it requires to generate
articial dataplane start/stop activity with stop/cont QMP
commands. In this case, seabios ends up looping for ever,
waiting for the virtio-scsi device to send a response to
a command it never received.
Add a helper that pairs all previously called bdrv_do_drained_begin()
with a bdrv_do_drained_end() and call it from bdrv_close().
While at it, update the "/bdrv-drain/graph-change/drain_all"
test in test-bdrv-drain so that it can catch the issue.
BugId: https://bugzilla.redhat.com/show_bug.cgi?id=1874441
Signed-off-by: Greg Kurz <groug@kaod.org>
Message-Id: <160346526998.272601.9045392804399803158.stgit@bahia.lan>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Since commit c8bb23cbdb when a write
request results in a new allocation QEMU first tries to see if the
rest of the cluster outside the written area contains only zeroes.
In that case, instead of doing a normal copy-on-write operation and
writing explicit zero buffers to disk, the code zeroes the whole
cluster efficiently using pwrite_zeroes() with BDRV_REQ_NO_FALLBACK.
This improves performance very significantly but it only happens when
we are writing to an area that was completely unallocated before. Zero
clusters (QCOW2_CLUSTER_ZERO_*) are treated like normal clusters and
are therefore slower to allocate.
This happens because the code uses bdrv_is_allocated_above() rather
bdrv_block_status_above(). The former is not as accurate for this
purpose but it is faster. However in the case of qcow2 the underlying
call does already report zero clusters just fine so there is no reason
why we cannot use that information.
After testing 4KB writes on an image that only contains zero clusters
this patch results in almost five times more IOPS.
Signed-off-by: Alberto Garcia <berto@igalia.com>
Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <6d77cab968c501c44d6e1089b9bc91b04170b49e.1603731354.git.berto@igalia.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
If a BlockDriverState supports backing files but has none then any
unallocated area reads back as zeroes.
bdrv_co_block_status() is only reporting this is if want_zero is true,
but this is an inexpensive test and there is no reason not to do it in
all cases.
Suggested-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Signed-off-by: Alberto Garcia <berto@igalia.com>
Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <66fa0914a0e2b727ab6d1b63ca773d7cd29a9a9e.1603731354.git.berto@igalia.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
bdrv_is_allocated_above wrongly handles short backing files: it reports
after-EOF space as UNALLOCATED which is wrong, as on read the data is
generated on the level of short backing file (if all overlays have
unallocated areas at that place).
Reusing bdrv_common_block_status_above fixes the issue and unifies code
path.
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Alberto Garcia <berto@igalia.com>
Message-id: 20200924194003.22080-5-vsementsov@virtuozzo.com
[Fix s/has/have/ as suggested by Eric Blake. Fix s/area/areas/.
--Stefan]
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
We are going to reuse bdrv_common_block_status_above in
bdrv_is_allocated_above. bdrv_is_allocated_above may be called with
include_base == false and still bs == base (for ex. from img_rebase()).
So, support this corner case.
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Alberto Garcia <berto@igalia.com>
Message-id: 20200924194003.22080-4-vsementsov@virtuozzo.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
In order to reuse bdrv_common_block_status_above in
bdrv_is_allocated_above, let's support include_base parameter.
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Alberto Garcia <berto@igalia.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-id: 20200924194003.22080-3-vsementsov@virtuozzo.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
bdrv_co_block_status_above has several design problems with handling
short backing files:
1. With want_zeros=true, it may return ret with BDRV_BLOCK_ZERO but
without BDRV_BLOCK_ALLOCATED flag, when actually short backing file
which produces these after-EOF zeros is inside requested backing
sequence.
2. With want_zero=false, it may return pnum=0 prior to actual EOF,
because of EOF of short backing file.
Fix these things, making logic about short backing files clearer.
With fixed bdrv_block_status_above we also have to improve is_zero in
qcow2 code, otherwise iotest 154 will fail, because with this patch we
stop to merge zeros of different types (produced by fully unallocated
in the whole backing chain regions vs produced by short backing files).
Note also, that this patch leaves for another day the general problem
around block-status: misuse of BDRV_BLOCK_ALLOCATED as is-fs-allocated
vs go-to-backing.
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Alberto Garcia <berto@igalia.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-id: 20200924194003.22080-2-vsementsov@virtuozzo.com
[Fix s/comes/come/ as suggested by Eric Blake
--Stefan]
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Like for read/write in a previous commit, drop extra indirection layer,
generate directly bdrv_readv_vmstate() and bdrv_writev_vmstate().
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-Id: <20200924185414.28642-8-vsementsov@virtuozzo.com>
Now that we are not maintaining boilerplate code for coroutine
wrappers, there is no more sense in keeping the extra indirection layer
of bdrv_prwv(). Let's drop it and instead generate pure bdrv_preadv()
and bdrv_pwritev().
Currently, bdrv_pwritev() and bdrv_preadv() are returning bytes on
success, auto generated functions will instead return zero, as their
_co_ prototype. Still, it's simple to make the conversion safe: the
only external user of bdrv_pwritev() is test-bdrv-drain, and it is
comfortable enough with bdrv_co_pwritev() instead. So prototypes are
moved to local block/coroutines.h. Next, the only internal use is
bdrv_pread() and bdrv_pwrite(), which are modified to return bytes on
success.
Of course, it would be great to convert bdrv_pread() and bdrv_pwrite()
to return 0 on success. But this requires audit (and probably
conversion) of all their users, let's leave it for another day
refactoring.
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-Id: <20200924185414.28642-7-vsementsov@virtuozzo.com>
Use code generation implemented in previous commit to generated
coroutine wrappers in block.c and block/io.c
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-Id: <20200924185414.28642-6-vsementsov@virtuozzo.com>
We are going to keep coroutine-wrappers code (structure-packing
parameters, BDRV_POLL wrapper functions) in separate auto-generated
files. So, we'll need a header with declaration of original _co_
functions, for those which are static now. As well, we'll need
declarations for wrapper functions. Do these declarations now, as a
preparation step.
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-Id: <20200924185414.28642-4-vsementsov@virtuozzo.com>
Most of our coroutine wrappers already follow this convention:
We have 'coroutine_fn bdrv_co_<something>(<normal argument list>)' as
the core function, and a wrapper 'bdrv_<something>(<same argument
list>)' which does parameter packing and calls bdrv_run_co().
The only outsiders are the bdrv_prwv_co and
bdrv_common_block_status_above wrappers. Let's refactor them to behave
as the others, it simplifies further conversion of coroutine wrappers.
This patch adds an indirection layer, but it will be compensated by
a further commit, which will drop bdrv_co_prwv together with the
is_write logic, to keep the read and write paths separate.
Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-Id: <20200924185414.28642-3-vsementsov@virtuozzo.com>
clang's C11 atomic_fetch_*() functions only take a C11 atomic type
pointer argument. QEMU uses direct types (int, etc) and this causes a
compiler error when a QEMU code calls these functions in a source file
that also included <stdatomic.h> via a system header file:
$ CC=clang CXX=clang++ ./configure ... && make
../util/async.c:79:17: error: address argument to atomic operation must be a pointer to _Atomic type ('unsigned int *' invalid)
Avoid using atomic_*() names in QEMU's atomic.h since that namespace is
used by <stdatomic.h>. Prefix QEMU's APIs with 'q' so that atomic.h
and <stdatomic.h> can co-exist. I checked /usr/include on my machine and
searched GitHub for existing "qatomic_" users but there seem to be none.
This patch was generated using:
$ git grep -h -o '\<atomic\(64\)\?_[a-z0-9_]\+' include/qemu/atomic.h | \
sort -u >/tmp/changed_identifiers
$ for identifier in $(</tmp/changed_identifiers); do
sed -i "s%\<$identifier\>%q$identifier%g" \
$(git grep -I -l "\<$identifier\>")
done
I manually fixed line-wrap issues and misaligned rST tables.
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Acked-by: Paolo Bonzini <pbonzini@redhat.com>
Message-Id: <20200923105646.47864-1-stefanha@redhat.com>
With bdrv_filter_bs(), we can easily handle this default filter behavior
in bdrv_co_block_status().
blkdebug wants to have an additional assertion, so it keeps its own
implementation, except bdrv_co_block_status_from_file() needs to be
inlined there.
Suggested-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Andrey Shinkevich <andrey.shinkevich@virtuozzo.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
