Move the various memalign-related functions out of osdep.h and into
their own header, which we include only where they are used.
While we're doing this, add some brief documentation comments.
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-id: 20220226180723.1706285-10-peter.maydell@linaro.org
Allow writable exports to get BLK_PERM_RESIZE permission
from creation, in fuse_export_create().
In this way, there is no need to give the permission in
fuse_do_truncate(), which might be run in an iothread.
Permissions should be set only in the main thread, so
in any case if an iothread tries to set RESIZE, it will
be blocked.
Also assert in fuse_do_truncate that if we give the
RESIZE permission we can then restore the original ones.
Signed-off-by: Emanuele Giuseppe Esposito <eesposit@redhat.com>
Reviewed-by: Hanna Reitz <hreitz@redhat.com>
Message-Id: <20220303151616.325444-7-eesposit@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
This function is currently just a wrapper for bdrv_invalidate_cache(),
but in future will contain the code of bdrv_co_invalidate_cache() that
has to always be protected by BQL, and leave the rest in the I/O
coroutine.
Replace all bdrv_invalidate_cache() invokations with bdrv_activate().
Signed-off-by: Emanuele Giuseppe Esposito <eesposit@redhat.com>
Reviewed-by: Hanna Reitz <hreitz@redhat.com>
Message-Id: <20220209105452.1694545-4-eesposit@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
When building on FreeBSD we get:
[816/6851] Compiling C object libblockdev.fa.p/block_export_fuse.c.o
../block/export/fuse.c:628:16: error: use of undeclared identifier 'FALLOC_FL_KEEP_SIZE'
if (mode & FALLOC_FL_KEEP_SIZE) {
^
../block/export/fuse.c:651:16: error: use of undeclared identifier 'FALLOC_FL_PUNCH_HOLE'
if (mode & FALLOC_FL_PUNCH_HOLE) {
^
../block/export/fuse.c:652:22: error: use of undeclared identifier 'FALLOC_FL_KEEP_SIZE'
if (!(mode & FALLOC_FL_KEEP_SIZE)) {
^
3 errors generated.
FAILED: libblockdev.fa.p/block_export_fuse.c.o
Meson indeed reported FALLOC_FL_PUNCH_HOLE is not available:
C compiler for the host machine: cc (clang 10.0.1 "FreeBSD clang version 10.0.1")
Checking for function "fallocate" : NO
Checking for function "posix_fallocate" : YES
Header <linux/falloc.h> has symbol "FALLOC_FL_PUNCH_HOLE" : NO
Header <linux/falloc.h> has symbol "FALLOC_FL_ZERO_RANGE" : NO
...
Similarly to commit 304332039 ("block/export/fuse.c: fix musl build"),
guard the code requiring FALLOC_FL_KEEP_SIZE / FALLOC_FL_PUNCH_HOLE
definitions under CONFIG_FALLOCATE_PUNCH_HOLE #ifdef'ry.
Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-Id: <20220201112655.344373-3-f4bug@amsat.org>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
In order to safely maintain a mixture of #ifdef'ry with if-else-if
ladder, rearrange the last statement (!mode) first. Since it is
mutually exclusive with the other conditions, checking it first
doesn't make any logical difference, but allows to add #ifdef'ry
around in a more cleanly way.
Suggested-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-Id: <20220201112655.344373-2-f4bug@amsat.org>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
The vhost-user-blk export runs requests asynchronously in their own
coroutine. When the vhost connection goes away and we want to stop the
vhost-user server, we need to wait for these coroutines to stop before
we can unmap the shared memory. Otherwise, they would still access the
unmapped memory and crash.
This introduces a refcount to VuServer which is increased when spawning
a new request coroutine and decreased before the coroutine exits. The
memory is only unmapped when the refcount reaches zero.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20220125151435.48792-1-kwolf@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Adaptive polling measures the execution time of the polling check plus
handlers called when a polled event becomes ready. Handlers can take a
significant amount of time, making it look like polling was running for
a long time when in fact the event handler was running for a long time.
For example, on Linux the io_submit(2) syscall invoked when a virtio-blk
device's virtqueue becomes ready can take 10s of microseconds. This
can exceed the default polling interval (32 microseconds) and cause
adaptive polling to stop polling.
By excluding the handler's execution time from the polling check we make
the adaptive polling calculation more accurate. As a result, the event
loop now stays in polling mode where previously it would have fallen
back to file descriptor monitoring.
The following data was collected with virtio-blk num-queues=2
event_idx=off using an IOThread. Before:
168k IOPS, IOThread syscalls:
9837.115 ( 0.020 ms): IO iothread1/620155 io_submit(ctx_id: 140512552468480, nr: 16, iocbpp: 0x7fcb9f937db0) = 16
9837.158 ( 0.002 ms): IO iothread1/620155 write(fd: 103, buf: 0x556a2ef71b88, count: 8) = 8
9837.161 ( 0.001 ms): IO iothread1/620155 write(fd: 104, buf: 0x556a2ef71b88, count: 8) = 8
9837.163 ( 0.001 ms): IO iothread1/620155 ppoll(ufds: 0x7fcb90002800, nfds: 4, tsp: 0x7fcb9f1342d0, sigsetsize: 8) = 3
9837.164 ( 0.001 ms): IO iothread1/620155 read(fd: 107, buf: 0x7fcb9f939cc0, count: 512) = 8
9837.174 ( 0.001 ms): IO iothread1/620155 read(fd: 105, buf: 0x7fcb9f939cc0, count: 512) = 8
9837.176 ( 0.001 ms): IO iothread1/620155 read(fd: 106, buf: 0x7fcb9f939cc0, count: 512) = 8
9837.209 ( 0.035 ms): IO iothread1/620155 io_submit(ctx_id: 140512552468480, nr: 32, iocbpp: 0x7fca7d0cebe0) = 32
174k IOPS (+3.6%), IOThread syscalls:
9809.566 ( 0.036 ms): IO iothread1/623061 io_submit(ctx_id: 140539805028352, nr: 32, iocbpp: 0x7fd0cdd62be0) = 32
9809.625 ( 0.001 ms): IO iothread1/623061 write(fd: 103, buf: 0x5647cfba5f58, count: 8) = 8
9809.627 ( 0.002 ms): IO iothread1/623061 write(fd: 104, buf: 0x5647cfba5f58, count: 8) = 8
9809.663 ( 0.036 ms): IO iothread1/623061 io_submit(ctx_id: 140539805028352, nr: 32, iocbpp: 0x7fd0d0388b50) = 32
Notice that ppoll(2) and eventfd read(2) syscalls are eliminated because
the IOThread stays in polling mode instead of falling back to file
descriptor monitoring.
As usual, polling is not implemented on Windows so this patch ignores
the new io_poll_read() callback in aio-win32.c.
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Stefano Garzarella <sgarzare@redhat.com>
Message-id: 20211207132336.36627-2-stefanha@redhat.com
[Fixed up aio_set_event_notifier() calls in
tests/unit/test-fdmon-epoll.c added after this series was queued.
--Stefan]
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Include linux/falloc.h if CONFIG_FALLOCATE_ZERO_RANGE is defined to fix
50482fda98
and avoid the following build failure on musl:
../block/export/fuse.c: In function 'fuse_fallocate':
../block/export/fuse.c:643:21: error: 'FALLOC_FL_ZERO_RANGE' undeclared (first use in this function)
643 | else if (mode & FALLOC_FL_ZERO_RANGE) {
| ^~~~~~~~~~~~~~~~~~~~
Fixes:
- http://autobuild.buildroot.org/results/be24433a429fda681fb66698160132c1c99bc53b
Fixes: 50482fda98 ("block/export/fuse.c: fix musl build")
Signed-off-by: Fabrice Fontaine <fontaine.fabrice@gmail.com>
Message-Id: <20211022095209.1319671-1-fontaine.fabrice@gmail.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Include linux/fs.h to avoid the following build failure on uclibc or
musl raised since version 6.0.0:
../block/export/fuse.c: In function 'fuse_lseek':
../block/export/fuse.c:641:19: error: 'SEEK_HOLE' undeclared (first use in this function)
641 | if (whence != SEEK_HOLE && whence != SEEK_DATA) {
| ^~~~~~~~~
../block/export/fuse.c:641:19: note: each undeclared identifier is reported only once for each function it appears in
../block/export/fuse.c:641:42: error: 'SEEK_DATA' undeclared (first use in this function); did you mean 'SEEK_SET'?
641 | if (whence != SEEK_HOLE && whence != SEEK_DATA) {
| ^~~~~~~~~
| SEEK_SET
Fixes:
- http://autobuild.buildroot.org/results/33c90ebf04997f4d3557cfa66abc9cf9a3076137
Signed-off-by: Fabrice Fontaine <fontaine.fabrice@gmail.com>
Message-Id: <20210827220301.272887-1-fontaine.fabrice@gmail.com>
Signed-off-by: Hanna Reitz <hreitz@redhat.com>
Fix the following build failure on musl raised since version 6.0.0 and
4ca37a96a7
because musl does not define FALLOC_FL_ZERO_RANGE:
../block/export/fuse.c: In function 'fuse_fallocate':
../block/export/fuse.c:563:23: error: 'FALLOC_FL_ZERO_RANGE' undeclared (first use in this function)
563 | } else if (mode & FALLOC_FL_ZERO_RANGE) {
| ^~~~~~~~~~~~~~~~~~~~
Fixes:
- http://autobuild.buildroot.org/results/b96e3d364fd1f8bbfb18904a742e73327d308f64
Signed-off-by: Fabrice Fontaine <fontaine.fabrice@gmail.com>
Message-Id: <20210809095101.1101336-1-fontaine.fabrice@gmail.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Reviewed-by: Denis V. Lunev <den@openvz.org>
Signed-off-by: Hanna Reitz <hreitz@redhat.com>
When invoking block-export-add with some iothread and
fixed-iothread=false, and changing the node's iothread fails, the error
is supposed to be ignored.
However, it is still stored in *errp, which is wrong. If a second error
occurs, the "*errp must be NULL" assertion in error_setv() fails:
qemu-system-x86_64: ../util/error.c:59: error_setv: Assertion
`*errp == NULL' failed.
So if fixed-iothread=false, we should ignore the error by passing NULL
to bdrv_try_set_aio_context().
Fixes: f51d23c80a
("block/export: add iothread and fixed-iothread options")
Signed-off-by: Max Reitz <mreitz@redhat.com>
Message-Id: <20210624083825.29224-2-mreitz@redhat.com>
Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Allow changing the file mode, UID, and GID through SETATTR.
Without allow_other, UID and GID are not allowed to be changed, because
it would not make sense. Also, changing group or others' permissions
is not allowed either.
For read-only exports, +w cannot be set.
Signed-off-by: Max Reitz <mreitz@redhat.com>
Message-Id: <20210625142317.271673-5-mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
In order to support changing other attributes than the file size in
fuse_setattr(), we have to give each its own independent branch. This
also applies to the only attribute we do support right now.
Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20210625142317.271673-4-mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Without the allow_other mount option, no user (not even root) but the
one who started qemu/the storage daemon can access the export. Allow
users to configure the export such that such accesses are possible.
While allow_other is probably what users want, we cannot make it an
unconditional default, because passing it is only possible (for non-root
users) if the global fuse.conf configuration file allows it. Thus, the
default is an 'auto' mode, in which we first try with allow_other, and
then fall back to without.
FuseExport.allow_other reports whether allow_other was actually used as
a mount option or not. Currently, this information is not used, but a
future patch will let this field decide whether e.g. an export's UID and
GID can be changed through chmod.
One notable thing about 'auto' mode is that libfuse may print error
messages directly to stderr, and so may fusermount (which it executes).
Our export code cannot really filter or hide them. Therefore, if 'auto'
fails its first attempt and has to fall back, fusermount will print an
error message that mounting with allow_other failed.
This behavior necessitates a change to iotest 308, namely we need to
filter out this error message (because if the first attempt at mounting
with allow_other succeeds, there will be no such message).
Furthermore, common.rc's _make_test_img should use allow-other=off for
FUSE exports, because iotests generally do not need to access images
from other users, so allow-other=on or allow-other=auto have no
advantage. OTOH, allow-other=on will not work on systems where
user_allow_other is disabled, and with allow-other=auto, we get said
error message that we would need to filter out again. Just disabling
allow-other is simplest.
Signed-off-by: Max Reitz <mreitz@redhat.com>
Message-Id: <20210625142317.271673-3-mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
We do not do any permission checks in fuse_open(), so let the kernel do
them. We already let fuse_getattr() report the proper UNIX permissions,
so this should work the way we want.
This causes a change in 308's reference output, because now opening a
non-writable export with O_RDWR fails already, instead of only actually
attempting to write to it. (That is an improvement.)
Signed-off-by: Max Reitz <mreitz@redhat.com>
Message-Id: <20210625142317.271673-2-mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
The checks in vu_blk_sect_range_ok() assume VIRTIO_BLK_SECTOR_SIZE is
equal to BDRV_SECTOR_SIZE. This is true, but let's add a
QEMU_BUILD_BUG_ON() to make it explicit.
We might as well check that the request buffer size is a multiple of
VIRTIO_BLK_SECTOR_SIZE while we're at it.
Suggested-by: Max Reitz <mreitz@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-Id: <20210331142727.391465-1-stefanha@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
The vhost-user in-flight shmfd feature has not been tested with
qemu-storage-daemon's vhost-user-blk server. Disable this optional
feature for now because it requires MFD_ALLOW_SEALING, which is not
available in some CI environments.
If we need this feature in the future it can be re-enabled after
testing.
Reported-by: Peter Maydell <peter.maydell@linaro.org>
Cc: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-Id: <20210309094106.196911-2-stefanha@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Check that the sector number and byte count are valid.
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-Id: <20210223144653.811468-13-stefanha@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Validate discard/write zeroes the same way we do for virtio-blk. Some of
these checks are mandated by the VIRTIO specification, others are
internal to QEMU.
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-Id: <20210223144653.811468-11-stefanha@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
The driver is supposed to honor the blk_size field but the protocol
still uses 512-byte sector numbers. It is incorrect to multiply
req->sector_num by blk_size.
VIRTIO 1.1 5.2.5 Device Initialization says:
blk_size can be read to determine the optimal sector size for the
driver to use. This does not affect the units used in the protocol
(always 512 bytes), but awareness of the correct value can affect
performance.
Fixes: 3578389bcf ("block/export: vhost-user block device backend server")
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-Id: <20210223144653.811468-10-stefanha@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Use VIRTIO_BLK_SECTOR_BITS and VIRTIO_BLK_SECTOR_SIZE when dealing with
virtio-blk sector numbers. Although the values happen to be the same as
BDRV_SECTOR_BITS and BDRV_SECTOR_SIZE, they are conceptually different.
This makes it clearer when we are dealing with virtio-blk sector units.
Use VIRTIO_BLK_SECTOR_BITS in vu_blk_initialize_config(). Later patches
will use it the new constants the virtqueue request processing code
path.
Suggested-by: Max Reitz <mreitz@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-Id: <20210223144653.811468-9-stefanha@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
The config->blk_size field is little-endian. Use the native-endian
blk_size variable to avoid double byteswapping.
Fixes: 11f60f7eae ("block/export: make vhost-user-blk config space little-endian")
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-Id: <20210223144653.811468-8-stefanha@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
The easiest spots to use QAPI_LIST_APPEND are where we already have an
obvious pointer to the tail of a list. While at it, consistently use
the variable name 'tail' for that purpose.
Signed-off-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Message-Id: <20210113221013.390592-5-eblake@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>
This is a relatively new feature in libfuse (available since 3.8.0,
which was released in November 2019), so we have to add a dedicated
check whether it is available before making use of it.
Signed-off-by: Max Reitz <mreitz@redhat.com>
Message-Id: <20201027190600.192171-7-mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
This allows allocating areas after the (old) EOF as part of a growing
resize, writing zeroes, and discarding.
Signed-off-by: Max Reitz <mreitz@redhat.com>
Message-Id: <20201027190600.192171-6-mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
These will behave more like normal files in that writes beyond the EOF
will automatically grow the export size.
As an optimization, keep the RESIZE permission for growable exports so
we do not have to take it for every post-EOF write. (This permission is
not released when the export is destroyed, because at that point the
BlockBackend is destroyed altogether anyway.)
Signed-off-by: Max Reitz <mreitz@redhat.com>
Message-Id: <20201027190600.192171-5-mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
This makes the export actually useful instead of only producing errors
whenever it is accessed.
Signed-off-by: Max Reitz <mreitz@redhat.com>
Message-Id: <20201027190600.192171-4-mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
block-export-add type=fuse allows mounting block graph nodes via FUSE on
some existing regular file. That file should then appears like a raw
disk image, and accesses to it result in accesses to the exported BDS.
Right now, we only implement the necessary block export functions to set
it up and shut it down. We do not implement any access functions, so
accessing the mount point only results in errors. This will be
addressed by a followup patch.
We keep a hash table of exported mount points, because we want to be
able to detect when users try to use a mount point twice. This is
because we invoke stat() to check whether the given mount point is a
regular file, but if that file is served by ourselves (because it is
already used as a mount point), then this stat() would have to be served
by ourselves, too, which is impossible to do while we (as the caller)
are waiting for it to settle. Therefore, keep track of mount point
paths to at least catch the most obvious instances of that problem.
Signed-off-by: Max Reitz <mreitz@redhat.com>
Message-Id: <20201027190600.192171-3-mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Do not validate input with g_return_val_if(). This API is intended for
checking programming errors and is compiled out with -DG_DISABLE_CHECKS.
Use an explicit if statement for input validation so it cannot
accidentally be compiled out.
Suggested-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-Id: <20201118091644.199527-5-stefanha@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
By making libvhost-user a subproject, check it builds
standalone (without the global QEMU cflags etc).
Note that the library still relies on QEMU include/qemu/atomic.h and
linux_headers/.
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Message-Id: <20201125100640.366523-6-marcandre.lureau@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
The --enable/disable-vhost-user-blk-server options were implemented in
./configure. There has been confusion about them and part of the problem
is that the shell syntax used for setting the default value is not easy
to read. Move the option over to meson where the conditions are easier
to understand:
have_vhost_user_blk_server = (targetos == 'linux')
if get_option('vhost_user_blk_server').enabled()
if targetos != 'linux'
error('vhost_user_blk_server requires linux')
endif
elif get_option('vhost_user_blk_server').disabled() or not have_system
have_vhost_user_blk_server = false
endif
This patch does not change behavior.
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-Id: <20201110171121.1265142-2-stefanha@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Tested-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Refuse get_config() requests in excess of sizeof(struct virtio_blk_config).
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-Id: <20201027173528.213464-5-stefanha@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
VIRTIO 1.0 devices have little-endian configuration space. The
vhost-user-blk-server.c code already uses little-endian for virtqueue
processing but not for the configuration space fields. Fix this so the
vhost-user-blk export works on big-endian hosts.
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-Id: <20201027173528.213464-4-stefanha@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Make it possible to compile out the vhost-user-blk server. It is enabled
by default on Linux.
Note that vhost-user-server.c depends on libvhost-user, which requires
CONFIG_LINUX. The CONFIG_VHOST_USER dependency was erroneous since that
option controls vhost-user frontends (previously known as "master") and
not device backends (previously known as "slave").
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-Id: <20201027173528.213464-3-stefanha@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Allow the number of queues to be configured using --export
vhost-user-blk,num-queues=N. This setting should match the QEMU --device
vhost-user-blk-pci,num-queues=N setting but QEMU vhost-user-blk.c lowers
its own value if the vhost-user-blk backend offers fewer queues than
QEMU.
The vhost-user-blk-server.c code is already capable of multi-queue. All
virtqueue processing runs in the same AioContext. No new locking is
needed.
Add the num-queues=N option and set the VIRTIO_BLK_F_MQ feature bit.
Note that the feature bit only announces the presence of the num_queues
configuration space field. It does not promise that there is more than 1
virtqueue, so we can set it unconditionally.
I tested multi-queue by running a random read fio test with numjobs=4 on
an -smp 4 guest. After the benchmark finished the guest /proc/interrupts
file showed activity on all 4 virtio-blk MSI-X. The /sys/block/vda/mq/
directory shows that Linux blk-mq has 4 queues configured.
An automated test is included in the next commit.
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Acked-by: Markus Armbruster <armbru@redhat.com>
Message-id: 20201001144604.559733-2-stefanha@redhat.com
[Fixed accidental tab characters as suggested by Markus Armbruster
--Stefan]
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Make it possible to specify the iothread where the export will run. By
default the block node can be moved to other AioContexts later and the
export will follow. The fixed-iothread option forces strict behavior
that prevents changing AioContext while the export is active. See the
QAPI docs for details.
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-id: 20200929125516.186715-5-stefanha@redhat.com
[Fix stray '#' character in block-export.json and add missing "(since:
5.2)" as suggested by Eric Blake.
--Stefan]
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Block exports are used by softmmu, qemu-storage-daemon, and qemu-nbd.
They are not used by other programs and are not otherwise needed in
libblock.
Undo the recent move of blockdev-nbd.c from blockdev_ss into block_ss.
Since bdrv_close_all() (libblock) calls blk_exp_close_all()
(libblockdev) a stub function is required..
Make qemu-nbd.c use signal handling utility functions instead of
duplicating the code. This helps because os-posix.c is in libblockdev
and it depends on a qemu_system_killed() symbol that qemu-nbd.c lacks.
Once we use the signal handling utility functions we also end up
providing the necessary symbol.
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-id: 20200929125516.186715-4-stefanha@redhat.com
[Fixed s/ndb/nbd/ typo in commit description as suggested by Eric Blake
--Stefan]
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Don't compile contrib/libvhost-user/libvhost-user.c again. Instead build
the static library once and then reuse it throughout QEMU.
Also switch from CONFIG_LINUX to CONFIG_VHOST_USER, which is what the
vhost-user tools (vhost-user-gpu, etc) do.
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-id: 20200924151549.913737-14-stefanha@redhat.com
[Added CONFIG_LINUX again because libvhost-user doesn't build on macOS.
--Stefan]
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Headers used by other subsystems are located in include/. Also add the
vhost-user-server and vhost-user-blk-server headers to MAINTAINERS.
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-id: 20200924151549.913737-13-stefanha@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Use the new QAPI block exports API instead of defining our own QOM
objects.
This is a large change because the lifecycle of VuBlockDev needs to
follow BlockExportDriver. QOM properties are replaced by QAPI options
objects.
VuBlockDev is renamed VuBlkExport and contains a BlockExport field.
Several fields can be dropped since BlockExport already has equivalents.
The file names and meson build integration will be adjusted in a future
patch. libvhost-user should probably be built as a static library that
is linked into QEMU instead of as a .c file that results in duplicate
compilation.
The new command-line syntax is:
$ qemu-storage-daemon \
--blockdev file,node-name=drive0,filename=test.img \
--export vhost-user-blk,node-name=drive0,id=export0,unix-socket=/tmp/vhost-user-blk.sock
Note that unix-socket is optional because we may wish to accept chardevs
too in the future.
Markus noted that supported address families are not explicit in the
QAPI schema. It is unlikely that support for more address families will
be added since file descriptor passing is required and few address
families support it. If a new address family needs to be added, then the
QAPI 'features' syntax can be used to advertize them.
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Acked-by: Markus Armbruster <armbru@redhat.com>
Message-id: 20200924151549.913737-12-stefanha@redhat.com
[Skip test on big-endian host architectures because this device doesn't
support them yet (as already mentioned in a code comment).
--Stefan]
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Propagate the flush return value since errors are possible.
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-id: 20200924151549.913737-11-stefanha@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
The vu_client_trip() coroutine is leaked during AioContext switching. It
is also unsafe to destroy the vu_dev in panic_cb() since its callers
still access it in some cases.
Rework the lifecycle to solve these safety issues.
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-id: 20200924151549.913737-10-stefanha@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
The device panic notifier callback is not used. Drop it.
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-id: 20200924151549.913737-7-stefanha@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Only one struct is needed per request. Drop req_data and the separate
VuBlockReq instance. Instead let vu_queue_pop() allocate everything at
once.
This fixes the req_data memory leak in vu_block_virtio_process_req().
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-id: 20200924151549.913737-6-stefanha@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
By making use of libvhost-user, block device drive can be shared to
the connected vhost-user client. Only one client can connect to the
server one time.
Since vhost-user-server needs a block drive to be created first, delay
the creation of this object.
Suggested-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Coiby Xu <coiby.xu@gmail.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Message-id: 20200918080912.321299-6-coiby.xu@gmail.com
[Shorten "vhost_user_blk_server" string to "vhost_user_blk" to avoid the
following compiler warning:
../block/export/vhost-user-blk-server.c:178:50: error: ‘%s’ directive output truncated writing 21 bytes into a region of size 20 [-Werror=format-truncation=]
and fix "Invalid size %ld ..." ssize_t format string arguments for
32-bit hosts.
--Stefan]
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
The 'writable' option is a basic option that will probably be applicable
to most if not all export types that we will implement. Move it from NBD
to the generic BlockExport layer.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-Id: <20200924152717.287415-26-kwolf@redhat.com>
Acked-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
This adds a simple QMP command to query the list of block exports.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-Id: <20200924152717.287415-25-kwolf@redhat.com>
Acked-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Every export type will need a BlockBackend, so creating it centrally in
blk_exp_add() instead of the .create driver callback avoids duplication.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-Id: <20200924152717.287415-24-kwolf@redhat.com>
Acked-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Every block export has a BlockBackend representing the disk that is
exported. It should live in BlockExport therefore.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-Id: <20200924152717.287415-23-kwolf@redhat.com>
Acked-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Clients may want to know when an export has finally disappeard
(block-export-del returns earlier than that in the general case), so add
a QAPI event for it.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20200924152717.287415-22-kwolf@redhat.com>
Acked-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Peter Maydell
Emanuele Giuseppe Esposito
Philippe Mathieu-Daudé
Kevin Wolf
Stefan Hajnoczi
Fabrice Fontaine
Max Reitz
Eric Blake
Marc-André Lureau
Coiby Xu