Currently, ublk_drv associates to each hardware queue (hctx) a unique
task (called the queue's ubq_daemon) which is allowed to issue
COMMIT_AND_FETCH commands against the hctx. If any other task attempts
to do so, the command fails immediately with EINVAL. When considered
together with the block layer architecture, the result is that for each
CPU C on the system, there is a unique ublk server thread which is
allowed to handle I/O submitted on CPU C. This can lead to suboptimal
performance under imbalanced load generation. For an extreme example,
suppose all the load is generated on CPUs mapping to a single ublk
server thread. Then that thread may be fully utilized and become the
bottleneck in the system, while other ublk server threads are totally
idle.
This issue can also be addressed directly in the ublk server without
kernel support by having threads dequeue I/Os and pass them around to
ensure even load. But this solution requires inter-thread communication
at least twice for each I/O (submission and completion), which is
generally a bad pattern for performance. The problem gets even worse
with zero copy, as more inter-thread communication would be required to
have the buffer register/unregister calls to come from the correct
thread.
Therefore, address this issue in ublk_drv by allowing each I/O to have
its own daemon task. Two I/Os in the same queue are now allowed to be
serviced by different daemon tasks - this was not possible before.
Imbalanced load can then be balanced across all ublk server threads by
having the ublk server threads issue FETCH_REQs in a round-robin manner.
As a small toy example, consider a system with a single ublk device
having 2 queues, each of depth 4. A ublk server having 4 threads could
issue its FETCH_REQs against this device as follows (where each entry is
the qid,tag pair that the FETCH_REQ targets):
ublk server thread: T0 T1 T2 T3
0,0 0,1 0,2 0,3
1,3 1,0 1,1 1,2
This setup allows for load that is concentrated on one hctx/ublk_queue
to be spread out across all ublk server threads, alleviating the issue
described above.
Add the new UBLK_F_PER_IO_DAEMON feature to ublk_drv, which ublk servers
can use to essentially test for the presence of this change and tailor
their behavior accordingly.
Signed-off-by: Uday Shankar <ushankar@purestorage.com>
Reviewed-by: Caleb Sander Mateos <csander@purestorage.com>
Reviewed-by: Ming Lei <ming.lei@redhat.com>
Reviewed-by: Jens Axboe <axboe@kernel.dk>
Link: https://lore.kernel.org/r/20250529-ublk_task_per_io-v8-1-e9d3b119336a@purestorage.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Pull block updates from Jens Axboe:
- ublk updates:
- Add support for updating the size of a ublk instance
- Zero-copy improvements
- Auto-registering of buffers for zero-copy
- Series simplifying and improving GET_DATA and request lookup
- Series adding quiesce support
- Lots of selftests additions
- Various cleanups
- NVMe updates via Christoph:
- add per-node DMA pools and use them for PRP/SGL allocations
(Caleb Sander Mateos, Keith Busch)
- nvme-fcloop refcounting fixes (Daniel Wagner)
- support delayed removal of the multipath node and optionally
support the multipath node for private namespaces (Nilay Shroff)
- support shared CQs in the PCI endpoint target code (Wilfred
Mallawa)
- support admin-queue only authentication (Hannes Reinecke)
- use the crc32c library instead of the crypto API (Eric Biggers)
- misc cleanups (Christoph Hellwig, Marcelo Moreira, Hannes
Reinecke, Leon Romanovsky, Gustavo A. R. Silva)
- MD updates via Yu:
- Fix that normal IO can be starved by sync IO, found by mkfs on
newly created large raid5, with some clean up patches for bdev
inflight counters
- Clean up brd, getting rid of atomic kmaps and bvec poking
- Add loop driver specifically for zoned IO testing
- Eliminate blk-rq-qos calls with a static key, if not enabled
- Improve hctx locking for when a plug has IO for multiple queues
pending
- Remove block layer bouncing support, which in turn means we can
remove the per-node bounce stat as well
- Improve blk-throttle support
- Improve delay support for blk-throttle
- Improve brd discard support
- Unify IO scheduler switching. This should also fix a bunch of lockdep
warnings we've been seeing, after enabling lockdep support for queue
freezing/unfreezeing
- Add support for block write streams via FDP (flexible data placement)
on NVMe
- Add a bunch of block helpers, facilitating the removal of a bunch of
duplicated boilerplate code
- Remove obsolete BLK_MQ pci and virtio Kconfig options
- Add atomic/untorn write support to blktrace
- Various little cleanups and fixes
* tag 'for-6.16/block-20250523' of git://git.kernel.dk/linux: (186 commits)
selftests: ublk: add test for UBLK_F_QUIESCE
ublk: add feature UBLK_F_QUIESCE
selftests: ublk: add test case for UBLK_U_CMD_UPDATE_SIZE
traceevent/block: Add REQ_ATOMIC flag to block trace events
ublk: run auto buf unregisgering in same io_ring_ctx with registering
io_uring: add helper io_uring_cmd_ctx_handle()
ublk: remove io argument from ublk_auto_buf_reg_fallback()
ublk: handle ublk_set_auto_buf_reg() failure correctly in ublk_fetch()
selftests: ublk: add test for covering UBLK_AUTO_BUF_REG_FALLBACK
selftests: ublk: support UBLK_F_AUTO_BUF_REG
ublk: support UBLK_AUTO_BUF_REG_FALLBACK
ublk: register buffer to local io_uring with provided buf index via UBLK_F_AUTO_BUF_REG
ublk: prepare for supporting to register request buffer automatically
ublk: convert to refcount_t
selftests: ublk: make IO & device removal test more stressful
nvme: rename nvme_mpath_shutdown_disk to nvme_mpath_remove_disk
nvme: introduce multipath_always_on module param
nvme-multipath: introduce delayed removal of the multipath head node
nvme-pci: derive and better document max segments limits
nvme-pci: use struct_size for allocation struct nvme_dev
...
UBLK_F_AUTO_BUF_REG requires that the buffer registered automatically
is unregistered in same `io_ring_ctx`, so check it explicitly.
Document this requirement for UBLK_F_AUTO_BUF_REG.
Drop WARN_ON_ONCE() which is triggered from userspace code path.
Fixes: 99c1e4eb6a ("ublk: register buffer to local io_uring with provided buf index via UBLK_F_AUTO_BUF_REG")
Reported-by: Caleb Sander Mateos <csander@purestorage.com>
Reviewed-by: Caleb Sander Mateos <csander@purestorage.com>
Signed-off-by: Ming Lei <ming.lei@redhat.com>
Link: https://lore.kernel.org/r/20250522152043.399824-3-ming.lei@redhat.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
For UBLK_F_AUTO_BUF_REG, buffer is registered to uring_cmd context
automatically with the provided buffer index. User may provide one wrong
buffer index, or the specified buffer is registered by application already.
Add UBLK_AUTO_BUF_REG_FALLBACK for supporting to auto buffer registering
fallback by completing the uring_cmd and telling ublk server the
register failure via UBLK_AUTO_BUF_REG_FALLBACK, then ublk server still
can register the buffer from userspace.
So we can provide reliable way for supporting auto buffer register.
Signed-off-by: Ming Lei <ming.lei@redhat.com>
Link: https://lore.kernel.org/r/20250520045455.515691-5-ming.lei@redhat.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Add UBLK_F_AUTO_BUF_REG for supporting to register buffer automatically
to local io_uring context with provided buffer index.
Add UAPI structure `struct ublk_auto_buf_reg` for holding user parameter
to register request buffer automatically, one 'flags' field is defined, and
there is still 32bit available for future extension, such as, adding one
io_ring FD field for registering buffer to external io_uring.
`struct ublk_auto_buf_reg` is populated from ublk uring_cmd's sqe->addr,
and all existing ublk commands are data-less, so it is just fine to reuse
sqe->addr for this purpose.
Signed-off-by: Ming Lei <ming.lei@redhat.com>
Link: https://lore.kernel.org/r/20250520045455.515691-4-ming.lei@redhat.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
UBLK_F_SUPPORT_ZERO_COPY requires ublk server to issue explicit buffer
register/unregister uring_cmd for each IO, this way is not only inefficient,
but also introduce dependency between buffer consumer and buffer register/
unregister uring_cmd, please see tools/testing/selftests/ublk/stripe.c
in which backing file IO has to be issued one by one by IOSQE_IO_LINK.
Prepare for adding feature UBLK_F_AUTO_BUF_REG for addressing the existing
zero copy limitation:
- register request buffer automatically to ublk uring_cmd's io_uring
context before delivering io command to ublk server
- unregister request buffer automatically from the ublk uring_cmd's
io_uring context when completing the request
- io_uring will unregister the buffer automatically when uring is
exiting, so we needn't worry about accident exit
For using this feature, ublk server has to create one sparse buffer table
Reviewed-by: Caleb Sander Mateos <csander@purestorage.com>
Signed-off-by: Ming Lei <ming.lei@redhat.com>
Link: https://lore.kernel.org/r/20250520045455.515691-3-ming.lei@redhat.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Commit:
f40139fde5 ("ublk: fix race between io_uring_cmd_complete_in_task and
ublk_cancel_cmd")
adds a request state check in ublk_cancel_cmd(), and if the request is
started, skips canceling this uring_cmd.
However, the current uring_cmd may be in ACTIVE state, without block
request coming to the uring command. Meantime, if the cached request in
tag_set.tags[tag] has been delivered to ublk server and reycycled, then
this uring_cmd can't be canceled.
ublk requests are aborted in ublk char device release handler, which
depends on canceling all ACTIVE uring_cmd. So it causes a dead loop.
Fix this issue by not taking a stale request into account when canceling
uring_cmd in ublk_cancel_cmd().
Reported-by: Shinichiro Kawasaki <shinichiro.kawasaki@wdc.com>
Closes: https://lore.kernel.org/linux-block/mruqwpf4tqenkbtgezv5oxwq7ngyq24jzeyqy4ixzvivatbbxv@4oh2wzz4e6qn/
Fixes: f40139fde5 ("ublk: fix race between io_uring_cmd_complete_in_task and ublk_cancel_cmd")
Signed-off-by: Ming Lei <ming.lei@redhat.com>
Link: https://lore.kernel.org/r/20250515162601.77346-1-ming.lei@redhat.com
[axboe: rewording of commit message]
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Merge 6.15 block fixes in, once again, to resolve conflicts with the
fixes for ublk that went into mainline and the 6.16 ublk updates.
* block-6.15:
nvmet-auth: always free derived key data
nvmet-tcp: don't restore null sk_state_change
nvmet-tcp: select CONFIG_TLS from CONFIG_NVME_TARGET_TCP_TLS
nvme-tcp: select CONFIG_TLS from CONFIG_NVME_TCP_TLS
nvme-tcp: fix premature queue removal and I/O failover
nvme-pci: add quirks for WDC Blue SN550 15b7:5009
nvme-pci: add quirks for device 126f:1001
nvme-pci: fix queue unquiesce check on slot_reset
ublk: remove the check of ublk_need_req_ref() from __ublk_check_and_get_req
ublk: enhance check for register/unregister io buffer command
ublk: decouple zero copy from user copy
selftests: ublk: fix UBLK_F_NEED_GET_DATA
Signed-off-by: Jens Axboe <axboe@kernel.dk>
A ublk_io is converted to a request in several places in the I/O path by
using blk_mq_tag_to_rq() to look up the (qid, tag) on the ublk device's
tagset. This involves a bunch of dereferences and a tag bounds check.
To make this conversion cheaper, store the request pointer in ublk_io.
Overlap this storage with the io_uring_cmd pointer. This is safe because
the io_uring_cmd pointer is only valid if UBLK_IO_FLAG_ACTIVE is set on
the ublk_io, the request pointer is valid if UBLK_IO_FLAG_OWNED_BY_SRV,
and these flags are mutually exclusive.
Suggested-by: Ming Lei <ming.lei@redhat.com>
Signed-off-by: Caleb Sander Mateos <csander@purestorage.com>
Reviewed-by: Ming Lei <ming.lei@redhat.com>
Link: https://lore.kernel.org/r/20250430225234.2676781-10-csander@purestorage.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
ublk_abort_queue() currently checks whether the UBLK_IO_FLAG_ACTIVE flag
is cleared to tell whether to abort each ublk_io in the queue. But it's
possible for a ublk_io to not be ACTIVE but also not have a request in
flight, such as when no fetch request has yet been submitted for a tag
or when a fetch request is cancelled. So ublk_abort_queue() must
additionally check for an inflight request.
Simplify this code by checking for UBLK_IO_FLAG_OWNED_BY_SRV instead,
which indicates precisely whether a request is currently inflight.
Signed-off-by: Caleb Sander Mateos <csander@purestorage.com>
Reviewed-by: Ming Lei <ming.lei@redhat.com>
Link: https://lore.kernel.org/r/20250430225234.2676781-9-csander@purestorage.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
ublk_dispatch_req() currently handles 3 different cases: incoming ublk
requests that don't need to wait for a data buffer, incoming requests
that do need to wait for a buffer, and resuming those requests once the
buffer is provided. But the call site that provides a data buffer
(UBLK_IO_NEED_GET_DATA) is separate from those for incoming requests.
So simplify the function by splitting the UBLK_IO_NEED_GET_DATA case
into its own function ublk_get_data(). This avoids several redundant
checks in the UBLK_IO_NEED_GET_DATA case, and streamlines the incoming
request cases.
Don't call ublk_fill_io_cmd() for UBLK_IO_NEED_GET_DATA, as it's no
longer necessary to set io->cmd or the UBLK_IO_FLAG_ACTIVE flag for
ublk_dispatch_req().
Since UBLK_IO_NEED_GET_DATA no longer relies on ublk_dispatch_req()
calling io_uring_cmd_done(), return the UBLK_IO_RES_OK status directly
from the ->uring_cmd() handler. If ublk_start_io() fails, don't complete
the UBLK_IO_NEED_GET_DATA command, matching the existing behavior.
Signed-off-by: Caleb Sander Mateos <csander@purestorage.com>
Reviewed-by: Ming Lei <ming.lei@redhat.com>
Link: https://lore.kernel.org/r/20250430225234.2676781-8-csander@purestorage.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Move the logic for the UBLK_IO_COMMIT_AND_FETCH_REQ opcode into its own
function. This also allows us to mark ublk_queue pointers as const for
that operation, which can help prevent data races since we may allow
concurrent operation on one ublk_queue in the future. Also open code
ublk_commit_completion in ublk_commit_and_fetch to reduce the number of
parameters/avoid a redundant lookup.
[Restore __ublk_ch_uring_cmd() req variable used in commit d6aa0c178b
("ublk: call ublk_dispatch_req() for handling UBLK_U_IO_NEED_GET_DATA")]
Suggested-by: Ming Lei <ming.lei@redhat.com>
Signed-off-by: Uday Shankar <ushankar@purestorage.com>
Reviewed-by: Caleb Sander Mateos <csander@purestorage.com>
Reviewed-by: Ming Lei <ming.lei@redhat.com>
Signed-off-by: Caleb Sander Mateos <csander@purestorage.com>
Link: https://lore.kernel.org/r/20250430225234.2676781-2-csander@purestorage.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
The simple check of UBLK_IO_FLAG_OWNED_BY_SRV can avoid incorrect
register/unregister io buffer easily, so check it before calling
starting to register/un-register io buffer.
Also only allow io buffer register/unregister uring_cmd in case of
UBLK_F_SUPPORT_ZERO_COPY.
Also mark argument 'ublk_queue *' of ublk_register_io_buf as const.
Reviewed-by: Caleb Sander Mateos <csander@purestorage.com>
Fixes: 1f6540e2aa ("ublk: zc register/unregister bvec")
Signed-off-by: Ming Lei <ming.lei@redhat.com>
Link: https://lore.kernel.org/r/20250429022941.1718671-4-ming.lei@redhat.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
UBLK_F_USER_COPY and UBLK_F_SUPPORT_ZERO_COPY are two different
features, and shouldn't be coupled together.
Commit 1f6540e2aa ("ublk: zc register/unregister bvec") enables
user copy automatically in case of UBLK_F_SUPPORT_ZERO_COPY, this way
isn't correct.
So decouple zero copy from user copy, and use independent helper to
check each one.
Fixes: 1f6540e2aa ("ublk: zc register/unregister bvec")
Signed-off-by: Ming Lei <ming.lei@redhat.com>
Link: https://lore.kernel.org/r/20250429022941.1718671-3-ming.lei@redhat.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Merge 6.15 block fixes - both to get the fixes causing issues with
XFS testing, but also to make it easier for 6.16 ublk patches to avoid
conflicts.
* block-6.15:
ublk: fix race between io_uring_cmd_complete_in_task and ublk_cancel_cmd
ublk: call ublk_dispatch_req() for handling UBLK_U_IO_NEED_GET_DATA
block: don't autoload drivers on blk-cgroup configuration
block: don't autoload drivers on stat
block: remove the backing_inode variable in bdev_statx
block: move blkdev_{get,put} _no_open prototypes out of blkdev.h
block: never reduce ra_pages in blk_apply_bdi_limits
selftests: ublk: common: fix _get_disk_dev_t for pre-9.0 coreutils
selftests: ublk: remove useless 'delay_us' from 'struct dev_ctx'
selftests: ublk: fix recover test
block: hoist block size validation code to a separate function
block: fix race between set_blocksize and read paths
nvmet: fix out-of-bounds access in nvmet_enable_port
We call io_uring_cmd_complete_in_task() to schedule task_work for handling
UBLK_U_IO_NEED_GET_DATA.
This way is really not necessary because the current context is exactly
the ublk queue context, so call ublk_dispatch_req() directly for handling
UBLK_U_IO_NEED_GET_DATA.
Fixes: 216c8f5ef0 ("ublk: replace monitor with cancelable uring_cmd")
Tested-by: Jared Holzman <jholzman@nvidia.com>
Signed-off-by: Ming Lei <ming.lei@redhat.com>
Link: https://lore.kernel.org/r/20250425013742.1079549-2-ming.lei@redhat.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
ublk_init_queues() ensures that all nr_hw_queues queues are initialized,
with each ublk_queue's q_id set to its index. And ublk_init_queues() is
called before ublk_add_chdev(), which creates the cdev. Is is therefore
impossible for the !ubq || ub_cmd->q_id != ubq->q_id condition to hit in
__ublk_ch_uring_cmd(). Remove it to avoids some branches in the I/O path.
Signed-off-by: Caleb Sander Mateos <csander@purestorage.com>
Reviewed-by: Ming Lei <ming.lei@redhat.com>
Reviewed-by: Uday Shankar <ushankar@purestorage.com>
Link: https://lore.kernel.org/r/20250416170154.3621609-1-csander@purestorage.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Currently ublk only allows the size of the ublkb block device to be
set via UBLK_CMD_SET_PARAMS before UBLK_CMD_START_DEV is triggered.
This does not provide support for extendable user-space block devices
without having to stop and restart the underlying ublkb block device
causing IO interruption.
This patch adds a new ublk command UBLK_U_CMD_UPDATE_SIZE to allow the
ublk block device to be resized on-the-fly.
Feature flag UBLK_F_UPDATE_SIZE is also added to indicate support.
Signed-off-by: Omri Mann <omri@nvidia.com>
Reviewed-by: Ming Lei <ming.lei@redhat.com>
Link: https://lore.kernel.org/r/2a370ab1-d85b-409d-b762-f9f3f6bdf705@nvidia.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Now ublk_abort_queue() is moved to ublk char device release handler,
meantime our request queue is "quiesced" because either ->canceling was
set from uring_cmd cancel function or all IOs are inflight and can't be
completed by ublk server, things becomes easy much:
- all uring_cmd are done, so we needn't to mark io as UBLK_IO_FLAG_ABORTED
for handling completion from uring_cmd
- ublk char device is closed, no one can hold IO request reference any more,
so we can simply complete this request or requeue it for ublk_nosrv_should_reissue_outstanding.
Reviewed-by: Uday Shankar <ushankar@purestorage.com>
Signed-off-by: Ming Lei <ming.lei@redhat.com>
Link: https://lore.kernel.org/r/20250416035444.99569-8-ming.lei@redhat.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Remove __ublk_quiesce_dev() and open code for updating device state as
QUIESCED.
We needn't to drain inflight requests in __ublk_quiesce_dev() any more,
because all inflight requests are aborted in ublk char device release
handler.
Also we needn't to set ->canceling in __ublk_quiesce_dev() any more
because it is done unconditionally now in ublk_ch_release().
Reviewed-by: Uday Shankar <ushankar@purestorage.com>
Signed-off-by: Ming Lei <ming.lei@redhat.com>
Link: https://lore.kernel.org/r/20250416035444.99569-7-ming.lei@redhat.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
There are currently two ways in which ublk server exit is detected by
ublk_drv:
1. uring_cmd cancellation. If there are any outstanding uring_cmds which
have not been completed to the ublk server when it exits, io_uring
calls the uring_cmd callback with a special cancellation flag as the
issuing task is exiting.
2. I/O timeout. This is needed in addition to the above to handle the
"saturated queue" case, when all I/Os for a given queue are in the
ublk server, and therefore there are no outstanding uring_cmds to
cancel when the ublk server exits.
There are a couple of issues with this approach:
- It is complex and inelegant to have two methods to detect the same
condition
- The second method detects ublk server exit only after a long delay
(~30s, the default timeout assigned by the block layer). This delays
the nosrv behavior from kicking in and potential subsequent recovery
of the device.
The second issue is brought to light with the new test_generic_06 which
will be added in following patch. It fails before this fix:
selftests: ublk: test_generic_06.sh
dev id is 0
dd: error writing '/dev/ublkb0': Input/output error
1+0 records in
0+0 records out
0 bytes copied, 30.0611 s, 0.0 kB/s
DEAD
dd took 31 seconds to exit (>= 5s tolerance)!
generic_06 : [FAIL]
Fix this by instead detecting and handling ublk server exit in the
character file release callback. This has several advantages:
- This one place can handle both saturated and unsaturated queues. Thus,
it replaces both preexisting methods of detecting ublk server exit.
- It runs quickly on ublk server exit - there is no 30s delay.
- It starts the process of removing task references in ublk_drv. This is
needed if we want to relax restrictions in the driver like letting
only one thread serve each queue
There is also the disadvantage that the character file release callback
can also be triggered by intentional close of the file, which is a
significant behavior change. Preexisting ublk servers (libublksrv) are
dependent on the ability to open/close the file multiple times. To
address this, only transition to a nosrv state if the file is released
while the ublk device is live. This allows for programs to open/close
the file multiple times during setup. It is still a behavior change if a
ublk server decides to close/reopen the file while the device is LIVE
(i.e. while it is responsible for serving I/O), but that would be highly
unusual. This behavior is in line with what is done by FUSE, which is
very similar to ublk in that a userspace daemon is providing services
traditionally provided by the kernel.
With this change in, the new test (and all other selftests, and all
ublksrv tests) pass:
selftests: ublk: test_generic_06.sh
dev id is 0
dd: error writing '/dev/ublkb0': Input/output error
1+0 records in
0+0 records out
0 bytes copied, 0.0376731 s, 0.0 kB/s
DEAD
generic_04 : [PASS]
Signed-off-by: Uday Shankar <ushankar@purestorage.com>
Signed-off-by: Ming Lei <ming.lei@redhat.com>
Link: https://lore.kernel.org/r/20250416035444.99569-6-ming.lei@redhat.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
ublk_ch_release() is called after ublk char device is closed, when all
uring_cmd are done, so it is perfect fine to move ublk device reset to
ublk_ch_release() from ublk_ctrl_start_recovery().
This way can avoid to grab the exiting daemon task_struct too long.
However, reset of the following ublk IO flags has to be moved until ublk
io_uring queues are ready:
- ubq->canceling
For requeuing IO in case of ublk_nosrv_dev_should_queue_io() before device
is recovered
- ubq->fail_io
For failing IO in case of UBLK_F_USER_RECOVERY_FAIL_IO before device is
recovered
- ublk_io->flags
For preventing using io->cmd
With this way, recovery is simplified a lot.
Signed-off-by: Ming Lei <ming.lei@redhat.com>
Link: https://lore.kernel.org/r/20250416035444.99569-5-ming.lei@redhat.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Now ublk deals with ublk_nosrv_dev_should_queue_io() by keeping request
queue as quiesced. This way is fragile because queue quiesce crosses syscalls
or process contexts.
Switch to rely on ubq->canceling for dealing with
ublk_nosrv_dev_should_queue_io(), because it has been used for this purpose
during io_uring context exiting, and it can be reused before recovering too.
In ublk_queue_rq(), the request will be added to requeue list without
kicking off requeue in case of ubq->canceling, and finally requests added in
requeue list will be dispatched from either ublk_stop_dev() or
ublk_ctrl_end_recovery().
Meantime we have to move reset of ubq->canceling from ublk_ctrl_start_recovery()
to ublk_ctrl_end_recovery(), when IO handling can be recovered completely.
Then blk_mq_quiesce_queue() and blk_mq_unquiesce_queue() are always used
in same context.
Signed-off-by: Ming Lei <ming.lei@redhat.com>
Reviewed-by: Uday Shankar <ushankar@purestorage.com>
Link: https://lore.kernel.org/r/20250416035444.99569-4-ming.lei@redhat.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Most uring_cmds issued against ublk character devices are serialized
because each command affects only one queue, and there is an early check
which only allows a single task (the queue's ubq_daemon) to issue
uring_cmds against that queue. However, this mechanism does not work for
FETCH_REQs, since they are expected before ubq_daemon is set. Since
FETCH_REQs are only used at initialization and not in the fast path,
serialize them using the per-ublk-device mutex. This fixes a number of
data races that were previously possible if a badly behaved ublk server
decided to issue multiple FETCH_REQs against the same qid/tag
concurrently.
Reported-by: Caleb Sander Mateos <csander@purestorage.com>
Signed-off-by: Uday Shankar <ushankar@purestorage.com>
Signed-off-by: Ming Lei <ming.lei@redhat.com>
Link: https://lore.kernel.org/r/20250416035444.99569-2-ming.lei@redhat.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Commit 8284066946 ("ublk: grab request reference when the request is handled
by userspace") doesn't grab request reference in case of recovery reissue.
Then the request can be requeued & re-dispatch & failed when canceling
uring command.
If it is one zc request, the request can be freed before io_uring
returns the zc buffer back, then cause kernel panic:
[ 126.773061] BUG: kernel NULL pointer dereference, address: 00000000000000c8
[ 126.773657] #PF: supervisor read access in kernel mode
[ 126.774052] #PF: error_code(0x0000) - not-present page
[ 126.774455] PGD 0 P4D 0
[ 126.774698] Oops: Oops: 0000 [#1] SMP NOPTI
[ 126.775034] CPU: 13 UID: 0 PID: 1612 Comm: kworker/u64:55 Not tainted 6.14.0_blk+ #182 PREEMPT(full)
[ 126.775676] Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS 1.16.3-1.fc39 04/01/2014
[ 126.776275] Workqueue: iou_exit io_ring_exit_work
[ 126.776651] RIP: 0010:ublk_io_release+0x14/0x130 [ublk_drv]
Fixes it by always grabbing request reference for aborting the request.
Reported-by: Caleb Sander Mateos <csander@purestorage.com>
Closes: https://lore.kernel.org/linux-block/CADUfDZodKfOGUeWrnAxcZiLT+puaZX8jDHoj_sfHZCOZwhzz6A@mail.gmail.com/
Fixes: 8284066946 ("ublk: grab request reference when the request is handled by userspace")
Signed-off-by: Ming Lei <ming.lei@redhat.com>
Link: https://lore.kernel.org/r/20250409011444.2142010-2-ming.lei@redhat.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Implement ->queue_rqs() for improving perf in case of MQ.
In this way, we just need to call io_uring_cmd_complete_in_task() once for
whole IO batch, then both io_uring and ublk server can get exact batch from
ublk frontend.
Follows IOPS improvement:
- tests
tools/testing/selftests/ublk/kublk add -t null -q 2 [-z]
fio/t/io_uring -p0 /dev/ublkb0
- results:
more than 10% IOPS boost observed
Pass all ublk selftests, especially the io dispatch order test.
Cc: Uday Shankar <ushankar@purestorage.com>
Signed-off-by: Ming Lei <ming.lei@redhat.com>
Link: https://lore.kernel.org/r/20250327095123.179113-9-ming.lei@redhat.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
