The __u32 len field cannot represent a 4GB buffer (0x100000000
overflows to 0). Change it to __u64 so buffers up to 4GB can be
registered. Add a reserved field for alignment and validate it
is zero.
The kernel enforces a default max of 4GB (UBLK_SHMEM_BUF_SIZE_MAX)
which may be increased in future.
Signed-off-by: Ming Lei <tom.leiming@gmail.com>
Link: https://patch.msgid.link/20260409133020.3780098-2-tom.leiming@gmail.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Add UBLK_F_SHMEM_ZC (1ULL << 19) to the UAPI header and UBLK_F_ALL.
Switch ublk_support_shmem_zc() and ublk_dev_support_shmem_zc() from
returning false to checking the actual flag, enabling the shared
memory zero-copy feature for devices that request it.
Signed-off-by: Ming Lei <ming.lei@redhat.com>
Link: https://patch.msgid.link/20260331153207.3635125-4-ming.lei@redhat.com
[axboe: ublk_buf_reg -> ublk_shmem_buf_reg errors]
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Add control commands for registering and unregistering shared memory
buffers for zero-copy I/O:
- UBLK_U_CMD_REG_BUF (0x18): pins pages from userspace, inserts PFN
ranges into a per-device maple tree for O(log n) lookup during I/O.
Buffer pointers are tracked in a per-device xarray. Returns the
assigned buffer index.
- UBLK_U_CMD_UNREG_BUF (0x19): removes PFN entries and unpins pages.
Queue freeze/unfreeze is handled internally so userspace need not
quiesce the device during registration.
Also adds:
- UBLK_IO_F_SHMEM_ZC flag and addr encoding helpers in UAPI header
(16-bit buffer index supporting up to 65536 buffers)
- Data structures (ublk_buf, ublk_buf_range) and xarray/maple tree
- __ublk_ctrl_reg_buf() helper for PFN insertion with error unwinding
- __ublk_ctrl_unreg_buf() helper for cleanup reuse
- ublk_support_shmem_zc() / ublk_dev_support_shmem_zc() stubs
(returning false — feature not enabled yet)
Signed-off-by: Ming Lei <ming.lei@redhat.com>
Link: https://patch.msgid.link/20260331153207.3635125-2-ming.lei@redhat.com
[axboe: fixup ublk_buf_reg -> ublk_shmem_buf_reg errors, comments]
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Add new feature UBLK_F_BATCH_IO which replaces the following two
per-io commands:
- UBLK_U_IO_FETCH_REQ
- UBLK_U_IO_COMMIT_AND_FETCH_REQ
with three per-queue batch io uring_cmd:
- UBLK_U_IO_PREP_IO_CMDS
- UBLK_U_IO_COMMIT_IO_CMDS
- UBLK_U_IO_FETCH_IO_CMDS
Then ublk can deliver batch io commands to ublk server in single
multishort uring_cmd, also allows to prepare & commit multiple
commands in batch style via single uring_cmd, communication cost is
reduced a lot.
This feature also doesn't limit task context any more for all supported
commands, so any allowed uring_cmd can be issued in any task context.
ublk server implementation becomes much easier.
Meantime load balance becomes much easier to support with this feature.
The command `UBLK_U_IO_FETCH_IO_CMDS` can be issued from multiple task
contexts, so each task can adjust this command's buffer length or number
of inflight commands for controlling how much load is handled by current
task.
Later, priority parameter will be added to command `UBLK_U_IO_FETCH_IO_CMDS`
for improving load balance support.
UBLK_U_IO_NEED_GET_DATA isn't supported in batch io yet, but it may be
enabled in future via its batch pair.
Reviewed-by: Caleb Sander Mateos <csander@purestorage.com>
Signed-off-by: Ming Lei <ming.lei@redhat.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Add UBLK_U_IO_FETCH_IO_CMDS command to enable efficient batch processing
of I/O requests. This multishot uring_cmd allows the ublk server to fetch
multiple I/O commands in a single operation, significantly reducing
submission overhead compared to individual FETCH_REQ* commands.
Key Design Features:
1. Multishot Operation: One UBLK_U_IO_FETCH_IO_CMDS can fetch many I/O
commands, with the batch size limited by the provided buffer length.
2. Dynamic Load Balancing: Multiple fetch commands can be submitted
simultaneously, but only one is active at any time. This enables
efficient load distribution across multiple server task contexts.
3. Implicit State Management: The implementation uses three key variables
to track state:
- evts_fifo: Queue of request tags awaiting processing
- fcmd_head: List of available fetch commands
- active_fcmd: Currently active fetch command (NULL = none active)
States are derived implicitly:
- IDLE: No fetch commands available
- READY: Fetch commands available, none active
- ACTIVE: One fetch command processing events
4. Lockless Reader Optimization: The active fetch command can read from
evts_fifo without locking (single reader guarantee), while writers
(ublk_queue_rq/ublk_queue_rqs) use evts_lock protection. The memory
barrier pairing plays key role for the single lockless reader
optimization.
Implementation Details:
- ublk_queue_rq() and ublk_queue_rqs() save request tags to evts_fifo
- __ublk_acquire_fcmd() selects an available fetch command when
events arrive and no command is currently active
- ublk_batch_dispatch() moves tags from evts_fifo to the fetch command's
buffer and posts completion via io_uring_mshot_cmd_post_cqe()
- State transitions are coordinated via evts_lock to maintain consistency
Reviewed-by: Caleb Sander Mateos <csander@purestorage.com>
Signed-off-by: Ming Lei <ming.lei@redhat.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Handle UBLK_U_IO_COMMIT_IO_CMDS by walking the uring_cmd fixed buffer:
- read each element into one temp buffer in batch style
- parse and apply each element for committing io result
Reviewed-by: Caleb Sander Mateos <csander@purestorage.com>
Signed-off-by: Ming Lei <ming.lei@redhat.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
This commit implements the handling of the UBLK_U_IO_PREP_IO_CMDS command,
which allows userspace to prepare a batch of I/O requests.
The core of this change is the `ublk_walk_cmd_buf` function, which iterates
over the elements in the uring_cmd fixed buffer. For each element, it parses
the I/O details, finds the corresponding `ublk_io` structure, and prepares it
for future dispatch.
Add per-io lock for protecting concurrent delivery and committing.
Reviewed-by: Caleb Sander Mateos <csander@purestorage.com>
Signed-off-by: Ming Lei <ming.lei@redhat.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Add new command UBLK_U_IO_PREP_IO_CMDS, which is the batch version of
UBLK_IO_FETCH_REQ.
Add new command UBLK_U_IO_COMMIT_IO_CMDS, which is for committing io command
result only, still the batch version.
The new command header type is `struct ublk_batch_io`.
This patch doesn't actually implement these commands yet, just validates the
SQE fields.
Reviewed-by: Caleb Sander Mateos <csander@purestorage.com>
Signed-off-by: Ming Lei <ming.lei@redhat.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Add a best-effort stop command, UBLK_CMD_TRY_STOP_DEV, which only stops a
ublk device when it has no active openers.
Unlike UBLK_CMD_STOP_DEV, this command does not disrupt existing users.
New opens are blocked only after disk_openers has reached zero; if the
device is busy, the command returns -EBUSY and leaves it running.
The ub->block_open flag is used only to close a race with an in-progress
open and does not otherwise change open behavior.
Advertise support via the UBLK_F_SAFE_STOP_DEV feature flag.
Signed-off-by: Yoav Cohen <yoav@nvidia.com>
Reviewed-by: Ming Lei <ming.lei@redhat.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Add a function ublk_copy_user_integrity() to copy integrity information
between a request and a user iov_iter. This mirrors the existing
ublk_copy_user_pages() but operates on request integrity data instead of
regular data. Check UBLKSRV_IO_INTEGRITY_FLAG in iocb->ki_pos in
ublk_user_copy() to choose between copying data or integrity data.
[csander: change offset units from data bytes to integrity data bytes,
fix CONFIG_BLK_DEV_INTEGRITY=n build, rebase on user copy refactor]
Signed-off-by: Stanley Zhang <stazhang@purestorage.com>
Signed-off-by: Caleb Sander Mateos <csander@purestorage.com>
Reviewed-by: Ming Lei <ming.lei@redhat.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Indicate to the ublk server when an incoming request has integrity data
by setting UBLK_IO_F_INTEGRITY in the ublksrv_io_desc's op_flags field.
Signed-off-by: Caleb Sander Mateos <csander@purestorage.com>
Reviewed-by: Ming Lei <ming.lei@redhat.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Add a feature flag UBLK_F_INTEGRITY for a ublk server to request
integrity/metadata support when creating a ublk device. The ublk server
can also check for the feature flag on the created device or the result
of UBLK_U_CMD_GET_FEATURES to tell if the ublk driver supports it.
UBLK_F_INTEGRITY requires UBLK_F_USER_COPY, as user copy is the only
data copy mode initially supported for integrity data.
Add UBLK_PARAM_TYPE_INTEGRITY and struct ublk_param_integrity to struct
ublk_params to specify the integrity params of a ublk device.
UBLK_PARAM_TYPE_INTEGRITY requires UBLK_F_INTEGRITY and a nonzero
metadata_size. The LBMD_PI_CAP_* and LBMD_PI_CSUM_* values from the
linux/fs.h UAPI header are used for the flags and csum_type fields.
If the UBLK_PARAM_TYPE_INTEGRITY flag is set, validate the integrity
parameters and apply them to the blk_integrity limits.
The struct ublk_param_integrity validations are based on the checks in
blk_validate_integrity_limits(). Any invalid parameters should be
rejected before being applied to struct blk_integrity.
[csander: drop redundant pi_tuple_size field, use block metadata UAPI
constants, add param validation]
Signed-off-by: Stanley Zhang <stazhang@purestorage.com>
Signed-off-by: Caleb Sander Mateos <csander@purestorage.com>
Reviewed-by: Ming Lei <ming.lei@redhat.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Currently, UBLK_IO_REGISTER_IO_BUF and UBLK_IO_UNREGISTER_IO_BUF are
only permitted on the ublk_io's daemon task. But this restriction is
unnecessary. ublk_register_io_buf() calls __ublk_check_and_get_req() to
look up the request from the tagset and atomically take a reference on
the request without accessing the ublk_io. ublk_unregister_io_buf()
doesn't use the q_id or tag at all.
So allow these opcodes even on tasks other than io->task.
Handle UBLK_IO_UNREGISTER_IO_BUF before obtaining the ubq and io since
the buffer index being unregistered is not necessarily related to the
specified q_id and tag.
Add a feature flag UBLK_F_BUF_REG_OFF_DAEMON that userspace can use to
determine whether the kernel supports off-daemon buffer registration.
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/20250620151008.3976463-10-csander@purestorage.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
UBLK_F_SUPPORT_ZERO_COPY has a very old comment describing the initial
idea for how zero-copy would be implemented. The actual implementation
added in commit 1f6540e2aa ("ublk: zc register/unregister bvec") uses
io_uring registered buffers rather than shared memory mapping.
Remove the inaccurate remarks about mapping ublk request memory into the
ublk server's address space and requiring 4K block size. Replace them
with a description of the current zero-copy mechanism.
Signed-off-by: Caleb Sander Mateos <csander@purestorage.com>
Reviewed-by: Ming Lei <ming.lei@redhat.com>
Link: https://lore.kernel.org/r/20250621171015.354932-1-csander@purestorage.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
When a C++ file compiled with -Wc++11-narrowing includes the UAPI header
linux/ublk_cmd.h, ublk_sqe_addr_to_auto_buf_reg()'s assignments of u64
values to u8, u16, and u32 fields result in compiler warnings. Add
explicit casts to the intended types to avoid these warnings. Drop the
unnecessary bitmasks.
Reported-by: Uday Shankar <ushankar@purestorage.com>
Signed-off-by: Caleb Sander Mateos <csander@purestorage.com>
Fixes: 99c1e4eb6a ("ublk: register buffer to local io_uring with provided buf index via UBLK_F_AUTO_BUF_REG")
Reviewed-by: Ming Lei <ming.lei@redhat.com>
Link: https://lore.kernel.org/r/20250621162842.337452-1-csander@purestorage.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
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>
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>
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>
IO split is usually bad in io_uring world, since -EAGAIN is caused and
IO handling may have to fallback to io-wq, this way does hurt performance.
ublk starts to support zero copy recently, for avoiding unnecessary IO
split, ublk driver's segment limit should be aligned with backend
device's segment limit.
Another reason is that io_buffer_register_bvec() needs to allocate bvecs,
which number is aligned with ublk request segment number, so that big
memory allocation can be avoided by setting reasonable max_segments limit.
So add segment parameter for providing ublk server chance to align
segment limit with backend, and keep it reasonable from implementation
viewpoint.
Signed-off-by: Ming Lei <ming.lei@redhat.com>
Link: https://lore.kernel.org/r/20250327095123.179113-7-ming.lei@redhat.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Pull block updates from Jens Axboe:
- Fixes for integrity handling
- NVMe pull request via Keith:
- Secure concatenation for TCP transport (Hannes)
- Multipath sysfs visibility (Nilay)
- Various cleanups (Qasim, Baruch, Wang, Chen, Mike, Damien, Li)
- Correct use of 64-bit BARs for pci-epf target (Niklas)
- Socket fix for selinux when used in containers (Peijie)
- MD pull request via Yu:
- fix recovery can preempt resync (Li Nan)
- fix md-bitmap IO limit (Su Yue)
- fix raid10 discard with REQ_NOWAIT (Xiao Ni)
- fix raid1 memory leak (Zheng Qixing)
- fix mddev uaf (Yu Kuai)
- fix raid1,raid10 IO flags (Yu Kuai)
- some refactor and cleanup (Yu Kuai)
- Series cleaning up and fixing bugs in the bad block handling code
- Improve support for write failure simulation in null_blk
- Various lock ordering fixes
- Fixes for locking for debugfs attributes
- Various ublk related fixes and improvements
- Cleanups for blk-rq-qos wait handling
- blk-throttle fixes
- Fixes for loop dio and sync handling
- Fixes and cleanups for the auto-PI code
- Block side support for hardware encryption keys in blk-crypto
- Various cleanups and fixes
* tag 'for-6.15/block-20250322' of git://git.kernel.dk/linux: (105 commits)
nvmet: replace max(a, min(b, c)) by clamp(val, lo, hi)
nvme-tcp: fix selinux denied when calling sock_sendmsg
nvmet: pci-epf: Always configure BAR0 as 64-bit
nvmet: Remove duplicate uuid_copy
nvme: zns: Simplify nvme_zone_parse_entry()
nvmet: pci-epf: Remove redundant 'flush_workqueue()' calls
nvmet-fc: Remove unused functions
nvme-pci: remove stale comment
nvme-fc: Utilise min3() to simplify queue count calculation
nvme-multipath: Add visibility for queue-depth io-policy
nvme-multipath: Add visibility for numa io-policy
nvme-multipath: Add visibility for round-robin io-policy
nvmet: add tls_concat and tls_key debugfs entries
nvmet-tcp: support secure channel concatenation
nvmet: Add 'sq' argument to alloc_ctrl_args
nvme-fabrics: reset admin connection for secure concatenation
nvme-tcp: request secure channel concatenation
nvme-keyring: add nvme_tls_psk_refresh()
nvme: add nvme_auth_derive_tls_psk()
nvme: add nvme_auth_generate_digest()
...
The in-tree ublk driver doesn't need DMA alignment limit because there
is one data copy between request pages and the userspace buffer.
However, ublk is going to support zero copy, then DMA alignment limit
is required, because same IO buffer is forwarded to backend which may
have specific buffer DMA alignment limit, so the limit has to be exposed
from the frontend driver to client application.
Cc: Keith Busch <kbusch@kernel.org>
Signed-off-by: Ming Lei <ming.lei@redhat.com>
Link: https://lore.kernel.org/r/20250227103707.2640014-1-ming.lei@redhat.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Ming LeiandJens Axboe
Ming LeiandJens Axboe
Yoav CohenandJens Axboe
Stanley ZhangandJens Axboe
Caleb Sander MateosandJens Axboe
Uday ShankarandJens Axboe
Omri MannandJens Axboe
Linus Torvalds