In order to provide a clearer, more symmetric API for pinning and
unpinning DMA pages. This way, pin_user_pages*() calls match up with
unpin_user_pages*() calls, and the API is a lot closer to being
self-explanatory.
Link: http://lkml.kernel.org/r/20200107224558.2362728-23-jhubbard@nvidia.com
Signed-off-by: John Hubbard <jhubbard@nvidia.com>
Reviewed-by: Jan Kara <jack@suse.cz>
Cc: Alex Williamson <alex.williamson@redhat.com>
Cc: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com>
Cc: Björn Töpel <bjorn.topel@intel.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: Dan Williams <dan.j.williams@intel.com>
Cc: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Cc: Ira Weiny <ira.weiny@intel.com>
Cc: Jason Gunthorpe <jgg@mellanox.com>
Cc: Jason Gunthorpe <jgg@ziepe.ca>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Jerome Glisse <jglisse@redhat.com>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Kirill A. Shutemov <kirill@shutemov.name>
Cc: Leon Romanovsky <leonro@mellanox.com>
Cc: Mauro Carvalho Chehab <mchehab@kernel.org>
Cc: Mike Rapoport <rppt@linux.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Convert fs/io_uring to use the new pin_user_pages() call, which sets
FOLL_PIN. Setting FOLL_PIN is now required for code that requires
tracking of pinned pages, and therefore for any code that calls
put_user_page().
In partial anticipation of this work, the io_uring code was already
calling put_user_page() instead of put_page(). Therefore, in order to
convert from the get_user_pages()/put_page() model, to the
pin_user_pages()/put_user_page() model, the only change required here is
to change get_user_pages() to pin_user_pages().
Link: http://lkml.kernel.org/r/20200107224558.2362728-17-jhubbard@nvidia.com
Signed-off-by: John Hubbard <jhubbard@nvidia.com>
Reviewed-by: Jens Axboe <axboe@kernel.dk>
Reviewed-by: Jan Kara <jack@suse.cz>
Cc: Alex Williamson <alex.williamson@redhat.com>
Cc: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com>
Cc: Björn Töpel <bjorn.topel@intel.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: Dan Williams <dan.j.williams@intel.com>
Cc: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Cc: Ira Weiny <ira.weiny@intel.com>
Cc: Jason Gunthorpe <jgg@mellanox.com>
Cc: Jason Gunthorpe <jgg@ziepe.ca>
Cc: Jerome Glisse <jglisse@redhat.com>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Kirill A. Shutemov <kirill@shutemov.name>
Cc: Leon Romanovsky <leonro@mellanox.com>
Cc: Mauro Carvalho Chehab <mchehab@kernel.org>
Cc: Mike Rapoport <rppt@linux.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
With nesting of anonymous unions and structs it's hard to
review layout changes. It's better to ask the compiler
for these things.
Signed-off-by: Stefan Metzmacher <metze@samba.org>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
It can be hard to know exactly what is registered with the ring.
Especially for credentials, it'd be handy to be able to see which
ones are registered, what personalities they have, and what the ID
of each of them is.
This adds support for showing information registered in the ring from
the fdinfo of the io_uring fd. Here's an example from a test case that
registers 4 files (two of them sparse), 4 buffers, and 2 personalities:
pos: 0
flags: 02000002
mnt_id: 14
UserFiles: 4
0: file-no-1
1: file-no-2
2: <none>
3: <none>
UserBufs: 4
0: 0x563817c46000/128
1: 0x563817c47000/256
2: 0x563817c48000/512
3: 0x563817c49000/1024
Personalities:
1
Uid: 0 0 0 0
Gid: 0 0 0 0
Groups: 0
CapEff: 0000003fffffffff
2
Uid: 0 0 0 0
Gid: 0 0 0 0
Groups: 0
CapEff: 0000003fffffffff
Suggested-by: Jann Horn <jannh@google.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
-----BEGIN PGP SIGNATURE-----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=ZZUI
-----END PGP SIGNATURE-----
Merge tag 'for-5.6/io_uring-vfs-2020-01-29' of git://git.kernel.dk/linux-block
Pull io_uring updates from Jens Axboe:
- Support for various new opcodes (fallocate, openat, close, statx,
fadvise, madvise, openat2, non-vectored read/write, send/recv, and
epoll_ctl)
- Faster ring quiesce for fileset updates
- Optimizations for overflow condition checking
- Support for max-sized clamping
- Support for probing what opcodes are supported
- Support for io-wq backend sharing between "sibling" rings
- Support for registering personalities
- Lots of little fixes and improvements
* tag 'for-5.6/io_uring-vfs-2020-01-29' of git://git.kernel.dk/linux-block: (64 commits)
io_uring: add support for epoll_ctl(2)
eventpoll: support non-blocking do_epoll_ctl() calls
eventpoll: abstract out epoll_ctl() handler
io_uring: fix linked command file table usage
io_uring: support using a registered personality for commands
io_uring: allow registering credentials
io_uring: add io-wq workqueue sharing
io-wq: allow grabbing existing io-wq
io_uring/io-wq: don't use static creds/mm assignments
io-wq: make the io_wq ref counted
io_uring: fix refcounting with batched allocations at OOM
io_uring: add comment for drain_next
io_uring: don't attempt to copy iovec for READ/WRITE
io_uring: honor IOSQE_ASYNC for linked reqs
io_uring: prep req when do IOSQE_ASYNC
io_uring: use labeled array init in io_op_defs
io_uring: optimise sqe-to-req flags translation
io_uring: remove REQ_F_IO_DRAINED
io_uring: file switch work needs to get flushed on exit
io_uring: hide uring_fd in ctx
...
We're not consistent in how the file table is grabbed and assigned if we
have a command linked that requires the use of it.
Add ->file_table to the io_op_defs[] array, and use that to determine
when to grab the table instead of having the handlers set it if they
need to defer. This also means we can kill the IO_WQ_WORK_NEEDS_FILES
flag. We always initialize work->files, so io-wq can just check for
that.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
For personalities previously registered via IORING_REGISTER_PERSONALITY,
allow any command to select them. This is done through setting
sqe->personality to the id returned from registration, and then flagging
sqe->flags with IOSQE_PERSONALITY.
Reviewed-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
If an application wants to use a ring with different kinds of
credentials, it can register them upfront. We don't lookup credentials,
the credentials of the task calling IORING_REGISTER_PERSONALITY is used.
An 'id' is returned for the application to use in subsequent personality
support.
Reviewed-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
If IORING_SETUP_ATTACH_WQ is set, it expects wq_fd in io_uring_params to
be a valid io_uring fd io-wq of which will be shared with the newly
created io_uring instance. If the flag is set but it can't share io-wq,
it fails.
This allows creation of "sibling" io_urings, where we prefer to keep the
SQ/CQ private, but want to share the async backend to minimize the amount
of overhead associated with having multiple rings that belong to the same
backend.
Reported-by: Jens Axboe <axboe@kernel.dk>
Reported-by: Daurnimator <quae@daurnimator.com>
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
We currently setup the io_wq with a static set of mm and creds. Even for
a single-use io-wq per io_uring, this is suboptimal as we have may have
multiple enters of the ring. For sharing the io-wq backend, it doesn't
work at all.
Switch to passing in the creds and mm when the work item is setup. This
means that async work is no longer deferred to the io_uring mm and creds,
it is done with the current mm and creds.
Flag this behavior with IORING_FEAT_CUR_PERSONALITY, so applications know
they can rely on the current personality (mm and creds) being the same
for direct issue and async issue.
Reviewed-by: Stefan Metzmacher <metze@samba.org>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
In case of out of memory the second argument of percpu_ref_put_many() in
io_submit_sqes() may evaluate into "nr - (-EAGAIN)", that is clearly
wrong.
Fixes: 2b85edfc0c ("io_uring: batch getting pcpu references")
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Draining the middle of a link is tricky, so leave a comment there
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
For the non-vectored variant of READV/WRITEV, we don't need to setup an
async io context, and we flag that appropriately in the io_op_defs
array. However, in fixing this for the 5.5 kernel in commit 74566df3a7
we didn't have these opcodes, so the check there was added just for the
READ_FIXED and WRITE_FIXED opcodes. Replace that check with just a
single check for needing async context, that covers all four of these
read/write variants that don't use an iovec.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
If we're sharing the ring across forks, then one process exiting means
that we cancel ALL work and prevent future work. This is overly
restrictive. As long as we cancel the work associated with the files
from the current task, it's safe to let others persist. Normal fd close
on exit will still wait (and cancel) pending work.
Fixes: fcb323cc53 ("io_uring: io_uring: add support for async work inheriting files")
Reported-by: Andres Freund <andres@anarazel.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
This ends up being too restrictive for tasks that willingly fork and
share the ring between forks. Andres reports that this breaks his
postgresql work. Since we're close to 5.5 release, revert this change
for now.
Cc: stable@vger.kernel.org
Fixes: 44d282796f ("io_uring: only allow submit from owning task")
Reported-by: Andres Freund <andres@anarazel.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
REQ_F_FORCE_ASYNC is checked only for the head of a link. Fix it.
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Whenever IOSQE_ASYNC is set, requests will be punted to async without
getting into io_issue_req() and without proper preparation done (e.g.
io_req_defer_prep()). Hence they will be left uninitialised.
Prepare them before punting.
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Don't rely on implicit ordering of IORING_OP_ and explicitly place them
at a right place in io_op_defs. Now former comments are now a part of
the code and won't ever outdate.
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
For each IOSQE_* flag there is a corresponding REQ_F_* flag. And there
is a repetitive pattern of their translation:
e.g. if (sqe->flags & SQE_FLAG*) req->flags |= REQ_F_FLAG*
Use same numeric values/bits for them and copy instead of manual
handling.
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
A request can get into the defer list only once, there is no need for
marking it as drained, so remove it. This probably was left after
extracting __need_defer() for use in timeouts.
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
We currently flush early, but if we have something in progress and a
new switch is scheduled, we need to ensure to flush after our teardown
as well.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
req->ring_fd and req->ring_file are used only during the prep stage
during submission, which is is protected by mutex. There is no need
to store them per-request, place them in ctx.
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
__io_commit_cqring() is almost always called when there is a change in
the rings, so the check is rather pessimising.
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Move setting ctx->drain_next to the only place it could be set, when it
got linked non-head requests. The same for checking it, it's interesting
only for a head of a link or a non-linked request.
No functional changes here. This removes some code from the common path
and also removes REQ_F_DRAIN_LINK flag, as it doesn't need it anymore.
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
The application currently has no way of knowing if a given opcode is
supported or not without having to try and issue one and see if we get
-EINVAL or not. And even this approach is fraught with peril, as maybe
we're getting -EINVAL due to some fields being missing, or maybe it's
just not that easy to issue that particular command without doing some
other leg work in terms of setup first.
This adds IORING_REGISTER_PROBE, which fills in a structure with info
on what it supported or not. This will work even with sparse opcode
fields, which may happen in the future or even today if someone
backports specific features to older kernels.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
We can't assume that the whole batch has fixed files in it. If it's a
mix, or none at all, then we can end up doing a ref put that either
messes up accounting, or causes an oops if we have no fixed files at
all.
Also ensure we free requests properly between inflight accounted and
normal requests.
Fixes: 82c721577011 ("io_uring: extend batch freeing to cover more cases")
Reported-by: Dmitrii Dolgov <9erthalion6@gmail.com>
Reported-by: Pavel Begunkov <asml.silence@gmail.com>
Tested-by: Dmitrii Dolgov <9erthalion6@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
For some test apps at least, user_data is just zeroes. So it's not a
good way to tell what the command actually is. Add the opcode to the
issue trace point.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Add support for the new openat2(2) system call. It's trivial to do, as
we can have openat(2) just be wrapped around it.
Suggested-by: Stefan Metzmacher <metze@samba.org>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
We only use it internally in the prep functions for both statx and
openat, so we don't need it to be persistent across the request.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
If an application is using eventfd notifications with poll to know when
new SQEs can be issued, it's expecting the following read/writes to
complete inline. And with that, it knows that there are events available,
and don't want spurious wakeups on the eventfd for those requests.
This adds IORING_REGISTER_EVENTFD_ASYNC, which works just like
IORING_REGISTER_EVENTFD, except it only triggers notifications for events
that happen from async completions (IRQ, or io-wq worker completions).
Any completions inline from the submission itself will not trigger
notifications.
Suggested-by: Mark Papadakis <markuspapadakis@icloud.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
In preparation for adding another one, which would make us spill into
another long (and hence bump the size of the ctx), change them to
bit fields.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
If an application attempts to register a set with unbounded requests
pending, we can be stuck here forever if they don't complete. We can
make this wait interruptible, and just abort if we get signaled.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Null check kfree is redundant, so remove it.
This is detected by coccinelle.
Signed-off-by: YueHaibing <yuehaibing@huawei.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
io_wq workers use io_issue_sqe() to forward sqes and never
io_queue_sqe(). Remove extra check for io_wq_current_is_worker()
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
It should be pretty rare to not submitting anything when there is
something in the ring. No need to keep heuristics for this case.
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
A user may ask to submit more than there is in the ring, and then
io_uring will submit as much as it can. However, in the last iteration
it will allocate an io_kiocb and immediately free it. It could do
better and adjust @to_submit to what is in the ring.
And since the ring's head is already checked here, there is no need to
do it in the loop, spamming with smp_load_acquire()'s barriers
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Make io_submit_sqes() to clamp @to_submit itself. It removes duplicated
code and prepares for following changes.
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Some applications like to start small in terms of ring size, and then
ramp up as needed. This is a bit tricky to do currently, since we don't
advertise the max ring size.
This adds IORING_SETUP_CLAMP. If set, and the values for SQ or CQ ring
size exceed what we support, then clamp them at the max values instead
of returning -EINVAL. Since we return the chosen ring sizes after setup,
no further changes are needed on the application side. io_uring already
changes the ring sizes if the application doesn't ask for power-of-two
sizes, for example.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Currently we only batch free if fixed files are used, no links, no aux
data, etc. This extends the batch freeing to only exclude the linked
case and fallback case, and make io_free_req_many() handle the other
cases just fine.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
percpu_ref_tryget() has its own overhead. Instead getting a reference
for each request, grab a bunch once per io_submit_sqes().
~5% throughput boost for a "submit and wait 128 nops" benchmark.
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
__io_req_free_empty() -> __io_req_do_free()
Signed-off-by: Jens Axboe <axboe@kernel.dk>
This adds support for doing madvise(2) through io_uring. We assume that
any operation can block, and hence punt everything async. This could be
improved, but hard to make bullet proof. The async punt ensures it's
safe.
Reviewed-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
This adds support for doing fadvise through io_uring. We assume that
WILLNEED doesn't block, but that DONTNEED may block.
Reviewed-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
This behaves like preadv2/pwritev2 with offset == -1, it'll use (and
update) the current file position. This obviously comes with the caveat
that if the application has multiple read/writes in flight, then the
end result will not be as expected. This is similar to threads sharing
a file descriptor and doing IO using the current file position.
Since this feature isn't easily detectable by doing a read or write,
add a feature flags, IORING_FEAT_RW_CUR_POS, to allow applications to
detect presence of this feature.
Reported-by: 李通洲 <carter.li@eoitek.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
For uses cases that don't already naturally have an iovec, it's easier
(or more convenient) to just use a buffer address + length. This is
particular true if the use case is from languages that want to create
a memory safe abstraction on top of io_uring, and where introducing
the need for the iovec may impose an ownership issue. For those cases,
they currently need an indirection buffer, which means allocating data
just for this purpose.
Add basic read/write that don't require the iovec.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
For busy IORING_OP_POLL_ADD workloads, we can have enough contention
on the completion lock that we fail the inline completion path quite
often as we fail the trylock on that lock. Add a list for deferred
completions that we can use in that case. This helps reduce the number
of async offloads we have to do, as if we get multiple completions in
a row, we'll piggy back on to the poll_llist instead of having to queue
our own offload.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
We currently check ->cq_overflow_list from both SQ and CQ context, which
causes some bouncing of that cache line. Add separate bits of state for
this instead, so that the SQ side can check using its own state, and
likewise for the CQ side.
This adds ->sq_check_overflow with the SQ state, and ->cq_check_overflow
with the CQ state. If we hit an overflow condition, both of these bits
are set. Likewise for overflow flush clear, we clear both bits. For the
fast path of just checking if there's an overflow condition on either
the SQ or CQ side, we can use our own private bit for this.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
We currently have various switch statements that check if an opcode needs
a file, mm, etc. These are hard to keep in sync as opcodes are added. Add
a struct io_op_def that holds all of this information, so we have just
one spot to update when opcodes are added.
This also enables us to NOT allocate req->io if a deferred command
doesn't need it, and corrects some mistakes we had in terms of what
commands need mm context.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
__io_free_req() and io_double_put_req() aren't used before they are
defined, so we can kill these two forwards.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Move io_queue_link_head() to links handling code in io_submit_sqe(),
so it wouldn't need extra checks and would have better data locality.
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Calling "prev" a head of a link is a bit misleading. Rename it
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
io_uring defaults to always doing inline submissions, if at all
possible. But for larger copies, even if the data is fully cached, that
can take a long time. Add an IOSQE_ASYNC flag that the application can
set on the SQE - if set, it'll ensure that we always go async for those
kinds of requests. Use the io-wq IO_WQ_WORK_CONCURRENT flag to ensure we
get the concurrency we desire for this case.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
We currently fully quiesce the ring before an unregister or update of
the fixed fileset. This is very expensive, and we can be a bit smarter
about this.
Add a percpu refcount for the file tables as a whole. Grab a percpu ref
when we use a registered file, and put it on completion. This is cheap
to do. Upon removal of a file from a set, switch the ref count to atomic
mode. When we hit zero ref on the completion side, then we know we can
drop the previously registered files. When the old files have been
dropped, switch the ref back to percpu mode for normal operation.
Since there's a period between doing the update and the kernel being
done with it, add a IORING_OP_FILES_UPDATE opcode that can perform the
same action. The application knows the update has completed when it gets
the CQE for it. Between doing the update and receiving this completion,
the application must continue to use the unregistered fd if submitting
IO on this particular file.
This takes the runtime of test/file-register from liburing from 14s to
about 0.7s.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
This works just like close(2), unsurprisingly. We remove the file
descriptor and post the completion inline, then offload the actual
(potential) last file put to async context.
Mark the async part of this work as uncancellable, as we really must
guarantee that the latter part of the close is run.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Not all work can be cancelled, some of it we may need to guarantee
that it runs to completion. Allow the caller to set IO_WQ_WORK_NO_CANCEL
on work that must not be cancelled. Note that the caller work function
must also check for IO_WQ_WORK_NO_CANCEL on work that is marked
IO_WQ_WORK_CANCEL.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
This works just like openat(2), except it can be performed async. For
the normal case of a non-blocking path lookup this will complete
inline. If we have to do IO to perform the open, it'll be done from
async context.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
fds field of struct io_uring_files_update is problematic with regards
to compat user space, as pointer size is different in 32-bit, 32-on-64-bit,
and 64-bit user space. In order to avoid custom handling of compat in
the syscall implementation, make fds __u64 and use u64_to_user_ptr in
order to retrieve it. Also, align the field naturally and check that
no garbage is passed there.
Fixes: c3a31e6056 ("io_uring: add support for IORING_REGISTER_FILES_UPDATE")
Signed-off-by: Eugene Syromiatnikov <esyr@redhat.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
If the credentials or the mm doesn't match, don't allow the task to
submit anything on behalf of this ring. The task that owns the ring can
pass the file descriptor to another task, but we don't want to allow
that task to submit an SQE that then assumes the ring mm and creds if
it needs to go async.
Cc: stable@vger.kernel.org
Suggested-by: Stefan Metzmacher <metze@samba.org>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
A previous commit moved the locking for the async sqthread, but didn't
take into account that the io-wq workers still need it. We can't use
req->in_async for this anymore as both the sqthread and io-wq workers
set it, gate the need for locking on io_wq_current_is_worker() instead.
Fixes: 8a4955ff1c ("io_uring: sqthread should grab ctx->uring_lock for submissions")
Reported-by: Bijan Mottahedeh <bijan.mottahedeh@oracle.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
req->result is cleared when io_issue_sqe() calls io_read/write_pre()
routines. Those routines however are not called when the sqe
argument is NULL, which is the case when io_issue_sqe() is called from
io_wq_submit_work(). io_issue_sqe() may then examine a stale result if
a polled request had previously failed with -EAGAIN:
if (ctx->flags & IORING_SETUP_IOPOLL) {
if (req->result == -EAGAIN)
return -EAGAIN;
io_iopoll_req_issued(req);
}
and in turn cause a subsequently completed request to be re-issued in
io_wq_submit_work().
Signed-off-by: Bijan Mottahedeh <bijan.mottahedeh@oracle.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
If we pass back dependent work in case of links, we need to always
ensure that we call the link setup and work prep handler. If not, we
might be missing some setup for the next work item.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
We don't need it, and if we have it, then the retry handler will attempt
to copy the non-existent iovec with the inline iovec, with a segment
count that doesn't make sense.
Fixes: f67676d160 ("io_uring: ensure async punted read/write requests copy iovec")
Reported-by: Jonathan Lemon <jonathan.lemon@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
We currently punt any short read on a regular file to async context,
but this fails if the short read is due to running into EOF. This is
especially problematic since we only do the single prep for commands
now, as we don't reset kiocb->ki_pos. This can result in a 4k read on
a 1k file returning zero, as we detect the short read and then retry
from async context. At the time of retry, the position is now 1k, and
we end up reading nothing, and hence return 0.
Instead of trying to patch around the fact that short reads can be
legitimate and won't succeed in case of retry, remove the logic to punt
a short read to async context. Simply return it.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
This moves the prep handlers outside of the opcode handlers, and allows
us to pass in the sqe directly. If the sqe is non-NULL, it means that
the request should be prepared for the first time.
With the opcode handlers not having access to the sqe at all, we are
guaranteed that the prep handler has setup the request fully by the
time we get there. As before, for opcodes that need to copy in more
data then the io_kiocb allows for, the io_async_ctx holds that info. If
a prep handler is invoked with req->io set, it must use that to retain
information for later.
Finally, we can remove io_kiocb->sqe as well.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
We currently have a mix of use cases. Most of the newer ones are pretty
uniform, but we have some older ones that use different calling
calling conventions. This is confusing.
For the opcodes that currently rely on the req->io->sqe copy saving
them from reuse, add a request type struct in the io_kiocb command
union to store the data they need.
Prepare for all opcodes having a standard prep method, so we can call
it in a uniform fashion and outside of the opcode handler. This is in
preparation for passing in the 'sqe' pointer, rather than storing it
in the io_kiocb. Once we have uniform prep handlers, we can leave all
the prep work to that part, and not even pass in the sqe to the opcode
handler. This ensures that we don't reuse sqe data inadvertently.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Add the count field to struct io_timeout, and ensure the prep handler
has read it. Timeout also needs an async context always, set it up
in the prep handler if we don't have one.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Add struct io_sr_msg in our io_kiocb per-command union, and ensure that
the send/recvmsg prep handlers have grabbed what they need from the SQE
by the time prep is done.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Add struct io_connect in our io_kiocb per-command union, and ensure
that io_connect_prep() has grabbed what it needs from the SQE.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Put the kiocb in struct io_rw, and add the addr/len for the request as
well. Use the kiocb->private field for the buffer index for fixed reads
and writes.
Any use of kiocb->ki_filp is flipped to req->file. It's the same thing,
and less confusing.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
We use it in some spots, but not consistently. Convert the rest over,
makes it easier to read as well.
No functional changes in this patch.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
I've been chasing a weird and obscure crash that was userspace stack
corruption, and finally narrowed it down to a bit flip that made a
stack address invalid. io_wq_submit_work() unconditionally flips
the req->rw.ki_flags IOCB_NOWAIT bit, but since it's a generic work
handler, this isn't valid. Normal read/write operations own that
part of the request, on other types it could be something else.
Move the IOCB_NOWAIT clear to the read/write handlers where it belongs.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
There is no reliable way to submit and wait in a single syscall, as
io_submit_sqes() may under-consume sqes (in case of an early error).
Then it will wait for not-yet-submitted requests, deadlocking the user
in most cases.
Don't wait/poll if can't submit all sqes
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Now that we have all the opcodes handled in terms of command prep and
SQE reuse, add a printk_once() to warn about any potentially new and
unhandled ones.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
If we defer a request, we can't be reading the opcode again. Ensure that
the user_data and opcode fields are stable. For the user_data we already
have a place for it, for the opcode we can fill a one byte hold and store
that as well. For both of them, assign them when we originally read the
SQE in io_get_sqring(). Any code that uses sqe->opcode or sqe->user_data
is switched to req->opcode and req->user_data.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
If we defer this command as part of a link, we have to make sure that
the SQE data has been read upfront. Integrate the timeout remove op into
the prep handling to make it safe for SQE reuse.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
If we defer this command as part of a link, we have to make sure that
the SQE data has been read upfront. Integrate the async cancel op into
the prep handling to make it safe for SQE reuse.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
If we defer these commands as part of a link, we have to make sure that
the SQE data has been read upfront. Integrate the poll add/remove into
the prep handling to make it safe for SQE reuse.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
The rules are as follows, if IOSQE_IO_HARDLINK is specified, then it's a
link and there is no need to set IOSQE_IO_LINK separately, though it
could be there. Add proper check and ensure that IOSQE_IO_HARDLINK
implies IOSQE_IO_LINK.
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
We're currently not retaining sqe data for accept, fsync, and
sync_file_range. None of these commands need data outside of what
is directly provided, hence it can't go stale when the request is
deferred. However, it can get reused, if an application reuses
SQE entries.
Ensure that we retain the information we need and only read the sqe
contents once, off the submission path. Most of this is just moving
code into a prep and finish function.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
We pass in req->sqe for all of them, no need to pass it in as the
request is always passed in. This is a necessary prep patch to be
able to cleanup/fix the request prep path.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Some of these code paths assume that any force_nonblock == true issue
is not prepped, but that's not true if we did prep as part of link setup
earlier. Check if we already have an async context allocate before
setting up a new one.
Cleanup the async context setup in general, we have a lot of duplicated
code there.
Fixes: 03b1230ca1 ("io_uring: ensure async punted sendmsg/recvmsg requests copy data")
Fixes: f67676d160 ("io_uring: ensure async punted read/write requests copy iovec")
Signed-off-by: Jens Axboe <axboe@kernel.dk>
If we have to punt the recvmsg to async context, we copy all the
context. But since the iovec used can be either on-stack (if small) or
dynamically allocated, if it's on-stack, then we need to ensure we reset
the iov pointer. If we don't, then we're reusing old stack data, and
that can lead to -EFAULTs if things get overwritten.
Ensure we retain the right pointers for the iov, and free it as well if
we end up having to go beyond UIO_FASTIOV number of vectors.
Fixes: 03b1230ca1 ("io_uring: ensure async punted sendmsg/recvmsg requests copy data")
Reported-by: 李通洲 <carter.li@eoitek.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
- Fix a few typos found while reading the code.
- Fix stale io_get_sqring comment referencing s->sqe, the 's' parameter
was renamed to 'req', but the comment still holds.
Signed-off-by: Brian Gianforcaro <b.gianfo@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
If we submit an unknown opcode and have fd == -1, io_op_needs_file()
will return true as we default to needing a file. Then when we go and
assign the file, we find the 'fd' invalid and return -EBADF. We really
should be returning -EINVAL for that case, as we normally do for
unsupported opcodes.
Change io_op_needs_file() to have the following return values:
0 - does not need a file
1 - does need a file
< 0 - error value
and use this to pass back the right value for this invalid case.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
In chasing a performance issue between using IORING_OP_RECVMSG and
IORING_OP_READV on sockets, tracing showed that we always punt the
socket reads to async offload. This is due to io_file_supports_async()
not checking for S_ISSOCK on the inode. Since sockets supports the
O_NONBLOCK (or MSG_DONTWAIT) flag just fine, add sockets to the list
of file types that we can do a non-blocking issue to.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
We hash regular files to avoid having multiple threads hammer on the
inode mutex, but it should not be needed on other types of files
(like sockets).
Signed-off-by: Jens Axboe <axboe@kernel.dk>
One major use case of linked commands is the ability to run the next
link inline, if at all possible. This is done correctly for async
offload, but somewhere along the line we lost the ability to do so when
we were able to complete a request without having to punt it. Ensure
that we do so correctly.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
This essentially reverts commit e944475e69. For high poll ops
workloads, like TAO, the dynamic allocation of the wait_queue
entry for IORING_OP_POLL_ADD adds considerable extra overhead.
Go back to embedding the wait_queue_entry, but keep the usage of
wait->private for the pointer stashing.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Don't just assign it from the main call path, that can miss the case
when we're called from issue deferral.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
We use the mutex to guard against registered file updates, for instance.
Ensure we're safe in accessing that state against concurrent updates.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Some commands will invariably end in a failure in the sense that the
completion result will be less than zero. One such example is timeouts
that don't have a completion count set, they will always complete with
-ETIME unless cancelled.
For linked commands, we sever links and fail the rest of the chain if
the result is less than zero. Since we have commands where we know that
will happen, add IOSQE_IO_HARDLINK as a stronger link that doesn't sever
regardless of the completion result. Note that the link will still sever
if we fail submitting the parent request, hard links are only resilient
in the presence of completion results for requests that did submit
correctly.
Cc: stable@vger.kernel.org # v5.4
Reviewed-by: Pavel Begunkov <asml.silence@gmail.com>
Reported-by: 李通洲 <carter.li@eoitek.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Links are created by chaining requests through req->list with an
exception that head uses req->link_list. (e.g. link_list->list->list)
Because of that, io_req_link_next() needs complex splicing to advance.
Link them all through list_list. Also, it seems to be simpler and more
consistent IMHO.
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
In case of an error io_submit_sqe() drops a request and continues
without it, even if the request was a part of a link. Not only it
doesn't cancel links, but also may execute wrong sequence of actions.
Stop consuming sqes, and let the user handle errors.
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
We recently changed this from a single list to an rbtree, but for some
real life workloads, the rbtree slows down the submission/insertion
case enough so that it's the top cycle consumer on the io_uring side.
In testing, using a hash table is a more well rounded compromise. It
is fast for insertion, and as long as it's sized appropriately, it
works well for the cancellation case as well. Running TAO with a lot
of network sockets, this removes io_poll_req_insert() from spending
2% of the CPU cycles.
Reported-by: Dan Melnic <dmm@fb.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
If we defer a timeout, we should ensure that we copy the timespec
when we have consumed the sqe. This is similar to commit f67676d160
for read/write requests. We already did this correctly for timeouts
deferred as links, but do it generally and use the infrastructure added
by commit 1a6b74fc87 instead of having the timeout deferral use its
own.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
There's really no reason why we forbid things like link/drain etc on
regular timeout commands. Enable the usual SQE flags on timeouts.
Reported-by: 李通洲 <carter.li@eoitek.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Right now we return it to userspace, which means the application has
to poll for the socket to be writeable. Let's just treat it like
-EAGAIN and have io_uring handle it internally, this makes it much
easier to use.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
If this flag is set, applications can be certain that any data for
async offload has been consumed when the kernel has consumed the
SQE.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Just like commit f67676d160 for read/write requests, this one ensures
that the sockaddr data has been copied for IORING_OP_CONNECT if we need
to punt the request to async context.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Just like commit f67676d160 for read/write requests, this one ensures
that the msghdr data is fully copied if we need to punt a recvmsg or
sendmsg system call to async context.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Currently we don't copy the iovecs when we punt to async context. This
can be problematic for applications that store the iovec on the stack,
as they often assume that it's safe to let the iovec go out of scope
as soon as IO submission has been called. This isn't always safe, as we
will re-copy the iovec once we're in async context.
Make this 100% safe by copying the iovec just once. With this change,
applications may safely store the iovec on the stack for all cases.
Reported-by: 李通洲 <carter.li@eoitek.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Right now we just copy the sqe for async offload, but we want to store
more context across an async punt. In preparation for doing so, put the
sqe copy inside a structure that we can expand. With this pointer added,
we can get rid of REQ_F_FREE_SQE, as that is now indicated by whether
req->io is NULL or not.
No functional changes in this patch.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
We should never return -ERESTARTSYS to userspace, transform it into
-EINTR.
Cc: stable@vger.kernel.org # v5.3+
Signed-off-by: Jens Axboe <axboe@kernel.dk>
-----BEGIN PGP SIGNATURE-----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=hnf4
-----END PGP SIGNATURE-----
Merge tag 'for-linus-20191129' of git://git.kernel.dk/linux-block
Pull block fixes from Jens Axboe:
"I wasn't going to send this one off so soon, but unfortunately one of
the fixes from the previous pull broke the build on some archs. So I'm
sending this sooner rather than later. This contains:
- Add highmem.h include for io_uring, because of the kmap() additions
from last round. For some reason the build bot didn't spot this
even though it sat for days.
- Three minor ';' removals
- Add support for the Beurer CD-on-a-chip device
- Make io_uring work on MMU-less archs"
* tag 'for-linus-20191129' of git://git.kernel.dk/linux-block:
io_uring: fix missing kmap() declaration on powerpc
ataflop: Remove unneeded semicolon
block: sunvdc: Remove unneeded semicolon
drbd: Remove unneeded semicolon
io_uring: add mapping support for NOMMU archs
sr_vendor: support Beurer GL50 evo CD-on-a-chip devices.
cdrom: respect device capabilities during opening action
Christophe reports that current master fails building on powerpc with
this error:
CC fs/io_uring.o
fs/io_uring.c: In function ‘loop_rw_iter’:
fs/io_uring.c:1628:21: error: implicit declaration of function ‘kmap’
[-Werror=implicit-function-declaration]
iovec.iov_base = kmap(iter->bvec->bv_page)
^
fs/io_uring.c:1628:19: warning: assignment makes pointer from integer
without a cast [-Wint-conversion]
iovec.iov_base = kmap(iter->bvec->bv_page)
^
fs/io_uring.c:1643:4: error: implicit declaration of function ‘kunmap’
[-Werror=implicit-function-declaration]
kunmap(iter->bvec->bv_page);
^
which is caused by a missing highmem.h include. Fix it by including
it.
Fixes: 311ae9e159 ("io_uring: fix dead-hung for non-iter fixed rw")
Reported-by: Christophe Leroy <christophe.leroy@c-s.fr>
Tested-by: Christophe Leroy <christophe.leroy@c-s.fr>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
That is a bit weird scenario but I find it interesting to run fio loads
using LKL linux, where MMU is disabled. Probably other real archs which
run uClinux can also benefit from this patch.
Signed-off-by: Roman Penyaev <rpenyaev@suse.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
In the quest to bring io_kiocb down to 3 cachelines, this one does
the trick. Make the wait_queue_entry for the poll command come out
of kmalloc instead of embedding it in struct io_poll_iocb, as the
latter is the largest member of io_kiocb. Once we trim this down a
bit, we're back at a healthy 192 bytes for struct io_kiocb.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Clean io_import_fixed() call site and make it return proper type.
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
There is no point left in keeping struct sqe_submit. Inline it
into struct io_kiocb, so any req->submit.field is now just req->field
- moves initialisation of ring_file into io_get_req()
- removes duplicated req->sequence.
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Timeouts' sequence offset (i.e. sqe->off) is stored in
req->submit.sequence under a false name. Keep it in timeout.data
instead. The unused space for sequence will be reclaimed in the
following patches.
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
This field contains a pointer to addrlen and checking to see if it's set
returns -EINVAL if the caller sets addr & addrlen pointers.
Fixes: 17f2fe35d0 ("io_uring: add support for IORING_OP_ACCEPT")
Signed-off-by: Hrvoje Zeba <zeba.hrvoje@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
We currently pass in 4 arguments outside of the bounded size. In
preparation for adding one more argument, let's bundle them up in
a struct to make it more readable.
No functional changes in this patch.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Read/write requests to devices without implemented read/write_iter
using fixed buffers can cause general protection fault, which totally
hangs a machine.
io_import_fixed() initialises iov_iter with bvec, but loop_rw_iter()
accesses it as iovec, dereferencing random address.
kmap() page by page in this case
Cc: stable@vger.kernel.org
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
This allows an application to call connect() in an async fashion. Like
other opcodes, we first try a non-blocking connect, then punt to async
context if we have to.
Note that we can still return -EINPROGRESS, and in that case the caller
should use IORING_OP_POLL_ADD to do an async wait for completion of the
connect request (just like for regular connect(2), except we can do it
async here too).
Signed-off-by: Jens Axboe <axboe@kernel.dk>
We return -EBUSY on submit when we have a CQ ring overflow backlog, but
that can be a bit problematic if the application is using pure userspace
poll of the CQ ring. For that case, if the ring briefly overflowed and
we have pending entries in the backlog, the submit flushes the backlog
successfully but still returns -EBUSY. If we're able to fully flush the
CQ ring backlog, let the submission proceed.
Reported-by: Dan Melnic <dmm@fb.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Pass only non-null @nxt to io_issue_sqe() and handle it at the caller's
side. And propagate it.
- kiocb_done() is only called from io_read() and io_write(), which are
only called from io_issue_sqe(), so it's @nxt != NULL
- io_put_req_find_next() is called either with explicitly non-null local
nxt, or from one of the functions in io_issue_sqe() switch (or their
callees).
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
"if (nxt)" is always true, as it was checked in the while's condition.
io_wq_current_is_worker() is unnecessary, as non-async callers don't
pass nxt, so io_queue_async_work() will be called for them anyway.
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Make io_req_find_next() and io_req_link_next() to accept only non-null
nxt, and handle it in callers.
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
There is only one one-liner user of io_free_req_find_next(). Inline it.
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
The number of SQEs to submit is specified by a user, so io_get_sqring()
in most of the cases succeeds. Hint compilers about that.
Checking ASM genereted by gcc 9.2.0 for x64, there is one branch
misprediction.
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
__io_submit_sqe() is issuing requests, so call it as
such. Moreover, it ends by calling io_iopoll_req_issued().
Rename it and make terminology clearer.
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
We don't have shadow requests anymore, so get rid of the shadow
argument. Add the user_data argument, as that's often useful to easily
match up requests, instead of having to look at request pointers.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
There's an issue with the shadow drain logic in that we drop the
completion lock after deciding to defer a request, then re-grab it later
and assume that the state is still the same. In the mean time, someone
else completing a request could have found and issued it. This can cause
a stall in the queue, by having a shadow request inserted that nobody is
going to drain.
Additionally, if we fail allocating the shadow request, we simply ignore
the drain.
Instead of using a shadow request, defer the next request/link instead.
This also has the following advantages:
- removes semi-duplicated code
- doesn't allocate memory for shadows
- works better if only the head marked for drain
- doesn't need complex synchronisation
On the flip side, it removes the shadow->seq ==
last_drain_in_in_link->seq optimization. That shouldn't be a common
case, and can always be added back, if needed.
Fixes: 4fe2c96315 ("io_uring: add support for link with drain")
Cc: Jackie Liu <liuyun01@kylinos.cn>
Reported-by: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
When we find new work to process within the work handler, we queue the
linked timeout before we have issued the new work. This can be
problematic for very short timeouts, as we have a window where the new
work isn't visible.
Allow the work handler to store a callback function for this in the work
item, and flag it with IO_WQ_WORK_CB if the caller has done so. If that
is set, then io-wq will call the callback when it has setup the new work
item.
Reported-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
We currently try and start the next link when we put the request, and
only if we were going to free it. This means that the optimization to
continue executing requests from the same context often fails, as we're
not putting the final reference.
Add REQ_F_LINK_NEXT to keep track of this, and allow io_uring to find the
next request more efficiently.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
We currently rely on the ring destroy on cleaning things up in case of
failure, but io_allocate_scq_urings() can leave things half initialized
if only parts of it fails.
Be nice and return with either everything setup in success, or return an
error with things nicely cleaned up.
Reported-by: syzbot+0d818c0d39399188f393@syzkaller.appspotmail.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Always mark requests with allocated sqe and deallocate it in
__io_free_req(). It's easier to follow and doesn't add edge cases.
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
We currently clear the linked timeout field if we cancel such a timeout,
but we should only attempt to cancel if it's the first one we see.
Others should simply be freed like other requests, as they haven't
been started yet.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
let have a dependant link: REQ -> LINK_TIMEOUT -> LINK_TIMEOUT
1. submission stage: submission references for REQ and LINK_TIMEOUT
are dropped. So, references respectively (1,1,2)
2. io_put(REQ) + FAIL_LINKS stage: calls io_fail_links(), which for all
linked timeouts will call cancel_timeout() and drop 1 reference.
So, references after: (0,0,1). That's a leak.
Make it treat only the first linked timeout as such, and pass others
through __io_double_put_req().
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Pass any IORING_OP_LINK_TIMEOUT request further, where it will
eventually fail in io_issue_sqe().
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
If io_req_defer() failed, it needs to cancel a dependant link.
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
We currently don't explicitly break links if a request is cancelled, but
we should. Add explicitly link breakage for all types of request
cancellations that we support.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Currently a poll request fills a completion entry of 0, even if it got
cancelled. This is odd, and it makes it harder to support with chains.
Ensure that it returns -ECANCELED in the completions events if it got
cancelled, and furthermore ensure that the linked timeout that triggered
it completes with -ETIME if we did indeed trigger the completions
through a timeout.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
With the conversion to io-wq, we no longer use that flag. Kill it.
Fixes: 561fb04a6a ("io_uring: replace workqueue usage with io-wq")
Signed-off-by: Jens Axboe <axboe@kernel.dk>
We have an issue with timeout links that are deeper in the submit chain,
because we only handle it upfront, not from later submissions. Move the
prep + issue of the timeout link to the async work prep handler, and do
it normally for non-async queue. If we validate and prepare the timeout
links upfront when we first see them, there's nothing stopping us from
supporting any sort of nesting.
Fixes: 2665abfd75 ("io_uring: add support for linked SQE timeouts")
Reported-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
There are a few reasons for this:
- As a prep to improving the linked timeout logic
- io_timeout is the biggest member in the io_kiocb opcode union
This also enables a few cleanups, like unifying the timer setup between
IORING_OP_TIMEOUT and IORING_OP_LINK_TIMEOUT, and not needing multiple
arguments to the link/prep helpers.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
If we don't use the normal completion path, we may skip killing links
that should be errored and freed. Add __io_double_put_req() for use
within the completion path itself, other calls should just use
io_double_put_req().
Signed-off-by: Jens Axboe <axboe@kernel.dk>
__io_queue_sqe(), io_queue_sqe(), io_queue_link_head() all return 0/err,
but the caller doesn't care since the errors are handled inline. Clean
these up and just make them void.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
If we have a linked request, this enables us to pass it back directly
without having to go through async context.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
-----BEGIN PGP SIGNATURE-----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=YZ9s
-----END PGP SIGNATURE-----
Merge tag 'for-5.5/io_uring-20191121' of git://git.kernel.dk/linux-block
Pull io_uring updates from Jens Axboe:
"A lot of stuff has been going on this cycle, with improving the
support for networked IO (and hence unbounded request completion
times) being one of the major themes. There's been a set of fixes done
this week, I'll send those out as well once we're certain we're fully
happy with them.
This contains:
- Unification of the "normal" submit path and the SQPOLL path (Pavel)
- Support for sparse (and bigger) file sets, and updating of those
file sets without needing to unregister/register again.
- Independently sized CQ ring, instead of just making it always 2x
the SQ ring size. This makes it more flexible for networked
applications.
- Support for overflowed CQ ring, never dropping events but providing
backpressure on submits.
- Add support for absolute timeouts, not just relative ones.
- Support for generic cancellations. This divorces io_uring from
workqueues as well, which additionally gets us one step closer to
generic async system call support.
- With cancellations, we can support grabbing the process file table
as well, just like we do mm context. This allows support for system
calls that create file descriptors, like accept4() support that's
built on top of that.
- Support for io_uring tracing (Dmitrii)
- Support for linked timeouts. These abort an operation if it isn't
completed by the time noted in the linke timeout.
- Speedup tracking of poll requests
- Various cleanups making the coder easier to follow (Jackie, Pavel,
Bob, YueHaibing, me)
- Update MAINTAINERS with new io_uring list"
* tag 'for-5.5/io_uring-20191121' of git://git.kernel.dk/linux-block: (64 commits)
io_uring: make POLL_ADD/POLL_REMOVE scale better
io-wq: remove now redundant struct io_wq_nulls_list
io_uring: Fix getting file for non-fd opcodes
io_uring: introduce req_need_defer()
io_uring: clean up io_uring_cancel_files()
io-wq: ensure free/busy list browsing see all items
io-wq: ensure we have a stable view of ->cur_work for cancellations
io_wq: add get/put_work handlers to io_wq_create()
io_uring: check for validity of ->rings in teardown
io_uring: fix potential deadlock in io_poll_wake()
io_uring: use correct "is IO worker" helper
io_uring: fix -ENOENT issue with linked timer with short timeout
io_uring: don't do flush cancel under inflight_lock
io_uring: flag SQPOLL busy condition to userspace
io_uring: make ASYNC_CANCEL work with poll and timeout
io_uring: provide fallback request for OOM situations
io_uring: convert accept4() -ERESTARTSYS into -EINTR
io_uring: fix error clear of ->file_table in io_sqe_files_register()
io_uring: separate the io_free_req and io_free_req_find_next interface
io_uring: keep io_put_req only responsible for release and put req
...
One of the obvious use cases for these commands is networking, where
it's not uncommon to have tons of sockets open and polled for. The
current implementation uses a list for insertion and lookup, which works
fine for file based use cases where the count is usually low, it breaks
down somewhat for higher number of files / sockets. A test case with
30k sockets being polled for and cancelled takes:
real 0m6.968s
user 0m0.002s
sys 0m6.936s
with the patch it takes:
real 0m0.233s
user 0m0.010s
sys 0m0.176s
If you go to 50k sockets, it gets even more abysmal with the current
code:
real 0m40.602s
user 0m0.010s
sys 0m40.555s
with the patch it takes:
real 0m0.398s
user 0m0.000s
sys 0m0.341s
Change is pretty straight forward, just replace the cancel_list with
a red/black tree instead.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
For timeout requests and bunch of others io_uring tries to grab a file
with specified fd, which is usually stdin/fd=0.
Update io_op_needs_file()
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
We don't use the return value anymore, drop it. Also drop the
unecessary double cancel_req value check.
Signed-off-by: Bob Liu <bob.liu@oracle.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
A test case was reported where two linked reads with registered buffers
failed the second link always. This is because we set the expected value
of a request in req->result, and if we don't get this result, then we
fail the dependent links. For some reason the registered buffer import
returned -ERROR/0, while the normal import returns -ERROR/length. This
broke linked commands with registered buffers.
Fix this by making io_import_fixed() correctly return the mapped length.
Cc: stable@vger.kernel.org # v5.3
Reported-by: 李通洲 <carter.li@eoitek.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
For timeout requests io_uring tries to grab a file with specified fd,
which is usually stdin/fd=0.
Update io_op_needs_file()
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
For cancellation, we need to ensure that the work item stays valid for
as long as ->cur_work is valid. Right now we can't safely dereference
the work item even under the wqe->lock, because while the ->cur_work
pointer will remain valid, the work could be completing and be freed
in parallel.
Only invoke ->get/put_work() on items we know that the caller queued
themselves. Add IO_WQ_WORK_INTERNAL for io-wq to use, which is needed
when we're queueing a flush item, for instance.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
We attempt to run the poll completion inline, but we're using trylock to
do so. This avoids a deadlock since we're grabbing the locks in reverse
order at this point, we already hold the poll wq lock and we're trying
to grab the completion lock, while the normal rules are the reverse of
that order.
IO completion for a timeout link will need to grab the completion lock,
but that's not safe from this context. Put the completion under the
completion_lock in io_poll_wake(), and mark the request as entering
the completion with the completion_lock already held.
Fixes: 2665abfd75 ("io_uring: add support for linked SQE timeouts")
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Since we switched to io-wq, the dependent link optimization for when to
pass back work inline has been broken. Fix this by providing a suitable
io-wq helper for io_uring to use to detect when to do this.
Fixes: 561fb04a6a ("io_uring: replace workqueue usage with io-wq")
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Currently we make sequence == 0 be the same as sequence == 1, but that's
not super useful if the intent is really to have a timeout that's just
a pure timeout.
If the user passes in sqe->off == 0, then don't apply any sequence logic
to the request, let it purely be driven by the timeout specified.
Reported-by: 李通洲 <carter.li@eoitek.com>
Reviewed-by: 李通洲 <carter.li@eoitek.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
If you prep a read (for example) that needs to get punted to async
context with a timer, if the timeout is sufficiently short, the timer
request will get completed with -ENOENT as it could not find the read.
The issue is that we prep and start the timer before we start the read.
Hence the timer can trigger before the read is even started, and the end
result is then that the timer completes with -ENOENT, while the read
starts instead of being cancelled by the timer.
Fix this by splitting the linked timer into two parts:
1) Prep and validate the linked timer
2) Start timer
The read is then started between steps 1 and 2, so we know that the
timer will always have a consistent view of the read request state.
Reported-by: Hrvoje Zeba <zeba.hrvoje@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
We can't safely cancel under the inflight lock. If the work hasn't been
started yet, then io_wq_cancel_work() simply marks the work as cancelled
and invokes the work handler. But if the work completion needs to grab
the inflight lock because it's grabbing user files, then we'll deadlock
trying to finish the work as we already hold that lock.
Instead grab a reference to the request, if it isn't already zero. If
it's zero, then we know it's going through completion anyway, and we
can safely ignore it. If it's not zero, then we can drop the lock and
attempt to cancel from there.
This also fixes a missing finish_wait() at the end of
io_uring_cancel_files().
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Now that we have backpressure, for SQPOLL, we have one more condition
that warrants flagging that the application needs to enter the kernel:
we failed to submit IO due to backpressure. Make sure we catch that
and flag it appropriately.
If we run into backpressure issues with the SQPOLL thread, flag it
as such to the application by setting IORING_SQ_NEED_WAKEUP. This will
cause the application to enter the kernel, and that will flush the
backlog and clear the condition.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
It's a little confusing that we have multiple types of command
cancellation opcodes now that we have a generic one. Make the generic
one work with POLL_ADD and TIMEOUT commands as well, that makes for an
easier to use API for the application. The fact that they currently
don't is a bit confusing.
Add a helper that takes care of it, so we can user it from both
IORING_OP_ASYNC_CANCEL and from the linked timeout cancellation.
Reported-by: Hrvoje Zeba <zeba.hrvoje@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
One thing that really sucks for userspace APIs is if the kernel passes
back -ENOMEM/-EAGAIN for resource shortages. The application really has
no idea of what to do in those cases. Should it try and reap
completions? Probably a good idea. Will it solve the issue? Who knows.
This patch adds a simple fallback mechanism if we fail to allocate
memory for a request. If we fail allocating memory from the slab for a
request, we punt to a pre-allocated request. There's just one of these
per io_ring_ctx, but the important part is if we ever return -EBUSY to
the application, the applications knows that it can wait for events and
make forward progress when events have completed. This is the important
part.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
If we cancel a pending accept operating with a signal, we get
-ERESTARTSYS returned. Turn that into -EINTR for userspace, we should
not be return -ERESTARTSYS.
Fixes: 17f2fe35d0 ("io_uring: add support for IORING_OP_ACCEPT")
Reported-by: Hrvoje Zeba <zeba.hrvoje@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Similar to the distinction between io_put_req and io_put_req_find_next,
io_free_req has been modified similarly, with no functional changes.
Signed-off-by: Jackie Liu <liuyun01@kylinos.cn>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
We already have io_put_req_find_next to find the next req of the link.
we should not use the io_put_req function to find them. They should be
functions of the same level.
Signed-off-by: Jackie Liu <liuyun01@kylinos.cn>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Many times, the core of the function is req, and req has already set
req->ctx at initialization time, so there is no need to pass in the
ctx from the caller.
Cleanup, no functional change.
Signed-off-by: Jackie Liu <liuyun01@kylinos.cn>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
With the recent flurry of additions and changes to io_uring, the
layout of io_ring_ctx has become a bit stale. We're right now at
704 bytes in size on my x86-64 build, or 11 cachelines. This
patch does two things:
- We have to completion structs embedded, that we only use for
quiesce of the ctx (or shutdown) and for sqthread init cases.
That 2x32 bytes right there, let's dynamically allocate them.
- Reorder the struct a bit with an eye on cachelines, use cases,
and holes.
With this patch, we're down to 512 bytes, or 8 cachelines.
Reviewed-by: Jackie Liu <liuyun01@kylinos.cn>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Now that io-wq supports separating the two request lifetime types, mark
the following IO as having unbounded runtimes:
- Any read/write to a non-regular file
- Any specific networked IO
- Any poll command
Signed-off-by: Jens Axboe <axboe@kernel.dk>
io_uring supports request types that basically have two different
lifetimes:
1) Bounded completion time. These are requests like disk reads or writes,
which we know will finish in a finite amount of time.
2) Unbounded completion time. These are generally networked IO, where we
have no idea how long they will take to complete. Another example is
POLL commands.
This patch provides support for io-wq to handle these differently, so we
don't starve bounded requests by tying up workers for too long. By default
all work is bounded, unless otherwise specified in the work item.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Currently we drop completion events, if the CQ ring is full. That's fine
for requests with bounded completion times, but it may make it harder or
impossible to use io_uring with networked IO where request completion
times are generally unbounded. Or with POLL, for example, which is also
unbounded.
After this patch, we never overflow the ring, we simply store requests
in a backlog for later flushing. This flushing is done automatically by
the kernel. To prevent the backlog from growing indefinitely, if the
backlog is non-empty, we apply back pressure on IO submissions. Any
attempt to submit new IO with a non-empty backlog will get an -EBUSY
return from the kernel. This is a signal to the application that it has
backlogged CQ events, and that it must reap those before being allowed
to submit more IO.
Note that if we do return -EBUSY, we will have filled whatever
backlogged events into the CQ ring first, if there's room. This means
the application can safely reap events WITHOUT entering the kernel and
waiting for them, they are already available in the CQ ring.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
This is in preparation for handling CQ ring overflow a bit smarter. We
should not have any functional changes in this patch. Most of the
changes are fairly straight forward, the only ones that stick out a bit
are the ones that change __io_free_req() to take the reference count
into account. If the request hasn't been submitted yet, we know it's
safe to simply ignore references and free it. But let's clean these up
too, as later patches will depend on the caller doing the right thing if
the completion logging grabs a reference to the request.
Reviewed-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
The rings can be derived from the ctx, and we need the ctx there for
a future change.
No functional changes in this patch.
Reviewed-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
While we have support for generic timeouts, we don't have a way to tie
a timeout to a specific SQE. The generic timeouts simply trigger wakeups
on the CQ ring.
This adds support for IORING_OP_LINK_TIMEOUT. This command is only valid
as a link to a previous command. The timeout specific can be either
relative or absolute, following the same rules as IORING_OP_TIMEOUT. If
the timeout triggers before the dependent command completes, it will
attempt to cancel that command. Likewise, if the dependent command
completes before the timeout triggers, it will cancel the timeout.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
We're going to need this helper in a future patch, so move it out
of io_async_cancel() and into its own separate function.
No functional changes in this patch.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Stack allocated struct sqe_submit is passed down to the submission path
along with a request (a.k.a. struct io_kiocb), and will be copied into
req->submit for async requests.
As space for it is already allocated, fill req->submit in the first
place instead of using on-stack one. As a result:
1. sqe->submit is the only place for sqe_submit and is always valid,
so we don't need to track which one to use.
2. don't need to copy in case of async
3. allows to simplify the code by not carrying it as an argument all
the way down
4. allows to reduce number of function arguments / potentially improve
spilling
The downside is that stack is most probably be cached, that's not true
for just allocated memory for a request. Another concern is cache
pollution. Though, a request would be touched and fetched along with
req->submit at some point anyway, so shouldn't be a problem.
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Let io_submit_sqes() to allocate io_kiocb before fetching an sqe.
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
After a call to io_submit_sqe(), it's already known whether it needs
to queue a link or not. Do it there, as it's simplier and doesn't keep
an extra variable across the loop.
Reviewed-by：Bob Liu <bob.liu@oracle.com>
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
io_submit_sqes() and io_ring_submit() are doing the same stuff with
a little difference. Deduplicate them.
Reviewed-by：Bob Liu <bob.liu@oracle.com>
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
We had no more use for this flag after the conversion to io-wq, kill it
off.
Fixes: 561fb04a6a ("io_uring: replace workqueue usage with io-wq")
Signed-off-by: Jens Axboe <axboe@kernel.dk>
If a request fails, we need to ensure we set REQ_F_FAIL_LINK on it if
REQ_F_LINK is set. Any failure in the chain should break the chain.
We were missing a few spots where this should be done. It might be nice
to generalize this somewhat at some point, as long as we factor in the
fact that failure looks different for each request type.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
As introduced by commit:
ba816ad61f ("io_uring: run dependent links inline if possible")
enable inline dependent link running for poll commands.
io_poll_complete_work() is the most important change, as it allows a
linked sequence of { POLL, READ } (for example) to proceed inline
instead of needing to get punted to another async context. The
submission side only potentially matters for sqthread, but may as well
include that bit.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
We currently don't have a completion event trace, add one of those. And
to better be able to match up submissions and completions, add user_data
to the submission trace as well.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
We didn't use -ERESTARTSYS to tell the application layer to restart the
system call, but instead return -EINTR. we can set -EINTR directly when
wakeup by the signal, which can help us save an assignment operation and
comparison operation.
Reviewed-by: Bob Liu <bob.liu@oracle.com>
Signed-off-by: Jackie Liu <liuyun01@kylinos.cn>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
This adds support for IORING_OP_ASYNC_CANCEL, which will attempt to
cancel requests that have been punted to async context and are now
in-flight. This works for regular read/write requests to files, as
long as they haven't been started yet. For socket based IO (or things
like accept4(2)), we can cancel work that is already running as well.
To cancel a request, the sqe must have ->addr set to the user_data of
the request it wishes to cancel. If the request is cancelled
successfully, the original request is completed with -ECANCELED
and the cancel request is completed with a result of 0. If the
request was already running, the original may or may not complete
in error. The cancel request will complete with -EALREADY for that
case. And finally, if the request to cancel wasn't found, the cancel
request is completed with -ENOENT.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
We use io_kiocb->result == -EAGAIN as a way to know if we need to
re-submit a polled request, as -EAGAIN reporting happens out-of-line
for IO submission failures. This field is cleared when we originally
allocate the request, but it isn't reset when we retry the submission
from async context. This can cause issues where we think something
needs a re-issue, but we're really just reading stale data.
Reset ->result whenever we re-prep a request for polled submission.
Cc: stable@vger.kernel.org
Fixes: 9e645e1105 ("io_uring: add support for sqe links")
Reported-by: Bijan Mottahedeh <bijan.mottahedeh@oracle.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
syzbot reported an issue where we crash at setup time if failslab is
used. The issue is that io_wq_create() returns an error pointer on
failure, not NULL. Hence io_uring thought the io-wq was setup just
fine, but in reality it's a garbage error pointer.
Use IS_ERR() instead of a NULL check, and assign ret appropriately.
Reported-by: syzbot+221cc24572a2fed23b6b@syzkaller.appspotmail.com
Fixes: 561fb04a6a ("io_uring: replace workqueue usage with io-wq")
Signed-off-by: Jens Axboe <axboe@kernel.dk>
If we get -1 from hrtimer_try_to_cancel(), we know that the timer
is running. Hence leave all completion to the timeout handler. If
we don't, we can corrupt the list and miss a completion.
Fixes: 11365043e5 ("io_uring: add support for canceling timeout requests")
Reported-by: Hrvoje Zeba <zeba.hrvoje@gmail.com>
Tested-by: Hrvoje Zeba <zeba.hrvoje@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
There's been a few requests for supporting more fixed files than 1024.
This isn't really tricky to do, we just need to split up the file table
into multiple tables and index appropriately. As we do so, reduce the
max single file table to 512. This enables us to do single page allocs
always for the tables, which is an improvement over the situation prior.
This patch adds support for up to 64K files, which should be enough for
everyone.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
We index the file tables with a user given value. After we check
it's within our limits, use array_index_nospec() to prevent any
spectre attacks here.
Suggested-by: Jann Horn <jannh@google.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
This allows an application to call accept4() in an async fashion. Like
other opcodes, we first try a non-blocking accept, then punt to async
context if we have to.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
This is in preparation for adding opcodes that need to add new files
in a process file table, system calls like open(2) or accept4(2).
If an opcode needs this, it must set IO_WQ_WORK_NEEDS_FILES in the work
item. If work that needs to get punted to async context have this
set, the async worker will assume the original task file table before
executing the work.
Note that opcodes that need access to the current files of an
application cannot be done through IORING_SETUP_SQPOLL.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Drop various work-arounds we have for workqueues:
- We no longer need the async_list for tracking sequential IO.
- We don't have to maintain our own mm tracking/setting.
- We don't need a separate workqueue for buffered writes. This didn't
even work that well to begin with, as it was suboptimal for multiple
buffered writers on multiple files.
- We can properly cancel pending interruptible work. This fixes
deadlocks with particularly socket IO, where we cannot cancel them
when the io_uring is closed. Hence the ring will wait forever for
these requests to complete, which may never happen. This is different
from disk IO where we know requests will complete in a finite amount
of time.
- Due to being able to cancel work interruptible work that is already
running, we can implement file table support for work. We need that
for supporting system calls that add to a process file table.
- It gets us one step closer to adding async support for any system
call.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Commit fb5ccc9878 ("io_uring: Fix broken links with offloading")
introduced a potential performance regression with unconditionally
taking mm even for READ/WRITE_FIXED operations.
Return the logic handling it back. mm-faulted requests will go through
the generic submission path, so honoring links and drains, but will
fail further on req->has_user check.
Fixes: fb5ccc9878 ("io_uring: Fix broken links with offloading")
Cc: stable@vger.kernel.org # v5.4
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
submit->index is used only for inbound check in submission path (i.e.
head < ctx->sq_entries). However, it always will be true, as
1. it's already validated by io_get_sqring()
2. ctx->sq_entries can't be changedd in between, because of held
ctx->uring_lock and ctx->refs.
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
To trace io_uring activity one can get an information from workqueue and
io trace events, but looks like some parts could be hard to identify via
this approach. Making what happens inside io_uring more transparent is
important to be able to reason about many aspects of it, hence introduce
the set of tracing events.
All such events could be roughly divided into two categories:
* those, that are helping to understand correctness (from both kernel
and an application point of view). E.g. a ring creation, file
registration, or waiting for available CQE. Proposed approach is to
get a pointer to an original structure of interest (ring context, or
request), and then find relevant events. io_uring_queue_async_work
also exposes a pointer to work_struct, to be able to track down
corresponding workqueue events.
* those, that provide performance related information. Mostly it's about
events that change the flow of requests, e.g. whether an async work
was queued, or delayed due to some dependencies. Another important
case is how io_uring optimizations (e.g. registered files) are
utilized.
Signed-off-by: Dmitrii Dolgov <9erthalion6@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
We might have cases where the need for a specific timeout is gone, add
support for canceling an existing timeout operation. This works like the
POLL_REMOVE command, where the application passes in the user_data of
the timeout it wishes to cancel in the sqe->addr field.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
This is a pretty trivial addition on top of the relative timeouts
we have now, but it's handy for ensuring tighter timing for those
that are building scheduling primitives on top of io_uring.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
There is no function change, just to clean up the code, use s->in_async
to make the code know where it is.
Signed-off-by: Jackie Liu <liuyun01@kylinos.cn>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
We currently size the CQ ring as twice the SQ ring, to allow some
flexibility in not overflowing the CQ ring. This is done because the
SQE life time is different than that of the IO request itself, the SQE
is consumed as soon as the kernel has seen the entry.
Certain application don't need a huge SQ ring size, since they just
submit IO in batches. But they may have a lot of requests pending, and
hence need a big CQ ring to hold them all. By allowing the application
to control the CQ ring size multiplier, we can cater to those
applications more efficiently.
If an application wants to define its own CQ ring size, it must set
IORING_SETUP_CQSIZE in the setup flags, and fill out
io_uring_params->cq_entries. The value must be a power of two.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Allows the application to remove/replace/add files to/from a file set.
Passes in a struct:
struct io_uring_files_update {
__u32 offset;
__s32 *fds;
};
that holds an array of fds, size of array passed in through the usual
nr_args part of the io_uring_register() system call. The logic is as
follows:
1) If ->fds[i] is -1, the existing file at i + ->offset is removed from
the set.
2) If ->fds[i] is a valid fd, the existing file at i + ->offset is
replaced with ->fds[i].
For case #2, is the existing file is currently empty (fd == -1), the
new fd is simply added to the array.
Reviewed-by: Jeff Moyer <jmoyer@redhat.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
This is in preparation for allowing updates to fixed file sets without
requiring a full unregister+register.
Reviewed-by: Jeff Moyer <jmoyer@redhat.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Currently any dependent link is executed from a new workqueue context,
which means that we'll be doing a context switch per link in the chain.
If we are running the completion of the current request from our async
workqueue and find that the next request is a link, then run it directly
from the workqueue context instead of forcing another switch.
This improves the performance of linked SQEs, and reduces the CPU
overhead.
Reviewed-by: Jackie Liu <liuyun01@kylinos.cn>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
syzkaller reported an issue where it looks like a malicious app can
trigger a use-after-free of reading the ctx ->sq_array and ->rings
value right after having installed the ring fd in the process file
table.
Defer ring fd installation until after we're done reading those
values.
Fixes: 75b28affdd ("io_uring: allocate the two rings together")
Reported-by: syzbot+6f03d895a6cd0d06187f@syzkaller.appspotmail.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
io_queue_link_head() owns shadow_req after taking it as an argument.
By not freeing it in case of an error, it can leak the request along
with taken ctx->refs.
Reviewed-by: Jackie Liu <liuyun01@kylinos.cn>
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
We currently assume that submissions from the sqthread are successful,
and if IO polling is enabled, we use that value for knowing how many
completions to look for. But if we overflowed the CQ ring or some
requests simply got errored and already completed, they won't be
available for polling.
For the case of IO polling and SQTHREAD usage, look at the pending
poll list. If it ever hits empty then we know that we don't have
anymore pollable requests inflight. For that case, simply reset
the inflight count to zero.
Reported-by: Pavel Begunkov <asml.silence@gmail.com>
Reviewed-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
We currently use the ring values directly, but that can lead to issues
if the application is malicious and changes these values on our behalf.
Created in-kernel cached versions of them, and just overwrite the user
side when we update them. This is similar to how we treat the sq/cq
ring tail/head updates.
Reported-by: Pavel Begunkov <asml.silence@gmail.com>
Reviewed-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
io_ring_submit() finalises with
1. io_commit_sqring(), which releases sqes to the userspace
2. Then calls to io_queue_link_head(), accessing released head's sqe
Reorder them.
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
io_sq_thread() processes sqes by 8 without considering links. As a
result, links will be randomely subdivided.
The easiest way to fix it is to call io_get_sqring() inside
io_submit_sqes() as do io_ring_submit().
Downsides:
1. This removes optimisation of not grabbing mm_struct for fixed files
2. It submitting all sqes in one go, without finer-grained sheduling
with cq processing.
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
There is a bug, where failed linked requests are returned not with
specified @user_data, but with garbage from a kernel stack.
The reason is that io_fail_links() uses req->user_data, which is
uninitialised when called from io_queue_sqe() on fail path.
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
The sequence number of the timeout req (req->sequence) indicate the
expected completion request. Because of each timeout req consume a
sequence number, so the sequence of each timeout req on the timeout
list shouldn't be the same. But now, we may get the same number (also
incorrect) if we insert a new entry before the last one, such as submit
such two timeout reqs on a new ring instance below.
req->sequence
req_1 (count = 2): 2
req_2 (count = 1): 2
Then, if we submit a nop req, req_2 will still timeout even the nop req
finished. This patch fix this problem by adjust the sequence number of
each reordered reqs when inserting a new entry.
Signed-off-by: zhangyi (F) <yi.zhang@huawei.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
The sequence number of reqs on the timeout_list before the timeout req
should be adjusted in io_timeout_fn(), because the current timeout req
will consumes a slot in the cq_ring and cq_tail pointer will be
increased, otherwise other timeout reqs may return in advance without
waiting for enough wait_nr.
Signed-off-by: zhangyi (F) <yi.zhang@huawei.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
There are cases where it isn't always safe to block for submission,
even if the caller asked to wait for events as well. Revert the
previous optimization of doing that.
This reverts two commits:
bf7ec93c64c576666863
Fixes: c576666863 ("io_uring: optimize submit_and_wait API")
Signed-off-by: Jens Axboe <axboe@kernel.dk>
-----BEGIN PGP SIGNATURE-----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=rvkw
-----END PGP SIGNATURE-----
Merge tag 'for-linus-2019-10-18' of git://git.kernel.dk/linux-block
Pull block fixes from Jens Axboe:
- NVMe pull request from Keith that address deadlocks, double resets,
memory leaks, and other regression.
- Fixup elv_support_iosched() for bio based devices (Damien)
- Fixup for the ahci PCS quirk (Dan)
- Socket O_NONBLOCK handling fix for io_uring (me)
- Timeout sequence io_uring fixes (yangerkun)
- MD warning fix for parameter default_layout (Song)
- blkcg activation fixes (Tejun)
- blk-rq-qos node deletion fix (Tejun)
* tag 'for-linus-2019-10-18' of git://git.kernel.dk/linux-block:
nvme-pci: Set the prp2 correctly when using more than 4k page
io_uring: fix logic error in io_timeout
io_uring: fix up O_NONBLOCK handling for sockets
md/raid0: fix warning message for parameter default_layout
libata/ahci: Fix PCS quirk application
blk-rq-qos: fix first node deletion of rq_qos_del()
blkcg: Fix multiple bugs in blkcg_activate_policy()
io_uring: consider the overflow of sequence for timeout req
nvme-tcp: fix possible leakage during error flow
nvmet-loop: fix possible leakage during error flow
block: Fix elv_support_iosched()
nvme-tcp: Initialize sk->sk_ll_usec only with NET_RX_BUSY_POLL
nvme: Wait for reset state when required
nvme: Prevent resets during paused controller state
nvme: Restart request timers in resetting state
nvme: Remove ADMIN_ONLY state
nvme-pci: Free tagset if no IO queues
nvme: retain split access workaround for capability reads
nvme: fix possible deadlock when nvme_update_formats fails
If ctx->cached_sq_head < nxt_sq_head, we should add UINT_MAX to tmp, not
tmp_nxt.
Fixes: 5da0fb1ab3 ("io_uring: consider the overflow of sequence for timeout req")
Signed-off-by: yangerkun <yangerkun@huawei.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
We've got two issues with the non-regular file handling for non-blocking
IO:
1) We don't want to re-do a short read in full for a non-regular file,
as we can't just read the data again.
2) For non-regular files that don't support non-blocking IO attempts,
we need to punt to async context even if the file is opened as
non-blocking. Otherwise the caller always gets -EAGAIN.
Add two new request flags to handle these cases. One is just a cache
of the inode S_ISREG() status, the other tells io_uring that we always
need to punt this request to async context, even if REQ_F_NOWAIT is set.
Cc: stable@vger.kernel.org
Reported-by: Hrvoje Zeba <zeba.hrvoje@gmail.com>
Tested-by: Hrvoje Zeba <zeba.hrvoje@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Now we recalculate the sequence of timeout with 'req->sequence =
ctx->cached_sq_head + count - 1', judge the right place to insert
for timeout_list by compare the number of request we still expected for
completion. But we have not consider about the situation of overflow:
1. ctx->cached_sq_head + count - 1 may overflow. And a bigger count for
the new timeout req can have a small req->sequence.
2. cached_sq_head of now may overflow compare with before req. And it
will lead the timeout req with small req->sequence.
This overflow will lead to the misorder of timeout_list, which can lead
to the wrong order of the completion of timeout_list. Fix it by reuse
req->submit.sequence to store the count, and change the logic of
inserting sort in io_timeout.
Signed-off-by: yangerkun <yangerkun@huawei.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
-----BEGIN PGP SIGNATURE-----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=PUeL
-----END PGP SIGNATURE-----
Merge tag 'for-linus-20191012' of git://git.kernel.dk/linux-block
Pull io_uring fix from Jens Axboe:
"Single small fix for a regression in the sequence logic for linked
commands"
* tag 'for-linus-20191012' of git://git.kernel.dk/linux-block:
io_uring: fix sequence logic for timeout requests
We have two ways a request can be deferred:
1) It's a regular request that depends on another one
2) It's a timeout that tracks completions
We have a shared helper to determine whether to defer, and that
attempts to make the right decision based on the request. But we
only have some of this information in the caller. Un-share the
two timeout/defer helpers so the caller can use the right one.
Fixes: 5262f56798 ("io_uring: IORING_OP_TIMEOUT support")
Reported-by: yangerkun <yangerkun@huawei.com>
Reviewed-by: Jackie Liu <liuyun01@kylinos.cn>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
We should not remove the workqueue, we just need to ensure that the
workqueues are synced. The workqueues are torn down on ctx removal.
Cc: stable@vger.kernel.org
Fixes: 6b06314c47 ("io_uring: add file set registration")
Reported-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Any changes interesting to tasks waiting in io_cqring_wait() are
commited with io_cqring_ev_posted(). However, io_ring_drop_ctx_refs()
also tries to do that but with no reason, that means spurious wakeups
every io_free_req() and io_uring_enter().
Just use percpu_ref_put() instead.
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
-----BEGIN PGP SIGNATURE-----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=vKVl
-----END PGP SIGNATURE-----
Merge tag 'for-linus-2019-10-03' of git://git.kernel.dk/linux-block
Pull block fixes from Jens Axboe:
- Mandate timespec64 for the io_uring timeout ABI (Arnd)
- Set of NVMe changes via Sagi:
- controller removal race fix from Balbir
- quirk additions from Gabriel and Jian-Hong
- nvme-pci power state save fix from Mario
- Add 64bit user commands (for 64bit registers) from Marta
- nvme-rdma/nvme-tcp fixes from Max, Mark and Me
- Minor cleanups and nits from James, Dan and John
- Two s390 dasd fixes (Jan, Stefan)
- Have loop change block size in DIO mode (Martijn)
- paride pg header ifdef guard (Masahiro)
- Two blk-mq queue scheduler tweaks, fixing an ordering issue on zoned
devices and suboptimal performance on others (Ming)
* tag 'for-linus-2019-10-03' of git://git.kernel.dk/linux-block: (22 commits)
block: sed-opal: fix sparse warning: convert __be64 data
block: sed-opal: fix sparse warning: obsolete array init.
block: pg: add header include guard
Revert "s390/dasd: Add discard support for ESE volumes"
s390/dasd: Fix error handling during online processing
io_uring: use __kernel_timespec in timeout ABI
loop: change queue block size to match when using DIO
blk-mq: apply normal plugging for HDD
blk-mq: honor IO scheduler for multiqueue devices
nvme-rdma: fix possible use-after-free in connect timeout
nvme: Move ctrl sqsize to generic space
nvme: Add ctrl attributes for queue_count and sqsize
nvme: allow 64-bit results in passthru commands
nvme: Add quirk for Kingston NVME SSD running FW E8FK11.T
nvmet-tcp: remove superflous check on request sgl
Added QUIRKs for ADATA XPG SX8200 Pro 512GB
nvme-rdma: Fix max_hw_sectors calculation
nvme: fix an error code in nvme_init_subsystem()
nvme-pci: Save PCI state before putting drive into deepest state
nvme-tcp: fix wrong stop condition in io_work
...
All system calls use struct __kernel_timespec instead of the old struct
timespec, but this one was just added with the old-style ABI. Change it
now to enforce the use of __kernel_timespec, avoiding ABI confusion and
the need for compat handlers on 32-bit architectures.
Any user space caller will have to use __kernel_timespec now, but this
is unambiguous and works for any C library regardless of the time_t
definition. A nicer way to specify the timeout would have been a less
ambiguous 64-bit nanosecond value, but I suppose it's too late now to
change that as this would impact both 32-bit and 64-bit users.
Fixes: 5262f56798 ("io_uring: IORING_OP_TIMEOUT support")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
-----BEGIN PGP SIGNATURE-----
iQJEBAABCAAuFiEEwPw5LcreJtl1+l5K99NY+ylx4KYFAl2OIu4QHGF4Ym9lQGtl
cm5lbC5kawAKCRD301j7KXHgpsedD/4h54330vuq66DsGqzFLonLwFl5YHC5NeJX
aV38j7pAUPHvr9CSnr3d2VwTk/ThBtPI50I9/d9SXh8n1oAAA5C/+nPf1XknME47
giKHr3eb0FNLOySt/Ry284gla8mO0GZM83zUbDMnF0N+tfwAtFvbvgCpsPFK9vdL
xNzMLsq6++pj7p9c6IXd+zv0nmJzDikZQz6PtU1KYlbfnU7hh/cP3CrIHIPtGbyk
c/7tfbKTB9UnbW5guGakZt3BzNViJpK28SKn+S6AlLiEOYpC+55dBbaZQIy5qxHv
CZsx0GJIw0Ya0Lw3UEFp/74krLHq2610jmx/va8P7MZCjZAR675G3mjxKUnC/+SY
mEdLo6vghMNAIqMBWNu59CFQOPnqa8sqRii0q6cRWXSqKiFr1FLN8mstb3Ghh9K0
kGVA/gw6ESWB/e/X6I+pD6pTm6O6BPWEqBzGAWSvavQQIP9YpIzf5j+k3JsRu03/
IIzR6gW9k9u4k0rFlOJKbp1+AO5sK3VtJFR8JGELiRwwgjD91w50gjPYak5OGM37
Mi7OHCxqtwFGTkSvT6RM6om6onBsizrveszkrPUO01bWYIHHbtu6ofLyQlfnEtpv
qbGZtLW6KYj9VsIKZNDfg99Ff79IAOiAZDbXWAu/JKyg/gu1Y9uOiVkNFPJGPNHV
8ourcldMGg==
=DEYH
-----END PGP SIGNATURE-----
Merge tag 'for-5.4/io_uring-2019-09-27' of git://git.kernel.dk/linux-block
Pull more io_uring updates from Jens Axboe:
"Just two things in here:
- Improvement to the io_uring CQ ring wakeup for batched IO (me)
- Fix wrong comparison in poll handling (yangerkun)
I realize the first one is a little late in the game, but it felt
pointless to hold it off until the next release. Went through various
testing and reviews with Pavel and peterz"
* tag 'for-5.4/io_uring-2019-09-27' of git://git.kernel.dk/linux-block:
io_uring: make CQ ring wakeups be more efficient
io_uring: compare cached_cq_tail with cq.head in_io_uring_poll
For batched IO, it's not uncommon for waiters to ask for more than 1
IO to complete before being woken up. This is a problem with
wait_event() since tasks will get woken for every IO that completes,
re-check condition, then go back to sleep. For batch counts on the
order of what you do for high IOPS, that can result in 10s of extra
wakeups for the waiting task.
Add a private wake function that checks for the wake up count criteria
being met before calling autoremove_wake_function(). Pavel reports that
one test case he has runs 40% faster with proper batching of wakeups.
Reported-by: Pavel Begunkov <asml.silence@gmail.com>
Tested-by: Pavel Begunkov <asml.silence@gmail.com>
Reviewed-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
-----BEGIN PGP SIGNATURE-----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=/w5G
-----END PGP SIGNATURE-----
Merge tag 'for-5.4/io_uring-2019-09-24' of git://git.kernel.dk/linux-block
Pull more io_uring updates from Jens Axboe:
"A collection of later fixes and additions, that weren't quite ready
for pushing out with the initial pull request.
This contains:
- Fix potential use-after-free of shadow requests (Jackie)
- Fix potential OOM crash in request allocation (Jackie)
- kmalloc+memcpy -> kmemdup cleanup (Jackie)
- Fix poll crash regression (me)
- Fix SQ thread not being nice and giving up CPU for !PREEMPT (me)
- Add support for timeouts, making it easier to do epoll_wait()
conversions, for instance (me)
- Ensure io_uring works without f_ops->read_iter() and
f_ops->write_iter() (me)"
* tag 'for-5.4/io_uring-2019-09-24' of git://git.kernel.dk/linux-block:
io_uring: correctly handle non ->{read,write}_iter() file_operations
io_uring: IORING_OP_TIMEOUT support
io_uring: use cond_resched() in sqthread
io_uring: fix potential crash issue due to io_get_req failure
io_uring: ensure poll commands clear ->sqe
io_uring: fix use-after-free of shadow_req
io_uring: use kmemdup instead of kmalloc and memcpy
Patch series "Make working with compound pages easier", v2.
These three patches add three helpers and convert the appropriate
places to use them.
This patch (of 3):
It's unnecessarily hard to find out the size of a potentially huge page.
Replace 'PAGE_SIZE << compound_order(page)' with page_size(page).
Link: http://lkml.kernel.org/r/20190721104612.19120-2-willy@infradead.org
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Acked-by: Michal Hocko <mhocko@suse.com>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: Ira Weiny <ira.weiny@intel.com>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
After 75b28af("io_uring: allocate the two rings together"), we compare
sq.head with cached_cq_tail to determine does there any cq invalid.
Actually, we should use cq.head.
Fixes: 75b28affdd ("io_uring: allocate the two rings together")
Signed-off-by: yangerkun <yangerkun@huawei.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Currently we just -EINVAL a read or write to an fd that isn't backed
by ->read_iter() or ->write_iter(). But we can handle them just fine,
as long as we punt fo async context first.
Implement a simple loop function for doing ->read() or ->write()
instead, and ensure we call it appropriately.
Reported-by: 李通洲 <carter.li@eoitek.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
There's been a few requests for functionality similar to io_getevents()
and epoll_wait(), where the user can specify a timeout for waiting on
events. I deliberately did not add support for this through the system
call initially to avoid overloading the args, but I can see that the use
cases for this are valid.
This adds support for IORING_OP_TIMEOUT. If a user wants to get woken
when waiting for events, simply submit one of these timeout commands
with your wait call (or before). This ensures that the application
sleeping on the CQ ring waiting for events will get woken. The timeout
command is passed in as a pointer to a struct timespec. Timeouts are
relative. The timeout command also includes a way to auto-cancel after
N events has passed.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
If preempt isn't enabled in the kernel, we can run into hang issues with
sqthread submissions. Use cond_resched() to play nice instead of
cpu_relax(), if we end up starting the loop and not having any events
pending for submissions.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Sometimes io_get_req will return a NUL, then we need to do the
correct error handling, otherwise it will cause the kernel null
pointer exception.
Fixes: 4fe2c96315 ("io_uring: add support for link with drain")
Signed-off-by: Jackie Liu <liuyun01@kylinos.cn>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
If we end up getting woken in poll (due to a signal), then we may need
to punt the poll request to an async worker. When we do that, we look up
the list to queue at, deferefencing req->submit.sqe, however that is
only set for requests we initially decided to queue async.
This fixes a crash with poll command usage and wakeups that need to punt
to async context.
Fixes: 54a91f3bb9 ("io_uring: limit parallelism of buffered writes")
Signed-off-by: Jens Axboe <axboe@kernel.dk>
There is a potential dangling pointer problem. we never clean
shadow_req, if there are multiple link lists in this series of
sqes, then the shadow_req will not reallocate, and continue to
use the last one. but in the previous, his memory has been
released, thus forming a dangling pointer. let's clean up him
and make sure that every new link list can reapply for a new
shadow_req.
Fixes: 4fe2c96315 ("io_uring: add support for link with drain")
Signed-off-by: Jackie Liu <liuyun01@kylinos.cn>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Some workloads can require far more than 4K oustanding entries. For
example memcached can have ~300K sockets over ~40 cores. Bumping the max
to 32K seems to work pretty well.
Reported-by: Dan Melnic <dmm@fb.com>
Signed-off-by: Daniel Xu <dxu@dxuuu.xyz>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
The way the logic is setup in io_uring_enter() means that you can't wake
up the SQ poller thread while at the same time waiting (or polling) for
completions afterwards. There's no reason for that to be the case.
Reported-by: Lewis Baker <lbaker@fb.com>
Reviewed-by: Jeff Moyer <jmoyer@redhat.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
We currently merge async work items if we see a strict sequential hit.
This helps avoid unnecessary workqueue switches when we don't need
them. We can extend this merging to cover cases where it's not a strict
sequential hit, but the IO still fits within the same page. If an
application is doing multiple requests within the same page, we don't
want separate workers waiting on the same page to complete IO. It's much
faster to let the first worker bring in the page, then operate on that
page from the same worker to complete the next request(s).
Reviewed-by: Jeff Moyer <jmoyer@redhat.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
All the popular filesystems need to grab the inode lock for buffered
writes. With io_uring punting buffered writes to async context, we
observe a lot of contention with all workers hamming this mutex.
For buffered writes, we generally don't need a lot of parallelism on
the submission side, as the flushing will take care of that for us.
Hence we don't need a deep queue on the write side, as long as we
can safely punt from the original submission context.
Add a workqueue with a limit of 2 that we can use for buffered writes.
This greatly improves the performance and efficiency of higher queue
depth buffered async writes with io_uring.
Reported-by: Andres Freund <andres@anarazel.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Add a helper for queueing a request for async execution, in preparation
for optimizing it.
No functional change in this patch.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
For some applications that end up using a submit-and-wait type of
approach for certain batches of IO, we can make that a bit more
efficient by allowing the application to block for the last IO
submission. This prevents an async when we don't need it, as the
application will be blocking for the completion event(s) anyway.
Typical use cases are using the liburing
io_uring_submit_and_wait() API, or just using io_uring_enter()
doing both submissions and completions. As a specific example,
RocksDB doing MultiGet() is sped up quite a bit with this
change.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
To support the link with drain, we need to do two parts.
There is an sqes:
0 1 2 3 4 5 6
+-----+-----+-----+-----+-----+-----+-----+
| N | L | L | L+D | N | N | N |
+-----+-----+-----+-----+-----+-----+-----+
First, we need to ensure that the io before the link is completed,
there is a easy way is set drain flag to the link list's head, so
all subsequent io will be inserted into the defer_list.
+-----+
(0) | N |
+-----+
| (2) (3) (4)
+-----+ +-----+ +-----+ +-----+
(1) | L+D | --> | L | --> | L+D | --> | N |
+-----+ +-----+ +-----+ +-----+
|
+-----+
(5) | N |
+-----+
|
+-----+
(6) | N |
+-----+
Second, ensure that the following IO will not be completed first,
an easy way is to create a mirror of drain io and insert it into
defer_list, in this way, as long as drain io is not processed, the
following io in the defer_list will not be actively process.
+-----+
(0) | N |
+-----+
| (2) (3) (4)
+-----+ +-----+ +-----+ +-----+
(1) | L+D | --> | L | --> | L+D | --> | N |
+-----+ +-----+ +-----+ +-----+
|
+-----+
('3) | D | <== This is a shadow of (3)
+-----+
|
+-----+
(5) | N |
+-----+
|
+-----+
(6) | N |
+-----+
Signed-off-by: Jackie Liu <liuyun01@kylinos.cn>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Sqo_thread will get sqring in batches, which will cause
ctx->cached_sq_head to be added in batches. if one of these
sqes is set with the DRAIN flag, then he will never get a
chance to process, and finally sqo_thread will not exit.
Fixes: de0617e467 ("io_uring: add support for marking commands as draining")
Signed-off-by: Jackie Liu <liuyun01@kylinos.cn>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
After commit 75b28affdd we can get by with just a single mmap to
map both the sq and cq ring. However, userspace doesn't know that.
Add a features variable to io_uring_params, and notify userspace
that the kernel has this ability. This can then be used in liburing
(or in applications directly) to avoid the second mmap.
Signed-off-by: Jens Axboe <axboe@kernel.dk>
