linux-block.git
4 years agoio_uring: remove file batch-get optimisation
Pavel Begunkov [Tue, 10 Aug 2021 13:52:47 +0000 (14:52 +0100)]
io_uring: remove file batch-get optimisation

For requests with non-fixed files, instead of grabbing just one
reference, we get by the number of left requests, so the following
requests using the same file can take it without atomics.

However, it's not all win. If there is one request in the middle
not using files or having a fixed file, we'll need to put back the left
references. Even worse if an application submits requests dealing with
different files, it will do a put for each new request, so doubling the
number of atomics needed. Also, even if not used, it's still takes some
cycles in the submission path.

If a file used many times, it rather makes sense to pre-register it, if
not, we may fall in the described pitfall. So, this optimisation is a
matter of use case. Go with the simpliest code-wise way, remove it.

Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
4 years agoio_uring: clean up tctx_task_work()
Pavel Begunkov [Tue, 10 Aug 2021 16:53:55 +0000 (17:53 +0100)]
io_uring: clean up tctx_task_work()

After recent fixes, tctx_task_work() always does proper spinlocking
before looking into ->task_list, so now we don't need atomics for
->task_state, replace it with non-atomic task_running using the critical
section.

Tide it up, combine two separate block with spinlocking, and always try
to splice in there, so we do less locking when new requests are arriving
during the function execution.

Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
[axboe: fix missing ->task_running reset on task_work_add() failure]
Signed-off-by: Jens Axboe <axboe@kernel.dk>
4 years agoio_uring: inline io_poll_remove_waitqs
Pavel Begunkov [Mon, 9 Aug 2021 19:18:13 +0000 (20:18 +0100)]
io_uring: inline io_poll_remove_waitqs

Inline io_poll_remove_waitqs() into its only user and clean it up.

Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Link: https://lore.kernel.org/r/2f1a91a19ffcd591531dc4c61e2f11c64a2d6a6d.1628536684.git.asml.silence@gmail.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
4 years agoio_uring: remove extra argument for overflow flush
Pavel Begunkov [Mon, 9 Aug 2021 19:18:12 +0000 (20:18 +0100)]
io_uring: remove extra argument for overflow flush

Unlike __io_cqring_overflow_flush(), nobody does forced flushing with
io_cqring_overflow_flush(), so removed the argument from it.

Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Link: https://lore.kernel.org/r/7594f869ca41b7cfb5a35a3c7c2d402242834e9e.1628536684.git.asml.silence@gmail.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
4 years agoio_uring: inline struct io_comp_state
Pavel Begunkov [Mon, 9 Aug 2021 19:18:11 +0000 (20:18 +0100)]
io_uring: inline struct io_comp_state

Inline struct io_comp_state into struct io_submit_state. They are
already coupled tightly, together with mixed responsibilities it
only brings confusion having them separately.

Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Link: https://lore.kernel.org/r/e55bba77426b399e3a2e54e3c6c267c6a0fc4b57.1628536684.git.asml.silence@gmail.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
4 years agoio_uring: use inflight_entry instead of compl.list
Pavel Begunkov [Mon, 9 Aug 2021 19:18:10 +0000 (20:18 +0100)]
io_uring: use inflight_entry instead of compl.list

req->compl.list is used to cache freed requests, and so can't overlap in
time with req->inflight_entry. So, use inflight_entry to link requests
and remove compl.list.

Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Link: https://lore.kernel.org/r/e430e79d22d70a190d718831bda7bfed1daf8976.1628536684.git.asml.silence@gmail.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
4 years agoio_uring: remove redundant args from cache_free
Pavel Begunkov [Mon, 9 Aug 2021 19:18:09 +0000 (20:18 +0100)]
io_uring: remove redundant args from cache_free

We don't use @tsk argument of io_req_cache_free(), remove it.

Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Link: https://lore.kernel.org/r/6a28b4a58ee0aaf0db98e2179b9c9f06f9b0cca1.1628536684.git.asml.silence@gmail.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
4 years agoio_uring: cache __io_free_req()'d requests
Pavel Begunkov [Mon, 9 Aug 2021 19:18:08 +0000 (20:18 +0100)]
io_uring: cache __io_free_req()'d requests

Don't kfree requests in __io_free_req() but put them back into the
internal request cache. That makes allocations more sustainable and will
be used for refcounting optimisations.

Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Link: https://lore.kernel.org/r/9f4950fbe7771c8d41799366d0a3a08ac3040236.1628536684.git.asml.silence@gmail.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
4 years agoio_uring: move io_fallback_req_func()
Pavel Begunkov [Mon, 9 Aug 2021 19:18:07 +0000 (20:18 +0100)]
io_uring: move io_fallback_req_func()

Move io_fallback_req_func() to kill yet another forward declaration.

Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Link: https://lore.kernel.org/r/d0a8f9d9a0057ed761d6237167d51c9378798d2d.1628536684.git.asml.silence@gmail.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
4 years agoio_uring: optimise putting task struct
Pavel Begunkov [Mon, 9 Aug 2021 12:04:20 +0000 (13:04 +0100)]
io_uring: optimise putting task struct

We cache all the reference to task + tctx, so if io_put_task() is
called by the corresponding task itself, we can save on atomics and
return the refs right back into the cache.

It's beneficial for all inline completions, and also iopolling, when
polling and submissions are done by the same task, including
SQPOLL|IOPOLL.

Note: io_uring_cancel_generic() can return refs to the cache as well,
so those should be flushed in the loop for tctx_inflight() to work
right.

Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Link: https://lore.kernel.org/r/6fe9646b3cb70e46aca1f58426776e368c8926b3.1628471125.git.asml.silence@gmail.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
4 years agoio_uring: drop exec checks from io_req_task_submit
Pavel Begunkov [Mon, 9 Aug 2021 12:04:19 +0000 (13:04 +0100)]
io_uring: drop exec checks from io_req_task_submit

In case of on-exec io_uring cancellations, tasks already wait for all
submitted requests to get completed/cancelled, so we don't need to check
for ->in_execve separately.

Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Link: https://lore.kernel.org/r/be8707049f10df9d20ca03dc4ca3316239b5e8e0.1628471125.git.asml.silence@gmail.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
4 years agoio_uring: kill unused IO_IOPOLL_BATCH
Pavel Begunkov [Mon, 9 Aug 2021 12:04:18 +0000 (13:04 +0100)]
io_uring: kill unused IO_IOPOLL_BATCH

IO_IOPOLL_BATCH is not used, delete it.

Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Link: https://lore.kernel.org/r/b2bdf19dbee2c9fc8865bbab9412135a14e24a64.1628471125.git.asml.silence@gmail.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
4 years agoio_uring: improve ctx hang handling
Pavel Begunkov [Mon, 9 Aug 2021 12:04:17 +0000 (13:04 +0100)]
io_uring: improve ctx hang handling

If io_ring_exit_work() can't get it done in 5 minutes, something is
going very wrong, don't keep spinning at HZ / 20 rate, it doesn't help
and it may take much of CPU time if there is a lot of workers stuck as
such.

Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Link: https://lore.kernel.org/r/9e2d1ca81d569f6bc628af1a42ff6663bff7ce9c.1628471125.git.asml.silence@gmail.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
4 years agoio_uring: deduplicate open iopoll check
Pavel Begunkov [Mon, 9 Aug 2021 12:04:16 +0000 (13:04 +0100)]
io_uring: deduplicate open iopoll check

Move IORING_SETUP_IOPOLL check into __io_openat_prep(), so both openat
and openat2 reuse it.

Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Link: https://lore.kernel.org/r/9a73ce83e4ee60d011180ef177eecef8e87ff2a2.1628471125.git.asml.silence@gmail.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
4 years agoio_uring: inline io_free_req_deferred
Pavel Begunkov [Mon, 9 Aug 2021 12:04:15 +0000 (13:04 +0100)]
io_uring: inline io_free_req_deferred

Inline io_free_req_deferred(), there is no reason to keep it separated.

Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Link: https://lore.kernel.org/r/ce04b7180d4eac0d69dd00677b227eefe80c2cc5.1628471125.git.asml.silence@gmail.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
4 years agoio_uring: move io_rsrc_node_alloc() definition
Pavel Begunkov [Mon, 9 Aug 2021 15:09:47 +0000 (09:09 -0600)]
io_uring: move io_rsrc_node_alloc() definition

Move the function together with io_rsrc_node_ref_zero() in the source
file as it is to get rid of forward declarations.

Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Link: https://lore.kernel.org/r/4d81f6f833e7d017860b24463a9a68b14a8a5ed2.1628471125.git.asml.silence@gmail.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
4 years agoio_uring: move io_put_task() definition
Pavel Begunkov [Mon, 9 Aug 2021 12:04:13 +0000 (13:04 +0100)]
io_uring: move io_put_task() definition

Move the function in the source file as it is to get rid of forward
declarations.

Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Link: https://lore.kernel.org/r/33d917d69e4206557c75a5b98fe22bcdf77ce47d.1628471125.git.asml.silence@gmail.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
4 years agoio_uring: extract a helper for ctx quiesce
Pavel Begunkov [Mon, 9 Aug 2021 12:04:12 +0000 (13:04 +0100)]
io_uring: extract a helper for ctx quiesce

Refactor __io_uring_register() by extracting a helper responsible for
ctx queisce. Looks better and will make it easier to add more
optimisations.

Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Link: https://lore.kernel.org/r/0339e0027504176be09237eefa7945bf9a6f153d.1628471125.git.asml.silence@gmail.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
4 years agoio_uring: optimise io_cqring_wait() hot path
Pavel Begunkov [Mon, 9 Aug 2021 15:07:32 +0000 (09:07 -0600)]
io_uring: optimise io_cqring_wait() hot path

Turns out we always init struct io_wait_queue in io_cqring_wait(), even
if it's not used after, i.e. there are already enough of CQEs. And often
it's exactly what happens, for instance, requests may have been
completed inline, or in case of io_uring_enter(submit=N, wait=1).

It shows up in my profiler, so optimise it by delaying the struct init.

Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Link: https://lore.kernel.org/r/6f1b81c60b947d165583dc333947869c3d85d037.1628471125.git.asml.silence@gmail.com
[axboe: fixed up for new cqring wait]
Signed-off-by: Jens Axboe <axboe@kernel.dk>
4 years agoio_uring: add more locking annotations for submit
Pavel Begunkov [Mon, 9 Aug 2021 12:04:10 +0000 (13:04 +0100)]
io_uring: add more locking annotations for submit

Add more annotations for submission path functions holding ->uring_lock.

Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Link: https://lore.kernel.org/r/128ec4185e26fbd661dd3a424aa66108ee8ff951.1628471125.git.asml.silence@gmail.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
4 years agoio_uring: don't halt iopoll too early
Pavel Begunkov [Mon, 9 Aug 2021 12:04:09 +0000 (13:04 +0100)]
io_uring: don't halt iopoll too early

IOPOLL users should care more about getting completions for requests
they submitted, but not in "device did/completed something". Currently,
io_do_iopoll() may return a positive number, which will instruct
io_iopoll_check() to break the loop and end the syscall, even if there
is not enough CQEs or none at all.

Don't return positive numbers, so io_iopoll_check() exits only when it
gets an actual error, need reschedule or got enough CQEs.

Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Link: https://lore.kernel.org/r/641a88f751623b6758303b3171f0a4141f06726e.1628471125.git.asml.silence@gmail.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
4 years agoio_uring: refactor io_alloc_req
Pavel Begunkov [Mon, 9 Aug 2021 12:04:08 +0000 (13:04 +0100)]
io_uring: refactor io_alloc_req

Replace the main if of io_flush_cached_reqs() with inverted condition +
goto, so all the cases are handled in the same way. And also extract
io_preinit_req() to make it cleaner and easier to refer to.

Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Link: https://lore.kernel.org/r/1abcba1f7b55dc53bf1dbe95036e345ffb1d5b01.1628471125.git.asml.silence@gmail.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
4 years agoio-wq: improve wq_list_add_tail()
Pavel Begunkov [Mon, 9 Aug 2021 12:04:07 +0000 (13:04 +0100)]
io-wq: improve wq_list_add_tail()

Prepare nodes that we're going to add before actually linking them, it's
always safer and costs us nothing.

Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Link: https://lore.kernel.org/r/f7e53f0c84c02ed6748c488ed0789b98f8cc6185.1628471125.git.asml.silence@gmail.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
4 years agoio_uring: remove unnecessary PF_EXITING check
Pavel Begunkov [Mon, 9 Aug 2021 12:04:06 +0000 (13:04 +0100)]
io_uring: remove unnecessary PF_EXITING check

We prefer nornal task_works even if it would fail requests inside. Kill
a PF_EXITING check in io_req_task_work_add(), task_work_add() handles
well dying tasks, i.e. return error when can't enqueue due to late
stages of do_exit().

Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Link: https://lore.kernel.org/r/fc14297e8441cd8f5d1743a2488cf0df09bf48ac.1628471125.git.asml.silence@gmail.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
4 years agoio_uring: clean io-wq callbacks
Pavel Begunkov [Mon, 9 Aug 2021 12:04:05 +0000 (13:04 +0100)]
io_uring: clean io-wq callbacks

Move io-wq callbacks closer to each other, so it's easier to work with
them, and rename io_free_work() into io_wq_free_work() for consistency.

Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Link: https://lore.kernel.org/r/851bbc7f0f86f206d8c1333efee8bcb9c26e419f.1628471125.git.asml.silence@gmail.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
4 years agoio_uring: avoid touching inode in rw prep
Pavel Begunkov [Mon, 9 Aug 2021 12:04:04 +0000 (13:04 +0100)]
io_uring: avoid touching inode in rw prep

If we use fixed files, we can be sure (almost) that REQ_F_ISREG is set.
However, for non-reg files io_prep_rw() still will look into inode to
double check, and that's expensive and can be avoided.

The only caveat is that it only currently works with 64+ bit
architectures, see FFS_ISREG, so we should consider that.

Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Link: https://lore.kernel.org/r/0a62780c491ca2522cd52db4ae3f16e03aafed0f.1628471125.git.asml.silence@gmail.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
4 years agoio_uring: rename io_file_supports_async()
Pavel Begunkov [Mon, 9 Aug 2021 12:04:03 +0000 (13:04 +0100)]
io_uring: rename io_file_supports_async()

io_file_supports_async() checks whether a file supports nowait
operations, so "async" in the name is misleading. Rename it.

Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Link: https://lore.kernel.org/r/33d55b5ce43aa1884c637c1957f1e30d30dc3bec.1628471125.git.asml.silence@gmail.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
4 years agoio_uring: inline fixed part of io_file_get()
Pavel Begunkov [Mon, 9 Aug 2021 12:04:02 +0000 (13:04 +0100)]
io_uring: inline fixed part of io_file_get()

Optimise io_file_get() with registered files, which is in a hot path,
by inlining parts of the function. Saves a function call, and
inefficiencies of passing arguments, e.g. evaluating
(sqe_flags & IOSQE_FIXED_FILE).

It couldn't have been done before as compilers were refusing to inline
it because of the function size.

Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Link: https://lore.kernel.org/r/52115cd6ce28f33bd0923149c0e6cb611084a0b1.1628471125.git.asml.silence@gmail.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
4 years agoio_uring: use kvmalloc for fixed files
Pavel Begunkov [Mon, 9 Aug 2021 12:04:01 +0000 (13:04 +0100)]
io_uring: use kvmalloc for fixed files

Instead of hand-coded two-level tables for registered files, allocate
them with kvmalloc(). In many cases small enough tables are enough, and
so can be kmalloc()'ed removing an extra memory load and a bunch of bit
logic instructions from the hot path. If the table is larger, we trade
off all the pros with a TLB-assisted memory lookup.

Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Link: https://lore.kernel.org/r/280421d3b48775dabab773006bb5588c7b2dabc0.1628471125.git.asml.silence@gmail.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
4 years agoio_uring: be smarter about waking multiple CQ ring waiters
Jens Axboe [Fri, 6 Aug 2021 20:04:31 +0000 (14:04 -0600)]
io_uring: be smarter about waking multiple CQ ring waiters

Currently we only wake the first waiter, even if we have enough entries
posted to satisfy multiple waiters. Improve that situation so that
every waiter knows how much the CQ tail has to advance before they can
be safely woken up.

With this change, if we have N waiters each asking for 1 event and we get
4 completions, then we wake up 4 waiters. If we have N waiters asking
for 2 completions and we get 4 completions, then we wake up the first
two. Previously, only the first waiter would've been woken up.

Signed-off-by: Jens Axboe <axboe@kernel.dk>
4 years agoio-wq: remove GFP_ATOMIC allocation off schedule out path
Jens Axboe [Wed, 4 Aug 2021 14:37:25 +0000 (08:37 -0600)]
io-wq: remove GFP_ATOMIC allocation off schedule out path

Daniel reports that the v5.14-rc4-rt4 kernel throws a BUG when running
stress-ng:

| [   90.202543] BUG: sleeping function called from invalid context at kernel/locking/spinlock_rt.c:35
| [   90.202549] in_atomic(): 1, irqs_disabled(): 1, non_block: 0, pid: 2047, name: iou-wrk-2041
| [   90.202555] CPU: 5 PID: 2047 Comm: iou-wrk-2041 Tainted: G        W         5.14.0-rc4-rt4+ #89
| [   90.202559] Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS 1.14.0-2 04/01/2014
| [   90.202561] Call Trace:
| [   90.202577]  dump_stack_lvl+0x34/0x44
| [   90.202584]  ___might_sleep.cold+0x87/0x94
| [   90.202588]  rt_spin_lock+0x19/0x70
| [   90.202593]  ___slab_alloc+0xcb/0x7d0
| [   90.202598]  ? newidle_balance.constprop.0+0xf5/0x3b0
| [   90.202603]  ? dequeue_entity+0xc3/0x290
| [   90.202605]  ? io_wqe_dec_running.isra.0+0x98/0xe0
| [   90.202610]  ? pick_next_task_fair+0xb9/0x330
| [   90.202612]  ? __schedule+0x670/0x1410
| [   90.202615]  ? io_wqe_dec_running.isra.0+0x98/0xe0
| [   90.202618]  kmem_cache_alloc_trace+0x79/0x1f0
| [   90.202621]  io_wqe_dec_running.isra.0+0x98/0xe0
| [   90.202625]  io_wq_worker_sleeping+0x37/0x50
| [   90.202628]  schedule+0x30/0xd0
| [   90.202630]  schedule_timeout+0x8f/0x1a0
| [   90.202634]  ? __bpf_trace_tick_stop+0x10/0x10
| [   90.202637]  io_wqe_worker+0xfd/0x320
| [   90.202641]  ? finish_task_switch.isra.0+0xd3/0x290
| [   90.202644]  ? io_worker_handle_work+0x670/0x670
| [   90.202646]  ? io_worker_handle_work+0x670/0x670
| [   90.202649]  ret_from_fork+0x22/0x30

which is due to the RT kernel not liking a GFP_ATOMIC allocation inside
a raw spinlock. Besides that not working on RT, doing any kind of
allocation from inside schedule() is kind of nasty and should be avoided
if at all possible.

This particular path happens when an io-wq worker goes to sleep, and we
need a new worker to handle pending work. We currently allocate a small
data item to hold the information we need to create a new worker, but we
can instead include this data in the io_worker struct itself and just
protect it with a single bit lock. We only really need one per worker
anyway, as we will have run pending work between to sleep cycles.

https://lore.kernel.org/lkml/20210804082418.fbibprcwtzyt5qax@beryllium.lan/
Reported-by: Daniel Wagner <dwagner@suse.de>
Tested-by: Daniel Wagner <dwagner@suse.de>
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
4 years agoMerge branch 'io_uring-5.14' into for-5.15/io_uring
Jens Axboe [Tue, 10 Aug 2021 17:25:47 +0000 (11:25 -0600)]
Merge branch 'io_uring-5.14' into for-5.15/io_uring

* io_uring-5.14:
  io_uring: fix ctx-exit io_rsrc_put_work() deadlock
  io_uring: drop ctx->uring_lock before flushing work item
  io-wq: fix IO_WORKER_F_FIXED issue in create_io_worker()
  io-wq: fix bug of creating io-wokers unconditionally
  io_uring: rsrc ref lock needs to be IRQ safe
  io_uring: Use WRITE_ONCE() when writing to sq_flags
  io_uring: clear TIF_NOTIFY_SIGNAL when running task work

4 years agoio_uring: fix ctx-exit io_rsrc_put_work() deadlock
Pavel Begunkov [Tue, 10 Aug 2021 01:44:23 +0000 (02:44 +0100)]
io_uring: fix ctx-exit io_rsrc_put_work() deadlock

__io_rsrc_put_work() might need ->uring_lock, so nobody should wait for
rsrc nodes holding the mutex. However, that's exactly what
io_ring_ctx_free() does with io_wait_rsrc_data().

Split it into rsrc wait + dealloc, and move the first one out of the
lock.

Cc: stable@vger.kernel.org
Fixes: b60c8dce33895 ("io_uring: preparation for rsrc tagging")
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Link: https://lore.kernel.org/r/0130c5c2693468173ec1afab714e0885d2c9c363.1628559783.git.asml.silence@gmail.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
4 years agoio_uring: drop ctx->uring_lock before flushing work item
Jens Axboe [Mon, 9 Aug 2021 14:15:50 +0000 (08:15 -0600)]
io_uring: drop ctx->uring_lock before flushing work item

Ammar reports that he's seeing a lockdep splat on running test/rsrc_tags
from the regression suite:

======================================================
WARNING: possible circular locking dependency detected
5.14.0-rc3-bluetea-test-00249-gc7d102232649 #5 Tainted: G           OE
------------------------------------------------------
kworker/2:4/2684 is trying to acquire lock:
ffff88814bb1c0a8 (&ctx->uring_lock){+.+.}-{3:3}, at: io_rsrc_put_work+0x13d/0x1a0

but task is already holding lock:
ffffc90001c6be70 ((work_completion)(&(&ctx->rsrc_put_work)->work)){+.+.}-{0:0}, at: process_one_work+0x1bc/0x530

which lock already depends on the new lock.

the existing dependency chain (in reverse order) is:

-> #1 ((work_completion)(&(&ctx->rsrc_put_work)->work)){+.+.}-{0:0}:
       __flush_work+0x31b/0x490
       io_rsrc_ref_quiesce.part.0.constprop.0+0x35/0xb0
       __do_sys_io_uring_register+0x45b/0x1060
       do_syscall_64+0x35/0xb0
       entry_SYSCALL_64_after_hwframe+0x44/0xae

-> #0 (&ctx->uring_lock){+.+.}-{3:3}:
       __lock_acquire+0x119a/0x1e10
       lock_acquire+0xc8/0x2f0
       __mutex_lock+0x86/0x740
       io_rsrc_put_work+0x13d/0x1a0
       process_one_work+0x236/0x530
       worker_thread+0x52/0x3b0
       kthread+0x135/0x160
       ret_from_fork+0x1f/0x30

other info that might help us debug this:

 Possible unsafe locking scenario:

       CPU0                    CPU1
       ----                    ----
  lock((work_completion)(&(&ctx->rsrc_put_work)->work));
                               lock(&ctx->uring_lock);
                               lock((work_completion)(&(&ctx->rsrc_put_work)->work));
  lock(&ctx->uring_lock);

 *** DEADLOCK ***

2 locks held by kworker/2:4/2684:
 #0: ffff88810004d938 ((wq_completion)events){+.+.}-{0:0}, at: process_one_work+0x1bc/0x530
 #1: ffffc90001c6be70 ((work_completion)(&(&ctx->rsrc_put_work)->work)){+.+.}-{0:0}, at: process_one_work+0x1bc/0x530

stack backtrace:
CPU: 2 PID: 2684 Comm: kworker/2:4 Tainted: G           OE     5.14.0-rc3-bluetea-test-00249-gc7d102232649 #5
Hardware name: Acer Aspire ES1-421/OLVIA_BE, BIOS V1.05 07/02/2015
Workqueue: events io_rsrc_put_work
Call Trace:
 dump_stack_lvl+0x6a/0x9a
 check_noncircular+0xfe/0x110
 __lock_acquire+0x119a/0x1e10
 lock_acquire+0xc8/0x2f0
 ? io_rsrc_put_work+0x13d/0x1a0
 __mutex_lock+0x86/0x740
 ? io_rsrc_put_work+0x13d/0x1a0
 ? io_rsrc_put_work+0x13d/0x1a0
 ? io_rsrc_put_work+0x13d/0x1a0
 ? process_one_work+0x1ce/0x530
 io_rsrc_put_work+0x13d/0x1a0
 process_one_work+0x236/0x530
 worker_thread+0x52/0x3b0
 ? process_one_work+0x530/0x530
 kthread+0x135/0x160
 ? set_kthread_struct+0x40/0x40
 ret_from_fork+0x1f/0x30

which is due to holding the ctx->uring_lock when flushing existing
pending work, while the pending work flushing may need to grab the uring
lock if we're using IOPOLL.

Fix this by dropping the uring_lock a bit earlier as part of the flush.

Cc: stable@vger.kernel.org
Link: https://github.com/axboe/liburing/issues/404
Tested-by: Ammar Faizi <ammarfaizi2@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
4 years agoio-wq: fix IO_WORKER_F_FIXED issue in create_io_worker()
Hao Xu [Sun, 8 Aug 2021 13:54:34 +0000 (21:54 +0800)]
io-wq: fix IO_WORKER_F_FIXED issue in create_io_worker()

There may be cases like:
        A                                 B
spin_lock(wqe->lock)
nr_workers is 0
nr_workers++
spin_unlock(wqe->lock)
                                     spin_lock(wqe->lock)
                                     nr_wokers is 1
                                     nr_workers++
                                     spin_unlock(wqe->lock)
create_io_worker()
  acct->worker is 1
                                     create_io_worker()
                                       acct->worker is 1

There should be one worker marked IO_WORKER_F_FIXED, but no one is.
Fix this by introduce a new agrument for create_io_worker() to indicate
if it is the first worker.

Fixes: 3d4e4face9c1 ("io-wq: fix no lock protection of acct->nr_worker")
Signed-off-by: Hao Xu <haoxu@linux.alibaba.com>
Link: https://lore.kernel.org/r/20210808135434.68667-3-haoxu@linux.alibaba.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
4 years agoio-wq: fix bug of creating io-wokers unconditionally
Hao Xu [Sun, 8 Aug 2021 13:54:33 +0000 (21:54 +0800)]
io-wq: fix bug of creating io-wokers unconditionally

The former patch to add check between nr_workers and max_workers has a
bug, which will cause unconditionally creating io-workers. That's
because the result of the check doesn't affect the call of
create_io_worker(), fix it by bringing in a boolean value for it.

Fixes: 21698274da5b ("io-wq: fix lack of acct->nr_workers < acct->max_workers judgement")
Signed-off-by: Hao Xu <haoxu@linux.alibaba.com>
Link: https://lore.kernel.org/r/20210808135434.68667-2-haoxu@linux.alibaba.com
[axboe: drop hunk that isn't strictly needed]
Signed-off-by: Jens Axboe <axboe@kernel.dk>
4 years agoio_uring: rsrc ref lock needs to be IRQ safe
Jens Axboe [Mon, 9 Aug 2021 13:49:41 +0000 (07:49 -0600)]
io_uring: rsrc ref lock needs to be IRQ safe

Nadav reports running into the below splat on re-enabling softirqs:

WARNING: CPU: 2 PID: 1777 at kernel/softirq.c:364 __local_bh_enable_ip+0xaa/0xe0
Modules linked in:
CPU: 2 PID: 1777 Comm: umem Not tainted 5.13.1+ #161
Hardware name: VMware, Inc. VMware Virtual Platform/440BX Desktop Reference Platform, BIOS 6.00 07/22/2020
RIP: 0010:__local_bh_enable_ip+0xaa/0xe0
Code: a9 00 ff ff 00 74 38 65 ff 0d a2 21 8c 7a e8 ed 1a 20 00 fb 66 0f 1f 44 00 00 5b 41 5c 5d c3 65 8b 05 e6 2d 8c 7a 85 c0 75 9a <0f> 0b eb 96 e8 2d 1f 20 00 eb a5 4c 89 e7 e8 73 4f 0c 00 eb ae 65
RSP: 0018:ffff88812e58fcc8 EFLAGS: 00010046
RAX: 0000000000000000 RBX: 0000000000000201 RCX: dffffc0000000000
RDX: 0000000000000007 RSI: 0000000000000201 RDI: ffffffff8898c5ac
RBP: ffff88812e58fcd8 R08: ffffffff8575dbbf R09: ffffed1028ef14f9
R10: ffff88814778a7c3 R11: ffffed1028ef14f8 R12: ffffffff85c9e9ae
R13: ffff88814778a000 R14: ffff88814778a7b0 R15: ffff8881086db890
FS:  00007fbcfee17700(0000) GS:ffff8881e0300000(0000) knlGS:0000000000000000
CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: 000000c0402a5008 CR3: 000000011c1ac003 CR4: 00000000003706e0
Call Trace:
 _raw_spin_unlock_bh+0x31/0x40
 io_rsrc_node_ref_zero+0x13e/0x190
 io_dismantle_req+0x215/0x220
 io_req_complete_post+0x1b8/0x720
 __io_complete_rw.isra.0+0x16b/0x1f0
 io_complete_rw+0x10/0x20

where it's clear we end up calling the percpu count release directly
from the completion path, as it's in atomic mode and we drop the last
ref. For file/block IO, this can be from IRQ context already, and the
softirq locking for rsrc isn't enough.

Just make the lock fully IRQ safe, and ensure we correctly safe state
from the release path as we don't know the full context there.

Reported-by: Nadav Amit <nadav.amit@gmail.com>
Tested-by: Nadav Amit <nadav.amit@gmail.com>
Link: https://lore.kernel.org/io-uring/C187C836-E78B-4A31-B24C-D16919ACA093@gmail.com/
Signed-off-by: Jens Axboe <axboe@kernel.dk>
4 years agoio_uring: Use WRITE_ONCE() when writing to sq_flags
Nadav Amit [Sun, 8 Aug 2021 00:13:42 +0000 (17:13 -0700)]
io_uring: Use WRITE_ONCE() when writing to sq_flags

The compiler should be forbidden from any strange optimization for async
writes to user visible data-structures. Without proper protection, the
compiler can cause write-tearing or invent writes that would confuse the
userspace.

However, there are writes to sq_flags which are not protected by
WRITE_ONCE(). Use WRITE_ONCE() for these writes.

This is purely a theoretical issue. Presumably, any compiler is very
unlikely to do such optimizations.

Fixes: 75b28affdd6a ("io_uring: allocate the two rings together")
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Link: https://lore.kernel.org/r/20210808001342.964634-3-namit@vmware.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
4 years agoio_uring: clear TIF_NOTIFY_SIGNAL when running task work
Nadav Amit [Sun, 8 Aug 2021 00:13:41 +0000 (17:13 -0700)]
io_uring: clear TIF_NOTIFY_SIGNAL when running task work

When using SQPOLL, the submission queue polling thread calls
task_work_run() to run queued work. However, when work is added with
TWA_SIGNAL - as done by io_uring itself - the TIF_NOTIFY_SIGNAL remains
set afterwards and is never cleared.

Consequently, when the submission queue polling thread checks whether
signal_pending(), it may always find a pending signal, if
task_work_add() was ever called before.

The impact of this bug might be different on different kernel versions.
It appears that on 5.14 it would only cause unnecessary calculation and
prevent the polling thread from sleeping. On 5.13, where the bug was
found, it stops the polling thread from finding newly submitted work.

Instead of task_work_run(), use tracehook_notify_signal() that clears
TIF_NOTIFY_SIGNAL. Test for TIF_NOTIFY_SIGNAL in addition to
current->task_works to avoid a race in which task_works is cleared but
the TIF_NOTIFY_SIGNAL is set.

Fixes: 685fe7feedb96 ("io-wq: eliminate the need for a manager thread")
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: Nadav Amit <namit@vmware.com>
Link: https://lore.kernel.org/r/20210808001342.964634-2-namit@vmware.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
4 years agoLinux 5.14-rc5 v5.14-rc5
Linus Torvalds [Sun, 8 Aug 2021 20:49:31 +0000 (13:49 -0700)]
Linux 5.14-rc5

4 years agoMerge tag 'timers-urgent-2021-08-08' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sun, 8 Aug 2021 18:53:30 +0000 (11:53 -0700)]
Merge tag 'timers-urgent-2021-08-08' of git://git./linux/kernel/git/tip/tip

Pull timer fix from Thomas Gleixner:
 "A single timer fix:

   - Prevent a memory ordering issue in the timer expiry code which
     makes it possible to observe falsely that the callback has been
     executed already while that's not the case, which violates the
     guarantee of del_timer_sync()"

* tag 'timers-urgent-2021-08-08' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  timers: Move clearing of base::timer_running under base:: Lock

4 years agoMerge tag 'sched-urgent-2021-08-08' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sun, 8 Aug 2021 18:50:07 +0000 (11:50 -0700)]
Merge tag 'sched-urgent-2021-08-08' of git://git./linux/kernel/git/tip/tip

Pull scheduler fix from Thomas Gleixner:
 "A single scheduler fix:

   - Prevent a double enqueue caused by rt_effective_prio() being
     invoked twice in __sched_setscheduler()"

* tag 'sched-urgent-2021-08-08' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  sched/rt: Fix double enqueue caused by rt_effective_prio

4 years agoMerge tag 'perf-urgent-2021-08-08' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Sun, 8 Aug 2021 18:46:13 +0000 (11:46 -0700)]
Merge tag 'perf-urgent-2021-08-08' of git://git./linux/kernel/git/tip/tip

Pull perf fixes from Thomas Gleixner:
 "A set of perf fixes:

   - Correct the permission checks for perf event which send SIGTRAP to
     a different process and clean up that code to be more readable.

   - Prevent an out of bound MSR access in the x86 perf code which
     happened due to an incomplete limiting to the actually available
     hardware counters.

   - Prevent access to the AMD64_EVENTSEL_HOSTONLY bit when running
     inside a guest.

   - Handle small core counter re-enabling correctly by issuing an ACK
     right before reenabling it to prevent a stale PEBS record being
     kept around"

* tag 'perf-urgent-2021-08-08' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  perf/x86/intel: Apply mid ACK for small core
  perf/x86/amd: Don't touch the AMD64_EVENTSEL_HOSTONLY bit inside the guest
  perf/x86: Fix out of bound MSR access
  perf: Refactor permissions check into perf_check_permission()
  perf: Fix required permissions if sigtrap is requested

4 years agoMerge tag 'char-misc-5.14-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/gregk...
Linus Torvalds [Sun, 8 Aug 2021 17:41:49 +0000 (10:41 -0700)]
Merge tag 'char-misc-5.14-rc5' of git://git./linux/kernel/git/gregkh/char-misc

Pull char/misc driver fixes from Greg KH:
 "Here are some small char/misc driver fixes for 5.14-rc5.

  They resolve a few regressions that people reported:

   - acrn driver fix

   - fpga driver fix

   - interconnect tiny driver fixes

  All have been in linux-next for a while with no reported issues"

* tag 'char-misc-5.14-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/char-misc:
  interconnect: Fix undersized devress_alloc allocation
  interconnect: qcom: icc-rpmh: Add BCMs to commit list in pre_aggregate
  interconnect: qcom: icc-rpmh: Ensure floor BW is enforced for all nodes
  fpga: dfl: fme: Fix cpu hotplug issue in performance reporting
  virt: acrn: Do hcall_destroy_vm() before resource release
  interconnect: Always call pre_aggregate before aggregate
  interconnect: Zero initial BW after sync-state

4 years agoMerge tag 'driver-core-5.14-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Sun, 8 Aug 2021 17:34:49 +0000 (10:34 -0700)]
Merge tag 'driver-core-5.14-rc5' of git://git./linux/kernel/git/gregkh/driver-core

Pull driver core fixes from Greg KH:
 "Here are three tiny driver core and firmware loader fixes for
  5.14-rc5. They are:

   - driver core fix for when probing fails

   - firmware loader fixes for reported problems.

  All have been in linux-next for a while with no reported issues"

* tag 'driver-core-5.14-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/driver-core:
  firmware_loader: fix use-after-free in firmware_fallback_sysfs
  firmware_loader: use -ETIMEDOUT instead of -EAGAIN in fw_load_sysfs_fallback
  drivers core: Fix oops when driver probe fails

4 years agoMerge tag 'staging-5.14-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh...
Linus Torvalds [Sun, 8 Aug 2021 17:29:39 +0000 (10:29 -0700)]
Merge tag 'staging-5.14-rc5' of git://git./linux/kernel/git/gregkh/staging

Pull staging driver fixes from Greg KH:
 "Here are a few small staging driver fixes for 5.14-rc5 to resolve some
  reported problems. They include:

   - mt7621 driver fix

   - rtl8723bs driver fixes

   - rtl8712 driver fixes.

  Nothing major, just small problems resolved.

  All have been in linux-next for a while with no reported issues"

* tag 'staging-5.14-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/staging:
  staging: mt7621-pci: avoid to re-disable clock for those pcies not in use
  staging: rtl8712: error handling refactoring
  staging: rtl8712: get rid of flush_scheduled_work
  staging: rtl8723bs: select CONFIG_CRYPTO_LIB_ARC4
  staging: rtl8723bs: Fix a resource leak in sd_int_dpc

4 years agoMerge tag 'tty-5.14-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/tty
Linus Torvalds [Sun, 8 Aug 2021 17:23:13 +0000 (10:23 -0700)]
Merge tag 'tty-5.14-rc5' of git://git./linux/kernel/git/gregkh/tty

Pull tty/serial fixes from Greg KH:
 "Here are some small tty/serial driver fixes for 5.14-rc5 to resolve a
  number of reported problems.

  They include:

   - mips serial driver fixes

   - 8250 driver fixes for reported problems

   - fsl_lpuart driver fixes

   - other tiny driver fixes

  All have been in linux-next for a while with no reported problems"

* tag 'tty-5.14-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/tty:
  serial: 8250_pci: Avoid irq sharing for MSI(-X) interrupts.
  serial: 8250_mtk: fix uart corruption issue when rx power off
  tty: serial: fsl_lpuart: fix the wrong return value in lpuart32_get_mctrl
  serial: 8250_pci: Enumerate Elkhart Lake UARTs via dedicated driver
  serial: 8250: fix handle_irq locking
  serial: tegra: Only print FIFO error message when an error occurs
  MIPS: Malta: Do not byte-swap accesses to the CBUS UART
  serial: 8250: Mask out floating 16/32-bit bus bits
  serial: max310x: Unprepare and disable clock in error path

4 years agoMerge tag 'usb-5.14-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usb
Linus Torvalds [Sun, 8 Aug 2021 17:20:05 +0000 (10:20 -0700)]
Merge tag 'usb-5.14-rc5' of git://git./linux/kernel/git/gregkh/usb

Pull USB driver fixes from Greg KH:
 "Here are some small USB driver fixes for 5.14-rc5. They resolve a
  number of small reported issues, including:

   - cdnsp driver fixes

   - usb serial driver fixes and device id updates

   - usb gadget hid fixes

   - usb host driver fixes

   - usb dwc3 driver fixes

   - other usb gadget driver fixes

  All of these have been in linux-next for a while with no reported
  issues"

* tag 'usb-5.14-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usb: (21 commits)
  usb: typec: tcpm: Keep other events when receiving FRS and Sourcing_vbus events
  usb: dwc3: gadget: Avoid runtime resume if disabling pullup
  usb: dwc3: gadget: Use list_replace_init() before traversing lists
  USB: serial: ftdi_sio: add device ID for Auto-M3 OP-COM v2
  USB: serial: pl2303: fix GT type detection
  USB: serial: option: add Telit FD980 composition 0x1056
  USB: serial: pl2303: fix HX type detection
  USB: serial: ch341: fix character loss at high transfer rates
  usb: cdnsp: Fix the IMAN_IE_SET and IMAN_IE_CLEAR macro
  usb: cdnsp: Fixed issue with ZLP
  usb: cdnsp: Fix incorrect supported maximum speed
  usb: cdns3: Fixed incorrect gadget state
  usb: gadget: f_hid: idle uses the highest byte for duration
  Revert "thunderbolt: Hide authorized attribute if router does not support PCIe tunnels"
  usb: otg-fsm: Fix hrtimer list corruption
  usb: host: ohci-at91: suspend/resume ports after/before OHCI accesses
  usb: musb: Fix suspend and resume issues for PHYs on I2C and SPI
  usb: gadget: f_hid: added GET_IDLE and SET_IDLE handlers
  usb: gadget: f_hid: fixed NULL pointer dereference
  usb: gadget: remove leaked entry from udc driver list
  ...

4 years agoMerge tag 'io_uring-5.14-2021-08-07' of git://git.kernel.dk/linux-block
Linus Torvalds [Sat, 7 Aug 2021 17:34:26 +0000 (10:34 -0700)]
Merge tag 'io_uring-5.14-2021-08-07' of git://git.kernel.dk/linux-block

Pull io_uring from Jens Axboe:
 "A few io-wq related fixes:

   - Fix potential nr_worker race and missing max_workers check from one
     path (Hao)

   - Fix race between worker exiting and new work queue (me)"

* tag 'io_uring-5.14-2021-08-07' of git://git.kernel.dk/linux-block:
  io-wq: fix lack of acct->nr_workers < acct->max_workers judgement
  io-wq: fix no lock protection of acct->nr_worker
  io-wq: fix race between worker exiting and activating free worker

4 years agoMerge tag 'block-5.14-2021-08-07' of git://git.kernel.dk/linux-block
Linus Torvalds [Sat, 7 Aug 2021 17:26:21 +0000 (10:26 -0700)]
Merge tag 'block-5.14-2021-08-07' of git://git.kernel.dk/linux-block

Pull block fixes from Jens Axboe:
 "A few minor fixes:

   - Fix ldm kernel-doc warning (Bart)

   - Fix adding offset twice for DMA address in n64cart (Christoph)

   - Fix use-after-free in dasd path handling (Stefan)

   - Order kyber insert trace correctly (Vincent)

   - raid1 errored write handling fix (Wei)

   - Fix blk-iolatency queue get failure handling (Yu)"

* tag 'block-5.14-2021-08-07' of git://git.kernel.dk/linux-block:
  kyber: make trace_block_rq call consistent with documentation
  block/partitions/ldm.c: Fix a kernel-doc warning
  blk-iolatency: error out if blk_get_queue() failed in iolatency_set_limit()
  n64cart: fix the dma address in n64cart_do_bvec
  s390/dasd: fix use after free in dasd path handling
  md/raid10: properly indicate failure when ending a failed write request

4 years agoMerge tag 'riscv-for-linus-5.14-rc5' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sat, 7 Aug 2021 17:15:04 +0000 (10:15 -0700)]
Merge tag 'riscv-for-linus-5.14-rc5' of git://git./linux/kernel/git/riscv/linux

Pull RISC-V fixes from Palmer Dabbelt:

 - avoid dereferencing a null task pointer while walking the stack

 - fix the memory size in the HiFive Unleashed device tree

 - disable stack protectors when randstruct is enabled, which results in
   non-deterministic offsets during module builds

 - a pair of fixes to avoid relying on a constant physical memory base
   for the non-XIP builds

* tag 'riscv-for-linus-5.14-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/riscv/linux:
  Revert "riscv: Remove CONFIG_PHYS_RAM_BASE_FIXED"
  riscv: Get rid of CONFIG_PHYS_RAM_BASE in kernel physical address conversion
  riscv: Disable STACKPROTECTOR_PER_TASK if GCC_PLUGIN_RANDSTRUCT is enabled
  riscv: dts: fix memory size for the SiFive HiFive Unmatched
  riscv: stacktrace: Fix NULL pointer dereference

4 years agoMerge tag 'kbuild-fixes-v5.14-2' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Sat, 7 Aug 2021 17:03:02 +0000 (10:03 -0700)]
Merge tag 'kbuild-fixes-v5.14-2' of git://git./linux/kernel/git/masahiroy/linux-kbuild

Pull Kbuild fixes from Masahiro Yamada:

 - Correct the Extended Regular Expressions in tools

 - Adjust scripts/checkversion.pl for the current Kbuild

 - Unset sub_make_done for 'make install' to make DKMS work again

* tag 'kbuild-fixes-v5.14-2' of git://git.kernel.org/pub/scm/linux/kernel/git/masahiroy/linux-kbuild:
  kbuild: cancel sub_make_done for the install target to fix DKMS
  scripts: checkversion: modernize linux/version.h search strings
  mips: Fix non-POSIX regexp
  x86/tools/relocs: Fix non-POSIX regexp

4 years agoRevert "riscv: Remove CONFIG_PHYS_RAM_BASE_FIXED"
Alexandre Ghiti [Wed, 21 Jul 2021 07:59:36 +0000 (09:59 +0200)]
Revert "riscv: Remove CONFIG_PHYS_RAM_BASE_FIXED"

This reverts commit 9b79878ced8f7ab85c57623f8b1f6882e484a316.

The removal of this config exposes CONFIG_PHYS_RAM_BASE for all kernel
types: this value being implementation-specific, this breaks the
genericity of the RISC-V kernel so revert it.

Signed-off-by: Alexandre Ghiti <alex@ghiti.fr>
Tested-by: Emil Renner Berthing <kernel@esmil.dk>
Reviewed-by: Jisheng Zhang <jszhang@kernel.org>
Cc: stable@vger.kernel.org
Signed-off-by: Palmer Dabbelt <palmerdabbelt@google.com>
4 years agoriscv: Get rid of CONFIG_PHYS_RAM_BASE in kernel physical address conversion
Alexandre Ghiti [Wed, 21 Jul 2021 07:59:35 +0000 (09:59 +0200)]
riscv: Get rid of CONFIG_PHYS_RAM_BASE in kernel physical address conversion

The usage of CONFIG_PHYS_RAM_BASE for all kernel types was a mistake:
this value is implementation-specific and this breaks the genericity of
the RISC-V kernel.

Fix this by introducing a new variable phys_ram_base that holds this
value at runtime and use it in the kernel physical address conversion
macro. Since this value is used only for XIP kernels, evaluate it only if
CONFIG_XIP_KERNEL is set which in addition optimizes this macro for
standard kernels at compile-time.

Signed-off-by: Alexandre Ghiti <alex@ghiti.fr>
Tested-by: Emil Renner Berthing <kernel@esmil.dk>
Reviewed-by: Jisheng Zhang <jszhang@kernel.org>
Fixes: 44c922572952 ("RISC-V: enable XIP")
Cc: stable@vger.kernel.org
Signed-off-by: Palmer Dabbelt <palmerdabbelt@google.com>
4 years agokyber: make trace_block_rq call consistent with documentation block-5.14-2021-08-07
Vincent Fu [Wed, 4 Aug 2021 19:49:23 +0000 (19:49 +0000)]
kyber: make trace_block_rq call consistent with documentation

The kyber ioscheduler calls trace_block_rq_insert() *after* the request
is added to the queue but the documentation for trace_block_rq_insert()
says that the call should be made *before* the request is added to the
queue.  Move the tracepoint for the kyber ioscheduler so that it is
consistent with the documentation.

Signed-off-by: Vincent Fu <vincent.fu@samsung.com>
Link: https://lore.kernel.org/r/20210804194913.10497-1-vincent.fu@samsung.com
Reviewed by: Adam Manzanares <a.manzanares@samsung.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
4 years agoMerge tag 'ext4_for_linus_stable' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Fri, 6 Aug 2021 19:55:07 +0000 (12:55 -0700)]
Merge tag 'ext4_for_linus_stable' of git://git./linux/kernel/git/tytso/ext4

Pull ext4 fixes from Ted Ts'o:
 "A regression fix, bug fix, and a comment cleanup for ext4"

* tag 'ext4_for_linus_stable' of git://git.kernel.org/pub/scm/linux/kernel/git/tytso/ext4:
  ext4: fix potential htree corruption when growing large_dir directories
  ext4: remove conflicting comment from __ext4_forget
  ext4: fix potential uninitialized access to retval in kmmpd

4 years agoMerge tag 'trace-v5.14-rc4-2' of git://git.kernel.org/pub/scm/linux/kernel/git/rosted...
Linus Torvalds [Fri, 6 Aug 2021 19:36:46 +0000 (12:36 -0700)]
Merge tag 'trace-v5.14-rc4-2' of git://git./linux/kernel/git/rostedt/linux-trace

Pull tracing fixes from Steven Rostedt:
 "Fix tracepoint race between static_call and callback data

  As callbacks to a tracepoint are paired with the data that is passed
  in when the callback is registered to the tracepoint, it must have
  that data passed to the callback when the tracepoint is triggered,
  else bad things will happen. To keep the two together, they are both
  assigned to a tracepoint structure and added to an array. The
  tracepoint call site will dereference the structure (via RCU) and call
  the callback in that structure along with the data in that structure.
  This keeps the callback and data tightly coupled.

  Because of the overhead that retpolines have on tracepoint callbacks,
  if there's only one callback attached to a tracepoint (a common case),
  then it is called via a static call (code modified to do a direct call
  instead of an indirect call). But to implement this, the data had to
  be decoupled from the callback, as now the callback is implemented via
  a direct call from the static call and not an indirect call from the
  dereferenced structure.

  Note, the static call only calls a callback used when there's a single
  callback attached to the tracepoint. If more than one callback is
  attached to the same tracepoint, then the static call will call an
  iterator function that goes back to dereferencing the structure
  keeping the callback and its data tightly coupled again.

  Issues can arise when going from 0 callbacks to one, as the static
  call is assigned to the callback, and it must take care that the data
  passed to it is loaded before the static call calls the callback.
  Going from 1 to 2 callbacks is not an issue, as long as the static
  call is updated to the iterator before the tracepoint structure array
  is updated via RCU. Going from 2 to more or back down to 2 is not an
  issue as the iterator can handle all theses cases. But going from 2 to
  1, care must be taken as the static call is now calling a callback and
  the data that is loaded must be the data for that callback.

  Care was taken to ensure the callback and data would be in-sync, but
  after a bug was reported, it became clear that not enough was done to
  make sure that was the case. These changes address this.

  The first change is to compare the old and new data instead of the old
  and new callback, as it's the data that can corrupt the callback, even
  if the callback is the same (something getting freed).

  The next change is to convert these transitions into states, to make
  it easier to know when a synchronization is needed, and to perform
  those synchronizations. The problem with this patch is that it slows
  down disabling all events from under a second, to making it take over
  10 seconds to do the same work. But that is addressed in the final
  patch.

  The final patch uses the RCU state functions to keep track of the RCU
  state between the transitions, and only needs to perform the
  synchronization if an RCU synchronization hasn't been done already.
  This brings the performance of disabling all events back to its
  original value. That's because no synchronization is required between
  disabling tracepoints but is required when enabling a tracepoint after
  its been disabled. If an RCU synchronization happens after the
  tracepoint is disabled, and before it is re-enabled, there's no need
  to do the synchronization again.

  Both the second and third patch have subtle complexities that they are
  separated into two patches. But because the second patch causes such a
  regression in performance, the third patch adds a "Fixes" tag to the
  second patch, such that the two must be backported together and not
  just the second patch"

* tag 'trace-v5.14-rc4-2' of git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/linux-trace:
  tracepoint: Use rcu get state and cond sync for static call updates
  tracepoint: Fix static call function vs data state mismatch
  tracepoint: static call: Compare data on transition from 2->1 callees

4 years agoMerge tag 'pm-5.14-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm
Linus Torvalds [Fri, 6 Aug 2021 19:30:29 +0000 (12:30 -0700)]
Merge tag 'pm-5.14-rc5' of git://git./linux/kernel/git/rafael/linux-pm

Pull power management fixes from Rafael Wysocki:
 "Fix a recent regression in the timer events oriented (TEO) cpuidle
  governor causing it to misbehave when idle state 0 is disabled and
  rename two local variables for improved clarity on top of that"

* tag 'pm-5.14-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm:
  cpuidle: teo: Rename two local variables in teo_select()
  cpuidle: teo: Fix alternative idle state lookup

4 years agoMerge tag 'acpi-5.14-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael...
Linus Torvalds [Fri, 6 Aug 2021 19:22:47 +0000 (12:22 -0700)]
Merge tag 'acpi-5.14-rc5' of git://git./linux/kernel/git/rafael/linux-pm

Pull ACPI fix from Rafael Wysocki:
 "Revert a recent ACPICA commit causing boot issues to appear on some
  systems"

* tag 'acpi-5.14-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm:
  Revert "ACPICA: Fix memory leak caused by _CID repair function"

4 years agoMerge tag 'soc-fixes-5.14-2' of git://git.kernel.org/pub/scm/linux/kernel/git/soc/soc
Linus Torvalds [Fri, 6 Aug 2021 18:41:12 +0000 (11:41 -0700)]
Merge tag 'soc-fixes-5.14-2' of git://git./linux/kernel/git/soc/soc

Pull ARM SoC fixes from Arnd Bergmann:
 "Lots of small fixes for Arm SoCs this time, nothing too worrying:

   - omap/beaglebone boot regression fix in gpt12 timer

   - revert for i.mx8 soc driver breaking as a platform_driver

   - kexec/kdump fixes for op-tee

   - various fixes for incorrect DT settings on imx, mvebu, omap, stm32,
     and tegra causing problems.

   - device tree fixes for static checks in nomadik, versatile, stm32

   - code fixes for issues found in build testing and with static
     checking on tegra, ixp4xx, imx, omap"

* tag 'soc-fixes-5.14-2' of git://git.kernel.org/pub/scm/linux/kernel/git/soc/soc: (36 commits)
  soc: ixp4xx/qmgr: fix invalid __iomem access
  soc: ixp4xx: fix printing resources
  ARM: ixp4xx: goramo_mlr depends on old PCI driver
  ARM: ixp4xx: fix compile-testing soc drivers
  soc/tegra: Make regulator couplers depend on CONFIG_REGULATOR
  ARM: dts: nomadik: Fix up interrupt controller node names
  ARM: dts: stm32: Fix touchscreen IRQ line assignment on DHCOM
  ARM: dts: stm32: Disable LAN8710 EDPD on DHCOM
  ARM: dts: stm32: Prefer HW RTC on DHCOM SoM
  omap5-board-common: remove not physically existing vdds_1v8_main fixed-regulator
  ARM: dts: am437x-l4: fix typo in can@0 node
  ARM: dts: am43x-epos-evm: Reduce i2c0 bus speed for tps65218
  bus: ti-sysc: AM3: RNG is GP only
  ARM: omap2+: hwmod: fix potential NULL pointer access
  arm64: dts: armada-3720-turris-mox: remove mrvl,i2c-fast-mode
  arm64: dts: armada-3720-turris-mox: fixed indices for the SDHC controllers
  ARM: dts: imx: Swap M53Menlo pinctrl_power_button/pinctrl_power_out pins
  ARM: imx: fix missing 3rd argument in macro imx_mmdc_perf_init
  ARM: dts: colibri-imx6ull: limit SDIO clock to 25MHz
  arm64: dts: ls1028: sl28: fix networking for variant 2
  ...

4 years agoMerge tag 'arm64-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/arm64/linux
Linus Torvalds [Fri, 6 Aug 2021 18:33:49 +0000 (11:33 -0700)]
Merge tag 'arm64-fixes' of git://git./linux/kernel/git/arm64/linux

Pull arm64 fixes from Will Deacon:
 "It's all pretty minor but the main fix is sorting out how we deal with
  return values from 32-bit system calls as audit expects error codes to
  be sign-extended to 64 bits

  Summary:

   - Fix extension/truncation of return values from 32-bit system calls

   - Fix interaction between unwinding and tracing

   - Fix spurious toolchain warning emitted during make

   - Fix Kconfig help text for RANDOMIZE_MODULE_REGION_FULL"

* tag 'arm64-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/arm64/linux:
  arm64: stacktrace: avoid tracing arch_stack_walk()
  arm64: stacktrace: fix comment
  arm64: fix the doc of RANDOMIZE_MODULE_REGION_FULL
  arm64: move warning about toolchains to archprepare
  arm64: fix compat syscall return truncation

4 years agoMerge tag 'mips-fixes_5.14_1' of git://git.kernel.org/pub/scm/linux/kernel/git/mips...
Linus Torvalds [Fri, 6 Aug 2021 18:31:11 +0000 (11:31 -0700)]
Merge tag 'mips-fixes_5.14_1' of git://git./linux/kernel/git/mips/linux

Pull MIPS fix from Thomas Bogendoerfer:
 "Fix PMD accounting change"

* tag 'mips-fixes_5.14_1' of git://git.kernel.org/pub/scm/linux/kernel/git/mips/linux:
  MIPS: check return value of pgtable_pmd_page_ctor

4 years agoMerge tag 'spi-fix-v5.14-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/brooni...
Linus Torvalds [Fri, 6 Aug 2021 18:15:02 +0000 (11:15 -0700)]
Merge tag 'spi-fix-v5.14-rc4' of git://git./linux/kernel/git/broonie/spi

Pull spi fixes from Mark Brown:
 "A small collection of fixes for SPI, small mostly driver specific
  things plus a fix for module autoloading which hadn't been working
  properly for DT systems"

* tag 'spi-fix-v5.14-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/broonie/spi:
  spi: cadence-quadspi: Fix check condition for DTR ops
  spi: mediatek: Fix fifo transfer
  spi: imx: mx51-ecspi: Fix CONFIGREG delay comment
  spi: imx: mx51-ecspi: Fix low-speed CONFIGREG delay calculation
  spi: update modalias_show after of_device_uevent_modalias support
  spi: meson-spicc: fix memory leak in meson_spicc_remove
  spi: spi-mux: Add module info needed for autoloading

4 years agoMerge tag 'dmaengine-fix-5.14' of git://git.kernel.org/pub/scm/linux/kernel/git/vkoul...
Linus Torvalds [Fri, 6 Aug 2021 18:08:24 +0000 (11:08 -0700)]
Merge tag 'dmaengine-fix-5.14' of git://git./linux/kernel/git/vkoul/dmaengine

Pull dmaengine fixes from Vinod Koul:
 "A bunch of driver fixes, notably:

   - idxd driver fixes for submission race, driver remove sequence,
     setup sequence for MSIXPERM, array index and updating descriptor
     vector

   - usb-dmac, pm reference leak fix

   - xilinx_dma, read-after-free fix

   - uniphier-xdmac fix for using atomic readl_poll_timeout_atomic()

   - of-dma, router_xlate to return

   - imx-dma, generic dma fix"

* tag 'dmaengine-fix-5.14' of git://git.kernel.org/pub/scm/linux/kernel/git/vkoul/dmaengine:
  dmaengine: imx-dma: configure the generic DMA type to make it work
  dmaengine: of-dma: router_xlate to return -EPROBE_DEFER if controller is not yet available
  dmaengine: stm32-dmamux: Fix PM usage counter unbalance in stm32 dmamux ops
  dmaengine: stm32-dma: Fix PM usage counter imbalance in stm32 dma ops
  dmaengine: uniphier-xdmac: Use readl_poll_timeout_atomic() in atomic state
  dmaengine: idxd: fix submission race window
  dmaengine: idxd: fix sequence for pci driver remove() and shutdown()
  dmaengine: idxd: fix desc->vector that isn't being updated
  dmaengine: idxd: fix setup sequence for MSIXPERM table
  dmaengine: idxd: fix array index when int_handles are being used
  dmaengine: usb-dmac: Fix PM reference leak in usb_dmac_probe()
  dmaengine: xilinx_dma: Fix read-after-free bug when terminating transfers

4 years agoMerge tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/rdma/rdma
Linus Torvalds [Fri, 6 Aug 2021 18:03:35 +0000 (11:03 -0700)]
Merge tag 'for-linus' of git://git./linux/kernel/git/rdma/rdma

Pull rdma fixes from Jason Gunthorpe:
 "Several small recent regressions - rather more than usual, but nothing
  too scary. Good to know people are testing.

   - Typo causing incorrect operation of the mlx5 mkey cache
     expiration

   - Revert a CM patch that is breaking some ULPs

   - Typo breaking SRQ in rxe

   - Revert a rxe patch breaking icrc calculation

   - Static checker warning about unbalanced locking in hns

   - Subtle cxgb4 regression from a recent atomic to refcount
     conversion"

* tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/rdma/rdma:
  RDMA/iw_cxgb4: Fix refcount underflow while destroying cqs.
  RDMA/hns: Fix the double unlock problem of poll_sem
  RDMA/rxe: Restore setting tot_len in the IPv4 header
  RDMA/rxe: Use the correct size of wqe when processing SRQ
  RDMA/cma: Revert INIT-INIT patch
  RDMA/mlx5: Delay emptying a cache entry when a new MR is added to it recently

4 years agoMerge tag 'sound-5.14-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai...
Linus Torvalds [Fri, 6 Aug 2021 17:50:26 +0000 (10:50 -0700)]
Merge tag 'sound-5.14-rc5' of git://git./linux/kernel/git/tiwai/sound

Pull sound fixes from Takashi Iwai:
 "A collection of small fixes:

   - A few regression fixes (PCM core fixes, USB-audio fixes)

   - Follow up fixes for the USB-audio mixer changes in this cycle

   - A long-standing ALSA sequencer race bug fix

   - Usual device-specific quirks for HD- and USB-audio"

* tag 'sound-5.14-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai/sound:
  ALSA: seq: Fix racy deletion of subscriber
  ALSA: memalloc: Fix regression with SNDRV_DMA_TYPE_CONTINUOUS
  ALSA: pcm - fix mmap capability check for the snd-dummy driver
  ALSA: usb-audio: Avoid unnecessary or invalid connector selection at resume
  ALSA: hda/realtek: add mic quirk for Acer SF314-42
  ALSA: usb-audio: Add registration quirk for JBL Quantum 600
  ALSA: hda/realtek: Fix headset mic for Acer SWIFT SF314-56 (ALC256)
  ALSA: usb-audio: Fix superfluous autosuspend recovery
  ALSA: usb-audio: fix incorrect clock source setting
  ALSA: scarlett2: Fix line out/speaker switching notifications
  ALSA: scarlett2: Correct channel mute status after mute button pressed
  ALSA: scarlett2: Fix Direct Monitor control name for 2i2
  ALSA: scarlett2: Fix Mute/Dim/MSD Mode control names

4 years agoMerge tag 'drm-fixes-2021-08-06' of git://anongit.freedesktop.org/drm/drm
Linus Torvalds [Fri, 6 Aug 2021 17:42:48 +0000 (10:42 -0700)]
Merge tag 'drm-fixes-2021-08-06' of git://anongit.freedesktop.org/drm/drm

Pull drm fixes from Dave Airlie:
 "Regular weekly fixes pull, live from a Brisbane lockdown with kids at
  home.

  A big bunch of scattered amdgpu fixes, but they are all pretty small,
  minor i915 fixes, kmb, and one vmwgfx regression fixes, all pretty
  quiet for this time.

  amdgpu:
   - Fix potential out-of-bounds read when updating GPUVM mapping
   - Renoir powergating fix
   - Yellow Carp updates
   - 8K fix for navi1x
   - Beige Goby updates and new DIDs
   - Fix DMUB firmware version output
   - EDP fix
   - pmops config fix

  i915:
   - Call i915_globals_exit if pci_register_device fails
   - (follow on fix for section mismatch)
   - Correct SFC_DONE register offset

  kmb:
   - DMA fix
   - driver date/version macros

  vmwgfx:
   - Fix I/O memory access on 64-bit systems"

* tag 'drm-fixes-2021-08-06' of git://anongit.freedesktop.org/drm/drm:
  drm/amdgpu: add DID for beige goby
  drm/amdgpu/display: fix DMUB firmware version info
  drm/amd/display: workaround for hard hang on HPD on native DP
  drm/amd/display: Fix resetting DCN3.1 HW when resuming from S4
  drm/amd/display: Increase stutter watermark for dcn303
  drm/amd/display: Fix Dynamic bpp issue with 8K30 with Navi 1X
  drm/amd/display: Assume LTTPR interop for DCN31+
  drm/amdgpu: fix checking pmops when PM_SLEEP is not enabled
  drm/amd/pm: update yellow carp pmfw interface version
  drm/i915: fix i915_globals_exit() section mismatch error
  drm/i915: Call i915_globals_exit() if pci_register_device() fails
  drm/i915: Correct SFC_DONE register offset
  drm/vmwgfx: Fix a 64bit regression on svga3
  drm/amdgpu: fix the doorbell missing when in CGPG issue for renoir.
  drm/amdgpu: Fix out-of-bounds read when update mapping
  drm/kmb: Define driver date and major/minor version
  drm/kmb: Enable LCD DMA for low TVDDCV

4 years agoext4: fix potential htree corruption when growing large_dir directories
Theodore Ts'o [Wed, 4 Aug 2021 18:23:55 +0000 (14:23 -0400)]
ext4: fix potential htree corruption when growing large_dir directories

Commit b5776e7524af ("ext4: fix potential htree index checksum
corruption) removed a required restart when multiple levels of index
nodes need to be split.  Fix this to avoid directory htree corruptions
when using the large_dir feature.

Cc: stable@kernel.org # v5.11
Cc: Благодаренко Артём <artem.blagodarenko@gmail.com>
Fixes: b5776e7524af ("ext4: fix potential htree index checksum corruption)
Reported-by: Denis <denis@voxelsoft.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
4 years agotracepoint: Use rcu get state and cond sync for static call updates
Mathieu Desnoyers [Thu, 5 Aug 2021 19:29:54 +0000 (15:29 -0400)]
tracepoint: Use rcu get state and cond sync for static call updates

State transitions from 1->0->1 and N->2->1 callbacks require RCU
synchronization. Rather than performing the RCU synchronization every
time the state change occurs, which is quite slow when many tracepoints
are registered in batch, instead keep a snapshot of the RCU state on the
most recent transitions which belong to a chain, and conditionally wait
for a grace period on the last transition of the chain if one g.p. has
not elapsed since the last snapshot.

This applies to both RCU and SRCU.

This brings the performance regression caused by commit 231264d6927f
("Fix: tracepoint: static call function vs data state mismatch") back to
what it was originally.

Before this commit:

  # trace-cmd start -e all
  # time trace-cmd start -p nop

  real 0m10.593s
  user 0m0.017s
  sys 0m0.259s

After this commit:

  # trace-cmd start -e all
  # time trace-cmd start -p nop

  real 0m0.878s
  user 0m0.000s
  sys 0m0.103s

Link: https://lkml.kernel.org/r/20210805192954.30688-1-mathieu.desnoyers@efficios.com
Link: https://lore.kernel.org/io-uring/4ebea8f0-58c9-e571-fd30-0ce4f6f09c70@samba.org/
Cc: stable@vger.kernel.org
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: "Paul E. McKenney" <paulmck@kernel.org>
Cc: Stefan Metzmacher <metze@samba.org>
Fixes: 231264d6927f ("Fix: tracepoint: static call function vs data state mismatch")
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Reviewed-by: Paul E. McKenney <paulmck@kernel.org>
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
4 years agoio-wq: fix lack of acct->nr_workers < acct->max_workers judgement io_uring-5.14-2021-08-07
Hao Xu [Thu, 5 Aug 2021 10:05:38 +0000 (18:05 +0800)]
io-wq: fix lack of acct->nr_workers < acct->max_workers judgement

There should be this judgement before we create an io-worker

Fixes: 685fe7feedb9 ("io-wq: eliminate the need for a manager thread")
Signed-off-by: Hao Xu <haoxu@linux.alibaba.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
4 years agoio-wq: fix no lock protection of acct->nr_worker
Hao Xu [Thu, 5 Aug 2021 10:05:37 +0000 (18:05 +0800)]
io-wq: fix no lock protection of acct->nr_worker

There is an acct->nr_worker visit without lock protection. Think about
the case: two callers call io_wqe_wake_worker(), one is the original
context and the other one is an io-worker(by calling
io_wqe_enqueue(wqe, linked)), on two cpus paralelly, this may cause
nr_worker to be larger than max_worker.
Let's fix it by adding lock for it, and let's do nr_workers++ before
create_io_worker. There may be a edge cause that the first caller fails
to create an io-worker, but the second caller doesn't know it and then
quit creating io-worker as well:

say nr_worker = max_worker - 1
        cpu 0                        cpu 1
   io_wqe_wake_worker()          io_wqe_wake_worker()
      nr_worker < max_worker
      nr_worker++
      create_io_worker()         nr_worker == max_worker
         failed                  return
      return

But the chance of this case is very slim.

Fixes: 685fe7feedb9 ("io-wq: eliminate the need for a manager thread")
Signed-off-by: Hao Xu <haoxu@linux.alibaba.com>
[axboe: fix unconditional create_io_worker() call]
Signed-off-by: Jens Axboe <axboe@kernel.dk>
4 years agoperf/x86/intel: Apply mid ACK for small core
Kan Liang [Tue, 3 Aug 2021 13:25:28 +0000 (06:25 -0700)]
perf/x86/intel: Apply mid ACK for small core

A warning as below may be occasionally triggered in an ADL machine when
these conditions occur:

 - Two perf record commands run one by one. Both record a PEBS event.
 - Both runs on small cores.
 - They have different adaptive PEBS configuration (PEBS_DATA_CFG).

  [ ] WARNING: CPU: 4 PID: 9874 at arch/x86/events/intel/ds.c:1743 setup_pebs_adaptive_sample_data+0x55e/0x5b0
  [ ] RIP: 0010:setup_pebs_adaptive_sample_data+0x55e/0x5b0
  [ ] Call Trace:
  [ ]  <NMI>
  [ ]  intel_pmu_drain_pebs_icl+0x48b/0x810
  [ ]  perf_event_nmi_handler+0x41/0x80
  [ ]  </NMI>
  [ ]  __perf_event_task_sched_in+0x2c2/0x3a0

Different from the big core, the small core requires the ACK right
before re-enabling counters in the NMI handler, otherwise a stale PEBS
record may be dumped into the later NMI handler, which trigger the
warning.

Add a new mid_ack flag to track the case. Add all PMI handler bits in
the struct x86_hybrid_pmu to track the bits for different types of
PMUs.  Apply mid ACK for the small cores on an Alder Lake machine.

The existing hybrid() macro has a compile error when taking address of
a bit-field variable. Add a new macro hybrid_bit() to get the
bit-field value of a given PMU.

Fixes: f83d2f91d259 ("perf/x86/intel: Add Alder Lake Hybrid support")
Reported-by: Ammy Yi <ammy.yi@intel.com>
Signed-off-by: Kan Liang <kan.liang@linux.intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Andi Kleen <ak@linux.intel.com>
Tested-by: Ammy Yi <ammy.yi@intel.com>
Link: https://lkml.kernel.org/r/1627997128-57891-1-git-send-email-kan.liang@linux.intel.com
4 years agoRDMA/iw_cxgb4: Fix refcount underflow while destroying cqs.
Dakshaja Uppalapati [Thu, 5 Aug 2021 12:43:32 +0000 (18:13 +0530)]
RDMA/iw_cxgb4: Fix refcount underflow while destroying cqs.

Previous atomic increment/decrement logic expects the atomic count to be
'0' after the final decrement.

Replacing atomic count with refcount does not allow that, as
refcount_dec() considers count of 1 as underflow and triggers a kernel
splat.

Fix the current refcount logic by using the usual pattern of decrementing
the refcount and test if it is '0' on the final deref in
c4iw_destroy_cq(). Use wait_for_completion() instead of wait_event().

Fixes: 7183451f846d ("RDMA/cxgb4: Use refcount_t instead of atomic_t for reference counting")
Link: https://lore.kernel.org/r/1628167412-12114-1-git-send-email-dakshaja@chelsio.com
Signed-off-by: Dakshaja Uppalapati <dakshaja@chelsio.com>
Reviewed-by: Potnuri Bharat Teja <bharat@chelsio.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
4 years agoMerge tag 'amd-drm-fixes-5.14-2021-08-05' of https://gitlab.freedesktop.org/agd5f...
Dave Airlie [Fri, 6 Aug 2021 01:22:08 +0000 (11:22 +1000)]
Merge tag 'amd-drm-fixes-5.14-2021-08-05' of https://gitlab.freedesktop.org/agd5f/linux into drm-fixes

amd-drm-fixes-5.14-2021-08-05:

amdgpu:
- Fix potential out-of-bounds read when updating GPUVM mapping
- Renoir powergating fix
- Yellow Carp updates
- 8K fix for navi1x
- Beige Goby updates and new DIDs
- Fix DMUB firmware version output
- EDP fix
- pmops config fix

Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Alex Deucher <alexander.deucher@amd.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20210806011635.1055841-1-alexander.deucher@amd.com
4 years agodrm/amdgpu: add DID for beige goby
Chengming Gui [Wed, 24 Feb 2021 03:48:23 +0000 (11:48 +0800)]
drm/amdgpu: add DID for beige goby

Add device ids.

Signed-off-by: Chengming Gui <Jack.Gui@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amdgpu/display: fix DMUB firmware version info
Shirish S [Tue, 3 Aug 2021 08:33:44 +0000 (14:03 +0530)]
drm/amdgpu/display: fix DMUB firmware version info

DMUB firmware info is printed before it gets initialized.
Correct this order to ensure true value is conveyed.

Signed-off-by: Shirish S <shirish.s@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org
4 years agodrm/amd/display: workaround for hard hang on HPD on native DP
Qingqing Zhuo [Thu, 22 Jul 2021 18:48:54 +0000 (14:48 -0400)]
drm/amd/display: workaround for hard hang on HPD on native DP

[Why]
HPD disable and enable sequences are not mutually exclusive
on Linux. For HPDs that spans over 1s (i.e. HPD low = 1s),
part of the disable sequence (specifically, a request to SMU
to lower refclk) could come right before the call to PHY
enable, causing DMUB to access an unresponsive PHY
and thus a hard hang on the system.

[How]
Disable 48mhz refclk off on native DP.

Reviewed-by: Hersen Wu <hersenxs.wu@amd.com>
Acked-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Signed-off-by: Qingqing Zhuo <qingqing.zhuo@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amd/display: Fix resetting DCN3.1 HW when resuming from S4
Jude Shih [Tue, 6 Jul 2021 10:04:11 +0000 (18:04 +0800)]
drm/amd/display: Fix resetting DCN3.1 HW when resuming from S4

[Why] On S4 resume we also need to fix detection of when to reload DMCUB
firmware because we're currently using the VBIOS version which isn't
compatible with the driver version.

[How] Update the hardware init check for DCN31 since it's the ASIC that
has this issue.

Reviewed-by: Nicholas Kazlauskas <Nicholas.Kazlauskas@amd.com>
Acked-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Signed-off-by: Jude Shih <jude.shih@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amd/display: Increase stutter watermark for dcn303
Bing Guo [Tue, 20 Jul 2021 19:13:38 +0000 (15:13 -0400)]
drm/amd/display: Increase stutter watermark for dcn303

[Why&How]
Hardware team suggested to use SRExitTime= 35.5us as w/a to prevent
underflow in certain modes.

Reviewed-by: Martin Leung <martin.leung@amd.com>
Acked-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Signed-off-by: Bing Guo <bing.guo@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amd/display: Fix Dynamic bpp issue with 8K30 with Navi 1X
Bing Guo [Mon, 19 Jul 2021 22:24:06 +0000 (18:24 -0400)]
drm/amd/display: Fix Dynamic bpp issue with 8K30 with Navi 1X

Why:
In DCN2x, HW doesn't automatically divide MASTER_UPDATE_LOCK_DB_X
by the number of pipes ODM Combined.

How:
Set MASTER_UPDATE_LOCK_DB_X to the value that is adjusted by the
number of pipes ODM Combined.

Reviewed-by: Martin Leung <martin.leung@amd.com>
Acked-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Signed-off-by: Bing Guo <bing.guo@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amd/display: Assume LTTPR interop for DCN31+
Wesley Chalmers [Mon, 19 Jul 2021 17:13:33 +0000 (13:13 -0400)]
drm/amd/display: Assume LTTPR interop for DCN31+

[WHY]
For DCN31 onward, LTTPR is to be enabled and set to Transparent by
VBIOS.  Driver is to assume that VBIOS has done this without needing to
check the VBIOS interop bit.

[HOW]
Add LTTPR enable and interop VBIOS bits into dc->caps, and force-set the
interop bit to true for DCN31+.

Reviewed-by: Jun Lei <Jun.Lei@amd.com>
Acked-by: Aurabindo Pillai <aurabindo.pillai@amd.com>
Signed-off-by: Wesley Chalmers <Wesley.Chalmers@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agodrm/amdgpu: fix checking pmops when PM_SLEEP is not enabled
Randy Dunlap [Fri, 30 Jul 2021 03:03:47 +0000 (20:03 -0700)]
drm/amdgpu: fix checking pmops when PM_SLEEP is not enabled

'pm_suspend_target_state' is only available when CONFIG_PM_SLEEP
is set/enabled. OTOH, when both SUSPEND and HIBERNATION are not set,
PM_SLEEP is not set, so this variable cannot be used.

../drivers/gpu/drm/amd/amdgpu/amdgpu_acpi.c: In function ‘amdgpu_acpi_is_s0ix_active’:
../drivers/gpu/drm/amd/amdgpu/amdgpu_acpi.c:1046:11: error: ‘pm_suspend_target_state’ undeclared (first use in this function); did you mean ‘__KSYM_pm_suspend_target_state’?
    return pm_suspend_target_state == PM_SUSPEND_TO_IDLE;
           ^~~~~~~~~~~~~~~~~~~~~~~
           __KSYM_pm_suspend_target_state

Also use shorter IS_ENABLED(CONFIG_foo) notation for checking the
2 config symbols.

Fixes: 91e273712ab8dd ("drm/amdgpu: Check pmops for desired suspend state")
Signed-off-by: Randy Dunlap <rdunlap@infradead.org>
Cc: Alex Deucher <alexander.deucher@amd.com>
Cc: Christian König <christian.koenig@amd.com>
Cc: "Pan, Xinhui" <Xinhui.Pan@amd.com>
Cc: amd-gfx@lists.freedesktop.org
Cc: dri-devel@lists.freedesktop.org
Cc: linux-next@vger.kernel.org
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org
4 years agodrm/amd/pm: update yellow carp pmfw interface version
Xiaomeng Hou [Thu, 29 Jul 2021 07:44:39 +0000 (15:44 +0800)]
drm/amd/pm: update yellow carp pmfw interface version

Correct yellow carp driver-PMFW interface version to v4.

Signed-off-by: Xiaomeng Hou <Xiaomeng.Hou@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
4 years agotracepoint: Fix static call function vs data state mismatch
Mathieu Desnoyers [Thu, 5 Aug 2021 13:27:16 +0000 (09:27 -0400)]
tracepoint: Fix static call function vs data state mismatch

On a 1->0->1 callbacks transition, there is an issue with the new
callback using the old callback's data.

Considering __DO_TRACE_CALL:

        do {                                                            \
                struct tracepoint_func *it_func_ptr;                    \
                void *__data;                                           \
                it_func_ptr =                                           \
                        rcu_dereference_raw((&__tracepoint_##name)->funcs); \
                if (it_func_ptr) {                                      \
                        __data = (it_func_ptr)->data;                   \

----> [ delayed here on one CPU (e.g. vcpu preempted by the host) ]

                        static_call(tp_func_##name)(__data, args);      \
                }                                                       \
        } while (0)

It has loaded the tp->funcs of the old callback, so it will try to use the old
data. This can be fixed by adding a RCU sync anywhere in the 1->0->1
transition chain.

On a N->2->1 transition, we need an rcu-sync because you may have a
sequence of 3->2->1 (or 1->2->1) where the element 0 data is unchanged
between 2->1, but was changed from 3->2 (or from 1->2), which may be
observed by the static call. This can be fixed by adding an
unconditional RCU sync in transition 2->1.

Note, this fixes a correctness issue at the cost of adding a tremendous
performance regression to the disabling of tracepoints.

Before this commit:

  # trace-cmd start -e all
  # time trace-cmd start -p nop

  real 0m0.778s
  user 0m0.000s
  sys 0m0.061s

After this commit:

  # trace-cmd start -e all
  # time trace-cmd start -p nop

  real 0m10.593s
  user 0m0.017s
  sys 0m0.259s

A follow up fix will introduce a more lightweight scheme based on RCU
get_state and cond_sync, that will return the performance back to what it
was. As both this change and the lightweight versions are complex on their
own, for bisecting any issues that this may cause, they are kept as two
separate changes.

Link: https://lkml.kernel.org/r/20210805132717.23813-3-mathieu.desnoyers@efficios.com
Link: https://lore.kernel.org/io-uring/4ebea8f0-58c9-e571-fd30-0ce4f6f09c70@samba.org/
Cc: stable@vger.kernel.org
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: "Paul E. McKenney" <paulmck@kernel.org>
Cc: Stefan Metzmacher <metze@samba.org>
Fixes: d25e37d89dd2 ("tracepoint: Optimize using static_call()")
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
4 years agotracepoint: static call: Compare data on transition from 2->1 callees
Mathieu Desnoyers [Thu, 5 Aug 2021 13:27:15 +0000 (09:27 -0400)]
tracepoint: static call: Compare data on transition from 2->1 callees

On transition from 2->1 callees, we should be comparing .data rather
than .func, because the same callback can be registered twice with
different data, and what we care about here is that the data of array
element 0 is unchanged to skip rcu sync.

Link: https://lkml.kernel.org/r/20210805132717.23813-2-mathieu.desnoyers@efficios.com
Link: https://lore.kernel.org/io-uring/4ebea8f0-58c9-e571-fd30-0ce4f6f09c70@samba.org/
Cc: stable@vger.kernel.org
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: "Paul E. McKenney" <paulmck@kernel.org>
Cc: Stefan Metzmacher <metze@samba.org>
Fixes: 547305a64632 ("tracepoint: Fix out of sync data passing by static caller")
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
4 years agoMerge tag 'net-5.14-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net
Linus Torvalds [Thu, 5 Aug 2021 19:26:00 +0000 (12:26 -0700)]
Merge tag 'net-5.14-rc5' of git://git./linux/kernel/git/netdev/net

Pull networking fixes from Jakub Kicinski:
 "Including fixes from ipsec.

  Current release - regressions:

   - sched: taprio: fix init procedure to avoid inf loop when dumping

   - sctp: move the active_key update after sh_keys is added

  Current release - new code bugs:

   - sparx5: fix build with old GCC & bitmask on 32-bit targets

  Previous releases - regressions:

   - xfrm: redo the PREEMPT_RT RCU vs hash_resize_mutex deadlock fix

   - xfrm: fixes for the compat netlink attribute translator

   - phy: micrel: Fix detection of ksz87xx switch

  Previous releases - always broken:

   - gro: set inner transport header offset in tcp/udp GRO hook to avoid
     crashes when such packets reach GSO

   - vsock: handle VIRTIO_VSOCK_OP_CREDIT_REQUEST, as required by spec

   - dsa: sja1105: fix static FDB entries on SJA1105P/Q/R/S and SJA1110

   - bridge: validate the NUD_PERMANENT bit when adding an extern_learn
     FDB entry

   - usb: lan78xx: don't modify phy_device state concurrently

   - usb: pegasus: check for errors of IO routines"

* tag 'net-5.14-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net: (48 commits)
  net: vxge: fix use-after-free in vxge_device_unregister
  net: fec: fix use-after-free in fec_drv_remove
  net: pegasus: fix uninit-value in get_interrupt_interval
  net: ethernet: ti: am65-cpsw: fix crash in am65_cpsw_port_offload_fwd_mark_update()
  bnx2x: fix an error code in bnx2x_nic_load()
  net: wwan: iosm: fix recursive lock acquire in unregister
  net: wwan: iosm: correct data protocol mask bit
  net: wwan: iosm: endianness type correction
  net: wwan: iosm: fix lkp buildbot warning
  net: usb: lan78xx: don't modify phy_device state concurrently
  docs: networking: netdevsim rules
  net: usb: pegasus: Remove the changelog and DRIVER_VERSION.
  net: usb: pegasus: Check the return value of get_geristers() and friends;
  net/prestera: Fix devlink groups leakage in error flow
  net: sched: fix lockdep_set_class() typo error for sch->seqlock
  net: dsa: qca: ar9331: reorder MDIO write sequence
  VSOCK: handle VIRTIO_VSOCK_OP_CREDIT_REQUEST
  mptcp: drop unused rcu member in mptcp_pm_addr_entry
  net: ipv6: fix returned variable type in ip6_skb_dst_mtu
  nfp: update ethtool reporting of pauseframe control
  ...

4 years agospi: cadence-quadspi: Fix check condition for DTR ops
Apurva Nandan [Fri, 16 Jul 2021 23:25:03 +0000 (23:25 +0000)]
spi: cadence-quadspi: Fix check condition for DTR ops

buswidth and dtr fields in spi_mem_op are only valid when the
corresponding spi_mem_op phase has a non-zero length. For example,
SPI NAND core doesn't set buswidth when using SPI_MEM_OP_NO_ADDR
phase.

Fix the dtr checks in set_protocol() and suppports_mem_op() to
ignore empty spi_mem_op phases, as checking for dtr field in
empty phase will result in false negatives.

Signed-off-by: Apurva Nandan <a-nandan@ti.com>
Link: https://lore.kernel.org/r/20210716232504.182-3-a-nandan@ti.com
Signed-off-by: Mark Brown <broonie@kernel.org>
4 years agoBluetooth: defer cleanup of resources in hci_unregister_dev()
Tetsuo Handa [Wed, 4 Aug 2021 10:26:56 +0000 (19:26 +0900)]
Bluetooth: defer cleanup of resources in hci_unregister_dev()

syzbot is hitting might_sleep() warning at hci_sock_dev_event() due to
calling lock_sock() with rw spinlock held [1].

It seems that history of this locking problem is a trial and error.

Commit b40df5743ee8 ("[PATCH] bluetooth: fix socket locking in
hci_sock_dev_event()") in 2.6.21-rc4 changed bh_lock_sock() to
lock_sock() as an attempt to fix lockdep warning.

Then, commit 4ce61d1c7a8e ("[BLUETOOTH]: Fix locking in
hci_sock_dev_event().") in 2.6.22-rc2 changed lock_sock() to
local_bh_disable() + bh_lock_sock_nested() as an attempt to fix the
sleep in atomic context warning.

Then, commit 4b5dd696f81b ("Bluetooth: Remove local_bh_disable() from
hci_sock.c") in 3.3-rc1 removed local_bh_disable().

Then, commit e305509e678b ("Bluetooth: use correct lock to prevent UAF
of hdev object") in 5.13-rc5 again changed bh_lock_sock_nested() to
lock_sock() as an attempt to fix CVE-2021-3573.

This difficulty comes from current implementation that
hci_sock_dev_event(HCI_DEV_UNREG) is responsible for dropping all
references from sockets because hci_unregister_dev() immediately
reclaims resources as soon as returning from
hci_sock_dev_event(HCI_DEV_UNREG).

But the history suggests that hci_sock_dev_event(HCI_DEV_UNREG) was not
doing what it should do.

Therefore, instead of trying to detach sockets from device, let's accept
not detaching sockets from device at hci_sock_dev_event(HCI_DEV_UNREG),
by moving actual cleanup of resources from hci_unregister_dev() to
hci_cleanup_dev() which is called by bt_host_release() when all
references to this unregistered device (which is a kobject) are gone.

Since hci_sock_dev_event(HCI_DEV_UNREG) no longer resets
hci_pi(sk)->hdev, we need to check whether this device was unregistered
and return an error based on HCI_UNREGISTER flag.  There might be subtle
behavioral difference in "monitor the hdev" functionality; please report
if you found something went wrong due to this patch.

Link: https://syzkaller.appspot.com/bug?extid=a5df189917e79d5e59c9
Reported-by: syzbot <syzbot+a5df189917e79d5e59c9@syzkaller.appspotmail.com>
Suggested-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Fixes: e305509e678b ("Bluetooth: use correct lock to prevent UAF of hdev object")
Acked-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
4 years agoMerge tag 'selinux-pr-20210805' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Thu, 5 Aug 2021 19:06:31 +0000 (12:06 -0700)]
Merge tag 'selinux-pr-20210805' of git://git./linux/kernel/git/pcmoore/selinux

Pull selinux fix from Paul Moore:
 "One small SELinux fix for a problem where an error code was not being
  propagated back up to userspace when a bogus SELinux policy is loaded
  into the kernel"

* tag 'selinux-pr-20210805' of git://git.kernel.org/pub/scm/linux/kernel/git/pcmoore/selinux:
  selinux: correct the return value when loads initial sids

4 years agoMerge branch 'for-v5.14' of git://git.kernel.org/pub/scm/linux/kernel/git/ebiederm...
Linus Torvalds [Thu, 5 Aug 2021 19:00:00 +0000 (12:00 -0700)]
Merge branch 'for-v5.14' of git://git./linux/kernel/git/ebiederm/user-namespace

Pull ucounts fix from Eric Biederman:
 "Fix a subtle locking versus reference counting bug in the ucount
  changes, found by syzbot"

* 'for-v5.14' of git://git.kernel.org/pub/scm/linux/kernel/git/ebiederm/user-namespace:
  ucounts: Fix race condition between alloc_ucounts and put_ucounts

4 years agoMerge tag 'trace-v5.14-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/rostedt...
Linus Torvalds [Thu, 5 Aug 2021 18:53:34 +0000 (11:53 -0700)]
Merge tag 'trace-v5.14-rc4' of git://git./linux/kernel/git/rostedt/linux-trace

Pull tracing fixes from Steven Rostedt:
 "Various tracing fixes:

   - Fix NULL pointer dereference caused by an error path

   - Give histogram calculation fields a size, otherwise it breaks
     synthetic creation based on them.

   - Reject strings being used for number calculations.

   - Fix recordmcount.pl warning on llvm building RISC-V allmodconfig

   - Fix the draw_functrace.py script to handle the new trace output

   - Fix warning of smp_processor_id() in preemptible code"

* tag 'trace-v5.14-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/linux-trace:
  tracing: Quiet smp_processor_id() use in preemptable warning in hwlat
  scripts/tracing: fix the bug that can't parse raw_trace_func
  scripts/recordmcount.pl: Remove check_objcopy() and $can_use_local
  tracing: Reject string operand in the histogram expression
  tracing / histogram: Give calculation hist_fields a size
  tracing: Fix NULL pointer dereference in start_creating

4 years agoMerge tag 's390-5.14-4' of git://git.kernel.org/pub/scm/linux/kernel/git/s390/linux
Linus Torvalds [Thu, 5 Aug 2021 18:46:24 +0000 (11:46 -0700)]
Merge tag 's390-5.14-4' of git://git./linux/kernel/git/s390/linux

Pull s390 fixes from Heiko Carstens:

 - fix zstd build for -march=z900 (undefined reference to __clzdi2)

 - add missing .got.plts to vdso linker scripts to fix kpatch build
   errors

 - update defconfigs

* tag 's390-5.14-4' of git://git.kernel.org/pub/scm/linux/kernel/git/s390/linux:
  s390: update defconfigs
  s390/boot: fix zstd build for -march=z900
  s390/vdso: add .got.plt in vdso linker script

4 years agoMerge tag 'for-linus' of git://git.kernel.org/pub/scm/virt/kvm/kvm
Linus Torvalds [Thu, 5 Aug 2021 18:23:09 +0000 (11:23 -0700)]
Merge tag 'for-linus' of git://git./virt/kvm/kvm

Pull kvm fixes from Paolo Bonzini:
 "Mostly bugfixes; plus, support for XMM arguments to Hyper-V hypercalls
  now obeys KVM_CAP_HYPERV_ENFORCE_CPUID.

  Both the XMM arguments feature and KVM_CAP_HYPERV_ENFORCE_CPUID are
  new in 5.14, and each did not know of the other"

* tag 'for-linus' of git://git.kernel.org/pub/scm/virt/kvm/kvm:
  KVM: x86/mmu: Fix per-cpu counter corruption on 32-bit builds
  KVM: selftests: fix hyperv_clock test
  KVM: SVM: improve the code readability for ASID management
  KVM: SVM: Fix off-by-one indexing when nullifying last used SEV VMCB
  KVM: Do not leak memory for duplicate debugfs directories
  KVM: selftests: Test access to XMM fast hypercalls
  KVM: x86: hyper-v: Check if guest is allowed to use XMM registers for hypercall input
  KVM: x86: Introduce trace_kvm_hv_hypercall_done()
  KVM: x86: hyper-v: Check access to hypercall before reading XMM registers
  KVM: x86: accept userspace interrupt only if no event is injected

4 years agoMerge branch 'pcmcia-next' of git://git.kernel.org/pub/scm/linux/kernel/git/brodo...
Linus Torvalds [Thu, 5 Aug 2021 18:16:02 +0000 (11:16 -0700)]
Merge branch 'pcmcia-next' of git://git./linux/kernel/git/brodo/linux

Pull pcmcia fix from Dominik Brodowski:
 "Zheyu Ma found and fixed a null pointer dereference bug in the device
  driver for the i82092 card reader"

* 'pcmcia-next' of git://git.kernel.org/pub/scm/linux/kernel/git/brodo/linux:
  pcmcia: i82092: fix a null pointer dereference bug

4 years agoblock/partitions/ldm.c: Fix a kernel-doc warning
Bart Van Assche [Thu, 5 Aug 2021 17:34:47 +0000 (10:34 -0700)]
block/partitions/ldm.c: Fix a kernel-doc warning

Fix the following kernel-doc warning that appears when building with W=1:

block/partitions/ldm.c:31: warning: expecting prototype for ldm().
Prototype was for ldm_debug() instead

Signed-off-by: Bart Van Assche <bvanassche@acm.org>
Link: https://lore.kernel.org/r/20210805173447.3249906-1-bvanassche@acm.org
Signed-off-by: Jens Axboe <axboe@kernel.dk>
4 years agopipe: increase minimum default pipe size to 2 pages
Alex Xu (Hello71) [Thu, 5 Aug 2021 14:40:47 +0000 (10:40 -0400)]
pipe: increase minimum default pipe size to 2 pages

This program always prints 4096 and hangs before the patch, and always
prints 8192 and exits successfully after:

  int main()
  {
      int pipefd[2];
      for (int i = 0; i < 1025; i++)
          if (pipe(pipefd) == -1)
              return 1;
      size_t bufsz = fcntl(pipefd[1], F_GETPIPE_SZ);
      printf("%zd\n", bufsz);
      char *buf = calloc(bufsz, 1);
      write(pipefd[1], buf, bufsz);
      read(pipefd[0], buf, bufsz-1);
      write(pipefd[1], buf, 1);
  }

Note that you may need to increase your RLIMIT_NOFILE before running the
program.

Fixes: 759c01142a ("pipe: limit the per-user amount of pages allocated in pipes")
Cc: <stable@vger.kernel.org>
Link: https://lore.kernel.org/lkml/1628086770.5rn8p04n6j.none@localhost/
Link: https://lore.kernel.org/lkml/1628127094.lxxn016tj7.none@localhost/
Signed-off-by: Alex Xu (Hello71) <alex_y_xu@yahoo.ca>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
4 years agoblk-iolatency: error out if blk_get_queue() failed in iolatency_set_limit()
Yu Kuai [Thu, 5 Aug 2021 12:46:45 +0000 (20:46 +0800)]
blk-iolatency: error out if blk_get_queue() failed in iolatency_set_limit()

If queue is dying while iolatency_set_limit() is in progress,
blk_get_queue() won't increment the refcount of the queue. However,
blk_put_queue() will still decrement the refcount later, which will
cause the refcout to be unbalanced.

Thus error out in such case to fix the problem.

Fixes: 8c772a9bfc7c ("blk-iolatency: fix IO hang due to negative inflight counter")
Signed-off-by: Yu Kuai <yukuai3@huawei.com>
Acked-by: Tejun Heo <tj@kernel.org>
Link: https://lore.kernel.org/r/20210805124645.543797-1-yukuai3@huawei.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
4 years agoMerge branch 'net-fix-use-after-free-bugs'
Jakub Kicinski [Thu, 5 Aug 2021 14:29:55 +0000 (07:29 -0700)]
Merge branch 'net-fix-use-after-free-bugs'

Pavel Skripkin says:

====================
net: fix use-after-free bugs

I've added new checker to smatch yesterday. It warns about using
netdev_priv() pointer after free_{netdev,candev}() call. I hope, it will
get into next smatch release.

Some of the reported bugs are fixed and upstreamed already, but Dan ran new
smatch with allmodconfig and found 2 more. Big thanks to Dan for doing it,
because I totally forgot to do it.
====================

Link: https://lore.kernel.org/r/cover.1628091954.git.paskripkin@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
4 years agonet: vxge: fix use-after-free in vxge_device_unregister
Pavel Skripkin [Wed, 4 Aug 2021 15:52:20 +0000 (18:52 +0300)]
net: vxge: fix use-after-free in vxge_device_unregister

Smatch says:
drivers/net/ethernet/neterion/vxge/vxge-main.c:3518 vxge_device_unregister() error: Using vdev after free_{netdev,candev}(dev);
drivers/net/ethernet/neterion/vxge/vxge-main.c:3518 vxge_device_unregister() error: Using vdev after free_{netdev,candev}(dev);
drivers/net/ethernet/neterion/vxge/vxge-main.c:3520 vxge_device_unregister() error: Using vdev after free_{netdev,candev}(dev);
drivers/net/ethernet/neterion/vxge/vxge-main.c:3520 vxge_device_unregister() error: Using vdev after free_{netdev,candev}(dev);

Since vdev pointer is netdev private data accessing it after free_netdev()
call can cause use-after-free bug. Fix it by moving free_netdev() call at
the end of the function

Fixes: 6cca200362b4 ("vxge: cleanup probe error paths")
Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Pavel Skripkin <paskripkin@gmail.com>
Reviewed-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
4 years agonet: fec: fix use-after-free in fec_drv_remove
Pavel Skripkin [Wed, 4 Aug 2021 15:51:51 +0000 (18:51 +0300)]
net: fec: fix use-after-free in fec_drv_remove

Smatch says:
drivers/net/ethernet/freescale/fec_main.c:3994 fec_drv_remove() error: Using fep after free_{netdev,candev}(ndev);
drivers/net/ethernet/freescale/fec_main.c:3995 fec_drv_remove() error: Using fep after free_{netdev,candev}(ndev);

Since fep pointer is netdev private data, accessing it after free_netdev()
call can cause use-after-free bug. Fix it by moving free_netdev() call at
the end of the function

Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Fixes: a31eda65ba21 ("net: fec: fix clock count mis-match")
Signed-off-by: Pavel Skripkin <paskripkin@gmail.com>
Reviewed-by: Joakim Zhang <qiangqing.zhang@nxp.com>
Reviewed-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>