Commit 90d9e367d4dc for kernel

commit 90d9e367d4dc3e4bec20c2919c1beab69a4d7431
Merge: a91337fb5e8e cf6542493e35
Author: Linus Torvalds <torvalds@linux-foundation.org>
Date:   Fri Nov 6 12:30:19 2026 -0800

    Merge tag 'io_uring-7.4-20261106' of git://git.kernel.org/pub/scm/linux/kernel/git/axboe/linux

    Pull io_uring fixes from Jens Axboe:

     - Only wait for in-flight requests on ring release if the closing task
       actually uses the ring.

     - SQPOLL fixes: do the return-to-userspace work in the SQPOLL thread,
       as memory.high reclaim and blkcg throttling otherwise only ran at
       thread exit, and don't set the stop bit on thread exit, which tripped
       a WARN_ON_ONCE() in the ring teardown.

     - Don't poll a writer for a read side hangup or for urgent data, as a
       half-closed peer or unread OOB data kept the async poll of a send
       ready on every re-arm.

     - Cancel all matching poll requests in one pass for CANCEL_ALL rather
       than restarting the hash walk per request.

     - Hold uring_lock across the put_user() for the index-alloc file
       install.

     - Don't arm an hrtimer and schedule for a wait whose timeout has
       already expired.

     - Add a feature flag for the dma-buf registered buffer support.

     - selftests 32-bit cast fix

    * tag 'io_uring-7.4-20261106' of git://git.kernel.org/pub/scm/linux/kernel/git/axboe/linux:
      io_uring: only wait for in-flight requests if the closing task uses the ring
      io_uring: add feature flag for dma-buf support
      io_uring/wait: don't sleep on an already expired timeout
      io_uring/rsrc: hold uring_lock across index-alloc put_user
      selftests/io_uring: cast optval through uintptr_t in mini_liburing.h
      io_uring/poll: don't poll a writer for urgent data
      io_uring/poll: don't poll a writer for a read side hangup
      io_uring/poll: cancel all matching requests in one pass
      io_uring/cancel: count all requests that a CANCEL_ALL cancels
      io_uring/sqpoll: do the return-to-userspace work in the SQPOLL thread
      io_uring/sqpoll: don't set IO_SQ_THREAD_SHOULD_STOP on thread exit