io_uring: don't rely on weak ->files references
authorJens Axboe <axboe@kernel.dk>
Sun, 13 Sep 2020 19:09:39 +0000 (13:09 -0600)
committerJens Axboe <axboe@kernel.dk>
Mon, 12 Oct 2020 20:59:33 +0000 (14:59 -0600)
commit2c82c90c7277f1fc1eccadf4098c335aa3178ad6
tree8a5d945a0cbbee78a7fe840e8f5f98ac40e10c2b
parent171129e7fdb4833e48c3c98d67d4247a01d769ad
io_uring: don't rely on weak ->files references

commit 0f2122045b946241a9e549c2a76cea54fa58a7ff upstream.

Grab actual references to the files_struct. To avoid circular references
issues due to this, we add a per-task note that keeps track of what
io_uring contexts a task has used. When the tasks execs or exits its
assigned files, we cancel requests based on this tracking.

With that, we can grab proper references to the files table, and no
longer need to rely on stashing away ring_fd and ring_file to check
if the ring_fd may have been closed.

Cc: stable@vger.kernel.org # v5.5+
Reviewed-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
fs/exec.c
fs/file.c
fs/io_uring.c
include/linux/io_uring.h [new file with mode: 0644]
include/linux/sched.h
init/init_task.c
kernel/fork.c