Hi Linus,
A few fixes in here, all based on reports and test cases from folks
using it. Most of it is stable material as well.
- Hashed work cancelation fix (Pavel)
- poll wakeup signalfd fix
- memlock accounting fix
- nonblocking poll retry fix
- Ensure we never return -ERESTARTSYS for reads
- Ensure offset == -1 is consistent with preadv2() as documented
- IOPOLL -EAGAIN handling fixes
- Remove useless task_work bounce for block based -EAGAIN retry
Please pull!
The following changes since commit 867a23eab52847d41a0a6eae41a64d76de7782a8:
io_uring: kill extra iovec=NULL in import_iovec() (2020-08-20 05:36:19 -0600)
are available in the Git repository at:
git://git.kernel.dk/linux-block.git tags/io_uring-5.9-2020-08-28
for you to fetch changes up to fdee946d0925f971f167d2606984426763355e4f:
io_uring: don't bounce block based -EAGAIN retry off task_work (2020-08-27 16:48:34 -0600)
----------------------------------------------------------------
io_uring-5.9-2020-08-28
----------------------------------------------------------------
Jens Axboe (9):
io_uring: don't recurse on tsk->sighand->siglock with signalfd
io_uring: revert consumed iov_iter bytes on error
io_uring: fix imbalanced sqo_mm accounting
io_uring: don't use poll handler if file can't be nonblocking read/written
io_uring: ensure read requests go through -ERESTART* transformation
io_uring: make offset == -1 consistent with preadv2/pwritev2
io_uring: clear req->result on IOPOLL re-issue
io_uring: fix IOPOLL -EAGAIN retries
io_uring: don't bounce block based -EAGAIN retry off task_work
Pavel Begunkov (1):
io-wq: fix hang after cancelling pending hashed work
fs/io-wq.c | 21 +++++++++++--
fs/io_uring.c | 99 ++++++++++++++++++++++++++++++++++-------------------------
2 files changed, 76 insertions(+), 44 deletions(-)
--
Jens Axboe
