Hi Linus,
A mix of fixes and a few stragglers. In detail:
- Revert the bogus __read_mostly that we discussed for the initial pull
request.
- Fix a merge window regression with fixed file registration error path
handling.
- Fix io-wq numa node affinities.
- Series abstracting out an io_identity struct, making it both easier to
see what the personality items are, and also easier to to adopt more.
Use this to cover audit logging.
- Fix for read-ahead disabled block condition in async buffered reads,
and using single page read-ahead to unify what
generic_file_buffer_read() path is used.
- Series for REQ_F_COMP_LOCKED fix and removal of it (Pavel)
- Poll fix (Pavel)
Please pull!
The following changes since commit 071a0578b0ce0b0e543d1e38ee6926b9cc21c198:
Merge tag 'ovl-update-5.10' of git://git.kernel.org/pub/scm/linux/kernel/git/mszeredi/vfs (2020-10-16 15:29:46 -0700)
are available in the Git repository at:
git://git.kernel.dk/linux-block.git tags/io_uring-5.10-2020-10-20
for you to fetch changes up to 9ba0d0c81284f4ec0b24529bdba2fc68b9d6a09a:
io_uring: use blk_queue_nowait() to check if NOWAIT supported (2020-10-19 07:32:36 -0600)
----------------------------------------------------------------
io_uring-5.10-2020-10-20
----------------------------------------------------------------
Colin Ian King (1):
io_uring: Fix sizeof() mismatch
Jeffle Xu (1):
io_uring: use blk_queue_nowait() to check if NOWAIT supported
Jens Axboe (13):
Revert "io_uring: mark io_uring_fops/io_op_defs as __read_mostly"
io_uring: fix error path cleanup in io_sqe_files_register()
io-wq: assign NUMA node locality if appropriate
io_uring: pass required context in as flags
io_uring: rely solely on work flags to determine personality.
io_uring: move io identity items into separate struct
io_uring: COW io_identity on mismatch
io_uring: store io_identity in io_uring_task
io_uring: assign new io_identity for task if members have changed
io_uring: use percpu counters to track inflight requests
io-wq: inherit audit loginuid and sessionid
mm: mark async iocb read as NOWAIT once some data has been copied
mm: use limited read-ahead to satisfy read
Pavel Begunkov (6):
io_uring: don't set COMP_LOCKED if won't put
io_uring: don't unnecessarily clear F_LINK_TIMEOUT
io_uring: don't put a poll req under spinlock
io_uring: dig out COMP_LOCK from deep call chain
io_uring: fix REQ_F_COMP_LOCKED by killing it
io_uring: fix double poll mask init
fs/io-wq.c | 51 ++--
fs/io-wq.h | 18 +-
fs/io_uring.c | 648 ++++++++++++++++++++++++++++-------------------
fs/proc/base.c | 4 +
include/linux/io_uring.h | 27 +-
mm/filemap.c | 8 +
mm/readahead.c | 20 +-
7 files changed, 475 insertions(+), 301 deletions(-)
--
Jens Axboe
