[PATCHSET,v2,0/11] io_uring epoll wait support

Message

Jens Axboe Feb. 4, 2025, 7:46 p.m. UTC

Hi,

One issue people consistently run into when converting legacy epoll
event loops with io_uring is that parts of the event loop still needs to
use epoll. And since event loops generally need to wait in one spot,
they add the io_uring fd to the epoll set and continue to use
epoll_wait(2) to wait on events. This is suboptimal on the io_uring
front as there's now an active poller on the ring, and it's suboptimal
as it doesn't give the application the batch waiting (with fine grained
timeouts) that io_uring provides.

This patchset adds support for IORING_OP_EPOLL_WAIT, which does an async
epoll_wait() operation. No sleeping or thread offload is involved, it
relies on the wait_queue_entry callback for retries. With that, then
the above event loops can continue to use epoll for certain parts, but
bundle it all under waiting on the ring itself rather than add the ring
fd to the epoll set.

Patches 1..4 are just prep patches, and patch 5 adds the epoll change
to allow io_uring to queue a callback, if no events are available.
Patches 6..7 are just prep patches on the io_uring side, and patch 8
finally adds IORING_OP_EPOLL_WAIT support. Patch 9 adds multishot
support, which further gets rid of repeated write_lock and list
manipulations on the struct eventpoll waitqueue head. This last bit
should be a nice win, having a persistent waitqueue entry rather
than needing to lock/add/unlock for each epoll_wait() equivalent
operation.

Patches can also be found here:

https://git.kernel.dk/cgit/linux/log/?h=io_uring-epoll-wait

and are against 6.14-rc1 + already pending io_uring patches.

Since v1:
- Add provided buffer support and require it for multishot
- Fixup various multishot issues
- Various other fixes

 fs/eventpoll.c                 | 155 ++++++++++++++++-------
 include/linux/eventpoll.h      |   8 ++
 include/linux/io_uring_types.h |   4 +
 include/uapi/linux/io_uring.h  |   7 ++
 io_uring/Makefile              |   9 +-
 io_uring/cancel.c              |   5 +
 io_uring/epoll.c               | 222 ++++++++++++++++++++++++++++++++-
 io_uring/epoll.h               |  22 ++++
 io_uring/io_uring.c            |   5 +
 io_uring/opdef.c               |  15 +++
 io_uring/poll.c                |  30 +----
 io_uring/poll.h                |  32 +++++
 12 files changed, 434 insertions(+), 80 deletions(-)