| Message ID | 3ad3c346-1415-45bd-bcb2-2f9b46164f30@kernel.dk (mailing list archive) |
|---|---|
| State | New |
| Headers | show |
| Series | [GIT,PULL] io_uring epoll reaping support | expand |
On 3/27/25 5:47 AM, Jens Axboe wrote: > Hi Linus, > > This sits on top of the recently sent out zero-copy rx pull request. > > This adds support for reading epoll events via io_uring. While this may > seem counter-intuitive (and/or productive), the reasoning here is that > quite a few existing epoll event loops can easily do a partial > conversion to a completion based model, but are still stuck with one (or > few) event types that remain readiness based. For that case, they then > need to add the io_uring fd to the epoll context, and continue to rely > on epoll_wait(2) for waiting on events. This misses out on the finer > grained waiting that io_uring can do, to reduce context switches and > wait for multiple events in one batch reliably. > > With adding support for reaping epoll events via io_uring, the whole > legacy readiness based event types can still be reaped via epoll, with > the overall waiting in the loop be driven by io_uring. > > Relies on a prep patch that went in via the VFS tree already. Please > pull! And even with having prepared this earlier, guess a big conference dinner was enough to still make me mess this up. The correct git location is of course: git://git.kernel.dk/linux.git for-6.15/io_uring-epoll-wait-20250325 which is what I get for creating the signed tag later than the actual PR write-up. Everything else in the original should be fine as-is, thanks!
The pull request you sent on Thu, 27 Mar 2025 05:47:56 -0600:
> git://git.kernel.dk/linux.git for-6.15/io_uring-rx-zc-20250325
has been merged into torvalds/linux.git:
https://git.kernel.org/torvalds/c/6e3da40ed6f3508dcf34b1539496bcccef7015ef
Thank you!
Hi Linus, This sits on top of the recently sent out zero-copy rx pull request. This adds support for reading epoll events via io_uring. While this may seem counter-intuitive (and/or productive), the reasoning here is that quite a few existing epoll event loops can easily do a partial conversion to a completion based model, but are still stuck with one (or few) event types that remain readiness based. For that case, they then need to add the io_uring fd to the epoll context, and continue to rely on epoll_wait(2) for waiting on events. This misses out on the finer grained waiting that io_uring can do, to reduce context switches and wait for multiple events in one batch reliably. With adding support for reaping epoll events via io_uring, the whole legacy readiness based event types can still be reaped via epoll, with the overall waiting in the loop be driven by io_uring. Relies on a prep patch that went in via the VFS tree already. Please pull! The following changes since commit 0511f4e6a16988e485a5e60727c89e2ca9f83f44: Merge patch series "epoll changes for io_uring wait support" (2025-02-20 10:18:47 +0100) are available in the Git repository at: git://git.kernel.dk/linux.git for-6.15/io_uring-rx-zc-20250325 for you to fetch changes up to 19f7e942732766aec8a51d217ab5fb4a7fe3bb0d: io_uring/epoll: add support for IORING_OP_EPOLL_WAIT (2025-02-20 07:59:56 -0700) ---------------------------------------------------------------- Jens Axboe (5): Merge branch 'for-6.15/io_uring' into for-6.15/io_uring-epoll-wait Merge branch 'for-6.15/io_uring-rx-zc' into for-6.15/io_uring-epoll-wait Merge branch 'vfs-6.15.eventpoll' of https://git.kernel.org/pub/scm/linux/kernel/git/vfs/vfs into for-6.15/io_uring-epoll-wait io_uring/epoll: remove CONFIG_EPOLL guards io_uring/epoll: add support for IORING_OP_EPOLL_WAIT include/uapi/linux/io_uring.h | 1 + io_uring/Makefile | 9 +++++---- io_uring/epoll.c | 35 +++++++++++++++++++++++++++++++++-- io_uring/epoll.h | 2 ++ io_uring/opdef.c | 14 ++++++++++++++ 5 files changed, 55 insertions(+), 6 deletions(-)