io_uring epoll wait support
Jens Axboe
maintainerv3 cover io_uring epoll wait support
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..2 are just prep patches, and patch 3 adds the epoll change to allow io_uring to queue a callback, if no events are available. Patches 5..6 are just prep patches on the io_uring side, and patch 7 finally adds IORING_OP_EPOLL_WAIT support
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 v2:
- Drop multishot support, to keep the initial version much simpler
- Drop provided buffers support, not required without multishot
- Cleanup epoll bits, notably adding a separate helper for queueing and checking for events
- Various other fixes and cleanups
Signature · 1 line
Jens Axboe
Jens Axboe
maintainerPATCH v3 1/7 eventpoll: abstract out ep_try_send_events() helper
In preparation for reusing this helper in another epoll setup helper, abstract it out.
--- fs/eventpoll.c | 28 ++++++++++++++++++---------- 1 file changed, 18 insertions(+), 10 deletions(-)
fs/eventpoll.c +18 -10
@@ -1980,6 +1980,22 @@ static int ep_autoremove_wake_function(struct wait_queue_entry *wq_entry,return ret; } +static int ep_try_send_events (struct eventpoll *ep, +struct epoll_event __user *events,int maxevents) +{ +int res; + +/* +* Try to transfer events to user space. In case we get 0 events and +* there's still timeout left over, we go trying again in search of +* more luck. +*/ + res =ep_send_events (ep, events, maxevents); +if (res >0 ) +ep_suspend_napi_irqs (ep); +return res; +} +/** * ep_poll - Retrieves ready events, and delivers them to the caller-supplied * event buffer.
@@ -2031,17 +2047,9 @@ static int ep_poll(struct eventpoll *ep, struct epoll_event __user *events,while (1 ) {if (eavail) { -/* -* Try to transfer events to user space. In case we get -* 0 events and there's still timeout left over, we go -* trying again in search of more luck. -*/ - res =ep_send_events (ep, events, maxevents); -if (res) { -if (res >0 ) -ep_suspend_napi_irqs (ep); + res =ep_try_send_events (ep, events, maxevents); +if (res)return res; - } }if (timed_out)
Signature · 1 line
2.47.2
Jens Axboe
maintainerPATCH v3 2/7 eventpoll: abstract out parameter sanity checking
In reply to Jens AxboeAdd a helper that checks the validity of the file descriptor and other parameters passed in to epoll_wait().
--- fs/eventpoll.c | 39 +++++++++++++++++++++++++-------------- 1 file changed, 25 insertions(+), 14 deletions(-)
fs/eventpoll.c +25 -14
@@ -2453,6 +2453,27 @@ SYSCALL_DEFINE4(epoll_ctl, int, epfd, int, op, int, fd,return do_epoll_ctl (epfd, op, fd, &epds, false); } +static int ep_check_params (struct file *file,struct epoll_event __user *evs, +int maxevents) +{ +/* The maximum number of event must be greater than zero */ +if (maxevents <=0 || maxevents >EP_MAX_EVENTS ) +return -EINVAL ; + +/* Verify that the area passed by the user is writeable */ +if (!access_ok (evs, maxevents *sizeof (struct epoll_event ))) +return -EFAULT ; + +/* +* We have to check that the file structure underneath the fd +* the user passed to us _is_ an eventpoll file. +*/ +if (!is_file_epoll (file)) +return -EINVAL ; + +return 0 ; +} +/* * Implement the event wait interface for the eventpoll file. It is the kernel * part of the user space epoll_wait(2).
@@ -2461,26 +2482,16 @@ static int do_epoll_wait(int epfd, struct epoll_event __user *events, int maxevents, struct timespec64 *to) { struct eventpoll *ep; - -/* The maximum number of event must be greater than zero */ -if (maxevents <=0 || maxevents >EP_MAX_EVENTS ) -return -EINVAL ; - -/* Verify that the area passed by the user is writeable */ -if (!access_ok (events, maxevents *sizeof (struct epoll_event ))) -return -EFAULT ; +int ret;/* Get the "struct file *" for the eventpoll file */ CLASS (fd, f)(epfd);if (fd_empty (f))return -EBADF ; -/* -* We have to check that the file structure underneath the fd -* the user passed to us _is_ an eventpoll file. -*/ -if (!is_file_epoll (fd_file (f))) -return -EINVAL ; + ret =ep_check_params (fd_file (f), events, maxevents); +if (unlikely (ret)) +return ret;/* * At this point it is safe to assume that the "private_data" contains
Signature · 1 line
2.47.2
Jens Axboe
maintainerPATCH v3 3/7 eventpoll: add epoll_queue() interface
In reply to Jens AxboeBasic interface that takes a wait_queue_entry rather than post one on the stack, which can be a persistent callback for when new events arrive.
Works like regular epoll_wait(), except it doesn't block. If events are available, they are returned. If none are available, the passed in wait_queue_entry is added to the callback list. The wait_queue_entry must be previously initialized, and the callback provided will be called when events are added to the epoll context.
--- fs/eventpoll.c | 39 +++++++++++++++++++++++++++++++++++++++ include/linux/eventpoll.h | 4 ++++ 2 files changed, 43 insertions(+)
fs/eventpoll.c +39 -0
@@ -1996,6 +1996,33 @@ static int ep_try_send_events(struct eventpoll *ep,return res; } +static int ep_poll_queue (struct eventpoll *ep, +struct epoll_event __user *events,int maxevents, +struct wait_queue_entry *wait) +{ +int res =0 , eavail; + +/* See ep_poll() for commentary */ + eavail =ep_events_available (ep); +while (1 ) { +if (eavail) { + res =ep_try_send_events (ep, events, maxevents); +if (res) +return res; + } +if (!list_empty_careful (&wait->entry)) +break ; +write_lock_irq (&ep->lock); + eavail =ep_events_available (ep); +if (!eavail) +__add_wait_queue_exclusive (&ep->wq, wait); +write_unlock_irq (&ep->lock); +if (!eavail) +break ; + } +return -EIOCBQUEUED ; +} +/** * ep_poll - Retrieves ready events, and delivers them to the caller-supplied * event buffer.
@@ -2474,6 +2501,18 @@ static int ep_check_params(struct file *file, struct epoll_event __user *evs,return 0 ; } +int epoll_queue (struct file *file,struct epoll_event __user *events, +int maxevents,struct wait_queue_entry *wait) +{ +int ret; + + ret =ep_check_params (file, events, maxevents); +if (unlikely (ret)) +return ret; + +return ep_poll_queue (file->private_data, events, maxevents, wait); +} +/* * Implement the event wait interface for the eventpoll file. It is the kernel * part of the user space epoll_wait(2).
include/linux/eventpoll.h +4 -0
@@ -25,6 +25,10 @@ struct file *get_epoll_tfile_raw_ptr(struct file *file, int tfd, unsigned long t/* Used to release the epoll bits inside the "struct file" */ void eventpoll_release_file (struct file *file); +/* Use to reap events, and/or queue for a callback on new events */ +int epoll_queue (struct file *file,struct epoll_event __user *events, +int maxevents,struct wait_queue_entry *wait); +/* * This is called from inside fs/file_table.c:__fput() to unlink files * from the eventpoll interface. We need to have this facility to cleanup
Signature · 1 line
2.47.2
Jens Axboe
maintainerPATCH v3 4/7 eventpoll: add helper to remove wait entry from wait queue head
In reply to Jens Axboe__epoll_wait_remove() is the core helper, it kills a given wait_queue_entry from the eventpoll wait_queue_head. Use it internally, and provide an overall helper, epoll_wait_remove(), which takes a struct file and provides the same functionality.
--- fs/eventpoll.c | 58 +++++++++++++++++++++++++-------------- include/linux/eventpoll.h | 3 ++ 2 files changed, 40 insertions(+), 21 deletions(-)
fs/eventpoll.c +37 -21
@@ -2023,6 +2023,42 @@ static int ep_poll_queue(struct eventpoll *ep,return -EIOCBQUEUED ; } +static int __epoll_wait_remove (struct eventpoll *ep, +struct wait_queue_entry *wait,int timed_out) +{ +int eavail; + +/* +* We were woken up, thus go and try to harvest some events. If timed +* out and still on the wait queue, recheck eavail carefully under +* lock, below. +*/ + eavail =1 ; + +if (!list_empty_careful (&wait->entry)) { +write_lock_irq (&ep->lock); +/* +* If the thread timed out and is not on the wait queue, it +* means that the thread was woken up after its timeout expired +* before it could reacquire the lock. Thus, when wait.entry is +* empty, it needs to harvest events. +*/ +if (timed_out) + eavail =list_empty (&wait->entry); +__remove_wait_queue (&ep->wq, wait); +write_unlock_irq (&ep->lock); + } + +return eavail; +} + +int epoll_wait_remove (struct file *file,struct wait_queue_entry *wait) +{ +if (is_file_epoll (file)) +return __epoll_wait_remove (file->private_data, wait, false); +return -EINVAL ; +} +/** * ep_poll - Retrieves ready events, and delivers them to the caller-supplied * event buffer.
@@ -2135,27 +2171,7 @@ static int ep_poll(struct eventpoll *ep, struct epoll_event __user *events, HRTIMER_MODE_ABS);__set_current_state (TASK_RUNNING ); -/* -* We were woken up, thus go and try to harvest some events. -* If timed out and still on the wait queue, recheck eavail -* carefully under lock, below. -*/ - eavail =1 ; - -if (!list_empty_careful (&wait.entry)) { -write_lock_irq (&ep->lock); -/* -* If the thread timed out and is not on the wait queue, -* it means that the thread was woken up after its -* timeout expired before it could reacquire the lock. -* Thus, when wait.entry is empty, it needs to harvest -* events. -*/ -if (timed_out) - eavail =list_empty (&wait.entry); -__remove_wait_queue (&ep->wq, &wait); -write_unlock_irq (&ep->lock); - } + eavail =__epoll_wait_remove (ep, &wait, timed_out); } }
include/linux/eventpoll.h +3 -0
@@ -29,6 +29,9 @@ void eventpoll_release_file(struct file *file);int epoll_queue (struct file *file,struct epoll_event __user *events,int maxevents,struct wait_queue_entry *wait); +/* Remove wait entry */ +int epoll_wait_remove (struct file *file,struct wait_queue_entry *wait); +/* * This is called from inside fs/file_table.c:__fput() to unlink files * from the eventpoll interface. We need to have this facility to cleanup
Signature · 1 line
2.47.2
Jens Axboe
maintainerPATCH v3 5/7 io_uring/epoll: remove CONFIG_EPOLL guards
In reply to Jens AxboeJust have the Makefile add the object if epoll is enabled, then it's not necessary to guard the entire epoll.c file inside an CONFIG_EPOLL ifdef.
--- io_uring/Makefile | 9 +++++---- io_uring/epoll.c | 2 -- 2 files changed, 5 insertions(+), 6 deletions(-)
io_uring/Makefile +5 -4
@@ -11,9 +11,10 @@ obj-$(CONFIG_IO_URING) += io_uring.o opdef.o kbuf.o rsrc.o notif.o \ eventfd.o uring_cmd.o openclose.o \ sqpoll.o xattr.o nop.o fs.o splice.o \ sync.o msg_ring.o advise.o openclose.o \ - epoll.o statx.o timeout.o fdinfo.o \ - cancel.o waitid.o register.o \ - truncate.o memmap.o alloc_cache.o + statx.o timeout.o fdinfo.o cancel.o \ + waitid.o register.o truncate.o \ + memmap.o alloc_cache.o obj-$(CONFIG_IO_WQ) += io-wq.o obj-$(CONFIG_FUTEX) += futex.o -obj-$(CONFIG_NET_RX_BUSY_POLL) += napi.o +obj-$(CONFIG_EPOLL) += epoll.o +obj-$(CONFIG_NET_RX_BUSY_POLL) += napi.o
io_uring/epoll.c +0 -2
@@ -12,7 +12,6 @@#include "io_uring.h" #include "epoll.h" -#if defined(CONFIG_EPOLL)struct io_epoll {struct file *file;int epfd;
@@ -58,4 +57,3 @@ int io_epoll_ctl(struct io_kiocb *req, unsigned int issue_flags)io_req_set_res (req, ret,0 );return IOU_OK ; } -#endif
Signature · 1 line
2.47.2
Jens Axboe
maintainerPATCH v3 6/7 io_uring/poll: pull ownership handling into poll.h
In reply to Jens AxboeIn preparation for using it from somewhere else. Rather than try and duplicate the functionality, just make it generically available to io_uring opcodes.
Note: would have to be used carefully, cannot be used by opcodes that can trigger poll logic.
--- io_uring/poll.c | 30 +----------------------------- io_uring/poll.h | 31 +++++++++++++++++++++++++++++++ 2 files changed, 32 insertions(+), 29 deletions(-)
io_uring/poll.c +1 -29
@@ -41,16 +41,6 @@ struct io_poll_table {__poll_t result_mask; }; -#define IO_POLL_CANCEL_FLAG BIT(31) -#define IO_POLL_RETRY_FLAG BIT(30) -#define IO_POLL_REF_MASK GENMASK(29, 0) - -/* -* We usually have 1-2 refs taken, 128 is more than enough and we want to -* maximise the margin between this amount and the moment when it overflows. -*/ -#define IO_POLL_REF_BIAS 128 -#define IO_WQE_F_DOUBLE 1 static int io_poll_wake(struct wait_queue_entry *wait, unsigned mode, int sync,
@@ -70,7 +60,7 @@ static inline bool wqe_is_double(struct wait_queue_entry *wqe)return priv &IO_WQE_F_DOUBLE ; } -static bool io_poll_get_ownership_slowpath (struct io_kiocb *req) +bool io_poll_get_ownership_slowpath (struct io_kiocb *req) {int v;
@@ -85,24 +75,6 @@ static bool io_poll_get_ownership_slowpath(struct io_kiocb *req)return !(atomic_fetch_inc (&req->poll_refs) &IO_POLL_REF_MASK ); } -/* -* If refs part of ->poll_refs (see IO_POLL_REF_MASK) is 0, it's free. We can -* bump it and acquire ownership. It's disallowed to modify requests while not -* owning it, that prevents from races for enqueueing task_work's and b/w -* arming poll and wakeups. -*/ -static inline bool io_poll_get_ownership (struct io_kiocb *req) -{ -if (unlikely (atomic_read (&req->poll_refs) >=IO_POLL_REF_BIAS )) -return io_poll_get_ownership_slowpath (req); -return !(atomic_fetch_inc (&req->poll_refs) &IO_POLL_REF_MASK ); -} - -static void io_poll_mark_cancelled (struct io_kiocb *req) -{ -atomic_or (IO_POLL_CANCEL_FLAG , &req->poll_refs); -} -static struct io_poll *io_poll_get_double (struct io_kiocb *req) {/* pure poll stashes this in ->async_data, poll driven retry elsewhere */
io_uring/poll.h +31 -0
@@ -21,6 +21,18 @@ struct async_poll {struct io_poll *double_poll; }; +#define IO_POLL_CANCEL_FLAG BIT(31) +#define IO_POLL_RETRY_FLAG BIT(30) +#define IO_POLL_REF_MASK GENMASK(29, 0) + +bool io_poll_get_ownership_slowpath (struct io_kiocb *req); + +/* +* We usually have 1-2 refs taken, 128 is more than enough and we want to +* maximise the margin between this amount and the moment when it overflows. +*/ +#define IO_POLL_REF_BIAS 128 +/* * Must only be called inside issue_flags & IO_URING_F_MULTISHOT, or * potentially other cases where we already "own" this poll request.
@@ -30,6 +42,25 @@ static inline void io_poll_multishot_retry(struct io_kiocb *req)atomic_inc (&req->poll_refs); } +/* +* If refs part of ->poll_refs (see IO_POLL_REF_MASK) is 0, it's free. We can +* bump it and acquire ownership. It's disallowed to modify requests while not +* owning it, that prevents from races for enqueueing task_work's and b/w +* arming poll and wakeups. +*/ +static inline bool io_poll_get_ownership (struct io_kiocb *req) +{ +if (unlikely (atomic_read (&req->poll_refs) >=IO_POLL_REF_BIAS )) +return io_poll_get_ownership_slowpath (req); +return !(atomic_fetch_inc (&req->poll_refs) &IO_POLL_REF_MASK ); +} + +static inline void io_poll_mark_cancelled (struct io_kiocb *req) +{ +atomic_or (IO_POLL_CANCEL_FLAG , &req->poll_refs); +} + +int io_poll_add_prep (struct io_kiocb *req,const struct io_uring_sqe *sqe);int io_poll_add (struct io_kiocb *req,unsigned int issue_flags);
Signature · 1 line
2.47.2
Jens Axboe
maintainerPATCH v3 7/7 io_uring/epoll: add support for IORING_OP_EPOLL_WAIT
In reply to Jens AxboeFor existing epoll event loops that can't fully convert to io_uring, the used approach is usually to add the io_uring fd to the epoll instance and use epoll_wait() to wait on both "legacy" and io_uring events. While this work, it isn't optimal as:
1) epoll_wait() is pretty limited in what it can do. It does not support partial reaping of events, or waiting on a batch of events.
2) When an io_uring ring is added to an epoll instance, it activates the io_uring "I'm being polled" logic which slows things down.
Rather than use this approach, with EPOLL_WAIT support added to io_uring, event loops can use the normal io_uring wait logic for everything, as long as an epoll wait request has been armed with io_uring.
Note that IORING_OP_EPOLL_WAIT does NOT take a timeout value, as this is an async request. Waiting on io_uring events in general has various timeout parameters, and those are the ones that should be used when waiting on any kind of request. If events are immediately available for reaping, then This opcode will return those immediately. If none are available, then it will post an async completion when they become available.
cqe->res will contain either an error code (< 0 value) for a malformed request, invalid epoll instance, etc. It will return a positive result indicating how many events were reaped.
IORING_OP_EPOLL_WAIT requests may be canceled using the normal io_uring cancelation infrastructure. The poll logic for managing ownership is adopted to guard the epoll side too.
--- include/linux/io_uring_types.h | 4 + include/uapi/linux/io_uring.h | 1 + io_uring/cancel.c | 5 ++ io_uring/epoll.c | 143 +++++++++++++++++++++++++++++++++ io_uring/epoll.h | 22 +++++ io_uring/io_uring.c | 5 ++ io_uring/opdef.c | 14 ++++ 7 files changed, 194 insertions(+)
include/linux/io_uring_types.h +4 -0
@@ -369,6 +369,10 @@ struct io_ring_ctx {struct io_alloc_cache futex_cache;#endif +#ifdef CONFIG_EPOLL +struct hlist_head epoll_list; +#endif +const struct cred *sq_creds;/* cred used for __io_sq_thread() */ struct io_sq_data *sq_data;/* if using sq thread polling */
include/uapi/linux/io_uring.h +1 -0
@@ -278,6 +278,7 @@ enum io_uring_op {IORING_OP_FTRUNCATE ,IORING_OP_BIND ,IORING_OP_LISTEN , +IORING_OP_EPOLL_WAIT ,/* this goes last, obviously */ IORING_OP_LAST,
io_uring/cancel.c +5 -0
@@ -17,6 +17,7 @@#include "timeout.h" #include "waitid.h" #include "futex.h" +#include "epoll.h" #include "cancel.h" struct io_cancel {
@@ -128,6 +129,10 @@ int io_try_cancel(struct io_uring_task *tctx, struct io_cancel_data *cd,if (ret != -ENOENT )return ret; + ret =io_epoll_wait_cancel (ctx, cd, issue_flags); +if (ret != -ENOENT ) +return ret; +spin_lock (&ctx->completion_lock);if (!(cd->flags &IORING_ASYNC_CANCEL_FD )) ret =io_timeout_cancel (ctx, cd);
io_uring/epoll.c +143 -0
158 lines not on this page · raw message
io_uring/epoll.h +22 -0
@@ -1,6 +1,28 @@// SPDX-License-Identifier: GPL-2.0 +#include "cancel.h" +#if defined(CONFIG_EPOLL ) +int io_epoll_wait_cancel (struct io_ring_ctx *ctx,struct io_cancel_data *cd, +unsigned int issue_flags); +bool io_epoll_wait_remove_all (struct io_ring_ctx *ctx,struct io_uring_task *tctx, +bool cancel_all); +int io_epoll_ctl_prep (struct io_kiocb *req,const struct io_uring_sqe *sqe);int io_epoll_ctl (struct io_kiocb *req,unsigned int issue_flags); +int io_epoll_wait_prep (struct io_kiocb *req,const struct io_uring_sqe *sqe); +int io_epoll_wait (struct io_kiocb *req,unsigned int issue_flags); +#else +static inline bool io_epoll_wait_remove_all (struct io_ring_ctx *ctx, +struct io_uring_task *tctx, +bool cancel_all) +{ +return false; +} +static inline int io_epoll_wait_cancel (struct io_ring_ctx *ctx, +struct io_cancel_data *cd, +unsigned int issue_flags) +{ +return 0 ; +}#endif
io_uring/io_uring.c +5 -0
@@ -93,6 +93,7 @@#include "notif.h" #include "waitid.h" #include "futex.h" +#include "epoll.h" #include "napi.h" #include "uring_cmd.h" #include "msg_ring.h"
@@ -356,6 +357,9 @@ static __cold struct io_ring_ctx *io_ring_ctx_alloc(struct io_uring_params *p)INIT_HLIST_HEAD (&ctx->waitid_list);#ifdef CONFIG_FUTEX INIT_HLIST_HEAD (&ctx->futex_list); +#endif +#ifdef CONFIG_EPOLL +INIT_HLIST_HEAD (&ctx->epoll_list);#endif INIT_DELAYED_WORK (&ctx->fallback_work, io_fallback_req_func);INIT_WQ_LIST (&ctx->submit_state.compl_reqs);
@@ -3079,6 +3083,7 @@ static __cold bool io_uring_try_cancel_requests(struct io_ring_ctx *ctx, ret |=io_poll_remove_all (ctx, tctx, cancel_all); ret |=io_waitid_remove_all (ctx, tctx, cancel_all); ret |=io_futex_remove_all (ctx, tctx, cancel_all); + ret |=io_epoll_wait_remove_all (ctx, tctx, cancel_all); ret |=io_uring_try_cancel_uring_cmd (ctx, tctx, cancel_all);mutex_unlock (&ctx->uring_lock); ret |=io_kill_timeouts (ctx, tctx, cancel_all);
io_uring/opdef.c +14 -0
@@ -514,6 +514,17 @@ const struct io_issue_def io_issue_defs[] = { .async_size = sizeof(struct io_async_msghdr), #else .prep = io_eopnotsupp_prep, +#endif + }, + [IORING_OP_EPOLL_WAIT ] = { + .needs_file =1 , + .unbound_nonreg_file =1 , + .audit_skip =1 , +#if defined(CONFIG_EPOLL) + .prep = io_epoll_wait_prep, + .issue = io_epoll_wait, +#else + .prep = io_eopnotsupp_prep, #endif }, };
@@ -745,6 +756,9 @@ const struct io_cold_def io_cold_defs[] = { [IORING_OP_LISTEN ] = { .name ="LISTEN" , }, + [IORING_OP_EPOLL_WAIT ] = { + .name ="EPOLL_WAIT" , + }, };const char *io_uring_get_opcode (u8 opcode)
Signature · 1 line
2.47.2
Re: [PATCH 7/7] io_uring/epoll: add support for IORING_OP_EPOLL_WAIT
Quoted 47 lines · Jens Axboe
For existing epoll event loops that can't fully convert to
io_uring, the used approach is usually to add theio_uringfd to the epoll instance and useepoll_wait()to wait on both "legacy" andio_uringevents. While this work, it isn't optimal as:1)
epoll_wait()is pretty limited in what it can do. It does not support partial reaping of events, or waiting on a batch of events.2) When an
io_uringring is added to an epoll instance, it activates theio_uring"I'm being polled" logic which slows things down.Rather than use this approach, with
EPOLL_WAITsupport added toio_uring, event loops can use the normalio_uringwait logic for everything, as long as an epoll wait request has been armed withio_uring.Note that
IORING_OP_EPOLL_WAITdoes NOT take a timeout value, as this is an async request. Waiting onio_uringevents in general has various timeout parameters, and those are the ones that should be used when waiting on any kind of request. If events are immediately available for reaping, then This opcode will return those immediately. If none are available, then it will post an async completion when they become available.
cqe->reswill contain either an error code (< 0 value) for a malformed request, invalid epoll instance, etc. It will return a positive result indicating how many events were reaped.
IORING_OP_EPOLL_WAITrequests may be canceled using the normalio_uringcancelation infrastructure. The poll logic for managing ownership is adopted to guard the epoll side too.Signed-off-by Jens Axboe <[email protected]>--- include/linux/io_uring_types.h | 4 + include/uapi/linux/io_uring.h | 1 + io_uring/cancel.c | 5 ++ io_uring/epoll.c | 143 +++++++++++++++++++++++++++++++++ io_uring/epoll.h | 22 +++++ io_uring/io_uring.c | 5 ++ io_uring/opdef.c | 14 ++++ 7 files changed, 194 insertions(+)include/linux/io_uring_types.h +0
-0@@ -369,6 +369,10 @@ struct io_ring_ctx {
...
Quoted 18 lines
include/linux/io_uring_types.h +18
-0+bool io_epoll_wait_remove_all (struct io_ring_ctx *ctx,struct io_uring_task *tctx, +bool cancel_all) +{ +return io_cancel_remove_all (ctx, tctx, &ctx->epoll_list, cancel_all, __io_epoll_wait_cancel); +} + +int io_epoll_wait_cancel (struct io_ring_ctx *ctx,struct io_cancel_data *cd, +unsigned int issue_flags) +{ +return io_cancel_remove (ctx, cd, issue_flags, &ctx->epoll_list, __io_epoll_wait_cancel); +} + +static void io_epoll_retry (struct io_kiocb *req,struct io_tw_state *ts) +{ +int v; + + do { + v = atomic_read(&req->poll_refs);
include/linux/io_uring_types.h
Quoted 10 lines
include/linux/io_uring_types.h +10
-0+if (unlikely (v !=1 )) { +if (WARN_ON_ONCE (!(v &IO_POLL_REF_MASK ))) +return ; +if (v &IO_POLL_CANCEL_FLAG ) { +__io_epoll_cancel (req); +return ; + } + } + v &=IO_POLL_REF_MASK ; + }while (atomic_sub_return (v, &req->poll_refs) &IO_POLL_REF_MASK );
I actually looked up the epoll code this time. If we disregard cancellations, you have only 1 wait entry, which should've been removed from the queue by io_epoll_wait_fn(), in which case the entire loop is doing nothing as there is no one to race with. ->hash_node is the only shared part, but it's sync'ed by the mutex.
As for cancellation, epoll_wait_remove() also removes the entry, and you can rely on it to tell if the entry was removed inside, from which you derive if you're the current owner.
Maybe this handling might be useful for the multishot mode, perhaps along the lines of:
io_epoll_retry()
{
do {
res = epoll_get_events();
if (one_shot || cancel) {
wq_remove();
unhash();
complete_req(res);
return;
}
post_cqe(res);
// now recheck if new events came while we were processing
// the previous batch.
} while (refs_drop(req->poll_refs));
}epoll_issue(issue_flags) {
queue_poll();
return;
}But it might be better to just poll the epoll fd, reuse all the io_uring polling machinery, and implement IO_URING_F_MULTISHOT for the epoll opcode.
epoll_issue(issue_flags) {
if (!(flags & IO_URING_F_MULTISHOT))
return -EAGAIN;
res = epoll_check_events();
post_cqe(res);
etc.
}I think that would make this patch quite trivial, including the multishot mode.
Signature · 1 line
Pavel Begunkov
Re: [PATCH 7/7] io_uring/epoll: add support for IORING_OP_EPOLL_WAIT
On 2/8/25 23:27, Pavel Begunkov wrote:
...
Quoted 8 lines
But it might be better to just poll the epoll fd, reuse all the
io_uringpolling machinery, and implementIO_URING_F_MULTISHOTfor the epoll opcode.
epoll_issue(issue_flags){
if (!(flags &IO_URING_F_MULTISHOT))
return-EAGAIN;
Quoted 7 lines
res =
epoll_check_events();
post_cqe(res);
etc.
}I think that would make this patch quite trivial, including the multishot mode.
Something like this instead of the last patch. Completely untested, the eventpoll.c hunk is dirty might be incorrect, need to pass the right mask for polling, and all that. At least it looks simpler, and probably doesn't need half of the prep patches.
fs/eventpoll.c +10 -31
@@ -1996,33 +1996,6 @@ static int ep_try_send_events(struct eventpoll *ep,return res; } -static int ep_poll_queue (struct eventpoll *ep, -struct epoll_event __user *events,int maxevents, -struct wait_queue_entry *wait) -{ -int res =0 , eavail; - -/* See ep_poll() for commentary */ - eavail =ep_events_available (ep); -while (1 ) { -if (eavail) { - res =ep_try_send_events (ep, events, maxevents); -if (res) -return res; - } -if (!list_empty_careful (&wait->entry)) -break ; -write_lock_irq (&ep->lock); - eavail =ep_events_available (ep); -if (!eavail) -__add_wait_queue_exclusive (&ep->wq, wait); -write_unlock_irq (&ep->lock); -if (!eavail) -break ; - } -return -EIOCBQUEUED ; -} -static int __epoll_wait_remove (struct eventpoll *ep,struct wait_queue_entry *wait,int timed_out) {
@@ -2517,16 +2490,22 @@ static int ep_check_params(struct file *file, struct epoll_event __user *evs,return 0 ; } -int epoll_queue (struct file *file,struct epoll_event __user *events, -int maxevents,struct wait_queue_entry *wait) +int epoll_sendevents (struct file *file,struct epoll_event __user *events, +int maxevents) { -int ret; +int res =0 , eavail; ret =ep_check_params (file, events, maxevents);if (unlikely (ret))return ret; -return ep_poll_queue (file->private_data, events, maxevents, wait); + eavail =ep_events_available (ep); +if (eavail) { + res =ep_try_send_events (ep, events, maxevents); +if (res) +return res; + } +return 0 ; }/*
include/linux/eventpoll.h +2 -3
@@ -25,9 +25,8 @@ struct file *get_epoll_tfile_raw_ptr(struct file *file, int tfd, unsigned long t/* Used to release the epoll bits inside the "struct file" */ void eventpoll_release_file (struct file *file); -/* Use to reap events, and/or queue for a callback on new events */ -int epoll_queue (struct file *file,struct epoll_event __user *events, -int maxevents,struct wait_queue_entry *wait); +int epoll_sendevents (struct file *file,struct epoll_event __user *events, +int maxevents);/* Remove wait entry */ int epoll_wait_remove (struct file *file,struct wait_queue_entry *wait);
include/uapi/linux/io_uring.h +1 -0
@@ -278,6 +278,7 @@ enum io_uring_op {IORING_OP_FTRUNCATE ,IORING_OP_BIND ,IORING_OP_LISTEN , +IORING_OP_EPOLL_WAIT ,/* this goes last, obviously */ IORING_OP_LAST,
io_uring/epoll.c +33 -0
@@ -20,6 +20,12 @@ struct io_epoll {struct epoll_event event; }; +struct io_epoll_wait { +struct file *file; +int maxevents; +struct epoll_event __user *events; +}; +int io_epoll_ctl_prep (struct io_kiocb *req,const struct io_uring_sqe *sqe) { struct io_epoll *epoll = io_kiocb_to_cmd(req, struct io_epoll);
@@ -57,3 +63,30 @@ int io_epoll_ctl(struct io_kiocb *req, unsigned int issue_flags)io_req_set_res (req, ret,0 );return IOU_OK ; } + +int io_epoll_wait_prep (struct io_kiocb *req,const struct io_uring_sqe *sqe) +{ + struct io_epoll_wait *iew = io_kiocb_to_cmd(req, struct io_epoll_wait); + +if (sqe->off || sqe->rw_flags || sqe->buf_index || sqe->splice_fd_in) +return -EINVAL ; + + iew->maxevents =READ_ONCE (sqe->len); + iew->events =u64_to_user_ptr (READ_ONCE (sqe->addr)); +return 0 ; +} + +int io_epoll_wait (struct io_kiocb *req,unsigned int issue_flags) +{ + struct io_epoll_wait *iew = io_kiocb_to_cmd(req, struct io_epoll_wait); +int ret; + + ret =epoll_sendevents (req->file, iew->events, iew->maxevents); +if (ret ==0 ) +return -EAGAIN ; +if (ret <0 ) +req_set_fail (req); + +io_req_set_res (req, ret,0 ); +return IOU_OK ; +}
io_uring/epoll.h +2 -0
@@ -3,4 +3,6 @@#if defined(CONFIG_EPOLL )int io_epoll_ctl_prep (struct io_kiocb *req,const struct io_uring_sqe *sqe);int io_epoll_ctl (struct io_kiocb *req,unsigned int issue_flags); +int io_epoll_wait_prep (struct io_kiocb *req,const struct io_uring_sqe *sqe); +int io_epoll_wait (struct io_kiocb *req,unsigned int issue_flags);#endif
io_uring/opdef.c +15 -0
@@ -514,6 +514,18 @@ const struct io_issue_def io_issue_defs[] = { .async_size = sizeof(struct io_async_msghdr), #else .prep = io_eopnotsupp_prep, +#endif + }, + [IORING_OP_EPOLL_WAIT ] = { + .needs_file =1 , + .audit_skip =1 , + .pollout =1 , + .pollin =1 , +#if defined(CONFIG_EPOLL) + .prep = io_epoll_wait_prep, + .issue = io_epoll_wait, +#else + .prep = io_eopnotsupp_prep, #endif }, };
@@ -745,6 +757,9 @@ const struct io_cold_def io_cold_defs[] = { [IORING_OP_LISTEN ] = { .name ="LISTEN" , }, + [IORING_OP_EPOLL_WAIT ] = { + .name ="EPOLL_WAIT" , + }, };const char *io_uring_get_opcode (u8 opcode)
Jens Axboe
maintainerRe: [PATCH 7/7] io_uring/epoll: add support for IORING_OP_EPOLL_WAIT
Quoted 15 lines · Pavel Begunkov
On 2/8/25 23:27, Pavel Begunkov wrote:
...Quoted 13 lines
But it might be better to just poll the epoll fd, reuse all the
io_uringpolling machinery, and implementIO_URING_F_MULTISHOTfor the epoll opcode.epoll_issue(issue_flags) { if (!(flags & IO_URING_F_MULTISHOT)) return -EAGAIN; res = epoll_check_events(); post_cqe(res); etc. }
Quoted 7 lines
Quoted 2 lines
I think that would make this patch quite trivial, including the multishot mode.
Something like this instead of the last patch. Completely untested, the
eventpoll.chunk is dirty might be incorrect, need to pass the right mask for polling, and all that. At least it looks simpler, and probably doesn't need half of the prep patches.
I like that idea! I'll roll with it and get it finalized and then do some testing.
Signature · 1 line
Jens Axboe