From: Willem de Bruijn <willemb@xxxxxxxxxx> The underlying hrtimer is programmed with nanosecond resolution. Use cases such as datacenter networking operate on timescales well below milliseconds. Setting shorter timeouts bounds tail latency. Add epoll_create1 flag EPOLL_NSTIMEO. When passed, this changes the interpretation of argument timeout in epoll_wait from msec to nsec. The new eventpoll state fits in existing 4B of padding when busy poll is compiled in (the default), and reads the same cacheline. Signed-off-by: Willem de Bruijn <willemb@xxxxxxxxxx> --- Selftest for now at github. Can follow-up for kselftests. https://github.com/wdebruij/kerneltools/blob/master/tests/epoll_nstimeo.c --- fs/eventpoll.c | 26 +++++++++++++++++++------- include/uapi/linux/eventpoll.h | 1 + 2 files changed, 20 insertions(+), 7 deletions(-) diff --git a/fs/eventpoll.c b/fs/eventpoll.c index 4df61129566d..1216b909d155 100644 --- a/fs/eventpoll.c +++ b/fs/eventpoll.c @@ -225,6 +225,9 @@ struct eventpoll { unsigned int napi_id; #endif + /* Accept timeout in ns resolution (EPOLL_NSTIMEO) */ + unsigned int nstimeout:1; + #ifdef CONFIG_DEBUG_LOCK_ALLOC /* tracks wakeup nests for lockdep validation */ u8 nests; @@ -1787,17 +1790,20 @@ static int ep_send_events(struct eventpoll *ep, return esed.res; } -static inline struct timespec64 ep_set_mstimeout(long ms) +static inline struct timespec64 ep_set_nstimeout(long ns) { - struct timespec64 now, ts = { - .tv_sec = ms / MSEC_PER_SEC, - .tv_nsec = NSEC_PER_MSEC * (ms % MSEC_PER_SEC), - }; + struct timespec64 now, ts; + ts = ns_to_timespec64(ns); ktime_get_ts64(&now); return timespec64_add_safe(now, ts); } +static inline struct timespec64 ep_set_mstimeout(long ms) +{ + return ep_set_nstimeout(NSEC_PER_MSEC * ms); +} + /** * ep_poll - Retrieves ready events, and delivers them to the caller supplied * event buffer. @@ -1826,7 +1832,10 @@ static int ep_poll(struct eventpoll *ep, struct epoll_event __user *events, lockdep_assert_irqs_enabled(); if (timeout > 0) { - struct timespec64 end_time = ep_set_mstimeout(timeout); + struct timespec64 end_time; + + end_time = ep->nstimeout ? ep_set_nstimeout(timeout) : + ep_set_mstimeout(timeout); slack = select_estimate_accuracy(&end_time); to = &expires; @@ -2046,7 +2055,7 @@ static int do_epoll_create(int flags) /* Check the EPOLL_* constant for consistency. */ BUILD_BUG_ON(EPOLL_CLOEXEC != O_CLOEXEC); - if (flags & ~EPOLL_CLOEXEC) + if (flags & ~(EPOLL_CLOEXEC | EPOLL_NSTIMEO)) return -EINVAL; /* * Create the internal data structure ("struct eventpoll"). @@ -2054,6 +2063,9 @@ static int do_epoll_create(int flags) error = ep_alloc(&ep); if (error < 0) return error; + + ep->nstimeout = !!(flags & EPOLL_NSTIMEO); + /* * Creates all the items needed to setup an eventpoll file. That is, * a file structure and a free file descriptor. diff --git a/include/uapi/linux/eventpoll.h b/include/uapi/linux/eventpoll.h index 8a3432d0f0dc..f6ef9c9f8ac2 100644 --- a/include/uapi/linux/eventpoll.h +++ b/include/uapi/linux/eventpoll.h @@ -21,6 +21,7 @@ /* Flags for epoll_create1. */ #define EPOLL_CLOEXEC O_CLOEXEC +#define EPOLL_NSTIMEO 0x1 /* Valid opcodes to issue to sys_epoll_ctl() */ #define EPOLL_CTL_ADD 1 -- 2.29.0.rc1.297.gfa9743e501-goog
