* [PATCH v2] io_uring: Add support for napi_busy_poll
@ 2022-02-27 17:20 Olivier Langlois
2022-02-27 20:21 ` kernel test robot
0 siblings, 1 reply; 2+ messages in thread
From: Olivier Langlois @ 2022-02-27 17:20 UTC (permalink / raw)
To: Jens Axboe; +Cc: Pavel Begunkov, Hao Xu, io-uring, linux-kernel, gwml
The sqpoll thread can be used for performing the napi busy poll in a
similar way that it does io polling for file systems supporting direct
access bypassing the page cache.
The other way that io_uring can be used for napi busy poll is by
calling io_uring_enter() to get events.
If the user specify a timeout value, it is distributed between polling
and sleeping by using the systemwide setting
/proc/sys/net/core/busy_poll.
The changes have been tested with this program:
https://github.com/lano1106/io_uring_udp_ping
and the result is:
Without sqpoll:
NAPI busy loop disabled:
rtt min/avg/max/mdev = 40.631/42.050/58.667/1.547 us
NAPI busy loop enabled:
rtt min/avg/max/mdev = 30.619/31.753/61.433/1.456 us
With sqpoll:
NAPI busy loop disabled:
rtt min/avg/max/mdev = 42.087/44.438/59.508/1.533 us
NAPI busy loop enabled:
rtt min/avg/max/mdev = 35.779/37.347/52.201/0.924 us
v2:
* Evaluate list_empty(&ctx->napi_list) outside io_napi_busy_loop() to keep
__io_sq_thread() execution as fast as possible
* In io_cqring_wait(), move up the sig block to avoid needless computation
if the block exits the function
* In io_cqring_wait(), protect ctx->napi_list from race condition by
splicing it into a local list
* In io_cqring_wait(), allow busy polling when uts is missing
* Fix kernel test robot issues
Co-developed-by: Hao Xu <[email protected]>
Signed-off-by: Hao Xu <[email protected]>
Signed-off-by: Olivier Langlois <[email protected]>
---
fs/io_uring.c | 238 ++++++++++++++++++++++++++++++++++++++++++++++++--
1 file changed, 229 insertions(+), 9 deletions(-)
diff --git a/fs/io_uring.c b/fs/io_uring.c
index 4715980e9015..2759a60339d2 100644
--- a/fs/io_uring.c
+++ b/fs/io_uring.c
@@ -63,6 +63,7 @@
#include <net/sock.h>
#include <net/af_unix.h>
#include <net/scm.h>
+#include <net/busy_poll.h>
#include <linux/anon_inodes.h>
#include <linux/sched/mm.h>
#include <linux/uaccess.h>
@@ -395,6 +396,10 @@ struct io_ring_ctx {
struct list_head sqd_list;
unsigned long check_cq_overflow;
+#ifdef CONFIG_NET_RX_BUSY_POLL
+ /* used to track busy poll napi_id */
+ struct list_head napi_list;
+#endif
struct {
unsigned cached_cq_tail;
@@ -1464,6 +1469,9 @@ static __cold struct io_ring_ctx *io_ring_ctx_alloc(struct io_uring_params *p)
INIT_WQ_LIST(&ctx->locked_free_list);
INIT_DELAYED_WORK(&ctx->fallback_work, io_fallback_req_func);
INIT_WQ_LIST(&ctx->submit_state.compl_reqs);
+#ifdef CONFIG_NET_RX_BUSY_POLL
+ INIT_LIST_HEAD(&ctx->napi_list);
+#endif
return ctx;
err:
kfree(ctx->dummy_ubuf);
@@ -5399,6 +5407,103 @@ IO_NETOP_FN(send);
IO_NETOP_FN(recv);
#endif /* CONFIG_NET */
+#ifdef CONFIG_NET_RX_BUSY_POLL
+
+#define NAPI_TIMEOUT (60 * SEC_CONVERSION)
+
+struct napi_entry {
+ struct list_head list;
+ unsigned int napi_id;
+ unsigned long timeout;
+};
+
+/*
+ * Add busy poll NAPI ID from sk.
+ */
+static void io_add_napi(struct file *file, struct io_ring_ctx *ctx)
+{
+ unsigned int napi_id;
+ struct socket *sock;
+ struct sock *sk;
+ struct napi_entry *ne;
+
+ if (!net_busy_loop_on())
+ return;
+
+ sock = sock_from_file(file);
+ if (!sock)
+ return;
+
+ sk = sock->sk;
+ if (!sk)
+ return;
+
+ napi_id = READ_ONCE(sk->sk_napi_id);
+
+ /* Non-NAPI IDs can be rejected */
+ if (napi_id < MIN_NAPI_ID)
+ return;
+
+ list_for_each_entry(ne, &ctx->napi_list, list) {
+ if (ne->napi_id == napi_id) {
+ ne->timeout = jiffies + NAPI_TIMEOUT;
+ return;
+ }
+ }
+
+ ne = kmalloc(sizeof(*ne), GFP_KERNEL);
+ if (!ne)
+ return;
+
+ ne->napi_id = napi_id;
+ ne->timeout = jiffies + NAPI_TIMEOUT;
+ list_add_tail(&ne->list, &ctx->napi_list);
+}
+
+static inline void io_check_napi_entry_timeout(struct napi_entry *ne)
+{
+ if (time_after(jiffies, ne->timeout)) {
+ list_del(&ne->list);
+ kfree(ne);
+ }
+}
+
+/*
+ * Busy poll if globally on and supporting sockets found
+ */
+static bool io_napi_busy_loop(struct list_head *napi_list)
+{
+ struct napi_entry *ne, *n;
+
+ list_for_each_entry_safe(ne, n, napi_list, list) {
+ napi_busy_loop(ne->napi_id, NULL, NULL, true,
+ BUSY_POLL_BUDGET);
+ io_check_napi_entry_timeout(ne);
+ }
+ return !list_empty(napi_list);
+}
+
+static void io_free_napi_list(struct io_ring_ctx *ctx)
+{
+ while (!list_empty(&ctx->napi_list)) {
+ struct napi_entry *ne =
+ list_first_entry(&ctx->napi_list, struct napi_entry,
+ list);
+
+ list_del(&ne->list);
+ kfree(ne);
+ }
+}
+#else
+static inline void io_add_napi(struct file *file, struct io_ring_ctx *ctx)
+{
+}
+
+static inline void io_free_napi_list(struct io_ring_ctx *ctx)
+{
+}
+#endif /* CONFIG_NET_RX_BUSY_POLL */
+
struct io_poll_table {
struct poll_table_struct pt;
struct io_kiocb *req;
@@ -5777,6 +5882,7 @@ static int __io_arm_poll_handler(struct io_kiocb *req,
__io_poll_execute(req, mask);
return 0;
}
+ io_add_napi(req->file, req->ctx);
/*
* Release ownership. If someone tried to queue a tw while it was
@@ -7519,7 +7625,11 @@ static int __io_sq_thread(struct io_ring_ctx *ctx, bool cap_entries)
!(ctx->flags & IORING_SETUP_R_DISABLED))
ret = io_submit_sqes(ctx, to_submit);
mutex_unlock(&ctx->uring_lock);
-
+#ifdef CONFIG_NET_RX_BUSY_POLL
+ if (!list_empty(&ctx->napi_list) &&
+ io_napi_busy_loop(&ctx->napi_list))
+ ++ret;
+#endif
if (to_submit && wq_has_sleeper(&ctx->sqo_sq_wait))
wake_up(&ctx->sqo_sq_wait);
if (creds)
@@ -7650,6 +7760,9 @@ struct io_wait_queue {
struct io_ring_ctx *ctx;
unsigned cq_tail;
unsigned nr_timeouts;
+#ifdef CONFIG_NET_RX_BUSY_POLL
+ unsigned busy_poll_to;
+#endif
};
static inline bool io_should_wake(struct io_wait_queue *iowq)
@@ -7711,6 +7824,87 @@ static inline int io_cqring_wait_schedule(struct io_ring_ctx *ctx,
return 1;
}
+#ifdef CONFIG_NET_RX_BUSY_POLL
+static void io_adjust_busy_loop_timeout(struct timespec64 *ts,
+ struct io_wait_queue *iowq)
+{
+ unsigned busy_poll_to = READ_ONCE(sysctl_net_busy_poll);
+ struct timespec64 pollto = ns_to_timespec64(1000 * (s64)busy_poll_to);
+
+ if (timespec64_compare(ts, &pollto) > 0) {
+ *ts = timespec64_sub(*ts, pollto);
+ iowq->busy_poll_to = busy_poll_to;
+ } else {
+ s64 to = timespec64_to_ns(ts);
+
+ do_div(to, 1000);
+ iowq->busy_poll_to = to;
+ ts->tv_sec = 0;
+ ts->tv_nsec = 0;
+ }
+}
+
+static inline bool io_busy_loop_timeout(unsigned long start_time,
+ unsigned long bp_usec)
+{
+ if (bp_usec) {
+ unsigned long end_time = start_time + bp_usec;
+ unsigned long now = busy_loop_current_time();
+
+ return time_after(now, end_time);
+ }
+ return true;
+}
+
+static bool io_busy_loop_end(void *p, unsigned long start_time)
+{
+ struct io_wait_queue *iowq = p;
+
+ return signal_pending(current) ||
+ io_should_wake(iowq) ||
+ io_busy_loop_timeout(start_time, iowq->busy_poll_to);
+}
+
+static void io_blocking_napi_busy_loop(struct list_head *napi_list,
+ struct io_wait_queue *iowq)
+{
+ unsigned long start_time =
+ list_is_singular(napi_list) ? 0 :
+ busy_loop_current_time();
+
+ do {
+ if (list_is_singular(napi_list)) {
+ struct napi_entry *ne =
+ list_first_entry(napi_list,
+ struct napi_entry, list);
+
+ napi_busy_loop(ne->napi_id, io_busy_loop_end, iowq,
+ true, BUSY_POLL_BUDGET);
+ io_check_napi_entry_timeout(ne);
+ break;
+ }
+ } while (io_napi_busy_loop(napi_list) &&
+ !io_busy_loop_end(iowq, start_time));
+}
+
+static void io_putback_napi_list(struct io_ring_ctx *ctx,
+ struct list_head *napi_list)
+{
+ struct napi_entry *cne, *lne;
+
+ mutex_lock(&ctx->uring_lock);
+ list_for_each_entry(cne, &ctx->napi_list, list)
+ list_for_each_entry(lne, napi_list, list)
+ if (cne->napi_id == lne->napi_id) {
+ list_del(&lne->list);
+ kfree(lne);
+ break;
+ }
+ list_splice(napi_list, &ctx->napi_list);
+ mutex_unlock(&ctx->uring_lock);
+}
+#endif /* CONFIG_NET_RX_BUSY_POLL */
+
/*
* Wait until events become available, if we don't already have some. The
* application must reap them itself, as they reside on the shared cq ring.
@@ -7723,6 +7917,9 @@ static int io_cqring_wait(struct io_ring_ctx *ctx, int min_events,
struct io_rings *rings = ctx->rings;
ktime_t timeout = KTIME_MAX;
int ret;
+#ifdef CONFIG_NET_RX_BUSY_POLL
+ LIST_HEAD(local_napi_list);
+#endif
do {
io_cqring_overflow_flush(ctx);
@@ -7732,14 +7929,6 @@ static int io_cqring_wait(struct io_ring_ctx *ctx, int min_events,
break;
} while (1);
- if (uts) {
- struct timespec64 ts;
-
- if (get_timespec64(&ts, uts))
- return -EFAULT;
- timeout = ktime_add_ns(timespec64_to_ktime(ts), ktime_get_ns());
- }
-
if (sig) {
#ifdef CONFIG_COMPAT
if (in_compat_syscall())
@@ -7753,6 +7942,30 @@ static int io_cqring_wait(struct io_ring_ctx *ctx, int min_events,
return ret;
}
+#ifdef CONFIG_NET_RX_BUSY_POLL
+ iowq.busy_poll_to = 0;
+ if (!(ctx->flags & IORING_SETUP_SQPOLL)) {
+ mutex_lock(&ctx->uring_lock);
+ list_splice_init(&ctx->napi_list, &local_napi_list);
+ mutex_unlock(&ctx->uring_lock);
+ }
+#endif
+ if (uts) {
+ struct timespec64 ts;
+
+ if (get_timespec64(&ts, uts))
+ return -EFAULT;
+#ifdef CONFIG_NET_RX_BUSY_POLL
+ if (!list_empty(&local_napi_list))
+ io_adjust_busy_loop_timeout(&ts, &iowq);
+#endif
+ timeout = ktime_add_ns(timespec64_to_ktime(ts), ktime_get_ns());
+ }
+#ifdef CONFIG_NET_RX_BUSY_POLL
+ else if (!list_empty(&local_napi_list))
+ iowq.busy_poll_to = READ_ONCE(sysctl_net_busy_poll);
+#endif
+
init_waitqueue_func_entry(&iowq.wq, io_wake_function);
iowq.wq.private = current;
INIT_LIST_HEAD(&iowq.wq.entry);
@@ -7761,6 +7974,12 @@ static int io_cqring_wait(struct io_ring_ctx *ctx, int min_events,
iowq.cq_tail = READ_ONCE(ctx->rings->cq.head) + min_events;
trace_io_uring_cqring_wait(ctx, min_events);
+#ifdef CONFIG_NET_RX_BUSY_POLL
+ if (iowq.busy_poll_to)
+ io_blocking_napi_busy_loop(&local_napi_list, &iowq);
+ if (!list_empty(&local_napi_list))
+ io_putback_napi_list(ctx, &local_napi_list);
+#endif
do {
/* if we can't even flush overflow, don't wait for more */
if (!io_cqring_overflow_flush(ctx)) {
@@ -9450,6 +9669,7 @@ static __cold void io_ring_ctx_free(struct io_ring_ctx *ctx)
__io_sqe_files_unregister(ctx);
if (ctx->rings)
__io_cqring_overflow_flush(ctx, true);
+ io_free_napi_list(ctx);
mutex_unlock(&ctx->uring_lock);
io_eventfd_unregister(ctx);
io_destroy_buffers(ctx);
--
2.35.1
^ permalink raw reply related [flat|nested] 2+ messages in thread
* Re: [PATCH v2] io_uring: Add support for napi_busy_poll
2022-02-27 17:20 [PATCH v2] io_uring: Add support for napi_busy_poll Olivier Langlois
@ 2022-02-27 20:21 ` kernel test robot
0 siblings, 0 replies; 2+ messages in thread
From: kernel test robot @ 2022-02-27 20:21 UTC (permalink / raw)
To: Olivier Langlois, Jens Axboe
Cc: llvm, kbuild-all, Pavel Begunkov, Hao Xu, io-uring, linux-kernel,
gwml
Hi Olivier,
Thank you for the patch! Perhaps something to improve:
[auto build test WARNING on linus/master]
[cannot apply to v5.17-rc5 next-20220225]
[If your patch is applied to the wrong git tree, kindly drop us a note.
And when submitting patch, we suggest to use '--base' as documented in
https://git-scm.com/docs/git-format-patch]
url: https://github.com/0day-ci/linux/commits/Olivier-Langlois/io_uring-Add-support-for-napi_busy_poll/20220228-012140
base: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git 2293be58d6a18cab800e25e42081bacb75c05752
config: mips-qi_lb60_defconfig (https://download.01.org/0day-ci/archive/20220228/[email protected]/config)
compiler: clang version 15.0.0 (https://github.com/llvm/llvm-project d271fc04d5b97b12e6b797c6067d3c96a8d7470e)
reproduce (this is a W=1 build):
wget https://raw.githubusercontent.com/intel/lkp-tests/master/sbin/make.cross -O ~/bin/make.cross
chmod +x ~/bin/make.cross
# install mips cross compiling tool for clang build
# apt-get install binutils-mips-linux-gnu
# https://github.com/0day-ci/linux/commit/65e72f78c66272f7cf0e87dfeef88f5b79de2d91
git remote add linux-review https://github.com/0day-ci/linux
git fetch --no-tags linux-review Olivier-Langlois/io_uring-Add-support-for-napi_busy_poll/20220228-012140
git checkout 65e72f78c66272f7cf0e87dfeef88f5b79de2d91
# save the config file to linux build tree
mkdir build_dir
COMPILER_INSTALL_PATH=$HOME/0day COMPILER=clang make.cross W=1 O=build_dir ARCH=mips SHELL=/bin/bash
If you fix the issue, kindly add following tag as appropriate
Reported-by: kernel test robot <[email protected]>
All warnings (new ones prefixed by >>):
>> fs/io_uring.c:7840:3: warning: comparison of distinct pointer types ('typeof ((to)) *' (aka 'long long *') and 'uint64_t *' (aka 'unsigned long long *')) [-Wcompare-distinct-pointer-types]
do_div(to, 1000);
^~~~~~~~~~~~~~~~
include/asm-generic/div64.h:222:28: note: expanded from macro 'do_div'
(void)(((typeof((n)) *)0) == ((uint64_t *)0)); \
~~~~~~~~~~~~~~~~~~ ^ ~~~~~~~~~~~~~~~
1 warning generated.
vim +7840 fs/io_uring.c
7826
7827 #ifdef CONFIG_NET_RX_BUSY_POLL
7828 static void io_adjust_busy_loop_timeout(struct timespec64 *ts,
7829 struct io_wait_queue *iowq)
7830 {
7831 unsigned busy_poll_to = READ_ONCE(sysctl_net_busy_poll);
7832 struct timespec64 pollto = ns_to_timespec64(1000 * (s64)busy_poll_to);
7833
7834 if (timespec64_compare(ts, &pollto) > 0) {
7835 *ts = timespec64_sub(*ts, pollto);
7836 iowq->busy_poll_to = busy_poll_to;
7837 } else {
7838 s64 to = timespec64_to_ns(ts);
7839
> 7840 do_div(to, 1000);
7841 iowq->busy_poll_to = to;
7842 ts->tv_sec = 0;
7843 ts->tv_nsec = 0;
7844 }
7845 }
7846
---
0-DAY CI Kernel Test Service, Intel Corporation
https://lists.01.org/hyperkitty/list/[email protected]
^ permalink raw reply [flat|nested] 2+ messages in thread
end of thread, other threads:[~2022-02-27 20:21 UTC | newest]
Thread overview: 2+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2022-02-27 17:20 [PATCH v2] io_uring: Add support for napi_busy_poll Olivier Langlois
2022-02-27 20:21 ` kernel test robot
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox