FAILED: patch "[PATCH] io_uring/sqpoll: be smarter on when to update the stime usage" failed to apply to 6.12-stable tree
Subsystems (2)
[email protected]
maintainerThe patch below does not apply to the 6.12-stable tree.
If someone wants it applied there, or to any other stable or longterm tree, then please email the backport, including the original git commit id to <[email protected]>.
To reproduce the conflict and resubmit, you may use the following commands:
git fetch https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/ linux-6.12.y
git checkout FETCH_HEAD
git cherry-pick -x a94e0657269c5b8e1a90b17aa2c048b3d276e16d
# <resolve conflicts, build, test, etc.>
git commit -s
git send-email --to '<[email protected]>' --in-reply-to '2025102618-plastic-eldercare-f957@gregkh' --subject-prefix 'PATCH 6.12.y' HEAD^..
Possible dependencies:
thanks,
greg k-h
------------------ original commit in Linus's tree ------------------
From a94e0657269c5b8e1a90b17aa2c048b3d276e16d Mon Sep 17 00:00:00 2001
From: Jens Axboe <[email protected]>
Date: Tue, 21 Oct 2025 11:44:39 -0600
Subject: [PATCH] io_uring/sqpoll: be smarter on when to update the stime usage
The current approach is a bit naive, and hence calls the time querying way too often. Only start the "doing work" timer when there's actual work to do, and then use that information to terminate (and account) the work time once done. This greatly reduces the frequency of these calls, when they cannot have changed anyway.
Running a basic random reader that is setup to use SQPOLL, a profile before this change shows these as the top cycle consumers:
+ 32.60% iou-sqp-1074 [kernel.kallsyms] [k] thread_group_cputime_adjusted + 19.97% iou-sqp-1074 [kernel.kallsyms] [k] thread_group_cputime + 12.20% io_uring io_uring [.] submitter_uring_fn + 4.13% iou-sqp-1074 [kernel.kallsyms] [k] getrusage + 2.45% iou-sqp-1074 [kernel.kallsyms] [k] io_submit_sqes + 2.18% iou-sqp-1074 [kernel.kallsyms] [k] __pi_memset_generic + 2.09% iou-sqp-1074 [kernel.kallsyms] [k] cputime_adjust
and after this change, top of profile looks as follows:
+ 36.23% io_uring io_uring [.] submitter_uring_fn
+ 23.26% iou-sqp-819 [kernel.kallsyms] [k] io_sq_thread
+ 10.14% iou-sqp-819 [kernel.kallsyms] [k] io_sq_tw
+ 6.52% iou-sqp-819 [kernel.kallsyms] [k] tctx_task_work_run
+ 4.82% iou-sqp-819 [kernel.kallsyms] [k] nvme_submit_cmds.part.0
+ 2.91% iou-sqp-819 [kernel.kallsyms] [k] io_submit_sqes
[...]
0.02% iou-sqp-819 [kernel.kallsyms] [k] cputime_adjustwhere it's spending the cycles on things that actually matter.
io_uring/sqpoll.c +32 -11
@@ -170,6 +170,11 @@ static inline bool io_sqd_events_pending(struct io_sq_data *sqd)return READ_ONCE (sqd->state); } +struct io_sq_time { +bool started; +u64 usec; +}; +u64 io_sq_cpu_usec (struct task_struct *tsk) {u64 utime, stime;
@@ -179,12 +184,24 @@ u64 io_sq_cpu_usec(struct task_struct *tsk)return stime; } -static void io_sq_update_worktime (struct io_sq_data *sqd,u64 usec) +static void io_sq_update_worktime (struct io_sq_data *sqd,struct io_sq_time *ist) { - sqd->work_time +=io_sq_cpu_usec (current) - usec; +if (!ist->started) +return ; + ist->started = false; + sqd->work_time +=io_sq_cpu_usec (current) - ist->usec; } -static int __io_sq_thread (struct io_ring_ctx *ctx,bool cap_entries) +static void io_sq_start_worktime (struct io_sq_time *ist) +{ +if (ist->started) +return ; + ist->started = true; + ist->usec =io_sq_cpu_usec (current); +} + +static int __io_sq_thread (struct io_ring_ctx *ctx,struct io_sq_data *sqd, +bool cap_entries,struct io_sq_time *ist) {unsigned int to_submit;int ret =0 ;
@@ -197,6 +214,8 @@ static int __io_sq_thread(struct io_ring_ctx *ctx, bool cap_entries)if (to_submit || !wq_list_empty (&ctx->iopoll_list)) {const struct cred *creds =NULL ; +io_sq_start_worktime (ist); +if (ctx->sq_creds !=current_cred ()) creds =override_creds (ctx->sq_creds);
@@ -278,7 +297,6 @@ static int io_sq_thread(void *data)unsigned long timeout =0 ;char buf[TASK_COMM_LEN ] = {};DEFINE_WAIT (wait); -u64 start;/* offload context creation failed, just exit */ if (!current->io_uring) {
@@ -313,6 +331,7 @@ static int io_sq_thread(void *data)mutex_lock (&sqd->lock);while (1 ) {bool cap_entries, sqt_spin = false; +struct io_sq_time ist = { };if (io_sqd_events_pending (sqd) ||signal_pending (current)) {if (io_sqd_handle_event (sqd))
@@ -321,9 +340,8 @@ static int io_sq_thread(void *data) } cap_entries = !list_is_singular (&sqd->ctx_list); - start =io_sq_cpu_usec (current);list_for_each_entry (ctx, &sqd->ctx_list, sqd_list) { -int ret =__io_sq_thread (ctx, cap_entries); +int ret =__io_sq_thread (ctx, sqd, cap_entries, &ist);if (!sqt_spin && (ret >0 || !wq_list_empty (&ctx->iopoll_list))) sqt_spin = true;
@@ -331,15 +349,18 @@ static int io_sq_thread(void *data)if (io_sq_tw (&retry_list,IORING_TW_CAP_ENTRIES_VALUE )) sqt_spin = true; -list_for_each_entry (ctx, &sqd->ctx_list, sqd_list) -if (io_napi (ctx)) +list_for_each_entry (ctx, &sqd->ctx_list, sqd_list) { +if (io_napi (ctx)) { +io_sq_start_worktime (&ist);io_napi_sqpoll_busy_poll (ctx); + } + } + +io_sq_update_worktime (sqd, &ist);if (sqt_spin || !time_after (jiffies, timeout)) { -if (sqt_spin) { -io_sq_update_worktime (sqd, start); +if (sqt_spin) timeout = jiffies + sqd->sq_thread_idle; - }if (unlikely (need_resched ())) {mutex_unlock (&sqd->lock);cond_resched ();
Jens Axboe
maintainerQuoted 6 lines · On 10/26/25 8:22 AM, [email protected] wrote:
The patch below does not apply to the 6.12-stable tree.
If someone wants it applied there, or to any other stable or longterm tree, then please email the backport, including the original git commit id to <[email protected]>.
Quoted 8 lines
To reproduce the conflict and resubmit, you may use the following commands:
git fetch https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/ linux-6.12.y
git checkoutFETCH_HEAD
git cherry-pick -xa94e0657269c5b8e1a90b17aa2c048b3d276e16d
# <resolve conflicts, build, test, etc.>
git commit -s
git send-email --to '<[email protected]>' --in-reply-to '2025102618-plastic-eldercare-f957@gregkh' --subject-prefix 'PATCH 6.12.y' HEAD^..
And here's this one, just a single fuzzed commit after patch 1 got applied.
Signature · 1 line
Jens Axboe
From bfe554574c418c0ef57cd234bed1bf31e9bb4f00 Mon Sep 17 00:00:00 2001
From: Jens Axboe <[email protected]>
Date: Tue, 21 Oct 2025 11:44:39 -0600
Subject: [PATCH 2/2] io_uring/sqpoll: be smarter on when to update the stime
usage
Commit a94e0657269c5b8e1a90b17aa2c048b3d276e16d upstream.
The current approach is a bit naive, and hence calls the time querying way too often. Only start the "doing work" timer when there's actual work to do, and then use that information to terminate (and account) the work time once done. This greatly reduces the frequency of these calls, when they cannot have changed anyway.
Running a basic random reader that is setup to use SQPOLL, a profile before this change shows these as the top cycle consumers:
+ 32.60% iou-sqp-1074 [kernel.kallsyms] [k] thread_group_cputime_adjusted + 19.97% iou-sqp-1074 [kernel.kallsyms] [k] thread_group_cputime + 12.20% io_uring io_uring [.] submitter_uring_fn + 4.13% iou-sqp-1074 [kernel.kallsyms] [k] getrusage + 2.45% iou-sqp-1074 [kernel.kallsyms] [k] io_submit_sqes + 2.18% iou-sqp-1074 [kernel.kallsyms] [k] __pi_memset_generic + 2.09% iou-sqp-1074 [kernel.kallsyms] [k] cputime_adjust
and after this change, top of profile looks as follows:
+ 36.23% io_uring io_uring [.] submitter_uring_fn
+ 23.26% iou-sqp-819 [kernel.kallsyms] [k] io_sq_thread
+ 10.14% iou-sqp-819 [kernel.kallsyms] [k] io_sq_tw
+ 6.52% iou-sqp-819 [kernel.kallsyms] [k] tctx_task_work_run
+ 4.82% iou-sqp-819 [kernel.kallsyms] [k] nvme_submit_cmds.part.0
+ 2.91% iou-sqp-819 [kernel.kallsyms] [k] io_submit_sqes
[...]
0.02% iou-sqp-819 [kernel.kallsyms] [k] cputime_adjustwhere it's spending the cycles on things that actually matter.
--- io_uring/sqpoll.c | 43 ++++++++++++++++++++++++++++++++----------- 1 file changed, 32 insertions(+), 11 deletions(-)
io_uring/sqpoll.c +32 -11
@@ -176,6 +176,11 @@ static inline bool io_sqd_events_pending(struct io_sq_data *sqd)return READ_ONCE (sqd->state); } +struct io_sq_time { +bool started; +u64 usec; +}; +u64 io_sq_cpu_usec (struct task_struct *tsk) {u64 utime, stime;
@@ -185,12 +190,24 @@ u64 io_sq_cpu_usec(struct task_struct *tsk)return stime; } -static void io_sq_update_worktime (struct io_sq_data *sqd,u64 usec) +static void io_sq_update_worktime (struct io_sq_data *sqd,struct io_sq_time *ist) +{ +if (!ist->started) +return ; + ist->started = false; + sqd->work_time +=io_sq_cpu_usec (current) - ist->usec; +} + +static void io_sq_start_worktime (struct io_sq_time *ist) { - sqd->work_time +=io_sq_cpu_usec (current) - usec; +if (ist->started) +return ; + ist->started = true; + ist->usec =io_sq_cpu_usec (current); } -static int __io_sq_thread (struct io_ring_ctx *ctx,bool cap_entries) +static int __io_sq_thread (struct io_ring_ctx *ctx,struct io_sq_data *sqd, +bool cap_entries,struct io_sq_time *ist) {unsigned int to_submit;int ret =0 ;
@@ -203,6 +220,8 @@ static int __io_sq_thread(struct io_ring_ctx *ctx, bool cap_entries)if (to_submit || !wq_list_empty (&ctx->iopoll_list)) {const struct cred *creds =NULL ; +io_sq_start_worktime (ist); +if (ctx->sq_creds !=current_cred ()) creds =override_creds (ctx->sq_creds);
@@ -284,7 +303,6 @@ static int io_sq_thread(void *data)unsigned long timeout =0 ;char buf[TASK_COMM_LEN ];DEFINE_WAIT (wait); -u64 start;/* offload context creation failed, just exit */ if (!current->io_uring) {
@@ -319,6 +337,7 @@ static int io_sq_thread(void *data)mutex_lock (&sqd->lock);while (1 ) {bool cap_entries, sqt_spin = false; +struct io_sq_time ist = { };if (io_sqd_events_pending (sqd) ||signal_pending (current)) {if (io_sqd_handle_event (sqd))
@@ -327,9 +346,8 @@ static int io_sq_thread(void *data) } cap_entries = !list_is_singular (&sqd->ctx_list); - start =io_sq_cpu_usec (current);list_for_each_entry (ctx, &sqd->ctx_list, sqd_list) { -int ret =__io_sq_thread (ctx, cap_entries); +int ret =__io_sq_thread (ctx, sqd, cap_entries, &ist);if (!sqt_spin && (ret >0 || !wq_list_empty (&ctx->iopoll_list))) sqt_spin = true;
@@ -337,15 +355,18 @@ static int io_sq_thread(void *data)if (io_sq_tw (&retry_list,IORING_TW_CAP_ENTRIES_VALUE )) sqt_spin = true; -list_for_each_entry (ctx, &sqd->ctx_list, sqd_list) -if (io_napi (ctx)) +list_for_each_entry (ctx, &sqd->ctx_list, sqd_list) { +if (io_napi (ctx)) { +io_sq_start_worktime (&ist);io_napi_sqpoll_busy_poll (ctx); + } + } + +io_sq_update_worktime (sqd, &ist);if (sqt_spin || !time_after (jiffies, timeout)) { -if (sqt_spin) { -io_sq_update_worktime (sqd, start); +if (sqt_spin) timeout = jiffies + sqd->sq_thread_idle; - }if (unlikely (need_resched ())) {mutex_unlock (&sqd->lock);cond_resched ();
Signature · 1 line
2.51.0
Greg KH
maintainerQuoted 16 lines · Jens Axboe
Quoted 6 lines · On 10/26/25 8:22 AM, [email protected] wrote:
The patch below does not apply to the 6.12-stable tree.
If someone wants it applied there, or to any other stable or longterm tree, then please email the backport, including the original git commit id to <[email protected]>.Quoted 8 lines
To reproduce the conflict and resubmit, you may use the following commands:
git fetch https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/ linux-6.12.y
git checkoutFETCH_HEAD
git cherry-pick -xa94e0657269c5b8e1a90b17aa2c048b3d276e16d
# <resolve conflicts, build, test, etc.>
git commit -s
git send-email --to '<[email protected]>' --in-reply-to '2025102618-plastic-eldercare-f957@gregkh' --subject-prefix 'PATCH 6.12.y' HEAD^..
Quoted 2 lines
And here's this one, just a single fuzzed commit after patch 1 got applied.
Both now queued up, thanks.
greg k-h