sched/deadline: Add more reschedule cases to prio_changed_dl()
Valentin Schneider
reviewerRFC PATCH 1/1 sched/deadline: Add more reschedule cases to prio_changed_dl()
I've been tracking down an issue on a ~5.17ish kernel where:
CPUx CPUy
<DL task p0 owns an rtmutex M>
<p0 depletes its runtime, gets throttled>
<rq switches to the idle task>
<DL task p1 blocks on M, boost/replenish p0> <No call to resched_curr() happens here>
[idle task keeps running here until *something*
accidentally sets TIF_NEED_RESCHED]
On that kernel, it is quite easy to trigger using rt-tests's deadline_test
[1] with the test running on isolated CPUs (this reduces the chance of something unrelated setting TIF_NEED_RESCHED on the idle tasks, making the issue even more obvious as the hung task detector chimes in).
I haven't been able to reproduce this using a mainline kernel, even if I revert
2972e3050e35 ("tracing: Make trace_marker{,_raw} stream-like")
which gets rid of the lock involved in the above test, *but* I cannot convince myself the issue isn't there from looking at the code.
Make prio_changed_dl() issue a reschedule if the current task isn't a deadline one. While at it, ensure a reschedule is emitted when a queued-but-not-current task gets boosted with an earlier deadline that current's.
[1]: https://git.kernel.org/pub/scm/utils/rt-tests/rt-tests.git
--- kernel/sched/deadline.c | 45 ++++++++++++++++++++++++++--------------- 1 file changed, 29 insertions(+), 16 deletions(-)
kernel/sched/deadline.c +29 -16
@@ -2663,17 +2663,28 @@ static void switched_to_dl(struct rq *rq, struct task_struct *p)static void prio_changed_dl (struct rq *rq,struct task_struct *p,int oldprio) { -if (task_on_rq_queued (p) ||task_current (rq, p)) { -#ifdef CONFIG_SMP -/* -* This might be too much, but unfortunately -* we don't have the old deadline value, and -* we can't argue if the task is increasing -* or lowering its prio, so... -*/ -if (!rq->dl.overloaded) -deadline_queue_pull_task (rq); +if (!task_on_rq_queued (p)) +return ; + +/* +* We don't know if p has a earlier or later deadline, so let's blindly +* set a (maybe not needed) rescheduling point. +*/ +if (!IS_ENABLED (CONFIG_SMP )) { +resched_curr (rq); +return ; + } +/* +* This might be too much, but unfortunately +* we don't have the old deadline value, and +* we can't argue if the task is increasing +* or lowering its prio, so... +*/ +if (!rq->dl.overloaded) +deadline_queue_pull_task (rq); + +if (task_current (rq, p)) {/* * If we now have a earlier deadline task than p, * then reschedule, provided p is still on this
@@ -2681,14 +2692,16 @@ static void prio_changed_dl(struct rq *rq, struct task_struct *p,*/ if (dl_time_before (rq->dl.earliest_dl.curr, p->dl.deadline))resched_curr (rq); -#else + }else {/* -* Again, we don't know if p has a earlier -* or later deadline, so let's blindly set a -* (maybe not needed) rescheduling point. +* Current may not be deadline in case p was throttled but we +* have just replenished it (e.g. rt_mutex_setprio()). +* +* Otherwise, if p was given an earlier deadline, reschedule. */ -resched_curr (rq); -#endif /* CONFIG_SMP */ +if (!dl_task (rq->curr) || +dl_time_before (p->dl.deadline, rq->curr->dl.deadline)) +resched_curr (rq); } }
Signature · 1 line
2.31.1
Juri Lelli
maintainerHi,
Quoted 23 lines · Valentin Schneider
I've been tracking down an issue on a ~5.17ish kernel where:
CPUx CPUy
<DL task p0 owns an rtmutex M>
<p0 depletes its runtime, gets throttled>
<rq switches to the idle task>
<DL task p1 blocks on M, boost/replenish p0> <No call toresched_curr()happens here>[idle task keeps running here until *something*
accidentally setsTIF_NEED_RESCHED]On that kernel, it is quite easy to trigger using rt-tests's
deadline_test
[1] with the test running on isolated CPUs (this reduces the chance of something unrelated settingTIF_NEED_RESCHEDon the idle tasks, making the issue even more obvious as the hung task detector chimes in).I haven't been able to reproduce this using a mainline kernel, even if I revert
2972e3050e35("tracing: Make trace_marker{,_raw} stream-like")
Quoted 7 lines
which gets rid of the lock involved in the above test, *but* I cannot convince myself the issue isn't there from looking at the code.
Make
prio_changed_dl()issue a reschedule if the current task isn't a deadline one. While at it, ensure a reschedule is emitted when a queued-but-not-current task gets boosted with an earlier deadline that current's.
As discussed offline I agree this needs fixing, but .. :)
Quoted 11 lines
[1]: https://git.kernel.org/pub/scm/utils/rt-tests/rt-tests.git
Signed-off-by Valentin Schneider <[email protected]>--- kernel/sched/deadline.c | 45 ++++++++++++++++++++++++++--------------- 1 file changed, 29 insertions(+), 16 deletions(-)kernel/sched/deadline.c +0
-0
kernel/sched/deadline.c static void switched_to_dl(struct rq *rq, struct task_struct *p) · line 2667
Quoted 5 lines
kernel/sched/deadline.c +0
-2@@ -2663,17 +2663,28 @@ static void switched_to_dl(struct rq *rq, struct task_struct *p)static void prio_changed_dl (struct rq *rq,struct task_struct *p,int oldprio) { -if (task_on_rq_queued (p) ||task_current (rq, p)) { -#ifdef CONFIG_SMP
Doesn't this break UP? Don't think earlierst_dl etc are defined in UP.
Thanks,
Juri
Valentin Schneider
reviewerQuoted 42 lines · Juri Lelli
Hi,
Quoted 23 lines · Valentin Schneider
I've been tracking down an issue on a ~5.17ish kernel where:
CPUx CPUy
<DL task p0 owns an rtmutex M>
<p0 depletes its runtime, gets throttled>
<rq switches to the idle task>
<DL task p1 blocks on M, boost/replenish p0> <No call toresched_curr()happens here>[idle task keeps running here until *something*
accidentally setsTIF_NEED_RESCHED]On that kernel, it is quite easy to trigger using rt-tests's
deadline_test
[1] with the test running on isolated CPUs (this reduces the chance of something unrelated settingTIF_NEED_RESCHEDon the idle tasks, making the issue even more obvious as the hung task detector chimes in).I haven't been able to reproduce this using a mainline kernel, even if I revert
2972e3050e35("tracing: Make trace_marker{,_raw} stream-like")Quoted 7 lines
which gets rid of the lock involved in the above test, *but* I cannot convince myself the issue isn't there from looking at the code.
Make
prio_changed_dl()issue a reschedule if the current task isn't a deadline one. While at it, ensure a reschedule is emitted when a queued-but-not-current task gets boosted with an earlier deadline that current's.As discussed offline I agree this needs fixing, but .. :)
Quoted 6 lines
[1]: https://git.kernel.org/pub/scm/utils/rt-tests/rt-tests.git
Signed-off-by Valentin Schneider <[email protected]>--- kernel/sched/deadline.c | 45 ++++++++++++++++++++++++++--------------- 1 file changed, 29 insertions(+), 16 deletions(-)
Quoted 13 lines
kernel/sched/deadline.c static void switched_to_dl(struct rq *rq, struct task_struct *p) · line 2667
Quoted 10 lines
kernel/sched/deadline.c +0
-2@@ -2663,17 +2663,28 @@ static void switched_to_dl(struct rq *rq, struct task_struct *p)static void prio_changed_dl (struct rq *rq,struct task_struct *p,int oldprio) { -if (task_on_rq_queued (p) ||task_current (rq, p)) { -#ifdef CONFIG_SMP Doesn't this break UP? Don't think
earlierst_dletc are defined in UP.
Indeed, I thought myself clever by getting rid of the ifdefs...
Quoted 2 lines
Thanks,
Juri
kernel test robot
build bot[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…
Hi Valentin,
[FYI, it's a private test report for your RFC patch.]
[auto build test ERROR on tip/sched/core]
[also build test ERROR on tip/master tip/auto-latest linus/master v6.2-rc6 next-20230203] [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#_base_tree_information]
url: https://github.com/intel-lab-lkp/linux/commits/Valentin-Schneider/sched-deadline-Add-more-reschedule-cases-to-prio_changed_dl/20230203-023449 patch link: https://lore.kernel.org/r/20230202182854.3696665-1-vschneid%40redhat.com patch subject: [RFC PATCH v1] sched/deadline: Add more reschedule cases to prio_changed_dl() config: x86_64-randconfig-a001 (https://download.01.org/0day-ci/archive/20230204/[email protected]/config) compiler: clang version 14.0.6 (https://github.com/llvm/llvm-project f28c006a5895fc0e329fe15fead81e37457cb1d1)
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
# https://github.com/intel-lab-lkp/linux/commit/83c5e4c04268bf314814c49022d874719a516dbe
git remote add linux-review https://github.com/intel-lab-lkp/linux
git fetch --no-tags linux-review Valentin-Schneider/sched-deadline-Add-more-reschedule-cases-to-prio_changed_dl/20230203-023449 git checkout 83c5e4c04268bf314814c49022d874719a516dbe
# save the config file
mkdir build_dir && cp config build_dir/.config
COMPILER_INSTALL_PATH=$HOME/0day COMPILER=clang make.cross W=1 O=build_dir ARCH=x86_64 olddefconfig COMPILER_INSTALL_PATH=$HOME/0day COMPILER=clang make.cross W=1 O=build_dir ARCH=x86_64 SHELL=/bin/bash
If you fix the issue, kindly add following tag where applicable | Reported-by: kernel test robot <[email protected]>
All errors (new ones prefixed by >>):
In file included from kernel/sched/build_policy.c:53:
>> kernel/sched/deadline.c:2684:14: error: no member named 'overloaded' in 'struct dl_rq'
if (!rq->dl.overloaded)
~~~~~~ ^
>> kernel/sched/deadline.c:2693:29: error: no member named 'earliest_dl' in 'struct dl_rq'
if (dl_time_before(rq->dl.earliest_dl.curr, p->dl.deadline))
~~~~~~ ^2 errors generated.
vim +2684 kernel/sched/deadline.c
aab03e05e8f7e2 Dario Faggioli 2013-11-28 2658
1baca4ce16b8cc Juri Lelli 2013-11-07 2659 /*
1baca4ce16b8cc Juri Lelli 2013-11-07 2660 * If the scheduling parameters of a -deadline task changed,
1baca4ce16b8cc Juri Lelli 2013-11-07 2661 * a push or pull operation might be needed.
1baca4ce16b8cc Juri Lelli 2013-11-07 2662 */
aab03e05e8f7e2 Dario Faggioli 2013-11-28 2663 static void prio_changed_dl(struct rq *rq, struct task_struct *p,
aab03e05e8f7e2 Dario Faggioli 2013-11-28 2664 int oldprio)
aab03e05e8f7e2 Dario Faggioli 2013-11-28 2665 {
83c5e4c04268bf Valentin Schneider 2023-02-02 2666 if (!task_on_rq_queued(p))
83c5e4c04268bf Valentin Schneider 2023-02-02 2667 return;
83c5e4c04268bf Valentin Schneider 2023-02-02 2668
83c5e4c04268bf Valentin Schneider 2023-02-02 2669 /*
83c5e4c04268bf Valentin Schneider 2023-02-02 2670 * We don't know if p has a earlier or later deadline, so let's blindly
83c5e4c04268bf Valentin Schneider 2023-02-02 2671 * set a (maybe not needed) rescheduling point.
83c5e4c04268bf Valentin Schneider 2023-02-02 2672 */
83c5e4c04268bf Valentin Schneider 2023-02-02 2673 if (!IS_ENABLED(CONFIG_SMP)) {
83c5e4c04268bf Valentin Schneider 2023-02-02 2674 resched_curr(rq);
83c5e4c04268bf Valentin Schneider 2023-02-02 2675 return;
83c5e4c04268bf Valentin Schneider 2023-02-02 2676 }
83c5e4c04268bf Valentin Schneider 2023-02-02 2677
1baca4ce16b8cc Juri Lelli 2013-11-07 2678 /*
1baca4ce16b8cc Juri Lelli 2013-11-07 2679 * This might be too much, but unfortunately
1baca4ce16b8cc Juri Lelli 2013-11-07 2680 * we don't have the old deadline value, and
1baca4ce16b8cc Juri Lelli 2013-11-07 2681 * we can't argue if the task is increasing
1baca4ce16b8cc Juri Lelli 2013-11-07 2682 * or lowering its prio, so...
1baca4ce16b8cc Juri Lelli 2013-11-07 2683 */
1baca4ce16b8cc Juri Lelli 2013-11-07 @2684 if (!rq->dl.overloaded)
02d8ec9456f47b Ingo Molnar 2018-03-03 2685 deadline_queue_pull_task(rq);
1baca4ce16b8cc Juri Lelli 2013-11-07 2686
83c5e4c04268bf Valentin Schneider 2023-02-02 2687 if (task_current(rq, p)) {
1baca4ce16b8cc Juri Lelli 2013-11-07 2688 /*
1baca4ce16b8cc Juri Lelli 2013-11-07 2689 * If we now have a earlier deadline task than p,
1baca4ce16b8cc Juri Lelli 2013-11-07 2690 * then reschedule, provided p is still on this
1baca4ce16b8cc Juri Lelli 2013-11-07 2691 * runqueue.
1baca4ce16b8cc Juri Lelli 2013-11-07 2692 */
9916e214998a4a Peter Zijlstra 2015-06-11 @2693 if (dl_time_before(rq->dl.earliest_dl.curr, p->dl.deadline))
8875125efe8402 Kirill Tkhai 2014-06-29 2694 resched_curr(rq);
83c5e4c04268bf Valentin Schneider 2023-02-02 2695 } else {
1baca4ce16b8cc Juri Lelli 2013-11-07 2696 /*
83c5e4c04268bf Valentin Schneider 2023-02-02 2697 * Current may not be deadline in case p was throttled but we
83c5e4c04268bf Valentin Schneider 2023-02-02 2698 * have just replenished it (e.g. rt_mutex_setprio()).
83c5e4c04268bf Valentin Schneider 2023-02-02 2699 *
83c5e4c04268bf Valentin Schneider 2023-02-02 2700 * Otherwise, if p was given an earlier deadline, reschedule.
1baca4ce16b8cc Juri Lelli 2013-11-07 2701 */
83c5e4c04268bf Valentin Schneider 2023-02-02 2702 if (!dl_task(rq->curr) ||
83c5e4c04268bf Valentin Schneider 2023-02-02 2703 dl_time_before(p->dl.deadline, rq->curr->dl.deadline))
8875125efe8402 Kirill Tkhai 2014-06-29 2704 resched_curr(rq);
801ccdbf018ca5 Peter Zijlstra 2016-02-25 2705 }
aab03e05e8f7e2 Dario Faggioli 2013-11-28 2706 }aab03e05e8f7e2 Dario Faggioli 2013-11-28 2707
Signature · 2 lines
0-DAY CI Kernel Test Service https://github.com/intel/lkp-tests
kernel test robot
build bot[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…
Hi Valentin,
[FYI, it's a private test report for your RFC patch.]
[auto build test ERROR on tip/sched/core]
[also build test ERROR on tip/master tip/auto-latest linus/master v6.2-rc6] [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#_base_tree_information]
url: https://github.com/intel-lab-lkp/linux/commits/Valentin-Schneider/sched-deadline-Add-more-reschedule-cases-to-prio_changed_dl/20230203-023449 patch link: https://lore.kernel.org/r/20230202182854.3696665-1-vschneid%40redhat.com patch subject: [RFC PATCH v1] sched/deadline: Add more reschedule cases to prio_changed_dl() config: arm-randconfig-r046-20230205 (https://download.01.org/0day-ci/archive/20230205/[email protected]/config) compiler: arm-linux-gnueabi-gcc (GCC) 12.1.0
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
# https://github.com/intel-lab-lkp/linux/commit/83c5e4c04268bf314814c49022d874719a516dbe
git remote add linux-review https://github.com/intel-lab-lkp/linux
git fetch --no-tags linux-review Valentin-Schneider/sched-deadline-Add-more-reschedule-cases-to-prio_changed_dl/20230203-023449 git checkout 83c5e4c04268bf314814c49022d874719a516dbe
# save the config file
mkdir build_dir && cp config build_dir/.config
COMPILER_INSTALL_PATH=$HOME/0day COMPILER=gcc-12.1.0 make.cross W=1 O=build_dir ARCH=arm olddefconfig COMPILER_INSTALL_PATH=$HOME/0day COMPILER=gcc-12.1.0 make.cross W=1 O=build_dir ARCH=arm SHELL=/bin/bash
If you fix the issue, kindly add following tag where applicable | Reported-by: kernel test robot <[email protected]>
All errors (new ones prefixed by >>):
In file included from kernel/sched/build_policy.c:53:
kernel/sched/deadline.c: In function 'prio_changed_dl':
>> kernel/sched/deadline.c:2684:20: error: 'struct dl_rq' has no member named 'overloaded'
2684 | if (!rq->dl.overloaded)
| ^
>> kernel/sched/deadline.c:2693:42: error: 'struct dl_rq' has no member named 'earliest_dl'
2693 | if (dl_time_before(rq->dl.earliest_dl.curr, p->dl.deadline))
| ^vim +2684 kernel/sched/deadline.c
aab03e05e8f7e2 Dario Faggioli 2013-11-28 2658
1baca4ce16b8cc Juri Lelli 2013-11-07 2659 /*
1baca4ce16b8cc Juri Lelli 2013-11-07 2660 * If the scheduling parameters of a -deadline task changed,
1baca4ce16b8cc Juri Lelli 2013-11-07 2661 * a push or pull operation might be needed.
1baca4ce16b8cc Juri Lelli 2013-11-07 2662 */
aab03e05e8f7e2 Dario Faggioli 2013-11-28 2663 static void prio_changed_dl(struct rq *rq, struct task_struct *p,
aab03e05e8f7e2 Dario Faggioli 2013-11-28 2664 int oldprio)
aab03e05e8f7e2 Dario Faggioli 2013-11-28 2665 {
83c5e4c04268bf Valentin Schneider 2023-02-02 2666 if (!task_on_rq_queued(p))
83c5e4c04268bf Valentin Schneider 2023-02-02 2667 return;
83c5e4c04268bf Valentin Schneider 2023-02-02 2668
83c5e4c04268bf Valentin Schneider 2023-02-02 2669 /*
83c5e4c04268bf Valentin Schneider 2023-02-02 2670 * We don't know if p has a earlier or later deadline, so let's blindly
83c5e4c04268bf Valentin Schneider 2023-02-02 2671 * set a (maybe not needed) rescheduling point.
83c5e4c04268bf Valentin Schneider 2023-02-02 2672 */
83c5e4c04268bf Valentin Schneider 2023-02-02 2673 if (!IS_ENABLED(CONFIG_SMP)) {
83c5e4c04268bf Valentin Schneider 2023-02-02 2674 resched_curr(rq);
83c5e4c04268bf Valentin Schneider 2023-02-02 2675 return;
83c5e4c04268bf Valentin Schneider 2023-02-02 2676 }
83c5e4c04268bf Valentin Schneider 2023-02-02 2677
1baca4ce16b8cc Juri Lelli 2013-11-07 2678 /*
1baca4ce16b8cc Juri Lelli 2013-11-07 2679 * This might be too much, but unfortunately
1baca4ce16b8cc Juri Lelli 2013-11-07 2680 * we don't have the old deadline value, and
1baca4ce16b8cc Juri Lelli 2013-11-07 2681 * we can't argue if the task is increasing
1baca4ce16b8cc Juri Lelli 2013-11-07 2682 * or lowering its prio, so...
1baca4ce16b8cc Juri Lelli 2013-11-07 2683 */
1baca4ce16b8cc Juri Lelli 2013-11-07 @2684 if (!rq->dl.overloaded)
02d8ec9456f47b Ingo Molnar 2018-03-03 2685 deadline_queue_pull_task(rq);
1baca4ce16b8cc Juri Lelli 2013-11-07 2686
83c5e4c04268bf Valentin Schneider 2023-02-02 2687 if (task_current(rq, p)) {
1baca4ce16b8cc Juri Lelli 2013-11-07 2688 /*
1baca4ce16b8cc Juri Lelli 2013-11-07 2689 * If we now have a earlier deadline task than p,
1baca4ce16b8cc Juri Lelli 2013-11-07 2690 * then reschedule, provided p is still on this
1baca4ce16b8cc Juri Lelli 2013-11-07 2691 * runqueue.
1baca4ce16b8cc Juri Lelli 2013-11-07 2692 */
9916e214998a4a Peter Zijlstra 2015-06-11 @2693 if (dl_time_before(rq->dl.earliest_dl.curr, p->dl.deadline))
8875125efe8402 Kirill Tkhai 2014-06-29 2694 resched_curr(rq);
83c5e4c04268bf Valentin Schneider 2023-02-02 2695 } else {
1baca4ce16b8cc Juri Lelli 2013-11-07 2696 /*
83c5e4c04268bf Valentin Schneider 2023-02-02 2697 * Current may not be deadline in case p was throttled but we
83c5e4c04268bf Valentin Schneider 2023-02-02 2698 * have just replenished it (e.g. rt_mutex_setprio()).
83c5e4c04268bf Valentin Schneider 2023-02-02 2699 *
83c5e4c04268bf Valentin Schneider 2023-02-02 2700 * Otherwise, if p was given an earlier deadline, reschedule.
1baca4ce16b8cc Juri Lelli 2013-11-07 2701 */
83c5e4c04268bf Valentin Schneider 2023-02-02 2702 if (!dl_task(rq->curr) ||
83c5e4c04268bf Valentin Schneider 2023-02-02 2703 dl_time_before(p->dl.deadline, rq->curr->dl.deadline))
8875125efe8402 Kirill Tkhai 2014-06-29 2704 resched_curr(rq);
801ccdbf018ca5 Peter Zijlstra 2016-02-25 2705 }
aab03e05e8f7e2 Dario Faggioli 2013-11-28 2706 }aab03e05e8f7e2 Dario Faggioli 2013-11-28 2707
Signature · 2 lines
0-DAY CI Kernel Test Service https://github.com/intel/lkp-tests