Skip to contentSkip to conversation

psi: do not require setsched permission from the trigger creator

PATCH In review Suren Baghdasaryan 3 participants 8 messages Jul 30, 2019 – Aug 6, 2019

Subsystems (4)
lore
Share & tools
lore

Suren Baghdasaryan

author maintainer
#0 lore raw

PATCH 1/1 psi: do not require setsched permission from the trigger creator

When a process creates a new trigger by writing into /proc/pressure/* files, permissions to write such a file should be used to determine whether the process is allowed to do so or not. Current implementation would also require such a process to have setsched capability. Setting of psi trigger thread's scheduling policy is an implementation detail and should not be exposed to the user level. Remove the permission check by using _nocheck version of the function.

---
 kernel/sched/psi.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
kernel/sched/psi.c +1 -1
index 7acc632c3b82..ed9a1d573cb1 100644
@@ -1061,7 +1061,7 @@ struct psi_trigger *psi_trigger_create(struct psi_group *group,
 			mutex_unlock(&group->trigger_lock);
 			return ERR_CAST(kworker);
 		}
-		sched_setscheduler(kworker->task, SCHED_FIFO, &param);
+		sched_setscheduler_nocheck(kworker->task, SCHED_FIFO, &param);
 		kthread_init_delayed_work(&group->poll_work,
 				psi_poll_work);
 		rcu_assign_pointer(group->poll_kworker, kworker);
Signature · 1 line
2.22.0.709.g102302147b-goog
maintainer
#1 lore raw
Quoted 19 lines · Suren Baghdasaryan

When a process creates a new trigger by writing into /proc/pressure/* files, permissions to write such a file should be used to determine whether the process is allowed to do so or not. Current implementation would also require such a process to have setsched capability. Setting of psi trigger thread's scheduling policy is an implementation detail and should not be exposed to the user level. Remove the permission check by using _nocheck version of the function.

---
 kernel/sched/psi.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
kernel/sched/psi.c +0 -0
index 7acc632c3b82..ed9a1d573cb1 100644

kernel/sched/psi.c struct psi_trigger *psi_trigger_create(struct psi_group *group, · line 1064

Quoted 5 lines
kernel/sched/psi.c +1 -1
@@ -1061,7 +1061,7 @@ struct psi_trigger *psi_trigger_create(struct psi_group *group,
 			mutex_unlock(&group->trigger_lock);
 			return ERR_CAST(kworker);
 		}
-		sched_setscheduler(kworker->task, SCHED_FIFO, &param);
+		sched_setscheduler_nocheck(kworker->task, SCHED_FIFO, &param);

ARGGH, wtf is there a FIFO-99!! thread here at all !?

Quoted 6 lines
kernel/sched/psi.c +0 -0
 		kthread_init_delayed_work(&group->poll_work,
 				psi_poll_work);
 		rcu_assign_pointer(group->poll_kworker, kworker);
Signature · 1 line
2.22.0.709.g102302147b-goog
author maintainer
#2 lore raw
Quoted 16 lines · Peter Zijlstra
Quoted 14 lines · Suren Baghdasaryan

When a process creates a new trigger by writing into /proc/pressure/* files, permissions to write such a file should be used to determine whether the process is allowed to do so or not. Current implementation would also require such a process to have setsched capability. Setting of psi trigger thread's scheduling policy is an implementation detail and should not be exposed to the user level. Remove the permission check by using _nocheck version of the function.

---
 kernel/sched/psi.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
Quoted 12 lines

kernel/sched/psi.c struct psi_trigger *psi_trigger_create(struct psi_group *group, · line 1064

Quoted 10 lines
kernel/sched/psi.c +1 -1
index 7acc632c3b82..ed9a1d573cb1 100644
@@ -1061,7 +1061,7 @@ struct psi_trigger *psi_trigger_create(struct psi_group *group,
                      mutex_unlock(&group->trigger_lock);
                      return ERR_CAST(kworker);
              }
-             sched_setscheduler(kworker->task, SCHED_FIFO, &param);
+             sched_setscheduler_nocheck(kworker->task, SCHED_FIFO, &param);

ARGGH, wtf is there a FIFO-99!! thread here at all !?

We need psi poll_kworker to be an rt-priority thread so that psi notifications are delivered to the userspace without delay even when the CPUs are very congested. Otherwise it's easy to delay psi notifications by running a simple CPU hogger executing "chrt -f 50 dd if=/dev/zero of=/dev/null". Because these notifications are time-critical for reacting to memory shortages we can't allow for such delays.
Notice that this kworker is created only if userspace creates a psi trigger. So unless you are using psi triggers you will never see this kthread created.

Quoted 10 lines
Quoted 6 lines
kernel/sched/psi.c +0 -1
              kthread_init_delayed_work(&group->poll_work,
                              psi_poll_work);
              rcu_assign_pointer(group->poll_kworker, kworker);
--

2.22.0.709.g102302147b-goog

--
To unsubscribe from this group and stop receiving emails from it, send an email to [email protected].

maintainer
#3 lore raw
Quoted 30 lines · Suren Baghdasaryan
Quoted 16 lines · Peter Zijlstra
Quoted 14 lines · Suren Baghdasaryan

When a process creates a new trigger by writing into /proc/pressure/* files, permissions to write such a file should be used to determine whether the process is allowed to do so or not. Current implementation would also require such a process to have setsched capability. Setting of psi trigger thread's scheduling policy is an implementation detail and should not be exposed to the user level. Remove the permission check by using _nocheck version of the function.

---
 kernel/sched/psi.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
Quoted 12 lines

kernel/sched/psi.c struct psi_trigger *psi_trigger_create(struct psi_group *group, · line 1064

Quoted 10 lines
kernel/sched/psi.c +1 -1
index 7acc632c3b82..ed9a1d573cb1 100644
@@ -1061,7 +1061,7 @@ struct psi_trigger *psi_trigger_create(struct psi_group *group,
                      mutex_unlock(&group->trigger_lock);
                      return ERR_CAST(kworker);
              }
-             sched_setscheduler(kworker->task, SCHED_FIFO, &param);
+             sched_setscheduler_nocheck(kworker->task, SCHED_FIFO, &param);

ARGGH, wtf is there a FIFO-99!! thread here at all !?

Quoted 1 line

We need psi poll_kworker to be an rt-priority thread so that psi

There is a giant difference between 'needs to be higher than OTHER' and FIFO-99.

Quoted 4 lines

notifications are delivered to the userspace without delay even when the CPUs are very congested. Otherwise it's easy to delay psi notifications by running a simple CPU hogger executing "chrt -f 50 dd if=/dev/zero of=/dev/null". Because these notifications are

So what; at that point that's exactly what you're asking for. Using RT is for those who know what they're doing.

Quoted 2 lines

time-critical for reacting to memory shortages we can't allow for such delays.

Furthermore, actual RT programs will have pre-allocated and locked any memory they rely on. They don't give a crap about your pressure nonsense.

Quoted 3 lines

Notice that this kworker is created only if userspace creates a psi trigger. So unless you are using psi triggers you will never see this kthread created.

By marking it FIFO-99 you're in effect saying that your stupid statistics gathering is more important than your life. It will preempt the task that's in control of the band-saw emergency break, it will preempt the task that's adjusting the electromagnetic field containing this plasma flow.

That's insane.

I'm going to queue a patch to reduce this to FIFO-1, that will preempt regular OTHER tasks but will not perturb (much) actual RT bits.

author maintainer
#4 lore raw

Hi Peter,
Thanks for sharing your thoughts. I understand your point and I tend to agree with it. I originally designed this using watchdog as the example of a critical system health signal and in the context of mobile device memory pressure is critical but I agree that there are more important things in life. I checked and your proposal to change it to FIFO-1 should still work for our purposes. Will test to make sure and reply to your patch. Couple clarifications in-line.

Quoted 45 lines · Peter Zijlstra
Quoted 30 lines · Suren Baghdasaryan
Quoted 16 lines · Peter Zijlstra
Quoted 14 lines · Suren Baghdasaryan

When a process creates a new trigger by writing into /proc/pressure/* files, permissions to write such a file should be used to determine whether the process is allowed to do so or not. Current implementation would also require such a process to have setsched capability. Setting of psi trigger thread's scheduling policy is an implementation detail and should not be exposed to the user level. Remove the permission check by using _nocheck version of the function.

---
 kernel/sched/psi.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
Quoted 12 lines

kernel/sched/psi.c struct psi_trigger *psi_trigger_create(struct psi_group *group, · line 1064

Quoted 10 lines
kernel/sched/psi.c +1 -1
index 7acc632c3b82..ed9a1d573cb1 100644
@@ -1061,7 +1061,7 @@ struct psi_trigger *psi_trigger_create(struct psi_group *group,
                      mutex_unlock(&group->trigger_lock);
                      return ERR_CAST(kworker);
              }
-             sched_setscheduler(kworker->task, SCHED_FIFO, &param);
+             sched_setscheduler_nocheck(kworker->task, SCHED_FIFO, &param);

ARGGH, wtf is there a FIFO-99!! thread here at all !?

Quoted 1 line

We need psi poll_kworker to be an rt-priority thread so that psi

There is a giant difference between 'needs to be higher than OTHER' and FIFO-99.

Quoted 4 lines

notifications are delivered to the userspace without delay even when the CPUs are very congested. Otherwise it's easy to delay psi notifications by running a simple CPU hogger executing "chrt -f 50 dd if=/dev/zero of=/dev/null". Because these notifications are

So what; at that point that's exactly what you're asking for. Using RT is for those who know what they're doing.

Quoted 7 lines
Quoted 2 lines

time-critical for reacting to memory shortages we can't allow for such delays.

Furthermore, actual RT programs will have pre-allocated and locked any memory they rely on. They don't give a crap about your pressure nonsense.

This signal is used not to protect other RT tasks but to monitor overall system memory health for the sake of system responsiveness.

Quoted 11 lines
Quoted 3 lines

Notice that this kworker is created only if userspace creates a psi trigger. So unless you are using psi triggers you will never see this kthread created.

By marking it FIFO-99 you're in effect saying that your stupid statistics gathering is more important than your life. It will preempt the task that's in control of the band-saw emergency break, it will preempt the task that's adjusting the electromagnetic field containing this plasma flow.

That's insane.

IMHO an opt-in feature stops being "stupid" as soon as the user opted in to use it, therefore explicitly indicating interest in it. However I assume you are using "stupid" here to indicate that it's "less important" rather than it's "useless".

Quoted 3 lines

I'm going to queue a patch to reduce this to FIFO-1, that will preempt regular OTHER tasks but will not perturb (much) actual RT bits.

Thanks for posting the fix.

Quoted 3 lines

--
To unsubscribe from this group and stop receiving emails from it, send an email to [email protected].

maintainer
#5 lore raw
Quoted 6 lines · Suren Baghdasaryan
Quoted 6 lines

By marking it FIFO-99 you're in effect saying that your stupid statistics gathering is more important than your life. It will preempt the task that's in control of the band-saw emergency break, it will preempt the task that's adjusting the electromagnetic field containing this plasma flow.

Quoted 6 lines
Quoted 1 line

That's insane.

IMHO an opt-in feature stops being "stupid" as soon as the user opted in to use it, therefore explicitly indicating interest in it. However I assume you are using "stupid" here to indicate that it's "less important" rather than it's "useless".

Quite; PSI does have its uses. RT just isn't one of them.

author maintainer
#6 lore raw
Quoted 8 lines · Peter Zijlstra
Quoted 6 lines · Suren Baghdasaryan
Quoted 6 lines

By marking it FIFO-99 you're in effect saying that your stupid statistics gathering is more important than your life. It will preempt the task that's in control of the band-saw emergency break, it will preempt the task that's adjusting the electromagnetic field containing this plasma flow.

Quoted 8 lines
Quoted 6 lines
Quoted 1 line

That's insane.

IMHO an opt-in feature stops being "stupid" as soon as the user opted in to use it, therefore explicitly indicating interest in it. However I assume you are using "stupid" here to indicate that it's "less important" rather than it's "useless".

Quite; PSI does have its uses. RT just isn't one of them.

Sorry about messing it up in the first place.
If you don't see any issues with my patch replacing sched_setscheduler() with sched_setscheduler_nocheck(), would you mind taking it too? I applied it over your patch onto Linus' ToT with no merge conflicts.
Thanks,
Suren.

Quoted 3 lines

--
To unsubscribe from this group and stop receiving emails from it, send an email to [email protected].

#7 lore raw

[tip:sched/urgent] sched/psi: Do not require setsched permission from the trigger creator

In reply to Suren Baghdasaryan
sched/psi: Do not require setsched permission from the trigger creator
Commit-ID:  04e048cf09d7b5fc995817cdc5ae1acd4482429c
Gitweb:     https://git.kernel.org/tip/04e048cf09d7b5fc995817cdc5ae1acd4482429c
Author:     Suren Baghdasaryan <[email protected]>
AuthorDate: Mon, 29 Jul 2019 18:33:10 -0700
Committer:  Peter Zijlstra <[email protected]>
CommitDate: Tue, 6 Aug 2019 12:49:18 +0200

sched/psi: Do not require setsched permission from the trigger creator

When a process creates a new trigger by writing into /proc/pressure/* files, permissions to write such a file should be used to determine whether the process is allowed to do so or not. Current implementation would also require such a process to have setsched capability. Setting of psi trigger thread's scheduling policy is an implementation detail and should not be exposed to the user level. Remove the permission check by using _nocheck version of the function.

---
 kernel/sched/psi.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
kernel/sched/psi.c +1 -1
index 7fe2c5fd26b5..23fbbcc414d5 100644
@@ -1061,7 +1061,7 @@ struct psi_trigger *psi_trigger_create(struct psi_group *group,
 			mutex_unlock(&group->trigger_lock);
 			return ERR_CAST(kworker);
 		}
-		sched_setscheduler(kworker->task, SCHED_FIFO, &param);
+		sched_setscheduler_nocheck(kworker->task, SCHED_FIFO, &param);
 		kthread_init_delayed_work(&group->poll_work,
 				psi_poll_work);
 		rcu_assign_pointer(group->poll_kworker, kworker);