Skip to contentSkip to conversation

sched/fair: Fix wrong cpu selecting from isolated domain

PATCH RESEND Applied Xunlei Pang 3 participants 4 messages Sep 24, 2020 – Sep 29, 2020 Scheduler

lore
Share & tools
lore
#0 lore raw

PATCH RESEND 1/1 sched/fair: Fix wrong cpu selecting from isolated domain

We've met problems that occasionally tasks with full cpumask (e.g. by putting it into a cpuset or setting to full affinity) were migrated to our isolated cpus in production environment.

After some analysis, we found that it is due to the current select_idle_smt() not considering the sched_domain mask.

Steps to reproduce on my 31-CPU hyperthreads machine:
1. with boot parameter: "isolcpus=domain,2-31" (thread lists: 0,16 and 1,17)
2. cgcreate -g cpu:test; cgexec -g cpu:test "test_threads"
3. some threads will be migrated to the isolated cpu16~17.

Fix it by checking the valid domain mask in select_idle_smt().

Fixes 10e2f1acd010 ("sched/core: Rewrite and improve select_idle_siblings())
---
 kernel/sched/fair.c | 9 +++++----
 1 file changed, 5 insertions(+), 4 deletions(-)
kernel/sched/fair.c +5 -4
index 1a68a05..fa942c4 100644
@@ -6075,7 +6075,7 @@ static int select_idle_core(struct task_struct *p, struct sched_domain *sd, int
 /*
  * Scan the local SMT mask for idle CPUs.
  */
-static int select_idle_smt(struct task_struct *p, int target)
+static int select_idle_smt(struct task_struct *p, struct sched_domain *sd, int target)
 {
 	int cpu;
 
@@ -6083,7 +6083,8 @@ static int select_idle_smt(struct task_struct *p, int target)
 		return -1;
 
 	for_each_cpu(cpu, cpu_smt_mask(target)) {
-		if (!cpumask_test_cpu(cpu, p->cpus_ptr))
+		if (!cpumask_test_cpu(cpu, p->cpus_ptr) ||
+		    !cpumask_test_cpu(cpu, sched_domain_span(sd)))
 			continue;
 		if (available_idle_cpu(cpu) || sched_idle_cpu(cpu))
 			return cpu;
@@ -6099,7 +6100,7 @@ static inline int select_idle_core(struct task_struct *p, struct sched_domain *s
 	return -1;
 }
 
-static inline int select_idle_smt(struct task_struct *p, int target)
+static inline int select_idle_smt(struct task_struct *p, struct sched_domain *sd, int target)
 {
 	return -1;
 }
@@ -6274,7 +6275,7 @@ static int select_idle_sibling(struct task_struct *p, int prev, int target)
 	if ((unsigned)i < nr_cpumask_bits)
 		return i;
 
-	i = select_idle_smt(p, target);
+	i = select_idle_smt(p, sd, target);
 	if ((unsigned)i < nr_cpumask_bits)
 		return i;
 
Signature · 1 line
1.8.3.1
maintainer Reviewed-by on v1 1/1
#1 lore raw
Quoted 14 lines · Xunlei Pang

We've met problems that occasionally tasks with full cpumask (e.g. by putting it into a cpuset or setting to full affinity) were migrated to our isolated cpus in production environment.

After some analysis, we found that it is due to the current select_idle_smt() not considering the sched_domain mask.

Steps to reproduce on my 31-CPU hyperthreads machine:
1. with boot parameter: "isolcpus=domain,2-31" (thread lists: 0,16 and 1,17)
2. cgcreate -g cpu:test; cgexec -g cpu:test "test_threads"
3. some threads will be migrated to the isolated cpu16~17.

Quoted 6 lines

Fix it by checking the valid domain mask in select_idle_smt().

Fixes 10e2f1acd010 ("sched/core: Rewrite and improve select_idle_siblings())
Quoted 48 lines
---
 kernel/sched/fair.c | 9 +++++----
 1 file changed, 5 insertions(+), 4 deletions(-)
kernel/sched/fair.c +5 -5
index 1a68a05..fa942c4 100644
@@ -6075,7 +6075,7 @@ static int select_idle_core(struct task_struct *p, struct sched_domain *sd, int
 /*
  * Scan the local SMT mask for idle CPUs.
  */
-static int select_idle_smt(struct task_struct *p, int target)
+static int select_idle_smt(struct task_struct *p, struct sched_domain *sd, int target)
 {
        int cpu;
 
@@ -6083,7 +6083,8 @@ static int select_idle_smt(struct task_struct *p, int target)
                return -1;
 
        for_each_cpu(cpu, cpu_smt_mask(target)) {
-               if (!cpumask_test_cpu(cpu, p->cpus_ptr))
+               if (!cpumask_test_cpu(cpu, p->cpus_ptr) ||
+                   !cpumask_test_cpu(cpu, sched_domain_span(sd)))
                        continue;
                if (available_idle_cpu(cpu) || sched_idle_cpu(cpu))
                        return cpu;
@@ -6099,7 +6100,7 @@ static inline int select_idle_core(struct task_struct *p, struct sched_domain *s
        return -1;
 }
 
-static inline int select_idle_smt(struct task_struct *p, int target)
+static inline int select_idle_smt(struct task_struct *p, struct sched_domain *sd, int target)
 {
        return -1;
 }
@@ -6274,7 +6275,7 @@ static int select_idle_sibling(struct task_struct *p, int prev, int target)
        if ((unsigned)i < nr_cpumask_bits)
                return i;
 
-       i = select_idle_smt(p, target);
+       i = select_idle_smt(p, sd, target);
        if ((unsigned)i < nr_cpumask_bits)
                return i;
 
--

1.8.3.1

#2 lore raw
Quoted 15 lines · Vincent Guittot
Quoted 14 lines · Xunlei Pang

We've met problems that occasionally tasks with full cpumask (e.g. by putting it into a cpuset or setting to full affinity) were migrated to our isolated cpus in production environment.

After some analysis, we found that it is due to the current select_idle_smt() not considering the sched_domain mask.

Steps to reproduce on my 31-CPU hyperthreads machine:
1. with boot parameter: "isolcpus=domain,2-31" (thread lists: 0,16 and 1,17)
2. cgcreate -g cpu:test; cgexec -g cpu:test "test_threads"
3. some threads will be migrated to the isolated cpu16~17.

Quoted 9 lines
Quoted 6 lines

Fix it by checking the valid domain mask in select_idle_smt().

Fixes 10e2f1acd010 ("sched/core: Rewrite and improve select_idle_siblings())

Thanks, Vincent :-)

merge bot Applied
#3 lore raw
In reply to Xunlei Pang
The following commit has been merged into the sched/core branch of tip:

The following commit has been merged into the sched/core branch of tip:

Commit-ID:     df3cb4ea1fb63ff326488efd671ba3c39034255e
Gitweb:        https://git.kernel.org/tip/df3cb4ea1fb63ff326488efd671ba3c39034255e
Author:        Xunlei Pang <[email protected]>
AuthorDate:    Thu, 24 Sep 2020 14:48:47 +08:00
Committer:     Peter Zijlstra <[email protected]>
CommitterDate: Fri, 25 Sep 2020 14:23:25 +02:00

sched/fair: Fix wrong cpu selecting from isolated domain

We've met problems that occasionally tasks with full cpumask (e.g. by putting it into a cpuset or setting to full affinity) were migrated to our isolated cpus in production environment.

After some analysis, we found that it is due to the current select_idle_smt() not considering the sched_domain mask.

Steps to reproduce on my 31-CPU hyperthreads machine:
1. with boot parameter: "isolcpus=domain,2-31" (thread lists: 0,16 and 1,17)
2. cgcreate -g cpu:test; cgexec -g cpu:test "test_threads"
3. some threads will be migrated to the isolated cpu16~17.

Fix it by checking the valid domain mask in select_idle_smt().

---
 kernel/sched/fair.c |  9 +++++----
 1 file changed, 5 insertions(+), 4 deletions(-)
kernel/sched/fair.c +5 -4
index a15deb2..9613e5d 100644
@@ -6080,7 +6080,7 @@ static int select_idle_core(struct task_struct *p, struct sched_domain *sd, int 
 /*
  * Scan the local SMT mask for idle CPUs.
  */
-static int select_idle_smt(struct task_struct *p, int target)
+static int select_idle_smt(struct task_struct *p, struct sched_domain *sd, int target)
 {
 	int cpu;
 
@@ -6088,7 +6088,8 @@ static int select_idle_smt(struct task_struct *p, int target)
 		return -1;
 
 	for_each_cpu(cpu, cpu_smt_mask(target)) {
-		if (!cpumask_test_cpu(cpu, p->cpus_ptr))
+		if (!cpumask_test_cpu(cpu, p->cpus_ptr) ||
+		    !cpumask_test_cpu(cpu, sched_domain_span(sd)))
 			continue;
 		if (available_idle_cpu(cpu) || sched_idle_cpu(cpu))
 			return cpu;
@@ -6104,7 +6105,7 @@ static inline int select_idle_core(struct task_struct *p, struct sched_domain *s
 	return -1;
 }
 
-static inline int select_idle_smt(struct task_struct *p, int target)
+static inline int select_idle_smt(struct task_struct *p, struct sched_domain *sd, int target)
 {
 	return -1;
 }
@@ -6279,7 +6280,7 @@ symmetric:
 	if ((unsigned)i < nr_cpumask_bits)
 		return i;
 
-	i = select_idle_smt(p, target);
+	i = select_idle_smt(p, sd, target);
 	if ((unsigned)i < nr_cpumask_bits)
 		return i;