diff options
author | Mike Galbraith <bitbucket@online.de> | 2014-04-07 04:55:15 -0400 |
---|---|---|
committer | Ingo Molnar <mingo@kernel.org> | 2014-04-11 04:39:15 -0400 |
commit | 60e69eed85bb7b5198ef70643b5895c26ad76ef7 (patch) | |
tree | 4fa65bf81edbd0a28d5bc160f878341657f5ad82 /kernel | |
parent | 467cbd207abdbfe29514b5804a22661ab6e80dc6 (diff) |
sched/numa: Fix task_numa_free() lockdep splat
Sasha reported that lockdep claims that the following commit:
made numa_group.lock interrupt unsafe:
156654f491dd ("sched/numa: Move task_numa_free() to __put_task_struct()")
While I don't see how that could be, given the commit in question moved
task_numa_free() from one irq enabled region to another, the below does
make both gripes and lockups upon gripe with numa=fake=4 go away.
Reported-by: Sasha Levin <sasha.levin@oracle.com>
Fixes: 156654f491dd ("sched/numa: Move task_numa_free() to __put_task_struct()")
Signed-off-by: Mike Galbraith <bitbucket@online.de>
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Cc: torvalds@linux-foundation.org
Cc: mgorman@suse.com
Cc: akpm@linux-foundation.org
Cc: Dave Jones <davej@redhat.com>
Link: http://lkml.kernel.org/r/1396860915.5170.5.camel@marge.simpson.net
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Diffstat (limited to 'kernel')
-rw-r--r-- | kernel/sched/fair.c | 13 | ||||
-rw-r--r-- | kernel/sched/sched.h | 9 |
2 files changed, 16 insertions, 6 deletions
diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c index 7e9bd0b1fa9e..4f14a656a720 100644 --- a/kernel/sched/fair.c +++ b/kernel/sched/fair.c | |||
@@ -1497,7 +1497,7 @@ static void task_numa_placement(struct task_struct *p) | |||
1497 | /* If the task is part of a group prevent parallel updates to group stats */ | 1497 | /* If the task is part of a group prevent parallel updates to group stats */ |
1498 | if (p->numa_group) { | 1498 | if (p->numa_group) { |
1499 | group_lock = &p->numa_group->lock; | 1499 | group_lock = &p->numa_group->lock; |
1500 | spin_lock(group_lock); | 1500 | spin_lock_irq(group_lock); |
1501 | } | 1501 | } |
1502 | 1502 | ||
1503 | /* Find the node with the highest number of faults */ | 1503 | /* Find the node with the highest number of faults */ |
@@ -1572,7 +1572,7 @@ static void task_numa_placement(struct task_struct *p) | |||
1572 | } | 1572 | } |
1573 | } | 1573 | } |
1574 | 1574 | ||
1575 | spin_unlock(group_lock); | 1575 | spin_unlock_irq(group_lock); |
1576 | } | 1576 | } |
1577 | 1577 | ||
1578 | /* Preferred node as the node with the most faults */ | 1578 | /* Preferred node as the node with the most faults */ |
@@ -1677,7 +1677,8 @@ static void task_numa_group(struct task_struct *p, int cpupid, int flags, | |||
1677 | if (!join) | 1677 | if (!join) |
1678 | return; | 1678 | return; |
1679 | 1679 | ||
1680 | double_lock(&my_grp->lock, &grp->lock); | 1680 | BUG_ON(irqs_disabled()); |
1681 | double_lock_irq(&my_grp->lock, &grp->lock); | ||
1681 | 1682 | ||
1682 | for (i = 0; i < NR_NUMA_HINT_FAULT_STATS * nr_node_ids; i++) { | 1683 | for (i = 0; i < NR_NUMA_HINT_FAULT_STATS * nr_node_ids; i++) { |
1683 | my_grp->faults[i] -= p->numa_faults_memory[i]; | 1684 | my_grp->faults[i] -= p->numa_faults_memory[i]; |
@@ -1691,7 +1692,7 @@ static void task_numa_group(struct task_struct *p, int cpupid, int flags, | |||
1691 | grp->nr_tasks++; | 1692 | grp->nr_tasks++; |
1692 | 1693 | ||
1693 | spin_unlock(&my_grp->lock); | 1694 | spin_unlock(&my_grp->lock); |
1694 | spin_unlock(&grp->lock); | 1695 | spin_unlock_irq(&grp->lock); |
1695 | 1696 | ||
1696 | rcu_assign_pointer(p->numa_group, grp); | 1697 | rcu_assign_pointer(p->numa_group, grp); |
1697 | 1698 | ||
@@ -1710,14 +1711,14 @@ void task_numa_free(struct task_struct *p) | |||
1710 | void *numa_faults = p->numa_faults_memory; | 1711 | void *numa_faults = p->numa_faults_memory; |
1711 | 1712 | ||
1712 | if (grp) { | 1713 | if (grp) { |
1713 | spin_lock(&grp->lock); | 1714 | spin_lock_irq(&grp->lock); |
1714 | for (i = 0; i < NR_NUMA_HINT_FAULT_STATS * nr_node_ids; i++) | 1715 | for (i = 0; i < NR_NUMA_HINT_FAULT_STATS * nr_node_ids; i++) |
1715 | grp->faults[i] -= p->numa_faults_memory[i]; | 1716 | grp->faults[i] -= p->numa_faults_memory[i]; |
1716 | grp->total_faults -= p->total_numa_faults; | 1717 | grp->total_faults -= p->total_numa_faults; |
1717 | 1718 | ||
1718 | list_del(&p->numa_entry); | 1719 | list_del(&p->numa_entry); |
1719 | grp->nr_tasks--; | 1720 | grp->nr_tasks--; |
1720 | spin_unlock(&grp->lock); | 1721 | spin_unlock_irq(&grp->lock); |
1721 | rcu_assign_pointer(p->numa_group, NULL); | 1722 | rcu_assign_pointer(p->numa_group, NULL); |
1722 | put_numa_group(grp); | 1723 | put_numa_group(grp); |
1723 | } | 1724 | } |
diff --git a/kernel/sched/sched.h b/kernel/sched/sched.h index c9007f28d3a2..456e492a3dca 100644 --- a/kernel/sched/sched.h +++ b/kernel/sched/sched.h | |||
@@ -1385,6 +1385,15 @@ static inline void double_lock(spinlock_t *l1, spinlock_t *l2) | |||
1385 | spin_lock_nested(l2, SINGLE_DEPTH_NESTING); | 1385 | spin_lock_nested(l2, SINGLE_DEPTH_NESTING); |
1386 | } | 1386 | } |
1387 | 1387 | ||
1388 | static inline void double_lock_irq(spinlock_t *l1, spinlock_t *l2) | ||
1389 | { | ||
1390 | if (l1 > l2) | ||
1391 | swap(l1, l2); | ||
1392 | |||
1393 | spin_lock_irq(l1); | ||
1394 | spin_lock_nested(l2, SINGLE_DEPTH_NESTING); | ||
1395 | } | ||
1396 | |||
1388 | static inline void double_raw_lock(raw_spinlock_t *l1, raw_spinlock_t *l2) | 1397 | static inline void double_raw_lock(raw_spinlock_t *l1, raw_spinlock_t *l2) |
1389 | { | 1398 | { |
1390 | if (l1 > l2) | 1399 | if (l1 > l2) |