Messages in this thread Patch in this message | | | Date | Thu, 5 Feb 2015 16:33:24 +0800 | From | Zefan Li <> | Subject | [PATCH] sched, autogroup: Fix failure when writing to cpu.rt_runtime_us |
| |
This is how to reproduce the bug:
int main() { struct sched_param param = {.sched_priority=1};
if (fork() > 0) exit(0);
setsid();
if (sched_setscheduler(0, SCHED_RR, ¶m) < 0){ perror("failed to sched_setscheduler()"); return -1; }
while(1) ; }
# ./test # mount -t cgroup -o cpu xxx /cgroup # cat /cgroup/cpu.rt_runtime_us 950000 # echo 940000 > /cgroup/cpu.rt_runtime_us Device or Resource busy
An autogroup has been created and there's an RT task in it. RT tasks in autogroups are redirected to the root group and task_group() should return &root_task_group, but it's broken and returns the autogroup.
We should reset p->sched_task_group when changing a normal task to an RT task which is in autogroup.
Fixes: 8323f26ce342 ("sched: Fix race in task_group()") Cc: <stable@vger.kernel.org> # 3.6+ Reported-by: Zhang Wei <zhangwei555@huawei.com> Signed-off-by: Zefan Li <lizefan@huawei.com> --- kernel/sched/core.c | 15 +++++++++++---- 1 file changed, 11 insertions(+), 4 deletions(-)
diff --git a/kernel/sched/core.c b/kernel/sched/core.c index 89e7283..fccde96 100644 --- a/kernel/sched/core.c +++ b/kernel/sched/core.c @@ -3414,6 +3414,7 @@ static int __sched_setscheduler(struct task_struct *p, const struct sched_class *prev_class; struct rq *rq; int reset_on_fork; + bool reset_task_group = false; /* may grab non-irq protected spin_locks */ BUG_ON(in_interrupt()); @@ -3546,10 +3547,13 @@ change: * assigned. */ if (rt_bandwidth_enabled() && rt_policy(policy) && - task_group(p)->rt_bandwidth.rt_runtime == 0 && - !task_group_is_autogroup(task_group(p))) { - task_rq_unlock(rq, p, &flags); - return -EPERM; + task_group(p)->rt_bandwidth.rt_runtime == 0) { + if (!task_group_is_autogroup(task_group(p))) { + task_rq_unlock(rq, p, &flags); + return -EPERM; + } else { + reset_task_group = true; + } } #endif #ifdef CONFIG_SMP @@ -3615,6 +3619,9 @@ change: prev_class = p->sched_class; __setscheduler(rq, p, attr); + if (reset_task_group) + p->sched_task_group = &root_task_group; + if (running) p->sched_class->set_curr_task(rq); if (queued) { -- 1.8.0.2
| |