From mboxrd@z Thu Jan 1 00:00:00 1970 From: Steven Rostedt Subject: Re: [PATCH v4 4/5] sched/core: Prevent race condition between cpuset and __sched_setscheduler() Date: Thu, 14 Jun 2018 16:11:42 -0400 Message-ID: <20180614161142.69f186a6@gandalf.local.home> References: <20180613121711.5018-1-juri.lelli@redhat.com> <20180613121711.5018-5-juri.lelli@redhat.com> Mime-Version: 1.0 Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: <20180613121711.5018-5-juri.lelli@redhat.com> Sender: linux-kernel-owner@vger.kernel.org List-ID: Content-Type: text/plain; charset="us-ascii" To: Juri Lelli Cc: peterz@infradead.org, mingo@redhat.com, linux-kernel@vger.kernel.org, luca.abeni@santannapisa.it, claudio@evidence.eu.com, tommaso.cucinotta@santannapisa.it, bristot@redhat.com, mathieu.poirier@linaro.org, lizefan@huawei.com, cgroups@vger.kernel.org On Wed, 13 Jun 2018 14:17:10 +0200 Juri Lelli wrote: > +/** > + * cpuset_lock - Grab the cpuset_mutex from another subsysytem > + */ > +int cpuset_lock(void) > +{ > + return mutex_trylock(&cpuset_mutex); > +} > + > +/** > + * cpuset_unlock - Release the cpuset_mutex from another subsysytem > + */ > +void cpuset_unlock(void) > +{ > + mutex_unlock(&cpuset_mutex); > +} > + > /** > * cpuset_cpus_allowed - return cpus_allowed mask from a tasks cpuset. > * @tsk: pointer to task_struct from which to obtain cpuset->cpus_allowed. > diff --git a/kernel/sched/core.c b/kernel/sched/core.c > index ca788f74259d..a5b0c6c25b44 100644 > --- a/kernel/sched/core.c > +++ b/kernel/sched/core.c > @@ -4218,6 +4218,14 @@ static int __sched_setscheduler(struct task_struct *p, > if (attr->sched_flags & SCHED_FLAG_SUGOV) > return -EINVAL; > > + /* > + * Make sure we don't race with the cpuset subsystem where root > + * domains can be rebuilt or modified while operations like DL > + * admission checks are carried out. > + */ > + if (!cpuset_lock()) > + return -EBUSY; > + How important is this for being a trylock? Reason I am asking is that it is making my deadline test fail to get the resources. What my test does is to create a bunch of threads, and they try to get the deadline resources as they are created. This happens in parallel. Thus, when one gets the cpuset_lock, the others are hitting this and returning -EBUSY. At first I thought I was over committing somehow but after adding a trace_printk() in cpuset_lock(), and this fail case it became obvious to what was happening: deadline_test-1376 [004] 107.179693: sys_enter_sched_setattr: pid: 0x00000000, uattr: 0x7f017fceeed0, flags: 0x00000000 deadline_test-1376 [004] 107.179697: bputs: cpuset_lock: cpuset_mutex trylock deadline_test-1377 [003] 107.179763: sys_exit_futex: 0x0 deadline_test-1377 [003] 107.179767: sys_enter_sched_setattr: pid: 0x00000000, uattr: 0x7f017f4eded0, flags: 0x00000000 deadline_test-1377 [003] 107.179771: bputs: cpuset_lock: cpuset_mutex trylock deadline_test-1377 [003] 107.179771: bputs: __sched_setscheduler: cpuset_lock deadline_test-1377 [003] 107.179773: sys_exit_sched_setattr: 0xfffffffffffffff0 -- Steve > retval = security_task_setscheduler(p); > if (retval) > return retval;