From mboxrd@z Thu Jan 1 00:00:00 1970 From: Li Zefan Subject: Re: [PATCH] sched: silence PROVE_RCU in sched_fork() Date: Wed, 23 Jun 2010 15:25:50 +0800 Message-ID: <4C21B6FE.8060801@cn.fujitsu.com> References: <20100608001929.GF2387@linux.vnet.ibm.com> <1275986441.5408.111.camel@twins> <1276004075.2987.208.camel@twins> <1276097068.1745.13.camel@laptop> <1276106229.1745.65.camel@laptop> <1276107311.1745.66.camel@laptop> <1276118458.1745.144.camel@laptop> <1277199893.1875.690.ca mel@laptop> Mime-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 7bit Cc: Miles Lane , paulmck@linux.vnet.ibm.com, Vivek Goyal , Eric Paris , Lai Jiangshan , Ingo Molnar , LKML , nauman@google.com, eric.dumazet@gmail.com, netdev@vger.kernel.org, Jens Axboe , Gui Jianfeng , Johannes Berg , Paul Menage To: Peter Zijlstra Return-path: In-Reply-To: <1277199893.1875.690.camel@laptop> Sender: linux-kernel-owner@vger.kernel.org List-Id: netdev.vger.kernel.org Peter Zijlstra wrote: > Paul Menage, Li Zefan, any comments? > > > --- > Because cgroup_fork() is ran before sched_fork() [ from copy_process() ] > and the child's pid is not yet visible the child is pinned to its > cgroup. Therefore we can silence this warning. > The explanation is correct. We silenced another warning according to the same reason. See freezer_fork() in kernel/cgroup_freezer.c > A nicer solution would be moving cgroup_fork() to right after > dup_task_struct() and exclude PF_STARTING from task_subsys_state(). > Seems PF_STARTING is set in copy_flags(), that's after dup_task_struct() and before cgroup_fork(). But it's cleared after copy_process(), that's after the task is linked into tasklist, so this doesn't seem work... > Signed-off-by: Peter Zijlstra For this patch: Reviewed-by: Li Zefan > --- > kernel/sched.c | 9 +++++++++ > 1 files changed, 9 insertions(+), 0 deletions(-) > > diff --git a/kernel/sched.c b/kernel/sched.c > index b697606..2e79518 100644 > --- a/kernel/sched.c > +++ b/kernel/sched.c > @@ -2561,7 +2561,16 @@ void sched_fork(struct task_struct *p, int clone_flags) > if (p->sched_class->task_fork) > p->sched_class->task_fork(p); > > + /* > + * The child is not yet in the pid-hash so no cgroup attach races, > + * and the cgroup is pinned to this child due to cgroup_fork() > + * is ran before sched_fork(). > + * > + * Silence PROVE_RCU. > + */ > + rcu_read_lock(); > set_task_cpu(p, cpu); > + rcu_read_unlock(); > > #if defined(CONFIG_SCHEDSTATS) || defined(CONFIG_TASK_DELAY_ACCT) > if (likely(sched_info_on())) > > >