Reply-To: xlpang@redhat.com
Subject: Re: [PATCH] sched/deadline/rtmutex: Fix a PI crash for deadline tasks
References: <56FE78E0.5060504@redhat.com> <20160401215143.GB2906@worktop>
 <57037974.1020002@redhat.com>
 <20160405091954.GI3448@twins.programming.kicks-ass.net>
 <20160405092954.GC24771@twins.programming.kicks-ass.net>
 <20160408122510.15978179@gandalf.local.home> <20160408173835.GC1087@worktop>
 <20160408145055.13c98a75@gandalf.local.home>
 <20160408185916.GQ3448@twins.programming.kicks-ass.net>
 <57087633.3020700@redhat.com>
 <20160409132935.GU3448@twins.programming.kicks-ass.net>
To: Peter Zijlstra <peterz@infradead.org>
Cc: Steven Rostedt <rostedt@goodmis.org>, linux-kernel@vger.kernel.org,
        Juri Lelli <juri.lelli@arm.com>, Ingo Molnar <mingo@redhat.com>,
        Thomas Gleixner <tglx@linutronix.de>
From: Xunlei Pang <xpang@redhat.com>
Message-ID: <570A0D37.8080400@redhat.com>
Date: Sun, 10 Apr 2016 16:22:15 +0800
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:38.0) Gecko/20100101
 Thunderbird/38.2.0
MIME-Version: 1.0
In-Reply-To: <20160409132935.GU3448@twins.programming.kicks-ass.net>
Content-Type: text/plain; charset=windows-1252
Content-Transfer-Encoding: 7bit
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 1445
Lines: 30

On 2016/04/09 at 21:29, Peter Zijlstra wrote:
> On Sat, Apr 09, 2016 at 11:25:39AM +0800, Xunlei Pang wrote:
>
>>> In any case, I just realized we do not in fact provide this guarantee
>>> (of pointing to a blocked task) that needs a bit more work.
>> Current patch calls rt_mutex_adjust_prio() before wake_up_q() the
>> wakee, at that moment the wakee has been removed by
>> rt_mutex_slowunlock()->mark_wakeup_next_waiter(), from current's
>> pi_waiters, rt_mutex_adjust_prio() won't see this wakee, so I think
>> this should not be problem.
> No, any wakeup after mark_wakeup_next_waiter() will make the task run.
> And since we must always consider spurious wakeups, we cannot rely on us
> (eg. our wake_up_q call) being the one and only.
>
> Therefore it is possible and the only thing that stands between us and
> doom is the fact that the wake_q stuff holds a task reference.
>
> But we cannot guarantee that the task we have a pointer to is in fact
> blocked.

Does that really matters? the pointer is accessed on behalf of current,  and current
will call rt_mutex_adjust_prio() very soon to update the right pointer.

Also the pointer is used to update current's deadline/runtime, we can restore these
params in rt_mutex_setprio() for deboost cases. I just checked current code,  it did
nothing to restore current's deadline/runtime when deboosting, maybe we can leave
this job to future deadline bandwidth inheritance?

Regards,
Xunlei