linux-kernel - Re: Latency histogram broken after "hrtimer: Set expiry time before switch_hrtimer

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite for Android: free password hash cracker in your pocket

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-ID: <53A69F09.9000202@osadl.org>
Date:	Sun, 22 Jun 2014 11:16:57 +0200
From:	Carsten Emde <C.Emde@...dl.org>
To:	Mike Galbraith <umgwanakikbuti@...il.com>,
	Ben Hutchings <ben@...adent.org.uk>
CC:	Thomas Gleixner <tglx@...utronix.de>,
	Uwe Kleine-König 
	<u.kleine-koenig@...gutronix.de>,
	LKML <linux-kernel@...r.kernel.org>,
	Steven Rostedt <rostedt@...dmis.org>
Subject: Re: Latency histogram broken after "hrtimer: Set expiry time before
 switch_hrtimer_base()"

On 06/22/2014 06:24 AM, Mike Galbraith wrote:
> [..]
> On Sun, 2014-06-22 at 06:10 +0200, Mike Galbraith wrote:
>> On Sun, 2014-06-22 at 02:04 +0100, Ben Hutchings wrote:
>>> In an rt-kernel with CONFIG_MISSED_TIMER_OFFSETS_HIST enabled,
>>> __hrtimer_start_range_ns() now crashes, as new_base is not assigned
>>> before it is used.
>>
>> Oh yeah, forgot about this.
>>
>>> I'm not sure how this should be fixed; is it:
>>
>> My (3.12-ish tree) merge resolution was the later, but it shouldn't
>> matter which you choose, what's stored where is unchanged pre/post.
I can confirm that 3.12.22-rt34 systems with 
CONFIG_MISSED_TIMER_OFFSETS_HIST enabled crash at an early boot stage. 
After applying Ben's second patch proposal, all is back to normal.

>>
>>> --- a/kernel/hrtimer.c
>>> +++ b/kernel/hrtimer.c
>>> @@ -1108,7 +1108,7 @@ int __hrtimer_start_range_ns(struct hrti
>>>
>>>   #ifdef CONFIG_MISSED_TIMER_OFFSETS_HIST
>>>   	{
>>> -		ktime_t now = new_base->get_time();
>>> +		ktime_t now = base->get_time();
>>>
>>>   		if (ktime_to_ns(tim) < ktime_to_ns(now))
>>>   			timer->praecox = now;
>>> --- END ---
>>>
>>> or:
Didn't use this one.

>>> --- a/kernel/hrtimer.c
>>> +++ b/kernel/hrtimer.c
>>> @@ -1106,6 +1106,11 @@ int __hrtimer_start_range_ns(struct hrti
>>>   #endif
>>>   	}
>>>
>>> +	hrtimer_set_expires_range_ns(timer, tim, delta_ns);
>>> +
>>> +	/* Switch the timer base, if necessary: */
>>> +	new_base = switch_hrtimer_base(timer, base, mode & HRTIMER_MODE_PINNED);
>>> +
>>>   #ifdef CONFIG_MISSED_TIMER_OFFSETS_HIST
>>>   	{
>>>   		ktime_t now = new_base->get_time();
>>> @@ -1117,11 +1122,6 @@ int __hrtimer_start_range_ns(struct hrti
>>>   	}
>>>   #endif
>>>
>>> -	hrtimer_set_expires_range_ns(timer, tim, delta_ns);
>>> -
>>> -	/* Switch the timer base, if necessary: */
>>> -	new_base = switch_hrtimer_base(timer, base, mode & HRTIMER_MODE_PINNED);
>>> -
>>>   	timer_stats_hrtimer_set_start_info(timer);
>>>
>>>   	leftmost = enqueue_hrtimer(timer, new_base);
>>> --- END ---
>>> or something else?
Tested-by: Carsten Emde <C.Emde@...dl.org>

Thanks,
	-Carsten.
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/