Re: [PATCH] hrtimer, add expiry time overflow check in hrtimer_interrupt

Re: [PATCH] hrtimer, add expiry time overflow check in hrtimer_interrupt

[Prarit Bhargava]

On 04/08/2013 04:19 PM, John Stultz wrote:
> On 04/08/2013 05:47 AM, Prarit Bhargava wrote:

>>
>> A simple check for an overflow can resolve this problem.  Using KTIME_MAX
>> instead of the overflow value will result in the hrtimer function being run,
>> and the reprogramming of the timer after that.
>>
>> Signed-off-by: Prarit Bhargava <prarit@redhat.com>
>> Cc: Thomas Gleixner <tglx@linutronix.de>
>> Cc: John Stultz <john.stultz@linaro.org>
> 
> Prarit: Should this be tagged for -stable?

John,

Yes, this should go to -stable.  cc'd.

P.

> 
> thanks
> -john
> 

Re: [PATCH] hrtimer, add expiry time overflow check in hrtimer_interrupt

[John Stultz]

On 04/08/2013 01:34 PM, Prarit Bhargava wrote:
>
> On 04/08/2013 04:19 PM, John Stultz wrote:
>> On 04/08/2013 05:47 AM, Prarit Bhargava wrote:
>>> A simple check for an overflow can resolve this problem.  Using KTIME_MAX
>>> instead of the overflow value will result in the hrtimer function being run,
>>> and the reprogramming of the timer after that.
>>>
>>> Signed-off-by: Prarit Bhargava <prarit@redhat.com>
>>> Cc: Thomas Gleixner <tglx@linutronix.de>
>>> Cc: John Stultz <john.stultz@linaro.org>
>> Prarit: Should this be tagged for -stable?
> John,
>
> Yes, this should go to -stable.  cc'd.
Also, added the cc to the commit.

thanks
-john

Re: [PATCH] hrtimer, add expiry time overflow check in hrtimer_interrupt

[Guenter Roeck]

On Mon, Apr 08, 2013 at 04:34:26PM -0400, Prarit Bhargava wrote:
> 
> 
> On 04/08/2013 04:19 PM, John Stultz wrote:
> > On 04/08/2013 05:47 AM, Prarit Bhargava wrote:
> 
> >>
> >> A simple check for an overflow can resolve this problem.  Using KTIME_MAX
> >> instead of the overflow value will result in the hrtimer function being run,
> >> and the reprogramming of the timer after that.
> >>
> >> Signed-off-by: Prarit Bhargava <prarit@redhat.com>
> >> Cc: Thomas Gleixner <tglx@linutronix.de>
> >> Cc: John Stultz <john.stultz@linaro.org>
> > 
> > Prarit: Should this be tagged for -stable?
> 
> John,
> 
> Yes, this should go to -stable.  cc'd.
> 
Hi,

I am a bit surprised that this patch has not found its way into mainline yet,
as everyone seems to agree that it is a candidate for -stable.

I hit this problem very reliably (ie with each boot) with 3.8.x on systems
which have no RTC and run systemd. Seen with Freescale P5040 as well as
a Broadcom MIPS based system.

Guenter

Re: [PATCH] hrtimer, add expiry time overflow check in hrtimer_interrupt

[John Stultz]

On 04/24/2013 03:42 PM, Guenter Roeck wrote:
> On Mon, Apr 08, 2013 at 04:34:26PM -0400, Prarit Bhargava wrote:
>>
>> On 04/08/2013 04:19 PM, John Stultz wrote:
>>> On 04/08/2013 05:47 AM, Prarit Bhargava wrote:
>>>> A simple check for an overflow can resolve this problem.  Using KTIME_MAX
>>>> instead of the overflow value will result in the hrtimer function being run,
>>>> and the reprogramming of the timer after that.
>>>>
>>>> Signed-off-by: Prarit Bhargava <prarit@redhat.com>
>>>> Cc: Thomas Gleixner <tglx@linutronix.de>
>>>> Cc: John Stultz <john.stultz@linaro.org>
>>> Prarit: Should this be tagged for -stable?
>> John,
>>
>> Yes, this should go to -stable.  cc'd.
>>
> Hi,
>
> I am a bit surprised that this patch has not found its way into mainline yet,
> as everyone seems to agree that it is a candidate for -stable.

It just has to land upstream first, which is likely in the next week or 
so when the 3.10 merge window opens. I'd have thought it would be sooner 
but 3.9 is taking longer to close then I expected (and I didn't think it 
was urgent enough to drop in at the last minute before the 3.9 release 
was made).

thanks
-john

Re: [PATCH] hrtimer, add expiry time overflow check in hrtimer_interrupt

[Guenter Roeck]

On Wed, Apr 24, 2013 at 05:05:03PM -0700, John Stultz wrote:
> On 04/24/2013 03:42 PM, Guenter Roeck wrote:
> >On Mon, Apr 08, 2013 at 04:34:26PM -0400, Prarit Bhargava wrote:
> >>
> >>On 04/08/2013 04:19 PM, John Stultz wrote:
> >>>On 04/08/2013 05:47 AM, Prarit Bhargava wrote:
> >>>>A simple check for an overflow can resolve this problem.  Using KTIME_MAX
> >>>>instead of the overflow value will result in the hrtimer function being run,
> >>>>and the reprogramming of the timer after that.
> >>>>
> >>>>Signed-off-by: Prarit Bhargava <prarit@redhat.com>
> >>>>Cc: Thomas Gleixner <tglx@linutronix.de>
> >>>>Cc: John Stultz <john.stultz@linaro.org>
> >>>Prarit: Should this be tagged for -stable?
> >>John,
> >>
> >>Yes, this should go to -stable.  cc'd.
> >>
> >Hi,
> >
> >I am a bit surprised that this patch has not found its way into mainline yet,
> >as everyone seems to agree that it is a candidate for -stable.
> 
> It just has to land upstream first, which is likely in the next week
> or so when the 3.10 merge window opens. I'd have thought it would be
> sooner but 3.9 is taking longer to close then I expected (and I
> didn't think it was urgent enough to drop in at the last minute
> before the 3.9 release was made).
> 
Guess I am a bit lost in process.

If this is going to be in -stable, it will presumably end up in 3.9.x as well as
in earlier releases. So why wasn't it pushed into 3.9-rcX to start with ?

Guenter

Re: [PATCH] hrtimer, add expiry time overflow check in hrtimer_interrupt

[John Stultz]

On 04/24/2013 05:35 PM, Guenter Roeck wrote:
> On Wed, Apr 24, 2013 at 05:05:03PM -0700, John Stultz wrote:
>> On 04/24/2013 03:42 PM, Guenter Roeck wrote:
>>> On Mon, Apr 08, 2013 at 04:34:26PM -0400, Prarit Bhargava wrote:
>>>> On 04/08/2013 04:19 PM, John Stultz wrote:
>>>>> On 04/08/2013 05:47 AM, Prarit Bhargava wrote:
>>>>>> A simple check for an overflow can resolve this problem.  Using KTIME_MAX
>>>>>> instead of the overflow value will result in the hrtimer function being run,
>>>>>> and the reprogramming of the timer after that.
>>>>>>
>>>>>> Signed-off-by: Prarit Bhargava <prarit@redhat.com>
>>>>>> Cc: Thomas Gleixner <tglx@linutronix.de>
>>>>>> Cc: John Stultz <john.stultz@linaro.org>
>>>>> Prarit: Should this be tagged for -stable?
>>>> John,
>>>>
>>>> Yes, this should go to -stable.  cc'd.
>>>>
>>> Hi,
>>>
>>> I am a bit surprised that this patch has not found its way into mainline yet,
>>> as everyone seems to agree that it is a candidate for -stable.
>> It just has to land upstream first, which is likely in the next week
>> or so when the 3.10 merge window opens. I'd have thought it would be
>> sooner but 3.9 is taking longer to close then I expected (and I
>> didn't think it was urgent enough to drop in at the last minute
>> before the 3.9 release was made).
>>
> Guess I am a bit lost in process.
>
> If this is going to be in -stable, it will presumably end up in 3.9.x as well as
> in earlier releases. So why wasn't it pushed into 3.9-rcX to start with ?

I usually only want to push changes to -rc6+ if they are really 
critical, affecting lots of folks and fixing issues introduced in the 
same cycle. By getting less critical fixes merged during a normal merge 
window, then backporting them to affected -stable trees, we get better 
test coverage and less chance for further bugs to be introduced at the 
last minute before the release is made.

Its maybe a bit overly conservative, but I'm less and less into 
late-night heroics these days. ;)

thanks
-john

Re: [PATCH] hrtimer, add expiry time overflow check in hrtimer_interrupt

[Li Zefan]

On 2013/4/25 6:42, Guenter Roeck wrote:
> On Mon, Apr 08, 2013 at 04:34:26PM -0400, Prarit Bhargava wrote:
>>
>>
>> On 04/08/2013 04:19 PM, John Stultz wrote:
>>> On 04/08/2013 05:47 AM, Prarit Bhargava wrote:
>>
>>>>
>>>> A simple check for an overflow can resolve this problem.  Using KTIME_MAX
>>>> instead of the overflow value will result in the hrtimer function being run,
>>>> and the reprogramming of the timer after that.
>>>>
>>>> Signed-off-by: Prarit Bhargava <prarit@redhat.com>
>>>> Cc: Thomas Gleixner <tglx@linutronix.de>
>>>> Cc: John Stultz <john.stultz@linaro.org>
>>>
>>> Prarit: Should this be tagged for -stable?
>>
>> John,
>>
>> Yes, this should go to -stable.  cc'd.
>>
> Hi,
> 
> I am a bit surprised that this patch has not found its way into mainline yet,
> as everyone seems to agree that it is a candidate for -stable.
> 
> I hit this problem very reliably (ie with each boot) with 3.8.x on systems
> which have no RTC and run systemd. Seen with Freescale P5040 as well as
> a Broadcom MIPS based system.
> 

FYI, we also hit this warning with 3.4-rt.

Re: [PATCH] hrtimer, add expiry time overflow check in hrtimer_interrupt

[Guenter Roeck]

On Thu, Apr 25, 2013 at 09:38:22AM +0800, Li Zefan wrote:
> On 2013/4/25 6:42, Guenter Roeck wrote:
> > On Mon, Apr 08, 2013 at 04:34:26PM -0400, Prarit Bhargava wrote:
> >>
> >>
> >> On 04/08/2013 04:19 PM, John Stultz wrote:
> >>> On 04/08/2013 05:47 AM, Prarit Bhargava wrote:
> >>
> >>>>
> >>>> A simple check for an overflow can resolve this problem.  Using KTIME_MAX
> >>>> instead of the overflow value will result in the hrtimer function being run,
> >>>> and the reprogramming of the timer after that.
> >>>>
> >>>> Signed-off-by: Prarit Bhargava <prarit@redhat.com>
> >>>> Cc: Thomas Gleixner <tglx@linutronix.de>
> >>>> Cc: John Stultz <john.stultz@linaro.org>
> >>>
> >>> Prarit: Should this be tagged for -stable?
> >>
> >> John,
> >>
> >> Yes, this should go to -stable.  cc'd.
> >>
> > Hi,
> > 
> > I am a bit surprised that this patch has not found its way into mainline yet,
> > as everyone seems to agree that it is a candidate for -stable.
> > 
> > I hit this problem very reliably (ie with each boot) with 3.8.x on systems
> > which have no RTC and run systemd. Seen with Freescale P5040 as well as
> > a Broadcom MIPS based system.
> > 
> 
> FYI, we also hit this warning with 3.4-rt.
> 
You are lucky if it is just a warning for you. In my case the system is
reliably dead. Guess there are not (yet) many users out there using
systemd (or something similar) on a system with no RTC.

While I am not too happy about the delay to get the patch integrated,
I am glad that Prarit found and fixed the problem. Saved me a lot of time.

Guenter