Re: [REGRESSION] 41c7f74 rtc: Disable the alarm in the hardware (v2)

[John Stultz <john.stultz@linaro.org>]

On 12/05/2013 03:51 AM, Brecht Machiels wrote:
> On Mon, 02 Dec 2013 22:19:58 +0100, Borislav Petkov wrote:
>
>> On Mon, Dec 02, 2013 at 12:47:17PM -0800, John Stultz wrote:
>>> Ok, sorry about this. I've been hoping we'd get some better insight
>>> into what's actually happening on these strange BIOSes where disabling
>>> the irq seems to cause it to scream (powering the system back on when
>>> its shutdown), in the hopes of having a proper workaround. But despite
>>> Borislav's efforts, he didn't seem to be able to root cause the issue.
>> Right, this bug is too nasty - you could generate good random numbers
>> just from how the hardware behaves. :) And I've been trying to make
>> sense of what happens but I failed, as you know. :(
>>
>> I consider it a huge waste of time and efforts having to deal with such
>> b0rked hardware instead of throwing it out of the window into the poring
>> rain while it is still powered.
>>
>>> Borislav, could you double check this patch still works on your
>>> hardware as well?
>> Well, we have the patch in SLES11:
>>
>> http://kernel.opensuse.org/cgit/kernel/commit/?h=SLE11-
> SP3&id=835398eb94dca7d55acd1a2628372e602ae3252a
>> and it passed testing.
>>
>> From what I see below, your version is equivalent to the one above with
>> the logic reversed so it should work. I'll still try to get that
>> affected box and run your version on it but it'll take a while.
> Hello John and Boris,
>
> Thank you for your quick response. And no need for an apology, I can 
> understand your frustration with the way some hardware behaves.
>
> I ran with John's patch for a couple of days, and it seems to work. 
> Curiously, the laptop did spontaneously boot the first time that I shut 
> it down with the patched kernel. I have no conclusive explanation for 
> this, but I have noticed that a manual power down is necessary after 
> booting with an unpatched kernel. Simply rebooting with a patched kernel 
> is not enough to stop the spontaneous boots. As far as I can remember, I 
> went directly from my custom kernel (with the v2 patch reverted) to a 
> kernel with your patch applied, so I'm not 100% convinced everything is 
> all right. I should say that I did experience some spontaneous boots when 
> running only Windows XP in the past, so there may be occasions where 
> drivers might not be able to help at all.

Yea. It seems almost like the RTC_AIE bit is inverted in the hardware or
something.

> Thankfully, after other shutdowns/hibernates (about 6 in total) the 
> laptop never booted spontaneously.
>
> As for killing alarm functionality on the affected systems, I did some 
> quick tests. With the patched kernel, I can set the RTC alarm by echoing 
> to /sys/class/rtc/rtc0/wakealarm, and the machine will boot at the 
> specified time. I have also tried setting the RTC alarm, and then 
> disabling it again by echoing '0' to /sys/class/rtc/rtc0/wakealarm. While 
> this sets the alrm_time to five minutes in the future, alarm_IRQ is set 
> to 'no' and the machine does *not* boot spontaneously 5 minutes after 
> shutting down. So, all seems well, as far as I can see. Unfortunately, I 
> don't know enough about the RTC driver to draw any conclusions from this.

Ok. Sounds like the patch works fairly well then. I'll go ahead and
queue it for 3.14 and -stable.

thanks
-john