Bug#626427: linux-image-2.6.38-2-686: Horrible Time Skew, Eventual Near-zero Responsiveness
On 12 May 2011, at 01:43, Ben Hutchings wrote:
> On Wed, 2011-05-11 at 22:55 +0100, Sabahattin Gucukoglu wrote:
>> Package: linux-2.6
>> Version: 2.6.38-5
>> Severity: important
>>
>> (reportbug script died, hope you've got everything.)
>>
>> The machine runs with *HORRIBLE* clock drift (see text). Eventually, the
>> system will slow right down, so that new processes or threads will not
>> start. It *seems* to be related to the use of BRLTTY. See this thread:
>> http://www.mail-archive.com/brltty@mielke.cc/msg05352.html
>>
>> And, yes, it's annoying having to hard-reboot the box every so often. It
>> would seem that after some period of time, touching my display or the
>> physical console triggers the lockup. I don't think this is a coincidence,
>> even though on the surface there are two different problems. The kernel
>> upgrade was sufficient to begin these issues.
>
> Please test whether any of these kernel parameters help (one at a time):
>
> clocksource=jiffies
> clocksource=pit
> idle=halt
Except for the kernel reporting on the fact that the pit clocksource was registered explicitly rather than chosen, nothing happens. (Interestingly, "tsc" fixed it in a previous encounter with this problem, where acpi_pm was default in Linux 2.6.18, on the same hardware, but with less drastic results, and is the one chosen now. Trying acpi_pm again just in case, as that was also a candidate, and it already seems more stable.)
> Could you also send a log from an earlier kernel version where this
> problem does not occur?
I'll let you know when I locate it.
Cheers,
Sabahattin
Reply to: