linux-kernel - Re: 2.6.19.2 sky2/acpi crashes

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: Windows password security audit tool. GUI, reports in PDF.

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

Date:	Wed, 24 Jan 2007 14:51:52 -0500
From:	Len Brown <lenb@...nel.org>
To:	Lionel Landwerlin <landwer@...e.fr>
Cc:	linux-kernel@...r.kernel.org
Subject: Re: 2.6.19.2 sky2/acpi crashes

On Wednesday 24 January 2007 07:23, Lionel Landwerlin wrote:
> Le mardi 23 janvier 2007 à 21:36 -0500, Len Brown a écrit :
> > > > > Apple Macbook 2GHz (x86, not amd64)
> > 
> > > > > > > Please try to remove processor module.
> > > > > >
> > > > > > Ok, that's done. Same problem.
> > > > >
> > > > > any difference with "idle=poll"?
> > > > > if yes, how about "idle=halt"?
> > > >
> > > > idle=poll seems to fix the problem (cpu fan is running almost at full
> > > > speed). Maybe I should run a longer test... For now it consists to run
> > > > about 15 torrents and watching HDTV through ethernet device.
> > > >
> > > > idle=halt does not :
> > > 
> > > It sounds like issues relative to Processor C state.
> > > Please enter a bug in ACPI category on bugzilla.kernel.org
> > 
> > Actually, the test above with the processor module removed proved
> > that it isn't ACPI C-states -- as they will not be available.
> > You should be able to observe that /proc/acpi/processor/*/power
> > does not indicate any C-state use when processor is unloaded.
> > 
> > My guess was that some deep C-state with long exit latency
> > was interfering with the device.  booting w/o the processor
> > module should have left you running the native mwait idle.
> > booting with idle=halt should have left you running the HLT idle.
> > booting with idle=poll is a busy spin loop that never enters
> > any hardware power saving state.
> > 
> > I'm quite puzzled that idle=halt was not sufficient to solve the issue,
> > because that should be the lowest exit latency idle loop.
> > So maybe I'm wrong about the cause -- though I can't then
> > explain why idle=poll helps...
> > 
> > All of the idle selection options cause the kernel to print
> > a line with the word "idle" in it.  Perhaps you could search
> > your dmesg for "idle" to verify that it is running what we
> > think it is?
> 
> Here I join the complete log for idle=halt

it is indeed doing what you asked it to
Jan 24 01:11:21 cocoduo kernel: [   30.741018] using halt in idle threads.

> I'm running idle=poll for more 1 hour now with heavy ethernet load, no
> crash. It usualy happens in 10~15mn with idle=halt and 4~5mn with no
> idle option.

I think my guess is wrong.  If idle=halt doesn't help, then the failure
doesn't have anything to do with the idle loop and power saving idle states.

I can't explain why idle=poll helps.

-Len
-
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/