Re: ATA + DMA still giving repeatable freezes

From: Robert Watson <rwatson_at_freebsd.org>
Date: Tue, 6 Jan 2004 14:46:31 -0500 (EST)
On Wed, 24 Dec 2003, Kirk Strauser wrote:

> At 2003-12-20T00:17:00Z, Robert Watson <rwatson_at_FreeBSD.org> writes:
> 
> > FYI, I had this problem on a notebook and foudn that disabling "device
> > apic" made things clear up.  Try it and see if it helps, and if it does,
> > post as much.  I'm busy testing to make sure my observation is right on my
> > system.  I also saw fairly frequent hangs starting or leaving XWindows,
> > and John Baldwin hypothesized there was an interaction between DRM and the
> > interrupt code.
> 
> Robert, I wanted to get in some more testing before reporting back.  I
> cvsup'ed again and built world and kernel.  This time, I started with
> plain GENERIC and made minor changes until I could reproduce the hangs. 
> Basically, commenting out "options WITNESS" and "options
> WITNESS_SKIPSPIN"  was enough to trigger the problem; I'm guessing that
> these options slow down the system enough that timing issues disappear. 
> 
> After adding this to my /boot/loader.conf.local:
> 
>     hw.ata.ata_dma="0"
>     hint.acpi.0.disabled="1"
> 
> I no longer get system hangs, although I do get these messages at random
> intervals during heavy ATA IO:
> 
>     ad0: TIMEOUT - READ_MUL retrying (2 retries left)
>     ata0: resetting devices ..
>     done

Not sure if you misread, but it looks like you're disabling ACPI rather
than apic (confusing, eh? :-).  I'm currently in a situation on my Dell
Notebook where leaving the new ACPI enable but "device apic" disabled
seems to get me to full functionality.  I discovered a BIOS upgrade on the
Dell web site, and need to try installing that to see if it helps.  I also
haven't tried plugging and playing DRI as yet. 

Robert N M Watson             FreeBSD Core Team, TrustedBSD Projects
robert_at_fledge.watson.org      Senior Research Scientist, McAfee Research
Received on Tue Jan 06 2004 - 10:47:56 UTC

This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:37:36 UTC