2002-03-18 13:53:08

by Richard Ems

[permalink] [raw]
Subject: 2.4.18 freezes on heavy IO

Hi all!

I'm seeing my system freeze on heavy IO. Only the reset button brings it
back to life again (ALT-SysRq-b also worked once). I'm running SuSE's
2.4.18-30 on a Pentium III (Coppermine) with 256 MB RAM (yes, I should
try vanilla 2.4.18, I will ...)
No SCSI, all IDE. LVM and ext3.
I don't get any oopses, no entries in /var/log/messages, nothing. I
mounted the ext3 partitions with the debug option but still no messages.
What options can I turn on to search for the problem? Any kernel boot
options? LVM/ext3 options?

Many thanks, Richard


--
Richard Ems
... e-mail: [email protected]
... Computer Science, University of Hamburg

Unix IS user friendly. It's just selective about who its friends are.


2002-03-18 14:10:51

by Chris Mason

[permalink] [raw]
Subject: Re: 2.4.18 freezes on heavy IO



On Monday, March 18, 2002 02:52:41 PM +0100 Richard Ems <[email protected]> wrote:

> Hi all!
>
> I'm seeing my system freeze on heavy IO. Only the reset button brings it
> back to life again (ALT-SysRq-b also worked once). I'm running SuSE's
> 2.4.18-30 on a Pentium III (Coppermine) with 256 MB RAM (yes, I should
> try vanilla 2.4.18, I will ...)
> No SCSI, all IDE. LVM and ext3.

Which rpm? k_i386, k_deflt, k_psmp, k_smp? This is probably the
(aa specific) ext3 deadlock recently fixed by andrea and andrew.

-chris


2002-03-18 14:13:01

by Clemens Schwaighofer

[permalink] [raw]
Subject: Re: 2.4.18 freezes on heavy IO

Hello Richard Ems

--On Monday, March 18, 2002 02:52:41 PM +0100 you wrote:

> Hi all!
>
> I'm seeing my system freeze on heavy IO. Only the reset button brings it
> back to life again (ALT-SysRq-b also worked once). I'm running SuSE's
> 2.4.18-30 on a Pentium III (Coppermine) with 256 MB RAM (yes, I should
> try vanilla 2.4.18, I will ...)

try ac3 patch, it worked for me and gave me a stable system

--
"Der Krieg ist ein Massaker von Leuten, die sich nicht kennen, zum
Nutzen von Leuten, die sich kennen, aber nicht massakrieren"
- Paul Val?ry (1871-1945)
mfg, Clemens Schwaighofer PIXELWINGS Medien GMBH
Kandlgasse 15/5, A-1070 Wien T: [+43 1] 524 58 50
JETZT NEU! MIT FEWA GEWASCHEN --> http://www.pixelwings.com

2002-03-18 14:16:41

by Richard Ems

[permalink] [raw]
Subject: Re: 2.4.18 freezes on heavy IO

So next freeze! (The 4th or 5th today!)

Now I got "hda:lost interrupt", ext3 partitions where all mounted with
the "debug" option and the console log level was set to 8 with
ALT-SysRq-8.
Both Caps-Lock and Scroll-Lock LEDs where flashing, what does this mean?
Again only the reset button worked ...

Thanks again, Richard

--
Richard Ems
... e-mail: [email protected]
... Computer Science, University of Hamburg

Unix IS user friendly. It's just selective about who its friends are.

2002-03-18 14:21:13

by Richard Ems

[permalink] [raw]
Subject: Re: 2.4.18 freezes on heavy IO


>Which rpm? k_i386, k_deflt, k_psmp, k_smp? This is probably the
>(aa specific) ext3 deadlock recently fixed by andrea and andrew.
>
>-chris

k_deflt-2.4.18-30.i386.rpm

Andrea's fix should already be there! (as read in
http://ftp.gwdg.de/pub/linux/suse/ftp.suse.com/people/mantel/next/lx_sus24.changes)

Richard


--
Richard Ems
... e-mail: [email protected]
... Computer Science, University of Hamburg

Unix IS user friendly. It's just selective about who its friends are.

2002-03-18 14:26:43

by Alan

[permalink] [raw]
Subject: Re: 2.4.18 freezes on heavy IO

> Now I got "hda:lost interrupt", ext3 partitions where all mounted with
> the "debug" option and the console log level was set to 8 with
> ALT-SysRq-8.
> Both Caps-Lock and Scroll-Lock LEDs where flashing, what does this mean?

Thats the code we added to detect a panic() when you are in a graphical mode

2002-03-18 15:49:10

by Joachim Breuer

[permalink] [raw]
Subject: Re: 2.4.18 freezes on heavy IO; SysRq question

Richard Ems <[email protected]> writes:

> Hi all!
>
> I'm seeing my system freeze on heavy IO. Only the reset button brings it
> back to life again (ALT-SysRq-b also worked once). I'm running SuSE's
> 2.4.18-30 on a Pentium III (Coppermine) with 256 MB RAM (yes, I should
> try vanilla 2.4.18, I will ...)
> No SCSI, all IDE. LVM and ext3.
> I don't get any oopses, no entries in /var/log/messages, nothing. I
> mounted the ext3 partitions with the debug option but still no messages.
> What options can I turn on to search for the problem? Any kernel boot
> options? LVM/ext3 options?

Seconded; happened to me with 2.4.18 from kernel.org + LVM 2.0 beta
1.1 + Trond's NFS ('current' for 2.4.18). IO was to a local ext3 fs;
no LVM on that machine (modules not loaded).

Lock-up was complete for me; no panic message and no IP either so I
couldn't poke around. Alt+Sysrq seemed dead; but it might have been on
one of the boxen where Linux does not detect Alt+Sysrq properly.

BTW, what's the status of the Sysrq entry key "decoder"? I've about 4
different types of keyboards here, as far as I could determine only
one of them sends Linux-parseable Alt-Sysrq. All do send *something*
on Alt-Sysrq, though; but I couldn't get them to work with the
procedure described in the SysRq manual. That was around 2.4.12.


So long,
Joe

--
"I use emacs, which might be thought of as a thermonuclear
word processor."
-- Neal Stephenson, "In the beginning... was the command line"

2002-03-18 16:43:31

by Randy.Dunlap

[permalink] [raw]
Subject: Re: 2.4.18 freezes on heavy IO; SysRq question

On Mon, 18 Mar 2002, Joachim Breuer wrote:

| Richard Ems <[email protected]> writes:
|
| > Hi all!
| >
| > I'm seeing my system freeze on heavy IO. Only the reset button brings it
| > back to life again (ALT-SysRq-b also worked once). I'm running SuSE's
| > 2.4.18-30 on a Pentium III (Coppermine) with 256 MB RAM (yes, I should
| > try vanilla 2.4.18, I will ...)
| > No SCSI, all IDE. LVM and ext3.
| > I don't get any oopses, no entries in /var/log/messages, nothing. I
| > mounted the ext3 partitions with the debug option but still no messages.
| > What options can I turn on to search for the problem? Any kernel boot
| > options? LVM/ext3 options?
|
| Seconded; happened to me with 2.4.18 from kernel.org + LVM 2.0 beta
| 1.1 + Trond's NFS ('current' for 2.4.18). IO was to a local ext3 fs;
| no LVM on that machine (modules not loaded).
|
| Lock-up was complete for me; no panic message and no IP either so I
| couldn't poke around. Alt+Sysrq seemed dead; but it might have been on
| one of the boxen where Linux does not detect Alt+Sysrq properly.
|
| BTW, what's the status of the Sysrq entry key "decoder"? I've about 4
| different types of keyboards here, as far as I could determine only
| one of them sends Linux-parseable Alt-Sysrq. All do send *something*
| on Alt-Sysrq, though; but I couldn't get them to work with the
| procedure described in the SysRq manual. That was around 2.4.12.

I've seen a couple of cheapo keyboards where some
Alt-SysRq-key combinations don't generate anything from the
keyboard. (I'm typing on one of them right now.)
For example, Alt-SysRq-5|6 works, but 1,2,3,4,7,8,9 don't.

I've been told that it's just keyboard manufacturers trying
to save 1/2 cent each keyboard e.g. :(

In any case, I have 2 patches that you might find useful.
One patch uses Alt-SysRq-Y to raise the console loglevel by 1
and Alt-SysRq-V to lower it by 1.
http://www.osdl.org/archive/rddunlap/patches/sysrq-logupdown.dif

The other patch implements magic sysrq keys via sysctl.
E.g., "echo 9 > /proc/sys/kernel/magickey" sets console loglevel
to 9, and "echo m > /proc/sys/kernel/magickey" calls show_mem().
http://www.osdl.org/archive/rddunlap/patches/sys-magic.dif

magickey updates go directly to the sysrq keyboard handler,
so all current keycodes are supported.

--
~Randy

2002-03-18 17:18:18

by David Woodhouse

[permalink] [raw]
Subject: Re: 2.4.18 freezes on heavy IO; SysRq question


[email protected] said:
> I've seen a couple of cheapo keyboards where some Alt-SysRq-key
> combinations don't generate anything from the keyboard. (I'm typing
> on one of them right now.) For example, Alt-SysRq-5|6 works, but
> 1,2,3,4,7,8,9 don't.

Try using the other Alt key.

--
dwmw2


2002-03-18 17:23:38

by Randy.Dunlap

[permalink] [raw]
Subject: Re: 2.4.18 freezes on heavy IO; SysRq question

On Mon, 18 Mar 2002, David Woodhouse wrote:

| [email protected] said:
| > I've seen a couple of cheapo keyboards where some Alt-SysRq-key
| > combinations don't generate anything from the keyboard. (I'm typing
| > on one of them right now.) For example, Alt-SysRq-5|6 works, but
| > 1,2,3,4,7,8,9 don't.
|
| Try using the other Alt key.

Same result with either Alt key.

--
~Randy

2002-03-19 11:46:54

by Joachim Breuer

[permalink] [raw]
Subject: Re: SysRq question

"Randy.Dunlap" <[email protected]> writes:
> On Mon, 18 Mar 2002, David Woodhouse wrote:
>
> | [email protected] said:
> | > I've seen a couple of cheapo keyboards where some Alt-SysRq-key
> | > combinations don't generate anything from the keyboard. (I'm typing
> | > on one of them right now.) For example, Alt-SysRq-5|6 works, but
> | > 1,2,3,4,7,8,9 don't.
> |
> | Try using the other Alt key.
>
> Same result with either Alt key.

BTW that's not the problem I'm seeing, Alt-SysRq-*ANYTHING* is not
caught from any of the offending keyboards because (IIRC) they send
Alt-SysRq as a two-code sequence, with the (apparently) Alt-Down code
repeated before the command keycode - thus, all the command keycode
parser sees (it seems it looks only at the first byte after Alt-SysRq)
is the Alt-Down keycode, which is not a command.

--
"I use emacs, which might be thought of as a thermonuclear
word processor."
-- Neal Stephenson, "In the beginning... was the command line"