Re: 2.5.65: *huge* interactivity problems

From: CaT (cat@zip.com.au)
Date: Tue Mar 25 2003 - 10:43:16 EST


On Mon, Mar 24, 2003 at 06:12:50PM -0800, Andrew Morton wrote:
> > Same here. After its straneous excercise the machine goes back to normal.
>
> Have you fiddled with all your power mgmt and bios options?
> Disabled acpi and apm? Tried an absolutely bare-bones kernel?

The fun so far. Standard build with 2.5.66 (config-big attachment):

With the kernel and mplayer compiling, mozilla tyring to load 40 or so
webpages at the same time and copying a 30meg file over nfs caused the
system to start severely fscking up. Just moving the mouse around caused
sever gerkyness in its progressiona nd eventually I believe it seriously
started to lose data from the mouse as it began acting as if I were
pressing the mouse buttons in extremely rapid succession whilst I was
only moving the mouse. Top was registering a load of about 4.3 and I was
about to check on /proc/interrups and vmstat when the system hung
completely. Couldn't so much as ping it.

Also, the X screen saver activated a few seconds after I did some
kboard+mouse activity instead of many minutes after any activity from
kboard and mouse. I also managed to get something logged from the erratic
mousyness that resulted:

Mar 25 17:18:25 theirongiant kernel: psmouse.c: Lost synchronization, throwing 1 bytes away.
Mar 25 17:18:27 theirongiant kernel: psmouse.c: Lost synchronization, throwing 2 bytes away.
Mar 25 17:18:40 theirongiant kernel: psmouse.c: Lost synchronization, throwing 1 bytes away.
Mar 25 17:18:42 theirongiant kernel: psmouse.c: Lost synchronization, throwing 2 bytes away.
Mar 25 17:18:47 theirongiant kernel: psmouse.c: Lost synchronization, throwing 2 bytes away.
Mar 25 17:18:49 theirongiant kernel: psmouse.c: Lost synchronization, throwing 1 bytes away.

I turned my laptop off and on and rebooted into the same kernel. ext3
journal restore was slow and at one stage seemed like it hung on one of
the partitions. It didn't and moved on to the next but then it did hang
and I had to turn the laptop off and back on as ctrl-alt-delete did not
work.

I rebooted into 2.5.63, restored my journals and booted back into 2.5.66
to write you this type of letter. During this I was recompiling the
kernel to get rid of some options I did not need. Whilst typing I
suffered severe keystroke loss (I could not type the kernels version
number even) which led to what looked like afreeze and then as soon as I
touched the mouse, the laptop shut down. Throughout all this I was
pinging the box on the local lan. The pings were erratic ranging from
0.2ms upto 7200ms. This follows a certain pattern iun that it would hit
0.2ms then jump to a high number like 2400ms and 7200ms and then the
pings would drop down by 1000ms/s until they hit 0.2ms and then jump
again.

Whilst 2.5.63 has severe problems with interactivity it does not crash
or lose keystrokes and pinging the box during a kernel compile resulst
in 'reasonable' results of 0.2ms to 0.3ms.

Same config but without anything in the Power Management menu selected:

Similar results.

A stripped down, minimal config (config-small attachment):

I couldn't really duplicate the problem. There were some interactivity
issues but that's more or less it.

Using Dave Hensen's report_lost_ticks patch did not yeild anything when
used with the first config (the big one) except for some results during
kernel initialisation (once during input dirver init and another waaay
early in the boot due to acpi - I can post these if need be).

Seeing as how the stripped down, minimal config basically worked I
created another that was smaller then the initial one but still useful
to me (I don't really want to run 2.5.63 if I can help it as its
interactivity issues are far worse then 2.6.66's (when it's not killing
my laptop :)). This is the .config-medium attachment.

The results for this config with attempting to compile the kernel, compile
mplayer, getting mozilla to load up about 40 websites in 3 windows and
constantly copying a 30MB file from the HD to an NFS server are the same
as for the minimal config (config-small). ie interactivity issues are
present but I got bored trying to type stuff and move the mouse around
in order to get the kernel to spew up on me.

I've removed all commented entries from the config files and did a diff
of the medium config versus the big one so that the diffs can easily
be seen. This is the config-diff attachment.

I'm going to stop recompiling the kernel now and crashing my laptop
(I've got 8 versions of .66 and god knows how many times I've run tests
:). I'm sure all the power downs and power ups aren't good for it but
if anyone wants me to try a new kernel with new config options to see
if it resumes being nasty, I'll happily do that. I'm just not clueful
enough to pick the right options myself without lots of trial and error
and I'm kinda tired. :) Also, if you have a patch you wish me to try,
chuck it at me and I'll see if it helps.

I do hope this helps. As usual, if you need more info or any help, please
holler or something. It sucks doing a bug report (especially if you've
put a fair bit of work into it) and getting nothing but silence back. :/

(oh yeah... I hope davej and alan don't mind being CCed - Andrew said
they'd be good at figuring out stuff like this so I put the email addies
into the CC line)

-- 
"Other countries of course, bear the same risk. But there's no doubt his
hatred is mainly directed at us. After all this is the guy who tried to
kill my dad."
        - George W. Bush Jr, Leader of the United States Regime
          September 26, 2002 (from a political fundraiser in Houston, Texas)





- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/



This archive was generated by hypermail 2b29 : Mon Mar 31 2003 - 22:00:20 EST