Re: Mysteriously disappearing 2.0.33

G. Sumner Hayes (zumner@collegium.adsl.net.cmu.edu)
Sun, 4 Jan 1998 13:08:12 -0500


On Sun, Jan 04, 1998 at 02:02:17PM +0000, Klaus Lichtenwalder wrote:
> Hi,
> this is not really a bug report, but more or less a shot in the dark, as
> I don't know what really happened. Here's the setup:
>
> Linux 2.0.33 fairly generic, gcc 2.7.2.1
> ISDN Subsystem (Elsa card), i82371PIIX enabled,
> AHA2940, 3 disks, 3c59x v.0.46C for 3c590 Vortex.
> 96MB Ram
[SNIP]
> The machine simply crashed after an uptime of 14 days. Nothing in the
> logs. It looks like one isdn interface and one asynchronous ppp
> interface were active at that time.
>
> Guess I recompile the kernel with the memory-leak patch? Anybody else
> has ideas?

I've had similar problems with 2.0.32 and 2.0.33; with 2.0.29 I had
uptimes of 50+ days (rebooted only to upgrade the kernel), no problems
with earlier 2.0.x or 1.2.x kernels. 1.2.13 (with AIC7xxx support
patch) was up for 100+ days without problems.

There are no logged messages and nothing appears on the console (I turned
off the screen blanker and left the machine on console while on vacation;
nothing appeared); the machine freezes completely, no Ctrl-Alt-Del, no
Shift-Scrollock, no vt switching. The biggest uptime I've managed with
2.0.33 is 5 days.

The only hardware similarity is that I have an Adaptec 2940U; if I recall
correctly there were updates to the 2940 drivers between 2.0.29 and
2.0.32. I have no reason to suspect the 2940 other than the fact that
you also have one and that other people seem to have problems with them
under Linux; I've run this one without problems since a year and a half
ago until now.

I also have an NE2000 Ethernet card, which hasn't been a problem in all the
time I've used Linux (since 1.1.54 or thereabouts).

Pentium Pro 180, 64MB, Adaptec 2940U, PS/2 mouse, NE2000 ISA Ethernet,
unpatched 2.0.33 kernel (2.0.32 crashed as well).

No hardware change; works fine with 2.0.29 but dies within two or three
days on average with 2.0.32 and 2.0.33 using modular and monolithic
kernels. gcc 2.7.2.3. (patching for egcs and compiling with egcs also
exhibits the problem)

Obviously this is an experimental system with hardware like that in it,
so I'm willing to try out experimental patches &c. It would be nice to
get it stable again, though. :)

-Sumner

-- 
rage, rage against the dying of the light