Re: Kernel 2.4.5-pre3

[Ken T. <ke...@we...>]

On Thu, 14 Jun 2001, Roman Zippel wrote:

> In your case I still didn't rule out bad memory. What is your memory
> configuration, how much swap do you have?

> It would help a lot if we would know that the base system is stable,
> before we go on to X.

Hello,

Not sure that I have a memory problem given the different behaviour under
2.2.10 and 2.4.5, I'll do any tests you suggest.

My system is A4000, 16 + 2M,  A2065, GVPIOExtender, A4091, CV64-3d,
060/604 + 64M. 

swapon -s reports

/dev/hda7                       partition       204776  0       0
/dev/fastram                    partition       16380   0       1

I've tried without fastram swap.

Makes no difference if I use the original virge or my modified virge
driver.

With and without A4091 SCSI.

I've posted all the following details before but just to collect them in
one place :

2.2.? is stable, never a problem fron consoles, just compiled 2.4.5 with
make -j 10 without a problem, which probably means memory and sawpping are
OK ... yes ?

Helix gnome has been unstable since I installed it, crashes always locked
up my system and required a reboot.  I remember KDE being more stable but
still played up occasionally.  As I've said previously, X crashes often
reported "Page fault in interrupt handler" in dmesgs, the call back trace
was always interrupt related but not consistent.  MagicSysReq was dead
too.  Sometimes it took several attempts, lockups and reboots to get X to
start.  When it did start, just running up and down menus a bit quickly
could crash it, if I 'nursed' it gently I could use it for a while but it
would crash eventually. Have tried many Xservers, all the same results.

Running 2.4.5, anything more than make -j 3 on a kernel causes the
compiler to report internal errors, illegal instructions and produce core
dumps, sometimes logging me out but not locking up (now I know what to
lookfor, maybe even -j 3 has problems).

But under 2.4.5, X is much more stable, probably as good or nearly as good
as another well known operating system. When it does crash, I've had about
6 so far, it returns to the initiating console and reports signal 4 and I
think I saw signal 11 on one occasion. One thing that happens is that
after X exits with a signal, the next command run, like ls, fails and
reports illegal instruction but after that all is OK, I can restart X, do
anything. 

On the 4091:

A4091 never worked reliably, still doesn't.  I spent ages  on it
with a very helpful and patient Richard Hirst.  Even mailed Dave Haynie,
he said there was an outstanding problem with ZORRO III, a bus arbitration
transition at the wrong time could lock up the bus state machine, but it
only affected those devices that did some sort of extended burst cycles,
the 4091 being the only one.

I hacked sim710 to drive the 4091, it behaved exactly like the 'big'
53c7xx driver.  

Presence or absence of 4091 driver doesn't affect the way kernels behave.

ext2 amd swap partitions are on IDE.

Ken.