random errors with bzip2

[Rodrigo Ventura]

        Hi. Let me first apologize if this message is at the wrong
mailing list, but in fact I haven't the faintest idea to where shall I
send this.

        Randomly bzip2 gives an crc error when checking a big bz2
file, e.g. linux-2.4.5.tar.bz2 (20M), in a particular machine (asus
a7v, athlon@1.2G, 256M). The same file (with the same md5sum, NFS file
system) gives no problem whatsoever on other machines. On the asus
machine, sometimes the file is checked ok, other times it fails (crc
error). When several bzip2 processes are launched at the same time for
the same file, some of them fail, others return ok. These have been my
latest observations:

        - avoiding shared IRQs (by moving PCI cards in different
slots) seems to reduce probability of error;

        - fresh reboot seems to solve the problem, but after some
hours of uptime, it comes again;

        - all machines are using 2.4.6-pre3, the file is local to the
asus machine but NFS for the others. The local filesystem is
*ReiserFS*;

        - bzip2 executable is ok (as far as md5sum);

        - it could be a memory problem, but if it were, lots of kernel
oops were expected, right?

        - could it be a process pre-emption problem? (FPU?) the kernel
is compiled for generic i586 (to be shared by several machines);

        - could it be a shared IRQ problem? I've been experiencing
several problems (freeze) related with IRQ sharing on asus boards, but
I'm not totally sure about this;

        - I've read some posts about file shifting, can it cause these
kinds of bzip problems?

        Cheers,

-- 

*** Rodrigo Martins de Matos Ventura <[e-mail]>
***  Web page: http://www.isr.ist.utl.pt/~yoda
***   Teaching Assistant and PhD Student at ISR:
***    Instituto de Sistemas e Robotica, Polo de Lisboa
***     Instituto Superior Tecnico, Lisboa, PORTUGAL
*** PGP fingerprint = 0119 AD13 9EEE 264A 3F10  31D3 89B3 C6C4 60C6 4585