G H wrote:
At the moment I don't have the board set up for using kgdb and it's
complicated by the fact that we only have one serial console port. But
I am looking into setting it up for kgdb now.
A single serial port will work fine.
As far as stressing the system, it doesn't have enough resources (
disk space ) to be able to compile the kernel, but we did write a
simple program that would stress the system by spawning multiple
threads, each one performing floating point calculations. With this
test, top reported a load average of over 400 and we have
seen no failure so far.
You can use an NFS mounted root fs to do native builds, assuming you
have a native toolchain. That's an excellent stress test.
Pete
*/Pete Popov <ppopov@mvista.com>/* wrote:
G H wrote:
> I've not had much response to this question so I would like to
> rephrase it :
>
> Can anyone think of any possible scenario where do_ri could
occur in
> blast_icache32() ??
>
> Is this possibly a cache synchronisation problem ??
>
Could be a hardware memory glitch. I would use kgdb to put a
breakpoint
there and see what the data in memory looks like when this happens --
look for memory corruption, etc.
Pete
> TIA
>
> >While testing out an amd au1500 based board I have been getting "
> do_ri " exceptions >that always occur in the cache flushing
routines.
> More often than not in >blast_icache_32().
>
> >So far this has mainly happened after running the board for
days on
> end while running >multiple telnet sessions t! o it. It has
sometimes (
> quite rarely ) happened after a few >hours to a day of multiple
telnet
> session use.
>
> __________________________________________________
> Do You Yahoo!?
> Tired of spam? Yahoo! Mail has the best spam protection around
> http://mail.yahoo.com
>
------------------------------------------------------------------------
Do you Yahoo!?
Yahoo! Mail Address AutoComplete
<http://us.rd.yahoo.com/mail_us/taglines/aac/*http://promotions.yahoo.com/new_mail/static/ease.html>
- You start. We finish.
|