Re: Weird lockups
Greg Johnson wrote:
> On Tue, Jan 18, 2000 at 09:14:49AM -0600, David Winchell wrote:
> > Since the machine doesn't respond to sysrq, it may be stuck in an interrupt handler.
> > If any of the machines in question have "halt" buttons (machines from
> > Compaq should have them) I would like to know if pressing halt
> > gets one back to the console ">>>" prompt. If it does, one can use
> > console commands to dump the pc, ra, sp. Also, the stack can be dumped
> > from the console. If its a multi-processor, one should halt all the processors at the console
> > and get the registers on the various processors.
>
> I have occasionally had the machine lock while not in X11. What happens
> is that I get the message "Machine check while in PALmode" and am
> returned to the SRM prompt. I have tried examining the registers and
> see no clear pattern. Sometimes the return address is inside the kernel
> (various places) and sometimes it isn't. How does one dump the stack
> from the console?
>
I'm not sure what to do with the Machine Check. It could be a hardware problem.
Are you running the latest firmware?
For the stack, one should be able to use the examine command - type
help examine - to look at memory starting at the address found for register
sp. Getting the sp for a machine that has been halted via halt button is
simply "examine sp" but getting the sp for the Machine Check
would probably mean looking at any data that is printed on the console
or saved in memory. There may be a console command that you can issue
to dump Machine Check data.
-Dave
>
> Greg
>
> >
> > Something like this:
> >
> > [hit halt button]
> > >> e pc
> > >> e ra
> > >> e sp
> > >> halt 1 /* if smp */
> > >> set host 1 (or set cpu 1 - I forget)
> > >> e pc
> > etc
> >
> > Now, under Compaq Unix one would eventually
> > do
> >
> > >> set cpu 0
> > >> crash
> >
> > and this would generate a crash dump. Crash dumps
> > under Linux are under development. I am working
> > one one and SGI has another.
> >
> > I will soon, within the next few days, release an "in-memory"
> > crash dump for Linux. I have not yet figured out how
> > to plug the crash command at the console into Linux kernel code.
> > If anyone knows how to do this, please let me know.
> >
> > -Dave
> >
> >
> >
> >
> > "Matthew R. Pavlovich" wrote:
> >
> > > > Just a "me too". I have had hard lockups with the following system:
> > > >
> > > > 164LX2 533MHz
> > >
> > > Another 21164 system. This is different from the PC164's. Anyone know
> > > what the difference is? Which sub-arch would the LX2 use?
> > >
> > > > 4GB Seagate IDE drive
> > > > 8MB ELSA Gloria Synergy video card.
> > > > DEC Tulip NIC (Kingston KNE100TX)
> > > > Debian 2.2
> > >
> > > Have you updated recently? The autobuilder is back and a good number of
> > > packages have been updated.
> > >
> > > > more often when running X. I have tried swapping all removable parts of
> > >
> > > I think we are seeing a lot of this... Here's what I've summarized:
> > >
> > > BUG-- HARD system locks with Alpha processors.
> > >
> > > Description- Unrecoverable (SysReq fails) system lockups at various,
> > > unpredictable, unrepeatable times. Users have estimated error could be
> > > with-
> > >
> > > 3 PC164's 1 LX2
> > > IDE controller support for alphas.
> > > CMD646 IDE controller specifically
> > > X - 2 ppl w/ G200 PCI's 1 w/ ELSA Gloria Synergy
> > > 2 w/ 3Com 905b's
> > > 2 w/ NCR 53C875 SCSI Controllers
> > >
> > > Diagnostics performed- All users have tried different hardware
> > > configurations, including replacing all IDE hardware w/ SCSI.
> > >
> > > Notes- Some report problems with module loading, or modular kernels.
> > >
> > > Kernel versions- Various 2.2.x.. including 2.2.13 and 2.2.14.
> > >
> > > Matthew R. Pavlovich
> >
> > --
> > David Winchell
> > Chief Technology Officer
> > Mission Critical Linux
> > http://www.missioncriticallinux.com
> >
> >
>
> --
> Greg Johnson gjohnson@physics.clarku.edu
> http://physics.clarku.edu/~gjohnson finger for PGP key
Reply to: