Port-sparc64 archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]

Re: SMP bugs ( was Re: Severe deadlock issues with 5.0/MP )

On Tue, 3 Feb 2009, Michael wrote:

> On Feb 2, 2009, at 11:32 AM, matthew green wrote:
> > when you get the SIR, do you get a prom prompt?
> Yes.
> > if so, please type ctrace at it and figure out the stack trace?
> Watchdog Reset
> Externally Initiated Reset
> {2} ok ctrace
> PC: Fast Instruction Access MMU Miss

That's not going to work too good.  When the system takes a reset it loses 
all the MMU mappings so ctrace won't work.

What you can do is dump the registers (.registers) and all the register 
windows (n .window).  Then take the address of the pc and use gdb on the 
kernel do figure out which instruction in locore.s it corresponds to.  
There's usually a comment in locore.s near the sir instruction that 
explains why the kernel was not able to recover from whatever the 
processor state was at the time.  That's when the real debugging starts, 
'cause you need to try to figure out how the kernel got all stuffed up, 
while trying to figure out what parts of the processor state are from the 
crash and what bits are left over from earlier or were overwritten as part 
of the SIR recovery operation.

Good luck.


Home | Main Index | Thread Index | Old Index