Re: smp crash under network load

Fri, Dec 06, 2002 at 10:34:11AM +0100, Frank van der Linden:
> On Thu, Dec 05, 2002 at 10:15:09PM +1030, Brett Lymn wrote:
> > 
> > Folks,
> >         I have just brought my dual PPro up in mp mode, now when I do
> > any moderately heavy network traffic the machine drops to ddb with a
> > message of:
> > 
> >  "ex0: uplistptr was 0"
> 
> That's not a fatal error, it won't drop you into DDB.. Please make sure
> that your sources are completely up to date, and if you're still seeing
> the problem, send a PR with tracebacks.

i had a number of these on my tyan S2466.  i doubled EX_NUPD

Index: sys/dev/ic/elinkxlreg.h
===================================================================
RCS file: /cvsroot/syssrc/sys/dev/ic/elinkxlreg.h,v
retrieving revision 1.11
diff -r1.11 elinkxlreg.h
217c217
< #define EX_NUPD               128
---
> #define EX_NUPD               256

and it seemed to stop, but panics in ex_intr() did not.  below is one
trace that i caught about 3 weeks ago, after which i stuck an fxp in
the box.  i dont think this SMP related.

login: kernel: page fault trap, code=0
Stopped at      ex_intr+0x1c1:  movl    %eax,         0(%edx)
db{1}> 
db{1}> trace
ex_intr(c164b000) at ex_intr+0x1c1
Xintrioapic4() at Xintrioapic4+0x69
--- interrupt ---
idle_loop+0xc:

db{1}> show reg
ds                0x10
es                0x10
fs                0x30
gs                0x10
edi         0xc164b000  end+0x12c7e64
esi         0xc1667e00  end+0x12e4c64
ebp         0xe3446f48
ebx                  0
edx                  0
ecx         0xc164cd10  end+0x12c9b74
eax         0xc164cd10  end+0x12c9b74
eip         0xc014f3b9  ex_intr+0x1c1
cs                 0x8
eflags         0x10282  gdt_desc+0xe1d6
esp         0xe3446f10
ss                0x10
ex_intr+0x1c1:  movl    %eax,         0(%edx)

db{1}> sync
syncing disks... ex0: uplistptr was 0
kernel: page fault trap, code=0
Stopped at      genfs_putpages+0x30f:   movl         0x10c(%eax),%edx