Re: Crashdump support in PPC?

[Donald Lee <MacPPC2%caution.icompute.com@localhost>]

>Hello,
>
>On Sat, 1 Oct 2011 19:03:59 -0500
>Donald Lee <MacPPC2%caution.icompute.com@localhost> wrote:
>
>> Do we have crashdump support in MacPPC NetBSD 5.0.2? (unmodified!)
>> 
>> I have a panic, (again).  It is a production machine, and I'd
>> like to take a closer look at the crash dump, but the panic/reboot did
>> not appear to produce one.  All I got was what I typed in from the
>> console:
>> 
>> trap: kernel read DSI trap @ 0xde0c5010 by 0x3ec85c (DSISR 0x40000000, 
>> err=14), lr 0x3ec850
>> panic: trap
>> Begin traceback...
>> 0xd45c7d00: at traap+0x108
>> 0xd45c7d90: kernel DSI read trap at 0xde0c5010 by turnstile_block+0x140: 
>> srr1=0x9032
>>      r1=0xd45c7e50 cr=0x84000044 xer=0 ctr=0x3d0778 dsisr=0x40000000
>> 0xd45c7e50: at turnstile_block+0xd8
>> 0xd45c7e80: at mutex_vector_enter+0x184
>> 0xd45c7eb0: at callout_softclock+0x1a8
>> 0xd45c7f10: at softint_thread+0xf4
>> 0xd45c7f40: at emptyidlespin+0x10
>> 0xd45c7f50: at aprint_error_internal+0x78
>> trap: pid 0.5 (system): kernel MCHK trap @ 0x49eba8 (SRR1=0x149030)
>> panic: trap
>> Faulted in mid-traceback; aborting... dumpsys: TBD
>
>Is this an SMP box? If it is, can you try to repeat this with only one CPU 
>online?

Sorry.  It's a single CPU.  Power Mac G4 "Quicksilver" machine.

I have seen this twice so far - I think.
Both times, I was using Netatalk to move large quantities of data
both to and from the box at the same time.  (two streams, one reading,
one writing to shared disks)

Since the disk handling has
proven quite stable, I'm betting it's the networking/ethernet that's
the most likely area with the problem.  I have a GigE card in a PCI slot.

mercy$ ifconfig -a
fwip0: flags=8802<BROADCAST,SIMPLEX,MULTICAST> mtu 1500
        address: 00:03:93:ff:fe:9e:f3:b2
gem0: flags=8863<UP,BROADCAST,NOTRAILERS,RUNNING,SIMPLEX,MULTICAST> mtu 1500
        capabilities=800<TCP4CSUM_Tx>
        enabled=0
        address: 00:03:93:9e:f3:b2
        media: Ethernet autoselect (1000baseT full-duplex)
        status: active
        inet 72.1.155.195 netmask 0xffffffe0 broadcast 72.1.155.223
        inet6 fe80::203:93ff:fe9e:f3b2%gem0 prefixlen 64 scopeid 0x2

and dmesg:

...
gem0 at pci2 dev 15 function 0: Apple Computer GMAC Ethernet (rev. 0x01)
gem0: interrupting at irq 41
brgphy0 at gem0 phy 0: BCM5421 1000BASE-T media interface, rev. 1
brgphy0: 10baseT, 10baseT-FDX, 100baseTX, 100baseTX-FDX, 1000baseT, 1000baseT-FD
X, auto
gem0: Ethernet address 00:03:93:9e:f3:b2, 10KB RX fifo, 4KB TX fifo
...



>> When this panic happens, the USB keyboard is useless, so
>> I can't ask for kernel debugger ops.  Whatever happens has to
>> happen automagically.
>
>This has been fixed in -current, although that probably won't help - the soft 
>interrupt code changed too.
>

Bummer.

>> Do I have to do something special to get this to happen?  Is it supported
>> in MacPPC NetBSD?
>
>There is no crashdump support in arch/powerpc.

Bummer again.

Maybe upgrading to 5.1 would help......

Thanks much for the help,

-dgl-