[parisc-linux] HPMC when doing sync on c3k (32 bit kernel)

[parisc-linux] HPMC when doing sync on c3k (32 bit kernel)

[Randolph Chung]

I noticed that when i do a manual 'sync' on my c3000 running a 32-bit
kernel, I get a HPMC:


High Priority Machine Check (HPMC): Code=1 regs=1029c080 (Addr=00000000)

     YZrvWESTHLNXBCVMcbcbcbcbOGFRQPDI
PSW: 00000000000001101111110100001100
r0-3     00000000 102a9010 1010e228 000000ff
r4-7     00000000 00000000 102ad010 10355074
r8-11    10355010 10354810 0004000e 102d4730
r12-15   10349810 10349010 103495dc 102ad010
r16-19   1032c580 f000017c f0000174 0000000e
r20-23   1036f140 000003fd 1010e20c 00000000
r24-27   000003fd fee003fd fee003fd 10292010
r28-31   00000000 00000205 1032cb00 1010cc2c
sr0-3    00000000 0000027c 00000000 00000009
sr4-7    00000000 00000000 00000000 00000000

IASQ: 00000000 00000000 IAOQ: 10207568 1020756c
 IIR: 0f40001c    ISR: a607fffb  IOR: 802003fd
 CPU:        0   CR30: 1032c000 CR31: 10330000
 ORIG_R28: 00000000

"pim hpmc" says:
A Data I/O Fetch Timeout occurred while CPU 0 was
requesting information from a device at the path 10/0/14/1/1 (built-in
PCI device)

apparently it died in _gsc_readb.

With a 64-bit kernel it works ok.

Anyone else seeing this/know what's going on?

randolph
-- 
   @..@                                         http://www.TauSq.org/
  (----)
 ( >__< )
 ^^ ~~ ^^

Re: [parisc-linux] HPMC when doing sync on c3k (32 bit kernel)

[Grant Grundler]

Randolph Chung wrote:
> I noticed that when i do a manual 'sync' on my c3000 running a 32-bit
> kernel, I get a HPMC:
...
> "pim hpmc" says:
> A Data I/O Fetch Timeout occurred while CPU 0 was
> requesting information from a device at the path 10/0/14/1/1 (built-in
> PCI device)

That's the SuckyIO serial port. Randolph and I exchanged notes offline
and concluded his prototype C3000 has PA8500 v2.0 CPU (and newer elroys
than mine). I can't reproduce the problem with a v2.3 chip.  I should be
able to upgrade his CPU to v2.3 and we'll see if the problem persists.
Lots of bugs in pre-2.3 chip revisions.
(BTW, does that answer your question Helge? :^/)

grant