Panic during install on Sparc64 - Only with large HDD

[Chris Gilbert]

Hi there,

Since FreeBSD 5.2 I've been seeing these strange intermittent "DMA WRITE 
TIMEOUT" messages. They seemed to go away for a time, but now they have 
returned again on my sparc64 machine. 

I do not get them with my 40GB disk, but I do get them with my new 200GB IDE 
disk. I have tested said disk with my x86 6-CURRENT machine and it seems to 
be okay.

Also, it starts spewing these timeout messages when it tries to 
disklabel/newfs a large disk slice in the installer. It successfully creates 
a few 2, 5, and 15 GB slices, then fails on my ~135GB slice.

The installer continued past these timeout messages, then while copying files 
during the install (post slice/label) it broke into the debugger.

I have the machine hooked up via a serial cable, and waiting in the debugger.

It is accessible via a shell if someone would like to look into it.

I've seen other reports of the same, or a similar error, and I would really 
like to see it fixed. So, if there is anything I can do to help please ask.

Here is the stack trace from the debugger:

db> trace
Tracing pid 77 tid 100035 td 0xfffff8006ed3fc80
panic() at panic+0x16c
ufs_dirbad() at ufs_dirbad+0x38
ufs_lookup() at ufs_lookup+0x2fc
VOP_CACHEDLOOKUP_APV() at VOP_CACHEDLOOKUP_APV+0xb4
vfs_cache_lookup() at vfs_cache_lookup+0xd8
VOP_LOOKUP_APV() at VOP_LOOKUP_APV+0xc0
lookup() at lookup+0x55c
namei() at namei+0x41c
kern_lstat() at kern_lstat+0x24
lstat() at lstat+0x14
syscall() at syscall+0x2dc
-- syscall (190, FreeBSD ELF64, lstat) %o7=0x11980c --
userland() at 0x20d308
user trace: trap %o7=0x11980c
pc 0x20d308, sp 0x7fdffffdab1
pc 0x11dd2c, sp 0x7fdffffdcb1
pc 0x100470, sp 0x7fdffffdd71
pc 0x100194, sp 0x7fdffffde31
pc 0, sp 0x7fdffffdef1
done
db>   

-- 
Thanks,
Chris (Lance) Gilbert
Ph: +16239791302 (UTC -0700)