Re: fifo panic (was: Panic when building the world under SMP kernel (HTT).)

>>> Just to report that i always encountered a panic when building the
>>> world using an SMP kernel for a CPU with HTT enable (in the BIOS)
>>> and the sysctl machdep.hlt_logical_cpus set to 0.
>>>
>>> Just cvsup'ed (kernel built at Fri Sep 9 18:39:41 CEST 2005).
>>> See the kernel trace and kernel configuration file attached, please.

>> Do you have a crash dump for this crash, and a kernel with debugging
>> symbols?  There have been occasional fifo panic reports from SMP boxes
>> (especially >=4 CPUs doing builds), and I'd like to try to track it
>> down, but based on prior experience, we'll need some amount of detailed
>> debugging information to get much further.

> (I should have say that this box follow the RELENG_6 branch.)
>
> Yes, i have a kernel with debugging symbols.  Sadly, it seems that the
> core file is unusable (maybe due to the "write_null" error on notebook
> disk, see below):
>
> /* From a serial console */
> db> panic
> panic: from debugger
> cpuid = 0
> Uptime: 22m7s
> Dumping 1021 MB (2 chunks)
> ad0: timeout waiting to issue command
> ad0: error issueing WRITE_MUL command
>   chunk 0: 1MB (159 pages)ad0: timeout waiting to issue command
> ad0: error issueing WRITE_MUL command
>  ... ok
>   chunk 1: 1021MB (261360 pages) 1005 989 973 957 941 925 909 893 877 861
> 845 829 813 797 781 765 749 733 717 701 685 669 653 637 621 605 589 573
> 557 541 525 509 493 477 461 445 429 413 397 381 365 349 333 317 301 285
> 269 253 237 221 205 189 173 157 141 125 109 93 77 61 45 29 13 ... ok
>
> /* After a reboot */
> # cd /var/crash
> # gzip -d vmcore.18.gz
> # kgdb /usr/obj/usr/src/sys/BOBOCHE/kernel.debug vmcore.18
> kgdb: cannot read IdlePTD
>
> I think this timeout is a "side effect" of some energy saving mecanism
> on the disk.  If someone can help me bypassing this timeout problem, i
> may provide a crash dump ASAP (i will try to reproduce the panic sitting
> near the machine to get a dump typing `panic' _just_ after the crash).

Some dozens minutes after having sent this, the box encountered another
panic: always during a buildworld test, it seems to be the same fifo
panic.  Fortunately, i had a crash dump this time.

-- 
-jpeg.