amd64: Fatal Trap 12 in high load situations

[ms80]

Am Saturday 06 February 2010 10:17:05 schrob ms80:
> Hi
> 
> I have a problem installing / upgrading FreeBSD 8.0-release on a new
>  machine.
> 
> The computers specs are:
> 
> cpu: AMD Phenom II X4
> board: Gigabyte MA790GPT-UD3H
> ram: 4x2GBytes DDR3/1333
> hdd: 2xMaxtor STM31000528AS
> nic: 4x Intel(R) PRO/1000
> 
> and I'm running
> FreeBSD phenom2.localnet 8.0-RELEASE FreeBSD 8.0-RELEASE #0: Sat Nov 21
> 15:02:08 UTC 2009    
>  root at mason.cse.buffalo.edu:/usr/obj/usr/src/sys/GENERIC amd64
> 
> During 'make buildworld' the machine regulary crashes with the following
> panic:
> 
> Fatal trap 12: page fault while in kernel mode
> cpuid = 0; apic id = 00
> fault virtual adress		= 0x80000
> fault code				= supervisor write data, page not present
> instruction pointer		= 0x20:0xffffffff80578591
> stack pointer			= 0x28:0xffffff80eab94700
> frame pointer			= 0x28:0xffffff80eab94720
> code segment			= base 0x0, limit 0xfffff, type 0x1b
> 					= DPL 0, pres 1, long 1, def32 0, gran 1
> processor eflags		= interrupt enabled, resume IOPL = 0
> current process		= 22039 (uudecode)
> trap number			= 12
> panic: pagefault
> cpuid = 0
> Uptime: 2h35m4s
> Physical memory: 8176 MB
> Dumping 2195 MB: 2180 2164 2148 2132 2116
> 
> 
> or this one, its from last night and the machine wrote a minidump before
> locking up:
> 
> Fatal trap 12: page fault while in kernel mode
> cpuid = 0; apic id = 00
> fault virtual address		= 0x80000
> fault code              		= supervisor write data, page not present
> instruction pointer  		= 0x20:0xffffffff80578591
> stack pointer           		= 0x28:0xffffff80eab21500
> frame pointer          	 	= 0x28:0xffffff80eab21520
> code segment        		= base 0x0, limit 0xfffff, type 0x1b
>                         		= DPL 0, pres 1, long 1, def32 0, gran 1
> processor eflags     		= interrupt enabled, resume, IOPL = 0
> current process      		= 5238 (objcopy)
> trap number           		= 12
> panic: page fault
> cpuid = 0
> Uptime: 1h15m45s
> Physical memory: 8176 MB
> Dumping 2148 MB: 2133 2117 2101 2085 2069 2053 2037 2021 2005 1989 1973
>  1957 1941 1925 1909 1893 1877 1861 1845 1829 1813 1797 1781 1765 1749 1733
>  1717 1701 1685 1669 1653 1637 1621 1605 1589 1573 1557 1541 1525 1509 1493
>  1477 1461 1445 1429 1413 1397 1381 1365 1349 1333 1317 1301 1285 1269 1253
>  1237 1221 1205 1189 1173 1157 1141 1125 1109 1093 1077 1061 1045 1029 1013
>  997 981 965 949 933 917 901 885 869 853 837 821 805 789 773 757 741 725
>  709 693 677 661 645 629 613 597 581 565 549 533 517 501 485 469 453 437
>  421 405 389 373 357 341 325 309 293 277 261 245 229 213 197 181 165 149
>  133 117 101 85 69 53 37 21 5
> 
[snip]

I know, its kind of stupid to reply to my own mails, but for reference:
I edited loader.conf to contain 
ahci_load="YES"

So far it works: The machine compiled all night and didn't crash.

I had the idea because yesterday while testing the proposal to lower the ddr3 
voltages, the machine crashed again. Additionally to the panic I'm already 
used to, I had a second panic in my core.txt.1: This was a fatal trap 1, 
referencing (current process) to irq 22. I checked what irq22 is and it is my 
atapci (ATI IXP700/800 SATA300 controller). Googling a bit around I found a 
tutorial how to activate ahci. I gave it a try and as said above: So far it 
seems to work.

regards

Sven
--
00