crash in fault.c

(apologies for the previous blank message)

I get this randomly and frequently, usually when booting, attempting to ftp
to the box, or occasionally when starting a new process.

Any hints or clues are appreciated.

I'm using kernel 2.4.4 running on a custom 8260 based board.

Oops: Exception in kernel mode, sig: 4
NIP: C000BE80 XER: 00000000 LR: C0003DC4 SP: C3D5FDC0 REGS: c3d5fd10

TRAP: 0700
MSR: 00089032 EE: 1 PR: 0 FP: 0 ME: 1 IR/DR: 11
TASK = c3d5e000[1] 'init' Last syscall: 13
last math c3d5e000 last altivec 00000000
GPR00: 00000400 C3D5FDC0 C3D5E000 C3D5FE80 0FF2FF1C 2000F932 1004CDE8 0FFDD688
GPR08: 0FFDE2D8 1004CDE8 00001000 7FFFF3C0 48444888 1004BB08 03FDF000 007FFF16
GPR16: 00000000 007FFF00 007FFEB0 03FD824C 00009032 03D5FE70 00000000 C0003DC4
GPR24: C000BE48 00000000 00030001 2000F932 C3D5FE80 0FF2FF1C 0FFE8248 C01270A0
Call backtrace:
C000BFD8 C0003DC4 0FF25DD0 0FF5B820 0FF5A56C 0FF5B47C 0FF57ED0 0FF8C440 0FF8C270 1000E1F4 1000E658 1000F30C 10027810 1002748C 0FEDC068 00000000

Kernel panic: Attempted to kill init!

closest to c000be80 :: c000be48 T do_page_fault (plus 56 of 904)
closest to c000bfd8 :: c000be48 T do_page_fault (plus 400 of 904)
closest to c0003dc4 :: c0003dc4 T ret_from_except (plus 0 of 40)

78:fault.c **** if (regs->trap == 0x400)

 967                    .stabn 68,0,78,.LM3-do_page_fault
 968                    .LM3:
 969 0018 801C00A0      lwz 0,160(28)
 970                    .stabn 68,0,66,.LM4-do_page_fault
 971                    .LM4:
 972 001c 3F400003      lis 26,0x3
 973                    .stabn 68,0,62,.LM5-do_page_fault
 974                    .LM5:
 975 0020 7C9D2378      mr 29,4
 976                    .stabn 68,0,78,.LM6-do_page_fault
 977                    .LM6:
 978 0024 2C000400      cmpwi 0,0,1024
 979                    .stabn 68,0,62,.LM7-do_page_fault
 980                    .LM7:
 981 0028 7CBB2B78      mr 27,5
 982                    .stabn 68,0,66,.LM8-do_page_fault
 983                    .LM8:
 984 002c 635A0001      ori 26,26,1
 985                    .stabn 68,0,70,.LM9-do_page_fault
 986                    .LM9:
 987 0030 3B200000      li 25,0
 988                    .stabn 68,0,78,.LM10-do_page_fault
 989                    .LM10:
 990 0034 4082000C      bc 4,2,.L868
  79:fault.c       ****                 error_code &= 0x48200000;
 991                    .stabn 68,0,79,.LM11-do_page_fault
 992                    .LM11:
 993 0038 777B4820      andis. 27,27,0x4820
 994 003c 48000008      b .L869
 995                    .L868:
  80:fault.c       ****         else
  81:fault.c       ****                 is_write = error_code &
0x02000000;
 996                    .stabn 68,0,81,.LM12-do_page_fault
 997                    .LM12:
 998 0040 5779018C      rlwinm 25,27,0,6,6
 999                    .L869:

And some more...

149:fault.c **** switch (handle_mm_fault(mm, vma,
address, is_write)) {

 1200                   .stabn 68,0,149,.LM53-do_page_fault
 1201                   .LM53:
 1202 017c 7C641B78     mr 4,3
 1203 0180 7F26CB78     mr 6,25
 1204 0184 7FE3FB78     mr 3,31
 1205 0188 7FA5EB78     mr 5,29
 1206 018c 48000001     bl handle_mm_fault
 1207 0190 2C030001     cmpwi 0,3,1
 1208 0194 41820020     bc 12,2,.L898
 1209 0198 41810010     bc 12,1,.L905
 1210 019c 2C030000     cmpwi 0,3,0
 1211 01a0 4182015C     bc 12,2,.L901
 1212 01a4 480000F4     b .L903
 1213                   .L905:
 1214 01a8 2C030002     cmpwi 0,3,2
 1215 01ac 41820018     bc 12,2,.L899
 1216 01b0 480000E8     b .L903
 1217                   .L898:
 150:fault.c       ****         case 1:
 151:fault.c       ****                 current->min_flt++;
 1218                   .stabn 68,0,151,.LM54-do_page_fault
 1219                   .LM54:
 1220 01b4 81220104     lwz 9,260(2)
 1221 01b8 39290001     addi 9,9,1
 1222 01bc 91220104     stw 9,260(2)

152:fault.c **** break;

--
Adam Wozniak (KG6GZR) COM DEV Wireless - Digital and Software Systems
[email protected] 3450 Broad St. 107, San Luis Obispo, CA 93401

                        http://www.comdev.cc
                        Voice: (805) 544-1089       Fax: (805) 544-2055

** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Does the 8260 have BATs ?

If not, can you check if the code for ret_from_except is entirely contained in a single page or spans over 2 pages ? (objdump should help here).

If it is the case, then you need some of my fixes to make the exception return path safe against TLB misses (or hash misses, dunno if you have a hash table on this CPU).

Ben.

** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Yes. It's a standard 603e core with BATs and no HPTE.

I suspect this is a hardware configuration problem in the memory controller or memory interface. As I have always said, random problems that occur when the network is used usually point to that. The reason is you can get back to back burst mode transfers to the memory, something you aren't likely to see with just normal CPU usage.

	-- Dan

** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Ok. I'm a little new to this. How do I start troubleshooting "a hardware configuration problem in the memory controller or memory interface" ?

I'm using PPCBoot for a boot monitor, and VisionProbe for a JTAG debugger. I've run a number of RAM tests, and don't seem to have any problems reading/writing to RAM. What kind of thing am I looking for?

--Adam
--
Adam Wozniak (KG6GZR) COM DEV Wireless - Digital and Software Systems
[email protected] 3450 Broad St. 107, San Luis Obispo, CA 93401

                        http://www.comdev.cc
                        Voice: (805) 544-1089       Fax: (805) 544-2055

** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/