I noticed that the BadVaddr in Oops info from do_fault didn't match the
faulting VA in the early part of the message:
---
vvvvvvvv
Unable to handle kernel paging request at virtual address 0000000c, epc
== 801c883c, ra == 801c8ae0
Oops in fault.c:do_page_fault, line 170:
$0 : 00000000 14001f00 8fde6a6c 8fde6a60 (zero,at,v0,v1)
$4 : 8fddb400 00000003 8fddc000 8d7a53e0 (a0-a3)
$8 : 00000003 00000000 8fe67810 8fde9a40 (t0-t3)
$12: 0000003c 8fe67820 8d7a5340 00000000 (t4-t7)
$16: 8d7a5340 8fddc000 00000001 8d7a5340 (s0-s3)
$20: 8fde2b50 8fde2ba8 8fde2940 8fde2bac (s4-s7)
$24: 8fde4000 2abb5340 8d7a5340 8fde2000 (t8,t9,k0,k1)
$28: 80106000 80107dd8 0000000d 801c8ae0 (gp,sp,fp,ra)
epc : 00000000801c883c
Status : 14001f03
Cause : 00808008
BadAddr: 000000008d7a5340 Process swapper (pid: 0, stackpage=80106000)
^^^^^^^^^^^^^^^^
---
Seems that the BadVaddr doesn't currently get stuffed into the pt_regs
on the stack for TLB exceptions. The following patches might be
reasonable:
--- tlbex-r3k.S.orig Thu Nov 1 10:50:29 2001
+++ tlbex-r3k.S Thu Nov 1 10:50:17 2001
@@ -92,6 +92,7 @@
.set macro; \
SAVE_ALL; \
mfc0 a2, CP0_BADVADDR; \
+ REG_S a2, PT_BVADDR(sp); \
STI; \
.set at; \
move a0, sp; \
--- tlbex-r4k.S.orig Thu Nov 1 10:50:44 2001
+++ tlbex-r4k.S Thu Nov 1 10:50:08 2001
@@ -337,6 +337,7 @@
.set noat; \
SAVE_ALL; \
mfc0 a2, CP0_BADVADDR; \
+ REG_S a2, PT_BVADDR(sp); \
STI; \
.set at; \
move a0, sp; \
--kip
