DO_FAULT patch

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:

@@ -92,6 +92,7 @@
        .set    macro; \
        SAVE_ALL; \
        mfc0    a2, CP0_BADVADDR; \
+       REG_S   a2, PT_BVADDR(sp); \
        STI; \
        .set    at; \
        move    a0, sp; \

@@ -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