powerpc/mm: Fix module instruction tlb fault handling on Book-E 64
authorKumar Gala <galak@kernel.crashing.org>
Wed, 17 Nov 2010 07:20:32 +0000 (07:20 +0000)
committerBenjamin Herrenschmidt <benh@kernel.crashing.org>
Thu, 18 Nov 2010 03:54:23 +0000 (14:54 +1100)
We were seeing oops like the following when we did an rmmod on a module:

Unable to handle kernel paging request for instruction fetch
Faulting instruction address: 0x8000000000008010
Oops: Kernel access of bad area, sig: 11 [#1]
SMP NR_CPUS=2 P5020 DS
last sysfs file: /sys/devices/qman-portals.2/qman-pool.9/uevent
Modules linked in: qman_tester(-)
NIP: 8000000000008010 LR: c000000000074858 CTR: 8000000000008010
REGS: c00000002e29bab0 TRAP: 0400   Not tainted
(2.6.34.6-00744-g2d21f14)
MSR: 0000000080029000 <EE,ME,CE>  CR: 24000448  XER: 00000000
TASK = c00000007a8be600[4987] 'rmmod' THREAD: c00000002e298000 CPU: 1
GPR00: 8000000000008010 c00000002e29bd30 8000000000012798 c00000000035fb28
GPR04: 0000000000000002 0000000000000002 0000000024022428 c000000000009108
GPR08: fffffffffffffffe 800000000000a618 c0000000003c13c8 0000000000000000
GPR12: 0000000022000444 c00000000fffed00 0000000000000000 0000000000000000
GPR16: 00000000100c0000 0000000000000000 00000000100dabc8 0000000010099688
GPR20: 0000000000000000 00000000100cfc28 0000000000000000 0000000010011a44
GPR24: 00000000100017b2 0000000000000000 0000000000000000 0000000000000880
GPR28: c00000000035fb28 800000000000a7b8 c000000000376d80 c0000000003cce50
NIP [8000000000008010] .test_exit+0x0/0x10 [qman_tester]
LR [c000000000074858] .SyS_delete_module+0x1f8/0x2f0
Call Trace:
[c00000002e29bd30] [c0000000000748b4] .SyS_delete_module+0x254/0x2f0 (unreliable)
[c00000002e29be30] [c000000000000580] syscall_exit+0x0/0x2c
Instruction dump:
XXXXXXXX XXXXXXXX XXXXXXXX XXXXXXXX XXXXXXXX XXXXXXXX XXXXXXXX XXXXXXXX
38600000 4e800020 60000000 60000000 <4e80002060000000 60000000 60000000
---[ end trace 4f57124939a84dc8 ]---

This appears to be due to checking the wrong permission bits in the
instruction_tlb_miss handling if the address that faulted was in vmalloc
space.  We need to look at the supervisor execute (_PAGE_BAP_SX) bit and
not the user bit (_PAGE_BAP_UX/_PAGE_EXEC).

Also removed a branch level since it did not appear to be used.

Reported-by: Jeffrey Ladouceur <Jeffrey.Ladouceur@freescale.com>
Signed-off-by: Kumar Gala <galak@kernel.crashing.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
arch/powerpc/mm/tlb_low_64e.S

index 8b04c54e596f2a5a1bf770cb44013808eaf27246..8526bd9d2aa33603be03dbf10bc3656c7a833448 100644 (file)
        cmpldi  cr0,r15,0                       /* Check for user region */
        std     r14,EX_TLB_ESR(r12)             /* write crazy -1 to frame */
        beq     normal_tlb_miss
+
+       li      r11,_PAGE_PRESENT|_PAGE_BAP_SX  /* Base perm */
+       oris    r11,r11,_PAGE_ACCESSED@h
        /* XXX replace the RMW cycles with immediate loads + writes */
-1:     mfspr   r10,SPRN_MAS1
+       mfspr   r10,SPRN_MAS1
        cmpldi  cr0,r15,8                       /* Check for vmalloc region */
        rlwinm  r10,r10,0,16,1                  /* Clear TID */
        mtspr   SPRN_MAS1,r10