RE: mm: kernel BUG at mm/rmap.c:530

From: Kirill A. Shutemov
Date: Wed Aug 20 2014 - 10:04:19 EST


Sasha Levin wrote:
> Hi all,
>
> While fuzzing with trinity inside a KVM tools guest running the latest -next
> kernel, I've stumbled on the following spew:
>
> [ 2581.180086] kernel BUG at mm/rmap.c:530!

Page is mapped where it shouldn't be. Or vma/struct page/pgtable is corrupted.
Basically, I have no idea what happend :-P

We really should dump page and vma info there. It's strange we don't have
dump_vma() helper yet.

> [ 2581.180086] invalid opcode: 0000 [#1]
> [ 2581.180086] PREEMPT SMP DEBUG_PAGEALLOC
> [ 2581.180086] Dumping ftrace buffer:
> [ 2581.180086] (ftrace buffer empty)
> [ 2581.180086] Modules linked in:
> [ 2581.180086] CPU: 13 PID: 8515 Comm: trinity-main Not tainted 3.16.0-next-20140815-sasha-00034-g615561b #1071
> [ 2581.180086] task: ffff8804c1b30000 ti: ffff8804bd9e4000 task.ti: ffff8804bd9e4000
> [ 2581.180086] RIP: rmap_walk (mm/rmap.c:530 mm/rmap.c:1675 mm/rmap.c:1707)
> [ 2581.180086] RSP: 0018:ffff8804bd9e7bb8 EFLAGS: 00010206
> [ 2581.180086] RAX: 0000000000000000 RBX: ffffea000b39e3c0 RCX: ffff8803c501fb18
> [ 2581.180086] RDX: 00007fffffffd000 RSI: 00000007fffffffd RDI: ffffea000b39e3c0
> [ 2581.180086] RBP: ffff8804bd9e7bf0 R08: ffff880254972200 R09: 0000000000000000
> [ 2581.180086] R10: 0000000000000001 R11: 0000000000000008 R12: ffff8804bd9e7c00
> [ 2581.180086] R13: ffff8801e2d40ff0 R14: 00000007fffffffd R15: ffff880254972200
> [ 2581.180086] FS: 00007fb53e50d700(0000) GS:ffff8804ca200000(0000) knlGS:0000000000000000
> [ 2581.180086] CS: 0010 DS: 0000 ES: 0000 CR0: 000000008005003b
> [ 2581.180086] CR2: 00000000025e5000 CR3: 00000004c19fa000 CR4: 00000000000006a0
> [ 2581.180086] Stack:
> [ 2581.180086] ffff8804c125a400 ffff8804bd9e7bf8 ffffea000b39e3c0 ffffea0020b86cc0
> [ 2581.180086] 0000000000000001 0000000000000000 0000000000000000 ffff8804bd9e7c30
> [ 2581.180086] ffffffffab30e7b1 ffffea0020b86cc0 ffffffffab30f050 0000000000000000
> [ 2581.180086] Call Trace:
> [ 2581.180086] remove_migration_ptes (mm/migrate.c:222)
> [ 2581.180086] ? __migration_entry_wait.isra.25 (mm/migrate.c:107)
> [ 2581.180086] ? remove_migration_pte (mm/migrate.c:193)
> [ 2581.180086] move_to_new_page (mm/migrate.c:785)
> [ 2581.180086] ? try_to_unmap (mm/rmap.c:1527)
> [ 2581.180086] ? try_to_unmap_nonlinear (mm/rmap.c:1124)
> [ 2581.180086] ? invalid_migration_vma (mm/rmap.c:1483)
> [ 2581.273353] ? page_remove_rmap (mm/rmap.c:1391)
> [ 2581.273353] migrate_pages (mm/migrate.c:916 mm/migrate.c:953 mm/migrate.c:1141)
> [ 2581.273353] ? buffer_migrate_lock_buffers (mm/migrate.c:1589)
> [ 2581.273353] migrate_misplaced_page (mm/migrate.c:1749)
> [ 2581.273353] handle_mm_fault (mm/memory.c:3175 mm/memory.c:3228 mm/memory.c:3341 mm/memory.c:3370)
> [ 2581.273353] ? __lock_is_held (kernel/locking/lockdep.c:3518)
> [ 2581.273353] __do_page_fault (arch/x86/mm/fault.c:1231)
> [ 2581.273353] ? put_lock_stats.isra.13 (./arch/x86/include/asm/preempt.h:98 kernel/locking/lockdep.c:254)
> [ 2581.273353] ? vtime_account_user (kernel/sched/cputime.c:687)
> [ 2581.273353] ? context_tracking_user_exit (kernel/context_tracking.c:184)
> [ 2581.273353] ? __this_cpu_preempt_check (lib/smp_processor_id.c:63)
> [ 2581.273353] ? trace_hardirqs_off_caller (kernel/locking/lockdep.c:2641 (discriminator 8))
> [ 2581.273353] trace_do_page_fault (arch/x86/mm/fault.c:1314 include/linux/jump_label.h:114 include/linux/context_tracking_state.h:27 include/linux/context_tracking.h:45 arch/x86/mm/fault.c:1315)
> [ 2581.273353] do_async_page_fault (arch/x86/kernel/kvm.c:279)
> [ 2581.273353] async_page_fault (arch/x86/kernel/entry_64.S:1313)
> [ 2581.273353] Code: 17 02 00 00 49 8b 14 24 4c 89 ee 48 89 df ff d0 48 8b 7d c8 89 45 d0 e8 77 24 ee ff 8b 45 d0 e9 38 01 00 00 0f 1f 80 00 00 00 00 <0f> 0b 66 0f 1f 44 00 00 e8 3b ff 00 00 4c 8b 73 10 85 c0 0f 85
> All code
> ========
> 0: 17 (bad)
> 1: 02 00 add (%rax),%al
> 3: 00 49 8b add %cl,-0x75(%rcx)
> 6: 14 24 adc $0x24,%al
> 8: 4c 89 ee mov %r13,%rsi
> b: 48 89 df mov %rbx,%rdi
> e: ff d0 callq *%rax
> 10: 48 8b 7d c8 mov -0x38(%rbp),%rdi
> 14: 89 45 d0 mov %eax,-0x30(%rbp)
> 17: e8 77 24 ee ff callq 0xffffffffffee2493
> 1c: 8b 45 d0 mov -0x30(%rbp),%eax
> 1f: e9 38 01 00 00 jmpq 0x15c
> 24: 0f 1f 80 00 00 00 00 nopl 0x0(%rax)
> 2b:* 0f 0b ud2 <-- trapping instruction
> 2d: 66 0f 1f 44 00 00 nopw 0x0(%rax,%rax,1)
> 33: e8 3b ff 00 00 callq 0xff73
> 38: 4c 8b 73 10 mov 0x10(%rbx),%r14
> 3c: 85 c0 test %eax,%eax
> 3e: 0f .byte 0xf
> 3f: 85 00 test %eax,(%rax)
>
> Code starting with the faulting instruction
> ===========================================
> 0: 0f 0b ud2
> 2: 66 0f 1f 44 00 00 nopw 0x0(%rax,%rax,1)
> 8: e8 3b ff 00 00 callq 0xff48
> d: 4c 8b 73 10 mov 0x10(%rbx),%r14
> 11: 85 c0 test %eax,%eax
> 13: 0f .byte 0xf
> 14: 85 00 test %eax,(%rax)
> [ 2581.273353] RIP rmap_walk (mm/rmap.c:530 mm/rmap.c:1675 mm/rmap.c:1707)
> [ 2581.273353] RSP <ffff8804bd9e7bb8>
>
>
> Thanks,
> Sasha

--
Kirill A. Shutemov
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/