Kernel panic during page-out

[Justin Hibbits]

Hi Peter


On Sat, Sep 7, 2013 at 7:23 PM, Peter Grehan <grehan at freebsd.org> wrote:

> Hi Justin,
>
>
>  I'm seeing a kernel panic on my G5, I think when it tries to page out to
>> swap, with the message:
>>
>> Lock vm object not exclusively locked @ .../aim/mmu_oea64.c:1529
>> backtrace:
>> .kdb_backtrace+0x5c
>> .vpanic+0x13c
>> .panic+0x38
>> .__rw_assert+0x184
>> .moea64_is_modified+0x54
>> .moea64_remove_all+0x78
>> .vm_pageout_object_deactivate_**pages+0x2e4
>> .vm_daemon+0x358
>> .fork_exit+0xdc
>> .fork_trampoline+0x10
>>
>> This machine has 12GiB RAM, running r254922.
>>
>
>  This is the same panic I saw on the g5 imac when I only had 512MB of RAM
> - 'twas much too easy to reproduce.
>
>  The fix is
>
> moea64_is_modified
>  ...
> -       VM_OBJECT_ASSERT_WLOCKED(m->**object);
> +       VM_OBJECT_ASSERT_LOCKED(m->**object);
>
>  This is consistent with locking higher up in the call chain. Apologies
> for not submitting that when I found it (and discussed it with Nathan).
>
>
> later,
>
> Peter.
>
>
If this is the correct fix, it should be pushed in through re@, and go into
10.  Is this correct?  I haven't checked the full locking on it.

- Justin