Re: [Cerowrt-devel] TFO crashes cerowrt 3.7.1-1

[Eric Dumazet <edumazet@google.com>]

[-- Attachment #1: Type: text/plain, Size: 2667 bytes --]

Some paths want to check a spinlock is held, others want to check if its
not held, it depends on the context.

So returning 1 on UP would break a bunch of code as well.


On Mon, Jan 14, 2013 at 12:18 AM, Jerry Chu <hkchu@google.com> wrote:

>
>
> On Sun, Jan 13, 2013 at 7:05 PM, Eric Dumazet <edumazet@google.com> wrote:
>
>> Oh well yes, this doesnt quite work on !SMP.
>>
>
> Strange - how would one assert a spin lock is held, and obviously only for
> SMP? (I almost think arch_spin_is_locked(lock) should be ((void)(lock),
> 1) for UP for the purpose of assertion...)
>
> Also it looks like there are bunch of other places spin_is_locked()
> assertion is made in the source tree. (Perhaps they are only configured for
> MP?)
>
> Thanks,
>
> Jerry
>
>
>> And this kind of bug is frequent....
>>
>> See following example :
>>
>> commit b9980cdcf2524c5fe15d8cbae9c97b3ed6385563
>> Author: Hugh Dickins <hughd@google.com>
>> Date:   Wed Feb 8 17:13:40 2012 -0800
>>
>>     mm: fix UP THP spin_is_locked BUGs
>>
>>     Fix CONFIG_TRANSPARENT_HUGEPAGE=y CONFIG_SMP=n CONFIG_DEBUG_VM=y
>>     CONFIG_DEBUG_SPINLOCK=n kernel: spin_is_locked() is then always false,
>>     and so triggers some BUGs in Transparent HugePage codepaths.
>>
>>     asm-generic/bug.h mentions this problem, and provides a
>> WARN_ON_SMP(x);
>>     but being too lazy to add VM_BUG_ON_SMP, BUG_ON_SMP, WARN_ON_SMP_ONCE,
>>     VM_WARN_ON_SMP_ONCE, just test NR_CPUS != 1 in the existing
>> VM_BUG_ONs.
>>
>>     Signed-off-by: Hugh Dickins <hughd@google.com>
>>     Cc: Andrea Arcangeli <aarcange@redhat.com>
>>     Cc: <stable@vger.kernel.org>
>>     Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
>>     Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
>>
>> diff --git a/mm/huge_memory.c b/mm/huge_memory.c
>> index b3ffc21..91d3efb 100644
>> --- a/mm/huge_memory.c
>> +++ b/mm/huge_memory.c
>> @@ -2083,7 +2083,7 @@ static void collect_mm_slot(struct mm_slot *mm_slot)
>>  {
>>         struct mm_struct *mm = mm_slot->mm;
>>
>> -       VM_BUG_ON(!spin_is_locked(&khugepaged_mm_lock));
>> +       VM_BUG_ON(NR_CPUS != 1 && !spin_is_locked(&khugepaged_mm_lock));
>>
>>
>>
>>
>> On Sun, Jan 13, 2013 at 1:39 PM, Felix Fietkau <nbd@openwrt.org> wrote:
>>
>>> On 2013-01-13 7:03 PM, Eric Dumazet wrote:
>>> > I suspect a bug in the spin_is_locked() implementation on your arch, as
>>> > he socket lock should be held at this point.
>>> I don't think this is an arch implementation bug, this probably happens
>>> on all !SMP systems. See this bit from include/linux/spinlock_up.h:
>>>
>>> #define arch_spin_is_locked(lock)   ((void)(lock), 0)
>>>
>>> - Felix
>>>
>>>
>>
>

[-- Attachment #2: Type: text/html, Size: 5150 bytes --]