Re: PROBLEM: Kernel 2.6.10 crashing repeatedly and hard

[Manfred Spraul]

> Did that last night. You are right -- it is so slow that it is no fun
> at all. So I started the test run last night and went to bed. 
>
>  
It was a crash in free_block - this means control structures of the slab 
allocator were corrupted.

You wrote that enabling everything caused a hard crash without anything 
on the console. Could you try what happens if slab debugging is enabled, 
but page alloc debugging is off? You should get a verbose BUG_ON() with 
kmem_cache_free/kmem_cache_alloc caller addresses and similar stuff.

> EFLAGS: 00010002   (2.6.10)
> EIP is at free_block+0x45/0xd0
> eax: 46484849   ebx: df2b1000   ecx: df2b1050   edx: df2ab000
> esi: c183cd80   edi: 00000001   ebp: 00000018   esp: c188fef8
> ds: 007b   es: 007b   ss: 0068

Could you compile your kernel without debug and send me mm/slab.o? The 
%eax value is odd: FHHI or IHHF.

Btw, if it's still too slow for you without page alloc debug: remove the 
forced poisoning:

 if ((size < 4096 || fls(size-1) == fls(size-1+3*BYTES_PER_WORD)))
         flags |= SLAB_RED_ZONE|SLAB_STORE_USER;
- if (!(flags & SLAB_DESTROY_BY_RCU))
-         flags |= SLAB_POISON;
#endif

--
	Manfred


-
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/