Re: [PATCH 1/1] MM: virtual address debug

[Jiri Slaby <jirislaby@gmail.com>]

Ingo Molnar napsal(a):
> * Ingo Molnar <mingo@elte.hu> wrote:
> 
>> * Jiri Slaby <jirislaby@gmail.com> wrote:
>>
>>> Add some (configurable) expensive sanity checking to catch wrong address
>>> translations on x86.
>>>
>>> - create linux/mmdebug.h file to be able include this file in
>>>   asm headers to not get unsolvable loops in header files
>>> - __phys_addr on x86_32 became a function in ioremap.c since
>>>   PAGE_OFFSET, is_vmalloc_addr and VMALLOC_* non-constasts are undefined
>>>   if declared in page_32.h
>>> - add __phys_addr_const for initializing doublefault_tss.__cr3
>> applied, thanks Jiri. I have created a new tip/x86/mm-debug topic for 
>> this because the patch touches mm/vmalloc.c and other MM bits.
> 
> -tip testing triggered an early boot crash and i have bisected it down 
> to your patch. The crash:
> 
> No NUMA configuration found
> Faking a node at 0000000000000000-000000003fff0000
> Entering add_active_range(0, 0, 159) 0 entries of 25600 used
> Entering add_active_range(0, 256, 262128) 1 entries of 25600 used
> Bootmem setup node 0 0000000000000000-000000003fff0000
>   NODE_DATA [000000000000a000 - 000000000003dfff]
> PANIC: early exception 06 rip 10:ffffffff80ba7531 error 0 cr2 f06f53

Expception 06 is an invalid opcode. This is maybe a false positive, probably 
a mistake in phys_to_nid, I'll look into that, thanks.

> Pid: 0, comm: swapper Not tainted 2.6.26-rc6 #7709
> 
> Call Trace:
>  [<ffffffff80b9c196>] early_idt_handler+0x56/0x6a
>  [<ffffffff80ba7531>] setup_node_bootmem+0x12a/0x2d4
>  [<ffffffff80ba7505>] setup_node_bootmem+0xfe/0x2d4
>  [<ffffffff80b9dd73>] setup_arch+0x2a2/0x3e7
>  [<ffffffff8024e858>] clockevents_register_notifier+0x2d/0x31
>  [<ffffffff80b9cb5d>] start_kernel+0x8d/0x30a
>  [<ffffffff80b9f87d>] reserve_early+0x16/0xad
>  [<ffffffff80b9c35f>] x86_64_start_kernel+0x16d/0x174
> 
> RIP 0x10