Re: [PATCH] edac: avoid mce decoding crash after edac driver unloaded

[Borislav Petkov <bp@amd64.org>]

+ Tony.

On Sat, May 05, 2012 at 09:20:36AM +0800, Chen Gong wrote:
> Some edac drivers register themselves as mce decoders via
> notifier_chain. But in current notifier_chain implementation logic,
> it doesn't accept same notifier registered twice. If so, it will be
> wrong when removing the element from the list. For example, on one
> SandyBridge platform, remove module sb_edac and then trigger one
> error, it will hit oops because it has no mce decoder registered
> but related notifier_chain still points to an invalid callback
> function. Here is an example:
> 
> Call Trace:
>  [<ffffffff8150ef6a>] atomic_notifier_call_chain+0x1a/0x20
>  [<ffffffff8102b936>] mce_log+0x46/0x180
>  [<ffffffff8102eaea>] apei_mce_report_mem_error+0x4a/0x60
>  [<ffffffff812e19d2>] ghes_do_proc+0x192/0x210
>  [<ffffffff812e2066>] ghes_proc+0x46/0x70
>  [<ffffffff812e20d8>] ghes_notify_sci+0x48/0x80
>  [<ffffffff8150ef05>] notifier_call_chain+0x55/0x80
>  [<ffffffff81076f1a>] __blocking_notifier_call_chain+0x5a/0x80
>  [<ffffffff812aea11>] ? acpi_os_wait_events_complete+0x23/0x23
>  [<ffffffff81076f56>] blocking_notifier_call_chain+0x16/0x20
>  [<ffffffff812ddc4d>] acpi_hed_notify+0x19/0x1b
>  [<ffffffff812b16bd>] acpi_device_notify+0x19/0x1b
>  [<ffffffff812beb38>] acpi_ev_notify_dispatch+0x67/0x7f
>  [<ffffffff812aea3a>] acpi_os_execute_deferred+0x29/0x36
>  [<ffffffff81069dc2>] process_one_work+0x132/0x450
>  [<ffffffff8106bbcb>] worker_thread+0x17b/0x3c0
>  [<ffffffff8106ba50>] ? manage_workers+0x120/0x120
>  [<ffffffff81070aee>] kthread+0x9e/0xb0
>  [<ffffffff81514724>] kernel_thread_helper+0x4/0x10
>  [<ffffffff81070a50>] ? kthread_freezable_should_stop+0x70/0x70
>  [<ffffffff81514720>] ? gs_change+0x13/0x13
> Code: f3 49 89 d4 45 85 ed 4d 89 c6 48 8b 0f 74 48 48 85 c9 75 17 eb 41
> 0f 1f 80 00 00 00 00 41 83 ed 01 4c 89 f9 74 22 4d 85 ff 74 1d <4c> 8b
> 79 08 4c 89 e2 48 89 de 48 89 cf ff 11 4d 85 f6 74 04 41
> RIP  [<ffffffff8150eef6>] notifier_call_chain+0x46/0x80
>  RSP <ffff88042868fb20>
> CR2: ffffffffa01af838
> ---[ end trace 0100930068e73e6f ]---
> BUG: unable to handle kernel paging request at fffffffffffffff8
> IP: [<ffffffff810705b0>] kthread_data+0x10/0x20
> PGD 1a0d067 PUD 1a0e067 PMD 0
> Oops: 0000 [#2] SMP
> 
> Only i7core_edac and sb_edac have such issues because they have more
> than one memory controller which means they have to register mce
> decoder many times.

Right, so the idea of the MCE notifier is to register on the chain
_once_ per whole driver, i.e. once per MCE error source. Having an EDAC
driver register memory controller every instance is simply not needed.

It could probably be done correctly by registering on the chain
at the end of i7core_init(), for the i7 edac driver for example,
after everything else has been successful. From the looks ot it,
i7core_mce_check_error() seems to select the proper node, i.e. memory
controller instance when invoked.

Thanks.

-- 
Regards/Gruss,
Boris.

Advanced Micro Devices GmbH
Einsteinring 24, 85609 Dornach
GM: Alberto Bozzo
Reg: Dornach, Landkreis Muenchen
HRB Nr. 43632 WEEE Registernr: 129 19551