boot failure on i7-3317u in Samsung 900x3c

boot failure on i7-3317u in Samsung 900x3c

[Mike Bakhterev]

Greetings.

When i'm trying to boot the subject maching in UEFI mode i get
sometimes machine check exception 0xbe2000000003110a (i was not able
to decode this, as some bits are not covered in Intel's Manuals). It
emerges when i am booting 3.4.x, 3.5.4 and 3.5.5 kernels from
ArchLinux. When i am booting from SSD i get this error all the time,
when i'm booting from USB stick i get this sometimes (quite frequently
and the probability to boot succesfully is increased when i specify
"verbose" kernel option). I'm booting with systemd. The sytems boots
always successfully, when i use the installation ArchLinux image,
which is shipped with 3.5 kernel and boots in BIOS mode with
traditional initscripts.

It looks like race condition when configuring memory or something like this.

It is not hardware error, because Windows 7 and 8 boot and run fine.
And this is my second unit, because the first one was broken
(something with motherboard) when running Windows 8, and Linux on that
broken machine reported the same MCE. Now Windows works OK (if you
could say, it works), and Linux reports the same MCEs.

Here is some guy playing with caching of PCI memory region and getting
the same MCE on the Sandy Bridge:
http://stackoverflow.com/questions/11254023/how-to-do-mmap-for-cacheable-pcie-bar

Sorry, but i have no time to bisect this :(

P.S. Excuse me my English.

P.P.S. Can this be fixed with kernel command line options?

Re: boot failure on i7-3317u in Samsung 900x3c

[Matt Fleming]

(Cc'ing some arch and MCE folks since my MCE-foo is weak)

On Mon, 2012-10-08 at 00:43 +0600, Mike Bakhterev wrote:
> Greetings.
> 
> When i'm trying to boot the subject maching in UEFI mode i get
> sometimes machine check exception 0xbe2000000003110a (i was not able
> to decode this, as some bits are not covered in Intel's Manuals). It
> emerges when i am booting 3.4.x, 3.5.4 and 3.5.5 kernels from
> ArchLinux. When i am booting from SSD i get this error all the time,
> when i'm booting from USB stick i get this sometimes (quite frequently
> and the probability to boot succesfully is increased when i specify
> "verbose" kernel option). I'm booting with systemd. The sytems boots
> always successfully, when i use the installation ArchLinux image,
> which is shipped with 3.5 kernel and boots in BIOS mode with
> traditional initscripts.

We've seen things like this before on Samsung laptops,

	https://bugzilla.kernel.org/show_bug.cgi?id=47121

We attributed it to a buggy UEFI implementation.
	
> It looks like race condition when configuring memory or something like this.
> 
> It is not hardware error, because Windows 7 and 8 boot and run fine.
> And this is my second unit, because the first one was broken
> (something with motherboard) when running Windows 8, and Linux on that
> broken machine reported the same MCE. Now Windows works OK (if you
> could say, it works), and Linux reports the same MCEs.

So both Windows 7 and 8 boot successfully on your new laptop, but every
OS failed with the *same* MCE error on your first? That is interesting. 

Does anyone have a clue how to diagnose this further? I still think it's
a buggy UEFI implementation, especially since it ships with a CSM. Is it
possible that these MCE errors are non-fatal on Win 7/8?

-- 
Matt Fleming, Intel Open Source Technology Center

Re: boot failure on i7-3317u in Samsung 900x3c

[Mike Bakhterev]

On Mon, Oct 8, 2012 at 4:09 PM, Matt Fleming <matt@console-pimps.org> wrote:
>
> We've seen things like this before on Samsung laptops,
>
>         https://bugzilla.kernel.org/show_bug.cgi?id=47121
>
> We attributed it to a buggy UEFI implementation.
>

Thanks for the link. But what kind of bug this could be? And could it
be worked around? How can i help?

> So both Windows 7 and 8 boot successfully on your new laptop, but every
> OS failed with the *same* MCE error on your first? That is interesting.
>

No. Windows-es didn't report any errors. The machine just broke
someday when Windows 8 booted. Different Linux kernels fail with the
same MCE. The same MCE rises sometimes (frequently, but not always)
when i'm trying to write something to /sys entry for keyboard
backlight or to rfkill bluetooth phy.

> Does anyone have a clue how to diagnose this further? I still think it's
> a buggy UEFI implementation, especially since it ships with a CSM. Is it
> possible that these MCE errors are non-fatal on Win 7/8?
>

Are there any info on how to decode this MCE code? It is not covered
in Intel's manuals. Hard to reason without any knowledge.

- MB, respectfully

Re: boot failure on i7-3317u in Samsung 900x3c

[Matt Fleming]

On Mon, 2012-10-08 at 21:17 +0600, Mike Bakhterev wrote:
> > Does anyone have a clue how to diagnose this further? I still think it's
> > a buggy UEFI implementation, especially since it ships with a CSM. Is it
> > possible that these MCE errors are non-fatal on Win 7/8?
> >
> 
> Are there any info on how to decode this MCE code? It is not covered
> in Intel's manuals. Hard to reason without any knowledge.

mcelog should be able to parse your MCE codes.

-- 
Matt Fleming, Intel Open Source Technology Center

Re: boot failure on i7-3317u in Samsung 900x3c

[Mike Bakhterev]

On Mon, Oct 8, 2012 at 10:45 PM, Matt Fleming <matt@console-pimps.org> wrote:
> mcelog should be able to parse your MCE codes.
>
> --
> Matt Fleming, Intel Open Source Technology Center
>

mcelog --ascii says:

echo -e 'CPU 3: Machine Check Exception: 5 Bank 6:
be2000000003110a\nTSC 2a124d1ca ADDR fe400 MISC f880004086' | mcelog
--ascii
CPU 3: Machine Check Exception: 5 Bank 6: be2000000003110a
TSC 2a124d1ca ADDR fe400 MISC f880004086
HARDWARE ERROR. This is *NOT* a software problem!
Please contact your hardware vendor
CPU 3 BANK 6 TSC 2a124d1ca
MISC f880004086 ADDR fe400
STATUS be2000000003110a MCGSTATUS 5

RE: boot failure on i7-3317u in Samsung 900x3c

[Luck, Tony]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain; charset="utf-8", Size: 482 bytes --]

> STATUS be2000000003110a MCGSTATUS 5

The SDM can help here. See volume 3A, section 15.9.2 "Compound Error Codes".

The low 16 bits of the status in this case are 0001 0001 0000 1010

This tells us that you have a cache error in L2 cache severe enough that the
processor has begun filtering further error reports.

-Tony
ÿôèº{.nÇ+‰·Ÿ®‰­†+%ŠËÿ±éݶ\x17¥Šwÿº{.nÇ+‰·¥Š{±þG«éÿŠ{ayº\x1dʇڙë,j\a­¢f£¢·hšïêÿ‘êçz_è®\x03(­éšŽŠÝ¢j"ú\x1a¶^[m§ÿÿ¾\a«þG«éÿ¢¸?™¨è­Ú&£ø§~á¶iO•æ¬z·švØ^\x14\x04\x1a¶^[m§ÿÿÃ\fÿ¶ìÿ¢¸?–I¥