Booting Linux using UEFI can brick Samsung laptops

headkase · 2013-01-31 03:36:17

Here is a situation that Arch users and developers may be interested in:

http://www.h-online.com/open/news/item/ … 93958.html

Basically, does Arch use the "samsung-laptop driver" as that seems to be the suspect at this time?

Edit: As mentioned by jasonwryan a few posts down, if you have experienced a Samsung bricking with Arch that would be very useful to post in this thread.

Last edited by headkase (2013-01-31 06:12:00)

WonderWoofy · 2013-01-31 03:39:48

Well, it is indeed in the kernel tree.

headkase · 2013-01-31 03:41:41

The severity of the problem I believe is what makes this an important issue: if this happens to you then you must send the device back to Samsung for repair. There is no software recovery as the machine will not even boot far enough to get to the firmware.

WonderWoofy · 2013-01-31 03:42:50

Maybe it might be wise to tell anyone with a samsung laptop to blacklist the samsung-laptop module for the time being.

jasonwryan · 2013-01-31 03:59:20

Yes it does use it. It hasn't bricked my laptop but it did throw some nasty panics when I tried to boot the Live image to install Arch; I blacklisted the module from the chroot and have modprobed it for the two kernel bumps since wondering if it would still panic (you need the module loaded for key backlighting and rfkill keys to work--hardly a deal breaker, but nice to have), and yes, it still freaks out.

# edit: it would be interesting to see if any Arch machines have been bricked, or if it is just an Ubuntu kernel issue. There are certainly a couple threads here from mid-to-late 2012 describing the panics, but no-one seems to have ended up with a coaster.

headkase · 2013-01-31 07:31:19

Here is the relevant posting on the LKML:

http://lkml.indiana.edu/hypermail/linux … 04361.html

65kid · 2013-01-31 10:44:19

I have a 900x4d which is, as far as I know, technically identical to the affected 900x4c. I also was affected by the kernel panic but it didn't brick my laptop either. The panic does not occur if you boot with UEFI-only because then the samsung-laptop module is not even loaded (it does however load and panic when you set the BIOS to support both UEFI and CSM and boot via UEFI. This is what the patch fixes as far as I understand).

little OT: The BIOS has another bug: the CPU doesn't report the AES-NI extension, not even under Windows, while according to Intel the CPU supports it. Strange thing is that I seem to be the only one who noticed this, I guess AES-NI just isn't really used by that many people. I wrote an E-Mail to the american samsung support, knowing that I probably wouldn't get a response. What I did get was an email that basically said "you have a german samsung model, so we can't help you. Go ask the german samsung guys.". Yeah right, as if these guys will help me with this...

edit: I hate how some people blame Linux for this. This is clearly Samsung's fault since it shouldn't even be possible for the kernel to brick the firmware.

Last edited by 65kid (2013-01-31 10:46:54)

mm7 · 2013-01-31 13:31:30

I got a 900x4d too and also noticed that aes-ni support isn't reported by the cpu. i was quite disappointed, because luks is slower without aes-ni support. i did'nt try to force use of aes-ni. if the bios turns it off while booting, the cpu will throw an exception. maybe a bios update will help.

65kid · 2013-01-31 13:38:51

mm7 wrote:

I got a 900x4d too and also noticed that aes-ni support isn't reported by the cpu. i was quite disappointed, because luks is slower without aes-ni support. i did'nt try to force use of aes-ni. if the bios turns it off while booting, the cpu will throw an exception. maybe a bios update will help.

nope, did a BIOS update last week, didn't help...

jasonwryan · 2013-01-31 21:19:36

65kid wrote:

The panic does not occur if you boot with UEFI-only because then the samsung-laptop module is not even loaded

Thanks for that. I still had it set to boot in compatibility mode as it was the only way I could get it to recognize the install USB.

bnavigator · 2013-01-31 22:01:19

I read the article on heise yesterday and it gave me a cold chill.

I have been running Arch Linux on my new Samsung NP900X4D for about a week. Installed it from USB Drive in UEFI only mode. I don't quite remember anymore but I think the first time I tried in CSR or combined UEFI and CSR mode and it didn't boot. Whether due to kernel panic or not, I don't know anymore. I then replaced the efi loader according to the wiki and booted in UEFI only mode. Now after install, the computer is booting through UEFI only and Grub.

After reading the article I checked for loaded kernel modules and was relieved to find samsung_laptop not loaded. I then blacklistet the module in /etc/modprobe.d/ make sure it does not get loaded after some kernel update.

headkase · 2013-02-01 09:48:50

The work around is being merged into all kernels now:

http://www.h-online.com/open/news/item/ … 95332.html

The next maintenance release of 3.7.x will have it and so will forthcoming 3.8.x's. Other kernels will be getting it too but the mentioned ones here are most important to Arch.

plasmoid · 2013-02-03 14:58:14

65kid could you keep us updated on the aes-ni issue (i.e. the reply from german support)? I think of buying one of the Samsung series 9 ultrabooks but especially this bug is a no-go...thanks

65kid · 2013-02-09 12:33:59

plasmoid wrote:

65kid could you keep us updated on the aes-ni issue (i.e. the reply from german support)? I think of buying one of the Samsung series 9 ultrabooks but especially this bug is a no-go...thanks

I actually just wrote them a few minutes ago, but I seriously doubt that I will get any helpful response.

headkase · 2013-02-09 13:43:13

The UEFI bug is definitely not fixed:

http://www.h-online.com/open/news/item/ … 00541.html

And it is now recommended that all Samsung owners of the affected models use BIOS compatibility mode at all times instead of UEFI until a firmware update is released.

blackout23 · 2013-02-09 13:53:40

Laptop bug not Linux specific - http://mjg59.dreamwidth.org/22855.html

65kid · 2013-02-11 10:40:58

so I actually got a response regarding the AES-NI issue.
loosely translated:

On some models AES-NI support has been deactivated. If TrueCrypt doesn't report AES-NI support, then this is also the case for your model. Unfortunately it is not possible to activate AES-NI. We hope for your understanding blablabla...

well... thank you for telling me what I already knew... idiots...

mm7 · 2013-02-11 11:50:37

I wonder why they disabled aes-ni. maybe a licencing problem with intel. one can argue that samsung does false advertising with core i5 3317U processors. can you ask the support if they plan to support aes-ni with a bios/uefi update?

headkase · 2013-02-11 12:01:08

My Linux computer doesn't have AES instructions, Pentium E6600 @ 3.06Ghz. My Windows computer does have AES instructions, Core i5 650 @ 3.2Ghz. Here's the thing: Truecrypt - which supports AES instructions - on my Linux machine (when it had Windows) did a full disk (1TB) hard disk encryption in 10 hours. My Windows machine, again the one that actually has AES and Truecrypt verified it was using them, had an estimated time of 30 hours to also encrypt a 1TB hard disk. I aborted that on the Windows machine and took Truecrypt off that one. I don't actually use Truecrypt, I was just futzing around.

65kid · 2013-02-12 18:11:44

mm7 wrote:

I wonder why they disabled aes-ni. maybe a licencing problem with intel. one can argue that samsung does false advertising with core i5 3317U processors. can you ask the support if they plan to support aes-ni with a bios/uefi update?

Well, I asked them again WHY aes-ni is deactivated because there is technically no reason for this. Yet another bullshit response:

Samsung reserves itself the right to provide devices without making all features available. Additionally there is no reason that this application is necessary for the device to function correctly. blablabla

yes, the email actually says "application", this might tell us something about how much this guy knows what he is talking about...

here the original:

Samsung behält sich vor die Geräte nicht mit alles zur Verfügung stehenden Funtkionen auszuliefern. Weiterhin gibt es keine zwingende Gründe dass dieses Programm zum reibungslosen Funktion des Notebooks notwendig ist.

I'm not gonna respond again, it's obviously hopeless.

Last edited by 65kid (2013-02-12 18:14:56)

mm7 · 2013-02-12 19:38:09

then samsung does misleading advertisement. they say, there is a core i5 3317U processor in their notebooks. ark.intel.com clearly states, that aes-ni support is one of this processor's features. aes support is definitly necessary to support the performance of samsung's ssds in this notebook. so the support is wrong in here too. a bios update will be necessary to support aes-ni. i am sure that the bios disables aes-ni support. (although the thing that i call bios is not really a bios but some kind of crippled firmware for an embedded device.)

one idea would be to change the bios to enable aes-ni. i guess, the bios disables it by writing to some msr. i googled and found, that some people were able to enabled aes-ni support on their processor by changing the bios code to write the correct value to the msr. the problem is, that i have no idea how i should flash the bios even if i find the code that disables aes-ni.

mm7 · 2013-02-12 20:01:17

use rdmsr from msr-tools to read msr 0x13c. it will output 1 on plattforms that support aes and have it enabled by the bios and 3 for plattforms that have it disabled.

ephemeric · 2013-02-17 18:59:23

Greetz,

Arch is fantastic, love it, thank you.

Now, I booted BIOS mode on a NP900X4D and saw samsung-laptop module loaded and proceeded to run pacstrap.

All good.

Now I booted USB ISO in UEFI and installed again and all good but no samsung-laptop module?

Has it been removed permanently if the host boots into UEFI? What are the future plans?

How does one get into /sysfs/ without this module?

I've installed samsung-tools and get some of the buttons working but I don't like the solution, seems messy

and overly complicated.

Even with the above it says no CPU fan control support.

Fn-F1 I assume is now somehow backlight hotkey but it blanks the display instead? Tried the acpi_backlight=vendor but no joy.

Ambient light sensor? I assume never working.

Why can't these vendors devote some resources to write some Linux code? They have so much money and resources on their hands.

Thank you.

Lone_Wolf · 2013-02-18 11:36:50

Ephemeric, check the links in headkase & blackout23 posts above.

Basicaly UEFI support on many samsung laptops is broken, and using it can brick your laptop.
NOTE : the bricking can also happen from within windows .

Samsung is the only one that can fix it, until they have done that stick to bios mode.

srs5694 · 2013-02-18 16:24:02

Lone_Wolf wrote:

Samsung is the only one that can fix it, until they have done that stick to bios mode.

Depending on the chipsets used, in theory you could use CoreBoot, perhaps with a TianoCore payload for UEFI support. This would require a lot of effort, though, and I expect most people would end up bricking their laptops trying to implement it. Thus, it's only likely to become practical if somebody with sufficient knowledge decides to create a replacement firmware image and distribute it. I only mention this because somebody (other than Samsung) might do so, particularly if Samsung drags their heels over this. As a practical matter, the advice to switch to BIOS-mode booting is certainly the safest.

Arch Linux

#1 2013-01-31 03:36:17

Booting Linux using UEFI can brick Samsung laptops

#2 2013-01-31 03:39:48

Re: Booting Linux using UEFI can brick Samsung laptops

#3 2013-01-31 03:41:41

Re: Booting Linux using UEFI can brick Samsung laptops

#4 2013-01-31 03:42:50

Re: Booting Linux using UEFI can brick Samsung laptops

#5 2013-01-31 03:59:20

Re: Booting Linux using UEFI can brick Samsung laptops

#6 2013-01-31 07:31:19

Re: Booting Linux using UEFI can brick Samsung laptops

#7 2013-01-31 10:44:19

Re: Booting Linux using UEFI can brick Samsung laptops

#8 2013-01-31 13:31:30

Re: Booting Linux using UEFI can brick Samsung laptops

#9 2013-01-31 13:38:51

Re: Booting Linux using UEFI can brick Samsung laptops

#10 2013-01-31 21:19:36

Re: Booting Linux using UEFI can brick Samsung laptops

#11 2013-01-31 22:01:19

Re: Booting Linux using UEFI can brick Samsung laptops

#12 2013-02-01 09:48:50

Re: Booting Linux using UEFI can brick Samsung laptops

#13 2013-02-03 14:58:14

Re: Booting Linux using UEFI can brick Samsung laptops

#14 2013-02-09 12:33:59

Re: Booting Linux using UEFI can brick Samsung laptops

#15 2013-02-09 13:43:13

Re: Booting Linux using UEFI can brick Samsung laptops

#16 2013-02-09 13:53:40

Re: Booting Linux using UEFI can brick Samsung laptops

#17 2013-02-11 10:40:58

Re: Booting Linux using UEFI can brick Samsung laptops

#18 2013-02-11 11:50:37

Re: Booting Linux using UEFI can brick Samsung laptops

#19 2013-02-11 12:01:08

Re: Booting Linux using UEFI can brick Samsung laptops

#20 2013-02-12 18:11:44

Re: Booting Linux using UEFI can brick Samsung laptops

#21 2013-02-12 19:38:09

Re: Booting Linux using UEFI can brick Samsung laptops

#22 2013-02-12 20:01:17

Re: Booting Linux using UEFI can brick Samsung laptops

#23 2013-02-17 18:59:23

Re: Booting Linux using UEFI can brick Samsung laptops

#24 2013-02-18 11:36:50

Re: Booting Linux using UEFI can brick Samsung laptops

#25 2013-02-18 16:24:02

Re: Booting Linux using UEFI can brick Samsung laptops

Board footer