About ten days ago I posted about a problem with Samsung laptops getting bricked when trying to boot Linux using UEFI. There was a fair amount of commentary on the problem, and we later updated the article to indicate that it appears the problem is in the NVRAM and that removing the CMOS battery and power is enough to clear the problem. That's certainly still a major bug, and the best thing to do is to simply avoid using UEFI with Linux on Samsung laptops. Except, the bug appears to be quite a bit deeper than just a Linux driver.

According to Matt Garrett, the problem is that the "UEFI Black Box" is supposed to take certain actions when specific conditions are met, but the UEFI code or some other aspect of the firmware is misbehaving. Bugs in firmware can be some of the most damaging, and in this case the bugs appear capable of bricking a laptop, even when what you're doing is done according to written specifications.

Matt explains, "This is pretty obviously a firmware bug. Writing UEFI variables is expressly permitted by the specification, and there should never be a situation in which an OS can fill the variable store in such a way that the firmware refuses to boot the system. We've seen similar bugs in Intel's reference code in the past, but they were all fixed early last year. For now the safest thing to do is not to use UEFI on any Samsung laptops. Unfortunately, if you're using Windows, that'll require you to reinstall it from scratch." Additional details are available in Matt's post.

What's not clear is whether the data that's written that causes the laptops to fail to boot is battery powered (i.e. removing the CMOS battery will clear the error) or if the problem is being written to NAND in some cases. If the latter is happening, the only way to fix the problem would be to send the unit to Samsung for service (or buy a new motherboard). We'll update if there's any additional information.

Source: Matthew Garrett

POST A COMMENT

15 Comments

View All Comments

  • Shrapnel09 - Sunday, February 10, 2013 - link

    Any chance HP laptops are also having this problem?
    I worked on an HP Envy dv6 last week. For compatibility with the organization, I was to install Windows 7 on it. I configured the UEFI to legacy mode and installed Windows 7. I resolved all the missing drivers and updated the system to a new UEFI version released back in November. I installed all of our standard apps and configured it like normal. I returned the laptop to the user and they used it for two days. On the third day, they plugged in a Canon USB printer and installed the driver. Upon restarting, the computer restarted with a BSoD as soon as Windows would start to load complaining that the BIOS were not ACPI compliant.

    I got the laptop back and would get the same BSoD complaint when trying to use Microsoft Deployment Toolkit to install Windows 7 again or the install media for Windows 8 and toggling legacy mode accordingly. The HP UEFI can only be flashed from within Windows but all forms of Windows that I tried to load (Windows Installer, LiveCD, and the installed OS) failed. I wiped the drive using a Gparted iso after reading that old drivers could be a source of the problem and also tried the NVRAM/CMOS trick with no luck. HP says the motherboard will have to be replaced.

    Any thoughts if this UEFI issue is related?
    Reply
  • JarredWalton - Sunday, February 10, 2013 - link

    Sounds like it could be a similar issue to what Samsung is experiencing. If Matt's post about the UEFI bug is correct, Intel had similar firmware issues about a year ago. Well, most of the OEMs get their base firmware from Intel, but they don't often update to every new release Intel puts out so they could have built off the old buggy firmware and thus still have problems. UEFI is a major change to the core of the I/O system so it's not too surprising to find some serious bugs, sadly. Reply
  • Alexvrb - Sunday, February 10, 2013 - link

    Out of curiousity, does anyone know if any AMD-based UEFI systems exhibit such problems? Reply
  • Wolfpup - Monday, February 11, 2013 - link

    I was wondering the same thing. Hopefully my friend's A6 Samsung notebook isn't affected, and maybe this would push me towards the DV6z. Reply
  • HibyPrime1 - Sunday, February 10, 2013 - link

    I realize that the lack of testing is an understandable consequence of the race to the bottom, but it still surprises the hell out of me that this problem managed to make it this far. It's easy to see the high-up execs not wanting to fund more testing than they deem necessary, but this should be an embarrassment for the engineers who worked on these systems. I think this hints at a larger problem than just itself.

    Imagine an engineer designing the internal components of a laptop, an industrial designer working with them to fit it all into a nice looking chassis and countless others putting a small piece of their lives into that laptop. Now imagine how you would react to the finished product if you were one of them, you'd play with it even on your own time. You would be the testing the product, for free.

    I have a feeling the problem lies in corporations like Samsung getting their engineers to work on one small thing on a large number of their products. For example, an LCD expert being asked to pick the best display to fit into 15 different products.

    There's no connection between the people and the product. It makes for more efficient production, but lacks in other areas, this being just one of them. I might get slammed for bringing this up, but I get the impression Apple doesn't work this way in their design process. Listening to Johnny Ive talk about the products he's designed gives me the impression he truly puts a small part of himself in his design work. Listening to others at Apple gives me the same impression.

    (even more off topic: I don't get the same impression from Tim Cook, which scares me for their future)
    Reply
  • neothe0ne - Sunday, February 10, 2013 - link

    ...but those of us who own Samsung laptops have known UEFI was broken even in Windows for a long time already. (which is why I bothered responding to the anti-Microsoft trolls in your post 10 days ago) Reply
  • Alexvrb - Sunday, February 10, 2013 - link

    I was just about to post and say that Neo already told us that Samsung's UEFI implementation was broken for more than just Linux. Where are these trolls now?

    "IT R TEH MIRCROSUCKS FALT DERRRRR"
    Reply
  • Beenthere - Sunday, February 10, 2013 - link

    Many electonics manufacturers constantly rush half-baked crap to market be it SSD, PCs, cellphones, mobos, DRAM, PSUs, etc.

    AFAIK, Asus mobos were the first to do this regularly because they reaped fortunes from first reviews duping the gullible sheeple. Once other mfgs. caught on to the financial value in shipping half-baked crap for rave PC reviews, other's marketeers have followed because it's extemely profitable to do so when the sheeple have more money than technical knowledge or good judgment. The saying is that there is a sucker born every second, in fact millions of suckers are born per second, especially in recent years.
    Reply
  • powerarmour - Monday, February 11, 2013 - link

    I think some UEFI implementations just aren't fully reliable yet, but it's not the fault of any OS if it gets corrupted, that's simply bad firmware design. Reply
  • tuxRoller - Monday, February 11, 2013 - link

    ...since I pointed this issue (and Matthew Garrett's blog) out to you in the previous article.
    No big deal, just crying a bit inside... :)
    Reply

Log in

Don't have an account? Sign up now