Slashdot Mirror
MoBo Manufacturer Foxconn Refuses To Support Linux
Noodlenose notes a thread up on the Ubuntu forums, where a user is questioning the practices of hardware manufacturer Foxconn. The user describes how his new Foxconn motherboard caused his Linux install to freeze and fire off weird kernel errors. He disassembles the BIOS and concludes that a faulty DSDT table is responsible for the errors. Even though the user makes Foxconn aware of the problem, they refuse to correct it, as 'it doesn't support Linux' and is only 'Microsoft certified.' The user speculates darkly on Foxconn's motives. Read the forum, read the code, and come to your own conclusions. "I disassembled my BIOS to have a look around, and while I won't post the results here, I'll tell you what I did find. They have several different tables, a group for Windows XP and Vista, a group for 2000, a group for NT, Me, 95, 98, etc. that just errors out, and one for LINUX. The one for Linux points to a badly written table that does not correspond to the board's ACPI implementation.' The worst part is Foxconn's insistence that the product is ACPI compliant because their tables passed to Windows work, and that Microsoft gave the the magic WHQL certification."
Return it and buy from a manufacturer... no need to disassemble the BIOS, your time is worth more than that.
It appears that within an hour there was a workaround posted on the same forum.
Have you Meta Moderated t
If you're planning on running a Linux OS on your machine, don't use Foxconn. If they don't want customers, that's their business.
...Windows hardware back. Seriously, who is stupid enough today not to support linux?
It takes a man to suffer ignorance and smile
Be yourself no matter what they say
Yeah, except for the part where the motherboard claims to be ACPI compliant when it really isn't. That's sort of false advertising.
In my workplace we run Windows, OS X, and Linux. I have the expectation of being able to use Linux on any x86 kit we buy. Absent an explanation or attitude change from this vendor, I won't recommend their kit here for Windows use either. That seems somewhat important so I'll repeat it:
I will not buy Foxconn kit for Windows use if Linux compatibility is impaired.
The trouble here isn't that it doesn't support Linux, it's that the motherboard appears to be actively sabotaging Linux. That's a really weird thing to do and deserves investigation.
The problem is that Foxconn says its ACPI compliant but its not. It also looks as if they botched Linux by pure purpouse. Why on earth would they have a Linux section in the bios when they dont support it? Something is very smelly here thats for sure. I will keep miles away from Foxconn at my departments no matter if my systems are intended for Windows or Linux.
HTTP/1.1 400
If you follow the link in the story, you would see that the poster claims the following:
1) Foxconn advertises its motherboard as ACPI compliant thus potentially misleading people into thinking that linux should be able to handle the board. The company does nothing to counter such possible misunderstandings. One could argue that Foxconn is not obliged to do anything of that sort but for customers it is not as simple as "doing homework" as you suggest. Foxconn doesn't say that things break on linux. They only say "works with windows" and "ACPI compliant". The only way to check is to buy and use (at least until this story).
2) The BIOS actively looks for the OS and passes a modified table to linux. It does not even ask the OS to identify itself and go along with that identification. It rather keeps on having random checks to ensure it is running on windows. I can't think of any good reason why they need to do that unless they want to actively break things for linux.
3) The poster smells something fishy in Foxconn's behavior. Right or wrong, I don't know. But if the poster is right in his suspicion (which s/he must believe), it would be a natural, rational and justified behavior to bitch and moan about it rather than just return the board for a refund. Society owes a lot to such "troublemakers".
Although this vendor seems definitely not trying to support Linux with it's BIOS, the hard truth is that it's not so easy even for those who try. For more information, there is currently a thread on the LKML disussing this and how to improve the situation.
In particular, latest kernels claim to be every versions of Windows at the same time, and not Linux! That's not easy to handle for the BIOS writer...
The point is that they advertised that they are ACPI compatible when they are not. And no, "it works on Windows" is not enough to claim ACPI compatibility.
No, actually.
File this under having done the world a service by publishing their findings.
Now we know that at least some Foxconn motherboards do not work with Linux, and Foxconn is not interested in doing anything about that. That's useful information.
From other posts, I gather that the motherboard actually has a table specifically targeted at Linux, which supplies broken settings. So it's more than Foxconn simply not supporting Linux; they've actually gone and broken things.
Finally, it seems there is already a workaround available. I guess Linux is willing to support Foxconn, even if Foxconn doesn't want it to. And, really, this is a case of "yay, open source!"
Please correct me if I got my facts wrong.
Foxconn has no obligation to support
They went out of their way and expended extra effort to prevent Linux from working on their system. This moved beyond "not supporting", to "breaking" hardware that should have functioned without any effort at all on foxconns part, using what was probably considerable effort on their part to detect what kernel was booting, then developing a fake ACPI table to show only when it detected linux.
The interesting part is that a year or so back, there was an article here about how Microsoft floated a letter around manufacturers asking how to make ACPI harder for Linux to implement. Everyone asserted that we were just paranoid and the only reason ACPI was hard for Linux was because "Linux developers suck", but now it seems we know.
Here is most of the original article.
The pesky junk filter meant I had to snip some of the code out - sorry.
Posting AC for the usual reason(s).
Foxconn deliberately sabotaging their BIOS to destroy Linux ACPI
Edit: Please tell Foxconn what you think of their behavior:
http://www.foxconnchannel.com/support/online.aspx
You need to put in an email, and then it will bring up a form, choose Complain/Suggest.
Edit: Welcome Digg, Reddit, and Slashdot.
http://digg.com/linux_unix/Foxconn_d..._destroy_Linux
http://www.reddit.com/comments/6tcv8...their_bios_to/
(Will add Slashdot when I know the final URL)
------------
I disassembled my BIOS to have a look around, and while I won't post the results here,I'll tell you what I did find.
They have several different tables, a group for Windws XP and Vista, a group for 2000, a group for NT, Me, 95, 98, etc. that just errors out, and one for LINUX.
The one for Linux points to a badly written table that does not correspond to the board's ACPI implementation, causing weird kernel errors, strange system freezing, no suspend or hibernate, and other problems, using my modifications below, I've gotten it down to just crashing on the next reboot after having suspended, the horrible thing about disassembling any program is that you have no commenting, so it's hard to tell which does what, but I'll be damned if I'm going to buy a copy of Vista just to get the crashing caused by Foxconn's BIOS to stop, I am not going to be terrorized.
-----
How to fix:
Get Intel's BIOS ACPI source compiler:
sudo apt-get install iasl
Dump your DSDT table:
sudo cat /sys/firmware/acpi/tables/DSDT > dsdt.dat
Disassemble it:
iasl -d dsdt.dat
Open it in Gedit:
gedit dsdt.dsl
Fix Foxconn sabotage:
Find, the section that starts out with
Code:
If (_OSI ("Windows 2000"))
{
Store (0x04, OSVR)
}
Go down til you get to the first
Code:
}
Else
{
Past that you should see Linux alongside Windows NT, which is above another Else that leads to Windows Me.
Should look like:
Code:
If (MCTH (_OS, "Linux"))
{
Store (0x3, OSVR)
}
Change it to:
Code:
If (_OSI ("Linux"))
{
Store (Zero, OSVR)
}
Copy the section, and remove it and the other characters (CAREFULLY PRESERVING SYNTAX!!!!)
Then move the Linux section to right underneath Windows 2006 section.
_Code removed to get past junk filter_
So there you have it!
Have to agree with you here. This is a case of false advertising if it isn't acpi compliant (there is no 90% compliant, or compliant if you use this-or-that software, all that is just non-compliant). I don't know about the slashdot readers that answer with "so what, just return it", but when I am looking for new hardware, I am very happy if people like him figure out who is trying to screw me with false claims, so I can skip these manufacturers from my list.
molmod.com - computing tips from a molecular modeling
Foxconn also accuses him of making "idle treats".
I want an idle treat.
Badass Resumes
Because the OS's have bugs in their ACPI implementations. So the BIOS provides a special version of function with a workaround for the bug in case it detects the specific OS version.
Let's note this is valid only for proprietary OS's (aka Windows). For F/OSS kernels, the BIOS writer can simply report a bug on non-ACPI compliance, and it's fixed soon after directly in the kernel.
If Windows, it sends the good tables. If Linux, it sends the deliberately faulty ones.
It's still more likely incompetence than conspiracy. Most motherboard manufacturers don't write their own BIOS - they buy a stock one from AMI/Award and make a few changes for their particular board.
What they most likely did was update the DSDT tables handed out to Windows to reflect their hardware, but didn't bother changing the others. So for Linux (and perhaps Win9x) it just has the generic tables that came with the BIOS, which of course don't work for their particular board.
Of course, a BIOS even having per-OS tables is indicative of poor design, since being OS-independent is kinda the whole point of ACPI. That's more of an issue with whoever wrote the BIOS in the first place, though.
While they're probably not out to actively sabotage Linux, it's still poor customer service to refuse to fix it and claim that everything is working fine. Sadly, getting most board manufacturers to fix their broken DSDT tables (and there are a lot out there) is akin to pulling teeth.
This is important and I want to expound on it. I work in a Microsoft shop. Really, it's IIS this, MSSQL that, .NET for all dev, and we've all got the latest and greatest Office suite. Strangely, we've heard rumors that our software is going to be tested Vista, but QA hasn't received a Vista machine, yet. With all of that out of the way, I use Linux in various ways on many of our test computers. Mostly, it is just boot CDs, such as Partimage Is Not Ghost and Ultimate Boot CD. So, just because hardware is meant for Windows doesn't mean that it will never see another OS. Hardware interoperability on the software level is necessary.
On another note, I've encountered Foxconn boards in the past... usually broken and being replaced.
This is more a case of *Microsoft* not being ACPI compliant. The different versions of Windows have historically broken ACPI in hilariously random and catastrophic ways. You can decompile any BIOS on the market and find a similar table. If you're willing to rule out malicious sabotage on the part of Foxconn (which would be a pretty ballsy move given that they manufacture Intel's reference motherboards), the fault can probably be traced back to their BIOS vendor - either AMI or Award, if memory serves.
Off-brand? They don't sell much under their own branding, but Foxconn is one of the biggest computer components manufacturer in the world. Lots of HP and Dells I've seen have Foxconn boards.
And how mature and professional is a support drone that says 'don't use linux, use windows vista'??
how long until
I don't work for FoxConn, but I do work for a hardware and software vendor. And here's some insight - as I have been in a situation similar to FoxConn - but being both the accuser and the guilty party at the same time ;)
WHQL is kind of a big deal for hardware vendors. The main attractive is being able to add the "certified compatible with Windows" to your product box. Honestly speaking - having the logo there gives you *some* cred with users - at least, with Windows users ;)
So your competitors are nibbling at you, the product has to ship, and you need to have the logo in the box. What do you do? If you're already late to market, you hack. You install all the different flavors of Windows, check if it works - if it doesn't and crashes, well, some of that can be attributed to Windows itself. As long as you can install the OS and pass the certification, you're good, the product ships, you get your bonus and a pat in the back for delivering on time.
So say that during testing you DO install Linux and crashes - time for a reality check. If the product spec said "Windows WHQL is a must", and making Linux happy means not passing WHQL - tough luck. Linux won't run.
Or if "fixing the product so it passes WHQL" means "screwing Linux users", well, let me think about that ;)
Many engineers working on any given product would like to ship the best possible product - the one that has a 100% compliant ACPI, APM, TPM, you-name-it implementation.
But when time is short and the management chain is breathing down your neck . . . you do whatever it has to be done to be able to ship. And hope that once the product is out there, you WILL be able to go back and clean up the mess - and ship a BIOS upgrade. Everyone is happy.
Sadly, by the time the product shipped, you've been reassigned to other product - and you will only go back to the first one if the Windows crowd complains.
The solution is easy - Linux users to boycott the brand. But then again: if the mobo was designed to be sold to another company to be used as the basis for a product that will only run Windows . . . It isn't like you care a lot about losing the Linux business. This is only the reality - hard as it might seem.
And to the guy that originally found the bug: next time, remember that maybe the guy at the other side of the email exchange also thinks the situation sucks, but he's powerless to change it.
Because if even if he was provided with a full working patch for the BIOS (that doesn't break Windows compatibility), he might need to reapply for WHQL if he patches the BIOS - which means more $$$ and time spent on a product that is already shipping. So.
It has not always been a myth. Sure, there may be no such thing as perfectly interchangeable hardware, but most hardware used to adhere to well-known standard interfaces.
Modems all supported RS-232 serial ports, and nearly all used the AT command set. Internal modems had no physical RS-232 port, but to rest of the computer, they looked like a serial port and still presented the same AT command interface. There were no "drivers" to speak of.
Sure, some modems added extensions / features, but as long as you stuck to the core AT commands (e.g. ATH, ATD, ATA, etc), the modem would work the same predictable way regardless of who made it or whether it was internal/external. You were free to choose your terminal software... hardware... operating system... even serial port hardware... the list goes on.
What was really nice about generic hardware was that it worked in a well-known, predictable way. If you were so inclined, you could write your own terminal software, operating system... even create your own hardware if you wanted. The information you needed to get everything working--UART documentation, AT command set, BIOS calls, X86 instruction set, etc... was widely available. The only limits in your way were the limits of your own ability to figure everything out.
You mention the C64 and its own proprietary modems. In fact, the user port on the C64 was RS-232 compatible, the main difference being voltage levels. Many companies designed RS-232 interface kits for the C64 allowing you to connect any standard modem you wanted. The specs for the user port were published in the Commodore 64 programmer's reference manual. If you were so inclined, you could actually build your own RS-232 interface from parts available at the electronics store.
With Windows-specific hardware, we no longer have that freedom. We've lost something -- Now, even if we want to write our own software stack or implement our own hardware, we're stuck -- the information needed to make the hardware work is hidden, locked away in a binary driver that only works on one platform. The only way to make it work elsewhere (e.g. Linux) is to reverse engineer the product -- much more difficult that working against an open spec.
Why do I have to reverse engineer my own hardware if it supposedly adheres to a published / well known specification?
No, it allegedly has a bunch of checks for Linux strewn about in random places which then give bad data upon detecting Linux.
The only person claiming that is the original poster in that thread, whose correspondence with Foxconn and the (!) FTC is instantly accusatory and full of assumption. It's almost as if he started with the premise that Foxconn was actively breaking Linux to be anticompetitive and looked for evidence to support it.
I'd have to see a full DSDT dump to be sure, but from the excerpts posted it looks like "active checking" is just matching against _OS instead of using _OSI, which is a mistake a naive BIOS writer unfamiliar with the spec could easily make. It doesn't help the issue that Linux lies about its identity in _OSI.
The "redundant checks" seem to be present for the Windows code path too, and look more to me like bad spaghetti code copied and pasted multiple places.
I also take issue with
Find and replace all seven occurences of Acquire (MUTE, 0x03E8) and replace with Acquire (MUTE, 0xFFFF), it appears they're trying to crash the kernel by locking a region of memory that shouldn't be locked, but without access to their source code comments, I can only speculate, this tells it to lock a memory address that is always reserved instead. ;)
It's obviously not trying to crash the kernel, that's not how Acquire() works. The second parameter is a timeout, not a memory address. 3E8 hex = 1000 decimal. The BIOS writer was trying to acquire a mutex with a 1 second timeout.
Changing it to 0xFFFF makes it wait forever, which could potentially cause worse problems as execution will get stuck if the mutex is already held. Multithreaded synchronization is a very tricky problem, and I'm not surprised to see they got it wrong. Without examining the code it's impossible to say what effect TheAlmightyCthulu's changes have, if they're correct or if they merely mask the problem.
Saying they're trying to deliberately crash the kernel is a bit ridiculous.
But then again I'm a BSD guy, so I don't start out with a chip on my shoulder and assume everyone's out to get me. Have seen a ton of shoddy BIOSes in my time though.
For reference, here is Bill Gates' email asking how they can make ACPI incompatible with Linux.