Linux/Athlon Bug Discovered

[Kosh Naranek]

check http://slashdot.org/articles/02/01/21/0750226.shtml for the full story.

Apparently this bug has been known since september 2000.

The question not answered in all the threads is .... is the bug still there ?

Is it in my Athlon XP ?

Since the bug affects AGP, how does this relate to the recent nVIDIA / VIA issues ?

[Wombat]

This bug IS responsible for nVidia problems under Linux. For further clarification, the bug was not publically "known" until just now. AMD knew about it in September 2000, and released a W2K patch for it that disabled the feature AMD incorrectly implemented. That's all that was said on the matter, and nobody knew Linux (or any other OS) needed a patch to workaround this CPU bug.

[Kosh Naranek]

Does Win XP need a patch ?

What CPU's are affected ?

Does anyone know, becuase AMD isn't talking.


It's not like I want to blame people or point fingers , I just want to know.

[Wombat]

Probably not.

Nobody knows.

AMD isn't being open about this.

[Rimfaxe]

There is a patch for WinXP and AMD K7. I don't know what will happend if you dont use it.

Windows update asked me to install it, so I did.

[Tempest]

Every processor has bugs, most of the bugs are just never brought to public...

16 INVLPG Instruction Does Not Flush Entire Four-Megabyte Page Properly with Certain Linear Addresses:
This particular bug might possibly cause memory corruption. This is because AMD tried to implement the same kind of memory handling that Intel did on its new chips, so that the OS could handle the memory in 4MB blocks instead of 4KB. But using 4MB blocks with the Athlon/Duron with an OS that takes advantage of this feature can in fact cause memory corruption. This bug should be corrected in Athlons of model 6 revision A5 (CPUID 662) and later.
AMD has not yet provided enough information for Linux users, so to be on the safe side they have to compile their kernels without Pentium optimizations.

[Wombat]

Originally posted by Tempest
Every processor has bugs, most of the bugs are just never brought to public...

This is very true. However, if a bug can be considered "customer visible," then a lot more should be done by the chip manufacturer to help people compensate. I'm not saying that AMD should send a postcard informing every person to ever purchase an AMD CPU, but the people that write this part of the Linux kernel do their homework; if they didn't find this information, and a workaround implementation, then AMD dropped the ball.
This is particularly bad for a company that has aspirations of breaking into the server market.

[TdB]

on intel processors you can patch the microcode, isn´t the same thing possible on amd cpus?