i'll summarize my off-list discussion with ruik here.
i've slightly, or moderately, edited some of my previous answers to ruik. in this email i've updated, corrected and integrated various information i gave.
On 07/17/2014 06:04 PM, Rudolf Marek wrote:
Hi Andrew,
I seen also some freezes and I investigate what it could be. In the meanwhile I fixed the errata #712 (this is related to freeze due to the problem in VGA powermanagement)
awesome!
I have also seen some freezes which ends up as soft lockup. I noticed that the microcode is not updateting to the version 061119 on the CPU.
i've also seen some strange lockups. i've done tests by leaving some machines idle over night.
one of the configurations i tested had a 10de:01d3 nvidia card, without a vga option rom. it was mostly unresponsive after ~13 hours. ssh worked for a while, but the 3rd ssh command or so did not complete.
So now I'm waiting for Sage to release a microcode update to coreboot.
great!
Strangely the freeze problem do not occur on my test system which has different APU.
i've been using a8-5600K processors. does your test system use the a10-5800k or the a10-5700? if so, i'd like to test it as well.
Please can you try with the latest coreboot to see if the fix for errata #712 fixed that?
i've been using builds which include that commit. i'll do future tests based off of today's git head.
a wheezy install without the vga option rom did well; it was up for ~28 hours. a wheezy machine with the internal graphics + the 6402 vga rom lasted ~40 hours. it's hard to see a clear pattern here, as there are very short up times for both configurations as well.
I would like to know also what is your microcde revision in the cpu (after linux microcode patch is applied)
i've attached microcode info. it's version 20130710, installed through debian sid.
a test using an external graphics card and updated microcode lasted over 2.5 days before someone rebooted it. other tests were all over the place, lasting less than an hour, or 9.5 hours, for instance.
interestingly enough, a version of coreboot from the fall of 2013, compiled by someone else, allegedly works quite well for someone using an external nvidia card and a trisquel install. their system only needs hard resetting about once every four days or so; when using internal graphics, they hard reset about once per day. i'm not sure if this is a best case scenario, or normal behavior.
it's very unlikely that they're using microcode; they use the same coreboot build when using internal and external graphics.
I dont know what VGA rom I'm using, I just extracted some random at the beginning. Those roms are most likely just patched with the proper ID so depending on your VGA ID we can try to compare them.
What is your VGA PCI ID? Mine are 9901 and 9993 I think.
the a8-5600k internal graphics pciid is 9904.
thanks for hacking on coreboot! :-D
Thanks Rudolf
-------- Forwarded Message -------- Subject: Re: f2a85-m vga binary revision Date: Tue, 22 Jul 2014 16:44:25 +0200 From: Rudolf Marek r.marek@assembler.cz To: Andrew Engelbrecht sudoman@ninthfloor.org
Hi
I did some decompression of the BIOS images and found that there are only two versions and two kinds of ROMs:
ATOMBIOSBK-AMD VER015.022.000.000.000000 which is present up to BIOS 5202 from that there is a newer version: VER015.031.000.000.000000
I think I used 15 22 0 in my tests. The EFI version has only atombios tables and no legacy x86 BIOS code (except for jmp blabla ... blabla: retf)
I also checked the dumped version from dd from system and it is different not only by the PCI IDs but also perhaps some atomtables are patched in runtime.
BIOS VERSION: F2A85-M-ASUS-5016 2f34e ATOMBIOSBK-AMD VER015.022.000.000.000000 ed66 ATOMBIOSBK-AMD VER015.022.000.000.000000 BIOS VERSION: F2A85-M-ASUS-5018 2f34e ATOMBIOSBK-AMD VER015.022.000.000.000000 ed66 ATOMBIOSBK-AMD VER015.022.000.000.000000 BIOS VERSION: F2A85-M-ASUS-5104 2f34e ATOMBIOSBK-AMD VER015.022.000.000.000000 ed66 ATOMBIOSBK-AMD VER015.022.000.000.000000 BIOS VERSION: F2A85-M-ASUS-5107 2f34e ATOMBIOSBK-AMD VER015.022.000.000.000000 ed66 ATOMBIOSBK-AMD VER015.022.000.000.000000 BIOS VERSION: F2A85-M-ASUS-5202 2f34e ATOMBIOSBK-AMD VER015.022.000.000.000000 ed66 ATOMBIOSBK-AMD VER015.022.000.000.000000 BIOS VERSION: F2A85-M-ASUS-6004 2f34e ATOMBIOSBK-AMD VER015.031.000.000.000000 2206 ATOMBIOSBK-AMD VER015.031.000.000.000000 BIOS VERSION: F2A85-M-ASUS-6102 2f34e ATOMBIOSBK-AMD VER015.031.000.000.000000 2206 ATOMBIOSBK-AMD VER015.031.000.000.000000 BIOS VERSION: F2A85-M-ASUS-6402 2f34e ATOMBIOSBK-AMD VER015.031.000.000.000000 2206 ATOMBIOSBK-AMD VER015.031.000.000.000000 BIOS VERSION: F2A85-M-ASUS-6404 2f34e ATOMBIOSBK-AMD VER015.031.000.000.000000 2206 ATOMBIOSBK-AMD VER015.031.000.000.000000 BIOS VERSION: F2A85-M-ASUS-6501 2f34e ATOMBIOSBK-AMD VER015.031.000.000.000000 2206 ATOMBIOSBK-AMD VER015.031.000.000.000000 BIOS VERSION: F2A85-M-ASUS-6502 2f34e ATOMBIOSBK-AMD VER015.031.000.000.000000 2206 ATOMBIOSBK-AMD VER015.031.000.000.000000 BIOS VERSION: F2A85-M-ASUS-6506 2f34e ATOMBIOSBK-AMD VER015.031.000.000.000000 2206 ATOMBIOSBK-AMD VER015.031.000.000.000000
Thanks Rudolf