BIOS update destroyed my KVM/vfio setup.
Printed From: ASRock.com
Category: Technical Support
Forum Name: AMD Motherboards
Forum Description: Question about ASRock AMD motherboards
URL: https://forum.asrock.com/forum_posts.asp?TID=11471
Printed Date: 05 Aug 2025 at 9:20pm
Software Version: Web Wiz Forums 12.04 - http://www.webwizforums.com
Topic: BIOS update destroyed my KVM/vfio setup.
Posted By: pantato
Subject: BIOS update destroyed my KVM/vfio setup.
Date Posted: 13 May 2019 at 8:33am
| How do I force my x470 Taichi to revert to a previous BIOS version? It appears that it's locked down on BIOS side of things. I want to revert to 1.60 after updating to 3.20. I am experiencing the (hopefully) infamous AGESA issue that is causing my VM not to work. I *replaced* my MSI x470 Gaming Plus motherboard with this Taichi motherboard because I bricked my MSI board trying to force the BIOS to revert on it for the same problem, then I was stupid enough to think the same issue would not occur on Asrock if I updated. I guess it has to do with AGESA 0.0.7.2 . Needless to say I'm really upset. I was finally able to get the VM to boot with [URL=https://clbin.com/GGnxm]this kernel hack[/URL] but it only boots once per host boot, and the performance is really bad. I don't want to brick my Taichi like I did my Gaming Plus, can someone *please* give me instructions for a workaround of this BIOS revert lockdown? It's truly unbelievable to me that all these companies would decide to force their users in this manner. I switched to AMD from Intel *just for the cores to power this KVM/vfio setup*. |
Replies:
Posted By: pantato
Date Posted: 13 May 2019 at 9:40am
| I tried using /O on the AFUDOS version, but it says it can't read flash information. I'd rather not use AFUWIN. |
Posted By: wardog
Date Posted: 13 May 2019 at 11:51am
|
As you've discovered with your MSI, major AGESA updates, 1.0.x to 0.0.x ComboAM4, cannot be back-flashed. Also, it is not the MB manuf's that req this, but AMD. |
Posted By: wardog
Date Posted: 13 May 2019 at 12:11pm
|
?? https://forum-en.msi.com/index.php?topic=302638.0 |
Posted By: pantato
Date Posted: 13 May 2019 at 12:55pm
| Apparently all the different motherboards (x470 vs x370, msi vs taichi) use different flags with their flashtools. Also I am told I have to flash down in steps (a later version, then an earlier, then an earlier). I am hoping to get explicit directions for this particular board because it's really easy to mess up. I was told that the last time AGESA had this problem it took over a year to fix...I think it's really really unfair that this BIOS update can't be reverted through normal means. Manufacturers should at least have official instructions for people who *really need* to go back to a different version. I understand it's AMD's policy to disallow updates because of people bricking their boards when they upgrade their CPUs...but come on at least give instructions or a workaround to people who are aware of that fact. |
Posted By: wardog
Date Posted: 13 May 2019 at 1:07pm
|
Wish me luck then! I'm fixing to flash my X370 Killer SLi/ac to the latest BIOS that is AGESA 0.0.7.2, 5.30, and after attempt to revert it back to 5.10 thru the EFI Shell. The Shell 'should' bypass the built-in protections that disallow a User to revert back. As to FLASH switchs, I'm not concerned. They are both AMI BIOSes and as such should accept the same commands, especially when executed from the EFI Shell. AMI would NOT include/omit a switch that would endanger flashing. Not in their best interest to do so. I'll post back after I develope the nads to try this on a working MB tonight. I fear alcohol may be req'd   
|
Posted By: wardog
Date Posted: 13 May 2019 at 2:24pm
|
OK. First, prior to flashing 5.30 that has 0.0.72 for my Killer SLI/ac I Disabled Fast Boot and Secure Boot in BIOS and disabled "Turn on fast startup" in Windows CP. The 'Turn on fast start' I suspect may be causing issues that are being reported here and elsewhere. Remember, the BIOS and Windows are now inextricably linked and do communicate between themselves. Disabling any cache that speeds up Windows is not a bonus when flashing a BIOS. Into BIOS and select the Instant Flash Util and up-flashed to 5.30(0.0.72) All is well up to this point. No sluggishness. Loads fast as ever has. I'll do some peeking under the hood before I attempt to revert back to 5.10(1.0.0.6). See where my mem in my sig will top out at as I didn't do any pre-flash benchs to compare. My mem, those G.Skill Trident-Z 4133's, I know inside out and upside down. More to come .............. |
Posted By: wardog
Date Posted: 13 May 2019 at 3:02pm
|
Ok, gonna pull the trigger and attempt a revert. I've chosen NOT to use the MSI method linked above and instead built a bootable USB EFI Shell environment from scratch as I understand it. Better as I have no one to blame, hahahaha, but myself should I brick my MB. Assuming this works, which I do, I'll zip up the files I used and put them on ZippyShare for all, with instructions to follow. All one will need to do is issue a command from the Shell and bingo. BRB |
Posted By: wardog
Date Posted: 13 May 2019 at 4:59pm
|
Oops. Let's hope I can recover it. |
Posted By: pantato
Date Posted: 13 May 2019 at 5:33pm
YEAH. Welcome to my world. |
wardog wrote:Posted By: pantato
Date Posted: 14 May 2019 at 12:08pm
Any luck waking it up or are you SOL? |
Posted By: pantato
Date Posted: 16 May 2019 at 2:01pm
| bump |
Posted By: Joykill3r
Date Posted: 19 May 2019 at 4:33am
|
Ah found this thread too, Exactly my problem x470 Taichi 2600x Ryzen. Upgraded from P2.0 to get the LED S5 fix and rekt my VM Qemu setup.. Wish I saw these posts before i went to do it.. |
Posted By: pantato
Date Posted: 19 May 2019 at 11:49am
|
There is a "fix" that you can use to patch the linux kernel that sort of works. I can boot my VM now once per host boot on my rx580 after applying the patch. https://clbin.com/GGnxm Save that as a .patch file and look up how to compile the kernel with a patch for your respective distro. You must edit the 2nd to last 2 lines on the patch for your PCI-E ids. eg. the "ls iommu" script given on the arch wiki (made a link here: http://ix.io/1Jqt ) shows my guest GPU PC id(s) like so: IOMMU Group 16 24:00.0 VGA compatible controller [0300]: Advanced Micro Devices, Inc. [AMD/ATI] Ellesmere [Radeon RX 470/480/570/570X/580/580X/590] [1002:67df] (rev e7) IOMMU Group 16 24:00.1 Audio device [0403]: Advanced Micro Devices, Inc. [AMD/ATI] Ellesmere HDMI Audio [Radeon RX 470/480 / 570/580/590] [1002:aaf0] So I had to edit the patch to look like this: [daniel@satori ~]$ cat build/linux/trunk/quirks.patch diff --git a/drivers/pci/quirks.c b/drivers/pci/quirks.c index e2a879e93d86..62b754253318 100644 --- a/drivers/pci/quirks.c +++ b/drivers/pci/quirks.c @@ -5118,3 +5118,7 @@ SWITCHTEC_QUIRK(0x8573); /* PFXI 48XG3 */ SWITCHTEC_QUIRK(0x8574); /* PFXI 64XG3 */ SWITCHTEC_QUIRK(0x8575); /* PFXI 80XG3 */ SWITCHTEC_QUIRK(0x8576); /* PFXI 96XG3 */ + +DECLARE_PCI_FIXUP_HEADER(0x1002, 0x67df, quirk_no_bus_reset); +DECLARE_PCI_FIXUP_HEADER(0x1002, 0xaaf0, quirk_no_bus_reset); +DECLARE_PCI_FIXUP_HEADER(0x1022, 0x145f, quirk_no_bus_reset); Let me know if you want more help with this, I know how painful this  mess is.
|
Posted By: pantato
Date Posted: 19 May 2019 at 11:51am
| For some people that patch I posted fixes it entirely. For me, it gave me a working VM with some small issues. If you apply it, let me know how it works out for you and what your GPU is. |
Posted By: pantato
Date Posted: 19 May 2019 at 7:40pm
| Actually, scratch those last couple replies. The fix is here: https://clbin.com/VCiYJ |
Posted By: Kannen_g
Date Posted: 20 Jun 2019 at 4:36am
|
Hallo, und hat das zurücksetzen des Bios von 5.30 auf 5.10 funktioniert. Gru? frank |
Posted By: lofe
Date Posted: 17 Nov 2019 at 6:48am
|
Hi, Anybody know anything about a fix for this problem? With the 3.50 bios version, passthrough is still not working, although I could found a video about a workaround to downgrade the bios, i don't want to take the risk to do that. |