Kernel 6.19.5 boot regression with IOMMU enabled

Xariann · 2026-03-02 12:57:28

I am reporting a boot regression in kernel 6.19.5 currently in [testing]. I cannot access the Arch GitLab to file an issue there.

ISSUE:
System fails to boot with kernel 6.19.5 (both linux and linux-zen) unless IOMMU is disabled via kernel parameter. Kernel 6.18.13 boots without issue. This is a transcription of the error:

[ 2.503040] AMD-Vi: Event logged [IO_PAGE_FAULT device=0000:11:00.0 domain=0x
0000 address=0x109a0440 flags=0x0070]
[ 2.504562] AMD-Vi: Event logged [IO_PAGE_FAULT device=0000:11:00.0 domain=0x
0000 address=0x109800440 flags=0x0070]
[ 2.507118] AMD-Vi: Event logged [IO_PAGE_FAULT device=0000:11:00.0 domain=0x
0000 address=0x1090c0440 flags=0x0070]
[ 2.553161] AMD-Vi: Event logged [IO_PAGE_FAULT device=0000:10:00.0 domain=0x
0000 address=0x102f05000 flags=0x0050]
[ 2.557813] AMD-Vi: Event logged [IO_PAGE_FAULT device=0000:13:00.3 domain=0x
0000 address=0x102f76000 flags=0x0050]
[ 2.559277] AMD-Vi: Event logged [IO_PAGE_FAULT device=0000:13:00.4 domain=0x
0000 address=0x102f00000 flags=0x0050]
[ 2.561592] AMD-Vi: Event logged [IO_PAGE_FAULT device=0000:14:00.0 domain=0x
0000 address=0x102fa7000 flags=0x0050]
[ 2.654533] AMD-Vi: Event logged [IO_PAGE_FAULT device=0000:10:00.0 domain=0x0000 address=0x102f32000 flags=0x0050]
[ 2.802265] AMD-Vi: Event logged [IO_PAGE_FAULT device=0000:11:00.0 domain=0x0000 address=0x109800000 flags=0x0050]
[ 3.159996] xhci_hcd 0000:13:00.3: PM: suspend_common(): xhci_pci_suspend returns -110
[ 3.160118] xhci_hcd 0000:13:00.3: can't suspend (hcd_pci_runtime_suspend returned -110)
[ 17.785815] xhci_hcd 0000:10:00.0: Abort failed to stop command ring: -110
[ 17.785817] xhci_hcd 0000:10:00.0: xHCI host controller not responding, assume dead
[ 17.785823] xhci_hcd 0000:10:00.0: HC died; cleaning up
[ 17.785923] xhci_hcd 0000:10:00.0: Error while assigning device slot ID: Command Aborted
[ 17.785949] ata1: softreset failed (1st FIS failed)
[ 17.786091] xhci_hcd 0000:10:00.0: Max number of devices this xHCI host supports is 127.
[ 17.786081] usb usb1-port4: couldn't allocate usb_device
[ 17.793553] AMD-Vi: Event logged [IO_PAGE_FAULT device=0000:11:00.0 domain=0x0000 address=0x109800440 flags=0x0070]
[ 18.090752] AMD-Vi: Event logged [IO_PAGE_FAULT device=0000:11:00.0 domain=0x0000 address=0x109800000 flags=0x0050]

AFFECTED HARDWARE:
Motherboard: B650M DS3H
CPU: Ryzen 9800x3D
GPU: AMD 9070

REGRESSION TIMELINE:
- 6.18.13: works fine
- 6.19-rc through 6.19.3: regression present, system unbootable
- 6.19.4: upstream appears to include fix (patch already present when attempting manual application)
- 6.19.5 in Arch [testing]: regression still present

FIX AVAILABLE:
https://lore.kernel.org/lkml/2026022708 … theas.dev/ resolves this issue. CachyOS kernel 6.19.5 includes this patch and boots successfully.

CONCERN:
The current number of signoffs in [testing] may push this to [core] with the regression intact. This affected multiple users as documented in CachyOS forums prior to kernel 6.19.5.

REQUEST:
Please apply the referenced patch before promoting 6.19.5 to [core] to prevent widespread boot failures.

I am an Arch tester currently awaiting GitLab access to file this formally. Once again apologies if this is not the correct channel.

Xariann

Mod edit: removed 'urgent' tag from subject. Please read https://wiki.archlinux.org/title/Genera … ow_to_post -- WorMzy

Last edited by WorMzy (2026-03-10 21:52:21)

cryptearth · 2026-03-02 14:05:13

hint: avoid wording as "urgent" or similar - in the past these were heavily abused and hence often cause topics with such names to be ignored
if something is actually important then this forum can handle please use good old mailing lists: https://lists.archlinux.org/mailman3/lists/
also: please use bbcode code tags for console output

gromit · 2026-03-02 14:57:40

Please file a bug here: https://gitlab.archlinux.org/archlinux/ … x/-/issues

seth · 2026-03-02 15:02:24

6.19.4: upstream appears to include fix (patch already present when attempting manual application)

https://github.com/gregkh/linux/commits … mu/iommu.c
https://github.com/gregkh/linux/commits … mu/iommu.c
https://github.com/torvalds/linux/commi … mu/iommu.c

Patch is from 27 Feb 2026 and afaict hasn't been picked up anywhere yet?

I am an Arch tester currently awaiting GitLab access to file this formally.

Ping back if you don't have an account tomorrow 7:00 UTC or after

@gromit, according to the OP they can't (unless you added their account)

Last edited by seth (2026-03-02 15:03:17)

Xariann · 2026-03-02 17:27:23

seth wrote:

6.19.4: upstream appears to include fix (patch already present when attempting manual application)
https://github.com/gregkh/linux/commits … mu/iommu.c
https://github.com/gregkh/linux/commits … mu/iommu.c
https://github.com/torvalds/linux/commi … mu/iommu.c
Patch is from 27 Feb 2026 and afaict hasn't been picked up anywhere yet?
I am an Arch tester currently awaiting GitLab access to file this formally.
Ping back if you don't have an account tomorrow 7:00 UTC or after
@gromit, according to the OP they can't (unless you added their account)

Indeed I cannot, it was the first thing I tried. When trying to sign up I got a message to contact a support email address which I have done.

I was under the impression that the patch existed in 6.19.4 because when I tried to apply the patch to it when compiling from source it looked like the patch was there, but new to this so probably easily confused.

@cryptearth, noted. I have indeed tried to use the mailing list and got bounced back. I need to set up a mailing client that can deal with them. I thought it would be urgent given the fact that this causes boot failures, it got marked as such by another distro (admittedly with different procedures) and it has been signed off by several testers and might get pushed to core imminently, and it is about the kernel. Obviously I was misguided. Given that I attempted other channels and was not able to access them I resorted to the forums.

If you would be willing to also receive some pointers, an acknowledgement of what I was trying to achieve in addition to everything I did wrong would have been nice to see. This wasn't me trying to get my favourite app to work on Arch Linux.

Last edited by Xariann (2026-03-02 17:29:06)

cryptearth · 2026-03-02 18:49:15

Xariann wrote:

@cryptearth, noted. I have indeed tried to use the mailing list and got bounced back. I need to set up a mailing client that can deal with them. I thought it would be urgent given the fact that this causes boot failures, it got marked as such by another distro (admittedly with different procedures) and it has been signed off by several testers and might get pushed to core imminently, and it is about the kernel. Obviously I was misguided. Given that I attempted other channels and was not able to access them I resorted to the forums.

well, yes, it is quite important - and thanks for bringing it up - but usually additions like "urgent", "important", "required fast" or similar are often abused by inexperienced users trying to get attention - usually for quite unimportant things OPs not even bothered to give one second the ask google about - and hence are often dismissed as spam
lucky for you this topic has at least are quite descriptive title so those who have to take a look at it did so anyway no matter the urgent "tag"
it's just a friendly hint: it's better to chose a good topic title rather than "spam" with stuff like "urgent" or "important" - a well chosen topic title will draw required attention way better - and enables users to get whether "do I have to bother to look into the topic to figure out I don't care?" or if it's something they can help with

as for mailing lists: usually you have to subscribe to a list upfront before you can post to it - so a bounce sounds like you tried to send to it without subscribing to it - this is normal and expected behaviour to prevent spam

seth · 2026-03-02 20:35:59

because when I tried to apply the patch to it

What source exactly and do you still have the output you ran into?

When trying to sign up I got a message to contact a support email address which I have done.

'key - idk whether gromit has picked up on this but otherwise

seth wrote:

Ping back if you don't have an account tomorrow 7:00 UTC or after

Xariann · 2026-03-07 17:39:58

I got access and posted the bug there.

When I tried to apply the patch to 6.19.4 I got a message asking me to revert it, so I took it to mean it was there.

I can't provide the exact message without reattempting it, but as it's not in 6.19.5 now, it's not really relevant and I took that part out on GitLab. Given that CachyOS had patched it and their kernel worked, in my mind that was, "Oh okay upstream probably added it in 6.19.4, and I expect the Arch version of 6.19.5 to work too", but it didn't. So now my conclusion is that the revert message probably means something else.

As I said new to this.

The kernel boots in a VM btw.

Anyway as I opened the issue on GitLab as I have access now, I guess the discussion can go there.

I see 6.19.6 has been pushed out of testing so I might have another look. I took a break from testing after this, there wasn't much I could do with the Arch install unless I booted LTS or the CachyOS kernel.

Last edited by Xariann (2026-03-07 17:41:43)

Kwarf · 2026-03-10 21:30:51

I just ran into this same issue now on 6.19.6.arch1-1 from core. I'm running a 9950X in a MSI MAG B850M Mortar WiFi (latest BIOS with AGESA PI pre-1.3.0.0), and a RX 9070 XT if that's relevant.

I installed linux-lts instead, which is 6.18.16-1 right now, and that works without any issues.

seth · 2026-03-10 22:50:24

https://gitlab.archlinux.org/archlinux/ … issues/182

masc2026 · 2026-03-11 05:16:42

Kann meine Kiste (9800X3D, MS-7E51 mit BIOS Release 7E51v1A81 vom 2026-03-11) nach dem Wechsel auf 6.19.6.arch1-1 auch nicht mehr booten.

Vor dem update:

Mär 10 10:28:24 archlinux kernel: Linux version 6.18.13-arch1-1 (linux@archlinux) (gcc (GCC) 15.2.1 20260209, GNU ld (GNU Binutils) 2.46) #1 SMP PREEMPT_DYNAMIC Wed, 25 Feb 2026 23:12:35 +0000

Um 12 Uhr update auf 6.19.6.arch1-1 und danach nur noch booten mit LTS:

Mär 10 15:30:34 archlinux kernel: Linux version 6.18.16-1-lts (linux-lts@archlinux) (gcc (GCC) 15.2.1 20260209, GNU ld (GNU Binutils) 2.46) #1 SMP PREEMPT_DYNAMIC Wed, 04 Mar 2026 18:09:01 +0000

Mit iommu=off (oder amd_iommu=on iommu=pt) bootet auch 6.19.6.arch1-1 :

Mär 10 18:05:58 archlinux kernel: Linux version 6.19.6-arch1-1 (linux@archlinux) (gcc (GCC) 15.2.1 20260209, GNU ld (GNU Binutils) 2.46) #1 SMP PREEMPT_DYNAMIC Wed, 04 Mar 2026 18:25:08 +0000

Last edited by masc2026 (Yesterday 18:27:41)

WorMzy · 2026-03-11 11:41:39

Welcome to the forums, masc2026. Please post in English outside of https://bbs.archlinux.org/viewforum.php?id=30. Using a free translation service is fine.

Arch Linux

#1 2026-03-02 12:57:28

Kernel 6.19.5 boot regression with IOMMU enabled

#2 2026-03-02 14:05:13

Re: Kernel 6.19.5 boot regression with IOMMU enabled

#3 2026-03-02 14:57:40

Re: Kernel 6.19.5 boot regression with IOMMU enabled

#4 2026-03-02 15:02:24

Re: Kernel 6.19.5 boot regression with IOMMU enabled

#5 2026-03-02 17:27:23

Re: Kernel 6.19.5 boot regression with IOMMU enabled

#6 2026-03-02 18:49:15

Re: Kernel 6.19.5 boot regression with IOMMU enabled

#7 2026-03-02 20:35:59

Re: Kernel 6.19.5 boot regression with IOMMU enabled

#8 2026-03-07 17:39:58

Re: Kernel 6.19.5 boot regression with IOMMU enabled

#9 2026-03-10 21:30:51

Re: Kernel 6.19.5 boot regression with IOMMU enabled

#10 2026-03-10 22:50:24

Re: Kernel 6.19.5 boot regression with IOMMU enabled

#11 2026-03-11 05:16:42

Re: Kernel 6.19.5 boot regression with IOMMU enabled

#12 2026-03-11 11:41:39

Re: Kernel 6.19.5 boot regression with IOMMU enabled

Board footer