Re: [REGRESSION] amdgpu error -22 first encountered in kernel 6.19, not fixed in 7.0

Re: [REGRESSION] amdgpu error -22 first encountered in kernel 6.19, not fixed in 7.0

[Linux regression tracking (Thorsten Leemhuis)]

On 4/26/26 01:22, Brandon Taylor wrote:
> I have an AMD Radeon R9 270X GPU (yes, I know it's old) which used to
> default to the `radeon` firmware

FYI, as using the wrong terms can lead to confusion while dealing with
bugs: that's not a firmware, that's the driver (sometimes also called
kernel module)

> on kernel versions up to and
> including 6.18. I could manually change this behavior to use the new
> `amdgpu` firmware by setting `radeon.si_support=0 amdgpu.si_support=1`
> as kernel parameters in the GRUB bootloader. Everything worked fine,
> and I was able to play my Windows games without any problems.
> 
> That was, until kernel version 6.19 came out.>
> 6.19, from what I was able to Google, was supposed to have AMD GPUs to
> default to the new `amdgpu` firmware. Unfortunately, when I updated
> the kernel and rebooted, I got a black screen.
> 
> Further Google searches led me to reboot, set `nomodeset` in the
> kernel parameters, switch to a TTY, login, and investigate the output
> of a `dmesg` command — and what I saw ... Well, let's just say it's
> the kind of thing that would make Linus go Chernobyl:
> 
> `amdgpu: probe with driver amdgpu failed with error -22`

Please provide a full kernel log, this is needed to full understand the
problem

> And this issue has not been fixed in kernel version 7.0, which I guess
> is set to be the next LTS kernel.

No, it's not.

> Now, I hope to God that this email finds whoever is responsible for
> breaking this `amdgpu` firmware before the same thing happens to Linus
> via Fedora 44, and he goes berserk. I realize that's not a heck of a
> lot of time, but God only knows HOW many Linux users with AMD GPUs are
> pissed off by this kind of issue — and I for one can only imagine how
> Linus' reaction will compare when it happens to HIM!

FWIW, you seem to assume that the problem hits everyone with a AMD GPUs.
That's not the case, we'd know this by now. Maybe it's not even
something that hits everyone with a card/gpu like yours, as some falires
only show up in certain environments, for example in combination with a
specific mainboard chipset or firmware.

Anyway: I CCed a few people that might be able to help. But they likely
will need at least the full kernel log mentioned above .

Ciao, Thorsten

Re: [REGRESSION] amdgpu error -22 first encountered in kernel 6.19, not fixed in 7.0

[Thorsten Leemhuis]

[note: please remove Timo Lindfors when replying to this mail and
include Timur Kristóf – sorry everyone and especially Timo, my mailers
address book tricked me]

On 4/27/26 11:35, Linux regression tracking (Thorsten Leemhuis) wrote:
> On 4/26/26 01:22, Brandon Taylor wrote:
>> I have an AMD Radeon R9 270X GPU (yes, I know it's old) which used to
>> default to the `radeon` firmware
> 
> FYI, as using the wrong terms can lead to confusion while dealing with
> bugs: that's not a firmware, that's the driver (sometimes also called
> kernel module)
> 
>> on kernel versions up to and
>> including 6.18. I could manually change this behavior to use the new
>> `amdgpu` firmware by setting `radeon.si_support=0 amdgpu.si_support=1`
>> as kernel parameters in the GRUB bootloader. Everything worked fine,
>> and I was able to play my Windows games without any problems.
>>
>> That was, until kernel version 6.19 came out.>
>> 6.19, from what I was able to Google, was supposed to have AMD GPUs to
>> default to the new `amdgpu` firmware. Unfortunately, when I updated
>> the kernel and rebooted, I got a black screen.
>>
>> Further Google searches led me to reboot, set `nomodeset` in the
>> kernel parameters, switch to a TTY, login, and investigate the output
>> of a `dmesg` command — and what I saw ... Well, let's just say it's
>> the kind of thing that would make Linus go Chernobyl:
>>
>> `amdgpu: probe with driver amdgpu failed with error -22`
> 
> Please provide a full kernel log, this is needed to full understand the
> problem
> 
>> And this issue has not been fixed in kernel version 7.0, which I guess
>> is set to be the next LTS kernel.
> 
> No, it's not.
> 
>> Now, I hope to God that this email finds whoever is responsible for
>> breaking this `amdgpu` firmware before the same thing happens to Linus
>> via Fedora 44, and he goes berserk. I realize that's not a heck of a
>> lot of time, but God only knows HOW many Linux users with AMD GPUs are
>> pissed off by this kind of issue — and I for one can only imagine how
>> Linus' reaction will compare when it happens to HIM!
> 
> FWIW, you seem to assume that the problem hits everyone with a AMD GPUs.
> That's not the case, we'd know this by now. Maybe it's not even
> something that hits everyone with a card/gpu like yours, as some falires
> only show up in certain environments, for example in combination with a
> specific mainboard chipset or firmware.
> 
> Anyway: I CCed a few people that might be able to help. But they likely
> will need at least the full kernel log mentioned above .
> 
> Ciao, Thorsten

Re: [REGRESSION] amdgpu error -22 first encountered in kernel 6.19, not fixed in 7.0

[Timur Kristóf]

Hi,

Thank you Thorsten for forwarding this to me, indeed I am working on improving 
the driver for these GPUs. See my replies below.

On Monday, April 27, 2026 11:37:53 AM Central European Summer Time Thorsten 
Leemhuis wrote:
> [note: please remove Timo Lindfors when replying to this mail and
> include Timur Kristóf – sorry everyone and especially Timo, my mailers
> address book tricked me]
> 
> On 4/27/26 11:35, Linux regression tracking (Thorsten Leemhuis) wrote:
> > On 4/26/26 01:22, Brandon Taylor wrote:
> >> I have an AMD Radeon R9 270X GPU (yes, I know it's old) which used to
> >> default to the `radeon` firmware
> > 
> > FYI, as using the wrong terms can lead to confusion while dealing with
> > bugs: that's not a firmware, that's the driver (sometimes also called
> > kernel module)

- "radeon" is the old kernel driver for these GPUs that didn't support Vulkan 
and therefore was mostly useless for gaming.
- "amdgpu" is the new driver, which supports Vulkan and offers better perf, 
this is the default since Linux 6.19.

I don't think this is a firmware issue.

> > 
> >> on kernel versions up to and
> >> including 6.18. I could manually change this behavior to use the new
> >> `amdgpu` firmware by setting `radeon.si_support=0 amdgpu.si_support=1`
> >> as kernel parameters in the GRUB bootloader. Everything worked fine,
> >> and I was able to play my Windows games without any problems.
> >> 
> >> That was, until kernel version 6.19 came out.>
> >> 6.19, from what I was able to Google, was supposed to have AMD GPUs to
> >> default to the new `amdgpu` firmware. Unfortunately, when I updated
> >> the kernel and rebooted, I got a black screen.
> >> 
> >> Further Google searches led me to reboot, set `nomodeset` in the
> >> kernel parameters, switch to a TTY, login, and investigate the output
> >> of a `dmesg` command — and what I saw ... Well, let's just say it's
> >> the kind of thing that would make Linus go Chernobyl:
> >> 
> >> `amdgpu: probe with driver amdgpu failed with error -22`
> > 

I am sorry this happened and understand the frustration.

Please open an issue here:
https://gitlab.freedesktop.org/drm/amd/-/work_items/new
and upload your full dmesg log. Please ping me personally @Venemo on that 
issue after you opened it so we can start investigating what is wrong there.

I got curious so I plugged in my R9 270X right now. I am typing this email on 
a computer with a Radeon R9 270X using Linux 6.19.12 on Fedora 43 and it works 
fine. We are going to need to need more details in order to diagnose what the 
problem really is. We might ask you to test some proposed fixes, considering 
that the issue doesn't happen on my R9 270X, I don't have a way to verify it.

> >> Now, I hope to God that this email finds whoever is responsible for
> >> breaking this `amdgpu` firmware before the same thing happens to Linus
> >> via Fedora 44, and he goes berserk. I realize that's not a heck of a
> >> lot of time, but God only knows HOW many Linux users with AMD GPUs are
> >> pissed off by this kind of issue — and I for one can only imagine how
> >> Linus' reaction will compare when it happens to HIM!

I'm happy to help, there is no need for personal insults.

(Note that the issue clearly doesn't affect all AMD GPUs and it clearly doesn't 
even affect all R9 270X GPUs.)

Thanks & best regards,
Timur

Re: [REGRESSION] amdgpu error -22 first encountered in kernel 6.19, not fixed in 7.0

[Brandon Taylor]

> I'm happy to help, there is no need for personal insults.

Sorry if I came across as insulting; that was most definitely not my
intention. I simply know Linus' penchant for going insane when an
error is found in his kernel. But I'm thinking Timur's probably right;
this may merely be a "me" problem, with my particular GPU.

At any rate, I've created the issue ticket as requested, and here it
is. https://gitlab.freedesktop.org/drm/amd/-/work-items/5220

Brandon Taylor

On Mon, Apr 27, 2026 at 9:01 AM Timur Kristóf <timur.kristof@gmail.com> wrote:
>
> Hi,
>
> Thank you Thorsten for forwarding this to me, indeed I am working on improving
> the driver for these GPUs. See my replies below.
>
> On Monday, April 27, 2026 11:37:53 AM Central European Summer Time Thorsten
> Leemhuis wrote:
> > [note: please remove Timo Lindfors when replying to this mail and
> > include Timur Kristóf – sorry everyone and especially Timo, my mailers
> > address book tricked me]
> >
> > On 4/27/26 11:35, Linux regression tracking (Thorsten Leemhuis) wrote:
> > > On 4/26/26 01:22, Brandon Taylor wrote:
> > >> I have an AMD Radeon R9 270X GPU (yes, I know it's old) which used to
> > >> default to the `radeon` firmware
> > >
> > > FYI, as using the wrong terms can lead to confusion while dealing with
> > > bugs: that's not a firmware, that's the driver (sometimes also called
> > > kernel module)
>
> - "radeon" is the old kernel driver for these GPUs that didn't support Vulkan
> and therefore was mostly useless for gaming.
> - "amdgpu" is the new driver, which supports Vulkan and offers better perf,
> this is the default since Linux 6.19.
>
> I don't think this is a firmware issue.
>
> > >
> > >> on kernel versions up to and
> > >> including 6.18. I could manually change this behavior to use the new
> > >> `amdgpu` firmware by setting `radeon.si_support=0 amdgpu.si_support=1`
> > >> as kernel parameters in the GRUB bootloader. Everything worked fine,
> > >> and I was able to play my Windows games without any problems.
> > >>
> > >> That was, until kernel version 6.19 came out.>
> > >> 6.19, from what I was able to Google, was supposed to have AMD GPUs to
> > >> default to the new `amdgpu` firmware. Unfortunately, when I updated
> > >> the kernel and rebooted, I got a black screen.
> > >>
> > >> Further Google searches led me to reboot, set `nomodeset` in the
> > >> kernel parameters, switch to a TTY, login, and investigate the output
> > >> of a `dmesg` command — and what I saw ... Well, let's just say it's
> > >> the kind of thing that would make Linus go Chernobyl:
> > >>
> > >> `amdgpu: probe with driver amdgpu failed with error -22`
> > >
>
> I am sorry this happened and understand the frustration.
>
> Please open an issue here:
> https://gitlab.freedesktop.org/drm/amd/-/work_items/new
> and upload your full dmesg log. Please ping me personally @Venemo on that
> issue after you opened it so we can start investigating what is wrong there.
>
> I got curious so I plugged in my R9 270X right now. I am typing this email on
> a computer with a Radeon R9 270X using Linux 6.19.12 on Fedora 43 and it works
> fine. We are going to need to need more details in order to diagnose what the
> problem really is. We might ask you to test some proposed fixes, considering
> that the issue doesn't happen on my R9 270X, I don't have a way to verify it.
>
> > >> Now, I hope to God that this email finds whoever is responsible for
> > >> breaking this `amdgpu` firmware before the same thing happens to Linus
> > >> via Fedora 44, and he goes berserk. I realize that's not a heck of a
> > >> lot of time, but God only knows HOW many Linux users with AMD GPUs are
> > >> pissed off by this kind of issue — and I for one can only imagine how
> > >> Linus' reaction will compare when it happens to HIM!
>
> I'm happy to help, there is no need for personal insults.
>
> (Note that the issue clearly doesn't affect all AMD GPUs and it clearly doesn't
> even affect all R9 270X GPUs.)
>
> Thanks & best regards,
> Timur
>
>
>

Re: [REGRESSION] amdgpu error -22 first encountered in kernel 6.19, not fixed in 7.0

[Brandon Taylor]

[-- Attachment #1: Type: text/plain, Size: 4657 bytes --]

Sometimes I hate copy-and-paste with a passion... Here's the correct link.
https://gitlab.freedesktop.org/drm/amd/-/work_items?show=eyJpaWQiOiI1MjIwIiwiZnVsbF9wYXRoIjoiZHJtL2FtZCIsImlkIjoxNTAxNzl9

On Mon, Apr 27, 2026, 12:17 PM Brandon Taylor <br.ta.2818@gmail.com> wrote:

> > I'm happy to help, there is no need for personal insults.
>
> Sorry if I came across as insulting; that was most definitely not my
> intention. I simply know Linus' penchant for going insane when an
> error is found in his kernel. But I'm thinking Timur's probably right;
> this may merely be a "me" problem, with my particular GPU.
>
> At any rate, I've created the issue ticket as requested, and here it
> is. https://gitlab.freedesktop.org/drm/amd/-/work-items/5220
>
> Brandon Taylor
>
> On Mon, Apr 27, 2026 at 9:01 AM Timur Kristóf <timur.kristof@gmail.com>
> wrote:
> >
> > Hi,
> >
> > Thank you Thorsten for forwarding this to me, indeed I am working on
> improving
> > the driver for these GPUs. See my replies below.
> >
> > On Monday, April 27, 2026 11:37:53 AM Central European Summer Time
> Thorsten
> > Leemhuis wrote:
> > > [note: please remove Timo Lindfors when replying to this mail and
> > > include Timur Kristóf – sorry everyone and especially Timo, my mailers
> > > address book tricked me]
> > >
> > > On 4/27/26 11:35, Linux regression tracking (Thorsten Leemhuis) wrote:
> > > > On 4/26/26 01:22, Brandon Taylor wrote:
> > > >> I have an AMD Radeon R9 270X GPU (yes, I know it's old) which used
> to
> > > >> default to the `radeon` firmware
> > > >
> > > > FYI, as using the wrong terms can lead to confusion while dealing
> with
> > > > bugs: that's not a firmware, that's the driver (sometimes also called
> > > > kernel module)
> >
> > - "radeon" is the old kernel driver for these GPUs that didn't support
> Vulkan
> > and therefore was mostly useless for gaming.
> > - "amdgpu" is the new driver, which supports Vulkan and offers better
> perf,
> > this is the default since Linux 6.19.
> >
> > I don't think this is a firmware issue.
> >
> > > >
> > > >> on kernel versions up to and
> > > >> including 6.18. I could manually change this behavior to use the new
> > > >> `amdgpu` firmware by setting `radeon.si_support=0
> amdgpu.si_support=1`
> > > >> as kernel parameters in the GRUB bootloader. Everything worked fine,
> > > >> and I was able to play my Windows games without any problems.
> > > >>
> > > >> That was, until kernel version 6.19 came out.>
> > > >> 6.19, from what I was able to Google, was supposed to have AMD GPUs
> to
> > > >> default to the new `amdgpu` firmware. Unfortunately, when I updated
> > > >> the kernel and rebooted, I got a black screen.
> > > >>
> > > >> Further Google searches led me to reboot, set `nomodeset` in the
> > > >> kernel parameters, switch to a TTY, login, and investigate the
> output
> > > >> of a `dmesg` command — and what I saw ... Well, let's just say it's
> > > >> the kind of thing that would make Linus go Chernobyl:
> > > >>
> > > >> `amdgpu: probe with driver amdgpu failed with error -22`
> > > >
> >
> > I am sorry this happened and understand the frustration.
> >
> > Please open an issue here:
> > https://gitlab.freedesktop.org/drm/amd/-/work_items/new
> > and upload your full dmesg log. Please ping me personally @Venemo on that
> > issue after you opened it so we can start investigating what is wrong
> there.
> >
> > I got curious so I plugged in my R9 270X right now. I am typing this
> email on
> > a computer with a Radeon R9 270X using Linux 6.19.12 on Fedora 43 and it
> works
> > fine. We are going to need to need more details in order to diagnose
> what the
> > problem really is. We might ask you to test some proposed fixes,
> considering
> > that the issue doesn't happen on my R9 270X, I don't have a way to
> verify it.
> >
> > > >> Now, I hope to God that this email finds whoever is responsible for
> > > >> breaking this `amdgpu` firmware before the same thing happens to
> Linus
> > > >> via Fedora 44, and he goes berserk. I realize that's not a heck of a
> > > >> lot of time, but God only knows HOW many Linux users with AMD GPUs
> are
> > > >> pissed off by this kind of issue — and I for one can only imagine
> how
> > > >> Linus' reaction will compare when it happens to HIM!
> >
> > I'm happy to help, there is no need for personal insults.
> >
> > (Note that the issue clearly doesn't affect all AMD GPUs and it clearly
> doesn't
> > even affect all R9 270X GPUs.)
> >
> > Thanks & best regards,
> > Timur
> >
> >
> >
>

[-- Attachment #2: Type: text/html, Size: 6144 bytes --]

Re: [REGRESSION] amdgpu error -22 first encountered in kernel 6.19, not fixed in 7.0

[Brandon Taylor]

Sometimes I hate copy-and-paste with a passion... Here's the corrected
link. https://gitlab.freedesktop.org/drm/amd/-/work_items?show=eyJpaWQiOiI1MjIwIiwiZnVsbF9wYXRoIjoiZHJtL2FtZCIsImlkIjoxNTAxNzl9

On Mon, Apr 27, 2026 at 12:17 PM Brandon Taylor <br.ta.2818@gmail.com> wrote:
>
> > I'm happy to help, there is no need for personal insults.
>
> Sorry if I came across as insulting; that was most definitely not my
> intention. I simply know Linus' penchant for going insane when an
> error is found in his kernel. But I'm thinking Timur's probably right;
> this may merely be a "me" problem, with my particular GPU.
>
> At any rate, I've created the issue ticket as requested, and here it
> is. https://gitlab.freedesktop.org/drm/amd/-/work-items/5220
>
> Brandon Taylor
>
> On Mon, Apr 27, 2026 at 9:01 AM Timur Kristóf <timur.kristof@gmail.com> wrote:
> >
> > Hi,
> >
> > Thank you Thorsten for forwarding this to me, indeed I am working on improving
> > the driver for these GPUs. See my replies below.
> >
> > On Monday, April 27, 2026 11:37:53 AM Central European Summer Time Thorsten
> > Leemhuis wrote:
> > > [note: please remove Timo Lindfors when replying to this mail and
> > > include Timur Kristóf – sorry everyone and especially Timo, my mailers
> > > address book tricked me]
> > >
> > > On 4/27/26 11:35, Linux regression tracking (Thorsten Leemhuis) wrote:
> > > > On 4/26/26 01:22, Brandon Taylor wrote:
> > > >> I have an AMD Radeon R9 270X GPU (yes, I know it's old) which used to
> > > >> default to the `radeon` firmware
> > > >
> > > > FYI, as using the wrong terms can lead to confusion while dealing with
> > > > bugs: that's not a firmware, that's the driver (sometimes also called
> > > > kernel module)
> >
> > - "radeon" is the old kernel driver for these GPUs that didn't support Vulkan
> > and therefore was mostly useless for gaming.
> > - "amdgpu" is the new driver, which supports Vulkan and offers better perf,
> > this is the default since Linux 6.19.
> >
> > I don't think this is a firmware issue.
> >
> > > >
> > > >> on kernel versions up to and
> > > >> including 6.18. I could manually change this behavior to use the new
> > > >> `amdgpu` firmware by setting `radeon.si_support=0 amdgpu.si_support=1`
> > > >> as kernel parameters in the GRUB bootloader. Everything worked fine,
> > > >> and I was able to play my Windows games without any problems.
> > > >>
> > > >> That was, until kernel version 6.19 came out.>
> > > >> 6.19, from what I was able to Google, was supposed to have AMD GPUs to
> > > >> default to the new `amdgpu` firmware. Unfortunately, when I updated
> > > >> the kernel and rebooted, I got a black screen.
> > > >>
> > > >> Further Google searches led me to reboot, set `nomodeset` in the
> > > >> kernel parameters, switch to a TTY, login, and investigate the output
> > > >> of a `dmesg` command — and what I saw ... Well, let's just say it's
> > > >> the kind of thing that would make Linus go Chernobyl:
> > > >>
> > > >> `amdgpu: probe with driver amdgpu failed with error -22`
> > > >
> >
> > I am sorry this happened and understand the frustration.
> >
> > Please open an issue here:
> > https://gitlab.freedesktop.org/drm/amd/-/work_items/new
> > and upload your full dmesg log. Please ping me personally @Venemo on that
> > issue after you opened it so we can start investigating what is wrong there.
> >
> > I got curious so I plugged in my R9 270X right now. I am typing this email on
> > a computer with a Radeon R9 270X using Linux 6.19.12 on Fedora 43 and it works
> > fine. We are going to need to need more details in order to diagnose what the
> > problem really is. We might ask you to test some proposed fixes, considering
> > that the issue doesn't happen on my R9 270X, I don't have a way to verify it.
> >
> > > >> Now, I hope to God that this email finds whoever is responsible for
> > > >> breaking this `amdgpu` firmware before the same thing happens to Linus
> > > >> via Fedora 44, and he goes berserk. I realize that's not a heck of a
> > > >> lot of time, but God only knows HOW many Linux users with AMD GPUs are
> > > >> pissed off by this kind of issue — and I for one can only imagine how
> > > >> Linus' reaction will compare when it happens to HIM!
> >
> > I'm happy to help, there is no need for personal insults.
> >
> > (Note that the issue clearly doesn't affect all AMD GPUs and it clearly doesn't
> > even affect all R9 270X GPUs.)
> >
> > Thanks & best regards,
> > Timur
> >
> >
> >