[PATCH v2 0/1] Correct Video Corruption for EFI Plus Bochs/Cirrus

[PATCH v2 0/1] Correct Video Corruption for EFI Plus Bochs/Cirrus

[Andrew Hamilton]

Disable building video_bochs and video_cirrus for EFI x86 platforms (i386_efi
and x86_64_efi). For x86 EFI platforms efi_gop or efi_uga should provide the
video support. Currently, in EFI platforms there is a risk that both the
legacy device is present and the EFI video interface - which can cause
unexpected video corruption when a mix of the two interfaces are used. This
can happen when all_video is loaded for simplicity in a GRUB config.

There is some additional analysis from the Qemu team in this thread:
https://gitlab.com/qemu-project/qemu/-/issues/2562

This is an issue on some major distros such as Debian on EFI
when running under Qemu and potentially other scenarios.

There is some discussion on the Debian side in this thread:
https://salsa.debian.org/kernel-team/linux/-/merge_requests/1453

This relates to this discussion from grub-devel also:
https://lists.gnu.org/archive/html/grub-devel/2025-09/msg00244.html

Following this change, for x86_64_efi the following modules would
be loaded by "all_video": "efi_gop" "efi_uga"

Link to v1 thread:
https://lists.gnu.org/archive/html/grub-devel/2025-09/msg00254.html

Fixes: https://savannah.gnu.org/bugs/index.php?66200

Changes v1 => v2:
Also remove video_cirrus when building for i386 EFI or x86-64 EFI

Andrew Hamilton (1):
  build: Disable building video_bochs/cirrus for EFI x86 platforms

 grub-core/Makefile.core.def | 12 ++++++++++--
 1 file changed, 10 insertions(+), 2 deletions(-)

-- 
2.47.3


_______________________________________________
Grub-devel mailing list
Grub-devel@gnu.org
https://lists.gnu.org/mailman/listinfo/grub-devel

[PATCH v2 1/1] build: Disable building video_bochs/cirrus for EFI x86

[Andrew Hamilton]

Disable building video_bochs and video_cirrus for EFI x86 platforms (i386_efi
and x86_64_efi). For x86 EFI platforms efi_gop or efi_uga should provide the
video support. Currently, in EFI platforms there is a risk that both the
legacy device is present and the EFI video interface - which can cause
unexpected video corruption when a mix of the two interfaces are used. This
can happen when all_video is loaded for simplicity in a GRUB config.

Fixes: https://savannah.gnu.org/bugs/index.php?66200

Signed-off-by: Andrew Hamilton <adhamilt@gmail.com>
---
 grub-core/Makefile.core.def | 12 ++++++++++--
 1 file changed, 10 insertions(+), 2 deletions(-)

diff --git a/grub-core/Makefile.core.def b/grub-core/Makefile.core.def
index 0fcf67f9d..fad15ca4e 100644
--- a/grub-core/Makefile.core.def
+++ b/grub-core/Makefile.core.def
@@ -2163,13 +2163,21 @@ module = {
 module = {
   name = video_cirrus;
   x86 = video/cirrus.c;
-  enable = x86;
+  enable = i386_pc;
+  enable = i386_qemu;
+  enable = i386_coreboot;
+  enable = i386_multiboot;
+  enable = i386_ieee1275;
 };
 
 module = {
   name = video_bochs;
   x86 = video/bochs.c;
-  enable = x86;
+  enable = i386_pc;
+  enable = i386_qemu;
+  enable = i386_coreboot;
+  enable = i386_multiboot;
+  enable = i386_ieee1275;
 };
 
 module = {
-- 
2.47.3


_______________________________________________
Grub-devel mailing list
Grub-devel@gnu.org
https://lists.gnu.org/mailman/listinfo/grub-devel

Re: [PATCH v2 0/1] Correct Video Corruption for EFI Plus Bochs/Cirrus

[Askar Safin via Grub-devel]

 ---- On Tue, 30 Sep 2025 06:23:39 +0400  Andrew Hamilton <adhamilt@gmail.com> wrote --- 
 > Disable building video_bochs and video_cirrus for EFI x86 platforms (i386_efi

I tested this patch, and it indeed solves the bug. (But I tested on x86_64 only.)

Tested-by: Askar Safin <safinaskar@zohomail.com>

-- 
Askar Safin
https://types.pl/@safinaskar


_______________________________________________
Grub-devel mailing list
Grub-devel@gnu.org
https://lists.gnu.org/mailman/listinfo/grub-devel

Re: [PATCH v2 0/1] Correct Video Corruption for EFI Plus Bochs/Cirrus

[Askar Safin]

Andrew Hamilton <adhamilt@gmail.com>:
> Disable building video_bochs and video_cirrus for EFI x86 platforms (i386_efi

Why this patch is not yet applied?

-- 
Askar Safin

_______________________________________________
Grub-devel mailing list
Grub-devel@gnu.org
https://lists.gnu.org/mailman/listinfo/grub-devel

Re: [PATCH v2 0/1] Correct Video Corruption for EFI Plus Bochs/Cirrus

[Andrew Hamilton]

[-- Attachment #1.1: Type: text/plain, Size: 409 bytes --]

Daniel - can you take a look at this one to see if we can get it pulled in
to help those affected by this issue?

Thanks
Andrew

On Tue, Oct 14, 2025 at 3:59 AM Askar Safin <safinaskar@gmail.com> wrote:

> Andrew Hamilton <adhamilt@gmail.com>:
> > Disable building video_bochs and video_cirrus for EFI x86 platforms
> (i386_efi
>
> Why this patch is not yet applied?
>
> --
> Askar Safin
>

[-- Attachment #1.2: Type: text/html, Size: 839 bytes --]

[-- Attachment #2: Type: text/plain, Size: 141 bytes --]

_______________________________________________
Grub-devel mailing list
Grub-devel@gnu.org
https://lists.gnu.org/mailman/listinfo/grub-devel

Re: [PATCH v2 0/1] Correct Video Corruption for EFI Plus Bochs/Cirrus

[Daniel Kiper via Grub-devel]

On Tue, Oct 14, 2025 at 10:17:45AM -0500, Andrew Hamilton wrote:
> Daniel - can you take a look at this one to see if we can get it pulled in to
> help those affected by this issue?

I am not entirely sure it [1] is correct. I have pinged and CC-ed Vladimir...

Daniel

[1] https://lore.kernel.org/grub-devel/20250930022340.146838-1-adhamilt@gmail.com/

_______________________________________________
Grub-devel mailing list
Grub-devel@gnu.org
https://lists.gnu.org/mailman/listinfo/grub-devel

Re: [PATCH v2 0/1] Correct Video Corruption for EFI Plus Bochs/Cirrus

[Gerd Hoffmann via Grub-devel]

On Wed, Oct 15, 2025 at 03:30:53PM +0200, Daniel Kiper via Grub-devel wrote:
> On Tue, Oct 14, 2025 at 10:17:45AM -0500, Andrew Hamilton wrote:
> > Daniel - can you take a look at this one to see if we can get it pulled in to
> > help those affected by this issue?
> 
> I am not entirely sure it [1] is correct.

Well.  Underlying problem is that there are two drivers active for one
piece of hardware.  The driver which comes with the EFI firmware, and
the grub driver.  Which is /never/ a good idea.  In this specific case
the effect is very visible, with other hardware the problems can be more
obscure or with luck not present.

The by far easiest way to handle this properly to allow only the efi_*
drivers on all EFI platforms.  The patch does just that (limited to
video drivers).

Alternatively you can check for conflicts.  Check which physical device
backs efi_gop, block that from being used by grub drivers.  Same for
all other device types.

The latter would handle the corner case that there is hardware which is
(a) supported by grub, and (b) not supported by the firmware, and (c)
essential to boot the system.  Not sure how relevant that is in
practice, most EFI firmware I've seen recently is pretty complete when
it comes to hardware support.  NIC drivers are missing sometimes, iPXE
to the rescue ;)

Note that the same conflict exists for filesystem drivers too, usually
both EFI and grub FAT drivers are active.  Luckily write access to
the ESP is very rare, so this isn't much of a problem in practice.

take care,
  Gerd


_______________________________________________
Grub-devel mailing list
Grub-devel@gnu.org
https://lists.gnu.org/mailman/listinfo/grub-devel

Re: [PATCH v2 0/1] Correct Video Corruption for EFI Plus Bochs/Cirrus

[Daniel Kiper via Grub-devel]

On Wed, Oct 15, 2025 at 05:16:05PM +0200, Gerd Hoffmann wrote:
> On Wed, Oct 15, 2025 at 03:30:53PM +0200, Daniel Kiper via Grub-devel wrote:
> > On Tue, Oct 14, 2025 at 10:17:45AM -0500, Andrew Hamilton wrote:
> > > Daniel - can you take a look at this one to see if we can get it pulled in to
> > > help those affected by this issue?
> >
> > I am not entirely sure it [1] is correct.
>
> Well.  Underlying problem is that there are two drivers active for one
> piece of hardware.  The driver which comes with the EFI firmware, and
> the grub driver.  Which is /never/ a good idea.  In this specific case
> the effect is very visible, with other hardware the problems can be more
> obscure or with luck not present.
>
> The by far easiest way to handle this properly to allow only the efi_*
> drivers on all EFI platforms.  The patch does just that (limited to
> video drivers).
>
> Alternatively you can check for conflicts.  Check which physical device
> backs efi_gop, block that from being used by grub drivers.  Same for
> all other device types.
>
> The latter would handle the corner case that there is hardware which is
> (a) supported by grub, and (b) not supported by the firmware, and (c)
> essential to boot the system.  Not sure how relevant that is in
> practice, most EFI firmware I've seen recently is pretty complete when
> it comes to hardware support.  NIC drivers are missing sometimes, iPXE
> to the rescue ;)
>
> Note that the same conflict exists for filesystem drivers too, usually
> both EFI and grub FAT drivers are active.  Luckily write access to
> the ESP is very rare, so this isn't much of a problem in practice.

I am not saying it is not a problem. I rather saying that I do not like
the solution. Or even we... I discussed the issue with Vladimir and he
agrees with me.

Anyway, why cannot you use GRUB_VIDEO_BACKEND config variable?
If it is not acceptable please propose a solution without disabling
drivers currently build and provided by the GRUB. We are OK with not
loading some drivers but we are against disabling them at build.

Daniel

_______________________________________________
Grub-devel mailing list
Grub-devel@gnu.org
https://lists.gnu.org/mailman/listinfo/grub-devel

Re: [PATCH v2 0/1] Correct Video Corruption for EFI Plus Bochs/Cirrus

[Askar Safin]

Fixing this problem is needed for Debian to enable
CONFIG_DRM_SIMPLEDRM for its kernels (this is already done by many
other distros).
(see this thread:
https://salsa.debian.org/kernel-team/linux/-/merge_requests/1453#note_603600
)
So I'm asking will GRUB merge at least a partial solution to
this problem in reasonable time?

On Wed, Oct 15, 2025 at 7:49 PM Daniel Kiper <daniel.kiper@oracle.com> wrote:
> Anyway, why cannot you use GRUB_VIDEO_BACKEND config variable?

I want GRUB to work by default. :)

> If it is not acceptable please propose a solution without disabling
> drivers currently build and provided by the GRUB. We are OK with not

Then make "all_video" depend on "efi_*" only on efi-x86_64.

-- 
Askar Safin

_______________________________________________
Grub-devel mailing list
Grub-devel@gnu.org
https://lists.gnu.org/mailman/listinfo/grub-devel

Re: [PATCH v2 0/1] Correct Video Corruption for EFI Plus Bochs/Cirrus

[Daniel Kiper via Grub-devel]

On Thu, Oct 16, 2025 at 10:07:36AM +0300, Askar Safin wrote:
> Fixing this problem is needed for Debian to enable
> CONFIG_DRM_SIMPLEDRM for its kernels (this is already done by many
> other distros).
> (see this thread:
> https://salsa.debian.org/kernel-team/linux/-/merge_requests/1453#note_603600
> )
> So I'm asking will GRUB merge at least a partial solution to
> this problem in reasonable time?

Propose an acceptable solution and we will take it...

> On Wed, Oct 15, 2025 at 7:49 PM Daniel Kiper <daniel.kiper@oracle.com> wrote:
> > Anyway, why cannot you use GRUB_VIDEO_BACKEND config variable?
>
> I want GRUB to work by default. :)
>
> > If it is not acceptable please propose a solution without disabling
> > drivers currently build and provided by the GRUB. We are OK with not
>
> Then make "all_video" depend on "efi_*" only on efi-x86_64.

I think it is much easier and sensible to tweak GRUB shell scripts to
make it work.

Daniel

_______________________________________________
Grub-devel mailing list
Grub-devel@gnu.org
https://lists.gnu.org/mailman/listinfo/grub-devel

Re: [PATCH v2 0/1] Correct Video Corruption for EFI Plus Bochs/Cirrus

[Daniel Kiper via Grub-devel]

On Thu, Oct 16, 2025 at 12:38:57PM +0200, Daniel Kiper wrote:
> On Thu, Oct 16, 2025 at 10:07:36AM +0300, Askar Safin wrote:
> > Fixing this problem is needed for Debian to enable
> > CONFIG_DRM_SIMPLEDRM for its kernels (this is already done by many
> > other distros).
> > (see this thread:
> > https://salsa.debian.org/kernel-team/linux/-/merge_requests/1453#note_603600
> > )
> > So I'm asking will GRUB merge at least a partial solution to
> > this problem in reasonable time?
>
> Propose an acceptable solution and we will take it...
>
> > On Wed, Oct 15, 2025 at 7:49 PM Daniel Kiper <daniel.kiper@oracle.com> wrote:
> > > Anyway, why cannot you use GRUB_VIDEO_BACKEND config variable?
> >
> > I want GRUB to work by default. :)
> >
> > > If it is not acceptable please propose a solution without disabling
> > > drivers currently build and provided by the GRUB. We are OK with not
> >
> > Then make "all_video" depend on "efi_*" only on efi-x86_64.
>
> I think it is much easier and sensible to tweak GRUB shell scripts to
> make it work.

To be precise, I am OK with a patch doing it properly. And it would be
nice to have a flag to enable original behavior just in case.

Daniel

_______________________________________________
Grub-devel mailing list
Grub-devel@gnu.org
https://lists.gnu.org/mailman/listinfo/grub-devel