* [PATCH] fbcon: Check font dimension limits
@ 2023-01-26 0:49 ` Samuel Thibault
0 siblings, 0 replies; 15+ messages in thread
From: Samuel Thibault @ 2023-01-26 0:49 UTC (permalink / raw)
To: gregkh, Daniel Vetter, Helge Deller
Cc: linux-fbdev, linux-kernel, stable, dri-devel, Sanan Hasanov,
Samuel Thibault
blit_x and blit_y are uint32_t, so fbcon currently cannot support fonts
larger than 32x32.
The 32x32 case also needs shifting an unsigned int, to properly set bit
31, otherwise we get "UBSAN: shift-out-of-bounds in fbcon_set_font",
as reported on
http://lore.kernel.org/all/IA1PR07MB98308653E259A6F2CE94A4AFABCE9@IA1PR07MB9830.namprd07.prod.outlook.com
Kernel Branch: 6.2.0-rc5-next-20230124
Kernel config: https://drive.google.com/file/d/1F-LszDAizEEH0ZX0HcSR06v5q8FPl2Uv/view?usp=sharing
Reproducer: https://drive.google.com/file/d/1mP1jcLBY7vWCNM60OMf-ogw-urQRjNrm/view?usp=sharing
Reported-by: Sanan Hasanov <sanan.hasanov@Knights.ucf.edu>
Signed-off-by: Samuel Thibault <samuel.thibault@ens-lyon.org>
Index: linux-6.0/drivers/video/fbdev/core/fbcon.c
===================================================================
--- linux-6.0.orig/drivers/video/fbdev/core/fbcon.c
+++ linux-6.0/drivers/video/fbdev/core/fbcon.c
@@ -2489,9 +2489,12 @@ static int fbcon_set_font(struct vc_data
h > FBCON_SWAP(info->var.rotate, info->var.yres, info->var.xres))
return -EINVAL;
+ if (font->width > 32 || font->height > 32)
+ return -EINVAL;
+
/* Make sure drawing engine can handle the font */
- if (!(info->pixmap.blit_x & (1 << (font->width - 1))) ||
- !(info->pixmap.blit_y & (1 << (font->height - 1))))
+ if (!(info->pixmap.blit_x & (1U << (font->width - 1))) ||
+ !(info->pixmap.blit_y & (1U << (font->height - 1))))
return -EINVAL;
/* Make sure driver can handle the font length */
^ permalink raw reply [flat|nested] 15+ messages in thread
* Re: [PATCH] fbcon: Check font dimension limits
2023-01-26 0:49 ` Samuel Thibault
@ 2023-01-26 0:51 ` Samuel Thibault
-1 siblings, 0 replies; 15+ messages in thread
From: Samuel Thibault @ 2023-01-26 0:51 UTC (permalink / raw)
To: gregkh, Daniel Vetter, Helge Deller
Cc: linux-fbdev, dri-devel, linux-kernel, stable, Sanan Hasanov
(FYI, the UB was already a problem before my big-font patch submission,
the checks I add here are already making sense in previous versions
anyway)
Samuel Thibault, le jeu. 26 janv. 2023 01:49:12 +0100, a ecrit:
> blit_x and blit_y are uint32_t, so fbcon currently cannot support fonts
> larger than 32x32.
>
> The 32x32 case also needs shifting an unsigned int, to properly set bit
> 31, otherwise we get "UBSAN: shift-out-of-bounds in fbcon_set_font",
> as reported on
>
> http://lore.kernel.org/all/IA1PR07MB98308653E259A6F2CE94A4AFABCE9@IA1PR07MB9830.namprd07.prod.outlook.com
> Kernel Branch: 6.2.0-rc5-next-20230124
> Kernel config: https://drive.google.com/file/d/1F-LszDAizEEH0ZX0HcSR06v5q8FPl2Uv/view?usp=sharing
> Reproducer: https://drive.google.com/file/d/1mP1jcLBY7vWCNM60OMf-ogw-urQRjNrm/view?usp=sharing
>
> Reported-by: Sanan Hasanov <sanan.hasanov@Knights.ucf.edu>
> Signed-off-by: Samuel Thibault <samuel.thibault@ens-lyon.org>
>
> Index: linux-6.0/drivers/video/fbdev/core/fbcon.c
> ===================================================================
> --- linux-6.0.orig/drivers/video/fbdev/core/fbcon.c
> +++ linux-6.0/drivers/video/fbdev/core/fbcon.c
> @@ -2489,9 +2489,12 @@ static int fbcon_set_font(struct vc_data
> h > FBCON_SWAP(info->var.rotate, info->var.yres, info->var.xres))
> return -EINVAL;
>
> + if (font->width > 32 || font->height > 32)
> + return -EINVAL;
> +
> /* Make sure drawing engine can handle the font */
> - if (!(info->pixmap.blit_x & (1 << (font->width - 1))) ||
> - !(info->pixmap.blit_y & (1 << (font->height - 1))))
> + if (!(info->pixmap.blit_x & (1U << (font->width - 1))) ||
> + !(info->pixmap.blit_y & (1U << (font->height - 1))))
> return -EINVAL;
>
> /* Make sure driver can handle the font length */
>
>
^ permalink raw reply [flat|nested] 15+ messages in thread
* Re: [PATCH] fbcon: Check font dimension limits
@ 2023-01-26 0:51 ` Samuel Thibault
0 siblings, 0 replies; 15+ messages in thread
From: Samuel Thibault @ 2023-01-26 0:51 UTC (permalink / raw)
To: gregkh, Daniel Vetter, Helge Deller
Cc: Sanan Hasanov, stable, linux-fbdev, linux-kernel, dri-devel
(FYI, the UB was already a problem before my big-font patch submission,
the checks I add here are already making sense in previous versions
anyway)
Samuel Thibault, le jeu. 26 janv. 2023 01:49:12 +0100, a ecrit:
> blit_x and blit_y are uint32_t, so fbcon currently cannot support fonts
> larger than 32x32.
>
> The 32x32 case also needs shifting an unsigned int, to properly set bit
> 31, otherwise we get "UBSAN: shift-out-of-bounds in fbcon_set_font",
> as reported on
>
> http://lore.kernel.org/all/IA1PR07MB98308653E259A6F2CE94A4AFABCE9@IA1PR07MB9830.namprd07.prod.outlook.com
> Kernel Branch: 6.2.0-rc5-next-20230124
> Kernel config: https://drive.google.com/file/d/1F-LszDAizEEH0ZX0HcSR06v5q8FPl2Uv/view?usp=sharing
> Reproducer: https://drive.google.com/file/d/1mP1jcLBY7vWCNM60OMf-ogw-urQRjNrm/view?usp=sharing
>
> Reported-by: Sanan Hasanov <sanan.hasanov@Knights.ucf.edu>
> Signed-off-by: Samuel Thibault <samuel.thibault@ens-lyon.org>
>
> Index: linux-6.0/drivers/video/fbdev/core/fbcon.c
> ===================================================================
> --- linux-6.0.orig/drivers/video/fbdev/core/fbcon.c
> +++ linux-6.0/drivers/video/fbdev/core/fbcon.c
> @@ -2489,9 +2489,12 @@ static int fbcon_set_font(struct vc_data
> h > FBCON_SWAP(info->var.rotate, info->var.yres, info->var.xres))
> return -EINVAL;
>
> + if (font->width > 32 || font->height > 32)
> + return -EINVAL;
> +
> /* Make sure drawing engine can handle the font */
> - if (!(info->pixmap.blit_x & (1 << (font->width - 1))) ||
> - !(info->pixmap.blit_y & (1 << (font->height - 1))))
> + if (!(info->pixmap.blit_x & (1U << (font->width - 1))) ||
> + !(info->pixmap.blit_y & (1U << (font->height - 1))))
> return -EINVAL;
>
> /* Make sure driver can handle the font length */
>
>
^ permalink raw reply [flat|nested] 15+ messages in thread
* Re: [PATCH] fbcon: Check font dimension limits
2023-01-26 0:49 ` Samuel Thibault
@ 2023-01-26 7:43 ` Greg KH
-1 siblings, 0 replies; 15+ messages in thread
From: Greg KH @ 2023-01-26 7:43 UTC (permalink / raw)
To: Samuel Thibault
Cc: Daniel Vetter, Helge Deller, linux-fbdev, dri-devel,
linux-kernel, stable, Sanan Hasanov
On Thu, Jan 26, 2023 at 01:49:12AM +0100, Samuel Thibault wrote:
> blit_x and blit_y are uint32_t, so fbcon currently cannot support fonts
> larger than 32x32.
"u32" you mean, right?
> The 32x32 case also needs shifting an unsigned int, to properly set bit
> 31, otherwise we get "UBSAN: shift-out-of-bounds in fbcon_set_font",
> as reported on
>
> http://lore.kernel.org/all/IA1PR07MB98308653E259A6F2CE94A4AFABCE9@IA1PR07MB9830.namprd07.prod.outlook.com
Odd blank line?
> Kernel Branch: 6.2.0-rc5-next-20230124
> Kernel config: https://drive.google.com/file/d/1F-LszDAizEEH0ZX0HcSR06v5q8FPl2Uv/view?usp=sharing
> Reproducer: https://drive.google.com/file/d/1mP1jcLBY7vWCNM60OMf-ogw-urQRjNrm/view?usp=sharing
What are all of these lines for?
>
> Reported-by: Sanan Hasanov <sanan.hasanov@Knights.ucf.edu>
> Signed-off-by: Samuel Thibault <samuel.thibault@ens-lyon.org>
What commit id does this fix? Should it go to stable kernels?
>
> Index: linux-6.0/drivers/video/fbdev/core/fbcon.c
> ===================================================================
> --- linux-6.0.orig/drivers/video/fbdev/core/fbcon.c
> +++ linux-6.0/drivers/video/fbdev/core/fbcon.c
> @@ -2489,9 +2489,12 @@ static int fbcon_set_font(struct vc_data
> h > FBCON_SWAP(info->var.rotate, info->var.yres, info->var.xres))
> return -EINVAL;
>
> + if (font->width > 32 || font->height > 32)
> + return -EINVAL;
> +
> /* Make sure drawing engine can handle the font */
> - if (!(info->pixmap.blit_x & (1 << (font->width - 1))) ||
> - !(info->pixmap.blit_y & (1 << (font->height - 1))))
> + if (!(info->pixmap.blit_x & (1U << (font->width - 1))) ||
> + !(info->pixmap.blit_y & (1U << (font->height - 1))))
Are you sure this is still needed with the above check added? If so,
why? What is the difference in the compiled code?
thanks,
greg k-h
^ permalink raw reply [flat|nested] 15+ messages in thread
* Re: [PATCH] fbcon: Check font dimension limits
@ 2023-01-26 7:43 ` Greg KH
0 siblings, 0 replies; 15+ messages in thread
From: Greg KH @ 2023-01-26 7:43 UTC (permalink / raw)
To: Samuel Thibault
Cc: linux-fbdev, Sanan Hasanov, Helge Deller, linux-kernel,
dri-devel, stable
On Thu, Jan 26, 2023 at 01:49:12AM +0100, Samuel Thibault wrote:
> blit_x and blit_y are uint32_t, so fbcon currently cannot support fonts
> larger than 32x32.
"u32" you mean, right?
> The 32x32 case also needs shifting an unsigned int, to properly set bit
> 31, otherwise we get "UBSAN: shift-out-of-bounds in fbcon_set_font",
> as reported on
>
> http://lore.kernel.org/all/IA1PR07MB98308653E259A6F2CE94A4AFABCE9@IA1PR07MB9830.namprd07.prod.outlook.com
Odd blank line?
> Kernel Branch: 6.2.0-rc5-next-20230124
> Kernel config: https://drive.google.com/file/d/1F-LszDAizEEH0ZX0HcSR06v5q8FPl2Uv/view?usp=sharing
> Reproducer: https://drive.google.com/file/d/1mP1jcLBY7vWCNM60OMf-ogw-urQRjNrm/view?usp=sharing
What are all of these lines for?
>
> Reported-by: Sanan Hasanov <sanan.hasanov@Knights.ucf.edu>
> Signed-off-by: Samuel Thibault <samuel.thibault@ens-lyon.org>
What commit id does this fix? Should it go to stable kernels?
>
> Index: linux-6.0/drivers/video/fbdev/core/fbcon.c
> ===================================================================
> --- linux-6.0.orig/drivers/video/fbdev/core/fbcon.c
> +++ linux-6.0/drivers/video/fbdev/core/fbcon.c
> @@ -2489,9 +2489,12 @@ static int fbcon_set_font(struct vc_data
> h > FBCON_SWAP(info->var.rotate, info->var.yres, info->var.xres))
> return -EINVAL;
>
> + if (font->width > 32 || font->height > 32)
> + return -EINVAL;
> +
> /* Make sure drawing engine can handle the font */
> - if (!(info->pixmap.blit_x & (1 << (font->width - 1))) ||
> - !(info->pixmap.blit_y & (1 << (font->height - 1))))
> + if (!(info->pixmap.blit_x & (1U << (font->width - 1))) ||
> + !(info->pixmap.blit_y & (1U << (font->height - 1))))
Are you sure this is still needed with the above check added? If so,
why? What is the difference in the compiled code?
thanks,
greg k-h
^ permalink raw reply [flat|nested] 15+ messages in thread
* Re: [PATCH] fbcon: Check font dimension limits
2023-01-26 7:43 ` Greg KH
@ 2023-01-26 9:08 ` Jiri Slaby
-1 siblings, 0 replies; 15+ messages in thread
From: Jiri Slaby @ 2023-01-26 9:08 UTC (permalink / raw)
To: Greg KH, Samuel Thibault
Cc: Daniel Vetter, Helge Deller, linux-fbdev, dri-devel,
linux-kernel, stable, Sanan Hasanov
On 26. 01. 23, 8:43, Greg KH wrote:
>> --- linux-6.0.orig/drivers/video/fbdev/core/fbcon.c
>> +++ linux-6.0/drivers/video/fbdev/core/fbcon.c
>> @@ -2489,9 +2489,12 @@ static int fbcon_set_font(struct vc_data
>> h > FBCON_SWAP(info->var.rotate, info->var.yres, info->var.xres))
>> return -EINVAL;
>>
>> + if (font->width > 32 || font->height > 32)
>> + return -EINVAL;
>> +
>> /* Make sure drawing engine can handle the font */
>> - if (!(info->pixmap.blit_x & (1 << (font->width - 1))) ||
>> - !(info->pixmap.blit_y & (1 << (font->height - 1))))
>> + if (!(info->pixmap.blit_x & (1U << (font->width - 1))) ||
>> + !(info->pixmap.blit_y & (1U << (font->height - 1))))
>
> Are you sure this is still needed with the above check added? If so,
> why? What is the difference in the compiled code?
For font->{width,height} == 32, definitely. IMO, 1 << 31 is undefined as
1 << 31 cannot be represented by an (signed) int.
--
js
suse labs
^ permalink raw reply [flat|nested] 15+ messages in thread* Re: [PATCH] fbcon: Check font dimension limits
@ 2023-01-26 9:08 ` Jiri Slaby
0 siblings, 0 replies; 15+ messages in thread
From: Jiri Slaby @ 2023-01-26 9:08 UTC (permalink / raw)
To: Greg KH, Samuel Thibault
Cc: linux-fbdev, Sanan Hasanov, Helge Deller, linux-kernel,
dri-devel, stable
On 26. 01. 23, 8:43, Greg KH wrote:
>> --- linux-6.0.orig/drivers/video/fbdev/core/fbcon.c
>> +++ linux-6.0/drivers/video/fbdev/core/fbcon.c
>> @@ -2489,9 +2489,12 @@ static int fbcon_set_font(struct vc_data
>> h > FBCON_SWAP(info->var.rotate, info->var.yres, info->var.xres))
>> return -EINVAL;
>>
>> + if (font->width > 32 || font->height > 32)
>> + return -EINVAL;
>> +
>> /* Make sure drawing engine can handle the font */
>> - if (!(info->pixmap.blit_x & (1 << (font->width - 1))) ||
>> - !(info->pixmap.blit_y & (1 << (font->height - 1))))
>> + if (!(info->pixmap.blit_x & (1U << (font->width - 1))) ||
>> + !(info->pixmap.blit_y & (1U << (font->height - 1))))
>
> Are you sure this is still needed with the above check added? If so,
> why? What is the difference in the compiled code?
For font->{width,height} == 32, definitely. IMO, 1 << 31 is undefined as
1 << 31 cannot be represented by an (signed) int.
--
js
suse labs
^ permalink raw reply [flat|nested] 15+ messages in thread
* Re: [PATCH] fbcon: Check font dimension limits
2023-01-26 7:43 ` Greg KH
@ 2023-01-29 15:13 ` Samuel Thibault
-1 siblings, 0 replies; 15+ messages in thread
From: Samuel Thibault @ 2023-01-29 15:13 UTC (permalink / raw)
To: Greg KH
Cc: Daniel Vetter, Helge Deller, linux-fbdev, dri-devel,
linux-kernel, stable, Sanan Hasanov
Greg KH, le jeu. 26 janv. 2023 08:43:02 +0100, a ecrit:
> On Thu, Jan 26, 2023 at 01:49:12AM +0100, Samuel Thibault wrote:
> > blit_x and blit_y are uint32_t, so fbcon currently cannot support fonts
> > larger than 32x32.
>
> "u32" you mean, right?
Right :)
> > The 32x32 case also needs shifting an unsigned int, to properly set bit
> > 31, otherwise we get "UBSAN: shift-out-of-bounds in fbcon_set_font",
> > as reported on
> >
> > http://lore.kernel.org/all/IA1PR07MB98308653E259A6F2CE94A4AFABCE9@IA1PR07MB9830.namprd07.prod.outlook.com
>
> Odd blank line?
Not sure what you mean? I found it easier to read to put a blank line
between the text and the references.
> > Kernel Branch: 6.2.0-rc5-next-20230124
> > Kernel config: https://drive.google.com/file/d/1F-LszDAizEEH0ZX0HcSR06v5q8FPl2Uv/view?usp=sharing
> > Reproducer: https://drive.google.com/file/d/1mP1jcLBY7vWCNM60OMf-ogw-urQRjNrm/view?usp=sharing
>
> What are all of these lines for?
They provide the references that were set in the original report
http://lore.kernel.org/all/IA1PR07MB98308653E259A6F2CE94A4AFABCE9@IA1PR07MB9830.namprd07.prod.outlook.com
Arguably the "branch" and "config" are not that useful since the bug has
been there for more than a dozen years, but notably the "Reproducer" is
useful to provide a userland program that triggers the UB.
> > Reported-by: Sanan Hasanov <sanan.hasanov@Knights.ucf.edu>
> > Signed-off-by: Samuel Thibault <samuel.thibault@ens-lyon.org>
>
> What commit id does this fix?
I though it was there forever, but it seems the check was added in 2007,
so I'll add it in next version.
> Should it go to stable kernels?
Yes. I added stable in Cc of the mail but missed adding it in the text,
I will add it.
> > Index: linux-6.0/drivers/video/fbdev/core/fbcon.c
> > ===================================================================
> > --- linux-6.0.orig/drivers/video/fbdev/core/fbcon.c
> > +++ linux-6.0/drivers/video/fbdev/core/fbcon.c
> > @@ -2489,9 +2489,12 @@ static int fbcon_set_font(struct vc_data
> > h > FBCON_SWAP(info->var.rotate, info->var.yres, info->var.xres))
> > return -EINVAL;
> >
> > + if (font->width > 32 || font->height > 32)
> > + return -EINVAL;
> > +
> > /* Make sure drawing engine can handle the font */
> > - if (!(info->pixmap.blit_x & (1 << (font->width - 1))) ||
> > - !(info->pixmap.blit_y & (1 << (font->height - 1))))
> > + if (!(info->pixmap.blit_x & (1U << (font->width - 1))) ||
> > + !(info->pixmap.blit_y & (1U << (font->height - 1))))
>
> Are you sure this is still needed with the above check added? If so,
> why? What is the difference in the compiled code?
As mentioned by Jiri, yes in the 32 case it's needed otherwise it's UB.
Samuel
^ permalink raw reply [flat|nested] 15+ messages in thread
* Re: [PATCH] fbcon: Check font dimension limits
@ 2023-01-29 15:13 ` Samuel Thibault
0 siblings, 0 replies; 15+ messages in thread
From: Samuel Thibault @ 2023-01-29 15:13 UTC (permalink / raw)
To: Greg KH
Cc: linux-fbdev, Sanan Hasanov, Helge Deller, linux-kernel,
dri-devel, stable
Greg KH, le jeu. 26 janv. 2023 08:43:02 +0100, a ecrit:
> On Thu, Jan 26, 2023 at 01:49:12AM +0100, Samuel Thibault wrote:
> > blit_x and blit_y are uint32_t, so fbcon currently cannot support fonts
> > larger than 32x32.
>
> "u32" you mean, right?
Right :)
> > The 32x32 case also needs shifting an unsigned int, to properly set bit
> > 31, otherwise we get "UBSAN: shift-out-of-bounds in fbcon_set_font",
> > as reported on
> >
> > http://lore.kernel.org/all/IA1PR07MB98308653E259A6F2CE94A4AFABCE9@IA1PR07MB9830.namprd07.prod.outlook.com
>
> Odd blank line?
Not sure what you mean? I found it easier to read to put a blank line
between the text and the references.
> > Kernel Branch: 6.2.0-rc5-next-20230124
> > Kernel config: https://drive.google.com/file/d/1F-LszDAizEEH0ZX0HcSR06v5q8FPl2Uv/view?usp=sharing
> > Reproducer: https://drive.google.com/file/d/1mP1jcLBY7vWCNM60OMf-ogw-urQRjNrm/view?usp=sharing
>
> What are all of these lines for?
They provide the references that were set in the original report
http://lore.kernel.org/all/IA1PR07MB98308653E259A6F2CE94A4AFABCE9@IA1PR07MB9830.namprd07.prod.outlook.com
Arguably the "branch" and "config" are not that useful since the bug has
been there for more than a dozen years, but notably the "Reproducer" is
useful to provide a userland program that triggers the UB.
> > Reported-by: Sanan Hasanov <sanan.hasanov@Knights.ucf.edu>
> > Signed-off-by: Samuel Thibault <samuel.thibault@ens-lyon.org>
>
> What commit id does this fix?
I though it was there forever, but it seems the check was added in 2007,
so I'll add it in next version.
> Should it go to stable kernels?
Yes. I added stable in Cc of the mail but missed adding it in the text,
I will add it.
> > Index: linux-6.0/drivers/video/fbdev/core/fbcon.c
> > ===================================================================
> > --- linux-6.0.orig/drivers/video/fbdev/core/fbcon.c
> > +++ linux-6.0/drivers/video/fbdev/core/fbcon.c
> > @@ -2489,9 +2489,12 @@ static int fbcon_set_font(struct vc_data
> > h > FBCON_SWAP(info->var.rotate, info->var.yres, info->var.xres))
> > return -EINVAL;
> >
> > + if (font->width > 32 || font->height > 32)
> > + return -EINVAL;
> > +
> > /* Make sure drawing engine can handle the font */
> > - if (!(info->pixmap.blit_x & (1 << (font->width - 1))) ||
> > - !(info->pixmap.blit_y & (1 << (font->height - 1))))
> > + if (!(info->pixmap.blit_x & (1U << (font->width - 1))) ||
> > + !(info->pixmap.blit_y & (1U << (font->height - 1))))
>
> Are you sure this is still needed with the above check added? If so,
> why? What is the difference in the compiled code?
As mentioned by Jiri, yes in the 32 case it's needed otherwise it's UB.
Samuel
^ permalink raw reply [flat|nested] 15+ messages in thread
* Re: [PATCH] fbcon: Check font dimension limits
2023-01-26 0:49 ` Samuel Thibault
@ 2023-01-26 9:02 ` Jiri Slaby
-1 siblings, 0 replies; 15+ messages in thread
From: Jiri Slaby @ 2023-01-26 9:02 UTC (permalink / raw)
To: Samuel Thibault, gregkh, Daniel Vetter, Helge Deller
Cc: linux-fbdev, dri-devel, linux-kernel, stable, Sanan Hasanov
On 26. 01. 23, 1:49, Samuel Thibault wrote:
> blit_x and blit_y are uint32_t, so fbcon currently cannot support fonts
> larger than 32x32.
>
> The 32x32 case also needs shifting an unsigned int, to properly set bit
> 31, otherwise we get "UBSAN: shift-out-of-bounds in fbcon_set_font",
> as reported on
>
> http://lore.kernel.org/all/IA1PR07MB98308653E259A6F2CE94A4AFABCE9@IA1PR07MB9830.namprd07.prod.outlook.com
> Kernel Branch: 6.2.0-rc5-next-20230124
> Kernel config: https://drive.google.com/file/d/1F-LszDAizEEH0ZX0HcSR06v5q8FPl2Uv/view?usp=sharing
> Reproducer: https://drive.google.com/file/d/1mP1jcLBY7vWCNM60OMf-ogw-urQRjNrm/view?usp=sharing
>
> Reported-by: Sanan Hasanov <sanan.hasanov@Knights.ucf.edu>
> Signed-off-by: Samuel Thibault <samuel.thibault@ens-lyon.org>
>
> Index: linux-6.0/drivers/video/fbdev/core/fbcon.c
> ===================================================================
> --- linux-6.0.orig/drivers/video/fbdev/core/fbcon.c
> +++ linux-6.0/drivers/video/fbdev/core/fbcon.c
> @@ -2489,9 +2489,12 @@ static int fbcon_set_font(struct vc_data
> h > FBCON_SWAP(info->var.rotate, info->var.yres, info->var.xres))
> return -EINVAL;
>
> + if (font->width > 32 || font->height > 32)
> + return -EINVAL;
> +
> /* Make sure drawing engine can handle the font */
> - if (!(info->pixmap.blit_x & (1 << (font->width - 1))) ||
> - !(info->pixmap.blit_y & (1 << (font->height - 1))))
> + if (!(info->pixmap.blit_x & (1U << (font->width - 1))) ||
> + !(info->pixmap.blit_y & (1U << (font->height - 1))))
So use BIT() properly then? That should be used in all these shifts
anyway. Exactly to avoid UB.
thanks,
--
js
suse labs
^ permalink raw reply [flat|nested] 15+ messages in thread
* Re: [PATCH] fbcon: Check font dimension limits
@ 2023-01-26 9:02 ` Jiri Slaby
0 siblings, 0 replies; 15+ messages in thread
From: Jiri Slaby @ 2023-01-26 9:02 UTC (permalink / raw)
To: Samuel Thibault, gregkh, Daniel Vetter, Helge Deller
Cc: Sanan Hasanov, stable, linux-fbdev, linux-kernel, dri-devel
On 26. 01. 23, 1:49, Samuel Thibault wrote:
> blit_x and blit_y are uint32_t, so fbcon currently cannot support fonts
> larger than 32x32.
>
> The 32x32 case also needs shifting an unsigned int, to properly set bit
> 31, otherwise we get "UBSAN: shift-out-of-bounds in fbcon_set_font",
> as reported on
>
> http://lore.kernel.org/all/IA1PR07MB98308653E259A6F2CE94A4AFABCE9@IA1PR07MB9830.namprd07.prod.outlook.com
> Kernel Branch: 6.2.0-rc5-next-20230124
> Kernel config: https://drive.google.com/file/d/1F-LszDAizEEH0ZX0HcSR06v5q8FPl2Uv/view?usp=sharing
> Reproducer: https://drive.google.com/file/d/1mP1jcLBY7vWCNM60OMf-ogw-urQRjNrm/view?usp=sharing
>
> Reported-by: Sanan Hasanov <sanan.hasanov@Knights.ucf.edu>
> Signed-off-by: Samuel Thibault <samuel.thibault@ens-lyon.org>
>
> Index: linux-6.0/drivers/video/fbdev/core/fbcon.c
> ===================================================================
> --- linux-6.0.orig/drivers/video/fbdev/core/fbcon.c
> +++ linux-6.0/drivers/video/fbdev/core/fbcon.c
> @@ -2489,9 +2489,12 @@ static int fbcon_set_font(struct vc_data
> h > FBCON_SWAP(info->var.rotate, info->var.yres, info->var.xres))
> return -EINVAL;
>
> + if (font->width > 32 || font->height > 32)
> + return -EINVAL;
> +
> /* Make sure drawing engine can handle the font */
> - if (!(info->pixmap.blit_x & (1 << (font->width - 1))) ||
> - !(info->pixmap.blit_y & (1 << (font->height - 1))))
> + if (!(info->pixmap.blit_x & (1U << (font->width - 1))) ||
> + !(info->pixmap.blit_y & (1U << (font->height - 1))))
So use BIT() properly then? That should be used in all these shifts
anyway. Exactly to avoid UB.
thanks,
--
js
suse labs
^ permalink raw reply [flat|nested] 15+ messages in thread
* Re: [PATCH] fbcon: Check font dimension limits
2023-01-26 9:02 ` Jiri Slaby
@ 2023-01-29 15:04 ` Samuel Thibault
-1 siblings, 0 replies; 15+ messages in thread
From: Samuel Thibault @ 2023-01-29 15:04 UTC (permalink / raw)
To: Jiri Slaby
Cc: gregkh, Daniel Vetter, Helge Deller, linux-fbdev, dri-devel,
linux-kernel, stable, Sanan Hasanov
Jiri Slaby, le jeu. 26 janv. 2023 10:02:55 +0100, a ecrit:
> On 26. 01. 23, 1:49, Samuel Thibault wrote:
> > Index: linux-6.0/drivers/video/fbdev/core/fbcon.c
> > ===================================================================
> > --- linux-6.0.orig/drivers/video/fbdev/core/fbcon.c
> > +++ linux-6.0/drivers/video/fbdev/core/fbcon.c
> > @@ -2489,9 +2489,12 @@ static int fbcon_set_font(struct vc_data
> > h > FBCON_SWAP(info->var.rotate, info->var.yres, info->var.xres))
> > return -EINVAL;
> > + if (font->width > 32 || font->height > 32)
> > + return -EINVAL;
> > +
> > /* Make sure drawing engine can handle the font */
> > - if (!(info->pixmap.blit_x & (1 << (font->width - 1))) ||
> > - !(info->pixmap.blit_y & (1 << (font->height - 1))))
> > + if (!(info->pixmap.blit_x & (1U << (font->width - 1))) ||
> > + !(info->pixmap.blit_y & (1U << (font->height - 1))))
>
> So use BIT() properly then? That should be used in all these shifts anyway.
> Exactly to avoid UB.
Right, I'll use that in next version.
Samuel
^ permalink raw reply [flat|nested] 15+ messages in thread
* Re: [PATCH] fbcon: Check font dimension limits
@ 2023-01-29 15:04 ` Samuel Thibault
0 siblings, 0 replies; 15+ messages in thread
From: Samuel Thibault @ 2023-01-29 15:04 UTC (permalink / raw)
To: Jiri Slaby
Cc: linux-fbdev, Sanan Hasanov, gregkh, Helge Deller, linux-kernel,
dri-devel, stable
Jiri Slaby, le jeu. 26 janv. 2023 10:02:55 +0100, a ecrit:
> On 26. 01. 23, 1:49, Samuel Thibault wrote:
> > Index: linux-6.0/drivers/video/fbdev/core/fbcon.c
> > ===================================================================
> > --- linux-6.0.orig/drivers/video/fbdev/core/fbcon.c
> > +++ linux-6.0/drivers/video/fbdev/core/fbcon.c
> > @@ -2489,9 +2489,12 @@ static int fbcon_set_font(struct vc_data
> > h > FBCON_SWAP(info->var.rotate, info->var.yres, info->var.xres))
> > return -EINVAL;
> > + if (font->width > 32 || font->height > 32)
> > + return -EINVAL;
> > +
> > /* Make sure drawing engine can handle the font */
> > - if (!(info->pixmap.blit_x & (1 << (font->width - 1))) ||
> > - !(info->pixmap.blit_y & (1 << (font->height - 1))))
> > + if (!(info->pixmap.blit_x & (1U << (font->width - 1))) ||
> > + !(info->pixmap.blit_y & (1U << (font->height - 1))))
>
> So use BIT() properly then? That should be used in all these shifts anyway.
> Exactly to avoid UB.
Right, I'll use that in next version.
Samuel
^ permalink raw reply [flat|nested] 15+ messages in thread
* Re: [PATCH] fbcon: Check font dimension limits
2023-01-26 0:49 ` Samuel Thibault
` (3 preceding siblings ...)
(?)
@ 2023-01-28 1:32 ` kernel test robot
-1 siblings, 0 replies; 15+ messages in thread
From: kernel test robot @ 2023-01-28 1:32 UTC (permalink / raw)
To: Samuel Thibault; +Cc: stable, oe-kbuild-all
Hi,
Thanks for your patch.
FYI: kernel test robot notices the stable kernel rule is not satisfied.
Rule: 'Cc: stable@vger.kernel.org' or 'commit <sha1> upstream.'
Subject: [PATCH] fbcon: Check font dimension limits
Link: https://lore.kernel.org/stable/20230126004921.616264824%40ens-lyon.org
The check is based on https://www.kernel.org/doc/html/latest/process/stable-kernel-rules.html
--
0-DAY CI Kernel Test Service
https://github.com/intel/lkp-tests
^ permalink raw reply [flat|nested] 15+ messages in thread