[PATCH v2] fbcon: Set fb_display[i]->mode to NULL when the mode is released

[PATCH v2] fbcon: Set fb_display[i]->mode to NULL when the mode is released

[Quanmin Yan]

Recently, we discovered the following issue through syzkaller:

BUG: KASAN: slab-use-after-free in fb_mode_is_equal+0x285/0x2f0
Read of size 4 at addr ff11000001b3c69c by task syz.xxx
...
Call Trace:
 <TASK>
 dump_stack_lvl+0xab/0xe0
 print_address_description.constprop.0+0x2c/0x390
 print_report+0xb9/0x280
 kasan_report+0xb8/0xf0
 fb_mode_is_equal+0x285/0x2f0
 fbcon_mode_deleted+0x129/0x180
 fb_set_var+0xe7f/0x11d0
 do_fb_ioctl+0x6a0/0x750
 fb_ioctl+0xe0/0x140
 __x64_sys_ioctl+0x193/0x210
 do_syscall_64+0x5f/0x9c0
 entry_SYSCALL_64_after_hwframe+0x76/0x7e

Based on experimentation and analysis, during framebuffer unregistration,
only the memory of fb_info->modelist is freed, without setting the
corresponding fb_display[i]->mode to NULL for the freed modes. This leads
to UAF issues during subsequent accesses. Here's an example of reproduction
steps:
1. With /dev/fb0 already registered in the system, load a kernel module
   to register a new device /dev/fb1;
2. Set fb1's mode to the global fb_display[] array (via FBIOPUT_CON2FBMAP);
3. Switch console from fb to VGA (to allow normal rmmod of the ko);
4. Unload the kernel module, at this point fb1's modelist is freed, leaving
   a wild pointer in fb_display[];
5. Trigger the bug via system calls through fb0 attempting to delete a mode
   from fb0.

Add a check in do_unregister_framebuffer(): if the mode to be freed exists
in fb_display[], set the corresponding mode pointer to NULL.

Signed-off-by: Quanmin Yan <yanquanmin1@huawei.com>
---

Changes from v1
(https://lore.kernel.org/all/20250923110608.3385083-1-yanquanmin1@huawei.com/)
- Focus on fixing the issue specifically in the framebuffer unregistration
  path, as other paths do not encounter this problem.
- Adjusted according to Thomas's suggestions.

 drivers/video/fbdev/core/fbcon.c | 19 +++++++++++++++++++
 drivers/video/fbdev/core/fbmem.c |  1 +
 include/linux/fbcon.h            |  2 ++
 3 files changed, 22 insertions(+)

diff --git a/drivers/video/fbdev/core/fbcon.c b/drivers/video/fbdev/core/fbcon.c
index 96cc9b389246..9bd3c3814b5c 100644
--- a/drivers/video/fbdev/core/fbcon.c
+++ b/drivers/video/fbdev/core/fbcon.c
@@ -2810,6 +2810,25 @@ int fbcon_mode_deleted(struct fb_info *info,
 	return found;
 }
 
+static void fbcon_delete_mode(struct fb_videomode *m)
+{
+	struct fbcon_display *p;
+
+	for (int i = first_fb_vc; i <= last_fb_vc; i++) {
+		p = &fb_display[i];
+		if (p->mode == m)
+			p->mode = NULL;
+	}
+}
+
+void fbcon_delete_modelist(struct list_head *head)
+{
+	struct fb_modelist *modelist;
+
+	list_for_each_entry(modelist, head, list)
+		fbcon_delete_mode(&modelist->mode);
+}
+
 #ifdef CONFIG_VT_HW_CONSOLE_BINDING
 static void fbcon_unbind(void)
 {
diff --git a/drivers/video/fbdev/core/fbmem.c b/drivers/video/fbdev/core/fbmem.c
index 53f1719b1ae1..eff757ebbed1 100644
--- a/drivers/video/fbdev/core/fbmem.c
+++ b/drivers/video/fbdev/core/fbmem.c
@@ -544,6 +544,7 @@ static void do_unregister_framebuffer(struct fb_info *fb_info)
 		fb_info->pixmap.addr = NULL;
 	}
 
+	fbcon_delete_modelist(&fb_info->modelist);
 	fb_destroy_modelist(&fb_info->modelist);
 	registered_fb[fb_info->node] = NULL;
 	num_registered_fb--;
diff --git a/include/linux/fbcon.h b/include/linux/fbcon.h
index 81f0e698acbf..f206370060e1 100644
--- a/include/linux/fbcon.h
+++ b/include/linux/fbcon.h
@@ -18,6 +18,7 @@ void fbcon_suspended(struct fb_info *info);
 void fbcon_resumed(struct fb_info *info);
 int fbcon_mode_deleted(struct fb_info *info,
 		       struct fb_videomode *mode);
+void fbcon_delete_modelist(struct list_head *head);
 void fbcon_new_modelist(struct fb_info *info);
 void fbcon_get_requirement(struct fb_info *info,
 			   struct fb_blit_caps *caps);
@@ -38,6 +39,7 @@ static inline void fbcon_suspended(struct fb_info *info) {}
 static inline void fbcon_resumed(struct fb_info *info) {}
 static inline int fbcon_mode_deleted(struct fb_info *info,
 				     struct fb_videomode *mode) { return 0; }
+static inline void fbcon_delete_modelist(struct list_head *head) {}
 static inline void fbcon_new_modelist(struct fb_info *info) {}
 static inline void fbcon_get_requirement(struct fb_info *info,
 					 struct fb_blit_caps *caps) {}
-- 
2.43.0

Re: [PATCH v2] fbcon: Set fb_display[i]->mode to NULL when the mode is released

[Thomas Zimmermann]

Am 10.10.25 um 10:16 schrieb Quanmin Yan:
> Recently, we discovered the following issue through syzkaller:
>
> BUG: KASAN: slab-use-after-free in fb_mode_is_equal+0x285/0x2f0
> Read of size 4 at addr ff11000001b3c69c by task syz.xxx
> ...
> Call Trace:
>   <TASK>
>   dump_stack_lvl+0xab/0xe0
>   print_address_description.constprop.0+0x2c/0x390
>   print_report+0xb9/0x280
>   kasan_report+0xb8/0xf0
>   fb_mode_is_equal+0x285/0x2f0
>   fbcon_mode_deleted+0x129/0x180
>   fb_set_var+0xe7f/0x11d0
>   do_fb_ioctl+0x6a0/0x750
>   fb_ioctl+0xe0/0x140
>   __x64_sys_ioctl+0x193/0x210
>   do_syscall_64+0x5f/0x9c0
>   entry_SYSCALL_64_after_hwframe+0x76/0x7e
>
> Based on experimentation and analysis, during framebuffer unregistration,
> only the memory of fb_info->modelist is freed, without setting the
> corresponding fb_display[i]->mode to NULL for the freed modes. This leads
> to UAF issues during subsequent accesses. Here's an example of reproduction
> steps:
> 1. With /dev/fb0 already registered in the system, load a kernel module
>     to register a new device /dev/fb1;
> 2. Set fb1's mode to the global fb_display[] array (via FBIOPUT_CON2FBMAP);
> 3. Switch console from fb to VGA (to allow normal rmmod of the ko);
> 4. Unload the kernel module, at this point fb1's modelist is freed, leaving
>     a wild pointer in fb_display[];
> 5. Trigger the bug via system calls through fb0 attempting to delete a mode
>     from fb0.
>
> Add a check in do_unregister_framebuffer(): if the mode to be freed exists
> in fb_display[], set the corresponding mode pointer to NULL.
>
> Signed-off-by: Quanmin Yan <yanquanmin1@huawei.com>

Reviewed-by: Thomas Zimmermann <tzimmermann@suse.de>

Thanks for fixing the bug.

> ---
>
> Changes from v1
> (https://lore.kernel.org/all/20250923110608.3385083-1-yanquanmin1@huawei.com/)
> - Focus on fixing the issue specifically in the framebuffer unregistration
>    path, as other paths do not encounter this problem.
> - Adjusted according to Thomas's suggestions.
>
>   drivers/video/fbdev/core/fbcon.c | 19 +++++++++++++++++++
>   drivers/video/fbdev/core/fbmem.c |  1 +
>   include/linux/fbcon.h            |  2 ++
>   3 files changed, 22 insertions(+)
>
> diff --git a/drivers/video/fbdev/core/fbcon.c b/drivers/video/fbdev/core/fbcon.c
> index 96cc9b389246..9bd3c3814b5c 100644
> --- a/drivers/video/fbdev/core/fbcon.c
> +++ b/drivers/video/fbdev/core/fbcon.c
> @@ -2810,6 +2810,25 @@ int fbcon_mode_deleted(struct fb_info *info,
>   	return found;
>   }
>   
> +static void fbcon_delete_mode(struct fb_videomode *m)
> +{
> +	struct fbcon_display *p;
> +
> +	for (int i = first_fb_vc; i <= last_fb_vc; i++) {
> +		p = &fb_display[i];
> +		if (p->mode == m)
> +			p->mode = NULL;
> +	}
> +}
> +
> +void fbcon_delete_modelist(struct list_head *head)
> +{
> +	struct fb_modelist *modelist;
> +
> +	list_for_each_entry(modelist, head, list)
> +		fbcon_delete_mode(&modelist->mode);
> +}
> +
>   #ifdef CONFIG_VT_HW_CONSOLE_BINDING
>   static void fbcon_unbind(void)
>   {
> diff --git a/drivers/video/fbdev/core/fbmem.c b/drivers/video/fbdev/core/fbmem.c
> index 53f1719b1ae1..eff757ebbed1 100644
> --- a/drivers/video/fbdev/core/fbmem.c
> +++ b/drivers/video/fbdev/core/fbmem.c
> @@ -544,6 +544,7 @@ static void do_unregister_framebuffer(struct fb_info *fb_info)
>   		fb_info->pixmap.addr = NULL;
>   	}
>   
> +	fbcon_delete_modelist(&fb_info->modelist);
>   	fb_destroy_modelist(&fb_info->modelist);
>   	registered_fb[fb_info->node] = NULL;
>   	num_registered_fb--;
> diff --git a/include/linux/fbcon.h b/include/linux/fbcon.h
> index 81f0e698acbf..f206370060e1 100644
> --- a/include/linux/fbcon.h
> +++ b/include/linux/fbcon.h
> @@ -18,6 +18,7 @@ void fbcon_suspended(struct fb_info *info);
>   void fbcon_resumed(struct fb_info *info);
>   int fbcon_mode_deleted(struct fb_info *info,
>   		       struct fb_videomode *mode);
> +void fbcon_delete_modelist(struct list_head *head);
>   void fbcon_new_modelist(struct fb_info *info);
>   void fbcon_get_requirement(struct fb_info *info,
>   			   struct fb_blit_caps *caps);
> @@ -38,6 +39,7 @@ static inline void fbcon_suspended(struct fb_info *info) {}
>   static inline void fbcon_resumed(struct fb_info *info) {}
>   static inline int fbcon_mode_deleted(struct fb_info *info,
>   				     struct fb_videomode *mode) { return 0; }
> +static inline void fbcon_delete_modelist(struct list_head *head) {}
>   static inline void fbcon_new_modelist(struct fb_info *info) {}
>   static inline void fbcon_get_requirement(struct fb_info *info,
>   					 struct fb_blit_caps *caps) {}

-- 
--
Thomas Zimmermann
Graphics Driver Developer
SUSE Software Solutions Germany GmbH
Frankenstrasse 146, 90461 Nuernberg, Germany
GF: Ivo Totev, Andrew Myers, Andrew McDonald, Boudien Moerman
HRB 36809 (AG Nuernberg)

Re: [PATCH v2] fbcon: Set fb_display[i]->mode to NULL when the mode is released

[Helge Deller]

On 10/10/25 10:16, Quanmin Yan wrote:
> Recently, we discovered the following issue through syzkaller:
> 
> BUG: KASAN: slab-use-after-free in fb_mode_is_equal+0x285/0x2f0
> Read of size 4 at addr ff11000001b3c69c by task syz.xxx
> ...
> Call Trace:
>   <TASK>
>   dump_stack_lvl+0xab/0xe0
>   print_address_description.constprop.0+0x2c/0x390
>   print_report+0xb9/0x280
>   kasan_report+0xb8/0xf0
>   fb_mode_is_equal+0x285/0x2f0
>   fbcon_mode_deleted+0x129/0x180
>   fb_set_var+0xe7f/0x11d0
>   do_fb_ioctl+0x6a0/0x750
>   fb_ioctl+0xe0/0x140
>   __x64_sys_ioctl+0x193/0x210
>   do_syscall_64+0x5f/0x9c0
>   entry_SYSCALL_64_after_hwframe+0x76/0x7e
> 
> Based on experimentation and analysis, during framebuffer unregistration,
> only the memory of fb_info->modelist is freed, without setting the
> corresponding fb_display[i]->mode to NULL for the freed modes. This leads
> to UAF issues during subsequent accesses. Here's an example of reproduction
> steps:
> 1. With /dev/fb0 already registered in the system, load a kernel module
>     to register a new device /dev/fb1;
> 2. Set fb1's mode to the global fb_display[] array (via FBIOPUT_CON2FBMAP);
> 3. Switch console from fb to VGA (to allow normal rmmod of the ko);
> 4. Unload the kernel module, at this point fb1's modelist is freed, leaving
>     a wild pointer in fb_display[];
> 5. Trigger the bug via system calls through fb0 attempting to delete a mode
>     from fb0.
> 
> Add a check in do_unregister_framebuffer(): if the mode to be freed exists
> in fb_display[], set the corresponding mode pointer to NULL.
> 
> Signed-off-by: Quanmin Yan <yanquanmin1@huawei.com>
> --- 
>   drivers/video/fbdev/core/fbcon.c | 19 +++++++++++++++++++
>   drivers/video/fbdev/core/fbmem.c |  1 +
>   include/linux/fbcon.h            |  2 ++
>   3 files changed, 22 insertions(+)

applied.

Thanks!
Helge