[PATCH] fbdev: udlfb: don't rebuild the mode list while the framebuffer is in use

[PATCH] fbdev: udlfb: don't rebuild the mode list while the framebuffer is in use

[Xiang Mei]

dlfb_setup_modes() frees info->modelist with fb_destroy_modelist() and
rebuilds it, but fbcon holds raw pointers into that list in
fb_display[i].mode and nothing un-publishes them the way
do_unregister_framebuffer() and store_modes() do with
fbcon_delete_modelist(). The function does refuse to install a new mode
while the framebuffer has users (dlfb->fb_count != 0), but only after the
list is already destroyed. Reached from the 0666 "edid" sysfs attribute,
an unprivileged write to /sys/class/graphics/fb0/edid therefore fails with
-EINVAL and still leaves every fb_display[i].mode dangling; the next
console switch reads the freed fb_videomode in fb_videomode_to_var().

Test fb_count before touching the list. fbcon takes a reference through
fbcon_open(), so that condition covers exactly the states in which
fb_display[] points into the list, and the errno returned to userspace is
unchanged.

  BUG: KASAN: slab-use-after-free in fb_videomode_to_var (drivers/video/fbdev/core/modedb.c:905)
  Read of size 4 at addr ffff8880107b449c by task kworker/1:1/47
  Workqueue: events console_callback
  Call Trace:
   fb_videomode_to_var (drivers/video/fbdev/core/modedb.c:905)
   display_to_var (drivers/video/fbdev/core/fbcon.c:998)
   fbcon_switch (drivers/video/fbdev/core/fbcon.c:2182)
   redraw_screen (drivers/tty/vt/vt.c:994)
   complete_change_console (drivers/tty/vt/vt_ioctl.c:1141)
   console_callback (drivers/tty/vt/vt.c:3358)
   process_one_work (kernel/workqueue.c:3396)
   worker_thread (kernel/workqueue.c:3479 kernel/workqueue.c:3560)
   kthread (kernel/kthread.c:436)
   ret_from_fork (arch/x86/kernel/process.c:158)
   ret_from_fork_asm (arch/x86/entry/entry_64.S:245)
  ...
  The buggy address is located 28 bytes inside of
   freed 96-byte region [ffff8880107b4480, ffff8880107b44e0)

Cc: stable@vger.kernel.org
Fixes: 7d9485e2c53c ("Staging: udlfb: Add functions to expose sysfs metrics and controls")
Reported-by: co+ed9902f2e44184d9@bugs.sh
Closes: https://lore.kernel.org/all/1jSCSNaDKmuUG7h40rTsSl1rMaSwdlJef4rp%40bugs.sh/
Assisted-by: Claude:claude-opus-5
Signed-off-by: Xiang Mei <xmei5@asu.edu>
---
 drivers/video/fbdev/udlfb.c | 5 +++++
 1 file changed, 5 insertions(+)

diff --git a/drivers/video/fbdev/udlfb.c b/drivers/video/fbdev/udlfb.c
index e78d6f95c9c5..92b352bc7e96 100644
--- a/drivers/video/fbdev/udlfb.c
+++ b/drivers/video/fbdev/udlfb.c
@@ -1293,6 +1293,11 @@ static int dlfb_setup_modes(struct dlfb_data *dlfb,
 		goto error;
 	}
 
+	if (dlfb->fb_count) {
+		result = -EINVAL;
+		goto error;
+	}
+
 	fb_destroy_modelist(&info->modelist);
 	memset(&info->monspecs, 0, sizeof(info->monspecs));
 
-- 
2.43.0

Re: [PATCH] fbdev: udlfb: don't rebuild the mode list while the framebuffer is in use

[sashiko-bot]

Thank you for your contribution! Sashiko AI review found 5 potential issue(s) to consider:

Pre-existing issues:
- [High] Missing locking in edid_store() leads to data races, double free, and memory corruption.
- [High] The patch fails to prevent a Use-After-Free of `fb_info->mode` via sysfs attributes.
- [High] The `edid` sysfs attribute is world-writable (0666), enabling unprivileged users to exploit multiple driver vulnerabilities.
- [Medium] Repeated calls to `dlfb_setup_modes()` leak the old `dlfb->edid` buffer.
- [Medium] Unbounded memory leak of the dynamically allocated `info->monspecs.modedb` array.
--

commit c422e1f65727b9cb9b5d809088416cf97c54aea9
Author: Xiang Mei <xmei5@asu.edu>

fbdev: udlfb: don't rebuild the mode list while the framebuffer is in use

This commit attempts to fix a use-after-free in fb_videomode_to_var() by
checking dlfb->fb_count before destroying the modelist in
dlfb_setup_modes(). This prevents fbcon from referencing a freed mode
list.

Link: https://lore.kernel.org/all/1jSCSNaDKmuUG7h40rTsSl1rMaSwdlJef4rp%40bugs.sh/

> diff --git a/drivers/video/fbdev/udlfb.c b/drivers/video/fbdev/udlfb.c
> index e78d6f95c9c5c..92b352bc7e969 100644
> --- a/drivers/video/fbdev/udlfb.c
> +++ b/drivers/video/fbdev/udlfb.c
> @@ -1293,6 +1293,11 @@ static int dlfb_setup_modes(struct dlfb_data *dlfb,
>  		goto error;
>  	}
>  
> +	if (dlfb->fb_count) {
> +		result = -EINVAL;
> +		goto error;
> +	}
> +
>  	fb_destroy_modelist(&info->modelist);

[Severity: High]
This isn't a bug introduced by this patch, but does writing to the mode sysfs
attribute increment dlfb->fb_count?

If an unprivileged user sets fb_info->mode to a node in info->modelist via
fbsysfs.c:store_mode(), and then writes to the edid sysfs attribute,
dlfb->fb_count remains 0. The check above passes, and fb_destroy_modelist()
is executed, leaving fb_info->mode as a dangling pointer.

Subsequent reads from the mode sysfs attribute would invoke show_mode(),
triggering a use-after-free.

>  	memset(&info->monspecs, 0, sizeof(info->monspecs));

[Severity: Medium]
This is a pre-existing issue, but does this memset() leak the dynamically
allocated info->monspecs.modedb array?

If fb_edid_to_monspecs() previously parsed the EDID and dynamically allocated
the array of struct fb_videomode, unconditionally zeroing the structure here
appears to erase the pointer before it can be freed, leaking the memory on
every sysfs write.

[Severity: Medium]
While not introduced by this patch, could this leak the old dlfb->edid
buffer?

When dlfb_setup_modes() succeeds:

drivers/video/fbdev/udlfb.c:dlfb_setup_modes() {
    ...
		if (info->monspecs.modedb_len > 0) {
			dlfb->edid = edid;
			dlfb->edid_size = i;
			break;
		}
    ...
}

Does this overwrite dlfb->edid with a newly allocated 128-byte edid buffer
without freeing the previous pointer?


[Severity: High]
This is a pre-existing issue, but is it safe for edid_store() to call
dlfb_ops_set_par() without holding info->lock?

drivers/video/fbdev/udlfb.c:edid_store() {
    ...
	ret = dlfb_ops_set_par(fb_info);
    ...
}

dlfb_ops_set_par() invokes dlfb_realloc_framebuffer() which performs memory
freeing and list_add() operations, explicitly relying on the lock:

drivers/video/fbdev/udlfb.c:dlfb_realloc_framebuffer() {
/*
 * Assumes &info->lock held by caller
 */
static int dlfb_realloc_framebuffer(struct dlfb_data *dlfb,
                                    struct fb_info *info, u32 new_len)
    ...
}

If an unprivileged user concurrently writes to the edid sysfs attribute,
could this lead to data races resulting in a double-free of
info->screen_buffer and list corruption on dlfb->deferred_free?


[Severity: High]
Finally, as a pre-existing issue, should the edid sysfs attribute be
world-writable?

drivers/video/fbdev/udlfb.c:edid_attr {
static const struct bin_attribute edid_attr = {
	.attr.name = "edid",
	.attr.mode = 0666,
	.size = EDID_LENGTH,
	.read = edid_show,
	.write = edid_store
};
}

This 0666 permission allows any unprivileged user to alter the physical
display configuration, and easily trigger the memory leaks, double frees,
and use-after-free vulnerabilities mentioned above.

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260912211512.2049147-1-xmei5@asu.edu?part=1

Re: [PATCH] fbdev: udlfb: don't rebuild the mode list while the framebuffer is in use

[Helge Deller]

On 9/12/26 23:15, Xiang Mei wrote:
> dlfb_setup_modes() frees info->modelist with fb_destroy_modelist() and
> rebuilds it, but fbcon holds raw pointers into that list in
> fb_display[i].mode and nothing un-publishes them the way
> do_unregister_framebuffer() and store_modes() do with
> fbcon_delete_modelist(). The function does refuse to install a new mode
> while the framebuffer has users (dlfb->fb_count != 0), but only after the
> list is already destroyed. Reached from the 0666 "edid" sysfs attribute,
> an unprivileged write to /sys/class/graphics/fb0/edid therefore fails with
> -EINVAL and still leaves every fb_display[i].mode dangling; the next
> console switch reads the freed fb_videomode in fb_videomode_to_var().
> 
> Test fb_count before touching the list. fbcon takes a reference through
> fbcon_open(), so that condition covers exactly the states in which
> fb_display[] points into the list, and the errno returned to userspace is
> unchanged.
> 
>    BUG: KASAN: slab-use-after-free in fb_videomode_to_var (drivers/video/fbdev/core/modedb.c:905)
>    Read of size 4 at addr ffff8880107b449c by task kworker/1:1/47
>    Workqueue: events console_callback
>    Call Trace:
>     fb_videomode_to_var (drivers/video/fbdev/core/modedb.c:905)
>     display_to_var (drivers/video/fbdev/core/fbcon.c:998)
>     fbcon_switch (drivers/video/fbdev/core/fbcon.c:2182)
>     redraw_screen (drivers/tty/vt/vt.c:994)
>     complete_change_console (drivers/tty/vt/vt_ioctl.c:1141)
>     console_callback (drivers/tty/vt/vt.c:3358)
>     process_one_work (kernel/workqueue.c:3396)
>     worker_thread (kernel/workqueue.c:3479 kernel/workqueue.c:3560)
>     kthread (kernel/kthread.c:436)
>     ret_from_fork (arch/x86/kernel/process.c:158)
>     ret_from_fork_asm (arch/x86/entry/entry_64.S:245)
>    ...
>    The buggy address is located 28 bytes inside of
>     freed 96-byte region [ffff8880107b4480, ffff8880107b44e0)
> 
> Cc: stable@vger.kernel.org
> Fixes: 7d9485e2c53c ("Staging: udlfb: Add functions to expose sysfs metrics and controls")
> Reported-by: co+ed9902f2e44184d9@bugs.sh
> Closes: https://lore.kernel.org/all/1jSCSNaDKmuUG7h40rTsSl1rMaSwdlJef4rp%40bugs.sh/
> Assisted-by: Claude:claude-opus-5
> Signed-off-by: Xiang Mei <xmei5@asu.edu>
> ---
>   drivers/video/fbdev/udlfb.c | 5 +++++
>   1 file changed, 5 insertions(+)

Looks good.
Applied.

Thanks!
Helge