diff mbox

video/logo: don't look for the logo after system boot

Message ID 1387549460-18501-1-git-send-email-bigeasy@linutronix.de (mailing list archive)
State New, archived
Headers show

Commit Message

Sebastian Andrzej Siewior Dec. 20, 2013, 2:24 p.m. UTC
If the primary GPU driver has been loaded _after_ system as a module
then this logo memory is no longer valid.
Managed to crash the system by booting a box without a GPU and then
hotpluggin => BOOM.

Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
---
 drivers/video/logo/logo.c | 3 +++
 1 file changed, 3 insertions(+)

Comments

Geert Uytterhoeven Dec. 20, 2013, 3:27 p.m. UTC | #1
On Fri, Dec 20, 2013 at 3:24 PM, Sebastian Andrzej Siewior
<bigeasy@linutronix.de> wrote:
> If the primary GPU driver has been loaded _after_ system as a module
> then this logo memory is no longer valid.
> Managed to crash the system by booting a box without a GPU and then
> hotpluggin => BOOM.

Which GPU driver is this?

drivers/video/fbmem.c:fb_prepare_logo() has protection against this:

        if (info->flags & FBINFO_MISC_TILEBLITTING ||
            info->flags & FBINFO_MODULE)
                return 0;

> Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
> ---
>  drivers/video/logo/logo.c | 3 +++
>  1 file changed, 3 insertions(+)
>
> diff --git a/drivers/video/logo/logo.c b/drivers/video/logo/logo.c
> index 080c35b..a26bb16 100644
> --- a/drivers/video/logo/logo.c
> +++ b/drivers/video/logo/logo.c
> @@ -36,6 +36,9 @@ const struct linux_logo * __init_refok fb_find_logo(int depth)
>         if (nologo)
>                 return NULL;
>
> +       if (system_state != SYSTEM_BOOTING)
> +               return NULL;
> +
>         if (depth >= 1) {
>  #ifdef CONFIG_LOGO_LINUX_MONO
>                 /* Generic Linux logo */

Gr{oetje,eeting}s,

                        Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
                                -- Linus Torvalds
--
To unsubscribe from this list: send the line "unsubscribe linux-fbdev" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Sebastian Andrzej Siewior Dec. 20, 2013, 3:31 p.m. UTC | #2
On 12/20/2013 04:27 PM, Geert Uytterhoeven wrote:
> On Fri, Dec 20, 2013 at 3:24 PM, Sebastian Andrzej Siewior
> <bigeasy@linutronix.de> wrote:
>> If the primary GPU driver has been loaded _after_ system as a module
>> then this logo memory is no longer valid.
>> Managed to crash the system by booting a box without a GPU and then
>> hotpluggin => BOOM.
> 
> Which GPU driver is this?

i915.

> 
> drivers/video/fbmem.c:fb_prepare_logo() has protection against this:
> 
>         if (info->flags & FBINFO_MISC_TILEBLITTING ||
>             info->flags & FBINFO_MODULE)
>                 return 0;

but gpu driver is built-in. I just add PCI device at run-time. The same
thing should happen if you go to sysfs and remove the PCI device and
then do
  echo 1 > /sys/bus/pci/rescan

Sebastian

--
To unsubscribe from this list: send the line "unsubscribe linux-fbdev" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Sebastian Andrzej Siewior Dec. 20, 2013, 3:49 p.m. UTC | #3
On 12/20/2013 04:31 PM, Sebastian Andrzej Siewior wrote:
> On 12/20/2013 04:27 PM, Geert Uytterhoeven wrote:
>> On Fri, Dec 20, 2013 at 3:24 PM, Sebastian Andrzej Siewior
>> <bigeasy@linutronix.de> wrote:
>>> If the primary GPU driver has been loaded _after_ system as a module
>>> then this logo memory is no longer valid.
>>> Managed to crash the system by booting a box without a GPU and then
>>> hotpluggin => BOOM.
>>
>> drivers/video/fbmem.c:fb_prepare_logo() has protection against this:
>>
>>         if (info->flags & FBINFO_MISC_TILEBLITTING ||
>>             info->flags & FBINFO_MODULE)
>>                 return 0;
> 
> but gpu driver is built-in. I just add PCI device at run-time. The same
> thing should happen if you go to sysfs and remove the PCI device and
> then do

No I see where you are going with this. My description of the problem
is wrong. The problem is nothing to do with the driver being a module.
If nobody objects this or suggests a different solution then I'm going
to provide a patch with a proper description.

Sebastian
--
To unsubscribe from this list: send the line "unsubscribe linux-fbdev" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Geert Uytterhoeven Dec. 20, 2013, 4:20 p.m. UTC | #4
On Fri, Dec 20, 2013 at 4:49 PM, Sebastian Andrzej Siewior
<bigeasy@linutronix.de> wrote:
> On 12/20/2013 04:31 PM, Sebastian Andrzej Siewior wrote:
>> On 12/20/2013 04:27 PM, Geert Uytterhoeven wrote:
>>> On Fri, Dec 20, 2013 at 3:24 PM, Sebastian Andrzej Siewior
>>> <bigeasy@linutronix.de> wrote:
>>>> If the primary GPU driver has been loaded _after_ system as a module
>>>> then this logo memory is no longer valid.
>>>> Managed to crash the system by booting a box without a GPU and then
>>>> hotpluggin => BOOM.
>>>
>>> drivers/video/fbmem.c:fb_prepare_logo() has protection against this:
>>>
>>>         if (info->flags & FBINFO_MISC_TILEBLITTING ||
>>>             info->flags & FBINFO_MODULE)
>>>                 return 0;
>>
>> but gpu driver is built-in. I just add PCI device at run-time. The same
>> thing should happen if you go to sysfs and remove the PCI device and
>> then do
>
> No I see where you are going with this. My description of the problem
> is wrong. The problem is nothing to do with the driver being a module.

Yes, I got confused by the "as a module".

> If nobody objects this or suggests a different solution then I'm going
> to provide a patch with a proper description.

Is "system_state != SYSTEM_BOOTING" the right check?
How long is is in the SYSTEM_BOOTING state?

Ah, kernel_init() does:

        free_initmem();
        mark_rodata_ro();
        system_state = SYSTEM_RUNNING;

So this looks OK to me.

We used to have an initmem_freed flag, but it was removed 15 years
ago. 11 years ago, we got system_state, for the same purpose ;-)

Looking at commit 70802c60379fb843c485dfd4cab9e8f527d8fe81
Author: Antonino A. Daplas <adaplas@gmail.com>
Date:   Tue May 8 00:38:14 2007 -0700

    fbdev: don't show logo if driver or fbcon are modular

it seems we were not aware of the existence of system_state....
I guess you can remove FBINFO_MODULE after your fix?

BTW, I'm wondering if modprobing newport_con.ko crashes, too, as
it seems to call fb_find_logo() without any protection.

Gr{oetje,eeting}s,

                        Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
                                -- Linus Torvalds
--
To unsubscribe from this list: send the line "unsubscribe linux-fbdev" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
diff mbox

Patch

diff --git a/drivers/video/logo/logo.c b/drivers/video/logo/logo.c
index 080c35b..a26bb16 100644
--- a/drivers/video/logo/logo.c
+++ b/drivers/video/logo/logo.c
@@ -36,6 +36,9 @@  const struct linux_logo * __init_refok fb_find_logo(int depth)
 	if (nologo)
 		return NULL;
 
+	if (system_state != SYSTEM_BOOTING)
+		return NULL;
+
 	if (depth >= 1) {
 #ifdef CONFIG_LOGO_LINUX_MONO
 		/* Generic Linux logo */