[RFC] Suspicious bug in module refcounting

On Wednesday 04 February 2009 20:41:30 Russell King wrote:
> On Wed, Feb 04, 2009 at 02:18:08PM +1030, Rusty Russell wrote:
> >   gameport.c, serio.c and input.c increment their own refcount, but to get
> > into those init functions someone must be holding a refcount already (ie. a
> > module depends on this module).  Ditto cyber2000fb.c, and MTD.
> 
> Err, wrong.  cyber2000fb.c does it in its module initialization function
> to prevent the module (when built for Shark) from being unloaded.  It
> does this because it's from the days of 2.2 kernels and no one bothered
> writing the module unload support for Shark.  I'm certainly not in a
> position to do that.

Thanks, here's the patch then:

Subject: cyber2000fb.c: use proper method for stopping unload if CONFIG_ARCH_SHARK

Russell explains the __module_get():
> cyber2000fb.c does it in its module initialization function
> to prevent the module (when built for Shark) from being unloaded.  It
> does this because it's from the days of 2.2 kernels and no one bothered
> writing the module unload support for Shark.

Since 2.4, the correct answer has been to not define an unload fn.

Cc: Russell King <rmk+lkml@arm.linux.org.uk>
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>

--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Adding Alexander Schulz who can comment on the correctness of your patch
for Shark.

On Wed, Feb 04, 2009 at 09:25:47PM +1030, Rusty Russell wrote:
> On Wednesday 04 February 2009 20:41:30 Russell King wrote:
> > On Wed, Feb 04, 2009 at 02:18:08PM +1030, Rusty Russell wrote:
> > >   gameport.c, serio.c and input.c increment their own refcount, but to get
> > > into those init functions someone must be holding a refcount already (ie. a
> > > module depends on this module).  Ditto cyber2000fb.c, and MTD.
> > 
> > Err, wrong.  cyber2000fb.c does it in its module initialization function
> > to prevent the module (when built for Shark) from being unloaded.  It
> > does this because it's from the days of 2.2 kernels and no one bothered
> > writing the module unload support for Shark.  I'm certainly not in a
> > position to do that.
> 
> Thanks, here's the patch then:
> 
> Subject: cyber2000fb.c: use proper method for stopping unload if CONFIG_ARCH_SHARK
> 
> Russell explains the __module_get():
> > cyber2000fb.c does it in its module initialization function
> > to prevent the module (when built for Shark) from being unloaded.  It
> > does this because it's from the days of 2.2 kernels and no one bothered
> > writing the module unload support for Shark.
> 
> Since 2.4, the correct answer has been to not define an unload fn.
> 
> Cc: Russell King <rmk+lkml@arm.linux.org.uk>
> Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
> 
> diff --git a/drivers/video/cyber2000fb.c b/drivers/video/cyber2000fb.c
> --- a/drivers/video/cyber2000fb.c
> +++ b/drivers/video/cyber2000fb.c
> @@ -1736,10 +1736,8 @@ static int __init cyber2000fb_init(void)
>  
>  #ifdef CONFIG_ARCH_SHARK
>  	err = cyberpro_vl_probe();
> -	if (!err) {
> +	if (!err)
>  		ret = 0;
> -		__module_get(THIS_MODULE);
> -	}
>  #endif
>  #ifdef CONFIG_PCI
>  	err = pci_register_driver(&cyberpro_driver);
> @@ -1749,14 +1747,15 @@ static int __init cyber2000fb_init(void)
>  
>  	return ret ? err : 0;
>  }
> +module_init(cyber2000fb_init);
>  
> +#ifndef CONFIG_ARCH_SHARK
>  static void __exit cyberpro_exit(void)
>  {
>  	pci_unregister_driver(&cyberpro_driver);
>  }
> -
> -module_init(cyber2000fb_init);
>  module_exit(cyberpro_exit);
> +#endif
>  
>  MODULE_AUTHOR("Russell King");
>  MODULE_DESCRIPTION("CyberPro 2000, 2010 and 5000 framebuffer driver");