Re: avoiding pci_disable_device()...

[Takashi Iwai <tiwai@suse.de>]

At Mon, 14 Feb 2005 21:05:58 -0500,
Jeff Garzik wrote:
> 
> Arjan van de Ven wrote:
> >>No.  You also need to consider situations such as out-of-tree drivers 
> >>for the same hardware (might not use PCI API), and situations where you 
> >>have peer devices discovered and used (PCI API doesn't have "hey, <this> 
> >>device is associated with <current driver>, too" capability)
> > 
> > 
> > 
> > there's not a lot you or anyone else can do about such broken (and often
> > proprietary) drivers.... if a device doesn't use the kernel API's its
> > end of game basically. Adding more new API's isn't going to help you ...
> 
> 
> This specific instance isn't about adding a new API, but using an 
> existing one correctly.
> 
> If pci_request_regions() fails, that implies another driver is using the 
> kernel API to let you know the region is unavailable.  You should honor 
> that, by not disabling the hardware in that case.

I guess an enable counter as Alan proposed would fix this problem
well, except for the case that an out-of-tree driver allocates the
resource without calling pci_enable_device().

OTOH this will introduce more buglets to broken drivers which don't
call pci_disable_device() properly.  Consequently, the ad hoc fix to
each driver like Jeff's patch might be most practical...


Takashi