Re: [PATCH] PCI: Only enable IO window if supported

[Lorenzo Pieralisi]

Hi Guenter,

On Thu, Jun 18, 2015 at 07:01:03PM +0100, Guenter Roeck wrote:
> On Thu, May 28, 2015 at 07:41:12AM -0500, Bjorn Helgaas wrote:
> > 
> > > > I'd like res->flags to reflect the capabilities of the hardware, not
> > > > whether the window is currently enabled.
> > > > 
> > > Flag bits seem to be all taken. Could we use IORESOURCE_DISABLED for that
> > > purpose, or could that cause conflicts elsewhere ?
> > 
> > Yes, I think IORESOURCE_DISABLED would be appropriate for any I/O windows
> > below a host bridge that doesn't support I/O space.
> > 
> I integrated Lorenzo's patch and tried to get this working.

Thanks. How do you want to proceed with this ? Are you taking my patch
and post it along with your updated series ? We need to extend test
coverage to platforms we could not test on, as you know my series
affects all archs but SPARC (I mean it should *not* affect them, this
has to be tested though, I do not have the HW needed, your coverage
for x86 and PowerPC is great but I do not think it can be deemed
sufficient).

Please let me know, thanks !

Lorenzo

> Problem is that the use of a resource is widely checked with "!res->flags"
> throughout the code. That would have to be changed to something like
> "(!res->flags || (res->flags & IORESOURCE_DISABLED))" whereever it is used.
> 
> I tried going with "!res->flags" instead, but have not been able to get it
> to work realiably; it is just very difficult to distinguish if "!res->flags"
> means that the resource has not yet been assigned or if it means that it is not
> supported.
> 
> The correct approach, in my opinion, would be to go with IORESOURCE_DISABLED
> and make the necessary changes whereever needed. Effectively this means to
> replace the "!res->flags" check with something like pci_res_used() [ pick
> your preferred name ] and define it as
> 
> #define pci_res_used(res) ((res)->flags && !((res)->flags & IORESOURCE_DISABLED))
> 
> Is that the right direction ?
> 
> Thanks,
> Guenter
> 
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
Please read the FAQ at  http://www.tux.org/lkml/