Re: [PATCH RFC v2 1/2] drivers: pci: fix pci_mmap_fits() implementation for procfs mmap

[Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>]

On Wed, Nov 12, 2014 at 07:23:49AM +0000, Benjamin Herrenschmidt wrote:
> On Mon, 2014-11-10 at 16:04 -0700, Bjorn Helgaas wrote:
> > But I'm not sure I have this right.  If the procfs offset is either
> > the
> > CPU physical address or the BAR value, then pci_resource_to_user()
> > should be (depending on the arch) either a no-op or use
> > pci_resource_to_bus().
> > 
> > But that's not how it's implemented.  Maybe it *could* be?  If
> > pci_resource_to_user() gives you something that's not a CPU physical
> > address and not a bus address, what *does* it give you, and why would
> > we
> > need this third kind of thing?
> > 
> > FWIW, I think the discussion leading up to pci_resource_to_user() is
> > here:
> > http://lkml.iu.edu/hypermail/linux/kernel/0504.3/0467.html
> 
> Oh, man... I remember that was all a giant trainwreck and some stuff
> just couldn't be made completely right due to broken assumptions by
> the proc code and users of it... but I don't remember all the details.
> 
> I think /proc users don't necessarily pass a BAR value but something
> they try to somewhat translates themselves via the "resources" file,
> which ends up working ... or not, depending on various factors such
> as 32 vs 64 bit etc...
> 
> I wonder who still uses this interface....

+1, even though I do not think that leaving it as it is is a good idea,
hence I posted this series.

I tried to fix it while fixing the way ARM pcibios code handles
pci_mmap_page_range() (for both procfs and sysfs mappings).

I will do what Bjorn suggested, more comments from arches maintainers
are welcome.

Lorenzo