Re: [RFT PATCH v3] PCI: move pci_read_bridge_bases to the generic PCI layer

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

On Thu, Jul 23, 2015 at 05:12:57PM +0100, Bjorn Helgaas wrote:
> On Thu, Jul 09, 2015 at 11:59:16AM +0100, Lorenzo Pieralisi wrote:
> > When a PCI bus is scanned, upon PCI bridge detection the kernel
> > has to read the bridge registers to set-up its resources so that
> > the PCI resource hierarchy can be validated properly.
> > 
> > Most if not all architectures read PCI bridge registers in the
> > pcibios_fixup_bus hook, that is called by the PCI generic layer
> > whenever a PCI bus is scanned.
> > 
> > Since pci_read_bridge_bases is an arch agnostic operation (and it
> > is carried out on all architectures) it can be moved to the generic
> > PCI layer in order to consolidate code and remove the respective
> > calls from the architectures back-ends.
> > 
> > The PCI_PROBE_ONLY flag is not checked before calling
> > pci_read_bridge_buses in the generic layer since reading the bridge
> > bases is not related to resources assignment; this implies that it
> > can be carried out safely on PCI_PROBE_ONLY systems too and should
> > not affect architectures (alpha, mips) that check the PCI_PROBE_ONLY
> > flag before reading the bridge bases.
> > 
> > Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
> 
> Applied to pci/resource for v4.3, thanks!
> 
> The PCI_PROBE_ONLY text seems backward to me: previously alpha and mips
> only called pci_read_bridge_bases() if PCI_PROBE_ONLY was set.  After this
> patch, alpha and mips systems that do not set PCI_PROBE_ONLY will also call
> pci_read_bridge_bases().

Yes you are right, your log is much clearer.

> I really don't know why alpha and mips were like that.  It seems backwards.
> It seems like we'd want to know the bridge windows if we were assigning
> things, but they only read them if they were *not* going to assign things.

I do not know either I will check again what issues this might trigger
(apart from some odd log messages) hopefully they are not reading the bases
because they are reassigning them anyway on !PCI_PROBE_ONLY, I hope that's
not subtler than that.

Actually I noticed on alpha at least, they claim resources only
on PCI_PROBE_ONLY and for that to succeed bridge bases must be
read before, that may be an explanation but I do not know the reason.

> I'm a little uneasy that we might break some alpha or mips system, since
> there must have been some reason this was done originally.  It'd be ideal
> if somebody could test a non-PCI_PROBE_ONLY system.  But maybe they're all
> obsolete.

I have no way to test them, that's the reason behind the RFT on this
patch, I am uneasy too I do not if anyone can help us test it
(maybe adding it to -next can help with that).

Please let me know.

Thanks !
Lorenzo

> 
> I amended the changelog like this:
> 
>     PCI: Call pci_read_bridge_bases() from core instead of arch code
>     
>     When we scan a PCI bus, we read PCI-PCI bridge window registers with
>     pci_read_bridge_bases() so we can validate the resource hierarchy.  Most
>     architectures call pci_read_bridge_bases() from pcibios_fixup_bus(), but
>     PCI-PCI bridges are not arch-specific, so this doesn't need to be in
>     arch-specific code.
>     
>     Call pci_read_bridge_bases() directly from the PCI core instead of from
>     arch code.
>     
>     For alpha and mips, we now call pci_read_bridge_bases() always; previously
>     we only called it if PCI_PROBE_ONLY was set.
>     
>     [bhelgaas: changelog]
>     Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
>     Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
>     ...
>