Re: [PATCH v8 3/6] PCI: endpoint: Add pci_epf_align_addr() helper for address alignment

[Niklas Cassel <cassel@kernel.org>]

On Tue, Nov 26, 2024 at 09:49:03AM +0530, Manivannan Sadhasivam wrote:
> On Mon, Nov 25, 2024 at 02:22:23PM -0500, Frank Li wrote:
> > On Sun, Nov 24, 2024 at 01:02:39PM +0530, Manivannan Sadhasivam wrote:
> > > On Sat, Nov 16, 2024 at 09:40:43AM -0500, Frank Li wrote:
> > > > +static inline int pci_epf_align_inbound_addr_lo_hi(struct pci_epf *epf, enum pci_barno bar,
> > > > +						   u32 low, u32 high, u64 *base, size_t *off)
> > >
> > > Why can't you just use pci_epf_align_inbound_addr() directly? Or the caller
> > > could pass u64 address directly.
> > 
> > 
> > msi message sperate low32 and high32.  (h << 32 | low) is quite easy to
> > cause build warning.  it should be ((u64) h << 32) | low. Avoid copy this
> > logic code at many EPF places.
> > 
> 
> There is absolutely no overhead in doing so. Also the concern for me is,
> pci_epf_align_inbound_addr() is exported but only used within
> pci_epf_align_inbound_addr_lo_hi(). This causes confusion. So I'd prefer to have
> a single exported API that is used by the callers.

Yes, other EPF drivers will need to copy the line:
pci_epf_align_inbound_addr(..., ((u64) h << 32) | low, ...)
instead of:
pci_epf_align_inbound_addr_lo_hi(..., low, high, ...)

which I think is fine to be honest.

Probably simplest thing is just to kill
pci_epf_align_inbound_addr_lo_hi().


Kind regards,
Niklas