On Thu, Mar 01, 2018 at 12:13:10PM -0700, Logan Gunthorpe wrote:
>
>
> On 01/03/18 11:02 AM, Bjorn Helgaas wrote:
> > > void pci_enable_acs(struct pci_dev *dev)
> > > {
> > > + if (pci_p2pdma_disable_acs(dev))
> > > + return;
> >
> > This doesn't read naturally to me. I do see that when
> > CONFIG_PCI_P2PDMA is not set, pci_p2pdma_disable_acs() does nothing
> > and returns 0, so we'll go ahead and try to enable ACS as before.
> >
> > But I think it would be clearer to have an #ifdef CONFIG_PCI_P2PDMA
> > right here so it's more obvious that we only disable ACS when it's
> > selected.
>
> I could do this... however, I wrote it this way because I've read Linus
> dislikes using #ifdef's inside function bodies and I personally agree with
> that sentiment.
I try to avoid #ifdefs too, but in this case the plain reading of the
code makes no sense (we're in a function called "enable_acs()", and
the first thing we do is call a function to "disable_acs()".
Disabling ACS is scary for all the security reasons mentioned
elsewhere, so a reader who knows what ACS does and cares about
virtualization and security *has* to go look up the definition of
pci_p2pdma_disable_acs().
If you put the #ifdef right here, then it's easier to read because we
can see that "oh, this is a special and uncommon case that I can
probably ignore".
Bjorn
