Re: [PATCH] PCI: s390: Fix use-after-free of PCI bus resources with s390 per-function hotplug

[Niklas Schnelle]

On Tue, 2023-02-14 at 10:46 +0100, Gerd Bayer wrote:
> Hi Niklas,
> 
> when comparing pci_bus_remove_resource with pci_bus_remove_resources,
> I find that the "single-resource" variant might be ending too early.
> 
> On Fri, 2023-02-03 at 12:48 +0100, Niklas Schnelle wrote:
> > +void pci_bus_remove_resource(struct pci_bus *bus, struct resource
> > *res)
> > +{
> > +       struct pci_bus_resource *bus_res, *tmp;
> > +       int i;
> > +
> > +       for (i = 0; i < PCI_BRIDGE_RESOURCE_NUM; i++) {
> > +               if (bus->resource[i] == res) {
> > +                       bus->resource[i] = NULL;
> > +                       return;
>                           ^^^^^^^
> Did you mean to "break" here, rather than end the routine?

No the return is intended. We're looking to remove a single resource
and if I understand things correctly then that resource can either be
in bus->resource[] (x)or in bus->resources depending on whether it is a
PCI bridge resource. Either way once found and removed from the
respective array/list we're done and can thus return immediately.

> > +               }
> > +       }
> > +
> > +       list_for_each_entry_safe(bus_res, tmp, &bus->resources, list)
> > {
> > +               if (bus_res->res == res) {
> > +                       list_del(&bus_res->list);
> > +                       kfree(bus_res);
> > +                       return;
>                           ^^^^^^^
> Here "break" and "return" have the same effect, but "break" would be
> "symmetric".
> > +               }
> > +       }
> > +       return;
> > +
> > +}
> 
> While this might be a nit, I'd like to better separate the "single-
> resource" variant's name. How about pci_bus_remove_one_resource - I
> know it's getting long...
> 
> Thanks,
> Gerd

I have no strong feelings on the name.