On Sun, May 14, 2023 at 11:12:22PM +0300, Dan Carpenter wrote:
> On Sun, May 14, 2023 at 07:06:52PM +0530, Manivannan Sadhasivam wrote:
> > On Thu, May 11, 2023 at 10:06:40AM +0300, Dan Carpenter wrote:
> > > Hello Damien Le Moal,
> > >
> > > The patch 01c68988addf: "PCI: endpoint: Automatically create a
> > > function specific attributes group" from Apr 15, 2023, leads to the
> > > following Smatch static checker warning:
> > >
> > > drivers/pci/endpoint/pci-ep-cfs.c:540 pci_ep_cfs_add_type_group()
> > > warn: 'group' isn't an ERR_PTR
> > >
> > > drivers/pci/endpoint/pci-ep-cfs.c
> > > 532 static void pci_ep_cfs_add_type_group(struct pci_epf_group *epf_group)
> > > 533 {
> > > 534 struct config_group *group;
> > > 535
> > > 536 group = pci_epf_type_add_cfs(epf_group->epf, &epf_group->group);
> > > 537 if (!group)
> > > 538 return;
> > > 539
> > > --> 540 if (IS_ERR(group)) {
> > >
> > > pci_epf_type_add_cfs() does not return error pointers currently. Which
> > > is fine. Presumably it will start returning them later. But could you
> > > add some comments next to the pci_epf_type_add_cfs() to explain what a
> > > NULL return means vs an error pointer return?
> > >
> >
> > pci_epf_type_add_cfs() may return ERR_PTR from add_cfs() callback.
> >
>
> Right, we presumably will add new implementations of ->add_cfs in the
> future which return error pointers. Currently in linux-next there are
> only two and neither of them fail. (Smatch looks at the returns from
> ->add_cfs when it says that these don't return error pointers. Of
> course, it depends on your .config as well. But a quick grep confirms
> that both implementations are included in my ARM allmodconfig).
>
Ah okay. I never thought Smatch would look into callback definitions. Good to
know!
- Mani
> > Regarding comments, it should be added as a part of kdoc for
> > pci_epf_type_add_cfs(). It already does for NULL part but not for ERR_PTR.
>
> Excelent thanks.
>
> regards,
> dan carpenter
>
--
மணிவண்ணன் சதாசிவம்
