RE: [PATCH] PCI: endpoint: Fix NULL pointer dereference for ->get_features()

[Sriram Dash]

> From: Kishon Vijay Abraham I <kishon@xxxxxx>
> Subject: Re: [PATCH] PCI: endpoint: Fix NULL pointer dereference for -
> >get_features()
> 
> Hi Sriram,
> 
> On 3/19/2020 7:06 PM, Sriram Dash wrote:
> >> From: Shradha Todi <shradha.t@xxxxxxxxxxx>
> >> Subject: [PATCH] PCI: endpoint: Fix NULL pointer dereference for -
> >>> get_features()
> >>
> >> get_features ops of pci_epc_ops may return NULL, causing NULL pointer
> >> dereference in pci_epf_test_bind function. Let us add a check for
> >> pci_epc_feature pointer in pci_epf_test_bind before we access it to
> >> avoid any such NULL pointer dereference and return -ENOTSUPP in case
> >> pci_epc_feature is not found.
> >>
> >> Reviewed-by: Pankaj Dubey <pankaj.dubey@xxxxxxxxxxx>
> >> Signed-off-by: Sriram Dash <sriram.dash@xxxxxxxxxxx>
> >> Signed-off-by: Shradha Todi <shradha.t@xxxxxxxxxxx>
> >> ---
> >
> > Hi Kishon,
> >
> > Any update on this?
> 
> Don't we access epc_features only after checking if epc_features is not NULL in
> pci_epf_test_bind() function? However we are accessing epc_features in
> multiple other functions all over pci-epf-test.

We access the epc_feature after checking the NULL condition in the bind function.
However, we do not stop if the epc_feature is NULL and proceed for allocation in the
pci_epf_test_alloc_space function, for example. During this allocation, we do not check
for NULL condition for epc_feature and hence, if any controller driver is not providing
the epc features, it will panic accessing epc_features.

> 
> So the patch itself is correct though the commit log has to be fixed. You should
> also check if all the endpoint controller drivers existing currently provides
> epc_features.

At the moment, there is no issue for existing controller drivers as I can see almost
all drivers are providing epc_features. But, this is not a mandatory feature and some
controller drivers may not have epc_features implemented, may be in the near future.
But because we are dealing with the configfs, the application need not bother about
the driver details underneath.

IMO, the code should be fixed regardless and should not cause panic in any case.

> 
> Thanks
> Kishon
> >
> >
> >>  drivers/pci/endpoint/functions/pci-epf-test.c | 15 +++++++++------
> >>  1 file changed, 9 insertions(+), 6 deletions(-)
> >>
> >> diff --git a/drivers/pci/endpoint/functions/pci-epf-test.c
> >> b/drivers/pci/endpoint/functions/pci-epf-test.c
> >> index c9121b1b9fa9..af4537a487bf 100644
> >> --- a/drivers/pci/endpoint/functions/pci-epf-test.c
> >> +++ b/drivers/pci/endpoint/functions/pci-epf-test.c
> >> @@ -510,14 +510,17 @@ static int pci_epf_test_bind(struct pci_epf *epf)
> >>  		return -EINVAL;
> >>
> >>  	epc_features = pci_epc_get_features(epc, epf->func_no);
> >> -	if (epc_features) {
> >> -		linkup_notifier = epc_features->linkup_notifier;
> >> -		msix_capable = epc_features->msix_capable;
> >> -		msi_capable = epc_features->msi_capable;
> >> -		test_reg_bar = pci_epc_get_first_free_bar(epc_features);
> >> -		pci_epf_configure_bar(epf, epc_features);
> >> +	if (!epc_features) {
> >> +		dev_err(dev, "epc_features not implemented\n");
> >> +		return -ENOTSUPP;
> >>  	}
> >>
> >> +	linkup_notifier = epc_features->linkup_notifier;
> >> +	msix_capable = epc_features->msix_capable;
> >> +	msi_capable = epc_features->msi_capable;
> >> +	test_reg_bar = pci_epc_get_first_free_bar(epc_features);
> >> +	pci_epf_configure_bar(epf, epc_features);
> >> +
> >>  	epf_test->test_reg_bar = test_reg_bar;
> >>  	epf_test->epc_features = epc_features;
> >>
> >> --
> >> 2.17.1
> >
> >