On Fri, Mar 29, 2024 at 08:19:47AM +0300, Aleksandr Mishin wrote:
> In ks_pcie_setup_rc_app_regs() resource_list_first_type() may return
> NULL which is later dereferenced. Fix this bug by adding NULL check.
>
> Found by Linux Verification Center (linuxtesting.org) with SVACE.
>
> Fixes: 0f71c60ffd26 ("PCI: dwc: Remove storing of PCI resources")
> Signed-off-by: Aleksandr Mishin <amishin@xxxxxxxxxx>
> ---
> drivers/pci/controller/dwc/pci-keystone.c | 6 +++++-
> 1 file changed, 5 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/pci/controller/dwc/pci-keystone.c b/drivers/pci/controller/dwc/pci-keystone.c
> index 844de4418724..00d616654171 100644
> --- a/drivers/pci/controller/dwc/pci-keystone.c
> +++ b/drivers/pci/controller/dwc/pci-keystone.c
> @@ -392,7 +392,11 @@ static void ks_pcie_setup_rc_app_regs(struct keystone_pcie *ks_pcie)
> struct resource *mem;
> int i;
>
> - mem = resource_list_first_type(&pp->bridge->windows, IORESOURCE_MEM)->res;
> + struct resource_entry *ft = resource_list_first_type(&pp->bridge->windows, IORESOURCE_MEM);
> + if (!ft)
> + return;
Like the other one
(https://lore.kernel.org/all/20240328180126.23574-1-amishin@xxxxxxxxxx/),
I think this potentially avoids a NULL pointer dereference (though I
didn't do the analysis in this case to see whether that's actually
possible), but fails to consider the implication of simply skipping
the rest of ks_pcie_setup_rc_app_regs().
"start" and "end" are used only for the loop about "Using Direct 1:1
mapping of RC <-> PCI memory space". If there's no IORESOURCE_MEM
resource, obviously that loop makes no sense. The function would
probably be improved by moving the resource_list_first_type() so it's
next to the loop that uses the result.
But the rest of the function doesn't depend on that IORESOURCE_MEM
resource, and it's not at all clear that the rest of the function
should be skipped.
> + mem = ft->res;
> start = mem->start;
> end = mem->end;
>
> --
> 2.30.2
>
