drivers/pci/controller/dwc/pcie-kirin.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)
In kirin_pcie_parse_port() pcie->num_slots is compared to
pcie->gpio_id_reset size (MAX_PCI_SLOTS). Need to fix
condition to pcie->num_slots + 1 >= MAX_PCI_SLOTS and move
pcie->num_slots increment lower to avoid out-of-bounds
array access.
Found by Linux Verification Center (linuxtesting.org) with SVACE.
Fixes: b22dbbb24571 ("PCI: kirin: Support PERST# GPIOs for HiKey970 external PEX 8606 bridge")
Signed-off-by: Alexandra Diupina <adiupina@astralinux.ru>
---
v2: some changes
drivers/pci/controller/dwc/pcie-kirin.c | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/drivers/pci/controller/dwc/pcie-kirin.c b/drivers/pci/controller/dwc/pcie-kirin.c
index d5523f302102..deab1e653b9a 100644
--- a/drivers/pci/controller/dwc/pcie-kirin.c
+++ b/drivers/pci/controller/dwc/pcie-kirin.c
@@ -412,12 +412,12 @@ static int kirin_pcie_parse_port(struct kirin_pcie *pcie,
if (pcie->gpio_id_reset[i] < 0)
continue;
- pcie->num_slots++;
- if (pcie->num_slots > MAX_PCI_SLOTS) {
+ if (pcie->num_slots + 1 >= MAX_PCI_SLOTS) {
dev_err(dev, "Too many PCI slots!\n");
ret = -EINVAL;
goto put_node;
}
+ pcie->num_slots++;
ret = of_pci_get_devfn(child);
if (ret < 0) {
--
2.30.2
Em Fri, 19 Jul 2024 15:21:53 +0300
Alexandra Diupina <adiupina@astralinux.ru> escreveu:
> In kirin_pcie_parse_port() pcie->num_slots is compared to
> pcie->gpio_id_reset size (MAX_PCI_SLOTS). Need to fix
> condition to pcie->num_slots + 1 >= MAX_PCI_SLOTS and move
> pcie->num_slots increment lower to avoid out-of-bounds
> array access.
>
> Found by Linux Verification Center (linuxtesting.org) with SVACE.
>
> Fixes: b22dbbb24571 ("PCI: kirin: Support PERST# GPIOs for HiKey970 external PEX 8606 bridge")
> Signed-off-by: Alexandra Diupina <adiupina@astralinux.ru>
LGTM.
Reviewed-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
> ---
> v2: some changes
> drivers/pci/controller/dwc/pcie-kirin.c | 4 ++--
> 1 file changed, 2 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/pci/controller/dwc/pcie-kirin.c b/drivers/pci/controller/dwc/pcie-kirin.c
> index d5523f302102..deab1e653b9a 100644
> --- a/drivers/pci/controller/dwc/pcie-kirin.c
> +++ b/drivers/pci/controller/dwc/pcie-kirin.c
> @@ -412,12 +412,12 @@ static int kirin_pcie_parse_port(struct kirin_pcie *pcie,
> if (pcie->gpio_id_reset[i] < 0)
> continue;
>
> - pcie->num_slots++;
> - if (pcie->num_slots > MAX_PCI_SLOTS) {
> + if (pcie->num_slots + 1 >= MAX_PCI_SLOTS) {
> dev_err(dev, "Too many PCI slots!\n");
> ret = -EINVAL;
> goto put_node;
> }
> + pcie->num_slots++;
>
> ret = of_pci_get_devfn(child);
> if (ret < 0) {
Thanks,
Mauro
just a friendly reminder
19/07/24 15:21, Alexandra Diupina пишет:
> In kirin_pcie_parse_port() pcie->num_slots is compared to
> pcie->gpio_id_reset size (MAX_PCI_SLOTS). Need to fix
> condition to pcie->num_slots + 1 >= MAX_PCI_SLOTS and move
> pcie->num_slots increment lower to avoid out-of-bounds
> array access.
>
> Found by Linux Verification Center (linuxtesting.org) with SVACE.
>
> Fixes: b22dbbb24571 ("PCI: kirin: Support PERST# GPIOs for HiKey970 external PEX 8606 bridge")
> Signed-off-by: Alexandra Diupina <adiupina@astralinux.ru>
> ---
> v2: some changes
> drivers/pci/controller/dwc/pcie-kirin.c | 4 ++--
> 1 file changed, 2 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/pci/controller/dwc/pcie-kirin.c b/drivers/pci/controller/dwc/pcie-kirin.c
> index d5523f302102..deab1e653b9a 100644
> --- a/drivers/pci/controller/dwc/pcie-kirin.c
> +++ b/drivers/pci/controller/dwc/pcie-kirin.c
> @@ -412,12 +412,12 @@ static int kirin_pcie_parse_port(struct kirin_pcie *pcie,
> if (pcie->gpio_id_reset[i] < 0)
> continue;
>
> - pcie->num_slots++;
> - if (pcie->num_slots > MAX_PCI_SLOTS) {
> + if (pcie->num_slots + 1 >= MAX_PCI_SLOTS) {
> dev_err(dev, "Too many PCI slots!\n");
> ret = -EINVAL;
> goto put_node;
> }
> + pcie->num_slots++;
>
> ret = of_pci_get_devfn(child);
> if (ret < 0) {
Hello, > just a friendly reminder Would you mind sending a v2 as a standalone patch? So that I can pick it up via Patchwork and such. I would be much obliged. I wonder if Mauro would be willing to have another review of the new iteration of the fix. Just something that would be nice to have his tag added. Krzysztof
© 2016 - 2025 Red Hat, Inc.