On Mon, 19 Apr 2021 18:41:33 +0200, Yizhuo Zhai wrote:

Inside the function uaa_to_xfi(), variable "l_timer" could be uninitialized if pci_bus_read_config_dword() returns PCIBIOS_BAD_REGISTER_NUMBER. However, it is write back to PCI config space via pci_write_config_dword(), which is potentially unsafe.

Thanks for the patch. But it cannot be taken as is.

First off, practically seen, reading PCI_LATENCY_TIMER would never fail. And, even if it were to fail, it's not safer at all to proceed with a value -1.

If any, it should return an error instead. Also, there are a few other similar PCI config reads and those have to be treated as well. But it's likely not worth. If it's a device that is used on VM, you might have more chance to break something unexpectedly, but on a real hardware, it's not. And it's about the init sequence, which won't bring to anything actually crashing the kernel.

Takashi

Signed-off-by: Yizhuo yzhai003@ucr.edu

sound/pci/ctxfi/cthw20k1.c | 2 +-  1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/sound/pci/ctxfi/cthw20k1.c b/sound/pci/ctxfi/cthw20k1.c index 0cea4982ed7d..7c3436499974 100644 --- a/sound/pci/ctxfi/cthw20k1.c +++ b/sound/pci/ctxfi/cthw20k1.c @@ -1789,7 +1789,7 @@ static struct capabilities hw_capabilities(struct hw *hw)  static int uaa_to_xfi(struct pci_dev *pci)  {         unsigned int bar0, bar1, bar2, bar3, bar4, bar5; -       unsigned int cmd, irq, cl_size, l_timer, pwr; +       unsigned int cmd, irq, cl_size, l_timer = ~0, pwr;         unsigned int is_uaa;         unsigned int data[4] = {0};         unsigned int io_base; --  2.31.1

-- Kind Regards,

Yizhuo Zhai

Computer Science, Graduate Student University of California, Riverside