[+cc Terry, Robert, CXL list, Smita, Dongdong]
On Fri, Sep 05, 2025 at 02:25:27AM +0800, Vernon Yang wrote:
> From: Vernon Yang <yanglinch...@kylinos.cn>
>
> The kzalloc(GFP_KERNEL) may return NULL, so all accesses to
> aer_info->xxx will result in kernel panic. Fix it.
>
> Signed-off-by: Vernon Yang <yanglinch...@kylinos.cn>
Applied to pci/aer for v6.18, thanks, Vernon!
Not directly related to this patch, but I'm concerned about some users
of dev->aer_cap.
Most users of dev->aer_cap either (a) check that it's set before using
it or (b) are called in paths obviously only reachable via an AER
interrupt.
But there are a few users of dev->aer_cap that use it without checking
it for zero, and it's not obvious to me that it must be valid:
- pci_aer_unmask_internal_errors(), added by b7e9392d5d46 ("PCI/AER:
Unmask RCEC internal errors to enable RCH downstream port error
handling")
- dpc_get_aer_uncorrect_severity(), added by 9f08a5d896ce ("PCI/DPC:
Fix print AER status in DPC event handling")
- dpc_is_surprise_removal(), added by 2ae8fbbe1cd4 ("PCI/DPC: Ignore
Surprise Down error on hot removal")
> ---
> drivers/pci/pcie/aer.c | 4 ++++
> 1 file changed, 4 insertions(+)
>
> diff --git a/drivers/pci/pcie/aer.c b/drivers/pci/pcie/aer.c
> index e286c197d716..aeb2534f50dd 100644
> --- a/drivers/pci/pcie/aer.c
> +++ b/drivers/pci/pcie/aer.c
> @@ -383,6 +383,10 @@ void pci_aer_init(struct pci_dev *dev)
> return;
>
> dev->aer_info = kzalloc(sizeof(*dev->aer_info), GFP_KERNEL);
> + if (!dev->aer_info) {
> + dev->aer_cap = 0;
> + return;
> + }
>
> ratelimit_state_init(&dev->aer_info->correctable_ratelimit,
> DEFAULT_RATELIMIT_INTERVAL,
> DEFAULT_RATELIMIT_BURST);
> --
> 2.51.0
>
