Re: [PATCH v9 07/11] PCI/CXL: Discover the CXL reset scope

[sashiko-bot@kernel.org]

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [High] CXL Reset scope validation is conditionally bypassed during actual reset execution, creating a TOCTOU vulnerability that fails to protect siblings.
--

--- Patch [7]: [PATCH v9 07/11] PCI/CXL: Discover the CXL reset scope ---
commit 363800e2ebd7dad78179683be8d0a027b5bd8a7e
Author: Srirangan Madhavan <smadhavan@nvidia.com>

PCI/CXL: Discover the CXL reset scope

This commit adds reset context support to validate that a CXL Reset is
function-scoped before advertising it as a PCI reset method. It uses the
Non-CXL Function Map, ARI/devfn rules, and CXL capability bits to reject
the reset if another same-scope function would be affected.

> diff --git a/drivers/cxl/core/reset.c b/drivers/cxl/core/reset.c
> index fb741b66e5022..cdfaa699583d2 100644
> --- a/drivers/cxl/core/reset.c
> +++ b/drivers/cxl/core/reset.c
[ ... ]
> @@ -890,6 +1037,7 @@ static int cxl_reset_execute(struct pci_dev *pdev, int dvsec)
>  int cxl_reset_function(struct pci_dev *pdev, bool probe)
>  {
>  	struct cxl_hdm_range_context range_ctx;
> +	struct cxl_reset_context ctx;
>  	int dvsec;
>  	int rc;
>  
> @@ -897,8 +1045,9 @@ int cxl_reset_function(struct pci_dev *pdev, bool probe)
>  	if (dvsec < 0)
>  		return dvsec;
>  
> +	cxl_reset_context_init(&ctx, pdev);
>  	if (probe)
> -		return 0;
> +		return cxl_reset_validate_function_scoped(&ctx);

[Severity: High]
Is it safe to conditionally bypass scope validation when probe is false?

During the early probe phase, when this is called from pci_init_capabilities()
for Function 0, higher-numbered sibling functions might not have been
enumerated and added to the bus list yet. As a result,
cxl_reset_validate_function_scoped() could find no siblings and incorrectly
report the reset as safe.

Later, when a user or driver actually triggers the reset and this function
is called with probe=0, it skips validation and executes the reset directly.
Could this inadvertently reset and destroy the state of active sibling
functions that were enumerated after the initial probe?

>  
>  	cxl_hdm_range_context_init(&range_ctx);

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260709010304.680422-1-smadhavan@nvidia.com?part=7