On Sat, May 23, 2026 at 11:52:14AM +0530, Ashwin Gundarapu wrote:
> From: Ashwin Gundarapu <[email protected]>
> Date: Sat, 23 May 2026 11:49:40 +0530
> Subject: [PATCH] e1000e: fix memory leak of msix_entries on MSI-X failure
> 
> When MSI-X initialization fails, the driver falls through to try
> MSI or legacy interrupts. However, the msix_entries array allocated
> earlier is not freed, causing a memory leak. Free it and set to
> NULL before falling through to the MSI fallback path.
> 
> Found by code inspection.
> 
> Signed-off-by: Ashwin Gundarapu <[email protected]>
> ---
>  drivers/net/ethernet/intel/e1000e/netdev.c | 10 ++++++----
>  1 file changed, 6 insertions(+), 4 deletions(-)
> 
> diff --git a/drivers/net/ethernet/intel/e1000e/netdev.c 
> b/drivers/net/ethernet/intel/e1000e/netdev.c
> index 7ce0cc8ab8f4..1526069d7fc1 100644
> --- a/drivers/net/ethernet/intel/e1000e/netdev.c
> +++ b/drivers/net/ethernet/intel/e1000e/netdev.c
> @@ -2065,10 +2065,12 @@ void e1000e_set_interrupt_capability(struct 
> e1000_adapter *adapter)
>                                                           a->num_vectors);
>                               if (err > 0)
>                                       return;
> -                     }
> -                     /* MSI-X failed, so fall through and try MSI */
> -                     e_err("Failed to initialize MSI-X interrupts.  Falling 
> back to MSI interrupts.\n");
> -                     e1000e_reset_interrupt_capability(adapter);
> +                        }
> +                        /* MSI-X failed, so fall through and try MSI */
> +                        e_err("Failed to initialize MSI-X interrupts.  
> Falling back to MSI interrupts.\n");
> +                        kfree(adapter->msix_entries);
> +                        adapter->msix_entries = NULL;
> +                        e1000e_reset_interrupt_capability(adapter);

Hi Ashwin,

e1000e_reset_interrupt_capability() already handles freeing (and NULLing)
msix_entries, and it does so after calling pci_disable_msix().

This patch seems to have the effect of bypassing the call to
pci_disable_msix().

Are you sure this is fixing a memory leak as described?

Flagged by https://sashiko.dev/ and https://netdev-ai.bots.linux.dev/sashiko/

>               }
>               adapter->int_mode = E1000E_INT_MODE_MSI;
>               fallthrough;
> --
> 2.43.0
> 
> 

Reply via email to