The current code does not check the return value of tg3_nvram_lock before
resetting the lock count (tp->nvram_lock_cnt = 0). This is dangerous
because if tg3_nvram_lock fails, the lock state may be inconsistent,
leading to potential race conditions or undefined behavior.

This patch adds a check for the return value of tg3_nvram_lock. If the
function fails, the error is propagated to the caller, ensuring that
the lock state remains consistent.

Signed-off-by: Wentao Liang <vulab@iscas.ac.cn>
---
 drivers/net/ethernet/broadcom/tg3.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/drivers/net/ethernet/broadcom/tg3.c b/drivers/net/ethernet/broadcom/tg3.c
index 9cc8db10a8d6..851d19b3f43c 100644
--- a/drivers/net/ethernet/broadcom/tg3.c
+++ b/drivers/net/ethernet/broadcom/tg3.c
@@ -9160,7 +9160,9 @@ static int tg3_chip_reset(struct tg3 *tp)
 	if (!pci_device_is_present(tp->pdev))
 		return -ENODEV;
 
-	tg3_nvram_lock(tp);
+	err = tg3_nvram_lock(tp);
+	if (err)
+		return err;
 
 	tg3_ape_lock(tp, TG3_APE_LOCK_GRC);
 
-- 
2.42.0.windows.2

On Tue, Feb 11, 2025 at 7:27 AM Wentao Liang <vulab@iscas.ac.cn> wrote:
>
> The current code does not check the return value of tg3_nvram_lock before
> resetting the lock count (tp->nvram_lock_cnt = 0). This is dangerous
> because if tg3_nvram_lock fails, the lock state may be inconsistent,
> leading to potential race conditions or undefined behavior.
>
> This patch adds a check for the return value of tg3_nvram_lock. If the
> function fails, the error is propagated to the caller, ensuring that
> the lock state remains consistent.
>
> Signed-off-by: Wentao Liang <vulab@iscas.ac.cn>
> ---
>  drivers/net/ethernet/broadcom/tg3.c | 4 +++-
>  1 file changed, 3 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/net/ethernet/broadcom/tg3.c b/drivers/net/ethernet/broadcom/tg3.c
> index 9cc8db10a8d6..851d19b3f43c 100644
> --- a/drivers/net/ethernet/broadcom/tg3.c
> +++ b/drivers/net/ethernet/broadcom/tg3.c
> @@ -9160,7 +9160,9 @@ static int tg3_chip_reset(struct tg3 *tp)
>         if (!pci_device_is_present(tp->pdev))
>                 return -ENODEV;
>
> -       tg3_nvram_lock(tp);
> +       err = tg3_nvram_lock(tp);
> +       if (err)
> +               return err;

It is correct not to check the return value here since we are about to
reset the chip.  The nvram lock is a hardware arbitration lock that is
granted by the hardware.  If the hardware is in a bad state, we may
not get the lock but we should still proceed to reset.  A few lines
below this we reset the tp->nvram_lock_cnt to 0 so it is safe.

On Tue, Feb 11, 2025 at 11:26:58PM +0800, Wentao Liang wrote:
> The current code does not check the return value of tg3_nvram_lock before
> resetting the lock count (tp->nvram_lock_cnt = 0). This is dangerous
> because if tg3_nvram_lock fails, the lock state may be inconsistent,
> leading to potential race conditions or undefined behavior.
> 
> This patch adds a check for the return value of tg3_nvram_lock. If the
> function fails, the error is propagated to the caller, ensuring that
> the lock state remains consistent.
> 
> Signed-off-by: Wentao Liang <vulab@iscas.ac.cn>
> ---
>  drivers/net/ethernet/broadcom/tg3.c | 4 +++-
>  1 file changed, 3 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/net/ethernet/broadcom/tg3.c b/drivers/net/ethernet/broadcom/tg3.c
> index 9cc8db10a8d6..851d19b3f43c 100644
> --- a/drivers/net/ethernet/broadcom/tg3.c
> +++ b/drivers/net/ethernet/broadcom/tg3.c
> @@ -9160,7 +9160,9 @@ static int tg3_chip_reset(struct tg3 *tp)
>  	if (!pci_device_is_present(tp->pdev))
>  		return -ENODEV;
>  
> -	tg3_nvram_lock(tp);
> +	err = tg3_nvram_lock(tp);
> +	if (err)
> +		return err;
>  
>  	tg3_ape_lock(tp, TG3_APE_LOCK_GRC);

A couple notes from me:

  1. Subject should say "PATCH net-next"
  2. Use --base=auto to generate a base-commit.
  3. code looks fine to me, tg3_nvram_lock is checked in all other
     invocations.
  4. that said, seems like tg3_ape_lock could have also gotten a
     check added at the same time?

I can see the argument that tg3_ape_lock cleanup should come
separately, since there are a few unchecked invocations of it other
than the one right next to the one you changed.

So, if you resend with 1 & 2 fixed, feel free to add:

Reviewed-by: Joe Damato <jdamato@fastly.com>