Re: [net PatchV3] Octeontx2-vf: Fix max packet length errors

[Hariprasad Kelam <hkelam@marvell.com>]

On 2025-07-21 at 19:28:05, Andrew Lunn (andrew@lunn.ch) wrote:
> On Mon, Jul 21, 2025 at 02:28:15PM +0530, Hariprasad Kelam wrote:
> > Implement packet length validation before submitting packets to
> > the hardware to prevent MAXLEN_ERR. Increment tx_dropped counter
> > on failure.
> 
> Sorry, i did not look at previous versions of this patch, so i might
> be asking a question some other Reviewer already asked.
> 
> How expensive is MAXLEN_ERR? What do you need to do when it happens?
>
  On error case, hardware raises the queue interrupts about max lenth errors 
  goes to hang state. Driver needs to execute reset to come out of the state.

 
> I would _guess_ that if ndev->mtu is set correctly, and any change to
> it validated, you are going to get very few packets which are too big.
> 
> Is it better to introduce this test on the hot path which effects
> every single packet, or just deal with MAXLEN_ERR if it ever actually
> happens, so leaving the hot path optimised for the common case?
> 
> Maybe you could include something about this in the commit message?
>
 ACK will update the commit description. 
> 	Andrew
>