[xdp-hints] Re: [PATCH bpf-next v8 4/5] igc: Refactor empty packet insertion into a reusable function

[Maciej Fijalkowski <maciej.fijalkowski@intel.com>]

On Wed, Feb 05, 2025 at 10:41:15AM +0800, Song Yoong Siang wrote:
> Refactor the code for inserting an empty packet into a new function
> igc_insert_empty_packet(). This change extracts the logic for inserting
> an empty packet from igc_xmit_frame_ring() into a separate function,
> allowing it to be reused in future implementations, such as the XDP
> zero copy transmit function.
> 
> This patch introduces no functional changes.
> 
> Signed-off-by: Song Yoong Siang <yoong.siang.song@intel.com>

Your SoB should be last in the set of tags.

> Reviewed-by: Faizal Rahim <faizal.abdul.rahim@linux.intel.com>
> ---
>  drivers/net/ethernet/intel/igc/igc_main.c | 42 ++++++++++++-----------
>  1 file changed, 22 insertions(+), 20 deletions(-)
> 
> diff --git a/drivers/net/ethernet/intel/igc/igc_main.c b/drivers/net/ethernet/intel/igc/igc_main.c
> index 21f318f12a8d..553d6d82af0d 100644
> --- a/drivers/net/ethernet/intel/igc/igc_main.c
> +++ b/drivers/net/ethernet/intel/igc/igc_main.c
> @@ -1566,6 +1566,26 @@ static bool igc_request_tx_tstamp(struct igc_adapter *adapter, struct sk_buff *s
>  	return false;
>  }
>  
> +static void igc_insert_empty_packet(struct igc_ring *tx_ring)
> +{
> +	struct igc_tx_buffer *empty_info;
> +	struct sk_buff *empty;
> +	void *data;
> +
> +	empty_info = &tx_ring->tx_buffer_info[tx_ring->next_to_use];
> +	empty = alloc_skb(IGC_EMPTY_FRAME_SIZE, GFP_ATOMIC);
> +	if (!empty)
> +		return;
> +
> +	data = skb_put(empty, IGC_EMPTY_FRAME_SIZE);
> +	memset(data, 0, IGC_EMPTY_FRAME_SIZE);
> +
> +	igc_tx_ctxtdesc(tx_ring, 0, false, 0, 0, 0);
> +
> +	if (igc_init_tx_empty_descriptor(tx_ring, empty, empty_info) < 0)
> +		dev_kfree_skb_any(empty);

I still don't like the fact igc_insert_empty_packet() doesn't communicate
to caller whether it successfully produced descriptors or not.

Look at this from igc_xmit_frame_ring() POV:
- at the beginning you peek at Tx ring whether there is required amount of
  descriptors free to be used
- but then here's your additional routine which might consume two more
  descs and you are not aware of the status
- then you continue to further produce descriptors assuming there is
  enough space in Tx ring

Right now igc_init_tx_empty_descriptor() returns -EBUSY when ring is full.
How can that happen in the first place + what if it would *really* happen
though? You just continue with your Tx flow.

What I'm trying to say here is, at least from correctness POV, you should
take into the account two potential descriptors for launchtime feature
when calling igc_maybe_stop_tx(). And igc_init_tx_empty_descriptor()
should not really care about space in ring, it should be a caller's job to
call it only when it will be sure it's safe to do so.

> +}
> +
>  static netdev_tx_t igc_xmit_frame_ring(struct sk_buff *skb,
>  				       struct igc_ring *tx_ring)
>  {
> @@ -1603,26 +1623,8 @@ static netdev_tx_t igc_xmit_frame_ring(struct sk_buff *skb,
>  	skb->tstamp = ktime_set(0, 0);
>  	launch_time = igc_tx_launchtime(tx_ring, txtime, &first_flag, &insert_empty);
>  
> -	if (insert_empty) {
> -		struct igc_tx_buffer *empty_info;
> -		struct sk_buff *empty;
> -		void *data;
> -
> -		empty_info = &tx_ring->tx_buffer_info[tx_ring->next_to_use];
> -		empty = alloc_skb(IGC_EMPTY_FRAME_SIZE, GFP_ATOMIC);
> -		if (!empty)
> -			goto done;
> -
> -		data = skb_put(empty, IGC_EMPTY_FRAME_SIZE);
> -		memset(data, 0, IGC_EMPTY_FRAME_SIZE);
> -
> -		igc_tx_ctxtdesc(tx_ring, 0, false, 0, 0, 0);
> -
> -		if (igc_init_tx_empty_descriptor(tx_ring,
> -						 empty,
> -						 empty_info) < 0)
> -			dev_kfree_skb_any(empty);
> -	}
> +	if (insert_empty)
> +		igc_insert_empty_packet(tx_ring);
>  
>  done:
>  	/* record the location of the first descriptor for this packet */
> -- 
> 2.34.1
>