[xdp-hints] Re: [PATCH iwl-next,v2 2/2] igc: Add Tx hardware timestamp request for AF_XDP zero-copy packet

[John Fastabend <john.fastabend@gmail.com>]

Song, Yoong Siang wrote:
> On Saturday, March 2, 2024 1:55 AM, John Fastabend <john.fastabend@gmail.com> wrote:
> >Song Yoong Siang wrote:
> >> This patch adds support to per-packet Tx hardware timestamp request to
> >> AF_XDP zero-copy packet via XDP Tx metadata framework. Please note that
> >> user needs to enable Tx HW timestamp capability via igc_ioctl() with
> >> SIOCSHWTSTAMP cmd before sending xsk Tx hardware timestamp request.
> >>
> >> Same as implementation in RX timestamp XDP hints kfunc metadata, Timer 0
> >> (adjustable clock) is used in xsk Tx hardware timestamp. i225/i226 have
> >> four sets of timestamping registers. Both *skb and *xsk_tx_buffer pointers
> >> are used to indicate whether the timestamping register is already occupied.
> >>
> >> Furthermore, a boolean variable named xsk_pending_ts is used to hold the
> >> transmit completion until the tx hardware timestamp is ready. This is
> >> because, for i225/i226, the timestamp notification event comes some time
> >> after the transmit completion event. The driver will retrigger hardware irq
> >> to clean the packet after retrieve the tx hardware timestamp.
> >>
> >> Besides, xsk_meta is added into struct igc_tx_timestamp_request as a hook
> >> to the metadata location of the transmit packet. When the Tx timestamp
> >> interrupt is fired, the interrupt handler will copy the value of Tx hwts
> >> into metadata location via xsk_tx_metadata_complete().
> >>
> >> Co-developed-by: Lai Peter Jun Ann <jun.ann.lai@intel.com>
> >> Signed-off-by: Lai Peter Jun Ann <jun.ann.lai@intel.com>
> >> Signed-off-by: Song Yoong Siang <yoong.siang.song@intel.com>
> >> ---
> >
> >[...]
> >
> >>
> >> +static void igc_xsk_request_timestamp(void *_priv)
> >> +{
> >> +	struct igc_metadata_request *meta_req = _priv;
> >> +	struct igc_ring *tx_ring = meta_req->tx_ring;
> >> +	struct igc_tx_timestamp_request *tstamp;
> >> +	u32 tx_flags = IGC_TX_FLAGS_TSTAMP;
> >> +	struct igc_adapter *adapter;
> >> +	unsigned long lock_flags;
> >> +	bool found = false;
> >> +	int i;
> >> +
> >> +	if (test_bit(IGC_RING_FLAG_TX_HWTSTAMP, &tx_ring->flags)) {
> >> +		adapter = netdev_priv(tx_ring->netdev);
> >> +
> >> +		spin_lock_irqsave(&adapter->ptp_tx_lock, lock_flags);
> >> +
> >> +		/* Search for available tstamp regs */
> >> +		for (i = 0; i < IGC_MAX_TX_TSTAMP_REGS; i++) {
> >> +			tstamp = &adapter->tx_tstamp[i];
> >> +
> >> +			if (tstamp->skb)
> >> +				continue;
> >> +
> >> +			found = true;
> >> +			break;
> >
> >Not how I would have written this loop construct seems a bit odd
> >to default break but it works.
> 
> Hi John,
> First of all, thank you for reviewing the patch.
> I agree that we can make the loop better. 
> How about I change the loop to below:

That is more natural to me, but whatever reads best for you
is probably ok.

> 
> for (i = 0; i < IGC_MAX_TX_TSTAMP_REGS; i++) {
> 	tstamp = &adapter->tx_tstamp[i];
> 
> 	if (!tstamp->skb) {
> 		found = true;
> 		break;
> 	}
> }
> 
> >
> >> +		}
> >> +
> >> +		/* Return if no available tstamp regs */
> >> +		if (!found) {
> >> +			adapter->tx_hwtstamp_skipped++;
> >> +			spin_unlock_irqrestore(&adapter->ptp_tx_lock,
> >> +					       lock_flags);
> >> +			return;
> >> +		}
> >
> >[...]
> >
> >>
> >> +static void igc_ptp_free_tx_buffer(struct igc_adapter *adapter,
> >> +				   struct igc_tx_timestamp_request *tstamp)
> >> +{
> >> +	if (tstamp->buffer_type == IGC_TX_BUFFER_TYPE_XSK) {
> >> +		/* Release the transmit completion */
> >> +		tstamp->xsk_tx_buffer->xsk_pending_ts = false;
> >> +		tstamp->xsk_tx_buffer = NULL;
> >> +		tstamp->buffer_type = 0;
> >> +
> >> +		/* Trigger txrx interrupt for transmit completion */
> >> +		igc_xsk_wakeup(adapter->netdev, tstamp->xsk_queue_index, 0);
> >
> >Just curious because I didn't find it. Fairly sure I just need to look more,
> >but don't you want to still 'tstamp->skb = NULL' in this path somewhere?
> >It looks like triggering the tx interrupt again with buffer_type == 0 wouldn't
> >do the null.
> >
> >I suspect I just missed it.
> 
> Since the timestamp register will only be used by either skb or xsk,
> So we make tstamp->xsk_tx_buffer and tstamp->skb as union,
> Then based on tstamp->buffer_type to decide whether
> tstamp->xsk_tx_buffer or tstamp->skb should be used.
> 
> My thought is, by setting tstamp->xsk_tx_buffer=NULL,
> tstamp->skb will become NULL as well, and vice versa.

Seems good to me. Maybe a comment though? Otherwise I suspect next
person to read the code will have to spend the extra time to track
it down as well.

> 
> Thanks & Regards
> Siang

Also feel free to carry my ack into the v2 if you make the couple
small nitpick changes.

Acked-by: John Fastabend <john.fastabend@gmail.com>