On Thu, May 1, 2025 at 11:16 AM Simon Horman <ho...@kernel.org> wrote:
>
> On Mon, Apr 28, 2025 at 07:55:32PM +0000, Brian Vazquez wrote:
> > Add a helper function to correctly handle the lockless
> > synchronization when the sender needs to block. The paradigm is
> >
> > if (no_resources()) {
> > stop_queue();
> > barrier();
> > if (!no_resources())
> > restart_queue();
> > }
> >
> > netif_subqueue_maybe_stop already handles the paradigm correctly, but
> > the code split the check for resources in three parts, the first one
> > (descriptors) followed the protocol, but the other two (completions and
> > tx_buf) were only doing the first part and so race prone.
> >
> > Luckily netif_subqueue_maybe_stop macro already allows you to use a
> > function to evaluate the start/stop conditions so the fix only requires
> > the right helper function to evaluate all the conditions at once.
> >
> > The patch removes idpf_tx_maybe_stop_common since it's no longer needed
> > and instead adjusts separately the conditions for singleq and splitq.
> >
> > Note that idpf_rx_buf_hw_update doesn't need to check for resources
> > since that will be covered in idpf_tx_splitq_frame.
>
> Should the above read idpf_tx_buf_hw_update() rather than
> idpf_rx_buf_hw_update()?
Nice catch, that's a typo indeed.
>
> If so, I see that this is true when idpf_tx_buf_hw_update() is called from
> idpf_tx_singleq_frame(). But is a check required in the case where
> idpf_rx_buf_hw_update() is called by idpf_tx_singleq_map()?
No, the check is not required. The call is at the end of
idpf_tx_singleq_map at that point you already checked for resources
and you're about to send the pkt.
>
> >
> > To reproduce:
> >
> > Reduce the threshold for pending completions to increase the chances of
> > hitting this pause by changing your kernel:
> >
> > drivers/net/ethernet/intel/idpf/idpf_txrx.h
> >
> > -#define IDPF_TX_COMPLQ_OVERFLOW_THRESH(txcq) ((txcq)->desc_count >> 1)
> > +#define IDPF_TX_COMPLQ_OVERFLOW_THRESH(txcq) ((txcq)->desc_count >> 4)
> >
> > Use pktgen to force the host to push small pkts very aggressively:
> >
> > ./pktgen_sample02_multiqueue.sh -i eth1 -s 100 -6 -d $IP -m $MAC \
> > -p 10000-10000 -t 16 -n 0 -v -x -c 64
> >
> > Fixes: 6818c4d5b3c2 ("idpf: add splitq start_xmit")
> > Signed-off-by: Josh Hay <joshua.a....@intel.com>
> > Signed-off-by: Brian Vazquez <bria...@google.com>
> > Signed-off-by: Luigi Rizzo <lri...@google.com>
>
> ...
