[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Mon, 06 Jun 2022 16:55:21 -0700
From: Vinicius Costa Gomes <vinicius.gomes@...el.com>
To: Kurt Kanzenbach <kurt@...utronix.de>,
Jesse Brandeburg <jesse.brandeburg@...el.com>,
Tony Nguyen <anthony.l.nguyen@...el.com>
Cc: "David S. Miller" <davem@...emloft.net>,
Eric Dumazet <edumazet@...gle.com>,
Jakub Kicinski <kuba@...nel.org>,
Paolo Abeni <pabeni@...hat.com>,
intel-wired-lan@...ts.osuosl.org, netdev@...r.kernel.org,
Kurt Kanzenbach <kurt@...utronix.de>
Subject: Re: [PATCH net-next] igc: Lift TAPRIO schedule restriction
Hi Kurt,
Kurt Kanzenbach <kurt@...utronix.de> writes:
> Add support for Qbv schedules where one queue stays open
> in consecutive entries. Currently that's not supported.
>
> Example schedule:
>
> |tc qdisc replace dev ${INTERFACE} handle 100 parent root taprio num_tc 3 \
> | map 2 2 1 0 2 2 2 2 2 2 2 2 2 2 2 2 \
> | queues 1@0 1@1 2@2 \
> | base-time ${BASETIME} \
> | sched-entry S 0x01 300000 \ # Stream High/Low
> | sched-entry S 0x06 500000 \ # Management and Best Effort
> | sched-entry S 0x04 200000 \ # Best Effort
> | flags 0x02
>
> Signed-off-by: Kurt Kanzenbach <kurt@...utronix.de>
> ---
> drivers/net/ethernet/intel/igc/igc_main.c | 23 +++++++++++++++++------
> 1 file changed, 17 insertions(+), 6 deletions(-)
>
> diff --git a/drivers/net/ethernet/intel/igc/igc_main.c b/drivers/net/ethernet/intel/igc/igc_main.c
> index ae17af44fe02..4758bdbe5df3 100644
> --- a/drivers/net/ethernet/intel/igc/igc_main.c
> +++ b/drivers/net/ethernet/intel/igc/igc_main.c
> @@ -5813,9 +5813,10 @@ static bool validate_schedule(struct igc_adapter *adapter,
> return false;
>
> for (n = 0; n < qopt->num_entries; n++) {
> - const struct tc_taprio_sched_entry *e;
> + const struct tc_taprio_sched_entry *e, *prev;
> int i;
>
> + prev = n ? &qopt->entries[n - 1] : NULL;
> e = &qopt->entries[n];
>
> /* i225 only supports "global" frame preemption
> @@ -5828,7 +5829,12 @@ static bool validate_schedule(struct igc_adapter *adapter,
> if (e->gate_mask & BIT(i))
> queue_uses[i]++;
>
> - if (queue_uses[i] > 1)
> + /* There are limitations: A single queue cannot be
> + * opened and closed multiple times per cycle unless the
> + * gate stays open. Check for it.
> + */
> + if (queue_uses[i] > 1 &&
> + !(prev->gate_mask & BIT(i)))
Perhaps I am missing something, I didn't try to run, but not checking if
'prev' can be NULL, could lead to crashes for some schedules, no?
What I have in mind is a schedule that queue 0 is mentioned multiple
times, for example:
| sched-entry S 0x01 300000 \ # Stream High/Low
| sched-entry S 0x03 500000 \ # Management and Best Effort
| sched-entry S 0x05 200000 \ # Best Effort
Anyway, looks much cleaner than what I had in mind when I wrote that
fixme.
Cheers,
--
Vinicius
Powered by blists - more mailing lists