| Message ID | 20230307064531.68840-1-takahiro.akashi@linaro.org |
|---|---|
| State | Accepted |
| Delegated to: | Anthony Nguyen |
| Headers | show |
| Series | igc: fix the validation logic for taprio's gate list | expand |
Kurt Kanzenbach <kurt@linutronix.de> writes: > On Tue Mar 07 2023, AKASHI Takahiro wrote: >> The check introduced in the commit a5fd39464a40 ("igc: Lift TAPRIO schedule >> restriction") can detect a false positive error in some corner case. >> For instance, >> tc qdisc replace ... taprio num_tc 4 >> ... >> sched-entry S 0x01 100000 # slot#1 >> sched-entry S 0x03 100000 # slot#2 >> sched-entry S 0x04 100000 # slot#3 >> sched-entry S 0x08 200000 # slot#4 >> flags 0x02 # hardware offload >> >> Here the queue#0 (the first queue) is on at the slot#1 and #2, >> and off at the slot#3 and #4. Under the current logic, when the slot#4 >> is examined, validate_schedule() returns *false* since the enablement >> count for the queue#0 is two and it is already off at the previous slot >> (i.e. #3). But this definition is truely correct. >> >> Let's fix the logic to enforce a strict validation for consecutively-opened >> slots. >> >> Fixes: a5fd39464a40 ("igc: Lift TAPRIO schedule restriction") >> Signed-off-by: AKASHI Takahiro <takahiro.akashi@linaro.org> > > Grml. Thanks! > > Reviewed-by: Kurt Kanzenbach <kurt@linutronix.de> Acked-by: Vinicius Costa Gomes <vinicius.gomes@intel.com> Cheers,
On 3/7/2023 08:45, AKASHI Takahiro wrote: > The check introduced in the commit a5fd39464a40 ("igc: Lift TAPRIO schedule > restriction") can detect a false positive error in some corner case. > For instance, > tc qdisc replace ... taprio num_tc 4 > ... > sched-entry S 0x01 100000 # slot#1 > sched-entry S 0x03 100000 # slot#2 > sched-entry S 0x04 100000 # slot#3 > sched-entry S 0x08 200000 # slot#4 > flags 0x02 # hardware offload > > Here the queue#0 (the first queue) is on at the slot#1 and #2, > and off at the slot#3 and #4. Under the current logic, when the slot#4 > is examined, validate_schedule() returns *false* since the enablement > count for the queue#0 is two and it is already off at the previous slot > (i.e. #3). But this definition is truely correct. > > Let's fix the logic to enforce a strict validation for consecutively-opened > slots. > > Fixes: a5fd39464a40 ("igc: Lift TAPRIO schedule restriction") > Signed-off-by: AKASHI Takahiro <takahiro.akashi@linaro.org> > --- > drivers/net/ethernet/intel/igc/igc_main.c | 20 ++++++++++---------- > 1 file changed, 10 insertions(+), 10 deletions(-) Tested-by: Naama Meir <naamax.meir@linux.intel.com>
diff --git a/drivers/net/ethernet/intel/igc/igc_main.c b/drivers/net/ethernet/intel/igc/igc_main.c index 2928a6c73692..25fc6c65209b 100644 --- a/drivers/net/ethernet/intel/igc/igc_main.c +++ b/drivers/net/ethernet/intel/igc/igc_main.c @@ -6010,18 +6010,18 @@ static bool validate_schedule(struct igc_adapter *adapter, if (e->command != TC_TAPRIO_CMD_SET_GATES) return false; - for (i = 0; i < adapter->num_tx_queues; i++) { - if (e->gate_mask & BIT(i)) + for (i = 0; i < adapter->num_tx_queues; i++) + if (e->gate_mask & BIT(i)) { queue_uses[i]++; - /* 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))) - return false; - } + /* 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))) + return false; + } } return true;
The check introduced in the commit a5fd39464a40 ("igc: Lift TAPRIO schedule restriction") can detect a false positive error in some corner case. For instance, tc qdisc replace ... taprio num_tc 4 ... sched-entry S 0x01 100000 # slot#1 sched-entry S 0x03 100000 # slot#2 sched-entry S 0x04 100000 # slot#3 sched-entry S 0x08 200000 # slot#4 flags 0x02 # hardware offload Here the queue#0 (the first queue) is on at the slot#1 and #2, and off at the slot#3 and #4. Under the current logic, when the slot#4 is examined, validate_schedule() returns *false* since the enablement count for the queue#0 is two and it is already off at the previous slot (i.e. #3). But this definition is truely correct. Let's fix the logic to enforce a strict validation for consecutively-opened slots. Fixes: a5fd39464a40 ("igc: Lift TAPRIO schedule restriction") Signed-off-by: AKASHI Takahiro <takahiro.akashi@linaro.org> --- drivers/net/ethernet/intel/igc/igc_main.c | 20 ++++++++++---------- 1 file changed, 10 insertions(+), 10 deletions(-)