On Mon Feb 24 2025, Loktionov, Aleksandr wrote: >> -----Original Message----- >> From: Intel-wired-lan <[email protected]> On Behalf Of >> Kurt Kanzenbach >> Sent: Monday, February 24, 2025 11:05 AM >> To: Nguyen, Anthony L <[email protected]>; Kitszel, Przemyslaw >> <[email protected]>; Faizal Rahim >> <[email protected]> >> Cc: Andrew Lunn <[email protected]>; David S. Miller >> <[email protected]>; Eric Dumazet <[email protected]>; Jakub >> Kicinski <[email protected]>; Paolo Abeni <[email protected]>; Sebastian >> Andrzej Siewior <[email protected]>; [email protected]; >> [email protected]; Kurt Kanzenbach <[email protected]> >> Subject: [Intel-wired-lan] [PATCH iwl-next v2] igc: Change Tx mode for >> MQPRIO offloading >> >> The current MQPRIO offload implementation uses the legacy TSN Tx mode. In >> this mode the hardware uses four packet buffers and considers queue >> priorities. >> >> In order to harmonize the TAPRIO implementation with MQPRIO, switch to >> the regular TSN Tx mode. In addition to the legacy mode, transmission is >> always coupled to Qbv. The driver already has mechanisms to use a dummy >> schedule of 1 second with all gates open for ETF. Simply use this for MQPRIO >> too. >> >> This reduces code and makes it easier to add support for frame preemption >> later. >> >> While at it limit the netdev_tc calls to MQPRIO only. >> >> Tested on i225 with real time application using high priority queue, iperf3 >> using low priority queue and network TAP device. >> >> Signed-off-by: Kurt Kanzenbach <[email protected]> >> --- >> Changes in v2: >> - Add comma to commit message (Faizal) >> - Simplify if condition (Faizal) >> - Link to v1: https://lore.kernel.org/r/20250217-igc_mqprio_tx_mode-v1-1- >> [email protected] >> --- >> drivers/net/ethernet/intel/igc/igc.h | 4 +--- >> drivers/net/ethernet/intel/igc/igc_main.c | 18 +++++++++++++- >> drivers/net/ethernet/intel/igc/igc_tsn.c | 40 >> ++----------------------------- >> 3 files changed, 20 insertions(+), 42 deletions(-) >> >> diff --git a/drivers/net/ethernet/intel/igc/igc.h >> b/drivers/net/ethernet/intel/igc/igc.h >> index >> cd1d7b6c1782352094f6867a31b6958c929bbbf4..16d85bdf55a7e9c412c4 >> 7acf727bca6bc7154c61 100644 >> --- a/drivers/net/ethernet/intel/igc/igc.h >> +++ b/drivers/net/ethernet/intel/igc/igc.h >> @@ -388,11 +388,9 @@ extern char igc_driver_name[]; >> #define IGC_FLAG_RX_LEGACY BIT(16) >> #define IGC_FLAG_TSN_QBV_ENABLED BIT(17) >> #define IGC_FLAG_TSN_QAV_ENABLED BIT(18) >> -#define IGC_FLAG_TSN_LEGACY_ENABLED BIT(19) >> >> #define IGC_FLAG_TSN_ANY_ENABLED \ >> - (IGC_FLAG_TSN_QBV_ENABLED | IGC_FLAG_TSN_QAV_ENABLED | >> \ >> - IGC_FLAG_TSN_LEGACY_ENABLED) >> + (IGC_FLAG_TSN_QBV_ENABLED | IGC_FLAG_TSN_QAV_ENABLED) >> >> #define IGC_FLAG_RSS_FIELD_IPV4_UDP BIT(6) >> #define IGC_FLAG_RSS_FIELD_IPV6_UDP BIT(7) >> diff --git a/drivers/net/ethernet/intel/igc/igc_main.c >> b/drivers/net/ethernet/intel/igc/igc_main.c >> index >> 3044392e8ded8619434040b9ccaa6b1babdbf685..0f44b0a6c166ae8aa798 >> 93ea87f706be5d94397c 100644 >> --- a/drivers/net/ethernet/intel/igc/igc_main.c >> +++ b/drivers/net/ethernet/intel/igc/igc_main.c >> @@ -6678,13 +6678,14 @@ static int igc_tsn_enable_mqprio(struct >> igc_adapter *adapter, >> struct tc_mqprio_qopt_offload *mqprio) { >> struct igc_hw *hw = &adapter->hw; >> - int i; >> + int err, i; >> >> if (hw->mac.type != igc_i225) >> return -EOPNOTSUPP; >> >> if (!mqprio->qopt.num_tc) { >> adapter->strict_priority_enable = false; >> + netdev_reset_tc(adapter->netdev); >> goto apply; >> } >> >> @@ -6715,6 +6716,21 @@ static int igc_tsn_enable_mqprio(struct >> igc_adapter *adapter, >> igc_save_mqprio_params(adapter, mqprio->qopt.num_tc, >> mqprio->qopt.offset); >> >> + err = netdev_set_num_tc(adapter->netdev, adapter->num_tc); >> + if (err) >> + return err; >> + >> + for (i = 0; i < adapter->num_tc; i++) { >> + err = netdev_set_tc_queue(adapter->netdev, i, 1, >> + adapter->queue_per_tc[i]); >> + if (err) >> + return err; >> + } >> + >> + /* In case the card is configured with less than four queues. */ >> + for (; i < IGC_MAX_TX_QUEUES; i++) >> + adapter->queue_per_tc[i] = i; >> + >> mqprio->qopt.hw = TC_MQPRIO_HW_OFFLOAD_TCS; >> >> apply: >> diff --git a/drivers/net/ethernet/intel/igc/igc_tsn.c >> b/drivers/net/ethernet/intel/igc/igc_tsn.c >> index >> 1e44374ca1ffbb86e9893266c590f318984ef574..7c28f3e7bb576f0e6a21c8 >> 83e934ede4d53096f4 100644 >> --- a/drivers/net/ethernet/intel/igc/igc_tsn.c >> +++ b/drivers/net/ethernet/intel/igc/igc_tsn.c >> @@ -37,18 +37,13 @@ static unsigned int igc_tsn_new_flags(struct >> igc_adapter *adapter) { >> unsigned int new_flags = adapter->flags & >> ~IGC_FLAG_TSN_ANY_ENABLED; >> >> - if (adapter->taprio_offload_enable) >> - new_flags |= IGC_FLAG_TSN_QBV_ENABLED; >> - >> - if (is_any_launchtime(adapter)) >> + if (adapter->taprio_offload_enable || is_any_launchtime(adapter) || >> + adapter->strict_priority_enable) > Isn't sequence of: > if (adapter->taprio_offload_enable || adapter->strict_priority_enable || > is_any_launchtime(adapter)) > faster statistically?
I don't think it's faster, because it is unlikely that strict_priority_enable is true. Most people do use taprio or etf. Thanks.
signature.asc
Description: PGP signature
