drivers/net/wireless/mediatek/mt76/mt76_connac_mac.c | 5 +++++ drivers/net/wireless/mediatek/mt76/mt7915/init.c | 3 ++- drivers/net/wireless/mediatek/mt76/mt792x_core.c | 3 +++ 3 files changed, 10 insertions(+), 1 deletion(-)
Supports IPv4 and IPv6 TCP + UDP
In various tests between MT7986 and Intel BE200, I observed a performance
boost ranging from 2 to 12%, with an average of 5.5%.
I did the tests on the MT7915, MT7981, MT7986, and MT7921 variants. The
MT7922, MT7925, and MT799x are untouched for now and still have
checksumming disabled.
Signed-off-by: Felix Fietkau <nbd@nbd.name>
[rebased and resolved conflicts]
Signed-off-by: Aleksander Jan Bajkowski <olek2@wp.pl>
---
drivers/net/wireless/mediatek/mt76/mt76_connac_mac.c | 5 +++++
drivers/net/wireless/mediatek/mt76/mt7915/init.c | 3 ++-
drivers/net/wireless/mediatek/mt76/mt792x_core.c | 3 +++
3 files changed, 10 insertions(+), 1 deletion(-)
diff --git a/drivers/net/wireless/mediatek/mt76/mt76_connac_mac.c b/drivers/net/wireless/mediatek/mt76/mt76_connac_mac.c
index 0db00efe88b0..e31b3e7e2038 100644
--- a/drivers/net/wireless/mediatek/mt76/mt76_connac_mac.c
+++ b/drivers/net/wireless/mediatek/mt76/mt76_connac_mac.c
@@ -372,6 +372,9 @@ mt76_connac2_mac_write_txwi_8023(__le32 *txwi, struct sk_buff *skb,
wmm = sta->wme;
}
+ val = FIELD_PREP(MT_TXD0_ETH_TYPE_OFFSET, 10);
+ txwi[0] |= cpu_to_le32(val);
+
val = FIELD_PREP(MT_TXD1_HDR_FORMAT, MT_HDR_FORMAT_802_3) |
FIELD_PREP(MT_TXD1_TID, tid);
@@ -391,6 +394,8 @@ mt76_connac2_mac_write_txwi_8023(__le32 *txwi, struct sk_buff *skb,
val = FIELD_PREP(MT_TXD7_TYPE, fc_type) |
FIELD_PREP(MT_TXD7_SUB_TYPE, fc_stype);
+ if (skb->ip_summed == CHECKSUM_PARTIAL)
+ val |= MT_TXD7_IP_SUM | MT_TXD7_UDP_TCP_SUM;
txwi[7] |= cpu_to_le32(val);
}
diff --git a/drivers/net/wireless/mediatek/mt76/mt7915/init.c b/drivers/net/wireless/mediatek/mt76/mt7915/init.c
index 5ea8b46e092e..a3bab240afb7 100644
--- a/drivers/net/wireless/mediatek/mt76/mt7915/init.c
+++ b/drivers/net/wireless/mediatek/mt76/mt7915/init.c
@@ -359,7 +359,8 @@ mt7915_init_wiphy(struct mt7915_phy *phy)
hw->queues = 4;
hw->max_rx_aggregation_subframes = IEEE80211_MAX_AMPDU_BUF_HE;
hw->max_tx_aggregation_subframes = IEEE80211_MAX_AMPDU_BUF_HE;
- hw->netdev_features = NETIF_F_RXCSUM;
+ hw->netdev_features = NETIF_F_RXCSUM |
+ NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM;
if (mtk_wed_device_active(&mdev->mmio.wed))
hw->netdev_features |= NETIF_F_HW_TC;
diff --git a/drivers/net/wireless/mediatek/mt76/mt792x_core.c b/drivers/net/wireless/mediatek/mt76/mt792x_core.c
index c0e56541a954..fe138d2e8e62 100644
--- a/drivers/net/wireless/mediatek/mt76/mt792x_core.c
+++ b/drivers/net/wireless/mediatek/mt76/mt792x_core.c
@@ -632,6 +632,9 @@ int mt792x_init_wiphy(struct ieee80211_hw *hw)
}
hw->netdev_features = NETIF_F_RXCSUM;
+ if (is_mt7921(&dev->mt76))
+ NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM;
+
hw->radiotap_timestamp.units_pos =
IEEE80211_RADIOTAP_TIMESTAMP_UNIT_US;
--
2.47.3Hi Aleksander,
kernel test robot noticed the following build warnings:
[auto build test WARNING on wireless-next/main]
[also build test WARNING on wireless/main linus/master v6.18-rc2 next-20251017]
[If your patch is applied to the wrong git tree, kindly drop us a note.
And when submitting patch, we suggest to use '--base' as documented in
https://git-scm.com/docs/git-format-patch#_base_tree_information]
url: https://github.com/intel-lab-lkp/linux/commits/Aleksander-Jan-Bajkowski/wifi-mt76-add-tx-checksum-offload-for-mt7915-mt7921-mt7981-and-mt7986/20251019-235515
base: https://git.kernel.org/pub/scm/linux/kernel/git/wireless/wireless-next.git main
patch link: https://lore.kernel.org/r/20251019155316.3537185-1-olek2%40wp.pl
patch subject: [PATCH] wifi: mt76: add tx checksum offload for mt7915, mt7921, mt7981 and mt7986
config: loongarch-allyesconfig (https://download.01.org/0day-ci/archive/20251020/202510201250.Hpsrohca-lkp@intel.com/config)
compiler: clang version 22.0.0git (https://github.com/llvm/llvm-project 754ebc6ebb9fb9fbee7aef33478c74ea74949853)
reproduce (this is a W=1 build): (https://download.01.org/0day-ci/archive/20251020/202510201250.Hpsrohca-lkp@intel.com/reproduce)
If you fix the issue in a separate patch/commit (i.e. not just a new version of
the same patch/commit), kindly add following tags
| Reported-by: kernel test robot <lkp@intel.com>
| Closes: https://lore.kernel.org/oe-kbuild-all/202510201250.Hpsrohca-lkp@intel.com/
All warnings (new ones prefixed by >>):
>> drivers/net/wireless/mediatek/mt76/mt792x_core.c:636:19: warning: expression result unused [-Wunused-value]
636 | NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM;
| ~~~~~~~~~~~~~~~ ^ ~~~~~~~~~~~~~~~~~
1 warning generated.
vim +636 drivers/net/wireless/mediatek/mt76/mt792x_core.c
618
619 int mt792x_init_wiphy(struct ieee80211_hw *hw)
620 {
621 struct mt792x_phy *phy = mt792x_hw_phy(hw);
622 struct mt792x_dev *dev = phy->dev;
623 struct wiphy *wiphy = hw->wiphy;
624
625 hw->queues = 4;
626 if (dev->has_eht) {
627 hw->max_rx_aggregation_subframes = IEEE80211_MAX_AMPDU_BUF_EHT;
628 hw->max_tx_aggregation_subframes = IEEE80211_MAX_AMPDU_BUF_EHT;
629 } else {
630 hw->max_rx_aggregation_subframes = IEEE80211_MAX_AMPDU_BUF_HE;
631 hw->max_tx_aggregation_subframes = IEEE80211_MAX_AMPDU_BUF_HE;
632 }
633 hw->netdev_features = NETIF_F_RXCSUM;
634
635 if (is_mt7921(&dev->mt76))
> 636 NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM;
637
638 hw->radiotap_timestamp.units_pos =
639 IEEE80211_RADIOTAP_TIMESTAMP_UNIT_US;
640
641 phy->slottime = 9;
642
643 hw->sta_data_size = sizeof(struct mt792x_sta);
644 hw->vif_data_size = sizeof(struct mt792x_vif);
645 hw->chanctx_data_size = sizeof(struct mt792x_chanctx);
646
647 if (dev->fw_features & MT792x_FW_CAP_CNM) {
648 wiphy->flags |= WIPHY_FLAG_HAS_REMAIN_ON_CHANNEL;
649 wiphy->iface_combinations = if_comb_chanctx;
650 wiphy->n_iface_combinations = ARRAY_SIZE(if_comb_chanctx);
651 } else {
652 wiphy->flags &= ~WIPHY_FLAG_HAS_REMAIN_ON_CHANNEL;
653 wiphy->iface_combinations = if_comb;
654 wiphy->n_iface_combinations = ARRAY_SIZE(if_comb);
655 }
656 wiphy->flags &= ~(WIPHY_FLAG_IBSS_RSN | WIPHY_FLAG_4ADDR_AP |
657 WIPHY_FLAG_4ADDR_STATION);
658 wiphy->interface_modes = BIT(NL80211_IFTYPE_STATION) |
659 BIT(NL80211_IFTYPE_AP) |
660 BIT(NL80211_IFTYPE_P2P_CLIENT) |
661 BIT(NL80211_IFTYPE_P2P_GO) |
662 BIT(NL80211_IFTYPE_P2P_DEVICE);
663 wiphy->max_remain_on_channel_duration = 5000;
664 wiphy->max_scan_ie_len = MT76_CONNAC_SCAN_IE_LEN;
665 wiphy->max_scan_ssids = 4;
666 wiphy->max_sched_scan_plan_interval =
667 MT76_CONNAC_MAX_TIME_SCHED_SCAN_INTERVAL;
668 wiphy->max_sched_scan_ie_len = IEEE80211_MAX_DATA_LEN;
669 wiphy->max_sched_scan_ssids = MT76_CONNAC_MAX_SCHED_SCAN_SSID;
670 wiphy->max_match_sets = MT76_CONNAC_MAX_SCAN_MATCH;
671 wiphy->max_sched_scan_reqs = 1;
672 wiphy->flags |= WIPHY_FLAG_HAS_CHANNEL_SWITCH |
673 WIPHY_FLAG_SPLIT_SCAN_6GHZ;
674
675 wiphy->features |= NL80211_FEATURE_SCHED_SCAN_RANDOM_MAC_ADDR |
676 NL80211_FEATURE_SCAN_RANDOM_MAC_ADDR;
677 wiphy_ext_feature_set(wiphy, NL80211_EXT_FEATURE_SET_SCAN_DWELL);
678 wiphy_ext_feature_set(wiphy, NL80211_EXT_FEATURE_BEACON_RATE_LEGACY);
679 wiphy_ext_feature_set(wiphy, NL80211_EXT_FEATURE_BEACON_RATE_HT);
680 wiphy_ext_feature_set(wiphy, NL80211_EXT_FEATURE_BEACON_RATE_VHT);
681 wiphy_ext_feature_set(wiphy, NL80211_EXT_FEATURE_BEACON_RATE_HE);
682 wiphy_ext_feature_set(wiphy, NL80211_EXT_FEATURE_ACK_SIGNAL_SUPPORT);
683 wiphy_ext_feature_set(wiphy, NL80211_EXT_FEATURE_CAN_REPLACE_PTK0);
684
685 ieee80211_hw_set(hw, SINGLE_SCAN_ON_ALL_BANDS);
686 ieee80211_hw_set(hw, HAS_RATE_CONTROL);
687 ieee80211_hw_set(hw, SUPPORTS_TX_ENCAP_OFFLOAD);
688 ieee80211_hw_set(hw, SUPPORTS_RX_DECAP_OFFLOAD);
689 ieee80211_hw_set(hw, WANT_MONITOR_VIF);
690 ieee80211_hw_set(hw, SUPPORTS_PS);
691 ieee80211_hw_set(hw, SUPPORTS_DYNAMIC_PS);
692 ieee80211_hw_set(hw, SUPPORTS_VHT_EXT_NSS_BW);
693 ieee80211_hw_set(hw, CONNECTION_MONITOR);
694 ieee80211_hw_set(hw, NO_VIRTUAL_MONITOR);
695 ieee80211_hw_set(hw, SUPPORTS_MULTI_BSSID);
696 ieee80211_hw_set(hw, SUPPORTS_ONLY_HE_MULTI_BSSID);
697
698 if (is_mt7921(&dev->mt76)) {
699 ieee80211_hw_set(hw, CHANCTX_STA_CSA);
700 }
701
702 if (dev->pm.enable)
703 ieee80211_hw_set(hw, CONNECTION_MONITOR);
704
705 hw->max_tx_fragments = 4;
706
707 return 0;
708 }
709 EXPORT_SYMBOL_GPL(mt792x_init_wiphy);
710
--
0-DAY CI Kernel Test Service
https://github.com/intel/lkp-tests/wiki
Hi Aleksander,
kernel test robot noticed the following build warnings:
[auto build test WARNING on wireless-next/main]
[also build test WARNING on wireless/main linus/master v6.18-rc1 next-20251017]
[If your patch is applied to the wrong git tree, kindly drop us a note.
And when submitting patch, we suggest to use '--base' as documented in
https://git-scm.com/docs/git-format-patch#_base_tree_information]
url: https://github.com/intel-lab-lkp/linux/commits/Aleksander-Jan-Bajkowski/wifi-mt76-add-tx-checksum-offload-for-mt7915-mt7921-mt7981-and-mt7986/20251019-235515
base: https://git.kernel.org/pub/scm/linux/kernel/git/wireless/wireless-next.git main
patch link: https://lore.kernel.org/r/20251019155316.3537185-1-olek2%40wp.pl
patch subject: [PATCH] wifi: mt76: add tx checksum offload for mt7915, mt7921, mt7981 and mt7986
config: i386-buildonly-randconfig-003-20251019 (https://download.01.org/0day-ci/archive/20251020/202510200105.T000lujD-lkp@intel.com/config)
compiler: gcc-14 (Debian 14.2.0-19) 14.2.0
reproduce (this is a W=1 build): (https://download.01.org/0day-ci/archive/20251020/202510200105.T000lujD-lkp@intel.com/reproduce)
If you fix the issue in a separate patch/commit (i.e. not just a new version of
the same patch/commit), kindly add following tags
| Reported-by: kernel test robot <lkp@intel.com>
| Closes: https://lore.kernel.org/oe-kbuild-all/202510200105.T000lujD-lkp@intel.com/
All warnings (new ones prefixed by >>):
In file included from include/linux/skbuff.h:29,
from drivers/net/wireless/mediatek/mt76/mt76.h:12,
from drivers/net/wireless/mediatek/mt76/mt76_connac.h:7,
from drivers/net/wireless/mediatek/mt76/mt76_connac_mcu.h:7,
from drivers/net/wireless/mediatek/mt76/mt792x.h:10,
from drivers/net/wireless/mediatek/mt76/mt792x_core.c:7:
drivers/net/wireless/mediatek/mt76/mt792x_core.c: In function 'mt792x_init_wiphy':
>> include/linux/netdev_features.h:104:33: warning: statement with no effect [-Wunused-value]
104 | #define __NETIF_F_BIT(bit) ((netdev_features_t)1 << (bit))
| ^
include/linux/netdev_features.h:105:33: note: in expansion of macro '__NETIF_F_BIT'
105 | #define __NETIF_F(name) __NETIF_F_BIT(NETIF_F_##name##_BIT)
| ^~~~~~~~~~~~~
include/linux/netdev_features.h:119:33: note: in expansion of macro '__NETIF_F'
119 | #define NETIF_F_IP_CSUM __NETIF_F(IP_CSUM)
| ^~~~~~~~~
drivers/net/wireless/mediatek/mt76/mt792x_core.c:636:17: note: in expansion of macro 'NETIF_F_IP_CSUM'
636 | NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM;
| ^~~~~~~~~~~~~~~
vim +104 include/linux/netdev_features.h
a19f2a6df28e0c Michał Mirosław 2011-11-15 102
a19f2a6df28e0c Michał Mirosław 2011-11-15 103 /* copy'n'paste compression ;) */
a19f2a6df28e0c Michał Mirosław 2011-11-15 @104 #define __NETIF_F_BIT(bit) ((netdev_features_t)1 << (bit))
a19f2a6df28e0c Michał Mirosław 2011-11-15 105 #define __NETIF_F(name) __NETIF_F_BIT(NETIF_F_##name##_BIT)
a19f2a6df28e0c Michał Mirosław 2011-11-15 106
--
0-DAY CI Kernel Test Service
https://github.com/intel/lkp-tests/wiki
On 10/19/25 08:51, Aleksander Jan Bajkowski wrote: > Supports IPv4 and IPv6 TCP + UDP > > In various tests between MT7986 and Intel BE200, I observed a performance > boost ranging from 2 to 12%, with an average of 5.5%. > > I did the tests on the MT7915, MT7981, MT7986, and MT7921 variants. The > MT7922, MT7925, and MT799x are untouched for now and still have > checksumming disabled. At least with 7996, tcp csum only worked on the first few vdevs created in our testing, so we had to add logic to disable that flag on subsequent vdevs. Have you tried creating a bunch of station and/or vap vdevs to see if all of them can still transmit TCP traffic? Thanks, Ben > > Signed-off-by: Felix Fietkau <nbd@nbd.name> > [rebased and resolved conflicts] > Signed-off-by: Aleksander Jan Bajkowski <olek2@wp.pl> > --- > drivers/net/wireless/mediatek/mt76/mt76_connac_mac.c | 5 +++++ > drivers/net/wireless/mediatek/mt76/mt7915/init.c | 3 ++- > drivers/net/wireless/mediatek/mt76/mt792x_core.c | 3 +++ > 3 files changed, 10 insertions(+), 1 deletion(-) > > diff --git a/drivers/net/wireless/mediatek/mt76/mt76_connac_mac.c b/drivers/net/wireless/mediatek/mt76/mt76_connac_mac.c > index 0db00efe88b0..e31b3e7e2038 100644 > --- a/drivers/net/wireless/mediatek/mt76/mt76_connac_mac.c > +++ b/drivers/net/wireless/mediatek/mt76/mt76_connac_mac.c > @@ -372,6 +372,9 @@ mt76_connac2_mac_write_txwi_8023(__le32 *txwi, struct sk_buff *skb, > wmm = sta->wme; > } > > + val = FIELD_PREP(MT_TXD0_ETH_TYPE_OFFSET, 10); > + txwi[0] |= cpu_to_le32(val); > + > val = FIELD_PREP(MT_TXD1_HDR_FORMAT, MT_HDR_FORMAT_802_3) | > FIELD_PREP(MT_TXD1_TID, tid); > > @@ -391,6 +394,8 @@ mt76_connac2_mac_write_txwi_8023(__le32 *txwi, struct sk_buff *skb, > > val = FIELD_PREP(MT_TXD7_TYPE, fc_type) | > FIELD_PREP(MT_TXD7_SUB_TYPE, fc_stype); > + if (skb->ip_summed == CHECKSUM_PARTIAL) > + val |= MT_TXD7_IP_SUM | MT_TXD7_UDP_TCP_SUM; > > txwi[7] |= cpu_to_le32(val); > } > diff --git a/drivers/net/wireless/mediatek/mt76/mt7915/init.c b/drivers/net/wireless/mediatek/mt76/mt7915/init.c > index 5ea8b46e092e..a3bab240afb7 100644 > --- a/drivers/net/wireless/mediatek/mt76/mt7915/init.c > +++ b/drivers/net/wireless/mediatek/mt76/mt7915/init.c > @@ -359,7 +359,8 @@ mt7915_init_wiphy(struct mt7915_phy *phy) > hw->queues = 4; > hw->max_rx_aggregation_subframes = IEEE80211_MAX_AMPDU_BUF_HE; > hw->max_tx_aggregation_subframes = IEEE80211_MAX_AMPDU_BUF_HE; > - hw->netdev_features = NETIF_F_RXCSUM; > + hw->netdev_features = NETIF_F_RXCSUM | > + NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM; > > if (mtk_wed_device_active(&mdev->mmio.wed)) > hw->netdev_features |= NETIF_F_HW_TC; > diff --git a/drivers/net/wireless/mediatek/mt76/mt792x_core.c b/drivers/net/wireless/mediatek/mt76/mt792x_core.c > index c0e56541a954..fe138d2e8e62 100644 > --- a/drivers/net/wireless/mediatek/mt76/mt792x_core.c > +++ b/drivers/net/wireless/mediatek/mt76/mt792x_core.c > @@ -632,6 +632,9 @@ int mt792x_init_wiphy(struct ieee80211_hw *hw) > } > hw->netdev_features = NETIF_F_RXCSUM; > > + if (is_mt7921(&dev->mt76)) > + NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM; > + > hw->radiotap_timestamp.units_pos = > IEEE80211_RADIOTAP_TIMESTAMP_UNIT_US; > -- Ben Greear <greearb@candelatech.com> Candela Technologies Inc http://www.candelatech.com
Hi Ben, On 10/19/25 18:26, Ben Greear wrote: > On 10/19/25 08:51, Aleksander Jan Bajkowski wrote: >> Supports IPv4 and IPv6 TCP + UDP >> >> In various tests between MT7986 and Intel BE200, I observed a >> performance >> boost ranging from 2 to 12%, with an average of 5.5%. >> >> I did the tests on the MT7915, MT7981, MT7986, and MT7921 variants. The >> MT7922, MT7925, and MT799x are untouched for now and still have >> checksumming disabled. > > At least with 7996, tcp csum only worked on the first few vdevs > created in our testing, so we had to add logic to disable that flag > on subsequent vdevs. > > Have you tried creating a bunch of station and/or vap vdevs to see if > all of them can still transmit TCP traffic? > > Thanks for the useful information. On all tested devices, I had a single AP configured per device. I will try to create several APs for each device. I also have a router with MT7996. A quick test shows that checksum offload doesn't work on this router. MT7996 is visible in the system as a single DBDC device. I have 3 APs configured there. Each on a separate band (2.4/5/6 GHz). Best regards, Aleksander
On 10/19/25 09:43, Aleksander Jan Bajkowski wrote: > Hi Ben, > > On 10/19/25 18:26, Ben Greear wrote: >> On 10/19/25 08:51, Aleksander Jan Bajkowski wrote: >>> Supports IPv4 and IPv6 TCP + UDP >>> >>> In various tests between MT7986 and Intel BE200, I observed a performance >>> boost ranging from 2 to 12%, with an average of 5.5%. >>> >>> I did the tests on the MT7915, MT7981, MT7986, and MT7921 variants. The >>> MT7922, MT7925, and MT799x are untouched for now and still have >>> checksumming disabled. >> >> At least with 7996, tcp csum only worked on the first few vdevs >> created in our testing, so we had to add logic to disable that flag >> on subsequent vdevs. >> >> Have you tried creating a bunch of station and/or vap vdevs to see if >> all of them can still transmit TCP traffic? >> >> > Thanks for the useful information. On all tested devices, I had a single > AP configured per device. I will try to create several APs for each device. We can try it out as well. > I also have a router with MT7996. A quick test shows that checksum > offload doesn't work on this router. MT7996 is visible in the system > as a single DBDC device. I have 3 APs configured there. Each on a > separate band (2.4/5/6 GHz). > > Best regards, > Aleksander We have it working on 7996, but the patch is tangled in some other changes we made with how wcid and such are allocated. We are still cleaning and testing patches on top of Felix's tree, should have something to post sometime soon. Thanks, Ben -- Ben Greear <greearb@candelatech.com> Candela Technologies Inc http://www.candelatech.com
On 10/19/25 20:18, Ben Greear wrote: > On 10/19/25 09:43, Aleksander Jan Bajkowski wrote: >> Hi Ben, >> >> On 10/19/25 18:26, Ben Greear wrote: >>> On 10/19/25 08:51, Aleksander Jan Bajkowski wrote: >>>> Supports IPv4 and IPv6 TCP + UDP >>>> >>>> In various tests between MT7986 and Intel BE200, I observed a >>>> performance >>>> boost ranging from 2 to 12%, with an average of 5.5%. >>>> >>>> I did the tests on the MT7915, MT7981, MT7986, and MT7921 variants. >>>> The >>>> MT7922, MT7925, and MT799x are untouched for now and still have >>>> checksumming disabled. >>> >>> At least with 7996, tcp csum only worked on the first few vdevs >>> created in our testing, so we had to add logic to disable that flag >>> on subsequent vdevs. >>> >>> Have you tried creating a bunch of station and/or vap vdevs to see if >>> all of them can still transmit TCP traffic? >>> >>> >> Thanks for the useful information. On all tested devices, I had a single >> AP configured per device. I will try to create several APs for each >> device. > > We can try it out as well. I've tested multiple VAPs (15x AP + 1x station). It seems that everything works fine on MT7915, MT7981, and MT7986. By connecting to the first and last AP, tx checksum offload works correctly. MT7921 only supports a single VAP, so I skipped testing it. > >> I also have a router with MT7996. A quick test shows that checksum >> offload doesn't work on this router. MT7996 is visible in the system >> as a single DBDC device. I have 3 APs configured there. Each on a >> separate band (2.4/5/6 GHz). >> >> Best regards, >> Aleksander > > We have it working on 7996, but the patch is tangled in some other > changes we made > with how wcid and such are allocated. We are still cleaning and > testing patches > on top of Felix's tree, should have something to post sometime soon. > Maybe only the MT7996 is affected?
Le 19/10/2025 à 17:51, Aleksander Jan Bajkowski a écrit : > Supports IPv4 and IPv6 TCP + UDP > > In various tests between MT7986 and Intel BE200, I observed a performance > boost ranging from 2 to 12%, with an average of 5.5%. > > I did the tests on the MT7915, MT7981, MT7986, and MT7921 variants. The > MT7922, MT7925, and MT799x are untouched for now and still have > checksumming disabled. > > Signed-off-by: Felix Fietkau <nbd@nbd.name> > [rebased and resolved conflicts] > Signed-off-by: Aleksander Jan Bajkowski <olek2@wp.pl> > --- > drivers/net/wireless/mediatek/mt76/mt76_connac_mac.c | 5 +++++ > drivers/net/wireless/mediatek/mt76/mt7915/init.c | 3 ++- > drivers/net/wireless/mediatek/mt76/mt792x_core.c | 3 +++ > 3 files changed, 10 insertions(+), 1 deletion(-) > > diff --git a/drivers/net/wireless/mediatek/mt76/mt76_connac_mac.c b/drivers/net/wireless/mediatek/mt76/mt76_connac_mac.c > index 0db00efe88b0..e31b3e7e2038 100644 > --- a/drivers/net/wireless/mediatek/mt76/mt76_connac_mac.c > +++ b/drivers/net/wireless/mediatek/mt76/mt76_connac_mac.c > @@ -372,6 +372,9 @@ mt76_connac2_mac_write_txwi_8023(__le32 *txwi, struct sk_buff *skb, > wmm = sta->wme; > } > > + val = FIELD_PREP(MT_TXD0_ETH_TYPE_OFFSET, 10); > + txwi[0] |= cpu_to_le32(val); > + > val = FIELD_PREP(MT_TXD1_HDR_FORMAT, MT_HDR_FORMAT_802_3) | > FIELD_PREP(MT_TXD1_TID, tid); > > @@ -391,6 +394,8 @@ mt76_connac2_mac_write_txwi_8023(__le32 *txwi, struct sk_buff *skb, > > val = FIELD_PREP(MT_TXD7_TYPE, fc_type) | > FIELD_PREP(MT_TXD7_SUB_TYPE, fc_stype); > + if (skb->ip_summed == CHECKSUM_PARTIAL) > + val |= MT_TXD7_IP_SUM | MT_TXD7_UDP_TCP_SUM; > > txwi[7] |= cpu_to_le32(val); > } > diff --git a/drivers/net/wireless/mediatek/mt76/mt7915/init.c b/drivers/net/wireless/mediatek/mt76/mt7915/init.c > index 5ea8b46e092e..a3bab240afb7 100644 > --- a/drivers/net/wireless/mediatek/mt76/mt7915/init.c > +++ b/drivers/net/wireless/mediatek/mt76/mt7915/init.c > @@ -359,7 +359,8 @@ mt7915_init_wiphy(struct mt7915_phy *phy) > hw->queues = 4; > hw->max_rx_aggregation_subframes = IEEE80211_MAX_AMPDU_BUF_HE; > hw->max_tx_aggregation_subframes = IEEE80211_MAX_AMPDU_BUF_HE; > - hw->netdev_features = NETIF_F_RXCSUM; > + hw->netdev_features = NETIF_F_RXCSUM | > + NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM; > > if (mtk_wed_device_active(&mdev->mmio.wed)) > hw->netdev_features |= NETIF_F_HW_TC; > diff --git a/drivers/net/wireless/mediatek/mt76/mt792x_core.c b/drivers/net/wireless/mediatek/mt76/mt792x_core.c > index c0e56541a954..fe138d2e8e62 100644 > --- a/drivers/net/wireless/mediatek/mt76/mt792x_core.c > +++ b/drivers/net/wireless/mediatek/mt76/mt792x_core.c > @@ -632,6 +632,9 @@ int mt792x_init_wiphy(struct ieee80211_hw *hw) > } > hw->netdev_features = NETIF_F_RXCSUM; > > + if (is_mt7921(&dev->mt76)) > + NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM; This looks strange. Maybe: netdev_features |= NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM; ? CJ > + > hw->radiotap_timestamp.units_pos = > IEEE80211_RADIOTAP_TIMESTAMP_UNIT_US; >
© 2016 - 2026 Red Hat, Inc.