OK, then it makes sense if the return is removed and most likely it is correct now. Probably this bug was the reason why someone set the fixed freqs 
I am in favour of just using performance governor and setting the NSS cores at max.
I cannot say for sure about the NSS cores but it seems that the difference in temp and power draw between performance and on demand is very little in normal use
How do we use the LUCI commits?
I've just taken wireless.js and zones.js files from your commits and added them via /files in my build.
But I get this error on wireless and firewall pages.
Update - I messed the file permissions. Corrected them and everything is OK.
Another thing I see is that a compiled build from your repo (using my own settings) I see almost 100 MB less free memory compared to my builds. The last one I've tried was from @asvio repo and my previous builds were based on @ACwifidude repo. What is using up that memory?
Just go into your feeds luci dir, add my repo as remote and cherry pick the commits.
You probably edited the files or somehow changed the rights on the files.
Regarding memory, if you look into /etc/init.d/bootcount you will see a few sysctl settings, that will cause extra mem usage e.g. vm.min_free_kbytes, you can tune them if you like.
2 Likes
Hi @Kong,
Can you revert the change suggested by @regae and try this official fix: https://git.kernel.org/pub/scm/linux/kernel/git/wireless/wireless-next.git/commit/?id=a8bca3e9371dc5e276af4168be099b2a05554c2a
2 Likes
Now throughput is even better than k 5.15.161.
@KONG better update your luci repo, those two commits seems to fix load average readings. Thank you for your work.
1 Like
I flashed my Linksys EA8500 from 23.5.4 Stable to Kong-NSS, basically to fix the TX wifi issue.
But unlike my gateway Netgear XR500 that has a Load Average of 0.40, the EA8500 - as an AP at idle - has a Load Average of 1.60.
Temperatures are ok, but LdAve remains consistantly very high.
EDIT - just read post above!
Looks like calculation of LA is wrong at present.
The NSS build has been updated. If you want to rebuild, you have to drop the 2 patches from https://github.com/openwrt/openwrt/pull/16007 into package/kernel/mac80211/patches/subsys/
5 Likes
I did that already here (screenshot above), results are better than regae's suggested change, but I also included this commit from morse-sophronia's pull
1 Like
Hi Kong,
I saw these dmesg logs when running an image built from your repo:
6,637,45516174,-;ipq8064-mdio 37000000.mdio eth0: nss_gmac_get_strset_count: Invalid string set
6,638,45516230,-;ipq8064-mdio 37000000.mdio eth1: nss_gmac_get_strset_count: Invalid string set
6,641,45609160,-;ipq8064-mdio 37000000.mdio eth0: nss_gmac_get_strset_count: Invalid string set
6,642,45700047,-;ipq8064-mdio 37000000.mdio eth1: nss_gmac_get_strset_count: Invalid string set
I did a fresh pull of your repo and saw you reverted the earlier commit.
I added both patches from: https://github.com/openwrt/openwrt/pull/16007 and can confirm it is working wireless throughput on 2.4 GHz is 106 Mb/s as expected (MCS15)
1 Like
I do not have it (I think) see my ipq8064-mdio:
root@R7800-1:~# logread | grep ipq8064-mdio
Fri Jul 26 15:12:23 2024 kern.info kernel: [ 5.432695] ipq8064-mdio 37000000.mdio (unnamed net_device) (uninitialized): nss_gmac_ipc_offload_init: enable Rx checksum
Fri Jul 26 15:12:23 2024 kern.info kernel: [ 5.435113] ipq8064-mdio 37000000.mdio eth0: Initialized NSS GMAC1 interface eth0: (base = 0x37200000, irq = 39, PhyId = 4, PollLink = 0)
Fri Jul 26 15:12:23 2024 kern.info kernel: [ 5.631849] ipq8064-mdio 37000000.mdio (unnamed net_device) (uninitialized): nss_gmac_ipc_offload_init: enable Rx checksum
Fri Jul 26 15:12:23 2024 kern.info kernel: [ 5.633380] ipq8064-mdio 37000000.mdio eth1: Initialized NSS GMAC2 interface eth1: (base = 0x37400000, irq = 40, PhyId = 0, PollLink = 0)
Fri Jul 26 15:12:23 2024 kern.info kernel: [ 8.360840] ipq8064-mdio 37000000.mdio eth1: nss_gmac_start_up: Force link up
Fri Jul 26 15:12:23 2024 kern.info kernel: [ 8.360879] ipq8064-mdio 37000000.mdio eth1: 1000 Mbps Full Duplex
Fri Jul 26 15:12:23 2024 kern.info kernel: [ 8.366960] ipq8064-mdio 37000000.mdio eth1: nss_gmac_dev_set_speed:NSS_ETH_CLK_DIV0(0xc) - 0x100
Fri Jul 26 15:12:23 2024 kern.info kernel: [ 8.373098] ipq8064-mdio 37000000.mdio eth1: nss_gmac_dev_set_speed: qsgmii_base(0xde8a0000) + PCS_MODE_CTL(0x68): 0x80008082
Fri Jul 26 15:12:23 2024 kern.info kernel: [ 8.382031] ipq8064-mdio 37000000.mdio eth1: nss_gmac_reset: eth1 resetting...
Fri Jul 26 15:12:23 2024 kern.info kernel: [ 9.440568] ipq8064-mdio 37000000.mdio eth1: GMAC reset completed in 105 jiffies; dma_bus_mode - 0x20100
Fri Jul 26 15:12:23 2024 kern.info kernel: [ 9.440847] ipq8064-mdio 37000000.mdio eth1: nss_gmac_change_mtu: Enabling Normal Frame MTU (Requested MTU [1500])
Fri Jul 26 15:12:23 2024 kern.info kernel: [ 9.449188] ipq8064-mdio 37000000.mdio eth1: nss_gmac_change_mtu: Enabling Jumbo Frame MTU for eth1 (Requested MTU [1500])
Fri Jul 26 15:12:23 2024 kern.info kernel: [ 9.459341] ipq8064-mdio 37000000.mdio eth1: Set addr 14:59:c0:5a:43:9d
Fri Jul 26 15:12:23 2024 kern.info kernel: [ 9.470368] ipq8064-mdio 37000000.mdio eth1: nss_gmac_ipc_offload_init: enable Rx checksum
Fri Jul 26 15:12:23 2024 kern.info kernel: [ 9.476807] ipq8064-mdio 37000000.mdio eth1: nss_gmac_config_flow_control:
Fri Jul 26 15:12:23 2024 kern.info kernel: [ 9.485206] ipq8064-mdio 37000000.mdio eth1: nss_gmac_rx_pause_disable: disable Rx flow control
Fri Jul 26 15:12:23 2024 kern.info kernel: [ 9.492047] ipq8064-mdio 37000000.mdio eth1: nss_gmac_tx_pause_disable: disable Tx flow control
Fri Jul 26 15:12:23 2024 kern.info kernel: [ 9.501138] ipq8064-mdio 37000000.mdio eth1: nss_gmac_linkup: data plane open command successfully issued
Fri Jul 26 15:12:29 2024 kern.info kernel: [ 35.364805] ipq8064-mdio 37000000.mdio eth1: nss_gmac_tx_pause_disable: disable Tx flow control
Fri Jul 26 15:12:29 2024 kern.info kernel: [ 35.365084] ipq8064-mdio 37000000.mdio eth1: nss_gmac_rx_pause_disable: disable Rx flow control
Fri Jul 26 15:12:29 2024 kern.info kernel: [ 35.372824] ipq8064-mdio 37000000.mdio eth1: nss_gmac_start_up: Force link up
Fri Jul 26 15:12:29 2024 kern.info kernel: [ 35.381099] ipq8064-mdio 37000000.mdio eth1: 1000 Mbps Full Duplex
Fri Jul 26 15:12:29 2024 kern.info kernel: [ 35.388359] ipq8064-mdio 37000000.mdio eth1: nss_gmac_dev_set_speed:NSS_ETH_CLK_DIV0(0xc) - 0x100
Fri Jul 26 15:12:29 2024 kern.info kernel: [ 35.394365] ipq8064-mdio 37000000.mdio eth1: nss_gmac_dev_set_speed: qsgmii_base(0xde8a0000) + PCS_MODE_CTL(0x68): 0x80008082
Fri Jul 26 15:12:29 2024 kern.info kernel: [ 35.403380] ipq8064-mdio 37000000.mdio eth1: Set addr 14:59:c0:5a:43:9d
Fri Jul 26 15:12:29 2024 kern.info kernel: [ 35.414657] ipq8064-mdio 37000000.mdio eth1: nss_gmac_ipc_offload_init: enable Rx checksum
Fri Jul 26 15:12:29 2024 kern.info kernel: [ 35.421060] ipq8064-mdio 37000000.mdio eth1: nss_gmac_config_flow_control:
Fri Jul 26 15:12:29 2024 kern.info kernel: [ 35.429401] ipq8064-mdio 37000000.mdio eth1: nss_gmac_rx_pause_disable: disable Rx flow control
Fri Jul 26 15:12:29 2024 kern.info kernel: [ 35.436203] ipq8064-mdio 37000000.mdio eth1: nss_gmac_tx_pause_disable: disable Tx flow control
Fri Jul 26 15:12:29 2024 kern.info kernel: [ 35.445935] ipq8064-mdio 37000000.mdio eth1: nss_gmac_linkup: data plane open command successfully issued
Fri Jul 26 15:12:29 2024 kern.info kernel: [ 35.615216] ipq8064-mdio 37000000.mdio eth0: nss_gmac_tx_pause_disable: disable Tx flow control
Fri Jul 26 15:12:29 2024 kern.info kernel: [ 35.615778] ipq8064-mdio 37000000.mdio eth0: nss_gmac_rx_pause_disable: disable Rx flow control
Fri Jul 26 15:12:29 2024 kern.info kernel: [ 35.640026] ipq8064-mdio 37000000.mdio eth0: nss_gmac_start_up: Force link up
Fri Jul 26 15:12:29 2024 kern.info kernel: [ 35.640066] ipq8064-mdio 37000000.mdio eth0: 1000 Mbps Full Duplex
Fri Jul 26 15:12:29 2024 kern.info kernel: [ 35.646229] ipq8064-mdio 37000000.mdio eth0: nss_gmac_dev_set_speed:NSS_ETH_CLK_DIV0(0xc) - 0x100
Fri Jul 26 15:12:29 2024 kern.info kernel: [ 35.652311] ipq8064-mdio 37000000.mdio eth0: nss_gmac_reset: eth0 resetting...
Fri Jul 26 15:12:30 2024 kern.info kernel: [ 36.717459] ipq8064-mdio 37000000.mdio eth0: GMAC reset completed in 106 jiffies; dma_bus_mode - 0x20100
Fri Jul 26 15:12:30 2024 kern.info kernel: [ 36.717656] ipq8064-mdio 37000000.mdio eth0: nss_gmac_change_mtu: Enabling Normal Frame MTU (Requested MTU [1500])
Fri Jul 26 15:12:30 2024 kern.info kernel: [ 36.726095] ipq8064-mdio 37000000.mdio eth0: Set addr 14:59:c0:5a:43:9e
Fri Jul 26 15:12:30 2024 kern.info kernel: [ 36.736277] ipq8064-mdio 37000000.mdio eth0: nss_gmac_ipc_offload_init: enable Rx checksum
Fri Jul 26 15:12:30 2024 kern.info kernel: [ 36.742960] ipq8064-mdio 37000000.mdio eth0: nss_gmac_config_flow_control:
Fri Jul 26 15:12:30 2024 kern.info kernel: [ 36.751111] ipq8064-mdio 37000000.mdio eth0: nss_gmac_rx_pause_disable: disable Rx flow control
Fri Jul 26 15:12:30 2024 kern.info kernel: [ 36.758067] ipq8064-mdio 37000000.mdio eth0: nss_gmac_tx_pause_disable: disable Tx flow control
Fri Jul 26 15:12:30 2024 kern.info kernel: [ 36.768598] ipq8064-mdio 37000000.mdio eth0: nss_gmac_linkup: data plane open command successfully issued
root@R7800-1:~#
I have not seen these messages on my unit, but I had a quick look what this is for.
The code where this is used prints statistics for gmac. Have you added additiona code, that's not included in my repo? Or do you read stats from nss gmac?
WiFi throughput fix just got merged into 23.05.
1 Like
Hi Kong,
My diffconfig uses the following NSS-related packages. I don't run any other app/script that reads stats from NSS gmac. Thanks.
CONFIG_PACKAGE_kmod-nss-ifb=y
CONFIG_PACKAGE_kmod-qca-nss-drv=y
CONFIG_PACKAGE_kmod-qca-nss-drv-pptp=y
CONFIG_PACKAGE_kmod-qca-nss-drv-igs=y
CONFIG_PACKAGE_kmod-qca-nss-drv-l2tpv2=y
CONFIG_PACKAGE_kmod-qca-nss-drv-pppoe=y
CONFIG_PACKAGE_kmod-qca-nss-drv-qdisc=y
CONFIG_PACKAGE_kmod-qca-nss-drv-tunipip6=y
CONFIG_PACKAGE_kmod-qca-nss-ecm-standard=y
CONFIG_PACKAGE_kmod-qca-nss-gmac=y
CONFIG_PACKAGE_kmod-qca-nss-drv-netlink=y
CONFIG_PACKAGE_MAC80211_NSS_SUPPORT=y
The kernel is also updated to 5.15.163.
Only difference I see is, that you have included
CONFIG_PACKAGE_kmod-qca-nss-drv-pptp=y
Try without it. If this doesn't change it, then the issue may be target specific.
Your NSS build dated 26/7 shows correct 5GHz wifi bitrate on my XR500, but the non-nss build seems TX limited to 6Mbps on my EA8500.
I tried the EA8500 NSS build - still only 6 Mbps on TX - and with Load Ave high at 1.60.
Thanks Kong. I will try it the next time I build a new image. Such messages only show up once after booting.
The device is Netgear R7500v2.