4282

FWIW, I've been running NSS builds from around 10/1 on 2x MX4300 dumb APs w/ almost no problems.

4283

Non NSS has other issues unfortunately. The whole QCA platform is just flawed at least driver wise and stock is completely outdated.

4284

hi @sppmaster and @anom3, would you please check on your 301w, the result of cat /proc/cpuinfo ?

on my build(which is from qosmio's repo), the result does not have "model name"

below is the screenshot of my build

image
image1048×602 26.8 KB

and below is the screenshot of a vanilla openwrt installed on the same 301w

image
image1534×440 24.6 KB

Thanks in advance!

1 Like
4285

Here it is on my NSS build.

root@QNAP:~# cat /proc/cpuinfo
processor       : 0
BogoMIPS        : 38.40
Features        : fp asimd evtstrm aes pmull sha1 sha2 crc32 cpuid
CPU implementer : 0x41
CPU architecture: 8
CPU variant     : 0x0
CPU part        : 0xd03
CPU revision    : 4

processor       : 1
BogoMIPS        : 38.40
Features        : fp asimd evtstrm aes pmull sha1 sha2 crc32 cpuid
CPU implementer : 0x41
CPU architecture: 8
CPU variant     : 0x0
CPU part        : 0xd03
CPU revision    : 4

processor       : 2
BogoMIPS        : 38.40
Features        : fp asimd evtstrm aes pmull sha1 sha2 crc32 cpuid
CPU implementer : 0x41
CPU architecture: 8
CPU variant     : 0x0
CPU part        : 0xd03
CPU revision    : 4

processor       : 3
BogoMIPS        : 38.40
Features        : fp asimd evtstrm aes pmull sha1 sha2 crc32 cpuid
CPU implementer : 0x41
CPU architecture: 8
CPU variant     : 0x0
CPU part        : 0xd03
CPU revision    : 4
4286

Thanks, then its confirmed that the NSS build does not show "model name"...

@qosmio would you mind take a look at this, and figure out why...

4287

do you use vlan/mDNS reflector ? do you have one ssid where your streaming devices are connected(VLAN1) and another SSID where your phone is connected(VLAN2) and can get it to reliably work?

1 Like
4288

I don't use mdns. I use multiple VLANs (with different SSID) and 802.11r. Wired backhaul.

4289

my setup is similar with mx4300(just with additiional mDNS reflector which causes issues when one vlan has streaming device and another vlan is trying to access airplay etc. i have a unifi AP too in the mix and it does not exhibit this issue). have noticed with the setup though, if you have management Ip's on your vlans(for the ap) defined, then AP does not like it when you try to access Luci using a different vlan management ip than the current client device vlan. its very slow (firewall allowing). similar for iperf tests

4290

For me STA/client mode is working on my ax3600 running an old AgustinLorenzo prebuilt (NSS-WiFi) dated 2024-08-26-2004
A few releases later the 2024-09-14-1545 build already showed the problem, i've not tested the inbetween releases.

4291

It might be related to the version of nss firmware.
12.x does not like mesh etc. for that you need to revert to the 11.4

4292

@qosmio can you add that to your repo? Maybe this finally solves the broadcast misery

https://lore.kernel.org/linux-wireless/Z2Q9POuV-6MIdzRf@pilgrim/T/#t

2 Likes
4293

Hi guys. I wish you happy holidays.
I want to report an unexpected return of an year old issue.

I haven't had any major issue for a year.

The above was true during the year and now suddenly when a torrent is run (on a PC) and is saved to SSD (attached to USB3 router port using ksmbd for SMB sharing), Luci web interface and SSH completely hang.
During the time the router has hanged, no DNS requests can be made (so there is no Internet), already established connections and traffic continue to work but no new connections are possible. No OOM or other crashes but it seems like all services do not respond.
Once I stop the torrent both Luci and SSH recover immediately. DNS and other services recover too and I can access the internet like nothing wrong has ever happened.

These errors can be seen in the log. Obviously a consequence of the services stall.

daemon.notice hostapd: nl80211: nl80211_recv_beacons->nl_recvmsgs failed: -5
daemon.warn collectd[4112]: Sleeping only 2s because the next interval is 112.108 seconds in the past!
daemon.notice hostapd: nl80211: wpa_driver_nl80211_event_receive->nl_recvmsgs failed: -5
daemon.warn collectd[4112]: plugin_read_thread: read-function of the `cpufreq' plugin took 518.148 seconds, which is above its read interval (30.000 seconds). You might want to adjust the `Interval' or `ReadThreads' settings.
daemon.warn collectd[4112]: plugin_read_thread: read-function of the `thermal' plugin took 516.716 seconds, which is above its read interval (30.000 seconds). You might want to adjust the `Interval' or `ReadThreads' settings.

If I check the statistics all the graphics have empty intervals when the hang occurred.

image
image932×565 7.14 KB

root@QNAP:~# nss_diag
     MODEL: QNAP 301w
   OPENWRT: r28489-408dbcb419
IPQ BRANCH: main-nss
IPQ COMMIT: 408dbcb419
  IPQ DATE: 2024-12-16
    NSS FW: NSS.FW.12.2-161-HK.R
  MAC80211: v6.11.2-0-g7aa21fec187b
 ATH11K FW: WLAN.HK.2.9.0.1-02146-QCAHKSWPL_SILICONZ-1
 INTERFACE: br-lan     tx-checksumming: on  rx-gro-list: off
            10g-1      tx-checksumming: on  rx-gro-list: off
            10g-2      tx-checksumming: on  rx-gro-list: off
            lan1       tx-checksumming: on  rx-gro-list: off
            lan2       tx-checksumming: on  rx-gro-list: off
            lan3       tx-checksumming: on  rx-gro-list: off
            lan4       tx-checksumming: on  rx-gro-list: off
            phy0-ap0   tx-checksumming: on  rx-gro-list: off
            phy1-ap0   tx-checksumming: on  rx-gro-list: off

  NSS PKGS: kmod-qca-mcs-6.6.65.12.5.2024.02.27~26d6424-r1 aarch64_cortex-a53 {feeds/nss_packages/qca-mcs} () [installed]
            kmod-qca-nss-dp-6.6.65.2024.04.16~5bf8b91e-r1 aarch64_cortex-a53 {feeds/base/kernel/qca-nss-dp} () [installed]
            kmod-qca-nss-drv-6.6.65.12.5.2024.04.06~53a0dc1-r15 aarch64_cortex-a53 {feeds/nss_packages/qca-nss-drv} () [installed]
            kmod-qca-nss-drv-bridge-mgr-6.6.65.12.5.2024.06.12~1bcef16-r7 aarch64_cortex-a53 {feeds/nss_packages/qca-nss-clients} () [installed]
            kmod-qca-nss-drv-igs-6.6.65.12.5.2024.06.12~1bcef16-r7 aarch64_cortex-a53 {feeds/nss_packages/qca-nss-clients} () [installed]
            kmod-qca-nss-drv-qdisc-6.6.65.12.5.2024.06.12~1bcef16-r7 aarch64_cortex-a53 {feeds/nss_packages/qca-nss-clients} () [installed]
            kmod-qca-nss-drv-vlan-mgr-6.6.65.12.5.2024.06.12~1bcef16-r7 aarch64_cortex-a53 {feeds/nss_packages/qca-nss-clients} () [installed]
            kmod-qca-nss-ecm-6.6.65.12.5.5.2024.09.02~bd5057b-r3 aarch64_cortex-a53 {feeds/nss_packages/qca-nss-ecm} () [installed]
            kmod-qca-ssdk-6.6.65.2024.06.13~c451136b-r3 aarch64_cortex-a53 {feeds/base/kernel/qca-ssdk} () [installed]
            nss-firmware-default-2024.08.04~794fe373-r1 aarch64_cortex-a53 {feeds/nss_packages/firmware/nss-firmware} () [installed]
            nss-firmware-ipq8074-2024.08.04~794fe373-r1 aarch64_cortex-a53 {feeds/nss_packages/firmware/nss-firmware} () [installed]

This happens only when the torrent is saved to a ksmbd share on a SSD connected to the router USB3 port.
If I save the torrent to a local PC disk drive this hang never occurs.
I cannot tell for sure if ksmbd causes the hang but watching htop (till the SSH hangs) I can see that several services like dnsmasq, https-dns-proxy cause high CPU usage and then the router stops responding.

4294

Hi @sppmaster,

Regarding the latter, OpenWRT updated the mac80211 backports to version 6.12.6, so we need a refresh of the NSS patches to be able to reapply them correctly.

In the meantime this is the last commit we can go to without fail: https://github.com/openwrt/openwrt/commit/de0c143742517d401c4730137f092be8fb7e882a

FYI @qosmio

Regards, Agustin

1 Like
4295

@qosmio Hello, are you interested in adapting the 6.12 kernel? I've already completed most of the work, and the system is running stably, but there are errors related to wireless functionality. If you're interested, you can check out https://github.com/LiBwrt/openwrt-6.x.git The related bug link is: https://github.com/breeze303/openwrt-ci/issues/35

5 Likes
4296

never mind, I managed to find the root cause...

in order to "let /proc/cpuinfo show model name", a patch is needed.

Ref: https://lore.kernel.org/lkml/1472461345-28219-1-git-send-email-sumitg@nvidia.com/

1 Like
4297

You can take this patch

If you want to go beyond mac80211 6.12 backport

2 Likes
4298

Here's what it looks like when it's on the machine

44ffaa9372a16eb62d99ec3dbbc926c4
44ffaa9372a16eb62d99ec3dbbc926c41213×829 29.1 KB

1 Like
4299

Hi @asvio,

Perfect, thank you

I didn't see that you had already done this task.

I am going to prepare a new build with the new backports + this patch (https://git.codelinaro.org/clo/qsdk/oss/system/feeds/wlan-open/-/blob/win.wlan_host_opensource.3.0.r24/patches/ath11k/350-ath11k-Revert-clear-the-keys-properly-when-DISABLE_K.patch) to try to fix the broadcast failure that @Crect mentions

@asvio can you try that patch? https://github.com/AgustinLorenzo/openwrt/commit/e09b9b1457e338be20e3a7a1356d50f9c19bfe01 (I have good feedback on the devices I currently compile.) @qosmio can you integrate?

Regards, Agustin

4300

I'm actually running on my nbg7815 this patch to see how It woks.

I will do after i see there is no problem with current patches.

4301

There seems to be a memory leak.
image