take your time, just thought if I should just do basic settings on current firmware until new one is done. :slight_smile:

So, at the end you will have a very simple distribution like 24.10 and you can install by Luci every app you want...

edit

testing version with apk:

Only basic package installed, you have to install everything you need... apk based

7 Likes

Ok, thanks again for share this info.

What do you think if I add it on script for irq?

Only this

mtk_eth-5

should be moved or do you think it is better to move 5 to 1 and 6 to 3?

Many many thanks

Here..

Here the log of the updated script...

Wed Nov 20 21:39:20 2024 user.notice root: smp_affinity: assigning 15100000.ethernet to CPU1 on IRQ 120
Wed Nov 20 21:39:20 2024 daemon.notice procd: /etc/rc.d/S24smp_affinity: root: smp_affinity: assigning 15100000.ethernet to CPU1 on IRQ 120
Wed Nov 20 21:39:20 2024 user.notice root: smp_affinity: assigning 15100000.ethernet to CPU3 on IRQ 121
Wed Nov 20 21:39:20 2024 daemon.notice procd: /etc/rc.d/S24smp_affinity: root: smp_affinity: assigning 15100000.ethernet to CPU3 on IRQ 121
Wed Nov 20 21:39:20 2024 user.notice root: smp_affinity: leave mt7915e to All CPUs on IRQ 133
Wed Nov 20 21:39:20 2024 daemon.notice procd: /etc/rc.d/S24smp_affinity: root: smp_affinity: leave mt7915e to All CPUs on IRQ 133
Wed Nov 20 21:39:20 2024 user.notice root: smp_affinity: mtk_eth-5 have: pid 524's current affinity list: 0-3 pid 524's new affinity list: 1
Wed Nov 20 21:39:20 2024 daemon.notice procd: /etc/rc.d/S24smp_affinity: root: smp_affinity: mtk_eth-5 have: pid 524's current affinity list: 0-3
Wed Nov 20 21:39:20 2024 daemon.notice procd: /etc/rc.d/S24smp_affinity: pid 524's new affinity list: 1
Wed Nov 20 21:39:20 2024 user.notice root: smp_affinity: mtk_eth-6 have: pid 525's current affinity list: 0-3 pid 525's new affinity list: 3
Wed Nov 20 21:39:20 2024 daemon.notice procd: /etc/rc.d/S24smp_affinity: root: smp_affinity: mtk_eth-6 have: pid 525's current affinity list: 0-3
Wed Nov 20 21:39:20 2024 daemon.notice procd: /etc/rc.d/S24smp_affinity: pid 525's new affinity list: 3

1 Like

thanks, is it possible to install software through luci? when trying to update the list getting error.

I'd say both, 5 and 6, should be out of core 0, because it seems kernel threads are running on 0.

Post a screen I've checked and I didn't get any issues by installing software...

also luci-adblock can be installed but won't run.

2 Likes

While I was working on checking the relationship between 'smp_affinity' and 'packet_steering', I noticed something.

root@GL-MT6000-LIGHT:~# /etc/init.d/smp_affinity start
root: smp_affinity: assigning 15100000.ethernet to CPU1 on IRQ 120
root: smp_affinity: assigning 15100000.ethernet to CPU3 on IRQ 121
root: smp_affinity: leave mt7915e to All CPUs on IRQ 133
root: smp_affinity: mtk_eth-5 have: pid 525's current affinity list: 0
pid 525's new affinity list: 1
root: smp_affinity: mtk_eth-6 have: pid 526's current affinity list: 0
pid 526's new affinity list: 3
root@GL-MT6000-LIGHT:~# /etc/init.d/packet_steering start
pid 525's current affinity list: 1
pid 525's new affinity list: 0
pid 526's current affinity list: 3
pid 526's new affinity list: 0
pid 1140's current affinity list: 3
pid 1140's new affinity list: 3
root@GL-MT6000-LIGHT:~# ps aux|fgrep 1140
root      1140  0.0  0.0      0     0 ?        S    02:11   0:02 [mt76-tx phy0]
root     10532  0.0  0.0   1408   896 pts/0    S+   04:57   0:00 fgrep 1140

The “packet_steering” script was targeting “mt76-tx”. Countermeasures may be required.

1 Like

I dont'understand why adblock-fast installing other package... Maybe you have this also on snapshot...

I've tried install sqm and openvpn and I didn't get issue...

Ok, checking...

edit 1

  • it's better to check packet steering status otherwise the setting are not applied...
1 Like

it does not install it, it was just an example that could not be installed, same with for example sqm, normal adblock can be installed but won't run.

in theory, we could move more than just mtk_eth-* off core 0.

untested so i do not know how the wifi related ones will react but here are the non 0:00 entries on one of my dumb aps after about 2 weeks of uptime.

all that being said, as i mentioned above: with mtk_eth-* moved off core 0, it no longer reaches anywhere near 100% for me regardless of traffic so there is probably no point.

root@OPENWRT-SALON:~# ps aux | grep \\[ | grep -v 0:00
root        14  0.0  0.0      0     0 ?        S    Nov12   0:51 [ksoftirqd/0]
root        15  0.0  0.0      0     0 ?        I    Nov12   1:52 [rcu_sched]
root        16  0.0  0.0      0     0 ?        S    Nov12   0:01 [migration/0]
root        20  0.0  0.0      0     0 ?        S    Nov12   0:44 [ksoftirqd/1]
root        25  0.0  0.0      0     0 ?        S    Nov12   0:52 [ksoftirqd/2]
root        30  0.0  0.0      0     0 ?        S    Nov12   0:36 [ksoftirqd/3]
root        38  0.0  0.0      0     0 ?        S    Nov12   0:34 [kcompactd0]
root        59  0.0  0.0      0     0 ?        S    Nov12   0:11 [kswapd0]
root       141  0.0  0.0      0     0 ?        S    Nov12   0:02 [hwrng]
root       524  0.0  0.0      0     0 ?        S    Nov12   5:37 [napi/mtk_eth-5]
root       525  0.1  0.0      0     0 ?        S    Nov12  18:43 [napi/mtk_eth-6]
root       933  0.0  0.0      0     0 ?        I<   Nov12   0:03 [kworker/u9:0-hci0]
root       942  0.0  0.0      0     0 ?        I<   Nov12   0:03 [kworker/u9:2-hci0]
root      1057  0.2  0.0      0     0 ?        S    Nov12  33:45 [napi/phy0-7]
root      1058  0.1  0.0      0     0 ?        S    Nov12  17:55 [napi/phy0-8]
root      1059  0.0  0.0      0     0 ?        S    Nov12   9:15 [napi/phy0-9]
root      1060  0.0  0.0      0     0 ?        S    Nov12   2:26 [napi/phy0-10]
root      1062  0.1  0.0      0     0 ?        S    Nov12  22:17 [napi/phy0-12]
root      1081  0.1  0.0      0     0 ?        S    Nov12  24:54 [mt76-tx phy0]
root     12025  0.0  0.0      0     0 ?        I    Nov20   0:03 [kworker/0:1-events]
root     14534  0.0  0.0      0     0 ?        I    Nov16   0:01 [kworker/3:0-events]
root     18668  0.0  0.0      0     0 ?        I    09:55   0:01 [kworker/u8:4-flush-7:0]
root     20518  0.0  0.0      0     0 ?        I    10:14   0:01 [kworker/u8:0-events_unbound]

I did a quick test yesterday and if someone don't use packet steering, there is no improvement, i would say i have even lower speeds than with the previous config. And in this case, other core will be fully loaded.

I've tried sqm and openvpn and it works, sorry

You get everytime the error on adblock have you read the output?

try to reset and restart and try again, let me know


regarding smp_affinity script I have postponed the run, now it work also if is enabled the packet steering script

START=26

1 Like

can anyone confirm the top wifi speed you can get with pesa builds?
with standard 24.10 (like old 23.05) i'm around 1.4G and just wanted to know if this thing can do anything better.. (some time ago i read here someone getting 1.7G which is amazing!)

Yes I reseted twice but same problems unfortunately. :frowning:

edit: reseted for third time and same problems, now I tried to install sqm and it worked, but when trying to install something else it fails, in this case stubby, but same thing happens with adblock.

edit2: tried with latest standard Openwrt 24.10 Snapshot and it works, but using opkg.

Apologies if my understanding is incorrect, but why the switch to apk alongside an attempt to build a stable version of 24.10? I presume the apk changes, currently in the main branch (post 24.10) has been pulled into the 24.10 version used by pesi?

Why are we not focusing on the stable 24.10 version using opkg, then once stable work on importing apk?

(again, my understanding might not be correct)

Check /etc/apk/arch do you have all listed?
If not add it:
echo "all" >> /etc/apk/arch

1 Like

Opkg sure is ok. Please check snapshot if you have time.
Thanks

1 Like

Sorry and thanks for testing.

upload testing 2 and removed old one.

2024-11-21_r28320-e01af5e8d6_next-r4.4.0.mtk

  • kernel 6.6.61
  • mt76: update to 2024.11
  • mac80211: updated with latest fix
  • adjusted smp_affinity script

testing version 2

3 Likes