I have 2 dumb AP, one is Asus TUF AX6000, the other one Asus TUF AX4200.
On both I have same issue, I've on both one wifi SSID in 2.4ghz which works just fine. Also have on both one SSID with multiple vlans (filtering vlan by client password). I did remove wpad-basic-mbedtls (which was the one installed on my router by default) and installed wpad-mbedtls in order to do the vlan filter by password.
Everything works, but when I'm connected to the 5ghz SSID on any of the routers, whenever I perform a speedtest, download test is fine, but on upload test wifi kicks my device out, and both wifi interfaces (2.4ghz and 5) in the router restart, so wifi dissapear for like 10 seconds.
I've tested turning off wifi in one of the dump ap, just in case there was any conflict with SSID, still same issue.
If I use one SSID for each band and only vlan, everything works as expected.
I tested with different clients, mainly iPhone15, Macbook pro OSX, and Windows 10 laptop, it happens on all of them. Should there be ack packets in that log?
I don't know if the issue is the same, I search on the forum before posting, the only post with exactly same issue I found is this one: BPI-R3 multi_ap WDS issue (device crash)
but Im not even using wds.
I don't have WED enabled. I have hardware offloading enabled on main router, but I have the problem on the dump aps attached to that router, not on the router itself, since I'm not doing multi vlan SSID on main router.
As I said, wds and wed are disabled in router and in dumb aps (no firewall on dumb aps). The config posted is from one of the dump ap, but this happens on both.
HW offloading is enabled in main router. Everything works, just wifi fails on upload speedtest in dumb aps, I can not reproduce it if I don't saturate wifi network.
It looks like issue 690, check that packet steering is enabled and actually yields almost same load on all CPU cores. Maybe irqbalance can help offload CPU0 if only that tops the lines.
I checked with htop, even with packet steering disabled, CPU cores doesnt even hit 15%, all cores are around 0-10% when im speed testing, so I don't think it's an overload problem in CPU0.
Try snapshot -> backup conf -> flash snapshot -> check
If no improvement -> sysupgrade back with conf reset, restore initial conf.
In htop press f2->disable hide kernel threads, enable cpu details. Should be minimal pink band representing softirq on the right side of currently red "kernel" cpu usage.
Okay, did everything.
Tried with snapshop, the problem persist, about the same results in htop.
I did sysupgrade back and then restore with the config, disabled hide kernel threads, and "enable cpu details" was not an option, I enabled Detailed cpu time (which I assume it is what u mean).
Interrupts and other kernel work is well balanced. nothing to add there. Somebody wiser should say what parameters to change to make driver emit more debug data in logs.