Oh... thanks for the info, might explain why some managed to get GIC-0 mode for ath10k_pci.
I stated before that it wasn't possible but some people said it was. I never got an explanation why that was, I thought it was because of different revisions of the router.
Some speculation...
From what I understand, MSI should lead to less lockups etc. However I doubt that is much of a problem for a router like the R7800?! Also the ARM architecture is a bit different too.
If we control what CPU wifi uses, in theory it should/could lead to better performance if you do it right, at least according to my logic.
A bit off topic now...
I tried your approach using stubby but I noticed that chrome is acting funky. Chrome first says the websites DNS name can't be found, then it loads the page after a second or two, once it's cached it loads fine. You seen this behavior?
Yea, I tried some debugging but I can't figure out why it happens.
Querys get refused or it returns NXDOMAIN. I think it's dnsmasq doing something weird. Because I don't see anything in the stubby log when it happens.
Oh well, sorry for offtopic.
EDIT:
I tried every solution I can figure out... it still reports NXDOMAIN.
I'm starting to think it's some weird dnsmasq bug...
Update
Ok, so it was the option strict order in dnsmasq that caused it, disable it and it should work.
hello,
just for information,
in addition to the Kong R7800 build,
there is also the Cezary Jackiewicz R7800 build with a full repository so you can use and install all kernel module and software, have fun
how did you manage to get that crashlog? that kind of crash looks to be very random. Also the error from ath10k are normal or it's the first time you notice them?
The thing is that yes from the bootlog is caused by the cpufreq driver (that scale the cpu freq) but the crash was caused by the krait notifier that is used to put the clk to a safe freq before switching... so it's from a deeper level and not from the cpufreq driver but by the clk driver that control the frequency (much deeper level)
I don't know but looking at the ath10k error and the strange kernel panic it looks like the system just fked up bad and crashed on the first operation it was doing... (but could really be a bug in the mux clk code that is hard to reproduce as it's very difficult to trigger)
Any idea how to reproduce this?
The crash log is from the ramoops pstore after the router rebooted.
The ath10k errors seems to be a recent thing from what I remember. I think I started noticing it from since early Dec 2021 after I made a new build to test out ar8337 IGMP snooping code. Since Dec 2021, I noticed that the router's WiFi is unstable. I sometimes need to restart the WiFi interface to get traffic flowing thru WiFi again. LAN traffic is fine tho.
Unfortunately I do not have any idea what caused the panic nor do I know how to re-produce it. I was reading forum posting and the router was connected to an OpenVPN TAP tunnel from a remote site when the router crashed.
mhhh it was under load? Can you share the config for the pstore? would be handy to track some problem with nss scaling (I also have some crash but as they are random I never manage to investigate them)
Enable pstore filesystem and ramoops using make menuconfig and it should start storing oops logs in /sys/fs/pstore when the router reboots.
Edit: You can test out the ramoops log by simulating a reboot using this command (if memory serves):
echo c > /proc/sysrq-trigger
For the reboots caused by the watchdog (which I assume is causing the spontaneous reboots of the NSS scaling), since it is not a kernel panic, it will not have any logs stored.
Curiously, but also I got a crash with my 21.02 build last week. First crash in ages.
Upstreaming and enabling ramoops for R7800 (and all ipq806x?) might be helpful. Mt7622 has it on by default, and it has helped a lot in the mt76 WiFi driver related crashes with e8450/rt3200.
Thing is that it would reduce ram space. I assume other target have better way to store these kind of stuff. Interesting that it does work and uboot doesn't clear this space. Wonder if I can try implementing the original way QCOM use to store kernel panics. Should be a funny project.
Potentially dumb question. Flashing to an image that has ramoops enabled can be done by sysupgrade? Or is RAM size/layout change necessary to use factory?