I can test pppoe, but you need to help me understanding how to build a NSS image with TWO repositories as source lol 
How to integrate the NSS packages in the owrt image?
Can i use a diffconfig from 5.10?
1 Like
I just had a look at your NSS packages repo and saw the issue you opened. When looking at the kernel panic, I get the idea you're not hitting an NSS firmware loading issue, but a frequency scaling issue, don't take my word for it though. I believe @Ansuel is very familiar with that, so perhaps he can have a look at that issue?
I believe you can use your own diffconfig from 5.10 but add this to it to select the 5.15 kernel when building:
CONFIG_LINUX_5_15=y
CONFIG_EXPERIMENTAL=y
CONFIG_TESTING_KERNEL=y
I too would like to contribute in testing 5.15 with NSS, but I'm also not sure on how to combine the contents of two repositories to get a build done. I believe the contents of NSS packages -> https://github.com/qosmio/nss-packages needs to be placed in openwrt/package/qca. And from the looks of it, the required patches to enable those are already in place in Kernel 5.15 (QSDK 11) -> https://github.com/qosmio/openwrt-ip806x.
1 Like
@ACwifidude ,
what's the status of @Ansuel's clock patches - I recall him saying he'd found issues (secondary mux) - have any fixes been checked in here or are we waiting for 5.15?
You can pull all recent Ansuel's commits to 5.15 and apply them on top of the current master commit HEAD and all existing DTS/NSS patches. No problem with the build.
I'm running such an image on R7800.
uname -a
Linux xxx 5.10.143 #0 SMP Sat Sep 17 19:02:52 2022 armv7l GNU/Linux
I installed the latest 5.10 base snapshot (MasterNSS) on the EA7500V1, it's stable and the router doesn't seem to reboot anymore.
The RAM is still acting weirdly though. Currently top reports 100MB used, 95MB free but only 5MB available. How can that be? I checked tmpfs mounts and they are barely occupied.
1 Like
You don't have to worry about the available memory. Linux uses lots of memory for buffer/cache and will dynamically change them accordingly. In Linux, unused memory is considered to be wasted memory.
Waiting for a little more testing and for people to try it out on 5.10. If everything looks good - I can build with the fixes.
Just posted a new master build (just a rebase, removed nlbwmon due to the log spam / memory hacks to try to get it to work, I haven’t applied the CPU changes yet). Master and 22.03 have some mac80211recent commits which might be interesting. Will see if I notice anything.
Will be posting a new 22.03, similar setup.
3 Likes
from the linux 'free' man page, which the memory status page mimics:
free displays the total amount of free and used physical and swap memory in the
system, as well as the buffers and caches used by the kernel. The information is
gathered by parsing /proc/meminfo. The displayed columns are:
total Total installed memory (MemTotal and SwapTotal in /proc/meminfo)
used Used memory (calculated as total - free - buffers - cache)
free Unused memory (MemFree and SwapFree in /proc/meminfo)
shared Memory used (mostly) by tmpfs (Shmem in /proc/meminfo)
buffers
Memory used by kernel buffers (Buffers in /proc/meminfo)
cache Memory used by the page cache and slabs (Cached and SReclaimable in /proc/mem‐
info)
buff/cache
Sum of buffers and cache
available
Estimation of how much memory is available for starting new applications, with‐
out swapping. Unlike the data provided by the cache or free fields, this field
takes into account page cache and also that not all reclaimable memory slabs
will be reclaimed due to items being in use (MemAvailable in /proc/meminfo,
available on kernels 3.14, emulated on kernels 2.6.27+, otherwise the same as
free)
it's only confusing if you think about it - and memory may be in use for buffers, cache but 'available' to programs should they need more data space.
Yeah, what confuses me is that LuCI says total available close to 0 or a "?" instead of a bar.
But anyway the uptime is now solid 12 hours with some services turned on and running fine, so i guess linux manpage is the one that's right
I'll keep it on for longer and throw some some torrents+i2p connections at it to see.
Just checking; one of the first patches from Ansuels PR splits the dts files from 5.10 to 5.15, most of the patches are then applied only to 5.15. Unless you managed to cherry-pick the correct patches from his PR, what is the current state your R7800 is running? Or did I mis something completely yesterday? Wouldn’t be the first time 
Just to note, after 18 hours uptime, i realized upnp was missing, so i installed luci-app-upnp and started the service. A few minutes later, reboot.
So upnp seems to be the main culprit here. Can anyone try this on their routers?
isn't upnp somewhat of a security risk? I was taught to only explicitly assign stuff on my network as opposed to allowing stuff to assign itself.
IGDV2 is fine afaik, those breaches were with V1.
It's easier for me, i use like 7 programs and a server that make use of upnp.
My understanding is that it is better to build the images with what you need 'baked' in vs adding later. It may not be the cause of your problem, though. I used to add minidlna to the builds before but now build my own images with everything needed as per @ACwifidude suggestion.
1 Like
what's new?
@ACwifidude - I would really like y'all to try a very simple patch and see if you can "feel" a difference, reducing NAPI_POLL_WAIT to 8, from it's present 64, across your build. multi-core A53s, in particular, could context switch fast enough to do this well, and it it seemed to be better to do less work, more often.
1 Like
interesting wonder if this can be applied generally...
Depends on what you mean by general. x86 - probably not. a72 - have no idea. In a virtualized environment, no. But bare metal quad core a53, works for me, and we've been trying to shorten the latencies in wifi (see link above) and it turns out that NAPI wasn't polling often enough. I think. So... could use more testers.
1 Like
Sounds interesting. I’ll be busy at work this week so I won’t have time to test or build - but would love to encourage some testing to see the results!
If anyone tweaks their custom build - post some results!
2 Likes