not really been following, the talk of kernel 5.4 causing crashes is confusing me, we've been on kernel 5.4 for almost a year now?
fwiw my build has been running for 17d now: https://github.com/facboy/openwrt/tree/hnyman-ct-nss
The last good build from me was on 24-Nov. It seems there are 2 commits to that file since then.
As we wait for @Ansuel to dig into it, is it sane to just revert those two or possibly just one of them against a fork of master, then rebase against his kernel5.4-nss-qsdk10.0 branch?
EDIT: that does not work.
The last good build for me was 27 November. That is the last build I have posted in my bin folder.
You should be able to git rebase —onto a commit on that date and be good until the patch is fixed.
Thanks for the info. The last commit that didn't affect hostapd was 451c1eb on 28-Nov. When I checkout that, I was able to rebase @Ansuel kernel5.4-nss-qsdk10.0 branch without the merge conflict to that file. Building now.
Hi @ACwifidude & @darksky,
The conflicting commit in master that currently breaks nss builds with latest master is below "hostapd: Add cell_density data rates option" commit added on Monday 30th Nov'20.
author Nick Lowe <nick.lowe@gmail.com>
Wed, 28 Oct 2020 19:04:32 +0800 (11:04 +0000)
committer David Bauer <mail@david-bauer.net>
Mon, 30 Nov 2020 16:31:15 +0800 (09:31 +0100)
commit 81ff23fc91dbbedc374e58afdb4b0b13146c0f15
tree afccf439cc7dac40d2a6a3f006b5dbb568c0aa8d tree | snapshot
parent 451c1eb8c2a87bbf486f2883518bbb8c1c06848f commit | diff
The above commit adds new "cell_density" parameter and related changes to hostapd.sh that conflicts with Ansuel "Enable vht on 2.4Ghz" patches that itself adds new "vendor_vht" parameter to hostapd.sh and related changes that are not found from master.
While waiting Ansuel to update his patches on top of that commit, more adventurous persons may "fix" the reported conflict in package/network/services/hostapd/files/hostapd.sh by removing the master commit changes.
After "fixing" the conflict, it compiles ok and seem to work ok including both 2.4G and 5G radios.
OpenWrt SNAPSHOT r15132+63-9c13513ce1 / LuCI Master git-20.336.62696-d80a171
Can you recommend a set of commits to revert?
If you want to build up to the previous master commit before Ansuel nss build breaking hostapd.sh changes that's 451c1eb8c2a87bbf486f2883518bbb8c1c06848f.
If you want to "fix" the conflict by removing the master changes from hostapd.sh, below are rough guidelines;
- Reported conflicts after I merged Ansuel nss repo on top of latest master today (fresh clone & build)
Unmerged paths:
(use "git add <file>..." to mark resolution)
both modified: package/network/services/hostapd/files/hostapd.sh
added by them: target/linux/ipq806x/files/include/net/netfilter/nf_conntrack_dscpremark_ext.h
added by them: target/linux/ipq806x/files/include/uapi/linux/tc_act/tc_nss_mirred.h
added by them: target/linux/ipq806x/files/net/netfilter/nf_conntrack_dscpremark_ext.c
For the last 3 there's no changes so I just git add... them. 1st one hostapd.sh you need to "fix" manually.
- Manually "fixing" the conflict by removing the master changes from hostapd.sh (
vi package/network/services/hostapd/files/hostapd.sh)
- Look for below changes in the file (there are several, below the 1st)
<<<<<<< HEAD
config_add_int cell_density
=======
config_add_boolean vendor_vht
>>>>>>> Ansuel/kernel5.4-nss-qsdk10.0
-
Depending on how you update your build the above may be reverse (vendor_vht under HEAD and cell_density below) but nevertheless you'll remove cell_density related changes by deleting those (or commenting out).
-
After deleting cell_density changes, below is the only what's left for that above section;
config_add_boolean vendor_vht -
Repeat for all the other found conflicts in hostapd.sh file.
- Save the changes &
git add package/network/services/hostapd/files/hostapd.sh&git commit& build away (your workflow may be different than mine for nss builds though)
@Gram - Thanks for the detailed reply. Can you post a link to your adjusted hostapd.sh so I can compare mine to your's? Here is mine. Building now.
Thanks, they are identical. As well, building the latest commit after rebasing against kernel5.4-nss-qsdk10.0 as you suggested gives a seemingly fully functional system. WiFI is up and works as expected.
1 Like
Should be ok until some other master changes will refer to these "cell_density" changes.
And obviously those hostapd "cell_density" changes as in that commit will not work after "fixing" them 
.
1 Like
Just a status update: My build pulled at Nov 26th 19:10ET, is up and stable for 8.5 days
This build has the 5.4.80 kernel, and I believe the new NSS driver albeit I believe without the wifi offload
Hope this helps with narrowing down the ongoing stability issues.
is it possible to download your build to be flashed on R7800 with Netgear stock firmware? thanks
I have a factory image that you can try. Just flash from your web interface.
1 Like
thank you, I will try it when I'm home again, do I need to do any changes or just flash and enjoy? I just want a stable build with NSS working. 
btw what is the difference between your image compared to Kongs from September?
I believe there is a an image in the build dir that you could flash from the stock, but I have not tested it.
I'm using dd-wrt / openwrt for a very long time, so I never pay attention to that image.
I would recommend trying @ACwifidude's build first, as mine has additional customizations
1 Like
Kong’s has more software and a couple custom configurations.
Mine has no custom configuration and a simple set of software. Just load the factory image and it should work right out of the box.
1 Like
@darksky @Gram I went the opposite way. I removed the “ Enable vht on 2.4Ghz” commit and I’m using master’s version. Will prevent further issues in the future.
@Ansuel VHT by definition is 802.11ac - which doesn’t exist in the 2.4ghz spectrum - so I’m confused about the origin of the patch. Non standard VHT implementation for specific devices?
thanks, I flashed the November image earlier and now the December one, thanks!
btw I just noticed after I flashed latest one it removed the addons (adblock, upnp) is it supposed to do that even though I choose keep current settings?
That is normal - my build does not have those packages so your device will “lose” those packages with each sysupgrade.
1 Like