Joining the infamous club of people having had WiFi stability issues with WRT32X since I first ran OpenWRT back in 2018. It's currently only collecting dust as reliability is just not workable. Now decided to finally make an account and offer any help needed to run tests on my WRT32X. Let me know if and how my device can help.
@adworacz I just spent some time testing r13954 and it was quick and easy to reproduce the wifi cutouts and was able to reproduce the issue approx. 8-9 times until I had enough.
Updated Builds List
r13342 (David's last build - May 2020) GOOD
r13583 (found on Wayback Machine - June 2020) GOOD
r13954 (July 27, 2020) BAD
r14566 (September 2020) BAD
r15791 (February 2021) BAD
21.02.x (All RCs and final stable) BAD
Therefore the issue is now narrowed down between:
r13583 (June 16, 2020) and r13954 (July 27, 2020)
For some reason, I had high hopes that r13954 would have been successful. So it would seem that the issue occurred not too long after David's last build.
1 Like
Not much to be seen in that window:
a4c0767fbc mac80211: ath9k: enable adding wireless noise to kernel entropy pool
a4248577a0 hostapd: fix compilation of wpa_supplicant
b933f9cf0c toolchain: remove gcc libssp and use libc variant
ba7ddae9a9 comgt-ncm: do not attempt to connect if the control device is invalid
8fe9940db6 openvpn: add generic hotplug mechanism
9c16c46e58 ath79: define CONFIG_MFD_RB4XX_CPLD to fix build breakage
8e98613f4d uclient: uclient-fetch: add option to read POST data from file
ef9cee4980 lantiq: drop "CONFIG_HZ_100 is not set"
27724c2f54 lantiq: switch to 5.4
90f930e0c2 kernel: bump 5.4 to 5.4.46
014c101e0b mvebu: fix bootloader kernel commandline mangling
and nothing leaps out from a quick check of the kernel change log.
1 Like
These two sound interesting to me:
hostapd: enter DFS state if no available channel is found · openwrt/openwrt@93bbd99 (github.com)
mac80211: create channel list for fixed channel operation · openwrt/openwrt@cfd2f3b (github.com)
Nothing else stands out to me right now either.
That would expand the window provided by the above:
./scripts/getver.sh r13583
c4037a5b9eee09c96e5c8108f1b7aa439f96cb6b
./scripts/getver.sh r13594
a4c0767fbc50ea50f1f0c5adfa35d1c8d347462b
git log --oneline c4037a5b..a4c0767f
r13954, not 13594
Should expand the window a bit.
I’ll work on a new build.
Thought that list looked a bit short, watching the game got the better of my attention.
@WildByDesign New build, r13648 is up: https://openwrt.austindw.com/linksys-wrt/wifi-hang-bisect/r13648/
2 Likes
Excellent. This new build is before mac80211 5.8 release. I think the previous build had mac80211 5.8 RC2 if I recall correctly. This is also something that I have been curious about.
I won’t be able to test until my kids are done virtual school stuff but later in the evening I will test this new build. If you have a new build each day, I can test each night and provide feedback around the same time each night.
We are definitely getting closer to finding this issue.
2 Likes
I now have this version on one of my spares running a unique SSID on 5GHz only that only my iPhone is connected to. It's AP only, no active WAN interface. With 21.02.1, I could connect with my iPhone and my Mac for a few minutes after boot and then it would stop passing traffic. Been running for about 15 minutes now, will monitor the iPhone to see if it ever stops working.
On a side note, I REALLY like the web interface on this build, a bit more modern than the stock interface!!
1 Like
@adworacz Build r13648 is good. No cutouts.
Immediately after flashing this build, I could tell that it was going to be good because everything was so responsive and smooth. The bad builds are really janky at all times even between the cutouts.
r13342 (David's last build - May 2020) GOOD
r13583 (found on Wayback Machine - June 2020) GOOD
r13648 (June 26, 2020) GOOD
r13954 (July 27, 2020) BAD
r14566 (September 2020) BAD
r15791 (February 2021) BAD
21.02.x (All RCs and final stable) BAD
Therefore the issue is now narrowed down between:
r13648 (June 26, 2020) and r13954 (July 27, 2020)
We are down to approx. one month now of potential commits to determine the source of this issue.
Excellent. I’ll cut another build tonight. I think we’re ~8 builds left until we nail the exact commit.
Edit: New build, r13801, is up here: https://openwrt.austindw.com/linksys-wrt/wifi-hang-bisect/r13801/
1 Like
@adworacz Just finished testing build r13801. It's a good build. No cutouts.
Updated Build Testing List
- r13342 (David's last build - May 2020) GOOD
- r13583 (found on Wayback Machine - June 2020) GOOD
- r13648 (June 26, 2020) GOOD
- r13801 (July 11, 2020) GOOD
- r13954 (July 27, 2020) BAD
- r14566 (September 2020) BAD
- r15791 (February 2021) BAD
- 21.02.x (All RCs and final stable) BAD
Targeted Issue Range:
r13801 (July 11, 2020) and r13954 (July 27, 2020)
2 Likes
I see that 21.02.1 is released, is this version safe for us to flash?
-Pyro
Roger that!
New build, r13877 is up: https://openwrt.austindw.com/linksys-wrt/wifi-hang-bisect/r13877/
2 Likes
sorry, does it fix?
They're just narrowing down where the problem in the code occurred. There's no fix yet.
1 Like
Kudos to the lads working so rigourously to nail this bug down!
3 Likes
Likely the end result is a commit doing either a kernel version bump or mac80211 bump (or mwlwifi bump, but I think that there is nothing on that range)
My personal guess is
2020-07-23 mac80211: Update to version 5.8-rc2-1
2020-07-23 mac80211: Update to version 5.7.5-1
Less likely possibilities:
2020-07-20 hostapd: enter DFS state if no available channel is...
2020-07-20 mac80211: create channel list for fixed channel operation
2 Likes
Excellent, thank you. I will test this later tonight as always and provide feedback at similar time.
If this build checks out as good, I would strongly suggest creating a build in between the commits above posted by @hnyman so that we can have separation between mac80211 5.8 commit and the hostapd/mac80211 DFS commits.
As a matter of fact, if you can, would you be willing to create another build now? I will push myself extra and test two builds tonight. If I’m already losing a bit of sleep, I might as well lose a bit more. But we’re close to this issue and it’s giving me more determination.
2 Likes