RE: [REGRESSION] Freeze on resume from S3 (bisected)

[Forty Five <mathewegeorge@gmail.com>]

[-- Attachment #1: Type: text/plain, Size: 1649 bytes --]


Hi Ping-Ke,

Sorry for the delay since my last response.

First off - I am unable to apply your debug patch
(https://lore.kernel.org/linux-wireless/8583c53fa42848c9855b2b425ac18ca4@realtek.com/)
on top of either bcbefbd032df or 5bbd9b249880; the patch application
fails and I'm not confident enough to try and manually apply it. So the
logs here do not include debug output.

> Your setup is very complicated, so I can't setup in my side easily,

Fair enough. For (my) future reference, let me just note down the
procedure I'm using to reproduce the system freeze at this point -

1. Have a working hotspot using hostapd, managed using hostapd.service.
   Start hostapd.service.

2. Suspend and resume the system.

3. Wait for hostapd to produce 'handle_probe_req: send failed' error
   messages. (My means of triggering these is by having another device
   attempt to connect to the hotspot)

4. Restart hostapd.service. The system will usually freeze at this
   point. If not, I repeat these steps a few times until it does.

> First problem is the culprit commit [1] that makes system frozen, and I still
> feel the patch [2] you have taken can fix it. Please use [1] as code base and
> apply patch [2] to see the result (#exp 1).

This does indeed appear to be the case - the patch does fix the issue on
that specific commit.

> Third (unsure) problem could be introduced by commits between [1] and [3].
> If first problem can be addressed by #exp 1, it could be possible to bisect
> the problem between [1] and [3]. Even if [1] is the only problem, revert
> the commit to see if it becomes good (#exp 2).

Here are the crash logs from #exp 2:

[-- Attachment #2: kdumpst-202407081539.zip --]
[-- Type: application/zip, Size: 32447 bytes --]

[-- Attachment #3: kdumpst-202407081543.zip --]
[-- Type: application/zip, Size: 33202 bytes --]

[-- Attachment #4: Type: text/plain, Size: 244 bytes --]

In the first log, the freeze happened at the end of step 4; in the
second, I had to repeat steps 1-4 a second time, after which the freeze
happened at the end of step 4.

I'll get to work on the bisection, and send the log here when it's done.