[Bug 298238] [rtw88] RTL8822CE TX ring stall under load, hw_scan returns -16, requires full poweroff to recover

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=298238

--- Comment #2 from llmhater <llmhater@protonmail.com> ---
(In reply to Bjoern A. Zeeb from comment #1)


I) Did I read the man page?

No, I hadn't read 'man 4 rtw88' carefully before filing the report. I only
checked general FreeBSD wiki pages and skimmed the driver info. After your
reply I read the BUGS section properly, and both caveats absolutely apply to my
setup:
1) More than 4GB RAM (I have 8GB)
2) RTL8822CE with the documented association / software scan fallback issue

Apologies for the noise, I should have caught this before opening the bug.
Lesson learned: always read the driver man page first.



II) Linux kernel version on Artix:

Currently running: 7.1.9-artix1-2

On Linux, the rtw88 driver works reliably on this exact hardware with no
special tunables, the card just works out of the box after boot, including
sustained downloads and heavy TCP traffic. Happy to provide more details from
the Linux side (dmesg, iwconfig, driver version, etc.) if that would help
comparison.



III) Other:

Current status: wifibox works perfectly

Following your suggestion, I set up wifibox with PCI passthrough. After some
initial trouble (I had to enable hw.vmm.amdvi.enable=1 — the kernel actually
prints a helpful hint about this in dmesg), everything now works flawlessly.
WiFi is completely stable under any load: pkg install, long downloads,
sustained TCP traffic, none of it causes any of the crashes I was seeing
before.

Also worth noting: the 'ivhd0: Unknown dev entry:0xf0' warning during boot
appears harmless on this HP firmware, passthrough works fine despite it.


About continuing to help:

I would very much like to keep helping when you get to rtw88. Please feel free
to CC me on this PR (or related ones) whenever it becomes relevant, I promise
not to disappear.

One honest caveat: since wifibox is now my only working internet on this
machine, I'd prefer not to run tests that could brick the passthrough setup or
destabilize bhyve while I need the network. But I'm completely available for:

1) Testing patches against the base rtw88 driver in scenarios where I can
safely fall back to wifibox afterwards
2) Enabling rtw88_debug_mask and collecting logs
3) Reproducing specific failure modes with controlled configs
4) Running structured reproduction tests (e.g. same config, N iterations, log
outcomes)
5) Anything else that doesn't risk breaking the passthrough / bhyve setup

If there's a safe way to test rtw88 changes without unloading vmm or touching
the passthrough config, I'm all in.


About the wifibox recommendation in docs:

Would it make sense to mention wifibox in the BUGS section of 'man 4 rtw88' as
a recommended workaround for affected chipsets? It saved my sanity, and I
suspect others hitting the same wall would appreciate a pointer.

Thanks again for the incredible amount of work you do on the wireless stack,
and for taking the time to answer in such detail.

-- 
You are receiving this mail because:
You are the assignee for the bug.