Hello,
I hope you're well.
I need testers for my last branch. Who feels like doing the test?
next-8864 branch
-
WPA3 is functional as ieee80211w.
=> The Marvel developers had the super bad idea of replacing the probResponse with that of the chip (something formatted).
Fortunately, they've added a feature to disable this behavior. This makes WPA functional. -
AMSDU is dropped for 8864/8897 due to high ping. Strangely enough, without the scheduler to validate packets passed through the mac80211 layer, AMSDU packets are not processed. This part needs to be reviewed.
-
Chipset code separation between 8897/8864 <=> 8997.
In terms of performance, the code must be as atomic as possible. -
(8897/8864) I added one spinlock per hardware queue instead of one spinlock for all hardware tails. The aim is to have 4 schedulers working in //.
-
I've also added some unlikely instructions to improve the code in terms of predicting the result.
-
Monitoring correction.
-
Improved interoperability between mac80211 layer and driver in terms of encryption.
I need the penultimate commit "xmit enhanced with inline". -
txpower (which is the last commit)
The problem with the last commit is that it improves the txpower, but I don't know what damage it can do to the wifi card, I've already damaged one.
Another notable problem with this txpower commit is a drop in throughput.
=> From memory, this improves performance for the RC branches of openwrt. I wonder if it's not the fact of shifting the txpower update action.
As far as testing is concerned, I've done it myself.
- Ping test
- Performance test (https://www.degrouptest.com/test-debit.php) (iperf3 between network machines)
- Wireshark analysis
- Streaming game (moonlight) to check network stability
Without real Marvel support, I'm coding a bit blindly.
Unfortunately, I still don't have a WRT32, so no study of this code. I can't afford to do anything without a minimum of testing on my side.
Anyway, to the bravest, don't hesitate ;).
All the best,
Jbsk