gagan sidhu

here is a sample, i didn't see any OOPs until debug, so maybe you can help. i will look into the file it is complaining about EDIT: just found out...

it's working correctly, for sure? then my client wifi card must be poor (BCM943602CDPAX) because i get only approx ~400Mbps download and ~ 500Mbps upload. i thought it should be...

for some reason it's download that is the problem. it is not hitting CPU. for upload, i see CPU usage go up right away and i can observe the speed...

it seems to me for some reason hook works better for upload rather than download. i experienced this problem with both hw_nat and natflow. i am not sure why this...

i was missing this patch: https://git.openwrt.org/?p=openwrt/openwrt.git;a=commitdiff;h=6786dc26a205da55ec2d9771693cdfb99e756e59 i am hoping by applying it, the results are better.

hellloooo my beautiful CHAINESE FRIEND @ptpt52. i have definitely got it working now, and natflow (maybe not superior, but possibly equivalent) to raeth+ra_hw_nat. my BCM943602CDPAX is probably the issue, but...

here is mpstat during download portion (5ghz radio assigned to cpu 3, just for your information): ``` root@DD-WRT:~# ./mpstat -P ALL 1 Linux 4.14.225 (DD-WRT) 03/15/21 _mips_ (4 CPU) 12:01:40...

do you think having a preempt kernel may be part of the issue? i thought a high-resolution timer plus preempt is superior to a period timer that is typically used.

Yes it should be high and I understand why. Just don't understand why Cpu usage for download much lower than upload. I've experienced too many issues with mt76 driver on...

```[16.389982] Loading modules backported from Linux version wireless-drivers-next-2021-02-12-0-g9d083348e938 [ 16.406641] Backport generated by backports.git v5.10.16-1-0-g21d2a1d2 [ 16.863297] bus=0x1, slot = 0x0, irq=0x0 [ 16.882479] ieee80211 phy0: Selected rate control algorithm...