Re: Question regarding CDC NCM and VNC performance issue

[Oliver Neukum <oneukum@suse.com>]

On 06.12.23 17:23, Hiago De Franco wrote:
> From: Hiago De Franco <hiago.franco@toradex.com>
> 

Good morning,

> Hi Oliver,
> 
> On 06.12.23 13:21, Oliver Neukum wrote:
>> Before we proceed, however, your patch raises a fundamental question:
>> Are we talking about gadget or host mode here?
>> Could you please describe your setup in detail?
> 
> Sure, I'm running a Apalis iMX6QD 512MB, which is basically an ARM 32-bits
> device with NXP iMX6QD. This device is running Linux with wayland, weston
> and the neatvnc server, so I can access my app using a vnc client on my
> host PC (also running Linux, Debian 12).

OK, you are using Linux on _both_ sides. Interesting, far from obvious, though.
(Putting Maciej into CC)  
> Finally, I open a VNC client on my host PC and connect to 192.168.11.1,
> where the issue begins with it being very slow/frozen. After applying the
> patch on the device (Apalis iMX6), now VNC runs smoothly.

For debugging purposes your first patch modifies the result of
USB_CDC_GET_NTB_PARAMETERS. Thereby both the host and the gadget change
their behavior. Hence we cannot tell which side has a problem with its
buffering at the old buffer size.

Can you tell us which side in your test case produces many small packets?

Furthermore, for testing purposes, could you decrease TX_TIMEOUT_NSECS in
f_ncm.c by an order of magnitude.

	Regards
		Oliver