Vojtěch Svědiroh

Could you please link the exact lines of code that are suspected of user token exfiltration? Otherwise this issue doesn't hold any water.

> I can't set any Higher value for manual mode than the one set in the hightest field, even if the check box is not checked. Let's say i set...

So it means that the % range used by dynamic mode is also supposed to limit the upper value of manual mode? Would it be possible to reflect this in...

Bug 4 seems to be caused by "insufficient kernel work area" for sockets, as all network functions return SYS_NET_ENOBUFS when it occurs. Not sure if that happens because of too...

Those are the values right after booting:  And after triggering bug 4, confirming that kernel connection table is full:  Things that were tried to decrease...

> Limitting files per second for passive mode under the magic value does bypass this issue, but it's impossible to calculate it on-the-fly. This "magic number" is actually tied to...

> Have you tried changing the thread priority or limiting the number of threads? https://github.com/aldostools/webMAN-MOD/blob/master/include/ftp.h#L1412-L1432 I'm not really sure if more time from CPU scheduler could help with kernel freeing...

FTP avg upload speed: 16.3 MiB/s FTP avg download speed: 22.9 MiB/s I don't think this is network limit, since if you remove call to cellFsWrite() [here](https://github.com/aldostools/webMAN-MOD/blob/f450445aeac71d69fe82ad6289a60da30c07630c/include/ftp.h#L449)... 43.36 MiB/s for...

It does indeed seem reading/writing in a separate thread would make a nice speedup: When transferring 324MiB file Average recv() time: 2.970ms Average cellFsWrite() time: 4.218ms This means there's around...

So, after implementing a separate write thread for file uploads, I found out few things: - The cost of thread synchronization with semaphores is practically immeasurable. - Using 2 separate...