Weak implementation of select() about elks HOT 4 CLOSED

from elks.

I don't know for now, as I am investigating and killing the bugs one after one 😄.

from elks.

To be exact, it is not the key push that wakes up the Ethernet driver. The driver is passive by nature, reacts to NE2K interrupts, and then wakes up the process (here ktcp) that was waiting for any incoming data during a blocking read() or select().

I am currently testing my fix to this select() weakness, and I can see now a correct workflow (no more endless looping) plus an improved performance with ktcp and telnetd, using busybox telnet 127.0.0.1 2323 on the host and QEMU for ELKS target.

But I can also see the third problem you reported : sometimes part of the output of the remote command is directed to the ELKS console in QEMU, and not to the remote client, even if not logged in ! I am going to open another issue to track that particular problem, because I guess it is not related to the select() implementation.

from elks.

Bad redirection tracked by #211.

from elks.