The 3c59x driver, version v0.30c-all, with the dev->tbusy=0 patch, seems
stable. That is to say, I would have expected several of the dreaded
"transmitter access conflict" errors by now on another version, or
complete Tx hang with v0.40, by now.
Just as a reminder, v0.30c-all seems to fail to note that the transmitter
is no longer busy (dev->tbusy flag), if it encounters a transmission
timeout. All hell is then free to break out.
Oh, and here's a new socket in the "2.0.30/pre31 leaves sockets in CLOSE"
fiasco:
tcp 0 0 *:3333 *:* CLOSE
on (4.77/0)
This timer decrements, looping from 0s to 10s, ceaselessly so far....
Chris