My configuration:
kernel 1.3.75
aha1542 + scsi HD + cdrom
WD80x3 ethernet card on a 2 machine net
On Mar 18, 10:16pm, "Leonard N. Zubkoff" wrote:
} There have been other reports of problems with recent 1.3.x kernels and I
} suspect there's a common thread here that's not yet understood. If interrupt
} latencies have actually gone through the roof as someone hypothesized, that
} could account for some of the timeout problems.
A while back, I remember someone complaining that they were noticing
large lags in response time in PPP. Just recently, I began to notice
this too--execpt it doesn't bother me, 'cause I don't watch what I type
anyway, and throughput has not gone down.
Also, I've noticed that sometimes my keystrokes on my dumb terminal
are not immediately acted upon--but I discounted this as either a minor
bug in screen. (It could also be a scheduling anomoly that affects screen.)
However, tonight, I noticed something that is a bit less acceptable.
I decided that I wanted to start with a fresh kernel source tree, so
I untar'ed patch level 60 (the only one I've got online) and applied
all the patches up to 74 in a loop... I was using screen on a serial
port, and everything seemed fine to me, but a user telneted in from a
dos machine complained that their session was totally locked up. I
checked, and it certainly was (even the NCSA telnet command keys didn't
work). When I stopped the patch process, their session immediately
recovered, and accepted all the keys pressed inbetween.
I got complaints several times more when I was actually compiling the
kernel later--but much more consistantly when applying the patches.
If I ctrl-z'ed the compile or patch or whatever, their session would
immediately unfreeze. Sounds to me like there's some kind of starvation
going on, but I've yet to be able to narrow it down more than that.
Steve
ssd@nevets.oau.org