Weird networking problems (caused by noapic?).
I'm using a cable modem connected to a NAT router connected to three
computers (2 Debian and 1 Windows), and I'm experiencing weird problems
with my main computer (Debian). I think I've eliminated the router
(since the other two computers always work and swapping them between
LAN ports doesn't affect the situation) and the network cable (by
replacing it). I get the same sporadic problem with the motherboard's
RJ45 and the PCI network card.
The problem is that occasionally the network -- or at least part of it
-- is unreachable. Sometimes I can ping a website host but the site
times out in my browser. Sometimes I can ping a POP3 server but not
download mail from it. Sometimes the internet is completely cut off.
Usually rebooting the router fixes it. Using "ifdown -a" and "ifup -a"
or "ifconfig eth0|eth1 up|down" sometimes makes things worse -- the
network can be unreachable after supposedly bringing it back up, and
sometimes (this is really embarrassing for a devoted GNU/Linux user)
rebooting the computer fixes the problem temporarily.
I occasionally find stuff like the following in /var/log/messages, but
not as often as I experience the problem.
Jan 23 11:43:19 garcia kernel: NETDEV WATCHDOG: eth0: transmit timed out
Jan 23 11:43:19 garcia kernel: eth0: Transmit timeout, status 00000004
I would be grateful for any suggestions, including further diagnostics
to run. One thing sticks in my mind: I recently started booting with
the "noapic" option, which is required to boot recent kernels on my
motherboard (ASRock K7S8X). (As discussed in the thread ``Kernel
installation woes on Athlon 1100: "hda: lost interrupt".'') Could this
affect the network this way?