With NET_RESTART="yes" in /etc/sysconfig/apmd only the ONBOOT=yes devices will be started after resume. This seems somewhat wrong, although I am not sure what a better alternative could be.
Probably what you want is to have whatever was active at suspend time restarted. That seems to me to be the most reasonable alternative. The following question seems related: Why does my pcmcia card eth0 start when pcmcia is started, whether ONBOOT is set or not?
The issue with only ONBOOT devices being started has been fixed in 3.0.2-3. I have no idea what's causing your eth0 to not come up, works perfectly for me. Can you still reproduce this in 3.0.2-3?
>I have no idea what's causing your eth0 to not come up, works perfectly for me. No, my point is, even if ONBOOT is no, eth0 *does* start at boot time--not when /etc/rc5.d/S10network is run, but later when S45pcmcia is run. On a colleague's machine with a built-in wireless card and a pcmcia network card, this causes both interfaces to start, even when he configures the pc-card interface not to. This has been noted elsewhere (on mailing lists, if not in bugzilla). When S10network runs on a laptop, the network fails to start if the interface is a pc-card. It starts later, when S45pcmcia runs, but it does so whether ONBOOT is yes or no. I recall that there was a reason for the order, but i can't recall what it was. But as it stands, there seems to be no way to prevent a pcmcia interface from starting at boot. Is this a bug? Should it be filed anew?
Please file this as a new bug in kernel-pcmcia-cs [for the pcmcia init script] - apmd can't do anything about startup order. Closing this one because the original issue is fixed.
Already there, as it turns out: 47529 and 10202 and 44177 (and maybe more...). Thanks.