I got a 'new' Thinkpad 560Z (Pentium II) to be my 'learn Debian' machine, partly so that all the desktop machines I administer can eventually become Debian (I have run RedHat for 8 years but RPM upgrade hell is trumping my inertia).
But the eth0 device vanishes for no reason that I can determine. (The X server also hangs the machine regularly but I'm hoping that running 4.2 from testing will fix it or maybe trying to improve the config file that I grabbed almost verbatim from the web.) eth0 is bound to a Xircom XEM5600 (combo 10/100 Ethernet + 56K Modem, not CardBus whatever that's called) and gets its network parameters via dhcp. Everything will be fine for maybe half an hour. Then I notice that new ssh connections don't work, and old ones are hung. So I run 'ifconfig' and find that eth0 is no longer listed. No errors in the system log or on the console. After 'ifup eth0' network connectivity is back. The card's Ethernet hardware is fine because in my TP 560 (2.4.18 kernel) I use it for days or weeks continuously with no problem. I noticed this problem when I installed the latest 3.0 release (stable), using the stock 2.2 kernel. So I compiled a kernel of my own (2.4.20), doing it the Debian wayd and deriving the config from one that I use happily for 2.4.18 on my TP 560 but adding CardBus support for example. After telling modules.conf that the module ds.o requires yenta_socket.o, cardmgr/cardctl etc. work fine, but alas, the eth0 problem is still there (but hibernation/suspend/standby work). Any thoughts? Is it a TP 560Z problem or a Debian problem or a general PCMCIA driver problem? I could do a controlled experiment by installing RH 7.3 on the 560Z (which is what the 560 runs) but if anyone has any quick thoughts I'd be grateful (and will summarize to the list). -Sanjoy
