Question: Machine Reboots on high volume network traffic 

Forum: PCMCIA Installation and Configuration Issues
Date: 1999, Nov 29
From: Jason M. Sullivan Jason0x21

Basic Information Texas Instruments TravelMate 5200 RedHat 6.0 Kernel Version 2.2.5-22 (i586) PCMCIA Version 3.0.9 D-Link DE-660 So I got the PCMCIA package and cardmgr running (I needed to turn off IO probing. It took a while to figure it out because the message that the pcmcia core was loading correctly didn't make it into the messages log before the cardmgr caused the machine to hang with the io probe). Now that I have the core and cardmgr running, I have a new problem. I can pop my Network card in, the core recongnizes it, I get two high beeps, and the network starts up fine. I can rlogin, mount, and open an ftp session. I can probably start a bunch of other things, but when the traffic starts getting heavy (like when I start to transfer a file: 'cat' in rlogin, 'get' in ftp, or simply 'cp' from a mounted directory), the machine immediately reboots. No warning, no messages. I tried a few options (fast_pci, and cmd_time), to no avail. The kernel did have some strange messages on insertion and removal, summarized below.. Nov 22 22:03:00 billbee kernel: eth0: NE2000 Compatible: port 0x1000, irq 10, hw_addr 00:80:C8:8C:0A:E5 Nov 22 22:03:00 billbee cardmgr[370]: executing: './network start eth0' Nov 22 22:03:01 billbee inet: inetd startup succeeded Nov 22 22:03:01 billbee kernel: Hw. address read/write mismap 3 Nov 22 22:03:01 billbee kernel: Hw. address read/write mismap 5 Nov 22 22:03:01 billbee nfs: Starting NFS services: succeeded Nov 22 22:03:01 billbee cardmgr[370]: + SIOCADDRT: Invalid argument ... Nov 22 22:03:57 billbee kernel: eth0: unexpected TX-done interrupt, lasttx=0. Nov 22 22:03:57 billbee last message repeated 3 times Nov 22 22:03:57 billbee kernel: eth0: Too much work at interrupt, status 0x20 Nov 22 22:03:58 billbee kernel: eth0: interrupt from stopped card Nov 22 22:03:58 billbee cardmgr[370]: shutting down socket 1 Nov 22 22:03:58 billbee cardmgr[370]: executing: './network stop eth0' Nov 22 22:03:58 billbee kernel: alloc_skb called nonatomically from interrupt c016aebd Nov 22 22:03:58 billbee kernel: alloc_skb called nonatomically from interrupt c01509b9 Nov 22 22:03:58 billbee cardmgr[370]: executing: 'rmmod pcnet_cs' Nov 22 22:03:58 billbee cardmgr[370]: executing: 'rmmod 8390' Any clues out there? Thanks in advance.

[ Next-in-Thread ]  [ Next Message ] 

Messages Inline: [ 0 ]  [ 1 ] 

Disabling IO probing is a bad idea Re: Question: Machine Reboots on high volume network traffic (Jason M. Sullivan) Date: 1999, Nov 29 From: David Hinds <dhinds@pcmcia.sourceforge.org>

> So I got the PCMCIA package and cardmgr running (I needed to turn off IO > probing. Disabling IO port probing is a very bad idea unless you're very sure about what you're doing. My guess is that you've now got an IO port conflict, which is causing reads and writes to the card to sometimes have unexpected results. Rather than disabling the IO probe, it would be much better to play with the IO port windows in /etc/pcmcia/config.opts to find a setting that doesn't cause a freeze. -- Dave

More: Working now, but one last question Re: Disabling IO probing is a bad idea (David Hinds) Date: 1999, Nov 30 From: Jason M. Sullivan Jason0x21

Ok, I got the card working with a minimal amount of frittering around, but one question remains. The port scan picked up a port that seemed to work (or worked much better than the previous ones), but my memory map lists that as in use by MIDI. I suppose I need to check and see if I can play a MIDI tune while accssing the nextwork, but the question is this: How does the port actually get asigned? Does it just look for a "free" port?, and tell the card where to respond? If that's the case, I'm guessing I should probably exclude the MIDI port? Thanks!

Clean resource allocation is hard Re: More: Working now, but one last question (Jason M. Sullivan) Date: 1999, Nov 30 From: David Hinds <dhinds@pcmcia.sourceforge.org>

The drivers try to avoid clashes with known devices, but the Linux kernel currently does not make this particularly easy. If you have a fairly recent PCMCIA package and compile it with PnP BIOS support enabled, it will generally do a better job of this, because the BIOS may know more about the resources used by your devices than the Linux kernel. However, PnP BIOS calls cause problems on some systems, so I don't enable this by default. You should exclude the MIDI ports if the PCMCIA drivers are trying to reuse them. -- Dave

More: Time for me to talk to TI/Acer Re: Clean resource allocation is hard (David Hinds) Date: 1999, Nov 30 From: Jason M. Sullivan Jason0x21

Thanks for the info. I did more checking on the websites that TI/Acer have for the TravelMate series, and in their doumentation for setting up a PCMCIA ethernet card, they say to use the very port that the linux pcmcia package is picking, so it looks like there's a conflict in their documentation as well. :-) Looks like I'm gonna be bugging Acer now. Thanks again!

[ Add Message ]  to: "Machine Reboots on high volume network traffic"

[ Members ]  [ Subscribe ]  [ Admin Mode ] 
[ Show Frames ]  [ Help for HyperNews at pcmcia-cs.sourceforge.net 1.10 ]