Re: problem on debian
Bogdan Costescu <[email protected]>
| Newsgroups | gmane.linux.drivers.vortex.devel |
|---|---|
| Message-ID | <Pine.LNX.4.44.0303281300310.15859-100000@kenzo.iwr.uni-heidelberg.de> |
On Thu, 27 Mar 2003, [iso-8859-1] Jürgen Sandner wrote: > kernel: eth0: Too much work in interrupt, status e401 Trying to use a search engine to look for the message (without the numerical value at the end) gives _lots_ of results. Even with the numerical value (which is 3c59x specific) you still get enough. As I often repeated on this list, the problem is that the system becomes too slow to process the network packets. This might mean that the system is too slow in general or that the network interrupt doesn't come through when it's raised, the CPU being blocked for too long in other interrupts. There are two ways of dealing with this: - try to make it happen less frequent: increase the "max_interrupt_work" module parameter, maybe coupled with increasing the Rx (and maybe Tx) ring sizes (this requires modifying the driver and recompiling) - try to fix the problem. Based on the interrupt list, my guess is that > 9: 875049 XT-PIC aic7xxx, usb-uhci is the place where the CPU is spending too much time. I would suggest to first try to get the two devices to not share an interrupt (suggestions for this were also given on this list and many others). Then try to find out why the device spends too much time there - changing the SCSI controller to a different brand (and driver) might help here. Saying the the 3Com card/driver is at fault is not realistic - as you can find through search engine results, these network cards are not the only ones giving this message... -- Bogdan Costescu IWR - Interdisziplinaeres Zentrum fuer Wissenschaftliches Rechnen Universitaet Heidelberg, INF 368, D-69120 Heidelberg, GERMANY Telephone: +49 6221 54 8869, Telefax: +49 6221 54 8868 E-mail: [email protected] _______________________________________________ vortex mailing list [email protected] http://www.scyld.com/mailman/listinfo/vortex