On Wed, 17 Dec 2003 13:32:23 +0100 "Huber, Matthias S-RD-CP1" <Matthias.Huber@heidelberg.com> wrote: > > > I still have got problems with the new 8139too NAPI driver. > > > On my network client, AMD K6-2 333MHz with 2.4.23 and > > 8139too NAPI, I no longer get "too much work at interrupt", > > but ifconfig still reports 1% (!) packet drops while cat'ing > > a file from nfs to /dev/null (help me: how do I better > > measure network performance?). From time to time I see a "nfs > > server not responding" in the syslog. Network performance is > > really slow, ~16 MBit/s while mii-tool says 100-TX-FD. > > > Corresponding server is a non-8139 Athlon XP 2000+ with > > 100MBit, in between a 100MBit switch. Server networking seems > > ok (tested from Windows, but with smb, not nfs). > > > Please, could anybody guess what my problem is? > > > > I had problems with NFS+8139 on an a similar machine, it was fixed > > by using PIO instead of MMIO. The help text mentions something about > > memory inconsistencies IIRC. The machine was the server, though. > > > > I tried PIO, but with no effect. I'm still really clueless... > Any other tips are wellcome! > Thanks, > Matthias You could try the experimental 64k receive ring to see if that helps. Also with NFS what are you can tune the rsize,wsize to reduce getting overrun. You might also check that the the switch and the board agree about running full/half duplex. - : send the line "unsubscribe linux-net" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
