>Hello all,
> I don't want to come off like I'm sounding a siren, but I have
>noticed something odd with kernels 2.0 since sublevel 18 or so. I get
>unreasonable numbers of errors in the ifconfig count for my ppp0
>interface. Some background: I've been using the same hardware/phone
>line/ISP since about 1.3.50 and have never seen this before. I'm using
>ppp-2.2.0f as a module. Only since 2.0.18 (or thereabouts) have I been
>seeing *any* errors on an ifconfig report. After only a few hours of
>uptime, I have a report like this:
>ppp0 Link encap:Point-Point Protocol
> inet addr:204.157.220.226 P-t-P:204.157.220.8 Mask:255.255.255.0
> UP POINTOPOINT RUNNING MTU:296 Metric:1
> RX packets:21426 errors:217 dropped:217 overruns:0
> TX packets:13815 errors:0 dropped:0 overruns:0
I see this too. I can't remember ever seeing this in the past.
First I thought it happened because of a serial board swap on
a machine but I suspect it has something to do with serial overruns /
interrupt latency problems when receiving data. I see it on my 2 PPP
router machines: (machine1: Cyrix 486 DX2/66, 4 MB, machine 2: Intel
486/33 20 MB) Both run 2.0.20 and have 16550As and are connected with 115
Kb to USR Courier modems. Machine 1 gets an occasional error. Lets say
a few per week. Machine 2 considerably more. Something like 25 a day.
This is interesting because the bulk of the traffic goes from machine
2 to machine 1. Maybe the small memory of machine 2 plays a role ?
Greetings,
Rob van Nieuwkerk