netdev
[Top] [All Lists]

Re: zerocopy changes in 3c59x.c

To: andrewm@xxxxxxxxxx (Andrew Morton)
Subject: Re: zerocopy changes in 3c59x.c
From: kuznet@xxxxxxxxxxxxx
Date: Fri, 26 Jan 2001 20:46:03 +0300 (MSK)
Cc: netdev@xxxxxxxxxxx
In-reply-to: <3A714788.82C064BD@xxxxxxxxxx> from "Andrew Morton" at Jan 26, 1 08:46:48 pm
Sender: owner-netdev@xxxxxxxxxxx
Hello!

>  /* Tx timeout interval (millisecs) */
> -static int watchdog = 400;
> +static int watchdog = 5000;
> 
> Five second transmit timeout.  Why is this?

I would ask: why 400msec? 8)

All the drivers use much larger values. 5 seconds is good one.


Actually, I changed this by _your_ advice? Do you remember? 8)
BTW it really helped to avoid watchdog, when card had some spurious stalls
due to excessive collisions or those link problems.

>             vp->stats.rx_compressed++;
> 
> each time we send a frame with hardware checksum?  This is the
> transmit path, not the receive path.  It seems we're abusing
> some ppp-related stats here.  What's up?

Is 3com ppp driver? 8) No? Then you can reuse this value
for any other purposes.

All this debugging statistics should be deteted. I removed it from acenic,
for example. It was interesting only at the first stage.

Alexey

<Prev in Thread] Current Thread [Next in Thread>