rl(4) TX_UNDERRUN fix

Joerg Sonnenberger joerg at britannica.bec.de
Mon Feb 21 11:45:17 PST 2005


On Mon, Feb 21, 2005 at 08:39:22PM +0100, Max Laier wrote:
> On Monday 21 February 2005 19:45, Joerg Sonnenberger wrote:
> > Hi all,
> > please test http://leaf.dragonflybsd.org/~joerg/if_rl.c.diff
> > with any kind of rl(4). This patch is a more aggressive version
> > of the recently commited patch in FreeBSD 5/6, the version
> > orginaly proposed in kern/61448.
> >
> > It shouldn't negatively impact performance or stability.
> 
> I choose the less aggressive one for a (couple of) reason(s):
>  - not all chips can RL_TXCFG_CLRABRT some even forbid it in the specs.
>  - the ifp->if_opackets++ just isn't right. You should only increase it if you 
> get RL_TXSTAT_TX_OK.
>  - removing the reset/init cycle *might* be a good thing, but only if you can 
> assure that it doesn't break the older chips.  Some notes in the specs 
> suggest different.

I know and I tried the other patch too. But it doesn't fix the hangs for me.
E.g. the card still got stuck in the TX state at random times.

Joerg





More information about the Submit mailing list