Re: tulip driver: errors instead TX packets?

[Adam Kropelin]

On Sat, Jan 10, 2004 at 09:28:45PM +0100, Piotr Kaczuba wrote:
> Jeff Garzik wrote:
> > Piotr Kaczuba wrote:
> > 
> >> I've got a ADMtek Centaur (3cSOHO100B-TX) running with the tulip 
> >> driver  on 2.6.1. I wonder if anyone has noticed that ifconfig shows 
> >> the  packets sent in the errors field instead of the TX packets field. 
> >> At  least, this is what I assume because it shows 0 TX packets and 
> >> 11756  errors.
> > 
> > This is an old error, but since packets show up, nobody bothers with the 
> > incorrect statistics...
> 
> It seems that the error lies in the following piece of code from 
> drivers/net/tulip/interrupt.c, function tulip_interrupt. I've inserted 
> an additional printk after the "if (status & 0x8000)" and it looks like 
> normal operation of the nic is considered as an major error by the 
> driver because my printk appeared in dmesg output right after bringing 
> the interface up. I assume that the else branch is never executed 
> although I didn't test what happens if an transmit error really happens. 
> I wonder if a fix for this problem consists of just changing the value 
> of the AND mask but I have no idea what the right value would be.
> 
> 
>    if (status & 0x8000) {

According to my PNIC docs (don't have true Tulip docs handy), this bit
is the logical OR of all the Tx error bits, so it appears a true tx
error has ocurred.

>            /* There was an major error, log it. */
> #ifndef final_version
>            if (tulip_debug > 1)
>               printk(KERN_DEBUG "%s: Transmit error, Tx status %8.8x.\n",
>                         dev->name, status);
> #endif

Enabling this printk and setting tulip_debug appropriately would tell us
what tx error you're experiencing.

>            tp->stats.tx_errors++;
>            if (status & 0x4104) tp->stats.tx_aborted_errors++;
>            if (status & 0x0C00) tp->stats.tx_carrier_errors++;
>            if (status & 0x0200) tp->stats.tx_window_errors++;
>            if (status & 0x0002) tp->stats.tx_fifo_errors++;
>            if ((status & 0x0080) && tp->full_duplex == 0)
>                    tp->stats.tx_heartbeat_errors++;

And this code bumps error counters based on the specific error that
ocurred.  It all seems perfectly sane to me. The actual problem lies
elsewhere, I think.

--Adam

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/