[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <1275537919.29413.55.camel@edumazet-laptop>
Date: Thu, 03 Jun 2010 06:05:19 +0200
From: Eric Dumazet <eric.dumazet@...il.com>
To: sonic zhang <sonic.adi@...il.com>
Cc: David Miller <davem@...emloft.net>,
netdev <netdev@...r.kernel.org>,
uclinux-dist-devel <uclinux-dist-devel@...ckfin.uclinux.org>
Subject: Re: [PATCH v2] netdev:bfin_mac: reclaim and free tx skb as soon as
possible after transfer
Le jeudi 03 juin 2010 à 11:48 +0800, sonic zhang a écrit :
> bfin_tx_hwtstamp(dev, skb);
>
> current_tx_ptr = current_tx_ptr->next;
> dev->stats.tx_packets++;
> dev->stats.tx_bytes += (skb->len);
> +
> + tx_reclaim_skb(lp);
> +
> return NETDEV_TX_OK;
> }
>
Not related to your patch, but reviewing it I see this driver still do
the "dev->stats.tx_packets++; dev->stats.tx_bytes += (skb->len);"
This is not necessary and expensive, since we update txq stats in core
network stack.
rc = ops->ndo_start_xmit(skb, dev);
if (rc == NETDEV_TX_OK)
txq_trans_update(txq); << here >>
Doing these changes in a network driver dirties yet another cache line
in a hot path.
I suspect many drivers could be changed to avoid this double accounting.
A driver providing a ndo_get_stats() method can use the
dev_txq_stats_fold() helper.
driver not provinding a ndo_get_stats() method has nothing special to
do, dev_get_stats() automatically calls dev_txq_stats_fold() to update
dev->stats.{tx_packets|tx_bytes|tx_dropped} using txq stats.
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Powered by blists - more mailing lists