From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from e5.ny.us.ibm.com (e5.ny.us.ibm.com [32.97.182.145]) (using TLSv1 with cipher DHE-RSA-AES256-SHA (256/256 bits)) (Client CN "e5.ny.us.ibm.com", Issuer "Equifax" (verified OK)) by ozlabs.org (Postfix) with ESMTP id 16E2267D00 for ; Wed, 11 Oct 2006 07:18:22 +1000 (EST) Received: from d01relay04.pok.ibm.com (d01relay04.pok.ibm.com [9.56.227.236]) by e5.ny.us.ibm.com (8.13.8/8.12.11) with ESMTP id k9ALIJfm015667 for ; Tue, 10 Oct 2006 17:18:19 -0400 Received: from d01av04.pok.ibm.com (d01av04.pok.ibm.com [9.56.224.64]) by d01relay04.pok.ibm.com (8.13.6/8.13.6/NCO v8.1.1) with ESMTP id k9ALIJFh108332 for ; Tue, 10 Oct 2006 17:18:19 -0400 Received: from d01av04.pok.ibm.com (loopback [127.0.0.1]) by d01av04.pok.ibm.com (8.12.11.20060308/8.13.3) with ESMTP id k9ALIJHn029251 for ; Tue, 10 Oct 2006 17:18:19 -0400 Date: Tue, 10 Oct 2006 16:18:18 -0500 To: akpm@osdl.org Subject: [PATCH 17/21]: powerpc/cell spidernet reduce DMA kicking Message-ID: <20061010211818.GN4381@austin.ibm.com> References: <20061010204946.GW4381@austin.ibm.com> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii In-Reply-To: <20061010204946.GW4381@austin.ibm.com> From: linas@austin.ibm.com (Linas Vepstas) Cc: jeff@garzik.org, Arnd Bergmann , netdev@vger.kernel.org, James K Lewis , linux-kernel@vger.kernel.org, linuxppc-dev@ozlabs.org List-Id: Linux on PowerPC Developers Mail List List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , The current code attempts to start the TX dma every time a packet is queued. This is too conservative, and wastes CPU time. This patch changes behaviour to call the kick-dma function less often, only when the tx queue is at risk of emptying. This reduces cpu usage, improves performance. Signed-off-by: Linas Vepstas Cc: James K Lewis Cc: Arnd Bergmann ---- drivers/net/spider_net.c | 19 +++++++++++++------ 1 file changed, 13 insertions(+), 6 deletions(-) Index: linux-2.6.18-mm2/drivers/net/spider_net.c =================================================================== --- linux-2.6.18-mm2.orig/drivers/net/spider_net.c 2006-10-10 13:21:41.000000000 -0500 +++ linux-2.6.18-mm2/drivers/net/spider_net.c 2006-10-10 13:28:43.000000000 -0500 @@ -698,7 +698,7 @@ spider_net_prepare_tx_descr(struct spide return 0; } -static void +static int spider_net_set_low_watermark(struct spider_net_card *card) { unsigned long flags; @@ -719,7 +719,7 @@ spider_net_set_low_watermark(struct spid /* If TX queue is short, don't even bother with interrupts */ if (cnt < card->tx_desc/4) - return; + return cnt; /* Set low-watermark 3/4th's of the way into the queue. */ descr = card->tx_chain.tail; @@ -735,6 +735,7 @@ spider_net_set_low_watermark(struct spid card->low_watermark->dmac_cmd_status & ~SPIDER_NET_DESCR_TXDESFLG; card->low_watermark = descr; spin_unlock_irqrestore(&card->tx_chain.lock, flags); + return cnt; } /** @@ -819,8 +820,12 @@ spider_net_release_tx_chain(struct spide * @card: card structure * @descr: descriptor address to enable TX processing at * - * spider_net_kick_tx_dma writes the current tx chain head as start address - * of the tx descriptor chain and enables the transmission DMA engine + * This routine will start the transmit DMA running if + * it is not already running. This routine ned only be + * called when queueing a new packet to an empty tx queue. + * Writes the current tx chain head as start address + * of the tx descriptor chain and enables the transmission + * DMA engine. */ static inline void spider_net_kick_tx_dma(struct spider_net_card *card) @@ -860,6 +865,7 @@ out: static int spider_net_xmit(struct sk_buff *skb, struct net_device *netdev) { + int cnt; struct spider_net_card *card = netdev_priv(netdev); struct spider_net_descr_chain *chain = &card->tx_chain; @@ -873,8 +879,9 @@ spider_net_xmit(struct sk_buff *skb, str return NETDEV_TX_BUSY; } - spider_net_set_low_watermark(card); - spider_net_kick_tx_dma(card); + cnt = spider_net_set_low_watermark(card); + if (cnt < 5) + spider_net_kick_tx_dma(card); return NETDEV_TX_OK; }