Thread (3 messages) 3 messages, 3 authors, 2008-07-29

RE: [Patch 1/1 net-next] S2io: fix statistics flush after a MTU change

From: Ramkrishna Vepa <hidden>
Date: 2008-07-24 17:45:48

Jeff,

The patch looks good. Please accept.

Thanks,
Ram
-----Original Message-----
From: Breno Leitao [mailto:leitao@linux.vnet.ibm.com]
Sent: Tuesday, July 22, 2008 12:27 PM
To: netdev@vger.kernel.org
Cc: jeff@garzik.org; ram.vepa@neterion.com; Rastapur Santosh;
Sivakumar
Subramani; Sreenivasa Honnur; Jay Vosburgh
Subject: [Patch 1/1 net-next] S2io: fix statistics flush after a MTU
change

On s2io driver, when you change the interface MTU, it invokes a card
reset, which flush some statistics.  This patch solves this problem,
and
quoted hunk ↗ jump to hunk
also set the net_device->stats as the default statistics structure,
instead of s2io_nic->stats.

To do that, s2io_nic->stats turned into a staging area, where is saved
statistics of the last hardware statistics query. So, the difference
between the current hardware statistics and s2io_nic->stats, is the
value that should be summed up, in order to get the correct statistics
value, even after a reset.

Signed-off-by: Breno Leitao <redacted>
Signed-off-by: Jay Vosburgh <redacted>

----
diff --git a/drivers/net/s2io.c b/drivers/net/s2io.c
index 9dae40c..f92108e 100644
--- a/drivers/net/s2io.c
+++ b/drivers/net/s2io.c
@@ -3139,7 +3139,7 @@ static void tx_intr_handler(struct fifo_info
*fifo_data)
 		pkt_cnt++;

 		/* Updating the statistics block */
-		nic->stats.tx_bytes += skb->len;
+		nic->dev->stats.tx_bytes += skb->len;
 		nic->mac_control.stats_info->sw_stat.mem_freed += skb-
quoted
truesize;
 		dev_kfree_skb_irq(skb);
@@ -4892,25 +4892,42 @@ static struct net_device_stats
*s2io_get_stats(struct net_device *dev)
 	/* Configure Stats for immediate updt */
 	s2io_updt_stats(sp);

+	/* Using sp->stats as a staging area, because reset (due to mtu
+	   change, for example) will clear some hardware counters */
+	dev->stats.tx_packets +=
+		le32_to_cpu(mac_control->stats_info->tmac_frms) -
+		sp->stats.tx_packets;
 	sp->stats.tx_packets =
 		le32_to_cpu(mac_control->stats_info->tmac_frms);
+	dev->stats.tx_errors +=
+		le32_to_cpu(mac_control->stats_info->tmac_any_err_frms)
-
+		sp->stats.tx_errors;
 	sp->stats.tx_errors =
 		le32_to_cpu(mac_control->stats_info->tmac_any_err_frms);
+	dev->stats.rx_errors +=
+		le64_to_cpu(mac_control->stats_info->rmac_drop_frms) -
+		sp->stats.rx_errors;
 	sp->stats.rx_errors =
 		le64_to_cpu(mac_control->stats_info->rmac_drop_frms);
+	dev->stats.multicast =
+		le32_to_cpu(mac_control->stats_info->rmac_vld_mcst_frms)
-
+		sp->stats.multicast;
 	sp->stats.multicast =
le32_to_cpu(mac_control->stats_info->rmac_vld_mcst_frms);
+	dev->stats.rx_length_errors =
+		le64_to_cpu(mac_control->stats_info->rmac_long_frms) -
+		sp->stats.rx_length_errors;
 	sp->stats.rx_length_errors =
 		le64_to_cpu(mac_control->stats_info->rmac_long_frms);

 	/* collect per-ring rx_packets and rx_bytes */
-	sp->stats.rx_packets = sp->stats.rx_bytes = 0;
+	dev->stats.rx_packets = dev->stats.rx_bytes = 0;
 	for (i = 0; i < config->rx_ring_num; i++) {
-		sp->stats.rx_packets +=
mac_control->rings[i].rx_packets;
-		sp->stats.rx_bytes += mac_control->rings[i].rx_bytes;
+		dev->stats.rx_packets +=
mac_control->rings[i].rx_packets;
quoted hunk ↗ jump to hunk
+		dev->stats.rx_bytes += mac_control->rings[i].rx_bytes;
 	}

-	return (&sp->stats);
+	return (&dev->stats);
 }

 /**
@@ -7413,7 +7430,7 @@ static int rx_osm_handler(struct ring_info
*ring_data, struct RxD_t * rxdp)
 		if (err_mask != 0x5) {
 			DBG_PRINT(ERR_DBG, "%s: Rx error Value: 0x%x\n",
 				dev->name, err_mask);
-			sp->stats.rx_crc_errors++;
+			dev->stats.rx_crc_errors++;
 			sp->mac_control.stats_info->sw_stat.mem_freed
 				+= skb->truesize;
 			dev_kfree_skb(skb);
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help