: Fix occasional counter corruption on 5708.

PATCH Applied Michael Chan 2 participants 2 messages Jun 4, 2007 – Jun 5, 2007

Subsystems (2)

Networking Network drivers

lore

Share & tools

lore

Michael Chan

author

2007-06-04 23:19 UTC #0 lore raw

PATCH 4/5 : Fix occasional counter corruption on 5708.

[BNX2]: Fix occasional counter corruption on 5708.

The statistics block DMA on 5708 can be messed up occasionally on the average of about once per hour. If the user is reading the counters within one second after the corruption, the counters will be all messed up. One second later, the counters will be ok again until the next corruption occurs.

The workaround is to disable the periodic statistics DMA. Instead, we manually trigger the DMA once a second in bnx2_timer(). This manual trigger of the DMA avoids the problem.

As a consequence, we can only allow 0 or 1 second settings for ethtool -C statistics block.

Thanks to Jean-Daniel Pauget <[email protected]> and CaT <[email protected]> for reporting this rare problem.

Signed-off-by Michael Chan <[email protected]>

drivers/net/bnx2.c +13 -1

index c03282c..5654cc7 100644
@@ -3783,7 +3783,10 @@ bnx2_init_chip(struct bnx2 *bp)
 	REG_WR(bp, BNX2_HC_CMD_TICKS,
 	       (bp->cmd_ticks_int << 16) | bp->cmd_ticks);
 
-	REG_WR(bp, BNX2_HC_STATS_TICKS, bp->stats_ticks & 0xffff00);
+	if (CHIP_NUM(bp) == CHIP_NUM_5708)
+		REG_WR(bp, BNX2_HC_STATS_TICKS, 0);
+	else
+		REG_WR(bp, BNX2_HC_STATS_TICKS, bp->stats_ticks & 0xffff00);
 	REG_WR(bp, BNX2_HC_STAT_COLLECT_TICKS, 0xbb8);  /* 3ms */
 
 	if (CHIP_ID(bp) == CHIP_ID_5706_A1)

@@ -4636,6 +4639,11 @@ bnx2_timer(unsigned long data)
 
 	bp->stats_blk->stat_FwRxDrop = REG_RD_IND(bp, BNX2_FW_RX_DROP_COUNT);
 
+	/* workaround occasional corrupted counters */
+	if (CHIP_NUM(bp) == CHIP_NUM_5708 && bp->stats_ticks)
+		REG_WR(bp, BNX2_HC_COMMAND, bp->hc_cmd |
+					    BNX2_HC_COMMAND_STATS_NOW);
+
 	if (bp->phy_flags & PHY_SERDES_FLAG) {
 		if (CHIP_NUM(bp) == CHIP_NUM_5706)
 			bnx2_5706_serdes_timer(bp);

@@ -5446,6 +5454,10 @@ bnx2_set_coalesce(struct net_device *dev, struct ethtool_coalesce *coal)
 		0xff;
 
 	bp->stats_ticks = coal->stats_block_coalesce_usecs;
+	if (CHIP_NUM(bp) == CHIP_NUM_5708) {
+		if (bp->stats_ticks != 0 && bp->stats_ticks != USEC_PER_SEC)
+			bp->stats_ticks = USEC_PER_SEC;
+	}
 	if (bp->stats_ticks > 0xffff00) bp->stats_ticks = 0xffff00;
 	bp->stats_ticks &= 0xffff00;

David Miller

maintainer Applied

2007-06-05 04:24 UTC #1 lore raw

From: "Michael Chan" <[email protected]>
Date: Mon, 04 Jun 2007 17:01:08 -0700

Quoted 12 lines

[BNX2]: Fix occasional counter corruption on 5708.
The statistics block DMA on 5708 can be messed up occasionally on the average of about once per hour. If the user is reading the counters within one second after the corruption, the counters will be all messed up. One second later, the counters will be ok again until the next corruption occurs.
The workaround is to disable the periodic statistics DMA. Instead, we manually trigger the DMA once a second in bnx2_timer(). This manual trigger of the DMA avoids the problem.

Quoted 7 lines

As a consequence, we can only allow 0 or 1 second settings for ethtool -C statistics block.
Thanks to Jean-Daniel Pauget <[email protected]> and CaT <[email protected]> for reporting this rare problem.
Signed-off-by Michael Chan <[email protected]>

Applied.