vxge: fix 64 bit access on 32 bit platforms
Need to add stat_sync wrapper around 64 bit statistic values.
Fix wraparound bug in lockup detector where it is unsafely comparing
64 bit value that is not atomic. Since only care about detecting activity
just looking at current low order bits will work.
Remove unused entries in old vxge_sw_stats structure.
Change the error counters to unsigned long since they won't grow so large
as to have to be 64 bits.
Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
diff --git a/drivers/net/vxge/vxge-main.c b/drivers/net/vxge/vxge-main.c
index e658edd..54ca748 100644
--- a/drivers/net/vxge/vxge-main.c
+++ b/drivers/net/vxge/vxge-main.c
@@ -296,11 +296,13 @@
skb_record_rx_queue(skb, ring->driver_id);
skb->protocol = eth_type_trans(skb, ring->ndev);
+ u64_stats_update_begin(&ring->stats.syncp);
ring->stats.rx_frms++;
ring->stats.rx_bytes += pkt_length;
if (skb->pkt_type == PACKET_MULTICAST)
ring->stats.rx_mcast++;
+ u64_stats_update_end(&ring->stats.syncp);
vxge_debug_rx(VXGE_TRACE,
"%s: %s:%d skb protocol = %d",
@@ -592,8 +594,10 @@
vxge_hw_fifo_txdl_free(fifo_hw, dtr);
/* Updating the statistics block */
+ u64_stats_update_begin(&fifo->stats.syncp);
fifo->stats.tx_frms++;
fifo->stats.tx_bytes += skb->len;
+ u64_stats_update_end(&fifo->stats.syncp);
*done_skb++ = skb;
@@ -2630,11 +2634,16 @@
struct vxge_vpath *vpath;
struct vxge_ring *ring;
int i;
+ unsigned long rx_frms;
for (i = 0; i < vdev->no_of_vpath; i++) {
ring = &vdev->vpaths[i].ring;
+
+ /* Truncated to machine word size number of frames */
+ rx_frms = ACCESS_ONCE(ring->stats.rx_frms);
+
/* Did this vpath received any packets */
- if (ring->stats.prev_rx_frms == ring->stats.rx_frms) {
+ if (ring->stats.prev_rx_frms == rx_frms) {
status = vxge_hw_vpath_check_leak(ring->handle);
/* Did it received any packets last time */
@@ -2654,7 +2663,7 @@
}
}
}
- ring->stats.prev_rx_frms = ring->stats.rx_frms;
+ ring->stats.prev_rx_frms = rx_frms;
ring->last_status = status;
}
@@ -3125,14 +3134,36 @@
/* net_stats already zeroed by caller */
for (k = 0; k < vdev->no_of_vpath; k++) {
- net_stats->rx_packets += vdev->vpaths[k].ring.stats.rx_frms;
- net_stats->rx_bytes += vdev->vpaths[k].ring.stats.rx_bytes;
- net_stats->rx_errors += vdev->vpaths[k].ring.stats.rx_errors;
- net_stats->multicast += vdev->vpaths[k].ring.stats.rx_mcast;
- net_stats->rx_dropped += vdev->vpaths[k].ring.stats.rx_dropped;
- net_stats->tx_packets += vdev->vpaths[k].fifo.stats.tx_frms;
- net_stats->tx_bytes += vdev->vpaths[k].fifo.stats.tx_bytes;
- net_stats->tx_errors += vdev->vpaths[k].fifo.stats.tx_errors;
+ struct vxge_ring_stats *rxstats = &vdev->vpaths[k].ring.stats;
+ struct vxge_fifo_stats *txstats = &vdev->vpaths[k].fifo.stats;
+ unsigned int start;
+ u64 packets, bytes, multicast;
+
+ do {
+ start = u64_stats_fetch_begin(&rxstats->syncp);
+
+ packets = rxstats->rx_frms;
+ multicast = rxstats->rx_mcast;
+ bytes = rxstats->rx_bytes;
+ } while (u64_stats_fetch_retry(&rxstats->syncp, start));
+
+ net_stats->rx_packets += packets;
+ net_stats->rx_bytes += bytes;
+ net_stats->multicast += multicast;
+
+ net_stats->rx_errors += rxstats->rx_errors;
+ net_stats->rx_dropped += rxstats->rx_dropped;
+
+ do {
+ start = u64_stats_fetch_begin(&txstats->syncp);
+
+ packets = txstats->tx_frms;
+ bytes = txstats->tx_bytes;
+ } while (u64_stats_fetch_retry(&txstats->syncp, start));
+
+ net_stats->tx_packets += packets;
+ net_stats->tx_bytes += bytes;
+ net_stats->tx_errors += txstats->tx_errors;
}
return net_stats;