| lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
|
Open Source and information security mailing list archives
| ||
|
Message-ID: <a80199422de16ae355e56ee1b2abc9b2bf91a7f6.1307029009.git.mst@redhat.com>
Date: Thu, 2 Jun 2011 18:43:17 +0300
From: "Michael S. Tsirkin" <mst@...hat.com>
To: linux-kernel@...r.kernel.org
Cc: Rusty Russell <rusty@...tcorp.com.au>,
Carsten Otte <cotte@...ibm.com>,
Christian Borntraeger <borntraeger@...ibm.com>,
linux390@...ibm.com, Martin Schwidefsky <schwidefsky@...ibm.com>,
Heiko Carstens <heiko.carstens@...ibm.com>,
Shirley Ma <xma@...ibm.com>, lguest@...ts.ozlabs.org,
linux-kernel@...r.kernel.org,
virtualization@...ts.linux-foundation.org, netdev@...r.kernel.org,
linux-s390@...r.kernel.org, kvm@...r.kernel.org,
Krishna Kumar <krkumar2@...ibm.com>,
Tom Lendacky <tahm@...ux.vnet.ibm.com>, steved@...ibm.com,
habanero@...ux.vnet.ibm.com
Subject: [PATCHv2 RFC 3/4] virtio_net: limit xmit polling
Current code might introduce a lot of latency variation
if there are many pending bufs at the time we
attempt to transmit a new one. This is bad for
real-time applications and can't be good for TCP either.
Free up just enough to both clean up all buffers
eventually and to be able to xmit the next packet.
Signed-off-by: Michael S. Tsirkin <mst@...hat.com>
---
drivers/net/virtio_net.c | 106 +++++++++++++++++++++++++++++-----------------
1 files changed, 67 insertions(+), 39 deletions(-)
diff --git a/drivers/net/virtio_net.c b/drivers/net/virtio_net.c
index a0ee78d..b25db1c 100644
--- a/drivers/net/virtio_net.c
+++ b/drivers/net/virtio_net.c
@@ -509,17 +509,33 @@ again:
return received;
}
-static void free_old_xmit_skbs(struct virtnet_info *vi)
+static bool free_old_xmit_skb(struct virtnet_info *vi)
{
struct sk_buff *skb;
unsigned int len;
- while ((skb = virtqueue_get_buf(vi->svq, &len)) != NULL) {
- pr_debug("Sent skb %p\n", skb);
- vi->dev->stats.tx_bytes += skb->len;
- vi->dev->stats.tx_packets++;
- dev_kfree_skb_any(skb);
- }
+ skb = virtqueue_get_buf(vi->svq, &len);
+ if (unlikely(!skb))
+ return false;
+ pr_debug("Sent skb %p\n", skb);
+ vi->dev->stats.tx_bytes += skb->len;
+ vi->dev->stats.tx_packets++;
+ dev_kfree_skb_any(skb);
+ return true;
+}
+
+/* Check capacity and try to free enough pending old buffers to enable queueing
+ * new ones. Return true if we can guarantee that a following
+ * virtqueue_add_buf will succeed. */
+static bool free_xmit_capacity(struct virtnet_info *vi)
+{
+ struct sk_buff *skb;
+ unsigned int len;
+
+ while (virtqueue_min_capacity(vi->svq) < MAX_SKB_FRAGS + 2)
+ if (unlikely(!free_old_xmit_skb))
+ return false;
+ return true;
}
static int xmit_skb(struct virtnet_info *vi, struct sk_buff *skb)
@@ -572,30 +588,34 @@ static int xmit_skb(struct virtnet_info *vi, struct sk_buff *skb)
static netdev_tx_t start_xmit(struct sk_buff *skb, struct net_device *dev)
{
struct virtnet_info *vi = netdev_priv(dev);
- int capacity;
-
- /* Free up any pending old buffers before queueing new ones. */
- free_old_xmit_skbs(vi);
-
- /* Try to transmit */
- capacity = xmit_skb(vi, skb);
-
- /* This can happen with OOM and indirect buffers. */
- if (unlikely(capacity < 0)) {
- if (net_ratelimit()) {
- if (likely(capacity == -ENOMEM)) {
- dev_warn(&dev->dev,
- "TX queue failure: out of memory\n");
- } else {
- dev->stats.tx_fifo_errors++;
+ int ret, i;
+
+ /* We normally do have space in the ring, so try to queue the skb as
+ * fast as possible. */
+ ret = xmit_skb(vi, skb);
+ if (unlikely(ret < 0)) {
+ /* This triggers on the first xmit after ring full condition.
+ * We need to free up some skbs first. */
+ if (likely(free_xmit_capacity(vi))) {
+ ret = xmit_skb(vi, skb);
+ /* This should never fail. Check, just in case. */
+ if (unlikely(ret < 0)) {
dev_warn(&dev->dev,
"Unexpected TX queue failure: %d\n",
- capacity);
+ ret);
+ dev->stats.tx_fifo_errors++;
+ dev->stats.tx_dropped++;
+ kfree_skb(skb);
+ return NETDEV_TX_OK;
}
+ } else {
+ /* Ring full: it might happen if we get a callback while
+ * the queue is still mostly full. This should be
+ * extremely rare. */
+ dev->stats.tx_dropped++;
+ kfree_skb(skb);
+ goto stop;
}
- dev->stats.tx_dropped++;
- kfree_skb(skb);
- return NETDEV_TX_OK;
}
virtqueue_kick(vi->svq);
@@ -603,18 +623,26 @@ static netdev_tx_t start_xmit(struct sk_buff *skb, struct net_device *dev)
skb_orphan(skb);
nf_reset(skb);
- /* Apparently nice girls don't return TX_BUSY; stop the queue
- * before it gets out of hand. Naturally, this wastes entries. */
- if (capacity < 2+MAX_SKB_FRAGS) {
- netif_stop_queue(dev);
- if (unlikely(!virtqueue_enable_cb_delayed(vi->svq))) {
- /* More just got used, free them then recheck. */
- free_old_xmit_skbs(vi);
- capacity = virtqueue_min_capacity(vi->svq);
- if (capacity >= 2+MAX_SKB_FRAGS) {
- netif_start_queue(dev);
- virtqueue_disable_cb(vi->svq);
- }
+ /* We transmit one skb, so try to free at least two pending skbs.
+ * This is so that we don't hog the skb memory unnecessarily. *
+ * Doing this after kick means there's a chance we'll free
+ * the skb we have just sent, which is hot in cache. */
+ for (i = 0; i < 2; i++)
+ free_old_xmit_skb(v);
+
+ if (likely(free_xmit_capacity(vi)))
+ return NETDEV_TX_OK;
+
+stop:
+ /* Apparently nice girls don't return TX_BUSY; check capacity and stop
+ * the queue before it gets out of hand.
+ * Naturally, this wastes entries. */
+ netif_stop_queue(dev);
+ if (unlikely(!virtqueue_enable_cb_delayed(vi->svq))) {
+ /* More just got used, free them and recheck. */
+ if (free_xmit_capacity(vi)) {
+ netif_start_queue(dev);
+ virtqueue_disable_cb(vi->svq);
}
}
--
1.7.5.53.gc233e
--
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