| 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
| ||
|
Date: Mon, 23 Apr 2012 15:18:39 +0300
From: Yevgeny Petrilin <yevgenyp@...lanox.co.il>
To: <davem@...emloft.net>
CC: <netdev@...r.kernel.org>, <yevgenyp@...lanox.co.il>
Subject: [PATCH net-next 2/3] mlx4_en: Moving to Interrupts for TX completions
Moving to interrupts instead of polling fpr TX completions
Avoiding situations where skb can be held in by the driver for
a long time (till timer expires).
The change is also necessary for supporting BQL.
Removing comp_lock that was required because we could handle TX
completions from several contexts: Interrupts, timer, polling.
Now there is only interrupts
Signed-off-by: Yevgeny Petrilin <yevgenyp@...lanox.co.il>
---
drivers/net/ethernet/mellanox/mlx4/en_cq.c | 14 +----
drivers/net/ethernet/mellanox/mlx4/en_netdev.c | 4 ++
drivers/net/ethernet/mellanox/mlx4/en_tx.c | 59 +-----------------------
drivers/net/ethernet/mellanox/mlx4/mlx4_en.h | 6 +--
4 files changed, 9 insertions(+), 74 deletions(-)
diff --git a/drivers/net/ethernet/mellanox/mlx4/en_cq.c b/drivers/net/ethernet/mellanox/mlx4/en_cq.c
index 00b8127..908a460 100644
--- a/drivers/net/ethernet/mellanox/mlx4/en_cq.c
+++ b/drivers/net/ethernet/mellanox/mlx4/en_cq.c
@@ -124,11 +124,7 @@ int mlx4_en_activate_cq(struct mlx4_en_priv *priv, struct mlx4_en_cq *cq,
cq->mcq.comp = cq->is_tx ? mlx4_en_tx_irq : mlx4_en_rx_irq;
cq->mcq.event = mlx4_en_cq_event;
- if (cq->is_tx) {
- init_timer(&cq->timer);
- cq->timer.function = mlx4_en_poll_tx_cq;
- cq->timer.data = (unsigned long) cq;
- } else {
+ if (!cq->is_tx) {
netif_napi_add(cq->dev, &cq->napi, mlx4_en_poll_rx_cq, 64);
napi_enable(&cq->napi);
}
@@ -151,16 +147,12 @@ void mlx4_en_destroy_cq(struct mlx4_en_priv *priv, struct mlx4_en_cq *cq)
void mlx4_en_deactivate_cq(struct mlx4_en_priv *priv, struct mlx4_en_cq *cq)
{
- struct mlx4_en_dev *mdev = priv->mdev;
-
- if (cq->is_tx)
- del_timer(&cq->timer);
- else {
+ if (!cq->is_tx) {
napi_disable(&cq->napi);
netif_napi_del(&cq->napi);
}
- mlx4_cq_free(mdev->dev, &cq->mcq);
+ mlx4_cq_free(priv->mdev->dev, &cq->mcq);
}
/* Set rx cq moderation parameters */
diff --git a/drivers/net/ethernet/mellanox/mlx4/en_netdev.c b/drivers/net/ethernet/mellanox/mlx4/en_netdev.c
index bfcfd8a..35ad097 100644
--- a/drivers/net/ethernet/mellanox/mlx4/en_netdev.c
+++ b/drivers/net/ethernet/mellanox/mlx4/en_netdev.c
@@ -667,6 +667,10 @@ int mlx4_en_start_port(struct net_device *dev)
mlx4_en_deactivate_cq(priv, cq);
goto tx_err;
}
+
+ /* Arm CQ for TX completions */
+ mlx4_en_arm_cq(priv, cq);
+
/* Set initial ownership of all Tx TXBBs to SW (1) */
for (j = 0; j < tx_ring->buf_size; j += STAMP_STRIDE)
*((u32 *) (tx_ring->buf + j)) = 0xffffffff;
diff --git a/drivers/net/ethernet/mellanox/mlx4/en_tx.c b/drivers/net/ethernet/mellanox/mlx4/en_tx.c
index d9bab53..2d49342 100644
--- a/drivers/net/ethernet/mellanox/mlx4/en_tx.c
+++ b/drivers/net/ethernet/mellanox/mlx4/en_tx.c
@@ -67,8 +67,6 @@ int mlx4_en_create_tx_ring(struct mlx4_en_priv *priv,
inline_thold = min(inline_thold, MAX_INLINE);
- spin_lock_init(&ring->comp_lock);
-
tmp = size * sizeof(struct mlx4_en_tx_info);
ring->tx_info = vmalloc(tmp);
if (!ring->tx_info)
@@ -377,41 +375,12 @@ void mlx4_en_tx_irq(struct mlx4_cq *mcq)
{
struct mlx4_en_cq *cq = container_of(mcq, struct mlx4_en_cq, mcq);
struct mlx4_en_priv *priv = netdev_priv(cq->dev);
- struct mlx4_en_tx_ring *ring = &priv->tx_ring[cq->ring];
- if (!spin_trylock(&ring->comp_lock))
- return;
mlx4_en_process_tx_cq(cq->dev, cq);
- mod_timer(&cq->timer, jiffies + 1);
- spin_unlock(&ring->comp_lock);
+ mlx4_en_arm_cq(priv, cq);
}
-void mlx4_en_poll_tx_cq(unsigned long data)
-{
- struct mlx4_en_cq *cq = (struct mlx4_en_cq *) data;
- struct mlx4_en_priv *priv = netdev_priv(cq->dev);
- struct mlx4_en_tx_ring *ring = &priv->tx_ring[cq->ring];
- u32 inflight;
-
- INC_PERF_COUNTER(priv->pstats.tx_poll);
-
- if (!spin_trylock_irq(&ring->comp_lock)) {
- mod_timer(&cq->timer, jiffies + MLX4_EN_TX_POLL_TIMEOUT);
- return;
- }
- mlx4_en_process_tx_cq(cq->dev, cq);
- inflight = (u32) (ring->prod - ring->cons - ring->last_nr_txbb);
-
- /* If there are still packets in flight and the timer has not already
- * been scheduled by the Tx routine then schedule it here to guarantee
- * completion processing of these packets */
- if (inflight && priv->port_up)
- mod_timer(&cq->timer, jiffies + MLX4_EN_TX_POLL_TIMEOUT);
-
- spin_unlock_irq(&ring->comp_lock);
-}
-
static struct mlx4_en_tx_desc *mlx4_en_bounce_to_desc(struct mlx4_en_priv *priv,
struct mlx4_en_tx_ring *ring,
u32 index,
@@ -440,25 +409,6 @@ static struct mlx4_en_tx_desc *mlx4_en_bounce_to_desc(struct mlx4_en_priv *priv,
return ring->buf + index * TXBB_SIZE;
}
-static inline void mlx4_en_xmit_poll(struct mlx4_en_priv *priv, int tx_ind)
-{
- struct mlx4_en_cq *cq = &priv->tx_cq[tx_ind];
- struct mlx4_en_tx_ring *ring = &priv->tx_ring[tx_ind];
- unsigned long flags;
-
- /* If we don't have a pending timer, set one up to catch our recent
- post in case the interface becomes idle */
- if (!timer_pending(&cq->timer))
- mod_timer(&cq->timer, jiffies + MLX4_EN_TX_POLL_TIMEOUT);
-
- /* Poll the CQ every mlx4_en_TX_MODER_POLL packets */
- if ((++ring->poll_cnt & (MLX4_EN_TX_POLL_MODER - 1)) == 0)
- if (spin_trylock_irqsave(&ring->comp_lock, flags)) {
- mlx4_en_process_tx_cq(priv->dev, cq);
- spin_unlock_irqrestore(&ring->comp_lock, flags);
- }
-}
-
static int is_inline(struct sk_buff *skb, void **pfrag)
{
void *ptr;
@@ -590,7 +540,6 @@ netdev_tx_t mlx4_en_xmit(struct sk_buff *skb, struct net_device *dev)
struct mlx4_en_priv *priv = netdev_priv(dev);
struct mlx4_en_dev *mdev = priv->mdev;
struct mlx4_en_tx_ring *ring;
- struct mlx4_en_cq *cq;
struct mlx4_en_tx_desc *tx_desc;
struct mlx4_wqe_data_seg *data;
struct skb_frag_struct *frag;
@@ -638,9 +587,6 @@ netdev_tx_t mlx4_en_xmit(struct sk_buff *skb, struct net_device *dev)
ring->blocked = 1;
priv->port_stats.queue_stopped++;
- /* Use interrupts to find out when queue opened */
- cq = &priv->tx_cq[tx_ind];
- mlx4_en_arm_cq(priv, cq);
return NETDEV_TX_BUSY;
}
@@ -788,9 +734,6 @@ netdev_tx_t mlx4_en_xmit(struct sk_buff *skb, struct net_device *dev)
iowrite32be(ring->doorbell_qpn, ring->bf.uar->map + MLX4_SEND_DOORBELL);
}
- /* Poll CQ here */
- mlx4_en_xmit_poll(priv, tx_ind);
-
return NETDEV_TX_OK;
tx_drop:
diff --git a/drivers/net/ethernet/mellanox/mlx4/mlx4_en.h b/drivers/net/ethernet/mellanox/mlx4/mlx4_en.h
index 3879c5e..ed8bdd9 100644
--- a/drivers/net/ethernet/mellanox/mlx4/mlx4_en.h
+++ b/drivers/net/ethernet/mellanox/mlx4/mlx4_en.h
@@ -121,7 +121,7 @@ enum {
#define MLX4_EN_RX_COAL_TARGET 44
#define MLX4_EN_RX_COAL_TIME 0x10
-#define MLX4_EN_TX_COAL_PKTS 5
+#define MLX4_EN_TX_COAL_PKTS 16
#define MLX4_EN_TX_COAL_TIME 0x80
#define MLX4_EN_RX_RATE_LOW 400000
@@ -254,7 +254,6 @@ struct mlx4_en_tx_ring {
unsigned long bytes;
unsigned long packets;
unsigned long tx_csum;
- spinlock_t comp_lock;
struct mlx4_bf bf;
bool bf_enabled;
};
@@ -307,8 +306,6 @@ struct mlx4_en_cq {
spinlock_t lock;
struct net_device *dev;
struct napi_struct napi;
- /* Per-core Tx cq processing support */
- struct timer_list timer;
int size;
int buf_size;
unsigned vector;
@@ -529,7 +526,6 @@ void mlx4_en_deactivate_cq(struct mlx4_en_priv *priv, struct mlx4_en_cq *cq);
int mlx4_en_set_cq_moder(struct mlx4_en_priv *priv, struct mlx4_en_cq *cq);
int mlx4_en_arm_cq(struct mlx4_en_priv *priv, struct mlx4_en_cq *cq);
-void mlx4_en_poll_tx_cq(unsigned long data);
void mlx4_en_tx_irq(struct mlx4_cq *mcq);
u16 mlx4_en_select_queue(struct net_device *dev, struct sk_buff *skb);
netdev_tx_t mlx4_en_xmit(struct sk_buff *skb, struct net_device *dev);
--
1.7.7
--
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