[PATCH net-next V2 0/4] mlx4 misc cleanups and improvements

[PATCH net-next V2 0/4] mlx4 misc cleanups and improvements

[Tariq Toukan]

Hi Dave,

This patchset contains some cleanups and improvements from the team
to the mlx4 Eth and core drivers.

Series generated against net-next commit:
5a7a5555a362 'net sched: stylistic cleanups'

Thanks,
Tariq.

V2:
* dropped patch 5, will be submitted separately to 'net'.

Jack Morgenstein (1):
  net/mlx4_core: Fix deadlock when switching between polling and event
    fw commands

Kamal Heib (1):
  net/mlx4_en: Fix wrong indentation

Leon Romanovsky (1):
  net/mlx4_core: Use RCU to perform radix tree lookup for SRQ

Tariq Toukan (1):
  net/mlx4_en: Add branch prediction hints in RX data-path

 drivers/net/ethernet/mellanox/mlx4/cmd.c   | 23 +++++++++++++++++------
 drivers/net/ethernet/mellanox/mlx4/en_rx.c | 28 +++++++++++++++-------------
 drivers/net/ethernet/mellanox/mlx4/mlx4.h  |  2 ++
 drivers/net/ethernet/mellanox/mlx4/srq.c   | 14 +++++---------
 4 files changed, 39 insertions(+), 28 deletions(-)

-- 
1.8.3.1

[PATCH net-next V2 1/4] net/mlx4_en: Add branch prediction hints in RX data-path

[Tariq Toukan]

Add likely/unlikely hints to improve branch predictions
in the RX data-path.

Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
---
 drivers/net/ethernet/mellanox/mlx4/en_rx.c | 26 ++++++++++++++------------
 1 file changed, 14 insertions(+), 12 deletions(-)

diff --git a/drivers/net/ethernet/mellanox/mlx4/en_rx.c b/drivers/net/ethernet/mellanox/mlx4/en_rx.c
index c80073e4947f..e60a86a9266d 100644
--- a/drivers/net/ethernet/mellanox/mlx4/en_rx.c
+++ b/drivers/net/ethernet/mellanox/mlx4/en_rx.c
@@ -72,7 +72,7 @@ static int mlx4_alloc_pages(struct mlx4_en_priv *priv,
 	}
 	dma = dma_map_page(priv->ddev, page, 0, PAGE_SIZE << order,
 			   frag_info->dma_dir);
-	if (dma_mapping_error(priv->ddev, dma)) {
+	if (unlikely(dma_mapping_error(priv->ddev, dma))) {
 		put_page(page);
 		return -ENOMEM;
 	}
@@ -108,7 +108,8 @@ static int mlx4_en_alloc_frags(struct mlx4_en_priv *priv,
 		    ring_alloc[i].page_size)
 			continue;
 
-		if (mlx4_alloc_pages(priv, &page_alloc[i], frag_info, gfp))
+		if (unlikely(mlx4_alloc_pages(priv, &page_alloc[i],
+					      frag_info, gfp)))
 			goto out;
 	}
 
@@ -585,7 +586,7 @@ static int mlx4_en_complete_rx_desc(struct mlx4_en_priv *priv,
 		frag_info = &priv->frag_info[nr];
 		if (length <= frag_info->frag_prefix_size)
 			break;
-		if (!frags[nr].page)
+		if (unlikely(!frags[nr].page))
 			goto fail;
 
 		dma = be64_to_cpu(rx_desc->data[nr].addr);
@@ -625,7 +626,7 @@ static struct sk_buff *mlx4_en_rx_skb(struct mlx4_en_priv *priv,
 	dma_addr_t dma;
 
 	skb = netdev_alloc_skb(priv->dev, SMALL_PACKET_SIZE + NET_IP_ALIGN);
-	if (!skb) {
+	if (unlikely(!skb)) {
 		en_dbg(RX_ERR, priv, "Failed allocating skb\n");
 		return NULL;
 	}
@@ -736,7 +737,8 @@ static int get_fixed_ipv6_csum(__wsum hw_checksum, struct sk_buff *skb,
 {
 	__wsum csum_pseudo_hdr = 0;
 
-	if (ipv6h->nexthdr == IPPROTO_FRAGMENT || ipv6h->nexthdr == IPPROTO_HOPOPTS)
+	if (unlikely(ipv6h->nexthdr == IPPROTO_FRAGMENT ||
+		     ipv6h->nexthdr == IPPROTO_HOPOPTS))
 		return -1;
 	hw_checksum = csum_add(hw_checksum, (__force __wsum)htons(ipv6h->nexthdr));
 
@@ -769,7 +771,7 @@ static int check_csum(struct mlx4_cqe *cqe, struct sk_buff *skb, void *va,
 		get_fixed_ipv4_csum(hw_checksum, skb, hdr);
 #if IS_ENABLED(CONFIG_IPV6)
 	else if (cqe->status & cpu_to_be16(MLX4_CQE_STATUS_IPV6))
-		if (get_fixed_ipv6_csum(hw_checksum, skb, hdr))
+		if (unlikely(get_fixed_ipv6_csum(hw_checksum, skb, hdr)))
 			return -1;
 #endif
 	return 0;
@@ -796,10 +798,10 @@ int mlx4_en_process_rx_cq(struct net_device *dev, struct mlx4_en_cq *cq, int bud
 	u64 timestamp;
 	bool l2_tunnel;
 
-	if (!priv->port_up)
+	if (unlikely(!priv->port_up))
 		return 0;
 
-	if (budget <= 0)
+	if (unlikely(budget <= 0))
 		return polled;
 
 	/* Protect accesses to: ring->xdp_prog, priv->mac_hash list */
@@ -902,16 +904,16 @@ int mlx4_en_process_rx_cq(struct net_device *dev, struct mlx4_en_cq *cq, int bud
 			case XDP_PASS:
 				break;
 			case XDP_TX:
-				if (!mlx4_en_xmit_frame(frags, dev,
+				if (likely(!mlx4_en_xmit_frame(frags, dev,
 							length, tx_index,
-							&doorbell_pending))
+							&doorbell_pending)))
 					goto consumed;
 				goto next; /* Drop on xmit failure */
 			default:
 				bpf_warn_invalid_xdp_action(act);
 			case XDP_ABORTED:
 			case XDP_DROP:
-				if (mlx4_en_rx_recycle(ring, frags))
+				if (likely(mlx4_en_rx_recycle(ring, frags)))
 					goto consumed;
 				goto next;
 			}
@@ -1015,7 +1017,7 @@ int mlx4_en_process_rx_cq(struct net_device *dev, struct mlx4_en_cq *cq, int bud
 
 		/* GRO not possible, complete processing here */
 		skb = mlx4_en_rx_skb(priv, rx_desc, frags, length);
-		if (!skb) {
+		if (unlikely(!skb)) {
 			ring->dropped++;
 			goto next;
 		}
-- 
1.8.3.1

[PATCH net-next V2 2/4] net/mlx4_en: Fix wrong indentation

[Tariq Toukan]

From: Kamal Heib <kamalh@mellanox.com>

Use tabs instead of spaces before if statement, no functional change.

Fixes: e7c1c2c46201 ("mlx4_en: Added self diagnostics test implementation")
Signed-off-by: Kamal Heib <kamalh@mellanox.com>
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
---
 drivers/net/ethernet/mellanox/mlx4/en_rx.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/net/ethernet/mellanox/mlx4/en_rx.c b/drivers/net/ethernet/mellanox/mlx4/en_rx.c
index e60a86a9266d..f0501ea867c7 100644
--- a/drivers/net/ethernet/mellanox/mlx4/en_rx.c
+++ b/drivers/net/ethernet/mellanox/mlx4/en_rx.c
@@ -1022,7 +1022,7 @@ int mlx4_en_process_rx_cq(struct net_device *dev, struct mlx4_en_cq *cq, int bud
 			goto next;
 		}
 
-                if (unlikely(priv->validate_loopback)) {
+		if (unlikely(priv->validate_loopback)) {
 			validate_loopback(priv, skb);
 			goto next;
 		}
-- 
1.8.3.1

[PATCH net-next V2 3/4] net/mlx4_core: Use RCU to perform radix tree lookup for SRQ

[Tariq Toukan]

From: Leon Romanovsky <leonro@mellanox.com>

Radix tree lookup can be performed without locking.

Fixes: 225c7b1feef1 ("IB/mlx4: Add a driver Mellanox ConnectX InfiniBand adapters")
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Suggested-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
---
 drivers/net/ethernet/mellanox/mlx4/srq.c | 14 +++++---------
 1 file changed, 5 insertions(+), 9 deletions(-)

diff --git a/drivers/net/ethernet/mellanox/mlx4/srq.c b/drivers/net/ethernet/mellanox/mlx4/srq.c
index 67146624eb58..f44d089e2ca6 100644
--- a/drivers/net/ethernet/mellanox/mlx4/srq.c
+++ b/drivers/net/ethernet/mellanox/mlx4/srq.c
@@ -45,15 +45,12 @@ void mlx4_srq_event(struct mlx4_dev *dev, u32 srqn, int event_type)
 	struct mlx4_srq_table *srq_table = &mlx4_priv(dev)->srq_table;
 	struct mlx4_srq *srq;
 
-	spin_lock(&srq_table->lock);
-
+	rcu_read_lock();
 	srq = radix_tree_lookup(&srq_table->tree, srqn & (dev->caps.num_srqs - 1));
+	rcu_read_unlock();
 	if (srq)
 		atomic_inc(&srq->refcount);
-
-	spin_unlock(&srq_table->lock);
-
-	if (!srq) {
+	else {
 		mlx4_warn(dev, "Async event for bogus SRQ %08x\n", srqn);
 		return;
 	}
@@ -301,12 +298,11 @@ struct mlx4_srq *mlx4_srq_lookup(struct mlx4_dev *dev, u32 srqn)
 {
 	struct mlx4_srq_table *srq_table = &mlx4_priv(dev)->srq_table;
 	struct mlx4_srq *srq;
-	unsigned long flags;
 
-	spin_lock_irqsave(&srq_table->lock, flags);
+	rcu_read_lock();
 	srq = radix_tree_lookup(&srq_table->tree,
 				srqn & (dev->caps.num_srqs - 1));
-	spin_unlock_irqrestore(&srq_table->lock, flags);
+	rcu_read_unlock();
 
 	return srq;
 }
-- 
1.8.3.1

Re: [PATCH net-next V2 3/4] net/mlx4_core: Use RCU to perform radix tree lookup for SRQ

[Sergei Shtylyov]

Hello.

On 9/20/2016 2:39 PM, Tariq Toukan wrote:

> From: Leon Romanovsky <leonro@mellanox.com>
>
> Radix tree lookup can be performed without locking.
>
> Fixes: 225c7b1feef1 ("IB/mlx4: Add a driver Mellanox ConnectX InfiniBand adapters")
> Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
> Suggested-by: Sagi Grimberg <sagi@grimberg.me>
> Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
> ---
>  drivers/net/ethernet/mellanox/mlx4/srq.c | 14 +++++---------
>  1 file changed, 5 insertions(+), 9 deletions(-)
>
> diff --git a/drivers/net/ethernet/mellanox/mlx4/srq.c b/drivers/net/ethernet/mellanox/mlx4/srq.c
> index 67146624eb58..f44d089e2ca6 100644
> --- a/drivers/net/ethernet/mellanox/mlx4/srq.c
> +++ b/drivers/net/ethernet/mellanox/mlx4/srq.c
> @@ -45,15 +45,12 @@ void mlx4_srq_event(struct mlx4_dev *dev, u32 srqn, int event_type)
>  	struct mlx4_srq_table *srq_table = &mlx4_priv(dev)->srq_table;
>  	struct mlx4_srq *srq;
>
> -	spin_lock(&srq_table->lock);
> -
> +	rcu_read_lock();
>  	srq = radix_tree_lookup(&srq_table->tree, srqn & (dev->caps.num_srqs - 1));
> +	rcu_read_unlock();
>  	if (srq)
>  		atomic_inc(&srq->refcount);
> -
> -	spin_unlock(&srq_table->lock);
> -
> -	if (!srq) {
> +	else {
>  		mlx4_warn(dev, "Async event for bogus SRQ %08x\n", srqn);
>  		return;
>  	}

    CodingStyle: all branches of the *if* statement should have {} if at least 
one has.

[...]

MBR, Sergei

Re: [PATCH net-next V2 3/4] net/mlx4_core: Use RCU to perform radix tree lookup for SRQ

[Leon Romanovsky]

[-- Attachment #1: Type: text/plain, Size: 2406 bytes --]

On Tue, Sep 20, 2016 at 03:47:49PM +0300, Sergei Shtylyov wrote:
> Hello.
>
> On 9/20/2016 2:39 PM, Tariq Toukan wrote:
>
> >From: Leon Romanovsky <leonro@mellanox.com>
> >
> >Radix tree lookup can be performed without locking.
> >
> >Fixes: 225c7b1feef1 ("IB/mlx4: Add a driver Mellanox ConnectX InfiniBand adapters")
> >Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
> >Suggested-by: Sagi Grimberg <sagi@grimberg.me>
> >Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
> >---
> > drivers/net/ethernet/mellanox/mlx4/srq.c | 14 +++++---------
> > 1 file changed, 5 insertions(+), 9 deletions(-)
> >
> >diff --git a/drivers/net/ethernet/mellanox/mlx4/srq.c b/drivers/net/ethernet/mellanox/mlx4/srq.c
> >index 67146624eb58..f44d089e2ca6 100644
> >--- a/drivers/net/ethernet/mellanox/mlx4/srq.c
> >+++ b/drivers/net/ethernet/mellanox/mlx4/srq.c
> >@@ -45,15 +45,12 @@ void mlx4_srq_event(struct mlx4_dev *dev, u32 srqn, int event_type)
> > 	struct mlx4_srq_table *srq_table = &mlx4_priv(dev)->srq_table;
> > 	struct mlx4_srq *srq;
> >
> >-	spin_lock(&srq_table->lock);
> >-
> >+	rcu_read_lock();
> > 	srq = radix_tree_lookup(&srq_table->tree, srqn & (dev->caps.num_srqs - 1));
> >+	rcu_read_unlock();
> > 	if (srq)
> > 		atomic_inc(&srq->refcount);
> >-
> >-	spin_unlock(&srq_table->lock);
> >-
> >-	if (!srq) {
> >+	else {
> > 		mlx4_warn(dev, "Async event for bogus SRQ %08x\n", srqn);
> > 		return;
> > 	}
>
>    CodingStyle: all branches of the *if* statement should have {} if at
> least one has.

I'm not arguing, but wanted to raise attention that checkpatch.pl
doesn't warn about it.

➜  /tmp wget https://patchwork.ozlabs.org/patch/672216/mbox/
➜  linux-rdma git:(master) ./scripts/checkpatch.pl /tmp/index.html
total: 0 errors, 0 warnings, 0 checks, 32 lines checked

/tmp/index.html has no obvious style problems and is ready for
submission.

And CodingStyle presents a little bit different case
----
This does not apply if only one branch of a conditional statement is a
single statement; in the latter case use braces in both branches:

        if (condition) {
                do_this();
                do_that();
        } else {
                otherwise();
        }

and not

        if (condition)
                do_that();
        else {
                otherwise();
		otherwise();
        }

Thanks

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 819 bytes --]

[PATCH net-next V2 4/4] net/mlx4_core: Fix deadlock when switching between polling and event fw commands

[Tariq Toukan]

From: Jack Morgenstein <jackm@dev.mellanox.co.il>

When switching from polling-based fw commands to event-based fw
commands, there is a race condition which could cause a fw command
in another task to hang: that task will keep waiting for the polling
sempahore, but may never be able to acquire it. This is due to
mlx4_cmd_use_events, which "down"s the sempahore back to 0.

During driver initialization, this is not a problem, since no other
tasks which invoke FW commands are active.

However, there is a problem if the driver switches to polling mode
and then back to event mode during normal operation.

The "test_interrupts" feature does exactly that.
Running "ethtool -t <eth device> offline" causes the PF driver to
temporarily switch to polling mode, and then back to event mode.
(Note that for VF drivers, such switching is not performed).

Fix this by adding a read-write semaphore for protection when
switching between modes.

Fixes: 225c7b1feef1 ("IB/mlx4: Add a driver Mellanox ConnectX InfiniBand adapters")
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
---
 drivers/net/ethernet/mellanox/mlx4/cmd.c  | 23 +++++++++++++++++------
 drivers/net/ethernet/mellanox/mlx4/mlx4.h |  2 ++
 2 files changed, 19 insertions(+), 6 deletions(-)

diff --git a/drivers/net/ethernet/mellanox/mlx4/cmd.c b/drivers/net/ethernet/mellanox/mlx4/cmd.c
index f04a423ff79d..a58d96cf1ed1 100644
--- a/drivers/net/ethernet/mellanox/mlx4/cmd.c
+++ b/drivers/net/ethernet/mellanox/mlx4/cmd.c
@@ -785,17 +785,23 @@ int __mlx4_cmd(struct mlx4_dev *dev, u64 in_param, u64 *out_param,
 		return mlx4_cmd_reset_flow(dev, op, op_modifier, -EIO);
 
 	if (!mlx4_is_mfunc(dev) || (native && mlx4_is_master(dev))) {
+		int ret;
+
 		if (dev->persist->state & MLX4_DEVICE_STATE_INTERNAL_ERROR)
 			return mlx4_internal_err_ret_value(dev, op,
 							  op_modifier);
+		down_read(&mlx4_priv(dev)->cmd.switch_sem);
 		if (mlx4_priv(dev)->cmd.use_events)
-			return mlx4_cmd_wait(dev, in_param, out_param,
-					     out_is_imm, in_modifier,
-					     op_modifier, op, timeout);
+			ret = mlx4_cmd_wait(dev, in_param, out_param,
+					    out_is_imm, in_modifier,
+					    op_modifier, op, timeout);
 		else
-			return mlx4_cmd_poll(dev, in_param, out_param,
-					     out_is_imm, in_modifier,
-					     op_modifier, op, timeout);
+			ret = mlx4_cmd_poll(dev, in_param, out_param,
+					    out_is_imm, in_modifier,
+					    op_modifier, op, timeout);
+
+		up_read(&mlx4_priv(dev)->cmd.switch_sem);
+		return ret;
 	}
 	return mlx4_slave_cmd(dev, in_param, out_param, out_is_imm,
 			      in_modifier, op_modifier, op, timeout);
@@ -2454,6 +2460,7 @@ int mlx4_cmd_init(struct mlx4_dev *dev)
 	int flags = 0;
 
 	if (!priv->cmd.initialized) {
+		init_rwsem(&priv->cmd.switch_sem);
 		mutex_init(&priv->cmd.slave_cmd_mutex);
 		sema_init(&priv->cmd.poll_sem, 1);
 		priv->cmd.use_events = 0;
@@ -2583,6 +2590,7 @@ int mlx4_cmd_use_events(struct mlx4_dev *dev)
 	if (!priv->cmd.context)
 		return -ENOMEM;
 
+	down_write(&priv->cmd.switch_sem);
 	for (i = 0; i < priv->cmd.max_cmds; ++i) {
 		priv->cmd.context[i].token = i;
 		priv->cmd.context[i].next  = i + 1;
@@ -2606,6 +2614,7 @@ int mlx4_cmd_use_events(struct mlx4_dev *dev)
 
 	down(&priv->cmd.poll_sem);
 	priv->cmd.use_events = 1;
+	up_write(&priv->cmd.switch_sem);
 
 	return err;
 }
@@ -2618,6 +2627,7 @@ void mlx4_cmd_use_polling(struct mlx4_dev *dev)
 	struct mlx4_priv *priv = mlx4_priv(dev);
 	int i;
 
+	down_write(&priv->cmd.switch_sem);
 	priv->cmd.use_events = 0;
 
 	for (i = 0; i < priv->cmd.max_cmds; ++i)
@@ -2626,6 +2636,7 @@ void mlx4_cmd_use_polling(struct mlx4_dev *dev)
 	kfree(priv->cmd.context);
 
 	up(&priv->cmd.poll_sem);
+	up_write(&priv->cmd.switch_sem);
 }
 
 struct mlx4_cmd_mailbox *mlx4_alloc_cmd_mailbox(struct mlx4_dev *dev)
diff --git a/drivers/net/ethernet/mellanox/mlx4/mlx4.h b/drivers/net/ethernet/mellanox/mlx4/mlx4.h
index c9d7fc5159f2..c128ba3ef014 100644
--- a/drivers/net/ethernet/mellanox/mlx4/mlx4.h
+++ b/drivers/net/ethernet/mellanox/mlx4/mlx4.h
@@ -46,6 +46,7 @@
 #include <linux/interrupt.h>
 #include <linux/spinlock.h>
 #include <net/devlink.h>
+#include <linux/rwsem.h>
 
 #include <linux/mlx4/device.h>
 #include <linux/mlx4/driver.h>
@@ -627,6 +628,7 @@ struct mlx4_cmd {
 	struct mutex		slave_cmd_mutex;
 	struct semaphore	poll_sem;
 	struct semaphore	event_sem;
+	struct rw_semaphore	switch_sem;
 	int			max_cmds;
 	spinlock_t		context_lock;
 	int			free_head;
-- 
1.8.3.1

Re: [PATCH net-next V2 0/4] mlx4 misc cleanups and improvements

[David Miller]

From: Tariq Toukan <tariqt@mellanox.com>
Date: Tue, 20 Sep 2016 14:39:38 +0300

> This patchset contains some cleanups and improvements from the team
> to the mlx4 Eth and core drivers.
> 
> Series generated against net-next commit:
> 5a7a5555a362 'net sched: stylistic cleanups'

Series applied, thanks.