public inbox for netdev@vger.kernel.org 
 help / color / mirror / Atom feed
From: Eric Dumazet <dada1@cosmosbay•com>
To: Yevgeny Petrilin <yevgenyp@mellanox•co.il>
Cc: David Miller <davem@davemloft•net>,
	netdev@vger•kernel.org, tziporet@mellanox•co.il
Subject: Re: [net-2.6 PATCH V2] mlx4_en: Fix a kernel panic when waking tx queue
Date: Tue, 26 May 2009 12:48:04 +0200	[thread overview]
Message-ID: <4A1BC8E4.1050009@cosmosbay.com> (raw)
In-Reply-To: <4A1B92D1.8070907@mellanox.co.il>

Yevgeny Petrilin a écrit :
> When the transmit queue gets full we enable interrupts for TX completions
> There was a race that we handled the TX queue both from the interrupt context
> and from the transmit function. Using "spin_trylock_irq()" ensures this
> doesn't happen.
> 
> Signed-off-by: Yevgeny Petrilin <yevgenyp@mellanox•co.il>
> ---
>  drivers/net/mlx4/en_tx.c |    8 ++++----
>  1 files changed, 4 insertions(+), 4 deletions(-)
> 
> diff --git a/drivers/net/mlx4/en_tx.c b/drivers/net/mlx4/en_tx.c
> index ac6fc49..e5c98a9 100644
> --- a/drivers/net/mlx4/en_tx.c
> +++ b/drivers/net/mlx4/en_tx.c
> @@ -426,7 +426,7 @@ void mlx4_en_poll_tx_cq(unsigned long data)
> 
>  	INC_PERF_COUNTER(priv->pstats.tx_poll);
> 
> -	if (!spin_trylock(&ring->comp_lock)) {
> +	if (!spin_trylock_irq(&ring->comp_lock)) {
>  		mod_timer(&cq->timer, jiffies + MLX4_EN_TX_POLL_TIMEOUT);
>  		return;
>  	}
> @@ -439,7 +439,7 @@ void mlx4_en_poll_tx_cq(unsigned long data)
>  	if (inflight && priv->port_up)
>  		mod_timer(&cq->timer, jiffies + MLX4_EN_TX_POLL_TIMEOUT);
> 
> -	spin_unlock(&ring->comp_lock);
> +	spin_unlock_irq(&ring->comp_lock);
>  }
> 
>  static struct mlx4_en_tx_desc *mlx4_en_bounce_to_desc(struct mlx4_en_priv *priv,
> @@ -482,9 +482,9 @@ static inline void mlx4_en_xmit_poll(struct mlx4_en_priv *priv, int tx_ind)
> 
>  	/* Poll the CQ every mlx4_en_TX_MODER_POLL packets */
>  	if ((++ring->poll_cnt & (MLX4_EN_TX_POLL_MODER - 1)) == 0)
> -		if (spin_trylock(&ring->comp_lock)) {
> +		if (spin_trylock_irq(&ring->comp_lock)) {
>  			mlx4_en_process_tx_cq(priv->dev, cq);
> -			spin_unlock(&ring->comp_lock);
> +			spin_unlock_irq(&ring->comp_lock);
>  		}
>  }
> 

Just curious.

Blocking hard IRQ while doing TX completion is quite nasty,
as freeing one hundred of skb take long.

Could you try something in process context instead ?

Is this timer driven tx completion thing really good for performance ?


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];

        /* 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(&ring->comp_lock)) {
                        mlx4_en_process_tx_cq(priv->dev, cq);
                        spin_unlock(&ring->comp_lock);
                }
}


One interesting thing is calling mlx4_en_process_tx_cq() once every 16 packets,
this means the producer cpu is also the cpu doing the TX completion, I like this.


  reply	other threads:[~2009-05-26 10:48 UTC|newest]

Thread overview: 39+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2009-04-20 14:24 [PATCH 1/5] mlx4_en: Fix error handling while activating RX rings Yevgeny Petrilin
2009-04-20 14:26 ` [PATCH 2/5]mlx4_en: Fix a race at restart task Yevgeny Petrilin
2009-04-21  4:32   ` [PATCH] mlx4_en: Fix cleanup if workqueue create in mlx4_en_add() fails Roland Dreier
2009-04-21  8:50     ` David Miller
2009-04-21  8:49   ` [PATCH 2/5]mlx4_en: Fix a race at restart task David Miller
2009-04-20 14:30 ` [PATCH 3/5] mlx4_en: Assign dummy event handler for TX queue Yevgeny Petrilin
2009-04-21  8:50   ` David Miller
2009-04-20 14:33 ` [PATCH 4/5] mlx4_en: use NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM for tx csum at initialization Yevgeny Petrilin
2009-04-20 14:34   ` [PATCH 5/5] mlx4_en: Move to SW counters for total bytes and packets Yevgeny Petrilin
2009-04-21  8:50     ` David Miller
2009-04-21  8:50   ` [PATCH 4/5] mlx4_en: use NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM for tx csum at initialization David Miller
2009-04-27  6:42   ` [PATCH 2/2] mlx4_en: Handle page allocation failure during receive Yevgeny Petrilin
2009-04-27  9:31     ` David Miller
2009-04-21  8:49 ` [PATCH 1/5] mlx4_en: Fix error handling while activating RX rings David Miller
2009-04-27  6:41 ` [PATCH 1/2] mlx4_en: Fix cleanup flow on cq activation Yevgeny Petrilin
2009-04-27  9:31   ` David Miller
2009-05-13 11:47 ` [PATCH] mlx4_en: Fix not deleted napi structures Yevgeny Petrilin
2009-05-18  3:49   ` David Miller
2009-05-24 13:16   ` [PATCH 1/2] mlx4_en: Removed redundant stride variable Yevgeny Petrilin
2009-05-25  7:36     ` David Miller
2009-05-24 13:17   ` [PATCH 2/2] mlx4_en: Fix partial rings feature Yevgeny Petrilin
2009-05-25  7:36     ` David Miller
2009-05-25  8:32     ` [net-2.6 PATCH] mlx4_en: Fix a kernel panic when waking tx queue Yevgeny Petrilin
2009-05-25  8:44       ` David Miller
2009-05-26  6:49         ` Yevgeny Petrilin
2009-06-02  9:20         ` [PATCH 2/8] mlx4_en: Moved all module parameters handling to en_main.c Yevgeny Petrilin
2009-06-02  9:21         ` [PATCH 3/8] mlx4_en renamed en_params.c to en_ethtool.c Yevgeny Petrilin
2009-06-02  9:22         ` [PATCH 4/8] mlx4_en: Work with part of the ports Yevgeny Petrilin
2009-06-02  9:23         ` [PATCH 5/8] mlx4_en: Coalescing target is equal for all mtu's Yevgeny Petrilin
2009-06-02  9:24         ` [PATCH 6/8] mlx4_en: multiqueue support Yevgeny Petrilin
2009-06-02  9:28         ` [PATCH 7/8] mlx4_en: Added vlan_features support Yevgeny Petrilin
2009-06-02  9:29         ` [PATCH 8/8] mlx4_en: Updated driver version Yevgeny Petrilin
2009-06-02  9:36           ` David Miller
2009-05-26  6:57       ` [net-2.6 PATCH V2] mlx4_en: Fix a kernel panic when waking tx queue Yevgeny Petrilin
2009-05-26 10:48         ` Eric Dumazet [this message]
2009-05-27  6:08           ` Yevgeny Petrilin
2009-05-30  5:00         ` David Miller
2009-06-02  6:27         ` [PATCH 1/8] mlx4_en: Giving interface name in debug messages Yevgeny Petrilin
2009-06-02  7:36           ` David Miller

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=4A1BC8E4.1050009@cosmosbay.com \
    --to=dada1@cosmosbay$(echo .)com \
    --cc=davem@davemloft$(echo .)net \
    --cc=netdev@vger$(echo .)kernel.org \
    --cc=tziporet@mellanox$(echo .)co.il \
    --cc=yevgenyp@mellanox$(echo .)co.il \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox