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
 
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date:	Wed, 13 Jun 2012 11:28:07 +0200
From:	Roland Stigge <stigge@...com.de>
To:	Eric Dumazet <eric.dumazet@...il.com>
CC:	David Miller <davem@...emloft.net>, netdev@...r.kernel.org,
	linux-kernel@...r.kernel.org, kevin.wells@....com,
	srinivas.bakki@....com, aletes.xgr@...il.com,
	linux-arm-kernel@...ts.infradead.org
Subject: Re: [PATCH 1/3] net: lpc_eth: Replace WARN() trace with simple pr_warn()

On 06/13/2012 08:16 AM, Eric Dumazet wrote:
> On Mon, 2012-06-11 at 21:18 +0200, Eric Dumazet wrote:
>> On Mon, 2012-06-11 at 11:26 +0200, Roland Stigge wrote:
>>
>>> Is it sensible at this point to increase the TX buffers anyway? For
>>> different reasons of course: We have enough SRAM available and TX
>>> buffers (16->32) are still more than RX buffers (48).
>>
>> I doubt it has any impact on performance for a 100Mbit link ?
>>
>> One thing that could be done would be to free skbs in
>> lpc_eth_hard_start_xmit() instead of __lpc_handle_xmit()
>>
> 
> Here is the patch I was thinking about
> 
> (on top of latest net-next)
> 
> Could you please test it ?
>
>  drivers/net/ethernet/nxp/lpc_eth.c |   10 +++++-----
>  1 file changed, 5 insertions(+), 5 deletions(-)
> 
> diff --git a/drivers/net/ethernet/nxp/lpc_eth.c b/drivers/net/ethernet/nxp/lpc_eth.c
> index 083d671..426f14c 100644
> --- a/drivers/net/ethernet/nxp/lpc_eth.c
> +++ b/drivers/net/ethernet/nxp/lpc_eth.c
> @@ -440,7 +440,7 @@ struct netdata_local {
>  	spinlock_t		lock;
>  	void __iomem		*net_base;
>  	u32			msg_enable;
> -	struct sk_buff		*skb[ENET_TX_DESC];
> +	unsigned int		skblen[ENET_TX_DESC];
>  	unsigned int		last_tx_idx;
>  	unsigned int		num_used_tx_buffs;
>  	struct mii_bus		*mii_bus;
> @@ -908,7 +908,7 @@ static void __lpc_handle_xmit(struct net_device *ndev)
>  
>  	txcidx = readl(LPC_ENET_TXCONSUMEINDEX(pldat->net_base));
>  	while (pldat->last_tx_idx != txcidx) {
> -		skb = pldat->skb[pldat->last_tx_idx];
> +		unsigned int skblen = pldat->skblen[pldat->last_tx_idx];
>  
>  		/* A buffer is available, get buffer status */
>  		ptxstat = &pldat->tx_stat_v[pldat->last_tx_idx];
> @@ -945,9 +945,8 @@ static void __lpc_handle_xmit(struct net_device *ndev)
>  		} else {
>  			/* Update stats */
>  			ndev->stats.tx_packets++;
> -			ndev->stats.tx_bytes += skb->len;
> +			ndev->stats.tx_bytes += skblen;
>  		}
> -		dev_kfree_skb_irq(skb);
>  
>  		txcidx = readl(LPC_ENET_TXCONSUMEINDEX(pldat->net_base));
>  	}
> @@ -1132,7 +1131,7 @@ static int lpc_eth_hard_start_xmit(struct sk_buff *skb, struct net_device *ndev)
>  	memcpy(pldat->tx_buff_v + txidx * ENET_MAXF_SIZE, skb->data, len);
>  
>  	/* Save the buffer and increment the buffer counter */
> -	pldat->skb[txidx] = skb;
> +	pldat->skblen[txidx] = len;
>  	pldat->num_used_tx_buffs++;
>  
>  	/* Start transmit */
> @@ -1147,6 +1146,7 @@ static int lpc_eth_hard_start_xmit(struct sk_buff *skb, struct net_device *ndev)
>  
>  	spin_unlock_irq(&pldat->lock);
>  
> +	dev_kfree_skb(skb);
>  	return NETDEV_TX_OK;
>  }

Works fine for a while now.

We can remove the unused variable skb from __lpc_handle_xmit() now,
maybe just do in your patch?

Thanks!

Tested-by: Roland Stigge <stigge@...com.de>
--
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