Re: [PATCH v2 net-next 2/2] net: introduce budget_squeeze to help us tune rx behavior

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On Tue, Mar 14, 2023 at 11:05:32AM +0800, Jason Xing wrote:
> From: Jason Xing <kernelxing@xxxxxxxxxxx>
> 
> When we encounter some performance issue and then get lost on how
> to tune the budget limit and time limit in net_rx_action() function,
> we can separately counting both of them to avoid the confusion.
> 
> Signed-off-by: Jason Xing <kernelxing@xxxxxxxxxxx>

As per my comment on patch 1/2, I'd drop the "/* keep it untouched */"
comment.

That notwithstanding:

Reviewed-by: Simon Horman <simon.horman@xxxxxxxxxxxx>

> ---
> v2:
> 1) change the coding style suggested by Stephen and Simon
> 2) Keep the display of the old data (time_squeeze) untouched suggested
> by Kui-Feng
> Link: https://lore.kernel.org/lkml/20230311163614.92296-1-kerneljasonxing@xxxxxxxxx/
> ---
>  include/linux/netdevice.h |  1 +
>  net/core/dev.c            | 12 ++++++++----
>  net/core/net-procfs.c     |  9 ++++++---
>  3 files changed, 15 insertions(+), 7 deletions(-)
> 
> diff --git a/include/linux/netdevice.h b/include/linux/netdevice.h
> index 6a14b7b11766..5736311a2133 100644
> --- a/include/linux/netdevice.h
> +++ b/include/linux/netdevice.h
> @@ -3157,6 +3157,7 @@ struct softnet_data {
>  	/* stats */
>  	unsigned int		processed;
>  	unsigned int		time_squeeze;
> +	unsigned int		budget_squeeze;
>  #ifdef CONFIG_RPS
>  	struct softnet_data	*rps_ipi_list;
>  #endif
> diff --git a/net/core/dev.c b/net/core/dev.c
> index 253584777101..1518a366783b 100644
> --- a/net/core/dev.c
> +++ b/net/core/dev.c
> @@ -6637,6 +6637,7 @@ static __latent_entropy void net_rx_action(struct softirq_action *h)
>  	unsigned long time_limit = jiffies +
>  		usecs_to_jiffies(READ_ONCE(netdev_budget_usecs));
>  	int budget = READ_ONCE(netdev_budget);
> +	bool done = false;
>  	LIST_HEAD(list);
>  	LIST_HEAD(repoll);
>  
> @@ -6644,7 +6645,7 @@ static __latent_entropy void net_rx_action(struct softirq_action *h)
>  	list_splice_init(&sd->poll_list, &list);
>  	local_irq_enable();
>  
> -	for (;;) {
> +	while (!done) {
>  		struct napi_struct *n;
>  
>  		skb_defer_free_flush(sd);
> @@ -6662,10 +6663,13 @@ static __latent_entropy void net_rx_action(struct softirq_action *h)
>  		 * Allow this to run for 2 jiffies since which will allow
>  		 * an average latency of 1.5/HZ.
>  		 */
> -		if (unlikely(budget <= 0 ||
> -			     time_after_eq(jiffies, time_limit))) {
> +		if (unlikely(budget <= 0)) {
> +			sd->budget_squeeze++;
> +			done = true;
> +		}
> +		if (unlikely(time_after_eq(jiffies, time_limit))) {
>  			sd->time_squeeze++;
> -			break;
> +			done = true;
>  		}
>  	}
>  
> diff --git a/net/core/net-procfs.c b/net/core/net-procfs.c
> index 2809b663e78d..25810ee46a04 100644
> --- a/net/core/net-procfs.c
> +++ b/net/core/net-procfs.c
> @@ -179,14 +179,17 @@ static int softnet_seq_show(struct seq_file *seq, void *v)
>  	 */
>  	seq_printf(seq,
>  		   "%08x %08x %08x %08x %08x %08x %08x %08x %08x %08x %08x %08x %08x "
> -		   "%08x %08x\n",
> -		   sd->processed, sd->dropped, sd->time_squeeze, 0,
> +		   "%08x %08x %08x %08x\n",
> +		   sd->processed, sd->dropped,
> +		   sd->time_squeeze + sd->budget_squeeze, /* keep it untouched */
> +		   0,
>  		   0, 0, 0, 0, /* was fastroute */
>  		   0,	/* was cpu_collision */
>  		   sd->received_rps, flow_limit_count,
>  		   softnet_backlog_len(sd),	/* keep it untouched */
>  		   (int)seq->index,
> -		   softnet_input_pkt_queue_len(sd), softnet_process_queue_len(sd));
> +		   softnet_input_pkt_queue_len(sd), softnet_process_queue_len(sd),
> +		   sd->time_squeeze, sd->budget_squeeze);
>  	return 0;
>  }
>  
> -- 
> 2.37.3
> 



[Index of Archives]     [Linux Samsung SoC]     [Linux Rockchip SoC]     [Linux Actions SoC]     [Linux for Synopsys ARC Processors]     [Linux NFS]     [Linux NILFS]     [Linux USB Devel]     [Video for Linux]     [Linux Audio Users]     [Yosemite News]     [Linux Kernel]     [Linux SCSI]


  Powered by Linux