Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > linux.kernel > #1478377 > unrolled thread

Re: [Intel-wired-lan] [PATCH 249/249] net:ethernet:intel:igb_main.c: Add Throttling disable option in order to decrease latency usually required by RT applications.

Back to article view | Back to linux.kernel

This discussion starts older than the indexed window; earlier articles aren't shown. The article labeled Started by below is the oldest one visible, not the original post.

Contents

  Re: [Intel-wired-lan] [PATCH 249/249]  net:ethernet:intel:igb_main.c: Add Throttling disable option in order to  decrease latency usually required by RT applications. "Duyck, Alexander H" <alexander.h.duyck@intel.com> - 2016-09-07 17:00 +0200

#1478377 — Re: [Intel-wired-lan] [PATCH 249/249] net:ethernet:intel:igb_main.c: Add Throttling disable option in order to decrease latency usually required by RT applications.

On Tue, 2016-09-06 at 14:46 +0000, Amir Yihie wrote:
> From e433f5bedd49a65b7337d025f3f204f0f21ad5e5 Mon Sep 17 00:00:00
> 2001
> From: amiry <amiry@waves.com>
> Date: Tue, 6 Sep 2016 14:58:38 +0300
> Subject: [PATCH 249/249] net:ethernet:intel:igb_main.c: Add
> Throttling disable
>  option in order to decrease latency usually required by RT
> applications.
> 
> Add an option to Disable Throttling algorithm , which will result
> with minimum latency on received frames .
> mainly targets for RT application that look for low larceny and less
> cares about throughput .
> How to use :
> On insmod , add parameter : InterruptThrottleRate=0
> 
> This patch is currently against a linux 4.1.27-rt31 kernel.
> 
> Signed-off-by: Amir Yihie <amiry@waves.com>
> ---
>  drivers/net/ethernet/intel/igb/igb_main.c | 11 +++++++++++
>  1 file changed, 11 insertions(+)
> 
> diff --git a/drivers/net/ethernet/intel/igb/igb_main.c
> b/drivers/net/ethernet/intel/igb/igb_main.c
> index 7dfbcde..9480850 100644
> --- a/drivers/net/ethernet/intel/igb/igb_main.c
> +++ b/drivers/net/ethernet/intel/igb/igb_main.c
> @@ -251,6 +251,14 @@ static int debug = -1;
>  module_param(debug, int, 0);
>  MODULE_PARM_DESC(debug, "Debug level (0=none,...,16=all)");
>  
> +/* Interrupt Throttle Rate (interrupts/sec)
> + *
> + * Valid Range: 0=off, 3=dynamic conservative
> + */
> +static int InterruptThrottleRate = 3;
> +module_param(InterruptThrottleRate, int, 0);
> +MODULE_PARM_DESC(InterruptThrottleRate, "InterruptThrottleRate
> (0=off, 3=dynamic conservative)");
> +

This is creating redundant functionality.  The same can be done via the
ethtool command "ethtool -C <iface> rx-usecs 0".  Is there some reason
why you couldn't use ethtool?

>  struct igb_reg_info {
>         u32 ofs;
>         char *name;
> @@ -2966,6 +2974,9 @@ static int igb_sw_init(struct igb_adapter
> *adapter)
>         adapter->rx_itr_setting = IGB_DEFAULT_ITR;
>         adapter->tx_itr_setting = IGB_DEFAULT_ITR;
>  
> +       if (InterruptThrottleRate == 0)
> +               adapter->rx_itr_setting = InterruptThrottleRate;
> +
>         /* set default work limits */
>         adapter->tx_work_limit = IGB_DEFAULT_TX_WORK;

If you are going to use InterruptThrottleRate you might as well use it
fully instead of making it a boolean value with 0 having one meaning
and the default of 3 being another.

Thanks.

- Alex

[toc] | [standalone]

Back to top | Article view | linux.kernel

csiph-web