[dpdk-dev] [PATCH] eal: fix TSC resolution in hz for ppc_64 architecture

[Jerin Jacob]

-----Original Message-----
> Date: Thu, 21 Sep 2017 17:11:26 +0530
> From: Gowrishankar <gowrishankar.m at linux.vnet.ibm.com>
> To: dev at dpdk.org
> CC: Chao Zhu <chaozhu at linux.vnet.ibm.com>, Bruce Richardson
>  <bruce.richardson at intel.com>, Konstantin Ananyev
>  <konstantin.ananyev at intel.com>, Jerin Jacob
>  <jerin.jacob at caviumnetworks.com>, Gowrishankar Muthukrishnan
>  <gowrishankar.m at linux.vnet.ibm.com>
> Subject: [PATCH] eal: fix TSC resolution in hz for ppc_64 architecture
> X-Mailer: git-send-email 1.9.1
> 
> From: Gowrishankar Muthukrishnan <gowrishankar.m at linux.vnet.ibm.com>
> 
> In ppc_64, rte_rdtsc() returns timebase register value which increments
> at independent timebase frequency and hence not related to lcore cpu
> frequency to derive into. In this patch, we fix get_tsc_freq() to not
> depend upon rte_rdtsc(), but obtain cpu current frequency from sysfs.
> 
> Signed-off-by: Gowrishankar Muthukrishnan <gowrishankar.m at linux.vnet.ibm.com>

Hi Gowrishankar,

> ---
>  Note:
>  * This patch would need minor port as per below patch (yet to upstream):
>    http://dpdk.org/dev/patchwork/patch/27527/
> 
>  lib/librte_eal/linuxapp/eal/eal_timer.c | 14 ++++++++++++++
>  1 file changed, 14 insertions(+)
> 
> diff --git a/lib/librte_eal/linuxapp/eal/eal_timer.c b/lib/librte_eal/linuxapp/eal/eal_timer.c
> index afa32f5..b8775cc 100644
> --- a/lib/librte_eal/linuxapp/eal/eal_timer.c
> +++ b/lib/librte_eal/linuxapp/eal/eal_timer.c
> @@ -55,8 +55,10 @@
>  
>  #include "eal_private.h"
>  #include "eal_internal_cfg.h"
> +#include "eal_filesystem.h"
>  
>  enum timer_source eal_timer_source = EAL_TIMER_HPET;
> +static const char sys_cpu_dir[] = "/sys/devices/system/cpu";
>  
>  #ifdef RTE_LIBEAL_USE_HPET
>  
> @@ -269,6 +271,17 @@ struct eal_hpet_regs {
>  uint64_t
>  get_tsc_freq(void)
>  {
> +#ifdef RTE_ARCH_PPC_64

I guess you can use the below series to avoid ifdef clutter in common code.
Right?

http://dpdk.org/ml/archives/dev/2017-August/072805.html.

If yes, then you could resend the series with ppc_64 change.i.e
removing following patch and add yours.
http://dpdk.org/dev/patchwork/patch/27527/


> +	unsigned long cpu_hz;
> +	char path[PATH_MAX];
> +
> +	snprintf(path, sizeof(path), "%s/cpu%d/cpufreq/cpuinfo_cur_freq",
> +			sys_cpu_dir, rte_get_master_lcore());
> +	if (eal_parse_sysfs_value(path, &cpu_hz) < 0)
> +		RTE_LOG(WARNING, EAL, "Unable to parse %s\n",
> +				path);
> +	return cpu_hz*1000;
> +#else
>  #ifdef CLOCK_MONOTONIC_RAW
>  #define NS_PER_SEC 1E9
>  
> @@ -290,6 +303,7 @@ struct eal_hpet_regs {
>  		return tsc_hz;
>  	}
>  #endif
> +#endif
>  	return 0;
>  }
>  
> -- 
> 1.9.1
>