Re: [dpdk-dev] [PATCH v5 1/2] test/power: add delay before checking cpuinfo cur freq

[Richael Zhuang <Richael.Zhuang@arm.com>]

> -----Original Message-----
> From: David Hunt <david.hunt@intel.com>
> Sent: Tuesday, April 20, 2021 8:39 PM
> To: Richael Zhuang <Richael.Zhuang@arm.com>; dev@dpdk.org
> Cc: nd <nd@arm.com>; alan.carew@intel.com; stable@dpdk.org; Pablo de
> Lara <pablo.de.lara.guarch@intel.com>
> Subject: Re: [PATCH v5 1/2] test/power: add delay before checking cpuinfo
> cur freq
> 
> 
> On 15/4/2021 6:59 AM, Richael Zhuang wrote:
> > For some platforms the newly-set frequency may not be effective
> > immediately. If we didn't get the right value from cpuinfo_cur_freq
> > immediately, add 10ms delay each time before rechecking until timeout.
> >
> >  From our test, for some arm platforms, it requires up to 700ms when
> > going from a minimum to a maximum frequency. And it's not the
> > driver/software issue.
> >
> > Fixes: ed7c51a6a680 ("app/test: vm power management")
> > Cc: alan.carew@intel.com
> > Cc: stable@dpdk.org
> >
> > Signed-off-by: Richael Zhuang <richael.zhuang@arm.com>
> > ---
> >   app/test/test_power_cpufreq.c | 27 ++++++++++++++++++++++-----
> >   1 file changed, 22 insertions(+), 5 deletions(-)
> >
> > diff --git a/app/test/test_power_cpufreq.c
> > b/app/test/test_power_cpufreq.c index 731c6b4dc..d47b3e0a1 100644
> > --- a/app/test/test_power_cpufreq.c
> > +++ b/app/test/test_power_cpufreq.c
> > @@ -8,6 +8,7 @@
> >   #include <limits.h>
> >   #include <string.h>
> >   #include <inttypes.h>
> > +#include <rte_cycles.h>
> >
> >   #include "test.h"
> >
> > @@ -44,11 +45,13 @@ static int
> >   check_cur_freq(unsigned lcore_id, uint32_t idx)
> >   {
> >   #define TEST_POWER_CONVERT_TO_DECIMAL 10
> > +#define MAX_LOOP 100
> >   	FILE *f;
> >   	char fullpath[PATH_MAX];
> >   	char buf[BUFSIZ];
> >   	uint32_t cur_freq;
> >   	int ret = -1;
> > +	int i;
> >
> >   	if (snprintf(fullpath, sizeof(fullpath),
> >   		TEST_POWER_SYSFILE_CUR_FREQ, lcore_id) < 0) { @@ -58,13
> +61,27 @@
> > check_cur_freq(unsigned lcore_id, uint32_t idx)
> >   	if (f == NULL) {
> >   		return 0;
> >   	}
> > -	if (fgets(buf, sizeof(buf), f) == NULL) {
> > -		goto fail_get_cur_freq;
> > +	for (i = 0; i < MAX_LOOP; i++) {
> > +		fflush(f);
> > +		if (fgets(buf, sizeof(buf), f) == NULL)
> > +			goto fail_all;
> > +
> > +		cur_freq = strtoul(buf, NULL,
> TEST_POWER_CONVERT_TO_DECIMAL);
> > +		ret = (freqs[idx] == cur_freq ? 0 : -1);
> > +
> > +		if (ret == 0)
> > +			break;
> > +
> > +		if (fseek(f, 0, SEEK_SET) < 0) {
> > +			printf("Fail to set file position indicator to 0\n");
> > +			goto fail_all;
> > +		}
> > +
> > +		/* wait for the value to be updated */
> > +		rte_delay_ms(10);
> >   	}
> > -	cur_freq = strtoul(buf, NULL, TEST_POWER_CONVERT_TO_DECIMAL);
> > -	ret = (freqs[idx] == cur_freq ? 0 : -1);
> >
> > -fail_get_cur_freq:
> > +fail_all:
> >   	fclose(f);
> >
> >   	return ret;
> 
> Hi Richael
> 
> On your system, is the current cpu frequency found in cpuinfo_cur_freq or in
> scaling_cur_freq? On my system, which uses intel_pstate driver, there is no
> file called cpuinfo_cur_freq, but the test works when I change
> TEST_POWER_SYSFILE_CUR_FREQ to scaling_cur_freq.
> 
> I know that's unrelated to your patch above, but it migth be worth using a file
> common to all platforms, or else attempting to open one, and if that fails, try
> open the other.
> 
> Rgds,
> Dave.
> 
Hi David,
Thanks for review. We have cpuinfo_cur_freq on our platform. For acpi_cpufreq it's also  cpuinfo_cur_freq. For pstate, the check is skipped because there is no such file. 

Best Regards,
Richael