mirror of https://lore.kernel.org/lkml/
 help / color / mirror / Atom feed
From: "tan.shaopeng@fujitsu.com" <tan.shaopeng@fujitsu.com>
To: 'Shuah Khan' <skhan@linuxfoundation.org>,
	Fenghua Yu <fenghua.yu@intel.com>,
	Reinette Chatre <reinette.chatre@intel.com>,
	Shuah Khan <shuah@kernel.org>
Cc: "linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>,
	"linux-kselftest@vger.kernel.org"
	<linux-kselftest@vger.kernel.org>
Subject: RE: [PATCH v3] selftests/resctrl: Print a message if the result of MBM&CMT tests is failed on Intel cpu
Date: Thu, 10 Mar 2022 08:18:50 +0000	[thread overview]
Message-ID: <TYAPR01MB63307F1C04270DD2D09ACF238B0B9@TYAPR01MB6330.jpnprd01.prod.outlook.com> (raw)
In-Reply-To: <9647f7ed-6a02-8603-0de4-3292d4d13157@linuxfoundation.org>

Hi Shuah,

> On 3/4/22 3:39 AM, Shaopeng Tan wrote:
> > According to "Intel Resource Director Technology (Intel RDT) on 2nd
> > Generation Intel Xeon Scalable Processors Reference Manual", When the
> > Intel Sub-NUMA Clustering(SNC) feature is enabled, Intel CMT and MBM
> > counters may not be accurate.
> >
> > However, there does not seem to be an architectural way to detect if
> > SNC is enabled.
> >
> > If the result of MBM&CMT test fails on Intel CPU, print a message to
> > let users know a possible cause of failure.
> >
> > Signed-off-by: Shaopeng Tan <tan.shaopeng@jp.fujitsu.com>
> > ---
> > Hello,
> >
> > In PATCH V2, I tried to detect whether SNC is enabled by NUMA info and
> > cpuinfo(socket_num), but it is not reliable and no future-proof.
> >
> > I just print a message to let users know a possible cause of "not ok",
> > When CMT or MBM test runs on Intel CPU, and the result is "not ok".
> >
> > This patch is based on v5.16.
> 
> Also need to be rebased on mainline latest

I will rebased on mainline latest in next version.

> >
> >   tools/testing/selftests/resctrl/resctrl_tests.c | 12 +++++++++---
> >   1 file changed, 9 insertions(+), 3 deletions(-)
> >
> > diff --git a/tools/testing/selftests/resctrl/resctrl_tests.c
> > b/tools/testing/selftests/resctrl/resctrl_tests.c
> > index 973f09a66e1e..ec2bdce7b85f 100644
> > --- a/tools/testing/selftests/resctrl/resctrl_tests.c
> > +++ b/tools/testing/selftests/resctrl/resctrl_tests.c
> > @@ -14,8 +14,9 @@
> >   #define BENCHMARK_ARG_SIZE	64
> >
> >   bool is_amd;
> > +bool is_intel;
> >
> 
> Why is this a global? I am not seeing a reason. These detect_*()s could be
> moved to resctrl.h and get rid of globals.
> 
> Instead of adding intel check to detect_amd() add detect_intel() or is_intel()
> and have ut return true of it detects intel.

"is_amd" and "is_intel" are called many times,
in this way, detect_vendor is called only once.

Do you mean in the following way?
"/proc/cpuinfo" will be open/closed many times,
since is_amd()/is_intel() is called in serval places.
So, I think current is better. Do you have any thoughts.
----
bool detect_vendor_str(const char *vendor_str )
{
        FILE *inf = fopen("/proc/cpuinfo", "r");
        char *res;
        bool found = false;

        if (!inf)
                return found;

        res = fgrep(inf, "vendor_id");

        if (res) {
                char *s = strchr(res, ':');

                found = s && !strcmp(s,vendor_str);
                free(res);
        }
        fclose(inf);

        return found;
}

bool is_amd()
{
        return detect_vendor_str(": AuthenticAMD\n");
}
bool is_intel()
{
        return detect_vendor_str(": GenuineIntel\n");
}
----

> > -void detect_amd(void)
> > +void detect_vendor(void)
> >   {
> >   	FILE *inf = fopen("/proc/cpuinfo", "r");
> >   	char *res;
> > @@ -29,6 +30,7 @@ void detect_amd(void)
> >   		char *s = strchr(res, ':');
> >
> >   		is_amd = s && !strcmp(s, ": AuthenticAMD\n");
> > +		is_intel = s && !strcmp(s, ": GenuineIntel\n");
> >   		free(res);
> >   	}
> >   	fclose(inf);
> > @@ -70,6 +72,8 @@ static void run_mbm_test(bool has_ben, char
> **benchmark_cmd, int span,
> >   		sprintf(benchmark_cmd[5], "%s", MBA_STR);
> >   	res = mbm_bw_change(span, cpu_no, bw_report, benchmark_cmd);
> >   	ksft_test_result(!res, "MBM: bw change\n");
> > +	if (is_intel && res)
> > +		ksft_print_msg("Intel CMT and MBM counters may be
> inaccurate when
> > +Sub-NUMA Clustering (SNC) is enabled. Ensure SNC is disabled in the
> > +BIOS if this system supports SNC.\n");
> 
> This message is rather long. Please make it concise.

I will use the following message in next version.
"Intel MBM may be inaccurate when Sub-NUMA Clustering is enabled. Check BIOS configuration."

> >   	mbm_test_cleanup();
> >   }
> >
> > @@ -106,6 +110,8 @@ static void run_cmt_test(bool has_ben, char
> **benchmark_cmd, int cpu_no)
> >   		sprintf(benchmark_cmd[5], "%s", CMT_STR);
> >   	res = cmt_resctrl_val(cpu_no, 5, benchmark_cmd);
> >   	ksft_test_result(!res, "CMT: test\n");
> > +	if (is_intel && res)
> > +		ksft_print_msg("Intel CMT and MBM counters may be
> inaccurate when
> > +Sub-NUMA Clustering (SNC) is enabled. Ensure SNC is disabled in the
> > +BIOS if this system supports SNC.\n");
> 
> This message is rather long. Please make it concise.

"Intel CMT may be inaccurate when Sub-NUMA Clustering is enabled. Check BIOS configuration."

> >   	cmt_test_cleanup();
> >   }
> >
> > @@ -207,8 +213,8 @@ int main(int argc, char **argv)
> >   	if (geteuid() != 0)
> >   		return ksft_exit_fail_msg("Not running as root, abort
> > testing.\n");
> >
> > -	/* Detect AMD vendor */
> > -	detect_amd();
> > +	/* Detect AMD/INTEL vendor */
> > +	detect_vendor();
> >
> >   	if (has_ben) {
> >   		/* Extract benchmark command from command line. */
> >

Best regards,
Tan Shaopeng

  reply	other threads:[~2022-03-10  8:20 UTC|newest]

Thread overview: 4+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2022-03-04 10:39 Shaopeng Tan
2022-03-08 21:38 ` Shuah Khan
2022-03-10  8:18   ` tan.shaopeng [this message]
2022-03-10 20:32     ` Shuah Khan

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=TYAPR01MB63307F1C04270DD2D09ACF238B0B9@TYAPR01MB6330.jpnprd01.prod.outlook.com \
    --to=tan.shaopeng@fujitsu.com \
    --cc=fenghua.yu@intel.com \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux-kselftest@vger.kernel.org \
    --cc=reinette.chatre@intel.com \
    --cc=shuah@kernel.org \
    --cc=skhan@linuxfoundation.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox

all inboxes | Powered by JetHome®