[GIT PULL] turbostat fixes

[GIT PULL] turbostat fixes

[Len Brown]

Hi Rafael,

Please pull these turbostat patches.

thanks!
Len Brown, Intel Open Source Technology Center

The following changes since commit 3c2993b8c6143d8a5793746a54eba8f86f95240f:

  Linux 4.12-rc4 (2017-06-04 16:47:43 -0700)

are available in the git repository at:

  git://git.kernel.org/pub/scm/linux/kernel/git/lenb/linux.git turbostat

for you to fetch changes up to f7d44a8f3fd7f13770470a306a233acbaad5e96d:

  tools/power turbostat: update version number (2017-06-24 20:03:42 -0700)

----------------------------------------------------------------
Len Brown (5):
      tools/power turbostat: hide SKL counters, when not requested
      tools/power turbostat: if  --debug, print sampling overhead
      tools/power turbostat: stop migrating, unless '-m'
      tools/power turbostat: decode MSR_IA32_MISC_ENABLE only on Intel
      tools/power turbostat: update version number

 tools/power/x86/turbostat/turbostat.c | 94 ++++++++++++++++++++++++++++-------
 1 file changed, 77 insertions(+), 17 deletions(-)

[PATCH 1/5] tools/power turbostat: hide SKL counters, when not requested

[Len Brown]

From: Len Brown <len.brown@intel.com>

Skylake has some new counters, and they were erroneously
exempt  from --show and --hide

eg.

turbostat  --quiet --show CPU
CPU	Totl%C0	Any%C0	GFX%C0	CPUGFX%
-	116.73	90.56	85.69	79.00
0	117.78	91.38	86.47	79.71
2
1
3

is now

CPU
-
0
2
1
3

Signed-off-by: Len Brown <len.brown@intel.com>
---
 tools/power/x86/turbostat/turbostat.c | 58 +++++++++++++++++++++++++++--------
 1 file changed, 45 insertions(+), 13 deletions(-)

diff --git a/tools/power/x86/turbostat/turbostat.c b/tools/power/x86/turbostat/turbostat.c
index b112947..2b25727 100644
--- a/tools/power/x86/turbostat/turbostat.c
+++ b/tools/power/x86/turbostat/turbostat.c
@@ -57,7 +57,6 @@ unsigned int list_header_only;
 unsigned int dump_only;
 unsigned int do_snb_cstates;
 unsigned int do_knl_cstates;
-unsigned int do_skl_residency;
 unsigned int do_slm_cstates;
 unsigned int use_c1_residency_msr;
 unsigned int has_aperf;
@@ -384,8 +383,14 @@ struct msr_counter bic[] = {
 	{ 0x0, "CPU" },
 	{ 0x0, "Mod%c6" },
 	{ 0x0, "sysfs" },
+	{ 0x0, "Totl%C0" },
+	{ 0x0, "Any%C0" },
+	{ 0x0, "GFX%C0" },
+	{ 0x0, "CPUGFX%" },
 };
 
+
+
 #define MAX_BIC (sizeof(bic) / sizeof(struct msr_counter))
 #define	BIC_Package	(1ULL << 0)
 #define	BIC_Avg_MHz	(1ULL << 1)
@@ -426,6 +431,10 @@ struct msr_counter bic[] = {
 #define	BIC_CPU		(1ULL << 36)
 #define	BIC_Mod_c6	(1ULL << 37)
 #define	BIC_sysfs	(1ULL << 38)
+#define	BIC_Totl_c0	(1ULL << 39)
+#define	BIC_Any_c0	(1ULL << 40)
+#define	BIC_GFX_c0	(1ULL << 41)
+#define	BIC_CPUGFX	(1ULL << 42)
 
 unsigned long long bic_enabled = 0xFFFFFFFFFFFFFFFFULL;
 unsigned long long bic_present = BIC_sysfs;
@@ -599,12 +608,14 @@ void print_header(char *delim)
 	if (DO_BIC(BIC_GFXMHz))
 		outp += sprintf(outp, "%sGFXMHz", (printed++ ? delim : ""));
 
-	if (do_skl_residency) {
+	if (DO_BIC(BIC_Totl_c0))
 		outp += sprintf(outp, "%sTotl%%C0", (printed++ ? delim : ""));
+	if (DO_BIC(BIC_Any_c0))
 		outp += sprintf(outp, "%sAny%%C0", (printed++ ? delim : ""));
+	if (DO_BIC(BIC_GFX_c0))
 		outp += sprintf(outp, "%sGFX%%C0", (printed++ ? delim : ""));
+	if (DO_BIC(BIC_CPUGFX))
 		outp += sprintf(outp, "%sCPUGFX%%", (printed++ ? delim : ""));
-	}
 
 	if (DO_BIC(BIC_Pkgpc2))
 		outp += sprintf(outp, "%sPkg%%pc2", (printed++ ? delim : ""));
@@ -912,12 +923,14 @@ int format_counters(struct thread_data *t, struct core_data *c,
 		outp += sprintf(outp, "%s%d", (printed++ ? delim : ""), p->gfx_mhz);
 
 	/* Totl%C0, Any%C0 GFX%C0 CPUGFX% */
-	if (do_skl_residency) {
+	if (DO_BIC(BIC_Totl_c0))
 		outp += sprintf(outp, "%s%.2f", (printed++ ? delim : ""), 100.0 * p->pkg_wtd_core_c0/tsc);
+	if (DO_BIC(BIC_Any_c0))
 		outp += sprintf(outp, "%s%.2f", (printed++ ? delim : ""), 100.0 * p->pkg_any_core_c0/tsc);
+	if (DO_BIC(BIC_GFX_c0))
 		outp += sprintf(outp, "%s%.2f", (printed++ ? delim : ""), 100.0 * p->pkg_any_gfxe_c0/tsc);
+	if (DO_BIC(BIC_CPUGFX))
 		outp += sprintf(outp, "%s%.2f", (printed++ ? delim : ""), 100.0 * p->pkg_both_core_gfxe_c0/tsc);
-	}
 
 	if (DO_BIC(BIC_Pkgpc2))
 		outp += sprintf(outp, "%s%.2f", (printed++ ? delim : ""), 100.0 * p->pc2/tsc);
@@ -1038,12 +1051,16 @@ delta_package(struct pkg_data *new, struct pkg_data *old)
 	int i;
 	struct msr_counter *mp;
 
-	if (do_skl_residency) {
+
+	if (DO_BIC(BIC_Totl_c0))
 		old->pkg_wtd_core_c0 = new->pkg_wtd_core_c0 - old->pkg_wtd_core_c0;
+	if (DO_BIC(BIC_Any_c0))
 		old->pkg_any_core_c0 = new->pkg_any_core_c0 - old->pkg_any_core_c0;
+	if (DO_BIC(BIC_GFX_c0))
 		old->pkg_any_gfxe_c0 = new->pkg_any_gfxe_c0 - old->pkg_any_gfxe_c0;
+	if (DO_BIC(BIC_CPUGFX))
 		old->pkg_both_core_gfxe_c0 = new->pkg_both_core_gfxe_c0 - old->pkg_both_core_gfxe_c0;
-	}
+
 	old->pc2 = new->pc2 - old->pc2;
 	if (DO_BIC(BIC_Pkgpc3))
 		old->pc3 = new->pc3 - old->pc3;
@@ -1292,12 +1309,14 @@ int sum_counters(struct thread_data *t, struct core_data *c,
 	if (!(t->flags & CPU_IS_FIRST_CORE_IN_PACKAGE))
 		return 0;
 
-	if (do_skl_residency) {
+	if (DO_BIC(BIC_Totl_c0))
 		average.packages.pkg_wtd_core_c0 += p->pkg_wtd_core_c0;
+	if (DO_BIC(BIC_Any_c0))
 		average.packages.pkg_any_core_c0 += p->pkg_any_core_c0;
+	if (DO_BIC(BIC_GFX_c0))
 		average.packages.pkg_any_gfxe_c0 += p->pkg_any_gfxe_c0;
+	if (DO_BIC(BIC_CPUGFX))
 		average.packages.pkg_both_core_gfxe_c0 += p->pkg_both_core_gfxe_c0;
-	}
 
 	average.packages.pc2 += p->pc2;
 	if (DO_BIC(BIC_Pkgpc3))
@@ -1357,12 +1376,14 @@ void compute_average(struct thread_data *t, struct core_data *c,
 	average.cores.c7 /= topo.num_cores;
 	average.cores.mc6_us /= topo.num_cores;
 
-	if (do_skl_residency) {
+	if (DO_BIC(BIC_Totl_c0))
 		average.packages.pkg_wtd_core_c0 /= topo.num_packages;
+	if (DO_BIC(BIC_Any_c0))
 		average.packages.pkg_any_core_c0 /= topo.num_packages;
+	if (DO_BIC(BIC_GFX_c0))
 		average.packages.pkg_any_gfxe_c0 /= topo.num_packages;
+	if (DO_BIC(BIC_CPUGFX))
 		average.packages.pkg_both_core_gfxe_c0 /= topo.num_packages;
-	}
 
 	average.packages.pc2 /= topo.num_packages;
 	if (DO_BIC(BIC_Pkgpc3))
@@ -1603,13 +1624,19 @@ int get_counters(struct thread_data *t, struct core_data *c, struct pkg_data *p)
 	if (!(t->flags & CPU_IS_FIRST_CORE_IN_PACKAGE))
 		return 0;
 
-	if (do_skl_residency) {
+	if (DO_BIC(BIC_Totl_c0)) {
 		if (get_msr(cpu, MSR_PKG_WEIGHTED_CORE_C0_RES, &p->pkg_wtd_core_c0))
 			return -10;
+	}
+	if (DO_BIC(BIC_Any_c0)) {
 		if (get_msr(cpu, MSR_PKG_ANY_CORE_C0_RES, &p->pkg_any_core_c0))
 			return -11;
+	}
+	if (DO_BIC(BIC_GFX_c0)) {
 		if (get_msr(cpu, MSR_PKG_ANY_GFXE_C0_RES, &p->pkg_any_gfxe_c0))
 			return -12;
+	}
+	if (DO_BIC(BIC_CPUGFX)) {
 		if (get_msr(cpu, MSR_PKG_BOTH_CORE_GFXE_C0_RES, &p->pkg_both_core_gfxe_c0))
 			return -13;
 	}
@@ -4198,7 +4225,12 @@ void process_cpuid()
 		BIC_PRESENT(BIC_Pkgpc10);
 	}
 	do_irtl_hsw = has_hsw_msrs(family, model);
-	do_skl_residency = has_skl_msrs(family, model);
+	if (has_skl_msrs(family, model)) {
+		BIC_PRESENT(BIC_Totl_c0);
+		BIC_PRESENT(BIC_Any_c0);
+		BIC_PRESENT(BIC_GFX_c0);
+		BIC_PRESENT(BIC_CPUGFX);
+	}
 	do_slm_cstates = is_slm(family, model);
 	do_knl_cstates  = is_knl(family, model);
 
-- 
2.7.4

[PATCH 2/5] tools/power turbostat: if --debug, print sampling overhead

[Len Brown]

From: Len Brown <len.brown@intel.com>

The --debug option now pre-pends each row with
the number  of micro-seconds [usec] to collect
the finishing snapshot for that row.

Signed-off-by: Len Brown <len.brown@intel.com>
---
 tools/power/x86/turbostat/turbostat.c | 21 +++++++++++++++++++--
 1 file changed, 19 insertions(+), 2 deletions(-)

diff --git a/tools/power/x86/turbostat/turbostat.c b/tools/power/x86/turbostat/turbostat.c
index 2b25727..6f7c64a 100644
--- a/tools/power/x86/turbostat/turbostat.c
+++ b/tools/power/x86/turbostat/turbostat.c
@@ -150,6 +150,8 @@ size_t cpu_present_setsize, cpu_affinity_setsize, cpu_subset_size;
 #define MAX_ADDED_COUNTERS 16
 
 struct thread_data {
+	struct timeval tv_begin;
+	struct timeval tv_end;
 	unsigned long long tsc;
 	unsigned long long aperf;
 	unsigned long long mperf;
@@ -530,6 +532,8 @@ void print_header(char *delim)
 	struct msr_counter *mp;
 	int printed = 0;
 
+	if (debug)
+		outp += sprintf(outp, "usec %s", delim);
 	if (DO_BIC(BIC_Package))
 		outp += sprintf(outp, "%sPackage", (printed++ ? delim : ""));
 	if (DO_BIC(BIC_Core))
@@ -782,6 +786,14 @@ int format_counters(struct thread_data *t, struct core_data *c,
 		(cpu_subset && !CPU_ISSET_S(t->cpu_id, cpu_subset_size, cpu_subset)))
 		return 0;
 
+	if (debug) {
+		/* on each row, print how many usec each timestamp took to gather */
+		struct timeval tv;
+
+		timersub(&t->tv_end, &t->tv_begin, &tv);
+		outp += sprintf(outp, "%5ld\t", tv.tv_sec * 1000000 + tv.tv_usec);
+	}
+
 	interval_float = tv_delta.tv_sec + tv_delta.tv_usec/1000000.0;
 
 	tsc = t->tsc * tsc_tweak;
@@ -1503,6 +1515,9 @@ int get_counters(struct thread_data *t, struct core_data *c, struct pkg_data *p)
 	struct msr_counter *mp;
 	int i;
 
+
+	gettimeofday(&t->tv_begin, (struct timezone *)NULL);
+
 	if (cpu_migrate(cpu)) {
 		fprintf(outf, "Could not migrate to CPU %d\n", cpu);
 		return -1;
@@ -1586,7 +1601,7 @@ int get_counters(struct thread_data *t, struct core_data *c, struct pkg_data *p)
 
 	/* collect core counters only for 1st thread in core */
 	if (!(t->flags & CPU_IS_FIRST_THREAD_IN_CORE))
-		return 0;
+		goto done;
 
 	if (DO_BIC(BIC_CPU_c3) && !do_slm_cstates && !do_knl_cstates) {
 		if (get_msr(cpu, MSR_CORE_C3_RESIDENCY, &c->c3))
@@ -1622,7 +1637,7 @@ int get_counters(struct thread_data *t, struct core_data *c, struct pkg_data *p)
 
 	/* collect package counters only for 1st core in package */
 	if (!(t->flags & CPU_IS_FIRST_CORE_IN_PACKAGE))
-		return 0;
+		goto done;
 
 	if (DO_BIC(BIC_Totl_c0)) {
 		if (get_msr(cpu, MSR_PKG_WEIGHTED_CORE_C0_RES, &p->pkg_wtd_core_c0))
@@ -1715,6 +1730,8 @@ int get_counters(struct thread_data *t, struct core_data *c, struct pkg_data *p)
 		if (get_mp(cpu, mp, &p->counter[i]))
 			return -10;
 	}
+done:
+	gettimeofday(&t->tv_end, (struct timezone *)NULL);
 
 	return 0;
 }
-- 
2.7.4

[PATCH 3/5] tools/power turbostat: stop migrating, unless '-m'

[Len Brown]

From: Len Brown <len.brown@intel.com>

Turbostat has the capability to set its own affinity to
each CPU so that its MSR accesses are on the local CPU.

However, using the in-kernel cross-call in  the msr driver
tends to be less invasive, so do that -- by-default.
'-m' remains to get the old behaviour.

Signed-off-by: Len Brown <len.brown@intel.com>
---
 tools/power/x86/turbostat/turbostat.c | 10 +++++++++-
 1 file changed, 9 insertions(+), 1 deletion(-)

diff --git a/tools/power/x86/turbostat/turbostat.c b/tools/power/x86/turbostat/turbostat.c
index 6f7c64a..1a3a5b4 100644
--- a/tools/power/x86/turbostat/turbostat.c
+++ b/tools/power/x86/turbostat/turbostat.c
@@ -92,6 +92,7 @@ unsigned int do_ring_perf_limit_reasons;
 unsigned int crystal_hz;
 unsigned long long tsc_hz;
 int base_cpu;
+int do_migrate;
 double discover_bclk(unsigned int family, unsigned int model);
 unsigned int has_hwp;	/* IA32_PM_ENABLE, IA32_HWP_CAPABILITIES */
 			/* IA32_HWP_REQUEST, IA32_HWP_STATUS */
@@ -302,6 +303,9 @@ int for_all_cpus(int (func)(struct thread_data *, struct core_data *, struct pkg
 
 int cpu_migrate(int cpu)
 {
+	if (!do_migrate)
+		return 0;
+
 	CPU_ZERO_S(cpu_affinity_setsize, cpu_affinity_set);
 	CPU_SET_S(cpu, cpu_affinity_setsize, cpu_affinity_set);
 	if (sched_setaffinity(0, cpu_affinity_setsize, cpu_affinity_set) == -1)
@@ -5000,6 +5004,7 @@ void cmdline(int argc, char **argv)
 		{"hide",	required_argument,	0, 'H'},	// meh, -h taken by --help
 		{"Joules",	no_argument,		0, 'J'},
 		{"list",	no_argument,		0, 'l'},
+		{"migrate",	no_argument,		0, 'm'},
 		{"out",		required_argument,	0, 'o'},
 		{"quiet",	no_argument,		0, 'q'},
 		{"show",	required_argument,	0, 's'},
@@ -5011,7 +5016,7 @@ void cmdline(int argc, char **argv)
 
 	progname = argv[0];
 
-	while ((opt = getopt_long_only(argc, argv, "+C:c:Ddhi:JM:m:o:qST:v",
+	while ((opt = getopt_long_only(argc, argv, "+C:c:Ddhi:Jmo:qST:v",
 				long_options, &option_index)) != -1) {
 		switch (opt) {
 		case 'a':
@@ -5054,6 +5059,9 @@ void cmdline(int argc, char **argv)
 			list_header_only++;
 			quiet++;
 			break;
+		case 'm':
+			do_migrate = 1;
+			break;
 		case 'o':
 			outf = fopen_or_die(optarg, "w");
 			break;
-- 
2.7.4

[PATCH 4/5] tools/power turbostat: decode MSR_IA32_MISC_ENABLE only on Intel

[Len Brown]

From: Len Brown <len.brown@intel.com>

otherwise, turbostat bails on on AMD Opteron boxes:

turbostat: cpu26: msr offset 0x1a0 read failed: Input/output error

Reported-by: Kamil Kolakowski <kkolakow@redhat.com>
Signed-off-by: Len Brown <len.brown@intel.com>
---
 tools/power/x86/turbostat/turbostat.c | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/tools/power/x86/turbostat/turbostat.c b/tools/power/x86/turbostat/turbostat.c
index 1a3a5b4..33992a9 100644
--- a/tools/power/x86/turbostat/turbostat.c
+++ b/tools/power/x86/turbostat/turbostat.c
@@ -3943,6 +3943,9 @@ void decode_misc_enable_msr(void)
 {
 	unsigned long long msr;
 
+	if (!genuine_intel)
+		return;
+
 	if (!get_msr(base_cpu, MSR_IA32_MISC_ENABLE, &msr))
 		fprintf(outf, "cpu%d: MSR_IA32_MISC_ENABLE: 0x%08llx (%sTCC %sEIST %sMWAIT %sPREFETCH %sTURBO)\n",
 			base_cpu, msr,
-- 
2.7.4

[PATCH 5/5] tools/power turbostat: update version number

[Len Brown]

From: Len Brown <len.brown@intel.com>

Signed-off-by: Len Brown <len.brown@intel.com>
---
 tools/power/x86/turbostat/turbostat.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/tools/power/x86/turbostat/turbostat.c b/tools/power/x86/turbostat/turbostat.c
index 33992a9..0dafba2 100644
--- a/tools/power/x86/turbostat/turbostat.c
+++ b/tools/power/x86/turbostat/turbostat.c
@@ -4634,7 +4634,7 @@ int get_and_dump_counters(void)
 }
 
 void print_version() {
-	fprintf(outf, "turbostat version 17.04.12"
+	fprintf(outf, "turbostat version 17.06.23"
 		" - Len Brown <lenb@kernel.org>\n");
 }
 
-- 
2.7.4

Re: [GIT PULL] turbostat fixes

[Rafael J. Wysocki]

On Sun, Jun 25, 2017 at 5:12 AM, Len Brown <lenb@kernel.org> wrote:
> Hi Rafael,

Hi,

> Please pull these turbostat patches.
>
> thanks!
> Len Brown, Intel Open Source Technology Center
>
> The following changes since commit 3c2993b8c6143d8a5793746a54eba8f86f95240f:
>
>   Linux 4.12-rc4 (2017-06-04 16:47:43 -0700)
>
> are available in the git repository at:
>
>   git://git.kernel.org/pub/scm/linux/kernel/git/lenb/linux.git turbostat
>
> for you to fetch changes up to f7d44a8f3fd7f13770470a306a233acbaad5e96d:
>
>   tools/power turbostat: update version number (2017-06-24 20:03:42 -0700)
>
> ----------------------------------------------------------------
> Len Brown (5):
>       tools/power turbostat: hide SKL counters, when not requested
>       tools/power turbostat: if  --debug, print sampling overhead
>       tools/power turbostat: stop migrating, unless '-m'
>       tools/power turbostat: decode MSR_IA32_MISC_ENABLE only on Intel
>       tools/power turbostat: update version number
>
>  tools/power/x86/turbostat/turbostat.c | 94 ++++++++++++++++++++++++++++-------
>  1 file changed, 77 insertions(+), 17 deletions(-)

Pulled, thanks!