[PATCH v2] powerpc/perf: Adds support for programming of Thresholding in P10

@@ -36,9 +36,9 @@ struct power_pmu {
 	unsigned long	test_adder;
 	int		(*compute_mmcr)(u64 events[], int n_ev,
 				unsigned int hwc[], struct mmcr_regs *mmcr,
-				struct perf_event *pevents[]);
+				struct perf_event *pevents[], u32 flags);
 	int		(*get_constraint)(u64 event_id, unsigned long *mskp,
-				unsigned long *valp);
+				unsigned long *valp, u64 event_config1);
 	int		(*get_alternatives)(u64 event_id, unsigned int flags,
 				u64 alt[]);
 	void		(*get_mem_data_src)(union perf_mem_data_src *dsrc,

@@ -83,6 +83,7 @@ struct power_pmu {
 #define PPMU_NO_SIAR		0x00000100 /* Do not use SIAR */
 #define PPMU_ARCH_31		0x00000200 /* Has MMCR3, SIER2 and SIER3 */
 #define PPMU_P10_DD1		0x00000400 /* Is power10 DD1 processor version */
+#define PPMU_HAS_ATTR_CONFIG1	0x00000800 /* Using config1 attribute */
 
 /*
  * Values for flags to get_alternatives()

@@ -36,6 +36,7 @@ struct cpu_hw_events {
 	u8  pmcs_enabled;
 	struct perf_event *event[MAX_HWEVENTS];
 	u64 events[MAX_HWEVENTS];
+	u64 events_config1[MAX_HWEVENTS];
 	unsigned int flags[MAX_HWEVENTS];
 	struct mmcr_regs mmcr;
 	struct perf_event *limited_counter[MAX_LIMITED_HWCOUNTERS];

@@ -938,7 +939,7 @@ static int power_check_constraints(struct cpu_hw_events *cpuhw,
 			event_id[i] = cpuhw->alternatives[i][0];
 		}
 		if (ppmu->get_constraint(event_id[i], &cpuhw->amasks[i][0],
-					 &cpuhw->avalues[i][0]))
+					 &cpuhw->avalues[i][0], cpuhw->events_config1[i]))
 			return -1;
 	}
 	value = mask = 0;

@@ -973,7 +974,8 @@ static int power_check_constraints(struct cpu_hw_events *cpuhw,
 		for (j = 1; j < n_alt[i]; ++j)
 			ppmu->get_constraint(cpuhw->alternatives[i][j],
 					     &cpuhw->amasks[i][j],
-					     &cpuhw->avalues[i][j]);
+					     &cpuhw->avalues[i][j],
+					     cpuhw->events_config1[i]);
 	}
 
 	/* enumerate all possibilities and see if any will work */

@@ -1391,7 +1393,7 @@ static void power_pmu_enable(struct pmu *pmu)
 	memset(&cpuhw->mmcr, 0, sizeof(cpuhw->mmcr));
 
 	if (ppmu->compute_mmcr(cpuhw->events, cpuhw->n_events, hwc_index,
-			       &cpuhw->mmcr, cpuhw->event)) {
+			       &cpuhw->mmcr, cpuhw->event, ppmu->flags)) {
 		/* shouldn't ever get here */
 		printk(KERN_ERR "oops compute_mmcr failed\n");
 		goto out;

@@ -1505,7 +1507,7 @@ static void power_pmu_enable(struct pmu *pmu)
 
 static int collect_events(struct perf_event *group, int max_count,
 			  struct perf_event *ctrs[], u64 *events,
-			  unsigned int *flags)
+			  unsigned int *flags, u64 *events_config1)
 {
 	int n = 0;
 	struct perf_event *event;

@@ -1515,6 +1517,7 @@ static int collect_events(struct perf_event *group, int max_count,
 			return -1;
 		ctrs[n] = group;
 		flags[n] = group->hw.event_base;
+		events_config1[n] = group->hw.extra_reg.config;
 		events[n++] = group->hw.config;
 	}
 	for_each_sibling_event(event, group) {

@@ -1524,6 +1527,7 @@ static int collect_events(struct perf_event *group, int max_count,
 				return -1;
 			ctrs[n] = event;
 			flags[n] = event->hw.event_base;
+			events_config1[n] = event->hw.extra_reg.config;
 			events[n++] = event->hw.config;
 		}
 	}

@@ -1556,6 +1560,7 @@ static int power_pmu_add(struct perf_event *event, int ef_flags)
 		goto out;
 	cpuhw->event[n0] = event;
 	cpuhw->events[n0] = event->hw.config;
+	cpuhw->events_config1[n0] = event->hw.extra_reg.config;
 	cpuhw->flags[n0] = event->hw.event_base;
 
 	/*

@@ -1582,6 +1587,7 @@ static int power_pmu_add(struct perf_event *event, int ef_flags)
 	if (power_check_constraints(cpuhw, cpuhw->events, cpuhw->flags, n0 + 1))
 		goto out;
 	event->hw.config = cpuhw->events[n0];
+	event->hw.extra_reg.config = cpuhw->events_config1[n0];
 
 nocheck:
 	ebb_event_add(event);

@@ -1629,6 +1635,7 @@ static void power_pmu_del(struct perf_event *event, int ef_flags)
 			while (++i < cpuhw->n_events) {
 				cpuhw->event[i-1] = cpuhw->event[i];
 				cpuhw->events[i-1] = cpuhw->events[i];
+				cpuhw->events_config1[i - 1] = cpuhw->events_config1[i];
 				cpuhw->flags[i-1] = cpuhw->flags[i];
 			}
 			--cpuhw->n_events;

@@ -1793,8 +1800,10 @@ static int power_pmu_commit_txn(struct pmu *pmu)
 	if (i < 0)
 		return -EAGAIN;
 
-	for (i = cpuhw->n_txn_start; i < n; ++i)
+	for (i = cpuhw->n_txn_start; i < n; ++i) {
 		cpuhw->event[i]->hw.config = cpuhw->events[i];
+		cpuhw->event[i]->hw.extra_reg.config = cpuhw->events_config1[i];
+	}
 
 	cpuhw->txn_flags = 0;
 	perf_pmu_enable(pmu);

@@ -1919,6 +1928,7 @@ static int power_pmu_event_init(struct perf_event *event)
 	unsigned long flags, irq_flags;
 	struct perf_event *ctrs[MAX_HWEVENTS];
 	u64 events[MAX_HWEVENTS];
+	u64 events_config1[MAX_HWEVENTS];
 	unsigned int cflags[MAX_HWEVENTS];
 	int n;
 	int err;

@@ -2014,11 +2024,12 @@ static int power_pmu_event_init(struct perf_event *event)
 	n = 0;
 	if (event->group_leader != event) {
 		n = collect_events(event->group_leader, ppmu->n_counter - 1,
-				   ctrs, events, cflags);
+				   ctrs, events, cflags, events_config1);
 		if (n < 0)
 			return -EINVAL;
 	}
 	events[n] = ev;
+	events_config1[n] = event->attr.config1;
 	ctrs[n] = event;
 	cflags[n] = flags;
 	if (check_excludes(ctrs, cflags, n, 1))

@@ -2048,6 +2059,7 @@ static int power_pmu_event_init(struct perf_event *event)
 		return -EINVAL;
 
 	event->hw.config = events[n];
+	event->hw.extra_reg.config = events_config1[n];
 	event->hw.event_base = cflags[n];
 	event->hw.last_period = event->hw.sample_period;
 	local64_set(&event->hw.period_left, event->hw.last_period);

@@ -108,12 +108,57 @@ static void mmcra_sdar_mode(u64 event, unsigned long *mmcra)
 		*mmcra |= MMCRA_SDAR_MODE_TLB;
 }
 
+static u64 p10_thresh_cmp_val(u64 value)
+{
+	int exp = 0;
+	u64 result = value;
+
+	if (!value)
+		return value;
+
+	/*
+	 * Incase of P10, thresh_cmp value is not part of raw event code
+	 * and provided via attr.config1 parameter. To program threshold in MMCRA,
+	 * take a 18 bit number N and shift right 2 places and increment
+	 * the exponent E by 1 until the upper 10 bits of N are zero.
+	 * Write E to the threshold exponent and write the lower 8 bits of N
+	 * to the threshold mantissa.
+	 * The max threshold that can be written is 261120.
+	 */
+	if (cpu_has_feature(CPU_FTR_ARCH_31)) {
+		if (value > 261120)
+			value = 261120;
+		while ((64 - __builtin_clzl(value)) > 8) {
+			exp++;
+			value >>= 2;
+		}
+
+		/*
+		 * Note that it is invalid to write a mantissa with the
+		 * upper 2 bits of mantissa being zero, unless the
+		 * exponent is also zero.
+		 */
+		if (!(value & 0xC0) && exp)
+			result = 0;
+		else
+			result = (exp << 8) | value;
+	}
+	return result;
+}
+
 static u64 thresh_cmp_val(u64 value)
 {
+	if (cpu_has_feature(CPU_FTR_ARCH_31))
+		value = p10_thresh_cmp_val(value);
+
+	/*
+	 * Since location of threshold compare bits in MMCRA
+	 * is different for p8, using different shift value.
+	 */
 	if (cpu_has_feature(CPU_FTR_ARCH_300))
 		return value << p9_MMCRA_THR_CMP_SHIFT;
-
-	return value << MMCRA_THR_CMP_SHIFT;
+	else
+		return value << MMCRA_THR_CMP_SHIFT;
 }
 
 static unsigned long combine_from_event(u64 event)

@@ -141,13 +186,17 @@ static bool is_thresh_cmp_valid(u64 event)
 {
 	unsigned int cmp, exp;
 
+	if (cpu_has_feature(CPU_FTR_ARCH_31)) {
+		if (!p10_thresh_cmp_val(event))
+			return false;
+		else
+			return true;
+	}
+
 	/*
 	 * Check the mantissa upper two bits are not zero, unless the
 	 * exponent is also zero. See the THRESH_CMP_MANTISSA doc.
-	 * Power10: thresh_cmp is replaced by l2_l3 event select.
 	 */
-	if (cpu_has_feature(CPU_FTR_ARCH_31))
-		return false;
 
 	cmp = (event >> EVENT_THR_CMP_SHIFT) & EVENT_THR_CMP_MASK;
 	exp = cmp >> 7;

@@ -256,7 +305,7 @@ void isa207_get_mem_weight(u64 *weight)
 		*weight = mantissa << (2 * exp);
 }
 
-int isa207_get_constraint(u64 event, unsigned long *maskp, unsigned long *valp)
+int isa207_get_constraint(u64 event, unsigned long *maskp, unsigned long *valp, u64 event_config1)
 {
 	unsigned int unit, pmc, cache, ebb;
 	unsigned long mask, value;

@@ -355,9 +404,11 @@ int isa207_get_constraint(u64 event, unsigned long *maskp, unsigned long *valp)
 	}
 
 	if (cpu_has_feature(CPU_FTR_ARCH_31)) {
-		if (event_is_threshold(event)) {
+		if (event_is_threshold(event) && is_thresh_cmp_valid(event_config1)) {
 			mask  |= CNST_THRESH_CTL_SEL_MASK;
 			value |= CNST_THRESH_CTL_SEL_VAL(event >> EVENT_THRESH_SHIFT);
+			mask  |= p10_CNST_THRESH_CMP_MASK;
+			value |= p10_CNST_THRESH_CMP_VAL(p10_thresh_cmp_val(event_config1));
 		}
 	} else if (cpu_has_feature(CPU_FTR_ARCH_300))  {
 		if (event_is_threshold(event) && is_thresh_cmp_valid(event)) {

@@ -411,7 +462,7 @@ int isa207_get_constraint(u64 event, unsigned long *maskp, unsigned long *valp)
 
 int isa207_compute_mmcr(u64 event[], int n_ev,
 			       unsigned int hwc[], struct mmcr_regs *mmcr,
-			       struct perf_event *pevents[])
+			       struct perf_event *pevents[], u32 flags)
 {
 	unsigned long mmcra, mmcr1, mmcr2, unit, combine, psel, cache, val;
 	unsigned long mmcr3;

@@ -504,6 +555,10 @@ int isa207_compute_mmcr(u64 event[], int n_ev,
 				val = (event[i] >> EVENT_THR_CMP_SHIFT) &
 					EVENT_THR_CMP_MASK;
 				mmcra |= thresh_cmp_val(val);
+			} else if (flags & PPMU_HAS_ATTR_CONFIG1) {
+				val = (pevents[i]->attr.config1 >> p10_EVENT_THR_CMP_SHIFT) &
+					p10_EVENT_THR_CMP_MASK;
+				mmcra |= thresh_cmp_val(val);
 			}
 		}
 

@@ -105,6 +105,10 @@
 #define p10_EVENT_RADIX_SCOPE_QUAL_MASK	0x1
 #define p10_MMCR1_RADIX_SCOPE_QUAL_SHIFT	45
 
+/* Event Threshold Compare bit constant for power10 in config1 attribute */
+#define p10_EVENT_THR_CMP_SHIFT        0
+#define p10_EVENT_THR_CMP_MASK 0x3FFFFull
+
 #define p10_EVENT_VALID_MASK		\
 	((p10_SDAR_MODE_MASK   << p10_SDAR_MODE_SHIFT		|	\
 	(p10_EVENT_THRESH_MASK  << EVENT_THRESH_SHIFT)		|	\

@@ -124,8 +128,8 @@
  *        60        56        52        48        44        40        36        32
  * | - - - - | - - - - | - - - - | - - - - | - - - - | - - - - | - - - - | - - - - |
  *   [   fab_match   ]         [       thresh_cmp      ] [   thresh_ctl    ] [   ]
- *                                                                             |
- *                                                                 thresh_sel -*
+ *                                          |                                  |
+ *                           [  thresh_cmp bits for p10]           thresh_sel -*
  *
  *        28        24        20        16        12         8         4         0
  * | - - - - | - - - - | - - - - | - - - - | - - - - | - - - - | - - - - | - - - - |

@@ -152,6 +156,9 @@
 #define CNST_THRESH_CTL_SEL_VAL(v)	(((v) & 0x7ffull) << 32)
 #define CNST_THRESH_CTL_SEL_MASK	CNST_THRESH_CTL_SEL_VAL(0x7ff)
 
+#define p10_CNST_THRESH_CMP_VAL(v) (((v) & 0x7ffull) << 43)
+#define p10_CNST_THRESH_CMP_MASK   p10_CNST_THRESH_CMP_VAL(0x7ff)
+
 #define CNST_EBB_VAL(v)		(((v) & EVENT_EBB_MASK) << 24)
 #define CNST_EBB_MASK		CNST_EBB_VAL(EVENT_EBB_MASK)
 

@@ -262,10 +269,10 @@
 #define PH(a, b)			(P(LVL, HIT) | P(a, b))
 #define PM(a, b)			(P(LVL, MISS) | P(a, b))
 
-int isa207_get_constraint(u64 event, unsigned long *maskp, unsigned long *valp);
+int isa207_get_constraint(u64 event, unsigned long *maskp, unsigned long *valp, u64 event_config1);
 int isa207_compute_mmcr(u64 event[], int n_ev,
 				unsigned int hwc[], struct mmcr_regs *mmcr,
-				struct perf_event *pevents[]);
+				struct perf_event *pevents[], u32 flags);
 void isa207_disable_pmc(unsigned int pmc, struct mmcr_regs *mmcr);
 int isa207_get_alternatives(u64 event, u64 alt[], int size, unsigned int flags,
 					const unsigned int ev_alt[][MAX_ALT]);

@@ -148,7 +148,7 @@ static u32 classbits[N_CLASSES - 1][2] = {
 };
 
 static int mpc7450_get_constraint(u64 event, unsigned long *maskp,
-				  unsigned long *valp)
+				  unsigned long *valp, u64 event_config1 __maybe_unused)
 {
 	int pmc, class;
 	u32 mask, value;

@@ -258,7 +258,8 @@ static const u32 pmcsel_mask[N_COUNTER] = {
  */
 static int mpc7450_compute_mmcr(u64 event[], int n_ev, unsigned int hwc[],
 				struct mmcr_regs *mmcr,
-				struct perf_event *pevents[])
+				struct perf_event *pevents[],
+				u32 flags __maybe_unused)
 {
 	u8 event_index[N_CLASSES][N_COUNTER];
 	int n_classevent[N_CLASSES];

@@ -216,6 +216,7 @@ PMU_FORMAT_ATTR(invert_bit,     "config:47");
 PMU_FORMAT_ATTR(src_mask,       "config:48-53");
 PMU_FORMAT_ATTR(src_match,      "config:54-59");
 PMU_FORMAT_ATTR(radix_scope,	"config:9");
+PMU_FORMAT_ATTR(thresh_cmp,     "config1:0-17");
 
 static struct attribute *power10_pmu_format_attr[] = {
 	&format_attr_event.attr,

@@ -236,6 +237,7 @@ static struct attribute *power10_pmu_format_attr[] = {
 	&format_attr_src_mask.attr,
 	&format_attr_src_match.attr,
 	&format_attr_radix_scope.attr,
+	&format_attr_thresh_cmp.attr,
 	NULL,
 };
 

@@ -550,7 +552,7 @@ static struct power_pmu power10_pmu = {
 	.get_mem_weight		= isa207_get_mem_weight,
 	.disable_pmc		= isa207_disable_pmc,
 	.flags			= PPMU_HAS_SIER | PPMU_ARCH_207S |
-				  PPMU_ARCH_31,
+				  PPMU_ARCH_31 | PPMU_HAS_ATTR_CONFIG1,
 	.n_generic		= ARRAY_SIZE(power10_generic_events),
 	.generic_events		= power10_generic_events,
 	.cache_events		= &power10_cache_events,

@@ -132,7 +132,7 @@ static unsigned long unit_cons[PM_LASTUNIT+1][2] = {
 };
 
 static int power5p_get_constraint(u64 event, unsigned long *maskp,
-				  unsigned long *valp)
+				  unsigned long *valp, u64 event_config1 __maybe_unused)
 {
 	int pmc, byte, unit, sh;
 	int bit, fmask;

@@ -451,7 +451,8 @@ static int power5p_marked_instr_event(u64 event)
 
 static int power5p_compute_mmcr(u64 event[], int n_ev,
 				unsigned int hwc[], struct mmcr_regs *mmcr,
-				struct perf_event *pevents[])
+				struct perf_event *pevents[],
+				u32 flags __maybe_unused)
 {
 	unsigned long mmcr1 = 0;
 	unsigned long mmcra = 0;

@@ -136,7 +136,7 @@ static unsigned long unit_cons[PM_LASTUNIT+1][2] = {
 };
 
 static int power5_get_constraint(u64 event, unsigned long *maskp,
-				 unsigned long *valp)
+				 unsigned long *valp, u64 event_config1 __maybe_unused)
 {
 	int pmc, byte, unit, sh;
 	int bit, fmask;

@@ -382,7 +382,8 @@ static int power5_marked_instr_event(u64 event)
 
 static int power5_compute_mmcr(u64 event[], int n_ev,
 			       unsigned int hwc[], struct mmcr_regs *mmcr,
-			       struct perf_event *pevents[])
+			       struct perf_event *pevents[],
+			       u32 flags __maybe_unused)
 {
 	unsigned long mmcr1 = 0;
 	unsigned long mmcra = MMCRA_SDAR_DCACHE_MISS | MMCRA_SDAR_ERAT_MISS;

@@ -173,7 +173,8 @@ static int power6_marked_instr_event(u64 event)
  * Assign PMC numbers and compute MMCR1 value for a set of events
  */
 static int p6_compute_mmcr(u64 event[], int n_ev,
-			   unsigned int hwc[], struct mmcr_regs *mmcr, struct perf_event *pevents[])
+			   unsigned int hwc[], struct mmcr_regs *mmcr, struct perf_event *pevents[],
+			   u32 flags __maybe_unused)
 {
 	unsigned long mmcr1 = 0;
 	unsigned long mmcra = MMCRA_SDAR_DCACHE_MISS | MMCRA_SDAR_ERAT_MISS;

@@ -266,7 +267,7 @@ static int p6_compute_mmcr(u64 event[], int n_ev,
  *	32-34	select field: nest (subunit) event selector
  */
 static int p6_get_constraint(u64 event, unsigned long *maskp,
-			     unsigned long *valp)
+			     unsigned long *valp, u64 event_config1 __maybe_unused)
 {
 	int pmc, byte, sh, subunit;
 	unsigned long mask = 0, value = 0;

@@ -81,7 +81,7 @@ enum {
  */
 
 static int power7_get_constraint(u64 event, unsigned long *maskp,
-				 unsigned long *valp)
+				 unsigned long *valp, u64 event_config1 __maybe_unused)
 {
 	int pmc, sh, unit;
 	unsigned long mask = 0, value = 0;

@@ -245,7 +245,8 @@ static int power7_marked_instr_event(u64 event)
 
 static int power7_compute_mmcr(u64 event[], int n_ev,
 			       unsigned int hwc[], struct mmcr_regs *mmcr,
-			       struct perf_event *pevents[])
+			       struct perf_event *pevents[],
+			       u32 flags __maybe_unused)
 {
 	unsigned long mmcr1 = 0;
 	unsigned long mmcra = MMCRA_SDAR_DCACHE_MISS | MMCRA_SDAR_ERAT_MISS;

@@ -190,7 +190,7 @@ static unsigned long unit_cons[PM_LASTUNIT+1][2] = {
 };
 
 static int p970_get_constraint(u64 event, unsigned long *maskp,
-			       unsigned long *valp)
+			       unsigned long *valp, u64 event_config1 __maybe_unused)
 {
 	int pmc, byte, unit, sh, spcsel;
 	unsigned long mask = 0, value = 0;

@@ -256,7 +256,8 @@ static int p970_get_alternatives(u64 event, unsigned int flags, u64 alt[])
 
 static int p970_compute_mmcr(u64 event[], int n_ev,
 			     unsigned int hwc[], struct mmcr_regs *mmcr,
-			     struct perf_event *pevents[])
+			     struct perf_event *pevents[],
+			     u32 flags __maybe_unused)
 {
 	unsigned long mmcr0 = 0, mmcr1 = 0, mmcra = 0;
 	unsigned int pmc, unit, byte, psel;