ia64: First draft of perfmon sampling-interval randomization support.

arch/ia64/kernel/perfmon.c

@@ -193,10 +193,12 @@ typedef enum {
  */
 typedef struct {
 	u64 val;	/* virtual 64bit counter value */
-	u64 ival;	/* initial value from user */
+	u64 lval;	/* last value */
 	u64 long_reset;	/* reset value on sampling overflow */
 	u64 short_reset;/* reset value on overflow */
 	u64 reset_pmds[4]; /* which other pmds to reset when this counter overflows */
+	u64 seed;	/* seed for random-number generator */
+	u64 mask;	/* mask for random-number generator */
 	int flags;	/* notify/do not notify */
 } pfm_counter_t;
 
@@ -1150,13 +1152,32 @@ pfm_context_create(struct task_struct *task, pfm_context_t *ctx, void *req, int
 	return ret;
 }
 
+static inline unsigned long
+new_counter_value (pfm_counter_t *reg, int is_long_reset)
+{
+	unsigned long val = is_long_reset ? reg->long_reset : reg->short_reset;
+	unsigned long new_seed, old_seed = reg->seed, mask = reg->mask;
+	extern unsigned long carta_random32 (unsigned long seed);
+
+	if (reg->flags & PFM_REGFL_RANDOM) {
+		new_seed = carta_random32(old_seed);
+		val -= (old_seed & mask);	/* counter values are negative numbers! */
+		if ((mask >> 32) != 0)
+			/* construct a full 64-bit random value: */
+			new_seed |= carta_random32(old_seed >> 32) << 32;
+		reg->seed = new_seed;
+	}
+	reg->lval = val;
+	return val;
+}
+
 static void
 pfm_reset_regs(pfm_context_t *ctx, unsigned long *ovfl_regs, int flag)
 {
 	unsigned long mask = ovfl_regs[0];
 	unsigned long reset_others = 0UL;
 	unsigned long val;
-	int i;
+	int i, is_long_reset = (flag & PFM_RELOAD_LONG_RESET);
 
 	DBprintk(("masks=0x%lx\n", mask));
 
@@ -1166,15 +1187,11 @@ pfm_reset_regs(pfm_context_t *ctx, unsigned long *ovfl_regs, int flag)
 	mask >>= PMU_FIRST_COUNTER;
 	for(i = PMU_FIRST_COUNTER; mask; i++, mask >>= 1) {
 		if (mask & 0x1) {
-			val  = flag == PFM_RELOAD_LONG_RESET ? 
-					ctx->ctx_soft_pmds[i].long_reset:
-					ctx->ctx_soft_pmds[i].short_reset;
-
+			val = new_counter_value(ctx->ctx_soft_pmds + i, is_long_reset);
 			reset_others |= ctx->ctx_soft_pmds[i].reset_pmds[0];
 
-			DBprintk(("[%d] %s reset soft_pmd[%d]=%lx\n", 
-			  	current->pid, 
-				flag == PFM_RELOAD_LONG_RESET ? "long" : "short", i, val));
+			DBprintk(("[%d] %s reset soft_pmd[%d]=%lx\n", current->pid,
+				  is_long_reset ? "long" : "short", i, val));
 
 			/* upper part is ignored on rval */
 			pfm_write_soft_counter(ctx, i, val);
@@ -1188,19 +1205,15 @@ pfm_reset_regs(pfm_context_t *ctx, unsigned long *ovfl_regs, int flag)
 
 		if ((reset_others & 0x1) == 0) continue;
 
-		val  = flag == PFM_RELOAD_LONG_RESET ? 
-					ctx->ctx_soft_pmds[i].long_reset:
-					ctx->ctx_soft_pmds[i].short_reset;
+		val = new_counter_value(ctx->ctx_soft_pmds + i, is_long_reset);
 
 		if (PMD_IS_COUNTING(i)) {
 			pfm_write_soft_counter(ctx, i, val);
 		} else {
 			ia64_set_pmd(i, val);
 		}
-
-		DBprintk(("[%d] %s reset_others pmd[%d]=%lx\n", 
-			  	current->pid, 
-				flag == PFM_RELOAD_LONG_RESET ? "long" : "short", i, val));
+		DBprintk(("[%d] %s reset_others pmd[%d]=%lx\n", current->pid,
+			  is_long_reset ? "long" : "short", i, val));
 	}
 	ia64_srlz_d();
 	/* just in case ! */
@@ -1283,6 +1296,9 @@ pfm_write_pmcs(struct task_struct *task, pfm_context_t *ctx, void *arg, int coun
 			ctx->ctx_soft_pmds[cnum].reset_pmds[1] = tmp.reg_reset_pmds[1];
 			ctx->ctx_soft_pmds[cnum].reset_pmds[2] = tmp.reg_reset_pmds[2];
 			ctx->ctx_soft_pmds[cnum].reset_pmds[3] = tmp.reg_reset_pmds[3];
+
+			if (tmp.reg_flags & PFM_REGFL_RANDOM)
+				ctx->ctx_soft_pmds[cnum].flags |= PFM_REGFL_RANDOM;
 		}
 		/*
 		 * execute write checker, if any
@@ -1369,11 +1385,12 @@ pfm_write_pmds(struct task_struct *task, pfm_context_t *ctx, void *arg, int coun
 
 		/* update virtualized (64bits) counter */
 		if (PMD_IS_COUNTING(cnum)) {
-			ctx->ctx_soft_pmds[cnum].ival = tmp.reg_value;
+			ctx->ctx_soft_pmds[cnum].lval = tmp.reg_value;
 			ctx->ctx_soft_pmds[cnum].val  = tmp.reg_value & ~pmu_conf.perf_ovfl_val;
 			ctx->ctx_soft_pmds[cnum].long_reset = tmp.reg_long_reset;
 			ctx->ctx_soft_pmds[cnum].short_reset = tmp.reg_short_reset;
-
+			ctx->ctx_soft_pmds[cnum].seed = tmp.reserved[0];
+			ctx->ctx_soft_pmds[cnum].mask = tmp.reserved[1];
 		}
 		/*
 		 * execute write checker, if any
@@ -2554,7 +2571,7 @@ pfm_record_sample(struct task_struct *task, pfm_context_t *ctx, unsigned long ov
 	 */
 	h->pid  = current->pid;
 	h->cpu  = smp_processor_id();
-	h->rate = 0; /* XXX: add the sampling rate used here */
+	h->last_reset_value = ovfl_mask ? ctx->ctx_soft_pmds[ffz(~ovfl_mask)].lval : 0;
 	h->ip   = regs ? regs->cr_iip : 0x0;	/* where did the fault happened */
 	h->regs = ovfl_mask; 			/* which registers overflowed */
 
@@ -3769,8 +3786,8 @@ pfm_inherit(struct task_struct *task, struct pt_regs *regs)
 	m = nctx->ctx_used_pmds[0] >> PMU_FIRST_COUNTER;
 	for(i = PMU_FIRST_COUNTER ; m ; m>>=1, i++) {
 		if ((m & 0x1) && pmu_conf.pmd_desc[i].type == PFM_REG_COUNTING) {
-			nctx->ctx_soft_pmds[i].val = nctx->ctx_soft_pmds[i].ival & ~pmu_conf.perf_ovfl_val;
-			thread->pmd[i]	      	   = nctx->ctx_soft_pmds[i].ival & pmu_conf.perf_ovfl_val;
+			nctx->ctx_soft_pmds[i].val = nctx->ctx_soft_pmds[i].lval & ~pmu_conf.perf_ovfl_val;
+			thread->pmd[i]	      	   = nctx->ctx_soft_pmds[i].lval & pmu_conf.perf_ovfl_val;
 		}
 		/* what about the other pmds? zero or keep as is */
 

arch/ia64/lib/Makefile

@@ -15,6 +15,7 @@ obj-y := __divsi3.o __udivsi3.o __modsi3.o __umodsi3.o			\
 
 obj-$(CONFIG_ITANIUM)	+= copy_page.o copy_user.o memcpy.o
 obj-$(CONFIG_MCKINLEY)	+= copy_page_mck.o memcpy_mck.o
+obj-$(CONFIG_PERFMON)	+= carta_random.o
 
 IGNORE_FLAGS_OBJS =	__divsi3.o __udivsi3.o __modsi3.o __umodsi3.o \
 			__divdi3.o __udivdi3.o __moddi3.o __umoddi3.o

arch/ia64/lib/carta_random.S

+/*
+ * Fast, simple, yet decent quality random number generator based on
+ * a paper by David G. Carta ("Two Fast Implementations of the
+ * `Minimal Standard' Random Number Generator," Communications of the
+ * ACM, January, 1990).
+ *
+ * Copyright (C) 2002 Hewlett-Packard Co
+ *	David Mosberger-Tang <davidm@hpl.hp.com>
+ */
+
+#include <asm/asmmacro.h>
+
+#define a	r2
+#define m	r3
+#define lo	r8
+#define hi	r9
+#define t0	r16
+#define t1	r17
+#define	seed	r32
+
+GLOBAL_ENTRY(carta_random32)
+	movl	a = (16807 << 16) | 16807
+	;;
+	pmpyshr2.u t0 = a, seed, 0
+	pmpyshr2.u t1 = a, seed, 16
+	;;
+	unpack2.l t0 = t1, t0
+	dep	m = -1, r0, 0, 31
+	;;
+	zxt4	lo = t0
+	shr.u	hi = t0, 32
+	;;
+	dep	t0 = 0, hi, 15, 49	// t0 = (hi & 0x7fff)
+	;;
+	shl	t0 = t0, 16		// t0 = (hi & 0x7fff) << 16
+	shr	t1 = hi, 15		// t1 = (hi >> 15)
+	;;
+	add	lo = lo, t0
+	;;
+	cmp.gtu	p6, p0 = lo, m
+	;;
+(p6)	and	lo = lo, m
+	;;
+(p6)	add	lo = 1, lo
+	;;
+	add	lo = lo, t1
+	;;
+	cmp.gtu p6, p0 = lo, m
+	;;
+(p6)	and	lo = lo, m
+	;;
+(p6)	add	lo = 1, lo
+	br.ret.sptk.many rp
+END(carta_random32)

include/asm-ia64/perfmon.h

@@ -45,6 +45,7 @@
  * PMC flags
  */
 #define PFM_REGFL_OVFL_NOTIFY	0x1	/* send notification on overflow */
+#define PFM_REGFL_RANDOM	0x2	/* randomize sampling interval */
 
 /*
  * PMD/PMC/IBR/DBR return flags (ignored on input)
@@ -132,28 +133,28 @@ typedef struct {
 #define PFM_VERSION_MINOR(x)	((x) & 0xffff)
 
 /*
- * Entry header in the sampling buffer.
- * The header is directly followed with the PMDS saved in increasing index 
- * order: PMD4, PMD5, .... How many PMDs are present is determined by the 
- * user program during context creation.
+ * Entry header in the sampling buffer.  The header is directly followed
+ * with the PMDs saved in increasing index order: PMD4, PMD5, .... How
+ * many PMDs are present is determined by the user program during
+ * context creation.
  *
- * XXX: in this version of the entry, only up to 64 registers can be recorded
- * This should be enough for quite some time. Always check sampling format
- * before parsing entries!
+ * XXX: in this version of the entry, only up to 64 registers can be
+ * recorded. This should be enough for quite some time. Always check
+ * sampling format before parsing entries!
  *
- * Inn the case where multiple counters have overflowed at the same time, the 
- * rate field indicate the initial value of the first PMD, based on the index.
- * For instance, if PMD2 and PMD5 have ovewrflowed for this entry, the rate field
- * will show the initial value of PMD2.
+ * In the case where multiple counters overflow at the same time, the
+ * last_reset_value member indicates the initial value of the PMD with
+ * the smallest index.  For instance, if PMD2 and PMD5 have overflowed,
+ * the last_reset_value member contains the initial value of PMD2.
  */
 typedef struct {
-	int		pid;		/* identification of process */
-	int		cpu;		/* which cpu was used */
-	unsigned long	rate;		/* initial value of overflowed counter */
-	unsigned long	stamp;		/* timestamp */
-	unsigned long	ip;		/* where did the overflow interrupt happened */
-	unsigned long	regs;		/* bitmask of which registers overflowed */
-	unsigned long   period;		/* sampling period used by overflowed counter (smallest pmd index) */
+	int		pid;			/* identification of process */
+	int		cpu;			/* which cpu was used */
+	unsigned long	last_reset_value;	/* initial value of counter that overflowed */
+	unsigned long	stamp;			/* timestamp */
+	unsigned long	ip;			/* where did the overflow interrupt happened */
+	unsigned long	regs;			/* bitmask of which registers overflowed */
+	unsigned long   period;			/* unused */
 } perfmon_smpl_entry_t;
 
 extern int perfmonctl(pid_t pid, int cmd, void *arg, int narg);