aboutsummaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
Diffstat
-rw-r--r--arch/powerpc/Kconfig.debug2
-rw-r--r--arch/powerpc/configs/pseries_defconfig2
-rw-r--r--arch/powerpc/include/asm/emulated_ops.h19
-rw-r--r--arch/powerpc/include/asm/hvcall.h2
-rw-r--r--arch/powerpc/include/asm/reg.h2
-rw-r--r--arch/powerpc/include/asm/trace.h133
-rw-r--r--arch/powerpc/kernel/align.c12
-rw-r--r--arch/powerpc/kernel/entry_64.S4
-rw-r--r--arch/powerpc/kernel/exceptions-64s.S3
-rw-r--r--arch/powerpc/kernel/irq.c6
-rw-r--r--arch/powerpc/kernel/perf_event.c2
-rw-r--r--arch/powerpc/kernel/power5+-pmu.c4
-rw-r--r--arch/powerpc/kernel/power5-pmu.c6
-rw-r--r--arch/powerpc/kernel/power6-pmu.c2
-rw-r--r--arch/powerpc/kernel/power7-pmu.c6
-rw-r--r--arch/powerpc/kernel/ppc970-pmu.c4
-rw-r--r--arch/powerpc/kernel/setup-common.c1
-rw-r--r--arch/powerpc/kernel/time.c6
-rw-r--r--arch/powerpc/kernel/traps.c18
-rw-r--r--arch/powerpc/lib/copypage_64.S4
-rw-r--r--arch/powerpc/platforms/pseries/hvCall.S132
-rw-r--r--arch/powerpc/platforms/pseries/hvCall_inst.c38
-rw-r--r--arch/powerpc/platforms/pseries/lpar.c33
-rw-r--r--include/linux/perf_counter.h2
-rw-r--r--include/linux/perf_event.h2
-rw-r--r--kernel/perf_event.c2
-rw-r--r--tools/perf/Documentation/perf-bench.txt120
-rw-r--r--tools/perf/Makefile7
-rw-r--r--tools/perf/bench/bench.h16
-rw-r--r--tools/perf/bench/sched-messaging.c336
-rw-r--r--tools/perf/bench/sched-pipe.c124
-rw-r--r--tools/perf/builtin-bench.c183
-rw-r--r--tools/perf/builtin.h1
-rw-r--r--tools/perf/command-list.txt1
-rw-r--r--tools/perf/design.txt2
-rw-r--r--tools/perf/perf.c1
-rw-r--r--tools/perf/util/parse-events.c4
37 files changed, 1146 insertions, 96 deletions
diff --git a/arch/powerpc/Kconfig.debug b/arch/powerpc/Kconfig.debug
index 3b1005185390..bf3382f1904d 100644
--- a/arch/powerpc/Kconfig.debug
+++ b/arch/powerpc/Kconfig.debug
@@ -46,7 +46,7 @@ config DEBUG_STACK_USAGE
46 46
47config HCALL_STATS 47config HCALL_STATS
48 bool "Hypervisor call instrumentation" 48 bool "Hypervisor call instrumentation"
49 depends on PPC_PSERIES && DEBUG_FS 49 depends on PPC_PSERIES && DEBUG_FS && TRACEPOINTS
50 help 50 help
51 Adds code to keep track of the number of hypervisor calls made and 51 Adds code to keep track of the number of hypervisor calls made and
52 the amount of time spent in hypervisor calls. Wall time spent in 52 the amount of time spent in hypervisor calls. Wall time spent in
diff --git a/arch/powerpc/configs/pseries_defconfig b/arch/powerpc/configs/pseries_defconfig
index f1889abb89b1..c568329723b8 100644
--- a/arch/powerpc/configs/pseries_defconfig
+++ b/arch/powerpc/configs/pseries_defconfig
@@ -1683,7 +1683,7 @@ CONFIG_HAVE_ARCH_KGDB=y
1683CONFIG_DEBUG_STACKOVERFLOW=y 1683CONFIG_DEBUG_STACKOVERFLOW=y
1684# CONFIG_DEBUG_STACK_USAGE is not set 1684# CONFIG_DEBUG_STACK_USAGE is not set
1685# CONFIG_DEBUG_PAGEALLOC is not set 1685# CONFIG_DEBUG_PAGEALLOC is not set
1686CONFIG_HCALL_STATS=y 1686# CONFIG_HCALL_STATS is not set
1687# CONFIG_CODE_PATCHING_SELFTEST is not set 1687# CONFIG_CODE_PATCHING_SELFTEST is not set
1688# CONFIG_FTR_FIXUP_SELFTEST is not set 1688# CONFIG_FTR_FIXUP_SELFTEST is not set
1689# CONFIG_MSI_BITMAP_SELFTEST is not set 1689# CONFIG_MSI_BITMAP_SELFTEST is not set
diff --git a/arch/powerpc/include/asm/emulated_ops.h b/arch/powerpc/include/asm/emulated_ops.h
index 9154e8526732..f0fb4fc1f6e6 100644
--- a/arch/powerpc/include/asm/emulated_ops.h
+++ b/arch/powerpc/include/asm/emulated_ops.h
@@ -19,6 +19,7 @@
19#define _ASM_POWERPC_EMULATED_OPS_H 19#define _ASM_POWERPC_EMULATED_OPS_H
20 20
21#include <asm/atomic.h> 21#include <asm/atomic.h>
22#include <linux/perf_event.h>
22 23
23 24
24#ifdef CONFIG_PPC_EMULATED_STATS 25#ifdef CONFIG_PPC_EMULATED_STATS
@@ -57,7 +58,7 @@ extern u32 ppc_warn_emulated;
57 58
58extern void ppc_warn_emulated_print(const char *type); 59extern void ppc_warn_emulated_print(const char *type);
59 60
60#define PPC_WARN_EMULATED(type) \ 61#define __PPC_WARN_EMULATED(type) \
61 do { \ 62 do { \
62 atomic_inc(&ppc_emulated.type.val); \ 63 atomic_inc(&ppc_emulated.type.val); \
63 if (ppc_warn_emulated) \ 64 if (ppc_warn_emulated) \
@@ -66,8 +67,22 @@ extern void ppc_warn_emulated_print(const char *type);
66 67
67#else /* !CONFIG_PPC_EMULATED_STATS */ 68#else /* !CONFIG_PPC_EMULATED_STATS */
68 69
69#define PPC_WARN_EMULATED(type) do { } while (0) 70#define __PPC_WARN_EMULATED(type) do { } while (0)
70 71
71#endif /* !CONFIG_PPC_EMULATED_STATS */ 72#endif /* !CONFIG_PPC_EMULATED_STATS */
72 73
74#define PPC_WARN_EMULATED(type, regs) \
75 do { \
76 perf_sw_event(PERF_COUNT_SW_EMULATION_FAULTS, \
77 1, 0, regs, 0); \
78 __PPC_WARN_EMULATED(type); \
79 } while (0)
80
81#define PPC_WARN_ALIGNMENT(type, regs) \
82 do { \
83 perf_sw_event(PERF_COUNT_SW_ALIGNMENT_FAULTS, \
84 1, 0, regs, regs->dar); \
85 __PPC_WARN_EMULATED(type); \
86 } while (0)
87
73#endif /* _ASM_POWERPC_EMULATED_OPS_H */ 88#endif /* _ASM_POWERPC_EMULATED_OPS_H */
diff --git a/arch/powerpc/include/asm/hvcall.h b/arch/powerpc/include/asm/hvcall.h
index 6251a4b10be7..c27caac47ad1 100644
--- a/arch/powerpc/include/asm/hvcall.h
+++ b/arch/powerpc/include/asm/hvcall.h
@@ -274,6 +274,8 @@ struct hcall_stats {
274 unsigned long num_calls; /* number of calls (on this CPU) */ 274 unsigned long num_calls; /* number of calls (on this CPU) */
275 unsigned long tb_total; /* total wall time (mftb) of calls. */ 275 unsigned long tb_total; /* total wall time (mftb) of calls. */
276 unsigned long purr_total; /* total cpu time (PURR) of calls. */ 276 unsigned long purr_total; /* total cpu time (PURR) of calls. */
277 unsigned long tb_start;
278 unsigned long purr_start;
277}; 279};
278#define HCALL_STAT_ARRAY_SIZE ((MAX_HCALL_OPCODE >> 2) + 1) 280#define HCALL_STAT_ARRAY_SIZE ((MAX_HCALL_OPCODE >> 2) + 1)
279 281
diff --git a/arch/powerpc/include/asm/reg.h b/arch/powerpc/include/asm/reg.h
index 6315edc205d8..bc8dd53f718a 100644
--- a/arch/powerpc/include/asm/reg.h
+++ b/arch/powerpc/include/asm/reg.h
@@ -489,6 +489,8 @@
489#define SPRN_MMCR1 798 489#define SPRN_MMCR1 798
490#define SPRN_MMCRA 0x312 490#define SPRN_MMCRA 0x312
491#define MMCRA_SDSYNC 0x80000000UL /* SDAR synced with SIAR */ 491#define MMCRA_SDSYNC 0x80000000UL /* SDAR synced with SIAR */
492#define MMCRA_SDAR_DCACHE_MISS 0x40000000UL
493#define MMCRA_SDAR_ERAT_MISS 0x20000000UL
492#define MMCRA_SIHV 0x10000000UL /* state of MSR HV when SIAR set */ 494#define MMCRA_SIHV 0x10000000UL /* state of MSR HV when SIAR set */
493#define MMCRA_SIPR 0x08000000UL /* state of MSR PR when SIAR set */ 495#define MMCRA_SIPR 0x08000000UL /* state of MSR PR when SIAR set */
494#define MMCRA_SLOT 0x07000000UL /* SLOT bits (37-39) */ 496#define MMCRA_SLOT 0x07000000UL /* SLOT bits (37-39) */
diff --git a/arch/powerpc/include/asm/trace.h b/arch/powerpc/include/asm/trace.h
new file mode 100644
index 000000000000..cbe2297d68b6
--- /dev/null
+++ b/arch/powerpc/include/asm/trace.h
@@ -0,0 +1,133 @@
1#undef TRACE_SYSTEM
2#define TRACE_SYSTEM powerpc
3
4#if !defined(_TRACE_POWERPC_H) || defined(TRACE_HEADER_MULTI_READ)
5#define _TRACE_POWERPC_H
6
7#include <linux/tracepoint.h>
8
9struct pt_regs;
10
11TRACE_EVENT(irq_entry,
12
13 TP_PROTO(struct pt_regs *regs),
14
15 TP_ARGS(regs),
16
17 TP_STRUCT__entry(
18 __field(struct pt_regs *, regs)
19 ),
20
21 TP_fast_assign(
22 __entry->regs = regs;
23 ),
24
25 TP_printk("pt_regs=%p", __entry->regs)
26);
27
28TRACE_EVENT(irq_exit,
29
30 TP_PROTO(struct pt_regs *regs),
31
32 TP_ARGS(regs),
33
34 TP_STRUCT__entry(
35 __field(struct pt_regs *, regs)
36 ),
37
38 TP_fast_assign(
39 __entry->regs = regs;
40 ),
41
42 TP_printk("pt_regs=%p", __entry->regs)
43);
44
45TRACE_EVENT(timer_interrupt_entry,
46
47 TP_PROTO(struct pt_regs *regs),
48
49 TP_ARGS(regs),
50
51 TP_STRUCT__entry(
52 __field(struct pt_regs *, regs)
53 ),
54
55 TP_fast_assign(
56 __entry->regs = regs;
57 ),
58
59 TP_printk("pt_regs=%p", __entry->regs)
60);
61
62TRACE_EVENT(timer_interrupt_exit,
63
64 TP_PROTO(struct pt_regs *regs),
65
66 TP_ARGS(regs),
67
68 TP_STRUCT__entry(
69 __field(struct pt_regs *, regs)
70 ),
71
72 TP_fast_assign(
73 __entry->regs = regs;
74 ),
75
76 TP_printk("pt_regs=%p", __entry->regs)
77);
78
79#ifdef CONFIG_PPC_PSERIES
80extern void hcall_tracepoint_regfunc(void);
81extern void hcall_tracepoint_unregfunc(void);
82
83TRACE_EVENT_FN(hcall_entry,
84
85 TP_PROTO(unsigned long opcode, unsigned long *args),
86
87 TP_ARGS(opcode, args),
88
89 TP_STRUCT__entry(
90 __field(unsigned long, opcode)
91 ),
92
93 TP_fast_assign(
94 __entry->opcode = opcode;
95 ),
96
97 TP_printk("opcode=%lu", __entry->opcode),
98
99 hcall_tracepoint_regfunc, hcall_tracepoint_unregfunc
100);
101
102TRACE_EVENT_FN(hcall_exit,
103
104 TP_PROTO(unsigned long opcode, unsigned long retval,
105 unsigned long *retbuf),
106
107 TP_ARGS(opcode, retval, retbuf),
108
109 TP_STRUCT__entry(
110 __field(unsigned long, opcode)
111 __field(unsigned long, retval)
112 ),
113
114 TP_fast_assign(
115 __entry->opcode = opcode;
116 __entry->retval = retval;
117 ),
118
119 TP_printk("opcode=%lu retval=%lu", __entry->opcode, __entry->retval),
120
121 hcall_tracepoint_regfunc, hcall_tracepoint_unregfunc
122);
123#endif
124
125#endif /* _TRACE_POWERPC_H */
126
127#undef TRACE_INCLUDE_PATH
128#undef TRACE_INCLUDE_FILE
129
130#define TRACE_INCLUDE_PATH asm
131#define TRACE_INCLUDE_FILE trace
132
133#include <trace/define_trace.h>
diff --git a/arch/powerpc/kernel/align.c b/arch/powerpc/kernel/align.c
index a5b632e52fae..3839839f83c7 100644
--- a/arch/powerpc/kernel/align.c
+++ b/arch/powerpc/kernel/align.c
@@ -732,7 +732,7 @@ int fix_alignment(struct pt_regs *regs)
732 732
733#ifdef CONFIG_SPE 733#ifdef CONFIG_SPE
734 if ((instr >> 26) == 0x4) { 734 if ((instr >> 26) == 0x4) {
735 PPC_WARN_EMULATED(spe); 735 PPC_WARN_ALIGNMENT(spe, regs);
736 return emulate_spe(regs, reg, instr); 736 return emulate_spe(regs, reg, instr);
737 } 737 }
738#endif 738#endif
@@ -786,7 +786,7 @@ int fix_alignment(struct pt_regs *regs)
786 flags |= SPLT; 786 flags |= SPLT;
787 nb = 8; 787 nb = 8;
788 } 788 }
789 PPC_WARN_EMULATED(vsx); 789 PPC_WARN_ALIGNMENT(vsx, regs);
790 return emulate_vsx(addr, reg, areg, regs, flags, nb); 790 return emulate_vsx(addr, reg, areg, regs, flags, nb);
791 } 791 }
792#endif 792#endif
@@ -794,7 +794,7 @@ int fix_alignment(struct pt_regs *regs)
794 * the exception of DCBZ which is handled as a special case here 794 * the exception of DCBZ which is handled as a special case here
795 */ 795 */
796 if (instr == DCBZ) { 796 if (instr == DCBZ) {
797 PPC_WARN_EMULATED(dcbz); 797 PPC_WARN_ALIGNMENT(dcbz, regs);
798 return emulate_dcbz(regs, addr); 798 return emulate_dcbz(regs, addr);
799 } 799 }
800 if (unlikely(nb == 0)) 800 if (unlikely(nb == 0))
@@ -804,7 +804,7 @@ int fix_alignment(struct pt_regs *regs)
804 * function 804 * function
805 */ 805 */
806 if (flags & M) { 806 if (flags & M) {
807 PPC_WARN_EMULATED(multiple); 807 PPC_WARN_ALIGNMENT(multiple, regs);
808 return emulate_multiple(regs, addr, reg, nb, 808 return emulate_multiple(regs, addr, reg, nb,
809 flags, instr, swiz); 809 flags, instr, swiz);
810 } 810 }
@@ -825,11 +825,11 @@ int fix_alignment(struct pt_regs *regs)
825 825
826 /* Special case for 16-byte FP loads and stores */ 826 /* Special case for 16-byte FP loads and stores */
827 if (nb == 16) { 827 if (nb == 16) {
828 PPC_WARN_EMULATED(fp_pair); 828 PPC_WARN_ALIGNMENT(fp_pair, regs);
829 return emulate_fp_pair(addr, reg, flags); 829 return emulate_fp_pair(addr, reg, flags);
830 } 830 }
831 831
832 PPC_WARN_EMULATED(unaligned); 832 PPC_WARN_ALIGNMENT(unaligned, regs);
833 833
834 /* If we are loading, get the data from user space, else 834 /* If we are loading, get the data from user space, else
835 * get it from register values 835 * get it from register values
diff --git a/arch/powerpc/kernel/entry_64.S b/arch/powerpc/kernel/entry_64.S
index 9763267e38b4..bdcb557d470a 100644
--- a/arch/powerpc/kernel/entry_64.S
+++ b/arch/powerpc/kernel/entry_64.S
@@ -551,7 +551,7 @@ restore:
551BEGIN_FW_FTR_SECTION 551BEGIN_FW_FTR_SECTION
552 ld r5,SOFTE(r1) 552 ld r5,SOFTE(r1)
553FW_FTR_SECTION_ELSE 553FW_FTR_SECTION_ELSE
554 b iseries_check_pending_irqs 554 b .Liseries_check_pending_irqs
555ALT_FW_FTR_SECTION_END_IFCLR(FW_FEATURE_ISERIES) 555ALT_FW_FTR_SECTION_END_IFCLR(FW_FEATURE_ISERIES)
5562: 5562:
557 TRACE_AND_RESTORE_IRQ(r5); 557 TRACE_AND_RESTORE_IRQ(r5);
@@ -623,7 +623,7 @@ ALT_FW_FTR_SECTION_END_IFCLR(FW_FEATURE_ISERIES)
623 623
624#endif /* CONFIG_PPC_BOOK3E */ 624#endif /* CONFIG_PPC_BOOK3E */
625 625
626iseries_check_pending_irqs: 626.Liseries_check_pending_irqs:
627#ifdef CONFIG_PPC_ISERIES 627#ifdef CONFIG_PPC_ISERIES
628 ld r5,SOFTE(r1) 628 ld r5,SOFTE(r1)
629 cmpdi 0,r5,0 629 cmpdi 0,r5,0
diff --git a/arch/powerpc/kernel/exceptions-64s.S b/arch/powerpc/kernel/exceptions-64s.S
index 1808876edcc9..c7eb4e0eb86c 100644
--- a/arch/powerpc/kernel/exceptions-64s.S
+++ b/arch/powerpc/kernel/exceptions-64s.S
@@ -185,12 +185,15 @@ END_FTR_SECTION_IFSET(CPU_FTR_REAL_LE)
185 * prolog code of the PerformanceMonitor one. A little 185 * prolog code of the PerformanceMonitor one. A little
186 * trickery is thus necessary 186 * trickery is thus necessary
187 */ 187 */
188performance_monitor_pSeries_1:
188 . = 0xf00 189 . = 0xf00
189 b performance_monitor_pSeries 190 b performance_monitor_pSeries
190 191
192altivec_unavailable_pSeries_1:
191 . = 0xf20 193 . = 0xf20
192 b altivec_unavailable_pSeries 194 b altivec_unavailable_pSeries
193 195
196vsx_unavailable_pSeries_1:
194 . = 0xf40 197 . = 0xf40
195 b vsx_unavailable_pSeries 198 b vsx_unavailable_pSeries
196 199
diff --git a/arch/powerpc/kernel/irq.c b/arch/powerpc/kernel/irq.c
index e5d121177984..02a334662cc0 100644
--- a/arch/powerpc/kernel/irq.c
+++ b/arch/powerpc/kernel/irq.c
@@ -70,6 +70,8 @@
70#include <asm/firmware.h> 70#include <asm/firmware.h>
71#include <asm/lv1call.h> 71#include <asm/lv1call.h>
72#endif 72#endif
73#define CREATE_TRACE_POINTS
74#include <asm/trace.h>
73 75
74int __irq_offset_value; 76int __irq_offset_value;
75static int ppc_spurious_interrupts; 77static int ppc_spurious_interrupts;
@@ -325,6 +327,8 @@ void do_IRQ(struct pt_regs *regs)
325 struct pt_regs *old_regs = set_irq_regs(regs); 327 struct pt_regs *old_regs = set_irq_regs(regs);
326 unsigned int irq; 328 unsigned int irq;
327 329
330 trace_irq_entry(regs);
331
328 irq_enter(); 332 irq_enter();
329 333
330 check_stack_overflow(); 334 check_stack_overflow();
@@ -348,6 +352,8 @@ void do_IRQ(struct pt_regs *regs)
348 timer_interrupt(regs); 352 timer_interrupt(regs);
349 } 353 }
350#endif 354#endif
355
356 trace_irq_exit(regs);
351} 357}
352 358
353void __init init_IRQ(void) 359void __init init_IRQ(void)
diff --git a/arch/powerpc/kernel/perf_event.c b/arch/powerpc/kernel/perf_event.c
index 87f1663584b0..1eb85fbf53a5 100644
--- a/arch/powerpc/kernel/perf_event.c
+++ b/arch/powerpc/kernel/perf_event.c
@@ -1165,7 +1165,7 @@ static void record_and_restart(struct perf_event *event, unsigned long val,
1165 */ 1165 */
1166 if (record) { 1166 if (record) {
1167 struct perf_sample_data data = { 1167 struct perf_sample_data data = {
1168 .addr = 0, 1168 .addr = ~0ULL,
1169 .period = event->hw.last_period, 1169 .period = event->hw.last_period,
1170 }; 1170 };
1171 1171
diff --git a/arch/powerpc/kernel/power5+-pmu.c b/arch/powerpc/kernel/power5+-pmu.c
index 0f4c1c73a6ad..199de527d411 100644
--- a/arch/powerpc/kernel/power5+-pmu.c
+++ b/arch/powerpc/kernel/power5+-pmu.c
@@ -73,10 +73,6 @@
73#define MMCR1_PMCSEL_MSK 0x7f 73#define MMCR1_PMCSEL_MSK 0x7f
74 74
75/* 75/*
76 * Bits in MMCRA
77 */
78
79/*
80 * Layout of constraint bits: 76 * Layout of constraint bits:
81 * 6666555555555544444444443333333333222222222211111111110000000000 77 * 6666555555555544444444443333333333222222222211111111110000000000
82 * 3210987654321098765432109876543210987654321098765432109876543210 78 * 3210987654321098765432109876543210987654321098765432109876543210
diff --git a/arch/powerpc/kernel/power5-pmu.c b/arch/powerpc/kernel/power5-pmu.c
index c351b3a57fbb..98b6a729a9dd 100644
--- a/arch/powerpc/kernel/power5-pmu.c
+++ b/arch/powerpc/kernel/power5-pmu.c
@@ -73,10 +73,6 @@
73#define MMCR1_PMCSEL_MSK 0x7f 73#define MMCR1_PMCSEL_MSK 0x7f
74 74
75/* 75/*
76 * Bits in MMCRA
77 */
78
79/*
80 * Layout of constraint bits: 76 * Layout of constraint bits:
81 * 6666555555555544444444443333333333222222222211111111110000000000 77 * 6666555555555544444444443333333333222222222211111111110000000000
82 * 3210987654321098765432109876543210987654321098765432109876543210 78 * 3210987654321098765432109876543210987654321098765432109876543210
@@ -390,7 +386,7 @@ static int power5_compute_mmcr(u64 event[], int n_ev,
390 unsigned int hwc[], unsigned long mmcr[]) 386 unsigned int hwc[], unsigned long mmcr[])
391{ 387{
392 unsigned long mmcr1 = 0; 388 unsigned long mmcr1 = 0;
393 unsigned long mmcra = 0; 389 unsigned long mmcra = MMCRA_SDAR_DCACHE_MISS | MMCRA_SDAR_ERAT_MISS;
394 unsigned int pmc, unit, byte, psel; 390 unsigned int pmc, unit, byte, psel;
395 unsigned int ttm, grp; 391 unsigned int ttm, grp;
396 int i, isbus, bit, grsel; 392 int i, isbus, bit, grsel;
diff --git a/arch/powerpc/kernel/power6-pmu.c b/arch/powerpc/kernel/power6-pmu.c
index ca399ba5034c..84a607bda8fb 100644
--- a/arch/powerpc/kernel/power6-pmu.c
+++ b/arch/powerpc/kernel/power6-pmu.c
@@ -178,7 +178,7 @@ static int p6_compute_mmcr(u64 event[], int n_ev,
178 unsigned int hwc[], unsigned long mmcr[]) 178 unsigned int hwc[], unsigned long mmcr[])
179{ 179{
180 unsigned long mmcr1 = 0; 180 unsigned long mmcr1 = 0;
181 unsigned long mmcra = 0; 181 unsigned long mmcra = MMCRA_SDAR_DCACHE_MISS | MMCRA_SDAR_ERAT_MISS;
182 int i; 182 int i;
183 unsigned int pmc, ev, b, u, s, psel; 183 unsigned int pmc, ev, b, u, s, psel;
184 unsigned int ttmset = 0; 184 unsigned int ttmset = 0;
diff --git a/arch/powerpc/kernel/power7-pmu.c b/arch/powerpc/kernel/power7-pmu.c
index 28a4daacdc02..852f7b7f6b40 100644
--- a/arch/powerpc/kernel/power7-pmu.c
+++ b/arch/powerpc/kernel/power7-pmu.c
@@ -51,10 +51,6 @@
51#define MMCR1_PMCSEL_MSK 0xff 51#define MMCR1_PMCSEL_MSK 0xff
52 52
53/* 53/*
54 * Bits in MMCRA
55 */
56
57/*
58 * Layout of constraint bits: 54 * Layout of constraint bits:
59 * 6666555555555544444444443333333333222222222211111111110000000000 55 * 6666555555555544444444443333333333222222222211111111110000000000
60 * 3210987654321098765432109876543210987654321098765432109876543210 56 * 3210987654321098765432109876543210987654321098765432109876543210
@@ -230,7 +226,7 @@ static int power7_compute_mmcr(u64 event[], int n_ev,
230 unsigned int hwc[], unsigned long mmcr[]) 226 unsigned int hwc[], unsigned long mmcr[])
231{ 227{
232 unsigned long mmcr1 = 0; 228 unsigned long mmcr1 = 0;
233 unsigned long mmcra = 0; 229 unsigned long mmcra = MMCRA_SDAR_DCACHE_MISS | MMCRA_SDAR_ERAT_MISS;
234 unsigned int pmc, unit, combine, l2sel, psel; 230 unsigned int pmc, unit, combine, l2sel, psel;
235 unsigned int pmc_inuse = 0; 231 unsigned int pmc_inuse = 0;
236 int i; 232 int i;
diff --git a/arch/powerpc/kernel/ppc970-pmu.c b/arch/powerpc/kernel/ppc970-pmu.c
index 479574413a93..8eff48e20dba 100644
--- a/arch/powerpc/kernel/ppc970-pmu.c
+++ b/arch/powerpc/kernel/ppc970-pmu.c
@@ -84,10 +84,6 @@ static short mmcr1_adder_bits[8] = {
84}; 84};
85 85
86/* 86/*
87 * Bits in MMCRA
88 */
89
90/*
91 * Layout of constraint bits: 87 * Layout of constraint bits:
92 * 6666555555555544444444443333333333222222222211111111110000000000 88 * 6666555555555544444444443333333333222222222211111111110000000000
93 * 3210987654321098765432109876543210987654321098765432109876543210 89 * 3210987654321098765432109876543210987654321098765432109876543210
diff --git a/arch/powerpc/kernel/setup-common.c b/arch/powerpc/kernel/setup-common.c
index 4271f7a655a3..845c72ab7357 100644
--- a/arch/powerpc/kernel/setup-common.c
+++ b/arch/powerpc/kernel/setup-common.c
@@ -660,6 +660,7 @@ late_initcall(check_cache_coherency);
660 660
661#ifdef CONFIG_DEBUG_FS 661#ifdef CONFIG_DEBUG_FS
662struct dentry *powerpc_debugfs_root; 662struct dentry *powerpc_debugfs_root;
663EXPORT_SYMBOL(powerpc_debugfs_root);
663 664
664static int powerpc_debugfs_init(void) 665static int powerpc_debugfs_init(void)
665{ 666{
diff --git a/arch/powerpc/kernel/time.c b/arch/powerpc/kernel/time.c
index a136a11c490d..36707dec94d7 100644
--- a/arch/powerpc/kernel/time.c
+++ b/arch/powerpc/kernel/time.c
@@ -54,6 +54,7 @@
54#include <linux/irq.h> 54#include <linux/irq.h>
55#include <linux/delay.h> 55#include <linux/delay.h>
56#include <linux/perf_event.h> 56#include <linux/perf_event.h>
57#include <asm/trace.h>
57 58
58#include <asm/io.h> 59#include <asm/io.h>
59#include <asm/processor.h> 60#include <asm/processor.h>
@@ -571,6 +572,8 @@ void timer_interrupt(struct pt_regs * regs)
571 struct clock_event_device *evt = &decrementer->event; 572 struct clock_event_device *evt = &decrementer->event;
572 u64 now; 573 u64 now;
573 574
575 trace_timer_interrupt_entry(regs);
576
574 /* Ensure a positive value is written to the decrementer, or else 577 /* Ensure a positive value is written to the decrementer, or else
575 * some CPUs will continuue to take decrementer exceptions */ 578 * some CPUs will continuue to take decrementer exceptions */
576 set_dec(DECREMENTER_MAX); 579 set_dec(DECREMENTER_MAX);
@@ -590,6 +593,7 @@ void timer_interrupt(struct pt_regs * regs)
590 now = decrementer->next_tb - now; 593 now = decrementer->next_tb - now;
591 if (now <= DECREMENTER_MAX) 594 if (now <= DECREMENTER_MAX)
592 set_dec((int)now); 595 set_dec((int)now);
596 trace_timer_interrupt_exit(regs);
593 return; 597 return;
594 } 598 }
595 old_regs = set_irq_regs(regs); 599 old_regs = set_irq_regs(regs);
@@ -620,6 +624,8 @@ void timer_interrupt(struct pt_regs * regs)
620 624
621 irq_exit(); 625 irq_exit();
622 set_irq_regs(old_regs); 626 set_irq_regs(old_regs);
627
628 trace_timer_interrupt_exit(regs);
623} 629}
624 630
625void wakeup_decrementer(void) 631void wakeup_decrementer(void)
diff --git a/arch/powerpc/kernel/traps.c b/arch/powerpc/kernel/traps.c
index 6f0ae1a9bfae..9d1f9354d6ca 100644
--- a/arch/powerpc/kernel/traps.c
+++ b/arch/powerpc/kernel/traps.c
@@ -759,7 +759,7 @@ static int emulate_instruction(struct pt_regs *regs)
759 759
760 /* Emulate the mfspr rD, PVR. */ 760 /* Emulate the mfspr rD, PVR. */
761 if ((instword & PPC_INST_MFSPR_PVR_MASK) == PPC_INST_MFSPR_PVR) { 761 if ((instword & PPC_INST_MFSPR_PVR_MASK) == PPC_INST_MFSPR_PVR) {
762 PPC_WARN_EMULATED(mfpvr); 762 PPC_WARN_EMULATED(mfpvr, regs);
763 rd = (instword >> 21) & 0x1f; 763 rd = (instword >> 21) & 0x1f;
764 regs->gpr[rd] = mfspr(SPRN_PVR); 764 regs->gpr[rd] = mfspr(SPRN_PVR);
765 return 0; 765 return 0;
@@ -767,7 +767,7 @@ static int emulate_instruction(struct pt_regs *regs)
767 767
768 /* Emulating the dcba insn is just a no-op. */ 768 /* Emulating the dcba insn is just a no-op. */
769 if ((instword & PPC_INST_DCBA_MASK) == PPC_INST_DCBA) { 769 if ((instword & PPC_INST_DCBA_MASK) == PPC_INST_DCBA) {
770 PPC_WARN_EMULATED(dcba); 770 PPC_WARN_EMULATED(dcba, regs);
771 return 0; 771 return 0;
772 } 772 }
773 773
@@ -776,7 +776,7 @@ static int emulate_instruction(struct pt_regs *regs)
776 int shift = (instword >> 21) & 0x1c; 776 int shift = (instword >> 21) & 0x1c;
777 unsigned long msk = 0xf0000000UL >> shift; 777 unsigned long msk = 0xf0000000UL >> shift;
778 778
779 PPC_WARN_EMULATED(mcrxr); 779 PPC_WARN_EMULATED(mcrxr, regs);
780 regs->ccr = (regs->ccr & ~msk) | ((regs->xer >> shift) & msk); 780 regs->ccr = (regs->ccr & ~msk) | ((regs->xer >> shift) & msk);
781 regs->xer &= ~0xf0000000UL; 781 regs->xer &= ~0xf0000000UL;
782 return 0; 782 return 0;
@@ -784,19 +784,19 @@ static int emulate_instruction(struct pt_regs *regs)
784 784
785 /* Emulate load/store string insn. */ 785 /* Emulate load/store string insn. */
786 if ((instword & PPC_INST_STRING_GEN_MASK) == PPC_INST_STRING) { 786 if ((instword & PPC_INST_STRING_GEN_MASK) == PPC_INST_STRING) {
787 PPC_WARN_EMULATED(string); 787 PPC_WARN_EMULATED(string, regs);
788 return emulate_string_inst(regs, instword); 788 return emulate_string_inst(regs, instword);
789 } 789 }
790 790
791 /* Emulate the popcntb (Population Count Bytes) instruction. */ 791 /* Emulate the popcntb (Population Count Bytes) instruction. */
792 if ((instword & PPC_INST_POPCNTB_MASK) == PPC_INST_POPCNTB) { 792 if ((instword & PPC_INST_POPCNTB_MASK) == PPC_INST_POPCNTB) {
793 PPC_WARN_EMULATED(popcntb); 793 PPC_WARN_EMULATED(popcntb, regs);
794 return emulate_popcntb_inst(regs, instword); 794 return emulate_popcntb_inst(regs, instword);
795 } 795 }
796 796
797 /* Emulate isel (Integer Select) instruction */ 797 /* Emulate isel (Integer Select) instruction */
798 if ((instword & PPC_INST_ISEL_MASK) == PPC_INST_ISEL) { 798 if ((instword & PPC_INST_ISEL_MASK) == PPC_INST_ISEL) {
799 PPC_WARN_EMULATED(isel); 799 PPC_WARN_EMULATED(isel, regs);
800 return emulate_isel(regs, instword); 800 return emulate_isel(regs, instword);
801 } 801 }
802 802
@@ -995,7 +995,7 @@ void SoftwareEmulation(struct pt_regs *regs)
995#ifdef CONFIG_MATH_EMULATION 995#ifdef CONFIG_MATH_EMULATION
996 errcode = do_mathemu(regs); 996 errcode = do_mathemu(regs);
997 if (errcode >= 0) 997 if (errcode >= 0)
998 PPC_WARN_EMULATED(math); 998 PPC_WARN_EMULATED(math, regs);
999 999
1000 switch (errcode) { 1000 switch (errcode) {
1001 case 0: 1001 case 0:
@@ -1018,7 +1018,7 @@ void SoftwareEmulation(struct pt_regs *regs)
1018#elif defined(CONFIG_8XX_MINIMAL_FPEMU) 1018#elif defined(CONFIG_8XX_MINIMAL_FPEMU)
1019 errcode = Soft_emulate_8xx(regs); 1019 errcode = Soft_emulate_8xx(regs);
1020 if (errcode >= 0) 1020 if (errcode >= 0)
1021 PPC_WARN_EMULATED(8xx); 1021 PPC_WARN_EMULATED(8xx, regs);
1022 1022
1023 switch (errcode) { 1023 switch (errcode) {
1024 case 0: 1024 case 0:
@@ -1129,7 +1129,7 @@ void altivec_assist_exception(struct pt_regs *regs)
1129 1129
1130 flush_altivec_to_thread(current); 1130 flush_altivec_to_thread(current);
1131 1131
1132 PPC_WARN_EMULATED(altivec); 1132 PPC_WARN_EMULATED(altivec, regs);
1133 err = emulate_altivec(regs); 1133 err = emulate_altivec(regs);
1134 if (err == 0) { 1134 if (err == 0) {
1135 regs->nip += 4; /* skip emulated instruction */ 1135 regs->nip += 4; /* skip emulated instruction */
diff --git a/arch/powerpc/lib/copypage_64.S b/arch/powerpc/lib/copypage_64.S
index 75f3267fdc30..e68beac0a171 100644
--- a/arch/powerpc/lib/copypage_64.S
+++ b/arch/powerpc/lib/copypage_64.S
@@ -26,11 +26,11 @@ BEGIN_FTR_SECTION
26 srd r8,r5,r11 26 srd r8,r5,r11
27 27
28 mtctr r8 28 mtctr r8
29setup: 29.Lsetup:
30 dcbt r9,r4 30 dcbt r9,r4
31 dcbz r9,r3 31 dcbz r9,r3
32 add r9,r9,r12 32 add r9,r9,r12
33 bdnz setup 33 bdnz .Lsetup
34END_FTR_SECTION_IFSET(CPU_FTR_CP_USE_DCBTZ) 34END_FTR_SECTION_IFSET(CPU_FTR_CP_USE_DCBTZ)
35 addi r3,r3,-8 35 addi r3,r3,-8
36 srdi r8,r5,7 /* page is copied in 128 byte strides */ 36 srdi r8,r5,7 /* page is copied in 128 byte strides */
diff --git a/arch/powerpc/platforms/pseries/hvCall.S b/arch/powerpc/platforms/pseries/hvCall.S
index c1427b3634ec..383a5d0e9818 100644
--- a/arch/powerpc/platforms/pseries/hvCall.S
+++ b/arch/powerpc/platforms/pseries/hvCall.S
@@ -14,68 +14,94 @@
14 14
15#define STK_PARM(i) (48 + ((i)-3)*8) 15#define STK_PARM(i) (48 + ((i)-3)*8)
16 16
17#ifdef CONFIG_HCALL_STATS 17#ifdef CONFIG_TRACEPOINTS
18
19 .section ".toc","aw"
20
21 .globl hcall_tracepoint_refcount
22hcall_tracepoint_refcount:
23 .llong 0
24
25 .section ".text"
26
18/* 27/*
19 * precall must preserve all registers. use unused STK_PARM() 28 * precall must preserve all registers. use unused STK_PARM()
20 * areas to save snapshots and opcode. 29 * areas to save snapshots and opcode. We branch around this
30 * in early init (eg when populating the MMU hashtable) by using an
31 * unconditional cpu feature.
21 */ 32 */
22#define HCALL_INST_PRECALL \ 33#define HCALL_INST_PRECALL(FIRST_REG) \
23 std r3,STK_PARM(r3)(r1); /* save opcode */ \
24 mftb r0; /* get timebase and */ \
25 std r0,STK_PARM(r5)(r1); /* save for later */ \
26BEGIN_FTR_SECTION; \ 34BEGIN_FTR_SECTION; \
27 mfspr r0,SPRN_PURR; /* get PURR and */ \ 35 b 1f; \
28 std r0,STK_PARM(r6)(r1); /* save for later */ \ 36END_FTR_SECTION(0, 1); \
29END_FTR_SECTION_IFSET(CPU_FTR_PURR); 37 ld r12,hcall_tracepoint_refcount@toc(r2); \
30 38 cmpdi r12,0; \
39 beq+ 1f; \
40 mflr r0; \
41 std r3,STK_PARM(r3)(r1); \
42 std r4,STK_PARM(r4)(r1); \
43 std r5,STK_PARM(r5)(r1); \
44 std r6,STK_PARM(r6)(r1); \
45 std r7,STK_PARM(r7)(r1); \
46 std r8,STK_PARM(r8)(r1); \
47 std r9,STK_PARM(r9)(r1); \
48 std r10,STK_PARM(r10)(r1); \
49 std r0,16(r1); \
50 addi r4,r1,STK_PARM(FIRST_REG); \
51 stdu r1,-STACK_FRAME_OVERHEAD(r1); \
52 bl .__trace_hcall_entry; \
53 addi r1,r1,STACK_FRAME_OVERHEAD; \
54 ld r0,16(r1); \
55 ld r3,STK_PARM(r3)(r1); \
56 ld r4,STK_PARM(r4)(r1); \
57 ld r5,STK_PARM(r5)(r1); \
58 ld r6,STK_PARM(r6)(r1); \
59 ld r7,STK_PARM(r7)(r1); \
60 ld r8,STK_PARM(r8)(r1); \
61 ld r9,STK_PARM(r9)(r1); \
62 ld r10,STK_PARM(r10)(r1); \
63 mtlr r0; \
641:
65
31/* 66/*
32 * postcall is performed immediately before function return which 67 * postcall is performed immediately before function return which
33 * allows liberal use of volatile registers. We branch around this 68 * allows liberal use of volatile registers. We branch around this
34 * in early init (eg when populating the MMU hashtable) by using an 69 * in early init (eg when populating the MMU hashtable) by using an
35 * unconditional cpu feature. 70 * unconditional cpu feature.
36 */ 71 */
37#define HCALL_INST_POSTCALL \ 72#define __HCALL_INST_POSTCALL \
38BEGIN_FTR_SECTION; \ 73BEGIN_FTR_SECTION; \
39 b 1f; \ 74 b 1f; \
40END_FTR_SECTION(0, 1); \ 75END_FTR_SECTION(0, 1); \
41 ld r4,STK_PARM(r3)(r1); /* validate opcode */ \ 76 ld r12,hcall_tracepoint_refcount@toc(r2); \
42 cmpldi cr7,r4,MAX_HCALL_OPCODE; \ 77 cmpdi r12,0; \
43 bgt- cr7,1f; \ 78 beq+ 1f; \
44 \ 79 mflr r0; \
45 /* get time and PURR snapshots after hcall */ \ 80 ld r6,STK_PARM(r3)(r1); \
46 mftb r7; /* timebase after */ \ 81 std r3,STK_PARM(r3)(r1); \
47BEGIN_FTR_SECTION; \ 82 mr r4,r3; \
48 mfspr r8,SPRN_PURR; /* PURR after */ \ 83 mr r3,r6; \
49 ld r6,STK_PARM(r6)(r1); /* PURR before */ \ 84 std r0,16(r1); \
50 subf r6,r6,r8; /* delta */ \ 85 stdu r1,-STACK_FRAME_OVERHEAD(r1); \
51END_FTR_SECTION_IFSET(CPU_FTR_PURR); \ 86 bl .__trace_hcall_exit; \
52 ld r5,STK_PARM(r5)(r1); /* timebase before */ \ 87 addi r1,r1,STACK_FRAME_OVERHEAD; \
53 subf r5,r5,r7; /* time delta */ \ 88 ld r0,16(r1); \
54 \ 89 ld r3,STK_PARM(r3)(r1); \
55 /* calculate address of stat structure r4 = opcode */ \ 90 mtlr r0; \
56 srdi r4,r4,2; /* index into array */ \
57 mulli r4,r4,HCALL_STAT_SIZE; \
58 LOAD_REG_ADDR(r7, per_cpu__hcall_stats); \
59 add r4,r4,r7; \
60 ld r7,PACA_DATA_OFFSET(r13); /* per cpu offset */ \
61 add r4,r4,r7; \
62 \
63 /* update stats */ \
64 ld r7,HCALL_STAT_CALLS(r4); /* count */ \
65 addi r7,r7,1; \
66 std r7,HCALL_STAT_CALLS(r4); \
67 ld r7,HCALL_STAT_TB(r4); /* timebase */ \
68 add r7,r7,r5; \
69 std r7,HCALL_STAT_TB(r4); \
70BEGIN_FTR_SECTION; \
71 ld r7,HCALL_STAT_PURR(r4); /* PURR */ \
72 add r7,r7,r6; \
73 std r7,HCALL_STAT_PURR(r4); \
74END_FTR_SECTION_IFSET(CPU_FTR_PURR); \
751: 911:
92
93#define HCALL_INST_POSTCALL_NORETS \
94 li r5,0; \
95 __HCALL_INST_POSTCALL
96
97#define HCALL_INST_POSTCALL(BUFREG) \
98 mr r5,BUFREG; \
99 __HCALL_INST_POSTCALL
100
76#else 101#else
77#define HCALL_INST_PRECALL 102#define HCALL_INST_PRECALL(FIRST_ARG)
78#define HCALL_INST_POSTCALL 103#define HCALL_INST_POSTCALL_NORETS
104#define HCALL_INST_POSTCALL(BUFREG)
79#endif 105#endif
80 106
81 .text 107 .text
@@ -86,11 +112,11 @@ _GLOBAL(plpar_hcall_norets)
86 mfcr r0 112 mfcr r0
87 stw r0,8(r1) 113 stw r0,8(r1)
88 114
89 HCALL_INST_PRECALL 115 HCALL_INST_PRECALL(r4)
90 116
91 HVSC /* invoke the hypervisor */ 117 HVSC /* invoke the hypervisor */
92 118
93 HCALL_INST_POSTCALL 119 HCALL_INST_POSTCALL_NORETS
94 120
95 lwz r0,8(r1) 121 lwz r0,8(r1)
96 mtcrf 0xff,r0 122 mtcrf 0xff,r0
@@ -102,7 +128,7 @@ _GLOBAL(plpar_hcall)
102 mfcr r0 128 mfcr r0
103 stw r0,8(r1) 129 stw r0,8(r1)
104 130
105 HCALL_INST_PRECALL 131 HCALL_INST_PRECALL(r5)
106 132
107 std r4,STK_PARM(r4)(r1) /* Save ret buffer */ 133 std r4,STK_PARM(r4)(r1) /* Save ret buffer */
108 134
@@ -121,7 +147,7 @@ _GLOBAL(plpar_hcall)
121 std r6, 16(r12) 147 std r6, 16(r12)
122 std r7, 24(r12) 148 std r7, 24(r12)
123 149
124 HCALL_INST_POSTCALL 150 HCALL_INST_POSTCALL(r12)
125 151
126 lwz r0,8(r1) 152 lwz r0,8(r1)
127 mtcrf 0xff,r0 153 mtcrf 0xff,r0
@@ -168,7 +194,7 @@ _GLOBAL(plpar_hcall9)
168 mfcr r0 194 mfcr r0
169 stw r0,8(r1) 195 stw r0,8(r1)
170 196
171 HCALL_INST_PRECALL 197 HCALL_INST_PRECALL(r5)
172 198
173 std r4,STK_PARM(r4)(r1) /* Save ret buffer */ 199 std r4,STK_PARM(r4)(r1) /* Save ret buffer */
174 200
@@ -196,7 +222,7 @@ _GLOBAL(plpar_hcall9)
196 std r11,56(r12) 222 std r11,56(r12)
197 std r0, 64(r12) 223 std r0, 64(r12)
198 224
199 HCALL_INST_POSTCALL 225 HCALL_INST_POSTCALL(r12)
200 226
201 lwz r0,8(r1) 227 lwz r0,8(r1)
202 mtcrf 0xff,r0 228 mtcrf 0xff,r0
diff --git a/arch/powerpc/platforms/pseries/hvCall_inst.c b/arch/powerpc/platforms/pseries/hvCall_inst.c
index 3631a4f277eb..2f58c71b7259 100644
--- a/arch/powerpc/platforms/pseries/hvCall_inst.c
+++ b/arch/powerpc/platforms/pseries/hvCall_inst.c
@@ -26,6 +26,7 @@
26#include <asm/hvcall.h> 26#include <asm/hvcall.h>
27#include <asm/firmware.h> 27#include <asm/firmware.h>
28#include <asm/cputable.h> 28#include <asm/cputable.h>
29#include <asm/trace.h>
29 30
30DEFINE_PER_CPU(struct hcall_stats[HCALL_STAT_ARRAY_SIZE], hcall_stats); 31DEFINE_PER_CPU(struct hcall_stats[HCALL_STAT_ARRAY_SIZE], hcall_stats);
31 32
@@ -100,6 +101,35 @@ static const struct file_operations hcall_inst_seq_fops = {
100#define HCALL_ROOT_DIR "hcall_inst" 101#define HCALL_ROOT_DIR "hcall_inst"
101#define CPU_NAME_BUF_SIZE 32 102#define CPU_NAME_BUF_SIZE 32
102 103
104
105static void probe_hcall_entry(unsigned long opcode, unsigned long *args)
106{
107 struct hcall_stats *h;
108
109 if (opcode > MAX_HCALL_OPCODE)
110 return;
111
112 h = &get_cpu_var(hcall_stats)[opcode / 4];
113 h->tb_start = mftb();
114 h->purr_start = mfspr(SPRN_PURR);
115}
116
117static void probe_hcall_exit(unsigned long opcode, unsigned long retval,
118 unsigned long *retbuf)
119{
120 struct hcall_stats *h;
121
122 if (opcode > MAX_HCALL_OPCODE)
123 return;
124
125 h = &__get_cpu_var(hcall_stats)[opcode / 4];
126 h->num_calls++;
127 h->tb_total = mftb() - h->tb_start;
128 h->purr_total = mfspr(SPRN_PURR) - h->purr_start;
129
130 put_cpu_var(hcall_stats);
131}
132
103static int __init hcall_inst_init(void) 133static int __init hcall_inst_init(void)
104{ 134{
105 struct dentry *hcall_root; 135 struct dentry *hcall_root;
@@ -110,6 +140,14 @@ static int __init hcall_inst_init(void)
110 if (!firmware_has_feature(FW_FEATURE_LPAR)) 140 if (!firmware_has_feature(FW_FEATURE_LPAR))
111 return 0; 141 return 0;
112 142
143 if (register_trace_hcall_entry(probe_hcall_entry))
144 return -EINVAL;
145
146 if (register_trace_hcall_exit(probe_hcall_exit)) {
147 unregister_trace_hcall_entry(probe_hcall_entry);
148 return -EINVAL;
149 }
150
113 hcall_root = debugfs_create_dir(HCALL_ROOT_DIR, NULL); 151 hcall_root = debugfs_create_dir(HCALL_ROOT_DIR, NULL);
114 if (!hcall_root) 152 if (!hcall_root)
115 return -ENOMEM; 153 return -ENOMEM;
diff --git a/arch/powerpc/platforms/pseries/lpar.c b/arch/powerpc/platforms/pseries/lpar.c
index 903eb9eec687..0707653612ba 100644
--- a/arch/powerpc/platforms/pseries/lpar.c
+++ b/arch/powerpc/platforms/pseries/lpar.c
@@ -39,6 +39,7 @@
39#include <asm/cputable.h> 39#include <asm/cputable.h>
40#include <asm/udbg.h> 40#include <asm/udbg.h>
41#include <asm/smp.h> 41#include <asm/smp.h>
42#include <asm/trace.h>
42 43
43#include "plpar_wrappers.h" 44#include "plpar_wrappers.h"
44#include "pseries.h" 45#include "pseries.h"
@@ -661,3 +662,35 @@ void arch_free_page(struct page *page, int order)
661EXPORT_SYMBOL(arch_free_page); 662EXPORT_SYMBOL(arch_free_page);
662 663
663#endif 664#endif
665
666#ifdef CONFIG_TRACEPOINTS
667/*
668 * We optimise our hcall path by placing hcall_tracepoint_refcount
669 * directly in the TOC so we can check if the hcall tracepoints are
670 * enabled via a single load.
671 */
672
673/* NB: reg/unreg are called while guarded with the tracepoints_mutex */
674extern long hcall_tracepoint_refcount;
675
676void hcall_tracepoint_regfunc(void)
677{
678 hcall_tracepoint_refcount++;
679}
680
681void hcall_tracepoint_unregfunc(void)
682{
683 hcall_tracepoint_refcount--;
684}
685
686void __trace_hcall_entry(unsigned long opcode, unsigned long *args)
687{
688 trace_hcall_entry(opcode, args);
689}
690
691void __trace_hcall_exit(long opcode, unsigned long retval,
692 unsigned long *retbuf)
693{
694 trace_hcall_exit(opcode, retval, retbuf);
695}
696#endif
diff --git a/include/linux/perf_counter.h b/include/linux/perf_counter.h
index 91a2b4309e7a..e3fb25606706 100644
--- a/include/linux/perf_counter.h
+++ b/include/linux/perf_counter.h
@@ -106,6 +106,8 @@ enum perf_sw_ids {
106 PERF_COUNT_SW_CPU_MIGRATIONS = 4, 106 PERF_COUNT_SW_CPU_MIGRATIONS = 4,
107 PERF_COUNT_SW_PAGE_FAULTS_MIN = 5, 107 PERF_COUNT_SW_PAGE_FAULTS_MIN = 5,
108 PERF_COUNT_SW_PAGE_FAULTS_MAJ = 6, 108 PERF_COUNT_SW_PAGE_FAULTS_MAJ = 6,
109 PERF_COUNT_SW_ALIGNMENT_FAULTS = 7,
110 PERF_COUNT_SW_EMULATION_FAULTS = 8,
109 111
110 PERF_COUNT_SW_MAX, /* non-ABI */ 112 PERF_COUNT_SW_MAX, /* non-ABI */
111}; 113};
diff --git a/include/linux/perf_event.h b/include/linux/perf_event.h
index ec3768a81058..df4e73e33774 100644
--- a/include/linux/perf_event.h
+++ b/include/linux/perf_event.h
@@ -102,6 +102,8 @@ enum perf_sw_ids {
102 PERF_COUNT_SW_CPU_MIGRATIONS = 4, 102 PERF_COUNT_SW_CPU_MIGRATIONS = 4,
103 PERF_COUNT_SW_PAGE_FAULTS_MIN = 5, 103 PERF_COUNT_SW_PAGE_FAULTS_MIN = 5,
104 PERF_COUNT_SW_PAGE_FAULTS_MAJ = 6, 104 PERF_COUNT_SW_PAGE_FAULTS_MAJ = 6,
105 PERF_COUNT_SW_ALIGNMENT_FAULTS = 7,
106 PERF_COUNT_SW_EMULATION_FAULTS = 8,
105 107
106 PERF_COUNT_SW_MAX, /* non-ABI */ 108 PERF_COUNT_SW_MAX, /* non-ABI */
107}; 109};
diff --git a/kernel/perf_event.c b/kernel/perf_event.c
index a69d4ed6a666..6f4ed3b4cd73 100644
--- a/kernel/perf_event.c
+++ b/kernel/perf_event.c
@@ -4274,6 +4274,8 @@ static const struct pmu *sw_perf_event_init(struct perf_event *event)
4274 case PERF_COUNT_SW_PAGE_FAULTS_MAJ: 4274 case PERF_COUNT_SW_PAGE_FAULTS_MAJ:
4275 case PERF_COUNT_SW_CONTEXT_SWITCHES: 4275 case PERF_COUNT_SW_CONTEXT_SWITCHES:
4276 case PERF_COUNT_SW_CPU_MIGRATIONS: 4276 case PERF_COUNT_SW_CPU_MIGRATIONS:
4277 case PERF_COUNT_SW_ALIGNMENT_FAULTS:
4278 case PERF_COUNT_SW_EMULATION_FAULTS:
4277 if (!event->parent) { 4279 if (!event->parent) {
4278 atomic_inc(&perf_swevent_enabled[event_id]); 4280 atomic_inc(&perf_swevent_enabled[event_id]);
4279 event->destroy = sw_perf_event_destroy; 4281 event->destroy = sw_perf_event_destroy;
diff --git a/tools/perf/Documentation/perf-bench.txt b/tools/perf/Documentation/perf-bench.txt
new file mode 100644
index 000000000000..ae525ac5a2ce
--- /dev/null
+++ b/tools/perf/Documentation/perf-bench.txt
@@ -0,0 +1,120 @@
1perf-bench(1)
2============
3
4NAME
5----
6perf-bench - General framework for benchmark suites
7
8SYNOPSIS
9--------
10[verse]
11'perf bench' [<common options>] <subsystem> <suite> [<options>]
12
13DESCRIPTION
14-----------
15This 'perf bench' command is general framework for benchmark suites.
16
17COMMON OPTIONS
18--------------
19-f::
20--format=::
21Specify format style.
22Current available format styles are,
23
24'default'::
25Default style. This is mainly for human reading.
26---------------------
27% perf bench sched pipe # with no style specify
28(executing 1000000 pipe operations between two tasks)
29 Total time:5.855 sec
30 5.855061 usecs/op
31 170792 ops/sec
32---------------------
33
34'simple'::
35This simple style is friendly for automated
36processing by scripts.
37---------------------
38% perf bench --format=simple sched pipe # specified simple
395.988
40---------------------
41
42SUBSYSTEM
43---------
44
45'sched'::
46 Scheduler and IPC mechanisms.
47
48SUITES FOR 'sched'
49~~~~~~~~~~~~~~~~~~
50*messaging*::
51Suite for evaluating performance of scheduler and IPC mechanisms.
52Based on hackbench by Rusty Russell.
53
54Options of *pipe*
55^^^^^^^^^^^^^^^^^
56-p::
57--pipe::
58Use pipe() instead of socketpair()
59
60-t::
61--thread::
62Be multi thread instead of multi process
63
64-g::
65--group=::
66Specify number of groups
67
68-l::
69--loop=::
70Specify number of loops
71
72Example of *messaging*
73^^^^^^^^^^^^^^^^^^^^^^
74
75---------------------
76% perf bench sched messaging # run with default
77options (20 sender and receiver processes per group)
78(10 groups == 400 processes run)
79
80 Total time:0.308 sec
81
82% perf bench sched messaging -t -g 20 # be multi-thread,with 20 groups
83(20 sender and receiver threads per group)
84(20 groups == 800 threads run)
85
86 Total time:0.582 sec
87---------------------
88
89*pipe*::
90Suite for pipe() system call.
91Based on pipe-test-1m.c by Ingo Molnar.
92
93Options of *pipe*
94^^^^^^^^^^^^^^^^^
95-l::
96--loop=::
97Specify number of loops.
98
99Example of *pipe*
100^^^^^^^^^^^^^^^^^
101
102---------------------
103% perf bench sched pipe
104(executing 1000000 pipe operations between two tasks)
105
106 Total time:8.091 sec
107 8.091833 usecs/op
108 123581 ops/sec
109
110% perf bench sched pipe -l 1000 # loop 1000
111(executing 1000 pipe operations between two tasks)
112
113 Total time:0.016 sec
114 16.948000 usecs/op
115 59004 ops/sec
116---------------------
117
118SEE ALSO
119--------
120linkperf:perf[1]
diff --git a/tools/perf/Makefile b/tools/perf/Makefile
index e6d42728a33a..f7cd89622cf4 100644
--- a/tools/perf/Makefile
+++ b/tools/perf/Makefile
@@ -421,6 +421,13 @@ LIB_OBJS += util/hist.o
421LIB_OBJS += util/data_map.o 421LIB_OBJS += util/data_map.o
422 422
423BUILTIN_OBJS += builtin-annotate.o 423BUILTIN_OBJS += builtin-annotate.o
424
425BUILTIN_OBJS += builtin-bench.o
426
427# Benchmark modules
428BUILTIN_OBJS += bench/sched-messaging.o
429BUILTIN_OBJS += bench/sched-pipe.o
430
424BUILTIN_OBJS += builtin-help.o 431BUILTIN_OBJS += builtin-help.o
425BUILTIN_OBJS += builtin-sched.o 432BUILTIN_OBJS += builtin-sched.o
426BUILTIN_OBJS += builtin-list.o 433BUILTIN_OBJS += builtin-list.o
diff --git a/tools/perf/bench/bench.h b/tools/perf/bench/bench.h
new file mode 100644
index 000000000000..9fbd8d745fa1
--- /dev/null
+++ b/tools/perf/bench/bench.h
@@ -0,0 +1,16 @@
1#ifndef BENCH_H
2#define BENCH_H
3
4extern int bench_sched_messaging(int argc, const char **argv, const char *prefix);
5extern int bench_sched_pipe(int argc, const char **argv, const char *prefix);
6
7#define BENCH_FORMAT_DEFAULT_STR "default"
8#define BENCH_FORMAT_DEFAULT 0
9#define BENCH_FORMAT_SIMPLE_STR "simple"
10#define BENCH_FORMAT_SIMPLE 1
11
12#define BENCH_FORMAT_UNKNOWN -1
13
14extern int bench_format;
15
16#endif
diff --git a/tools/perf/bench/sched-messaging.c b/tools/perf/bench/sched-messaging.c
new file mode 100644
index 000000000000..605a2a959aa8
--- /dev/null
+++ b/tools/perf/bench/sched-messaging.c
@@ -0,0 +1,336 @@
1/*
2 *
3 * builtin-bench-messaging.c
4 *
5 * messaging: Benchmark for scheduler and IPC mechanisms
6 *
7 * Based on hackbench by Rusty Russell <rusty@rustcorp.com.au>
8 * Ported to perf by Hitoshi Mitake <mitake@dcl.info.waseda.ac.jp>
9 *
10 */
11
12#include "../perf.h"
13#include "../util/util.h"
14#include "../util/parse-options.h"
15#include "../builtin.h"
16#include "bench.h"
17
18/* Test groups of 20 processes spraying to 20 receivers */
19#include <pthread.h>
20#include <stdio.h>
21#include <stdlib.h>
22#include <string.h>
23#include <errno.h>
24#include <unistd.h>
25#include <sys/types.h>
26#include <sys/socket.h>
27#include <sys/wait.h>
28#include <sys/time.h>
29#include <sys/poll.h>
30#include <limits.h>
31
32#define DATASIZE 100
33
34static int use_pipes = 0;
35static unsigned int loops = 100;
36static unsigned int thread_mode = 0;
37static unsigned int num_groups = 10;
38
39struct sender_context {
40 unsigned int num_fds;
41 int ready_out;
42 int wakefd;
43 int out_fds[0];
44};
45
46struct receiver_context {
47 unsigned int num_packets;
48 int in_fds[2];
49 int ready_out;
50 int wakefd;
51};
52
53static void barf(const char *msg)
54{
55 fprintf(stderr, "%s (error: %s)\n", msg, strerror(errno));
56 exit(1);
57}
58
59static void fdpair(int fds[2])
60{
61 if (use_pipes) {
62 if (pipe(fds) == 0)
63 return;
64 } else {
65 if (socketpair(AF_UNIX, SOCK_STREAM, 0, fds) == 0)
66 return;
67 }
68
69 barf(use_pipes ? "pipe()" : "socketpair()");
70}
71
72/* Block until we're ready to go */
73static void ready(int ready_out, int wakefd)
74{
75 char dummy;
76 struct pollfd pollfd = { .fd = wakefd, .events = POLLIN };
77
78 /* Tell them we're ready. */
79 if (write(ready_out, &dummy, 1) != 1)
80 barf("CLIENT: ready write");
81
82 /* Wait for "GO" signal */
83 if (poll(&pollfd, 1, -1) != 1)
84 barf("poll");
85}
86
87/* Sender sprays loops messages down each file descriptor */
88static void *sender(struct sender_context *ctx)
89{
90 char data[DATASIZE];
91 unsigned int i, j;
92
93 ready(ctx->ready_out, ctx->wakefd);
94
95 /* Now pump to every receiver. */
96 for (i = 0; i < loops; i++) {
97 for (j = 0; j < ctx->num_fds; j++) {
98 int ret, done = 0;
99
100again:
101 ret = write(ctx->out_fds[j], data + done,
102 sizeof(data)-done);
103 if (ret < 0)
104 barf("SENDER: write");
105 done += ret;
106 if (done < DATASIZE)
107 goto again;
108 }
109 }
110
111 return NULL;
112}
113
114
115/* One receiver per fd */
116static void *receiver(struct receiver_context* ctx)
117{
118 unsigned int i;
119
120 if (!thread_mode)
121 close(ctx->in_fds[1]);
122
123 /* Wait for start... */
124 ready(ctx->ready_out, ctx->wakefd);
125
126 /* Receive them all */
127 for (i = 0; i < ctx->num_packets; i++) {
128 char data[DATASIZE];
129 int ret, done = 0;
130
131again:
132 ret = read(ctx->in_fds[0], data + done, DATASIZE - done);
133 if (ret < 0)
134 barf("SERVER: read");
135 done += ret;
136 if (done < DATASIZE)
137 goto again;
138 }
139
140 return NULL;
141}
142
143static pthread_t create_worker(void *ctx, void *(*func)(void *))
144{
145 pthread_attr_t attr;
146 pthread_t childid;
147 int err;
148
149 if (!thread_mode) {
150 /* process mode */
151 /* Fork the receiver. */
152 switch (fork()) {
153 case -1:
154 barf("fork()");
155 break;
156 case 0:
157 (*func) (ctx);
158 exit(0);
159 break;
160 default:
161 break;
162 }
163
164 return (pthread_t)0;
165 }
166
167 if (pthread_attr_init(&attr) != 0)
168 barf("pthread_attr_init:");
169
170#ifndef __ia64__
171 if (pthread_attr_setstacksize(&attr, PTHREAD_STACK_MIN) != 0)
172 barf("pthread_attr_setstacksize");
173#endif
174
175 err = pthread_create(&childid, &attr, func, ctx);
176 if (err != 0) {
177 fprintf(stderr, "pthread_create failed: %s (%d)\n",
178 strerror(err), err);
179 exit(-1);
180 }
181 return childid;
182}
183
184static void reap_worker(pthread_t id)
185{
186 int proc_status;
187 void *thread_status;
188
189 if (!thread_mode) {
190 /* process mode */
191 wait(&proc_status);
192 if (!WIFEXITED(proc_status))
193 exit(1);
194 } else {
195 pthread_join(id, &thread_status);
196 }
197}
198
199/* One group of senders and receivers */
200static unsigned int group(pthread_t *pth,
201 unsigned int num_fds,
202 int ready_out,
203 int wakefd)
204{
205 unsigned int i;
206 struct sender_context *snd_ctx = malloc(sizeof(struct sender_context)
207 + num_fds * sizeof(int));
208
209 if (!snd_ctx)
210 barf("malloc()");
211
212 for (i = 0; i < num_fds; i++) {
213 int fds[2];
214 struct receiver_context *ctx = malloc(sizeof(*ctx));
215
216 if (!ctx)
217 barf("malloc()");
218
219
220 /* Create the pipe between client and server */
221 fdpair(fds);
222
223 ctx->num_packets = num_fds * loops;
224 ctx->in_fds[0] = fds[0];
225 ctx->in_fds[1] = fds[1];
226 ctx->ready_out = ready_out;
227 ctx->wakefd = wakefd;
228
229 pth[i] = create_worker(ctx, (void *)receiver);
230
231 snd_ctx->out_fds[i] = fds[1];
232 if (!thread_mode)
233 close(fds[0]);
234 }
235
236 /* Now we have all the fds, fork the senders */
237 for (i = 0; i < num_fds; i++) {
238 snd_ctx->ready_out = ready_out;
239 snd_ctx->wakefd = wakefd;
240 snd_ctx->num_fds = num_fds;
241
242 pth[num_fds+i] = create_worker(snd_ctx, (void *)sender);
243 }
244
245 /* Close the fds we have left */
246 if (!thread_mode)
247 for (i = 0; i < num_fds; i++)
248 close(snd_ctx->out_fds[i]);
249
250 /* Return number of children to reap */
251 return num_fds * 2;
252}
253
254static const struct option options[] = {
255 OPT_BOOLEAN('p', "pipe", &use_pipes,
256 "Use pipe() instead of socketpair()"),
257 OPT_BOOLEAN('t', "thread", &thread_mode,
258 "Be multi thread instead of multi process"),
259 OPT_INTEGER('g', "group", &num_groups,
260 "Specify number of groups"),
261 OPT_INTEGER('l', "loop", &loops,
262 "Specify number of loops"),
263 OPT_END()
264};
265
266static const char * const bench_sched_message_usage[] = {
267 "perf bench sched messaging <options>",
268 NULL
269};
270
271int bench_sched_messaging(int argc, const char **argv,
272 const char *prefix __used)
273{
274 unsigned int i, total_children;
275 struct timeval start, stop, diff;
276 unsigned int num_fds = 20;
277 int readyfds[2], wakefds[2];
278 char dummy;
279 pthread_t *pth_tab;
280
281 argc = parse_options(argc, argv, options,
282 bench_sched_message_usage, 0);
283
284 pth_tab = malloc(num_fds * 2 * num_groups * sizeof(pthread_t));
285 if (!pth_tab)
286 barf("main:malloc()");
287
288 fdpair(readyfds);
289 fdpair(wakefds);
290
291 total_children = 0;
292 for (i = 0; i < num_groups; i++)
293 total_children += group(pth_tab+total_children, num_fds,
294 readyfds[1], wakefds[0]);
295
296 /* Wait for everyone to be ready */
297 for (i = 0; i < total_children; i++)
298 if (read(readyfds[0], &dummy, 1) != 1)
299 barf("Reading for readyfds");
300
301 gettimeofday(&start, NULL);
302
303 /* Kick them off */
304 if (write(wakefds[1], &dummy, 1) != 1)
305 barf("Writing to start them");
306
307 /* Reap them all */
308 for (i = 0; i < total_children; i++)
309 reap_worker(pth_tab[i]);
310
311 gettimeofday(&stop, NULL);
312
313 timersub(&stop, &start, &diff);
314
315 switch (bench_format) {
316 case BENCH_FORMAT_DEFAULT:
317 printf("# %d sender and receiver %s per group\n",
318 num_fds, thread_mode ? "threads" : "processes");
319 printf("# %d groups == %d %s run\n\n",
320 num_groups, num_groups * 2 * num_fds,
321 thread_mode ? "threads" : "processes");
322 printf(" %14s: %lu.%03lu [sec]\n", "Total time",
323 diff.tv_sec, diff.tv_usec/1000);
324 break;
325 case BENCH_FORMAT_SIMPLE:
326 printf("%lu.%03lu\n", diff.tv_sec, diff.tv_usec/1000);
327 break;
328 default:
329 /* reaching here is something disaster */
330 fprintf(stderr, "Unknown format:%d\n", bench_format);
331 exit(1);
332 break;
333 }
334
335 return 0;
336}
diff --git a/tools/perf/bench/sched-pipe.c b/tools/perf/bench/sched-pipe.c
new file mode 100644
index 000000000000..238185f97977
--- /dev/null
+++ b/tools/perf/bench/sched-pipe.c
@@ -0,0 +1,124 @@
1/*
2 *
3 * builtin-bench-pipe.c
4 *
5 * pipe: Benchmark for pipe()
6 *
7 * Based on pipe-test-1m.c by Ingo Molnar <mingo@redhat.com>
8 * http://people.redhat.com/mingo/cfs-scheduler/tools/pipe-test-1m.c
9 * Ported to perf by Hitoshi Mitake <mitake@dcl.info.waseda.ac.jp>
10 *
11 */
12
13#include "../perf.h"
14#include "../util/util.h"
15#include "../util/parse-options.h"
16#include "../builtin.h"
17#include "bench.h"
18
19#include <unistd.h>
20#include <stdio.h>
21#include <stdlib.h>
22#include <signal.h>
23#include <sys/wait.h>
24#include <linux/unistd.h>
25#include <string.h>
26#include <errno.h>
27#include <assert.h>
28#include <sys/time.h>
29#include <sys/types.h>
30
31#define LOOPS_DEFAULT 1000000
32static int loops = LOOPS_DEFAULT;
33
34static const struct option options[] = {
35 OPT_INTEGER('l', "loop", &loops,
36 "Specify number of loops"),
37 OPT_END()
38};
39
40static const char * const bench_sched_pipe_usage[] = {
41 "perf bench sched pipe <options>",
42 NULL
43};
44
45int bench_sched_pipe(int argc, const char **argv,
46 const char *prefix __used)
47{
48 int pipe_1[2], pipe_2[2];
49 int m = 0, i;
50 struct timeval start, stop, diff;
51 unsigned long long result_usec = 0;
52
53 /*
54 * why does "ret" exist?
55 * discarding returned value of read(), write()
56 * causes error in building environment for perf
57 */
58 int ret, wait_stat;
59 pid_t pid, retpid;
60
61 argc = parse_options(argc, argv, options,
62 bench_sched_pipe_usage, 0);
63
64 assert(!pipe(pipe_1));
65 assert(!pipe(pipe_2));
66
67 pid = fork();
68 assert(pid >= 0);
69
70 gettimeofday(&start, NULL);
71
72 if (!pid) {
73 for (i = 0; i < loops; i++) {
74 ret = read(pipe_1[0], &m, sizeof(int));
75 ret = write(pipe_2[1], &m, sizeof(int));
76 }
77 } else {
78 for (i = 0; i < loops; i++) {
79 ret = write(pipe_1[1], &m, sizeof(int));
80 ret = read(pipe_2[0], &m, sizeof(int));
81 }
82 }
83
84 gettimeofday(&stop, NULL);
85 timersub(&stop, &start, &diff);
86
87 if (pid) {
88 retpid = waitpid(pid, &wait_stat, 0);
89 assert((retpid == pid) && WIFEXITED(wait_stat));
90 return 0;
91 }
92
93 switch (bench_format) {
94 case BENCH_FORMAT_DEFAULT:
95 printf("# Extecuted %d pipe operations between two tasks\n\n",
96 loops);
97
98 result_usec = diff.tv_sec * 1000000;
99 result_usec += diff.tv_usec;
100
101 printf(" %14s: %lu.%03lu [sec]\n\n", "Total time",
102 diff.tv_sec, diff.tv_usec/1000);
103
104 printf(" %14lf usecs/op\n",
105 (double)result_usec / (double)loops);
106 printf(" %14d ops/sec\n",
107 (int)((double)loops /
108 ((double)result_usec / (double)1000000)));
109 break;
110
111 case BENCH_FORMAT_SIMPLE:
112 printf("%lu.%03lu\n",
113 diff.tv_sec, diff.tv_usec / 1000);
114 break;
115
116 default:
117 /* reaching here is something disaster */
118 fprintf(stderr, "Unknown format:%d\n", bench_format);
119 exit(1);
120 break;
121 }
122
123 return 0;
124}
diff --git a/tools/perf/builtin-bench.c b/tools/perf/builtin-bench.c
new file mode 100644
index 000000000000..90c39baae0de
--- /dev/null
+++ b/tools/perf/builtin-bench.c
@@ -0,0 +1,183 @@
1/*
2 *
3 * builtin-bench.c
4 *
5 * General benchmarking subsystem provided by perf
6 *
7 * Copyright (C) 2009, Hitoshi Mitake <mitake@dcl.info.waseda.ac.jp>
8 *
9 */
10
11/*
12 *
13 * Available subsystem list:
14 * sched ... scheduler and IPC mechanism
15 *
16 */
17
18#include "perf.h"
19#include "util/util.h"
20#include "util/parse-options.h"
21#include "builtin.h"
22#include "bench/bench.h"
23
24#include <stdio.h>
25#include <stdlib.h>
26#include <string.h>
27
28struct bench_suite {
29 const char *name;
30 const char *summary;
31 int (*fn)(int, const char **, const char *);
32};
33
34static struct bench_suite sched_suites[] = {
35 { "messaging",
36 "Benchmark for scheduler and IPC mechanisms",
37 bench_sched_messaging },
38 { "pipe",
39 "Flood of communication over pipe() between two processes",
40 bench_sched_pipe },
41 { NULL,
42 NULL,
43 NULL }
44};
45
46struct bench_subsys {
47 const char *name;
48 const char *summary;
49 struct bench_suite *suites;
50};
51
52static struct bench_subsys subsystems[] = {
53 { "sched",
54 "scheduler and IPC mechanism",
55 sched_suites },
56 { NULL,
57 NULL,
58 NULL }
59};
60
61static void dump_suites(int subsys_index)
62{
63 int i;
64
65 printf("List of available suites for %s...\n\n",
66 subsystems[subsys_index].name);
67
68 for (i = 0; subsystems[subsys_index].suites[i].name; i++)
69 printf("\t%s: %s\n",
70 subsystems[subsys_index].suites[i].name,
71 subsystems[subsys_index].suites[i].summary);
72
73 printf("\n");
74 return;
75}
76
77static char *bench_format_str;
78int bench_format = BENCH_FORMAT_DEFAULT;
79
80static const struct option bench_options[] = {
81 OPT_STRING('f', "format", &bench_format_str, "default",
82 "Specify format style"),
83 OPT_END()
84};
85
86static const char * const bench_usage[] = {
87 "perf bench [<common options>] <subsystem> <suite> [<options>]",
88 NULL
89};
90
91static void print_usage(void)
92{
93 int i;
94
95 printf("Usage: \n");
96 for (i = 0; bench_usage[i]; i++)
97 printf("\t%s\n", bench_usage[i]);
98 printf("\n");
99
100 printf("List of available subsystems...\n\n");
101
102 for (i = 0; subsystems[i].name; i++)
103 printf("\t%s: %s\n",
104 subsystems[i].name, subsystems[i].summary);
105 printf("\n");
106}
107
108static int bench_str2int(char *str)
109{
110 if (!str)
111 return BENCH_FORMAT_DEFAULT;
112
113 if (!strcmp(str, BENCH_FORMAT_DEFAULT_STR))
114 return BENCH_FORMAT_DEFAULT;
115 else if (!strcmp(str, BENCH_FORMAT_SIMPLE_STR))
116 return BENCH_FORMAT_SIMPLE;
117
118 return BENCH_FORMAT_UNKNOWN;
119}
120
121int cmd_bench(int argc, const char **argv, const char *prefix __used)
122{
123 int i, j, status = 0;
124
125 if (argc < 2) {
126 /* No subsystem specified. */
127 print_usage();
128 goto end;
129 }
130
131 argc = parse_options(argc, argv, bench_options, bench_usage,
132 PARSE_OPT_STOP_AT_NON_OPTION);
133
134 bench_format = bench_str2int(bench_format_str);
135 if (bench_format == BENCH_FORMAT_UNKNOWN) {
136 printf("Unknown format descriptor:%s\n", bench_format_str);
137 goto end;
138 }
139
140 if (argc < 1) {
141 print_usage();
142 goto end;
143 }
144
145 for (i = 0; subsystems[i].name; i++) {
146 if (strcmp(subsystems[i].name, argv[0]))
147 continue;
148
149 if (argc < 2) {
150 /* No suite specified. */
151 dump_suites(i);
152 goto end;
153 }
154
155 for (j = 0; subsystems[i].suites[j].name; j++) {
156 if (strcmp(subsystems[i].suites[j].name, argv[1]))
157 continue;
158
159 if (bench_format == BENCH_FORMAT_DEFAULT)
160 printf("# Running %s/%s benchmark...\n",
161 subsystems[i].name,
162 subsystems[i].suites[j].name);
163 status = subsystems[i].suites[j].fn(argc - 1,
164 argv + 1, prefix);
165 goto end;
166 }
167
168 if (!strcmp(argv[1], "-h") || !strcmp(argv[1], "--help")) {
169 dump_suites(i);
170 goto end;
171 }
172
173 printf("Unknown suite:%s for %s\n", argv[1], argv[0]);
174 status = 1;
175 goto end;
176 }
177
178 printf("Unknown subsystem:%s\n", argv[0]);
179 status = 1;
180
181end:
182 return status;
183}
diff --git a/tools/perf/builtin.h b/tools/perf/builtin.h
index e11d8d231c3b..f0cd5b139b7c 100644
--- a/tools/perf/builtin.h
+++ b/tools/perf/builtin.h
@@ -15,6 +15,7 @@ extern int read_line_with_nul(char *buf, int size, FILE *file);
15extern int check_pager_config(const char *cmd); 15extern int check_pager_config(const char *cmd);
16 16
17extern int cmd_annotate(int argc, const char **argv, const char *prefix); 17extern int cmd_annotate(int argc, const char **argv, const char *prefix);
18extern int cmd_bench(int argc, const char **argv, const char *prefix);
18extern int cmd_help(int argc, const char **argv, const char *prefix); 19extern int cmd_help(int argc, const char **argv, const char *prefix);
19extern int cmd_sched(int argc, const char **argv, const char *prefix); 20extern int cmd_sched(int argc, const char **argv, const char *prefix);
20extern int cmd_list(int argc, const char **argv, const char *prefix); 21extern int cmd_list(int argc, const char **argv, const char *prefix);
diff --git a/tools/perf/command-list.txt b/tools/perf/command-list.txt
index 00326e230d87..981c40b9a5e2 100644
--- a/tools/perf/command-list.txt
+++ b/tools/perf/command-list.txt
@@ -3,6 +3,7 @@
3# command name category [deprecated] [common] 3# command name category [deprecated] [common]
4# 4#
5perf-annotate mainporcelain common 5perf-annotate mainporcelain common
6perf-bench mainporcelain common
6perf-list mainporcelain common 7perf-list mainporcelain common
7perf-sched mainporcelain common 8perf-sched mainporcelain common
8perf-record mainporcelain common 9perf-record mainporcelain common
diff --git a/tools/perf/design.txt b/tools/perf/design.txt
index fdd42a824c98..f000c30877ac 100644
--- a/tools/perf/design.txt
+++ b/tools/perf/design.txt
@@ -137,6 +137,8 @@ enum sw_event_ids {
137 PERF_COUNT_SW_CPU_MIGRATIONS = 4, 137 PERF_COUNT_SW_CPU_MIGRATIONS = 4,
138 PERF_COUNT_SW_PAGE_FAULTS_MIN = 5, 138 PERF_COUNT_SW_PAGE_FAULTS_MIN = 5,
139 PERF_COUNT_SW_PAGE_FAULTS_MAJ = 6, 139 PERF_COUNT_SW_PAGE_FAULTS_MAJ = 6,
140 PERF_COUNT_SW_ALIGNMENT_FAULTS = 7,
141 PERF_COUNT_SW_EMULATION_FAULTS = 8,
140}; 142};
141 143
142Counters of the type PERF_TYPE_TRACEPOINT are available when the ftrace event 144Counters of the type PERF_TYPE_TRACEPOINT are available when the ftrace event
diff --git a/tools/perf/perf.c b/tools/perf/perf.c
index 601f403fbdac..8936786b42ea 100644
--- a/tools/perf/perf.c
+++ b/tools/perf/perf.c
@@ -289,6 +289,7 @@ static void handle_internal_command(int argc, const char **argv)
289 { "list", cmd_list, 0 }, 289 { "list", cmd_list, 0 },
290 { "record", cmd_record, 0 }, 290 { "record", cmd_record, 0 },
291 { "report", cmd_report, 0 }, 291 { "report", cmd_report, 0 },
292 { "bench", cmd_bench, 0 },
292 { "stat", cmd_stat, 0 }, 293 { "stat", cmd_stat, 0 },
293 { "timechart", cmd_timechart, 0 }, 294 { "timechart", cmd_timechart, 0 },
294 { "top", cmd_top, 0 }, 295 { "top", cmd_top, 0 },
diff --git a/tools/perf/util/parse-events.c b/tools/perf/util/parse-events.c
index 097938a96d74..0faf4f2bb5ca 100644
--- a/tools/perf/util/parse-events.c
+++ b/tools/perf/util/parse-events.c
@@ -48,6 +48,8 @@ static struct event_symbol event_symbols[] = {
48 { CSW(PAGE_FAULTS_MAJ), "major-faults", "" }, 48 { CSW(PAGE_FAULTS_MAJ), "major-faults", "" },
49 { CSW(CONTEXT_SWITCHES), "context-switches", "cs" }, 49 { CSW(CONTEXT_SWITCHES), "context-switches", "cs" },
50 { CSW(CPU_MIGRATIONS), "cpu-migrations", "migrations" }, 50 { CSW(CPU_MIGRATIONS), "cpu-migrations", "migrations" },
51 { CSW(ALIGNMENT_FAULTS), "alignment-faults", "" },
52 { CSW(EMULATION_FAULTS), "emulation-faults", "" },
51}; 53};
52 54
53#define __PERF_EVENT_FIELD(config, name) \ 55#define __PERF_EVENT_FIELD(config, name) \
@@ -76,6 +78,8 @@ static const char *sw_event_names[] = {
76 "CPU-migrations", 78 "CPU-migrations",
77 "minor-faults", 79 "minor-faults",
78 "major-faults", 80 "major-faults",
81 "alignment-faults",
82 "emulation-faults",
79}; 83};
80 84
81#define MAX_ALIASES 8 85#define MAX_ALIASES 8
generated by cgit v1.2.2 (git 2.25.0) at 2024-11-23 06:57:54 -0500