aboutsummaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
Diffstat
-rw-r--r--arch/x86/include/asm/perf_event.h2
-rw-r--r--arch/x86/kernel/cpu/perf_event.h2
-rw-r--r--arch/x86/kernel/cpu/perf_event_intel.c92
-rw-r--r--arch/x86/kernel/cpu/perf_event_intel_uncore.c1316
-rw-r--r--arch/x86/kernel/cpu/perf_event_intel_uncore.h207
-rw-r--r--kernel/events/uprobes.c213
-rw-r--r--kernel/sched/core.c2
-rw-r--r--mm/mmap.c6
-rw-r--r--tools/lib/traceevent/.gitignore1
-rw-r--r--tools/lib/traceevent/Makefile14
-rw-r--r--tools/perf/Makefile4
-rw-r--r--tools/perf/builtin-test.c4
-rw-r--r--tools/perf/builtin-top.c2
-rw-r--r--tools/perf/ui/browsers/hists.c4
-rw-r--r--tools/perf/util/annotate.c15
-rw-r--r--tools/perf/util/dso-test-data.c153
-rw-r--r--tools/perf/util/evlist.c2
-rw-r--r--tools/perf/util/header.c6
-rw-r--r--tools/perf/util/hist.c7
-rw-r--r--tools/perf/util/map.c41
-rw-r--r--tools/perf/util/map.h1
-rw-r--r--tools/perf/util/parse-events.c2
-rw-r--r--tools/perf/util/session.c5
-rw-r--r--tools/perf/util/symbol.c447
-rw-r--r--tools/perf/util/symbol.h54
-rw-r--r--tools/perf/util/target.c11
26 files changed, 2228 insertions, 385 deletions
diff --git a/arch/x86/include/asm/perf_event.h b/arch/x86/include/asm/perf_event.h
index c78f14a0df00..dab39350e51e 100644
--- a/arch/x86/include/asm/perf_event.h
+++ b/arch/x86/include/asm/perf_event.h
@@ -234,7 +234,7 @@ extern struct perf_guest_switch_msr *perf_guest_get_msrs(int *nr);
234extern void perf_get_x86_pmu_capability(struct x86_pmu_capability *cap); 234extern void perf_get_x86_pmu_capability(struct x86_pmu_capability *cap);
235extern void perf_check_microcode(void); 235extern void perf_check_microcode(void);
236#else 236#else
237static inline perf_guest_switch_msr *perf_guest_get_msrs(int *nr) 237static inline struct perf_guest_switch_msr *perf_guest_get_msrs(int *nr)
238{ 238{
239 *nr = 0; 239 *nr = 0;
240 return NULL; 240 return NULL;
diff --git a/arch/x86/kernel/cpu/perf_event.h b/arch/x86/kernel/cpu/perf_event.h
index a15df4be151f..821d53b696d1 100644
--- a/arch/x86/kernel/cpu/perf_event.h
+++ b/arch/x86/kernel/cpu/perf_event.h
@@ -374,7 +374,7 @@ struct x86_pmu {
374 /* 374 /*
375 * Intel DebugStore bits 375 * Intel DebugStore bits
376 */ 376 */
377 int bts :1, 377 unsigned int bts :1,
378 bts_active :1, 378 bts_active :1,
379 pebs :1, 379 pebs :1,
380 pebs_active :1, 380 pebs_active :1,
diff --git a/arch/x86/kernel/cpu/perf_event_intel.c b/arch/x86/kernel/cpu/perf_event_intel.c
index 7a8b9d0abcaa..382366977d4c 100644
--- a/arch/x86/kernel/cpu/perf_event_intel.c
+++ b/arch/x86/kernel/cpu/perf_event_intel.c
@@ -138,6 +138,84 @@ static u64 intel_pmu_event_map(int hw_event)
138 return intel_perfmon_event_map[hw_event]; 138 return intel_perfmon_event_map[hw_event];
139} 139}
140 140
141#define SNB_DMND_DATA_RD (1ULL << 0)
142#define SNB_DMND_RFO (1ULL << 1)
143#define SNB_DMND_IFETCH (1ULL << 2)
144#define SNB_DMND_WB (1ULL << 3)
145#define SNB_PF_DATA_RD (1ULL << 4)
146#define SNB_PF_RFO (1ULL << 5)
147#define SNB_PF_IFETCH (1ULL << 6)
148#define SNB_LLC_DATA_RD (1ULL << 7)
149#define SNB_LLC_RFO (1ULL << 8)
150#define SNB_LLC_IFETCH (1ULL << 9)
151#define SNB_BUS_LOCKS (1ULL << 10)
152#define SNB_STRM_ST (1ULL << 11)
153#define SNB_OTHER (1ULL << 15)
154#define SNB_RESP_ANY (1ULL << 16)
155#define SNB_NO_SUPP (1ULL << 17)
156#define SNB_LLC_HITM (1ULL << 18)
157#define SNB_LLC_HITE (1ULL << 19)
158#define SNB_LLC_HITS (1ULL << 20)
159#define SNB_LLC_HITF (1ULL << 21)
160#define SNB_LOCAL (1ULL << 22)
161#define SNB_REMOTE (0xffULL << 23)
162#define SNB_SNP_NONE (1ULL << 31)
163#define SNB_SNP_NOT_NEEDED (1ULL << 32)
164#define SNB_SNP_MISS (1ULL << 33)
165#define SNB_NO_FWD (1ULL << 34)
166#define SNB_SNP_FWD (1ULL << 35)
167#define SNB_HITM (1ULL << 36)
168#define SNB_NON_DRAM (1ULL << 37)
169
170#define SNB_DMND_READ (SNB_DMND_DATA_RD|SNB_LLC_DATA_RD)
171#define SNB_DMND_WRITE (SNB_DMND_RFO|SNB_LLC_RFO)
172#define SNB_DMND_PREFETCH (SNB_PF_DATA_RD|SNB_PF_RFO)
173
174#define SNB_SNP_ANY (SNB_SNP_NONE|SNB_SNP_NOT_NEEDED| \
175 SNB_SNP_MISS|SNB_NO_FWD|SNB_SNP_FWD| \
176 SNB_HITM)
177
178#define SNB_DRAM_ANY (SNB_LOCAL|SNB_REMOTE|SNB_SNP_ANY)
179#define SNB_DRAM_REMOTE (SNB_REMOTE|SNB_SNP_ANY)
180
181#define SNB_L3_ACCESS SNB_RESP_ANY
182#define SNB_L3_MISS (SNB_DRAM_ANY|SNB_NON_DRAM)
183
184static __initconst const u64 snb_hw_cache_extra_regs
185 [PERF_COUNT_HW_CACHE_MAX]
186 [PERF_COUNT_HW_CACHE_OP_MAX]
187 [PERF_COUNT_HW_CACHE_RESULT_MAX] =
188{
189 [ C(LL ) ] = {
190 [ C(OP_READ) ] = {
191 [ C(RESULT_ACCESS) ] = SNB_DMND_READ|SNB_L3_ACCESS,
192 [ C(RESULT_MISS) ] = SNB_DMND_READ|SNB_L3_MISS,
193 },
194 [ C(OP_WRITE) ] = {
195 [ C(RESULT_ACCESS) ] = SNB_DMND_WRITE|SNB_L3_ACCESS,
196 [ C(RESULT_MISS) ] = SNB_DMND_WRITE|SNB_L3_MISS,
197 },
198 [ C(OP_PREFETCH) ] = {
199 [ C(RESULT_ACCESS) ] = SNB_DMND_PREFETCH|SNB_L3_ACCESS,
200 [ C(RESULT_MISS) ] = SNB_DMND_PREFETCH|SNB_L3_MISS,
201 },
202 },
203 [ C(NODE) ] = {
204 [ C(OP_READ) ] = {
205 [ C(RESULT_ACCESS) ] = SNB_DMND_READ|SNB_DRAM_ANY,
206 [ C(RESULT_MISS) ] = SNB_DMND_READ|SNB_DRAM_REMOTE,
207 },
208 [ C(OP_WRITE) ] = {
209 [ C(RESULT_ACCESS) ] = SNB_DMND_WRITE|SNB_DRAM_ANY,
210 [ C(RESULT_MISS) ] = SNB_DMND_WRITE|SNB_DRAM_REMOTE,
211 },
212 [ C(OP_PREFETCH) ] = {
213 [ C(RESULT_ACCESS) ] = SNB_DMND_PREFETCH|SNB_DRAM_ANY,
214 [ C(RESULT_MISS) ] = SNB_DMND_PREFETCH|SNB_DRAM_REMOTE,
215 },
216 },
217};
218
141static __initconst const u64 snb_hw_cache_event_ids 219static __initconst const u64 snb_hw_cache_event_ids
142 [PERF_COUNT_HW_CACHE_MAX] 220 [PERF_COUNT_HW_CACHE_MAX]
143 [PERF_COUNT_HW_CACHE_OP_MAX] 221 [PERF_COUNT_HW_CACHE_OP_MAX]
@@ -235,16 +313,16 @@ static __initconst const u64 snb_hw_cache_event_ids
235 }, 313 },
236 [ C(NODE) ] = { 314 [ C(NODE) ] = {
237 [ C(OP_READ) ] = { 315 [ C(OP_READ) ] = {
238 [ C(RESULT_ACCESS) ] = -1, 316 [ C(RESULT_ACCESS) ] = 0x01b7,
239 [ C(RESULT_MISS) ] = -1, 317 [ C(RESULT_MISS) ] = 0x01b7,
240 }, 318 },
241 [ C(OP_WRITE) ] = { 319 [ C(OP_WRITE) ] = {
242 [ C(RESULT_ACCESS) ] = -1, 320 [ C(RESULT_ACCESS) ] = 0x01b7,
243 [ C(RESULT_MISS) ] = -1, 321 [ C(RESULT_MISS) ] = 0x01b7,
244 }, 322 },
245 [ C(OP_PREFETCH) ] = { 323 [ C(OP_PREFETCH) ] = {
246 [ C(RESULT_ACCESS) ] = -1, 324 [ C(RESULT_ACCESS) ] = 0x01b7,
247 [ C(RESULT_MISS) ] = -1, 325 [ C(RESULT_MISS) ] = 0x01b7,
248 }, 326 },
249 }, 327 },
250 328
@@ -1964,6 +2042,8 @@ __init int intel_pmu_init(void)
1964 case 58: /* IvyBridge */ 2042 case 58: /* IvyBridge */
1965 memcpy(hw_cache_event_ids, snb_hw_cache_event_ids, 2043 memcpy(hw_cache_event_ids, snb_hw_cache_event_ids,
1966 sizeof(hw_cache_event_ids)); 2044 sizeof(hw_cache_event_ids));
2045 memcpy(hw_cache_extra_regs, snb_hw_cache_extra_regs,
2046 sizeof(hw_cache_extra_regs));
1967 2047
1968 intel_pmu_lbr_init_snb(); 2048 intel_pmu_lbr_init_snb();
1969 2049
diff --git a/arch/x86/kernel/cpu/perf_event_intel_uncore.c b/arch/x86/kernel/cpu/perf_event_intel_uncore.c
index 19faffc60886..7563fda9f033 100644
--- a/arch/x86/kernel/cpu/perf_event_intel_uncore.c
+++ b/arch/x86/kernel/cpu/perf_event_intel_uncore.c
@@ -18,6 +18,7 @@ static struct event_constraint constraint_empty =
18 EVENT_CONSTRAINT(0, 0, 0); 18 EVENT_CONSTRAINT(0, 0, 0);
19 19
20DEFINE_UNCORE_FORMAT_ATTR(event, event, "config:0-7"); 20DEFINE_UNCORE_FORMAT_ATTR(event, event, "config:0-7");
21DEFINE_UNCORE_FORMAT_ATTR(event_ext, event, "config:0-7,21");
21DEFINE_UNCORE_FORMAT_ATTR(umask, umask, "config:8-15"); 22DEFINE_UNCORE_FORMAT_ATTR(umask, umask, "config:8-15");
22DEFINE_UNCORE_FORMAT_ATTR(edge, edge, "config:18"); 23DEFINE_UNCORE_FORMAT_ATTR(edge, edge, "config:18");
23DEFINE_UNCORE_FORMAT_ATTR(tid_en, tid_en, "config:19"); 24DEFINE_UNCORE_FORMAT_ATTR(tid_en, tid_en, "config:19");
@@ -33,10 +34,81 @@ DEFINE_UNCORE_FORMAT_ATTR(filter_tid, filter_tid, "config1:0-4");
33DEFINE_UNCORE_FORMAT_ATTR(filter_nid, filter_nid, "config1:10-17"); 34DEFINE_UNCORE_FORMAT_ATTR(filter_nid, filter_nid, "config1:10-17");
34DEFINE_UNCORE_FORMAT_ATTR(filter_state, filter_state, "config1:18-22"); 35DEFINE_UNCORE_FORMAT_ATTR(filter_state, filter_state, "config1:18-22");
35DEFINE_UNCORE_FORMAT_ATTR(filter_opc, filter_opc, "config1:23-31"); 36DEFINE_UNCORE_FORMAT_ATTR(filter_opc, filter_opc, "config1:23-31");
36DEFINE_UNCORE_FORMAT_ATTR(filter_brand0, filter_brand0, "config1:0-7"); 37DEFINE_UNCORE_FORMAT_ATTR(filter_band0, filter_band0, "config1:0-7");
37DEFINE_UNCORE_FORMAT_ATTR(filter_brand1, filter_brand1, "config1:8-15"); 38DEFINE_UNCORE_FORMAT_ATTR(filter_band1, filter_band1, "config1:8-15");
38DEFINE_UNCORE_FORMAT_ATTR(filter_brand2, filter_brand2, "config1:16-23"); 39DEFINE_UNCORE_FORMAT_ATTR(filter_band2, filter_band2, "config1:16-23");
39DEFINE_UNCORE_FORMAT_ATTR(filter_brand3, filter_brand3, "config1:24-31"); 40DEFINE_UNCORE_FORMAT_ATTR(filter_band3, filter_band3, "config1:24-31");
41
42static u64 uncore_msr_read_counter(struct intel_uncore_box *box, struct perf_event *event)
43{
44 u64 count;
45
46 rdmsrl(event->hw.event_base, count);
47
48 return count;
49}
50
51/*
52 * generic get constraint function for shared match/mask registers.
53 */
54static struct event_constraint *
55uncore_get_constraint(struct intel_uncore_box *box, struct perf_event *event)
56{
57 struct intel_uncore_extra_reg *er;
58 struct hw_perf_event_extra *reg1 = &event->hw.extra_reg;
59 struct hw_perf_event_extra *reg2 = &event->hw.branch_reg;
60 unsigned long flags;
61 bool ok = false;
62
63 /*
64 * reg->alloc can be set due to existing state, so for fake box we
65 * need to ignore this, otherwise we might fail to allocate proper
66 * fake state for this extra reg constraint.
67 */
68 if (reg1->idx == EXTRA_REG_NONE ||
69 (!uncore_box_is_fake(box) && reg1->alloc))
70 return NULL;
71
72 er = &box->shared_regs[reg1->idx];
73 raw_spin_lock_irqsave(&er->lock, flags);
74 if (!atomic_read(&er->ref) ||
75 (er->config1 == reg1->config && er->config2 == reg2->config)) {
76 atomic_inc(&er->ref);
77 er->config1 = reg1->config;
78 er->config2 = reg2->config;
79 ok = true;
80 }
81 raw_spin_unlock_irqrestore(&er->lock, flags);
82
83 if (ok) {
84 if (!uncore_box_is_fake(box))
85 reg1->alloc = 1;
86 return NULL;
87 }
88
89 return &constraint_empty;
90}
91
92static void uncore_put_constraint(struct intel_uncore_box *box, struct perf_event *event)
93{
94 struct intel_uncore_extra_reg *er;
95 struct hw_perf_event_extra *reg1 = &event->hw.extra_reg;
96
97 /*
98 * Only put constraint if extra reg was actually allocated. Also
99 * takes care of event which do not use an extra shared reg.
100 *
101 * Also, if this is a fake box we shouldn't touch any event state
102 * (reg->alloc) and we don't care about leaving inconsistent box
103 * state either since it will be thrown out.
104 */
105 if (uncore_box_is_fake(box) || !reg1->alloc)
106 return;
107
108 er = &box->shared_regs[reg1->idx];
109 atomic_dec(&er->ref);
110 reg1->alloc = 0;
111}
40 112
41/* Sandy Bridge-EP uncore support */ 113/* Sandy Bridge-EP uncore support */
42static struct intel_uncore_type snbep_uncore_cbox; 114static struct intel_uncore_type snbep_uncore_cbox;
@@ -64,18 +136,15 @@ static void snbep_uncore_pci_enable_box(struct intel_uncore_box *box)
64 pci_write_config_dword(pdev, box_ctl, config); 136 pci_write_config_dword(pdev, box_ctl, config);
65} 137}
66 138
67static void snbep_uncore_pci_enable_event(struct intel_uncore_box *box, 139static void snbep_uncore_pci_enable_event(struct intel_uncore_box *box, struct perf_event *event)
68 struct perf_event *event)
69{ 140{
70 struct pci_dev *pdev = box->pci_dev; 141 struct pci_dev *pdev = box->pci_dev;
71 struct hw_perf_event *hwc = &event->hw; 142 struct hw_perf_event *hwc = &event->hw;
72 143
73 pci_write_config_dword(pdev, hwc->config_base, hwc->config | 144 pci_write_config_dword(pdev, hwc->config_base, hwc->config | SNBEP_PMON_CTL_EN);
74 SNBEP_PMON_CTL_EN);
75} 145}
76 146
77static void snbep_uncore_pci_disable_event(struct intel_uncore_box *box, 147static void snbep_uncore_pci_disable_event(struct intel_uncore_box *box, struct perf_event *event)
78 struct perf_event *event)
79{ 148{
80 struct pci_dev *pdev = box->pci_dev; 149 struct pci_dev *pdev = box->pci_dev;
81 struct hw_perf_event *hwc = &event->hw; 150 struct hw_perf_event *hwc = &event->hw;
@@ -83,8 +152,7 @@ static void snbep_uncore_pci_disable_event(struct intel_uncore_box *box,
83 pci_write_config_dword(pdev, hwc->config_base, hwc->config); 152 pci_write_config_dword(pdev, hwc->config_base, hwc->config);
84} 153}
85 154
86static u64 snbep_uncore_pci_read_counter(struct intel_uncore_box *box, 155static u64 snbep_uncore_pci_read_counter(struct intel_uncore_box *box, struct perf_event *event)
87 struct perf_event *event)
88{ 156{
89 struct pci_dev *pdev = box->pci_dev; 157 struct pci_dev *pdev = box->pci_dev;
90 struct hw_perf_event *hwc = &event->hw; 158 struct hw_perf_event *hwc = &event->hw;
@@ -92,14 +160,15 @@ static u64 snbep_uncore_pci_read_counter(struct intel_uncore_box *box,
92 160
93 pci_read_config_dword(pdev, hwc->event_base, (u32 *)&count); 161 pci_read_config_dword(pdev, hwc->event_base, (u32 *)&count);
94 pci_read_config_dword(pdev, hwc->event_base + 4, (u32 *)&count + 1); 162 pci_read_config_dword(pdev, hwc->event_base + 4, (u32 *)&count + 1);
163
95 return count; 164 return count;
96} 165}
97 166
98static void snbep_uncore_pci_init_box(struct intel_uncore_box *box) 167static void snbep_uncore_pci_init_box(struct intel_uncore_box *box)
99{ 168{
100 struct pci_dev *pdev = box->pci_dev; 169 struct pci_dev *pdev = box->pci_dev;
101 pci_write_config_dword(pdev, SNBEP_PCI_PMON_BOX_CTL, 170
102 SNBEP_PMON_BOX_CTL_INT); 171 pci_write_config_dword(pdev, SNBEP_PCI_PMON_BOX_CTL, SNBEP_PMON_BOX_CTL_INT);
103} 172}
104 173
105static void snbep_uncore_msr_disable_box(struct intel_uncore_box *box) 174static void snbep_uncore_msr_disable_box(struct intel_uncore_box *box)
@@ -112,7 +181,6 @@ static void snbep_uncore_msr_disable_box(struct intel_uncore_box *box)
112 rdmsrl(msr, config); 181 rdmsrl(msr, config);
113 config |= SNBEP_PMON_BOX_CTL_FRZ; 182 config |= SNBEP_PMON_BOX_CTL_FRZ;
114 wrmsrl(msr, config); 183 wrmsrl(msr, config);
115 return;
116 } 184 }
117} 185}
118 186
@@ -126,12 +194,10 @@ static void snbep_uncore_msr_enable_box(struct intel_uncore_box *box)
126 rdmsrl(msr, config); 194 rdmsrl(msr, config);
127 config &= ~SNBEP_PMON_BOX_CTL_FRZ; 195 config &= ~SNBEP_PMON_BOX_CTL_FRZ;
128 wrmsrl(msr, config); 196 wrmsrl(msr, config);
129 return;
130 } 197 }
131} 198}
132 199
133static void snbep_uncore_msr_enable_event(struct intel_uncore_box *box, 200static void snbep_uncore_msr_enable_event(struct intel_uncore_box *box, struct perf_event *event)
134 struct perf_event *event)
135{ 201{
136 struct hw_perf_event *hwc = &event->hw; 202 struct hw_perf_event *hwc = &event->hw;
137 struct hw_perf_event_extra *reg1 = &hwc->extra_reg; 203 struct hw_perf_event_extra *reg1 = &hwc->extra_reg;
@@ -150,68 +216,15 @@ static void snbep_uncore_msr_disable_event(struct intel_uncore_box *box,
150 wrmsrl(hwc->config_base, hwc->config); 216 wrmsrl(hwc->config_base, hwc->config);
151} 217}
152 218
153static u64 snbep_uncore_msr_read_counter(struct intel_uncore_box *box,
154 struct perf_event *event)
155{
156 struct hw_perf_event *hwc = &event->hw;
157 u64 count;
158
159 rdmsrl(hwc->event_base, count);
160 return count;
161}
162
163static void snbep_uncore_msr_init_box(struct intel_uncore_box *box) 219static void snbep_uncore_msr_init_box(struct intel_uncore_box *box)
164{ 220{
165 unsigned msr = uncore_msr_box_ctl(box); 221 unsigned msr = uncore_msr_box_ctl(box);
222
166 if (msr) 223 if (msr)
167 wrmsrl(msr, SNBEP_PMON_BOX_CTL_INT); 224 wrmsrl(msr, SNBEP_PMON_BOX_CTL_INT);
168} 225}
169 226
170static struct event_constraint * 227static int snbep_uncore_hw_config(struct intel_uncore_box *box, struct perf_event *event)
171snbep_uncore_get_constraint(struct intel_uncore_box *box,
172 struct perf_event *event)
173{
174 struct intel_uncore_extra_reg *er;
175 struct hw_perf_event_extra *reg1 = &event->hw.extra_reg;
176 unsigned long flags;
177 bool ok = false;
178
179 if (reg1->idx == EXTRA_REG_NONE || (box->phys_id >= 0 && reg1->alloc))
180 return NULL;
181
182 er = &box->shared_regs[reg1->idx];
183 raw_spin_lock_irqsave(&er->lock, flags);
184 if (!atomic_read(&er->ref) || er->config1 == reg1->config) {
185 atomic_inc(&er->ref);
186 er->config1 = reg1->config;
187 ok = true;
188 }
189 raw_spin_unlock_irqrestore(&er->lock, flags);
190
191 if (ok) {
192 if (box->phys_id >= 0)
193 reg1->alloc = 1;
194 return NULL;
195 }
196 return &constraint_empty;
197}
198
199static void snbep_uncore_put_constraint(struct intel_uncore_box *box,
200 struct perf_event *event)
201{
202 struct intel_uncore_extra_reg *er;
203 struct hw_perf_event_extra *reg1 = &event->hw.extra_reg;
204
205 if (box->phys_id < 0 || !reg1->alloc)
206 return;
207
208 er = &box->shared_regs[reg1->idx];
209 atomic_dec(&er->ref);
210 reg1->alloc = 0;
211}
212
213static int snbep_uncore_hw_config(struct intel_uncore_box *box,
214 struct perf_event *event)
215{ 228{
216 struct hw_perf_event *hwc = &event->hw; 229 struct hw_perf_event *hwc = &event->hw;
217 struct hw_perf_event_extra *reg1 = &hwc->extra_reg; 230 struct hw_perf_event_extra *reg1 = &hwc->extra_reg;
@@ -221,14 +234,16 @@ static int snbep_uncore_hw_config(struct intel_uncore_box *box,
221 SNBEP_CBO_MSR_OFFSET * box->pmu->pmu_idx; 234 SNBEP_CBO_MSR_OFFSET * box->pmu->pmu_idx;
222 reg1->config = event->attr.config1 & 235 reg1->config = event->attr.config1 &
223 SNBEP_CB0_MSR_PMON_BOX_FILTER_MASK; 236 SNBEP_CB0_MSR_PMON_BOX_FILTER_MASK;
224 } else if (box->pmu->type == &snbep_uncore_pcu) {
225 reg1->reg = SNBEP_PCU_MSR_PMON_BOX_FILTER;
226 reg1->config = event->attr.config1 &
227 SNBEP_PCU_MSR_PMON_BOX_FILTER_MASK;
228 } else { 237 } else {
229 return 0; 238 if (box->pmu->type == &snbep_uncore_pcu) {
239 reg1->reg = SNBEP_PCU_MSR_PMON_BOX_FILTER;
240 reg1->config = event->attr.config1 & SNBEP_PCU_MSR_PMON_BOX_FILTER_MASK;
241 } else {
242 return 0;
243 }
230 } 244 }
231 reg1->idx = 0; 245 reg1->idx = 0;
246
232 return 0; 247 return 0;
233} 248}
234 249
@@ -272,10 +287,19 @@ static struct attribute *snbep_uncore_pcu_formats_attr[] = {
272 &format_attr_thresh5.attr, 287 &format_attr_thresh5.attr,
273 &format_attr_occ_invert.attr, 288 &format_attr_occ_invert.attr,
274 &format_attr_occ_edge.attr, 289 &format_attr_occ_edge.attr,
275 &format_attr_filter_brand0.attr, 290 &format_attr_filter_band0.attr,
276 &format_attr_filter_brand1.attr, 291 &format_attr_filter_band1.attr,
277 &format_attr_filter_brand2.attr, 292 &format_attr_filter_band2.attr,
278 &format_attr_filter_brand3.attr, 293 &format_attr_filter_band3.attr,
294 NULL,
295};
296
297static struct attribute *snbep_uncore_qpi_formats_attr[] = {
298 &format_attr_event_ext.attr,
299 &format_attr_umask.attr,
300 &format_attr_edge.attr,
301 &format_attr_inv.attr,
302 &format_attr_thresh8.attr,
279 NULL, 303 NULL,
280}; 304};
281 305
@@ -314,15 +338,20 @@ static struct attribute_group snbep_uncore_pcu_format_group = {
314 .attrs = snbep_uncore_pcu_formats_attr, 338 .attrs = snbep_uncore_pcu_formats_attr,
315}; 339};
316 340
341static struct attribute_group snbep_uncore_qpi_format_group = {
342 .name = "format",
343 .attrs = snbep_uncore_qpi_formats_attr,
344};
345
317static struct intel_uncore_ops snbep_uncore_msr_ops = { 346static struct intel_uncore_ops snbep_uncore_msr_ops = {
318 .init_box = snbep_uncore_msr_init_box, 347 .init_box = snbep_uncore_msr_init_box,
319 .disable_box = snbep_uncore_msr_disable_box, 348 .disable_box = snbep_uncore_msr_disable_box,
320 .enable_box = snbep_uncore_msr_enable_box, 349 .enable_box = snbep_uncore_msr_enable_box,
321 .disable_event = snbep_uncore_msr_disable_event, 350 .disable_event = snbep_uncore_msr_disable_event,
322 .enable_event = snbep_uncore_msr_enable_event, 351 .enable_event = snbep_uncore_msr_enable_event,
323 .read_counter = snbep_uncore_msr_read_counter, 352 .read_counter = uncore_msr_read_counter,
324 .get_constraint = snbep_uncore_get_constraint, 353 .get_constraint = uncore_get_constraint,
325 .put_constraint = snbep_uncore_put_constraint, 354 .put_constraint = uncore_put_constraint,
326 .hw_config = snbep_uncore_hw_config, 355 .hw_config = snbep_uncore_hw_config,
327}; 356};
328 357
@@ -485,8 +514,13 @@ static struct intel_uncore_type snbep_uncore_qpi = {
485 .num_counters = 4, 514 .num_counters = 4,
486 .num_boxes = 2, 515 .num_boxes = 2,
487 .perf_ctr_bits = 48, 516 .perf_ctr_bits = 48,
517 .perf_ctr = SNBEP_PCI_PMON_CTR0,
518 .event_ctl = SNBEP_PCI_PMON_CTL0,
519 .event_mask = SNBEP_QPI_PCI_PMON_RAW_EVENT_MASK,
520 .box_ctl = SNBEP_PCI_PMON_BOX_CTL,
521 .ops = &snbep_uncore_pci_ops,
488 .event_descs = snbep_uncore_qpi_events, 522 .event_descs = snbep_uncore_qpi_events,
489 SNBEP_UNCORE_PCI_COMMON_INIT(), 523 .format_group = &snbep_uncore_qpi_format_group,
490}; 524};
491 525
492 526
@@ -603,10 +637,8 @@ static void snbep_pci2phy_map_init(void)
603} 637}
604/* end of Sandy Bridge-EP uncore support */ 638/* end of Sandy Bridge-EP uncore support */
605 639
606
607/* Sandy Bridge uncore support */ 640/* Sandy Bridge uncore support */
608static void snb_uncore_msr_enable_event(struct intel_uncore_box *box, 641static void snb_uncore_msr_enable_event(struct intel_uncore_box *box, struct perf_event *event)
609 struct perf_event *event)
610{ 642{
611 struct hw_perf_event *hwc = &event->hw; 643 struct hw_perf_event *hwc = &event->hw;
612 644
@@ -616,20 +648,11 @@ static void snb_uncore_msr_enable_event(struct intel_uncore_box *box,
616 wrmsrl(hwc->config_base, SNB_UNC_CTL_EN); 648 wrmsrl(hwc->config_base, SNB_UNC_CTL_EN);
617} 649}
618 650
619static void snb_uncore_msr_disable_event(struct intel_uncore_box *box, 651static void snb_uncore_msr_disable_event(struct intel_uncore_box *box, struct perf_event *event)
620 struct perf_event *event)
621{ 652{
622 wrmsrl(event->hw.config_base, 0); 653 wrmsrl(event->hw.config_base, 0);
623} 654}
624 655
625static u64 snb_uncore_msr_read_counter(struct intel_uncore_box *box,
626 struct perf_event *event)
627{
628 u64 count;
629 rdmsrl(event->hw.event_base, count);
630 return count;
631}
632
633static void snb_uncore_msr_init_box(struct intel_uncore_box *box) 656static void snb_uncore_msr_init_box(struct intel_uncore_box *box)
634{ 657{
635 if (box->pmu->pmu_idx == 0) { 658 if (box->pmu->pmu_idx == 0) {
@@ -648,15 +671,15 @@ static struct attribute *snb_uncore_formats_attr[] = {
648}; 671};
649 672
650static struct attribute_group snb_uncore_format_group = { 673static struct attribute_group snb_uncore_format_group = {
651 .name = "format", 674 .name = "format",
652 .attrs = snb_uncore_formats_attr, 675 .attrs = snb_uncore_formats_attr,
653}; 676};
654 677
655static struct intel_uncore_ops snb_uncore_msr_ops = { 678static struct intel_uncore_ops snb_uncore_msr_ops = {
656 .init_box = snb_uncore_msr_init_box, 679 .init_box = snb_uncore_msr_init_box,
657 .disable_event = snb_uncore_msr_disable_event, 680 .disable_event = snb_uncore_msr_disable_event,
658 .enable_event = snb_uncore_msr_enable_event, 681 .enable_event = snb_uncore_msr_enable_event,
659 .read_counter = snb_uncore_msr_read_counter, 682 .read_counter = uncore_msr_read_counter,
660}; 683};
661 684
662static struct event_constraint snb_uncore_cbox_constraints[] = { 685static struct event_constraint snb_uncore_cbox_constraints[] = {
@@ -697,12 +720,10 @@ static void nhm_uncore_msr_disable_box(struct intel_uncore_box *box)
697 720
698static void nhm_uncore_msr_enable_box(struct intel_uncore_box *box) 721static void nhm_uncore_msr_enable_box(struct intel_uncore_box *box)
699{ 722{
700 wrmsrl(NHM_UNC_PERF_GLOBAL_CTL, 723 wrmsrl(NHM_UNC_PERF_GLOBAL_CTL, NHM_UNC_GLOBAL_CTL_EN_PC_ALL | NHM_UNC_GLOBAL_CTL_EN_FC);
701 NHM_UNC_GLOBAL_CTL_EN_PC_ALL | NHM_UNC_GLOBAL_CTL_EN_FC);
702} 724}
703 725
704static void nhm_uncore_msr_enable_event(struct intel_uncore_box *box, 726static void nhm_uncore_msr_enable_event(struct intel_uncore_box *box, struct perf_event *event)
705 struct perf_event *event)
706{ 727{
707 struct hw_perf_event *hwc = &event->hw; 728 struct hw_perf_event *hwc = &event->hw;
708 729
@@ -744,7 +765,7 @@ static struct intel_uncore_ops nhm_uncore_msr_ops = {
744 .enable_box = nhm_uncore_msr_enable_box, 765 .enable_box = nhm_uncore_msr_enable_box,
745 .disable_event = snb_uncore_msr_disable_event, 766 .disable_event = snb_uncore_msr_disable_event,
746 .enable_event = nhm_uncore_msr_enable_event, 767 .enable_event = nhm_uncore_msr_enable_event,
747 .read_counter = snb_uncore_msr_read_counter, 768 .read_counter = uncore_msr_read_counter,
748}; 769};
749 770
750static struct intel_uncore_type nhm_uncore = { 771static struct intel_uncore_type nhm_uncore = {
@@ -769,8 +790,1041 @@ static struct intel_uncore_type *nhm_msr_uncores[] = {
769}; 790};
770/* end of Nehalem uncore support */ 791/* end of Nehalem uncore support */
771 792
772static void uncore_assign_hw_event(struct intel_uncore_box *box, 793/* Nehalem-EX uncore support */
773 struct perf_event *event, int idx) 794#define __BITS_VALUE(x, i, n) ((typeof(x))(((x) >> ((i) * (n))) & \
795 ((1ULL << (n)) - 1)))
796
797DEFINE_UNCORE_FORMAT_ATTR(event5, event, "config:1-5");
798DEFINE_UNCORE_FORMAT_ATTR(counter, counter, "config:6-7");
799DEFINE_UNCORE_FORMAT_ATTR(mm_cfg, mm_cfg, "config:63");
800DEFINE_UNCORE_FORMAT_ATTR(match, match, "config1:0-63");
801DEFINE_UNCORE_FORMAT_ATTR(mask, mask, "config2:0-63");
802
803static void nhmex_uncore_msr_init_box(struct intel_uncore_box *box)
804{
805 wrmsrl(NHMEX_U_MSR_PMON_GLOBAL_CTL, NHMEX_U_PMON_GLOBAL_EN_ALL);
806}
807
808static void nhmex_uncore_msr_disable_box(struct intel_uncore_box *box)
809{
810 unsigned msr = uncore_msr_box_ctl(box);
811 u64 config;
812
813 if (msr) {
814 rdmsrl(msr, config);
815 config &= ~((1ULL << uncore_num_counters(box)) - 1);
816 /* WBox has a fixed counter */
817 if (uncore_msr_fixed_ctl(box))
818 config &= ~NHMEX_W_PMON_GLOBAL_FIXED_EN;
819 wrmsrl(msr, config);
820 }
821}
822
823static void nhmex_uncore_msr_enable_box(struct intel_uncore_box *box)
824{
825 unsigned msr = uncore_msr_box_ctl(box);
826 u64 config;
827
828 if (msr) {
829 rdmsrl(msr, config);
830 config |= (1ULL << uncore_num_counters(box)) - 1;
831 /* WBox has a fixed counter */
832 if (uncore_msr_fixed_ctl(box))
833 config |= NHMEX_W_PMON_GLOBAL_FIXED_EN;
834 wrmsrl(msr, config);
835 }
836}
837
838static void nhmex_uncore_msr_disable_event(struct intel_uncore_box *box, struct perf_event *event)
839{
840 wrmsrl(event->hw.config_base, 0);
841}
842
843static void nhmex_uncore_msr_enable_event(struct intel_uncore_box *box, struct perf_event *event)
844{
845 struct hw_perf_event *hwc = &event->hw;
846
847 if (hwc->idx >= UNCORE_PMC_IDX_FIXED)
848 wrmsrl(hwc->config_base, NHMEX_PMON_CTL_EN_BIT0);
849 else if (box->pmu->type->event_mask & NHMEX_PMON_CTL_EN_BIT0)
850 wrmsrl(hwc->config_base, hwc->config | NHMEX_PMON_CTL_EN_BIT22);
851 else
852 wrmsrl(hwc->config_base, hwc->config | NHMEX_PMON_CTL_EN_BIT0);
853}
854
855#define NHMEX_UNCORE_OPS_COMMON_INIT() \
856 .init_box = nhmex_uncore_msr_init_box, \
857 .disable_box = nhmex_uncore_msr_disable_box, \
858 .enable_box = nhmex_uncore_msr_enable_box, \
859 .disable_event = nhmex_uncore_msr_disable_event, \
860 .read_counter = uncore_msr_read_counter
861
862static struct intel_uncore_ops nhmex_uncore_ops = {
863 NHMEX_UNCORE_OPS_COMMON_INIT(),
864 .enable_event = nhmex_uncore_msr_enable_event,
865};
866
867static struct attribute *nhmex_uncore_ubox_formats_attr[] = {
868 &format_attr_event.attr,
869 &format_attr_edge.attr,
870 NULL,
871};
872
873static struct attribute_group nhmex_uncore_ubox_format_group = {
874 .name = "format",
875 .attrs = nhmex_uncore_ubox_formats_attr,
876};
877
878static struct intel_uncore_type nhmex_uncore_ubox = {
879 .name = "ubox",
880 .num_counters = 1,
881 .num_boxes = 1,
882 .perf_ctr_bits = 48,
883 .event_ctl = NHMEX_U_MSR_PMON_EV_SEL,
884 .perf_ctr = NHMEX_U_MSR_PMON_CTR,
885 .event_mask = NHMEX_U_PMON_RAW_EVENT_MASK,
886 .box_ctl = NHMEX_U_MSR_PMON_GLOBAL_CTL,
887 .ops = &nhmex_uncore_ops,
888 .format_group = &nhmex_uncore_ubox_format_group
889};
890
891static struct attribute *nhmex_uncore_cbox_formats_attr[] = {
892 &format_attr_event.attr,
893 &format_attr_umask.attr,
894 &format_attr_edge.attr,
895 &format_attr_inv.attr,
896 &format_attr_thresh8.attr,
897 NULL,
898};
899
900static struct attribute_group nhmex_uncore_cbox_format_group = {
901 .name = "format",
902 .attrs = nhmex_uncore_cbox_formats_attr,
903};
904
905static struct intel_uncore_type nhmex_uncore_cbox = {
906 .name = "cbox",
907 .num_counters = 6,
908 .num_boxes = 8,
909 .perf_ctr_bits = 48,
910 .event_ctl = NHMEX_C0_MSR_PMON_EV_SEL0,
911 .perf_ctr = NHMEX_C0_MSR_PMON_CTR0,
912 .event_mask = NHMEX_PMON_RAW_EVENT_MASK,
913 .box_ctl = NHMEX_C0_MSR_PMON_GLOBAL_CTL,
914 .msr_offset = NHMEX_C_MSR_OFFSET,
915 .pair_ctr_ctl = 1,
916 .ops = &nhmex_uncore_ops,
917 .format_group = &nhmex_uncore_cbox_format_group
918};
919
920static struct uncore_event_desc nhmex_uncore_wbox_events[] = {
921 INTEL_UNCORE_EVENT_DESC(clockticks, "event=0xff,umask=0"),
922 { /* end: all zeroes */ },
923};
924
925static struct intel_uncore_type nhmex_uncore_wbox = {
926 .name = "wbox",
927 .num_counters = 4,
928 .num_boxes = 1,
929 .perf_ctr_bits = 48,
930 .event_ctl = NHMEX_W_MSR_PMON_CNT0,
931 .perf_ctr = NHMEX_W_MSR_PMON_EVT_SEL0,
932 .fixed_ctr = NHMEX_W_MSR_PMON_FIXED_CTR,
933 .fixed_ctl = NHMEX_W_MSR_PMON_FIXED_CTL,
934 .event_mask = NHMEX_PMON_RAW_EVENT_MASK,
935 .box_ctl = NHMEX_W_MSR_GLOBAL_CTL,
936 .pair_ctr_ctl = 1,
937 .event_descs = nhmex_uncore_wbox_events,
938 .ops = &nhmex_uncore_ops,
939 .format_group = &nhmex_uncore_cbox_format_group
940};
941
942static int nhmex_bbox_hw_config(struct intel_uncore_box *box, struct perf_event *event)
943{
944 struct hw_perf_event *hwc = &event->hw;
945 struct hw_perf_event_extra *reg1 = &hwc->extra_reg;
946 struct hw_perf_event_extra *reg2 = &hwc->branch_reg;
947 int ctr, ev_sel;
948
949 ctr = (hwc->config & NHMEX_B_PMON_CTR_MASK) >>
950 NHMEX_B_PMON_CTR_SHIFT;
951 ev_sel = (hwc->config & NHMEX_B_PMON_CTL_EV_SEL_MASK) >>
952 NHMEX_B_PMON_CTL_EV_SEL_SHIFT;
953
954 /* events that do not use the match/mask registers */
955 if ((ctr == 0 && ev_sel > 0x3) || (ctr == 1 && ev_sel > 0x6) ||
956 (ctr == 2 && ev_sel != 0x4) || ctr == 3)
957 return 0;
958
959 if (box->pmu->pmu_idx == 0)
960 reg1->reg = NHMEX_B0_MSR_MATCH;
961 else
962 reg1->reg = NHMEX_B1_MSR_MATCH;
963 reg1->idx = 0;
964 reg1->config = event->attr.config1;
965 reg2->config = event->attr.config2;
966 return 0;
967}
968
969static void nhmex_bbox_msr_enable_event(struct intel_uncore_box *box, struct perf_event *event)
970{
971 struct hw_perf_event *hwc = &event->hw;
972 struct hw_perf_event_extra *reg1 = &hwc->extra_reg;
973 struct hw_perf_event_extra *reg2 = &hwc->branch_reg;
974
975 if (reg1->idx != EXTRA_REG_NONE) {
976 wrmsrl(reg1->reg, reg1->config);
977 wrmsrl(reg1->reg + 1, reg2->config);
978 }
979 wrmsrl(hwc->config_base, NHMEX_PMON_CTL_EN_BIT0 |
980 (hwc->config & NHMEX_B_PMON_CTL_EV_SEL_MASK));
981}
982
983/*
984 * The Bbox has 4 counters, but each counter monitors different events.
985 * Use bits 6-7 in the event config to select counter.
986 */
987static struct event_constraint nhmex_uncore_bbox_constraints[] = {
988 EVENT_CONSTRAINT(0 , 1, 0xc0),
989 EVENT_CONSTRAINT(0x40, 2, 0xc0),
990 EVENT_CONSTRAINT(0x80, 4, 0xc0),
991 EVENT_CONSTRAINT(0xc0, 8, 0xc0),
992 EVENT_CONSTRAINT_END,
993};
994
995static struct attribute *nhmex_uncore_bbox_formats_attr[] = {
996 &format_attr_event5.attr,
997 &format_attr_counter.attr,
998 &format_attr_match.attr,
999 &format_attr_mask.attr,
1000 NULL,
1001};
1002
1003static struct attribute_group nhmex_uncore_bbox_format_group = {
1004 .name = "format",
1005 .attrs = nhmex_uncore_bbox_formats_attr,
1006};
1007
1008static struct intel_uncore_ops nhmex_uncore_bbox_ops = {
1009 NHMEX_UNCORE_OPS_COMMON_INIT(),
1010 .enable_event = nhmex_bbox_msr_enable_event,
1011 .hw_config = nhmex_bbox_hw_config,
1012 .get_constraint = uncore_get_constraint,
1013 .put_constraint = uncore_put_constraint,
1014};
1015
1016static struct intel_uncore_type nhmex_uncore_bbox = {
1017 .name = "bbox",
1018 .num_counters = 4,
1019 .num_boxes = 2,
1020 .perf_ctr_bits = 48,
1021 .event_ctl = NHMEX_B0_MSR_PMON_CTL0,
1022 .perf_ctr = NHMEX_B0_MSR_PMON_CTR0,
1023 .event_mask = NHMEX_B_PMON_RAW_EVENT_MASK,
1024 .box_ctl = NHMEX_B0_MSR_PMON_GLOBAL_CTL,
1025 .msr_offset = NHMEX_B_MSR_OFFSET,
1026 .pair_ctr_ctl = 1,
1027 .num_shared_regs = 1,
1028 .constraints = nhmex_uncore_bbox_constraints,
1029 .ops = &nhmex_uncore_bbox_ops,
1030 .format_group = &nhmex_uncore_bbox_format_group
1031};
1032
1033static int nhmex_sbox_hw_config(struct intel_uncore_box *box, struct perf_event *event)
1034{
1035 struct hw_perf_event_extra *reg1 = &event->hw.extra_reg;
1036 struct hw_perf_event_extra *reg2 = &event->hw.branch_reg;
1037
1038 if (event->attr.config & NHMEX_S_PMON_MM_CFG_EN) {
1039 reg1->config = event->attr.config1;
1040 reg2->config = event->attr.config2;
1041 } else {
1042 reg1->config = ~0ULL;
1043 reg2->config = ~0ULL;
1044 }
1045
1046 if (box->pmu->pmu_idx == 0)
1047 reg1->reg = NHMEX_S0_MSR_MM_CFG;
1048 else
1049 reg1->reg = NHMEX_S1_MSR_MM_CFG;
1050
1051 reg1->idx = 0;
1052
1053 return 0;
1054}
1055
1056static void nhmex_sbox_msr_enable_event(struct intel_uncore_box *box, struct perf_event *event)
1057{
1058 struct hw_perf_event *hwc = &event->hw;
1059 struct hw_perf_event_extra *reg1 = &hwc->extra_reg;
1060 struct hw_perf_event_extra *reg2 = &hwc->branch_reg;
1061
1062 wrmsrl(reg1->reg, 0);
1063 if (reg1->config != ~0ULL || reg2->config != ~0ULL) {
1064 wrmsrl(reg1->reg + 1, reg1->config);
1065 wrmsrl(reg1->reg + 2, reg2->config);
1066 wrmsrl(reg1->reg, NHMEX_S_PMON_MM_CFG_EN);
1067 }
1068 wrmsrl(hwc->config_base, hwc->config | NHMEX_PMON_CTL_EN_BIT22);
1069}
1070
1071static struct attribute *nhmex_uncore_sbox_formats_attr[] = {
1072 &format_attr_event.attr,
1073 &format_attr_umask.attr,
1074 &format_attr_edge.attr,
1075 &format_attr_inv.attr,
1076 &format_attr_thresh8.attr,
1077 &format_attr_mm_cfg.attr,
1078 &format_attr_match.attr,
1079 &format_attr_mask.attr,
1080 NULL,
1081};
1082
1083static struct attribute_group nhmex_uncore_sbox_format_group = {
1084 .name = "format",
1085 .attrs = nhmex_uncore_sbox_formats_attr,
1086};
1087
1088static struct intel_uncore_ops nhmex_uncore_sbox_ops = {
1089 NHMEX_UNCORE_OPS_COMMON_INIT(),
1090 .enable_event = nhmex_sbox_msr_enable_event,
1091 .hw_config = nhmex_sbox_hw_config,
1092 .get_constraint = uncore_get_constraint,
1093 .put_constraint = uncore_put_constraint,
1094};
1095
1096static struct intel_uncore_type nhmex_uncore_sbox = {
1097 .name = "sbox",
1098 .num_counters = 4,
1099 .num_boxes = 2,
1100 .perf_ctr_bits = 48,
1101 .event_ctl = NHMEX_S0_MSR_PMON_CTL0,
1102 .perf_ctr = NHMEX_S0_MSR_PMON_CTR0,
1103 .event_mask = NHMEX_PMON_RAW_EVENT_MASK,
1104 .box_ctl = NHMEX_S0_MSR_PMON_GLOBAL_CTL,
1105 .msr_offset = NHMEX_S_MSR_OFFSET,
1106 .pair_ctr_ctl = 1,
1107 .num_shared_regs = 1,
1108 .ops = &nhmex_uncore_sbox_ops,
1109 .format_group = &nhmex_uncore_sbox_format_group
1110};
1111
1112enum {
1113 EXTRA_REG_NHMEX_M_FILTER,
1114 EXTRA_REG_NHMEX_M_DSP,
1115 EXTRA_REG_NHMEX_M_ISS,
1116 EXTRA_REG_NHMEX_M_MAP,
1117 EXTRA_REG_NHMEX_M_MSC_THR,
1118 EXTRA_REG_NHMEX_M_PGT,
1119 EXTRA_REG_NHMEX_M_PLD,
1120 EXTRA_REG_NHMEX_M_ZDP_CTL_FVC,
1121};
1122
1123static struct extra_reg nhmex_uncore_mbox_extra_regs[] = {
1124 MBOX_INC_SEL_EXTAR_REG(0x0, DSP),
1125 MBOX_INC_SEL_EXTAR_REG(0x4, MSC_THR),
1126 MBOX_INC_SEL_EXTAR_REG(0x5, MSC_THR),
1127 MBOX_INC_SEL_EXTAR_REG(0x9, ISS),
1128 /* event 0xa uses two extra registers */
1129 MBOX_INC_SEL_EXTAR_REG(0xa, ISS),
1130 MBOX_INC_SEL_EXTAR_REG(0xa, PLD),
1131 MBOX_INC_SEL_EXTAR_REG(0xb, PLD),
1132 /* events 0xd ~ 0x10 use the same extra register */
1133 MBOX_INC_SEL_EXTAR_REG(0xd, ZDP_CTL_FVC),
1134 MBOX_INC_SEL_EXTAR_REG(0xe, ZDP_CTL_FVC),
1135 MBOX_INC_SEL_EXTAR_REG(0xf, ZDP_CTL_FVC),
1136 MBOX_INC_SEL_EXTAR_REG(0x10, ZDP_CTL_FVC),
1137 MBOX_INC_SEL_EXTAR_REG(0x16, PGT),
1138 MBOX_SET_FLAG_SEL_EXTRA_REG(0x0, DSP),
1139 MBOX_SET_FLAG_SEL_EXTRA_REG(0x1, ISS),
1140 MBOX_SET_FLAG_SEL_EXTRA_REG(0x5, PGT),
1141 MBOX_SET_FLAG_SEL_EXTRA_REG(0x6, MAP),
1142 EVENT_EXTRA_END
1143};
1144
1145static bool nhmex_mbox_get_shared_reg(struct intel_uncore_box *box, int idx, u64 config)
1146{
1147 struct intel_uncore_extra_reg *er;
1148 unsigned long flags;
1149 bool ret = false;
1150 u64 mask;
1151
1152 if (idx < EXTRA_REG_NHMEX_M_ZDP_CTL_FVC) {
1153 er = &box->shared_regs[idx];
1154 raw_spin_lock_irqsave(&er->lock, flags);
1155 if (!atomic_read(&er->ref) || er->config == config) {
1156 atomic_inc(&er->ref);
1157 er->config = config;
1158 ret = true;
1159 }
1160 raw_spin_unlock_irqrestore(&er->lock, flags);
1161
1162 return ret;
1163 }
1164 /*
1165 * The ZDP_CTL_FVC MSR has 4 fields which are used to control
1166 * events 0xd ~ 0x10. Besides these 4 fields, there are additional
1167 * fields which are shared.
1168 */
1169 idx -= EXTRA_REG_NHMEX_M_ZDP_CTL_FVC;
1170 if (WARN_ON_ONCE(idx >= 4))
1171 return false;
1172
1173 /* mask of the shared fields */
1174 mask = NHMEX_M_PMON_ZDP_CTL_FVC_MASK;
1175 er = &box->shared_regs[EXTRA_REG_NHMEX_M_ZDP_CTL_FVC];
1176
1177 raw_spin_lock_irqsave(&er->lock, flags);
1178 /* add mask of the non-shared field if it's in use */
1179 if (__BITS_VALUE(atomic_read(&er->ref), idx, 8))
1180 mask |= NHMEX_M_PMON_ZDP_CTL_FVC_EVENT_MASK(idx);
1181
1182 if (!atomic_read(&er->ref) || !((er->config ^ config) & mask)) {
1183 atomic_add(1 << (idx * 8), &er->ref);
1184 mask = NHMEX_M_PMON_ZDP_CTL_FVC_MASK |
1185 NHMEX_M_PMON_ZDP_CTL_FVC_EVENT_MASK(idx);
1186 er->config &= ~mask;
1187 er->config |= (config & mask);
1188 ret = true;
1189 }
1190 raw_spin_unlock_irqrestore(&er->lock, flags);
1191
1192 return ret;
1193}
1194
1195static void nhmex_mbox_put_shared_reg(struct intel_uncore_box *box, int idx)
1196{
1197 struct intel_uncore_extra_reg *er;
1198
1199 if (idx < EXTRA_REG_NHMEX_M_ZDP_CTL_FVC) {
1200 er = &box->shared_regs[idx];
1201 atomic_dec(&er->ref);
1202 return;
1203 }
1204
1205 idx -= EXTRA_REG_NHMEX_M_ZDP_CTL_FVC;
1206 er = &box->shared_regs[EXTRA_REG_NHMEX_M_ZDP_CTL_FVC];
1207 atomic_sub(1 << (idx * 8), &er->ref);
1208}
1209
1210u64 nhmex_mbox_alter_er(struct perf_event *event, int new_idx, bool modify)
1211{
1212 struct hw_perf_event *hwc = &event->hw;
1213 struct hw_perf_event_extra *reg1 = &hwc->extra_reg;
1214 int idx, orig_idx = __BITS_VALUE(reg1->idx, 0, 8);
1215 u64 config = reg1->config;
1216
1217 /* get the non-shared control bits and shift them */
1218 idx = orig_idx - EXTRA_REG_NHMEX_M_ZDP_CTL_FVC;
1219 config &= NHMEX_M_PMON_ZDP_CTL_FVC_EVENT_MASK(idx);
1220 if (new_idx > orig_idx) {
1221 idx = new_idx - orig_idx;
1222 config <<= 3 * idx;
1223 } else {
1224 idx = orig_idx - new_idx;
1225 config >>= 3 * idx;
1226 }
1227
1228 /* add the shared control bits back */
1229 config |= NHMEX_M_PMON_ZDP_CTL_FVC_MASK & reg1->config;
1230 if (modify) {
1231 /* adjust the main event selector */
1232 if (new_idx > orig_idx)
1233 hwc->config += idx << NHMEX_M_PMON_CTL_INC_SEL_SHIFT;
1234 else
1235 hwc->config -= idx << NHMEX_M_PMON_CTL_INC_SEL_SHIFT;
1236 reg1->config = config;
1237 reg1->idx = ~0xff | new_idx;
1238 }
1239 return config;
1240}
1241
1242static struct event_constraint *
1243nhmex_mbox_get_constraint(struct intel_uncore_box *box, struct perf_event *event)
1244{
1245 struct hw_perf_event_extra *reg1 = &event->hw.extra_reg;
1246 struct hw_perf_event_extra *reg2 = &event->hw.branch_reg;
1247 int i, idx[2], alloc = 0;
1248 u64 config1 = reg1->config;
1249
1250 idx[0] = __BITS_VALUE(reg1->idx, 0, 8);
1251 idx[1] = __BITS_VALUE(reg1->idx, 1, 8);
1252again:
1253 for (i = 0; i < 2; i++) {
1254 if (!uncore_box_is_fake(box) && (reg1->alloc & (0x1 << i)))
1255 idx[i] = 0xff;
1256
1257 if (idx[i] == 0xff)
1258 continue;
1259
1260 if (!nhmex_mbox_get_shared_reg(box, idx[i],
1261 __BITS_VALUE(config1, i, 32)))
1262 goto fail;
1263 alloc |= (0x1 << i);
1264 }
1265
1266 /* for the match/mask registers */
1267 if ((uncore_box_is_fake(box) || !reg2->alloc) &&
1268 !nhmex_mbox_get_shared_reg(box, reg2->idx, reg2->config))
1269 goto fail;
1270
1271 /*
1272 * If it's a fake box -- as per validate_{group,event}() we
1273 * shouldn't touch event state and we can avoid doing so
1274 * since both will only call get_event_constraints() once
1275 * on each event, this avoids the need for reg->alloc.
1276 */
1277 if (!uncore_box_is_fake(box)) {
1278 if (idx[0] != 0xff && idx[0] != __BITS_VALUE(reg1->idx, 0, 8))
1279 nhmex_mbox_alter_er(event, idx[0], true);
1280 reg1->alloc |= alloc;
1281 reg2->alloc = 1;
1282 }
1283 return NULL;
1284fail:
1285 if (idx[0] != 0xff && !(alloc & 0x1) &&
1286 idx[0] >= EXTRA_REG_NHMEX_M_ZDP_CTL_FVC) {
1287 /*
1288 * events 0xd ~ 0x10 are functional identical, but are
1289 * controlled by different fields in the ZDP_CTL_FVC
1290 * register. If we failed to take one field, try the
1291 * rest 3 choices.
1292 */
1293 BUG_ON(__BITS_VALUE(reg1->idx, 1, 8) != 0xff);
1294 idx[0] -= EXTRA_REG_NHMEX_M_ZDP_CTL_FVC;
1295 idx[0] = (idx[0] + 1) % 4;
1296 idx[0] += EXTRA_REG_NHMEX_M_ZDP_CTL_FVC;
1297 if (idx[0] != __BITS_VALUE(reg1->idx, 0, 8)) {
1298 config1 = nhmex_mbox_alter_er(event, idx[0], false);
1299 goto again;
1300 }
1301 }
1302
1303 if (alloc & 0x1)
1304 nhmex_mbox_put_shared_reg(box, idx[0]);
1305 if (alloc & 0x2)
1306 nhmex_mbox_put_shared_reg(box, idx[1]);
1307 return &constraint_empty;
1308}
1309
1310static void nhmex_mbox_put_constraint(struct intel_uncore_box *box, struct perf_event *event)
1311{
1312 struct hw_perf_event_extra *reg1 = &event->hw.extra_reg;
1313 struct hw_perf_event_extra *reg2 = &event->hw.branch_reg;
1314
1315 if (uncore_box_is_fake(box))
1316 return;
1317
1318 if (reg1->alloc & 0x1)
1319 nhmex_mbox_put_shared_reg(box, __BITS_VALUE(reg1->idx, 0, 8));
1320 if (reg1->alloc & 0x2)
1321 nhmex_mbox_put_shared_reg(box, __BITS_VALUE(reg1->idx, 1, 8));
1322 reg1->alloc = 0;
1323
1324 if (reg2->alloc) {
1325 nhmex_mbox_put_shared_reg(box, reg2->idx);
1326 reg2->alloc = 0;
1327 }
1328}
1329
1330static int nhmex_mbox_extra_reg_idx(struct extra_reg *er)
1331{
1332 if (er->idx < EXTRA_REG_NHMEX_M_ZDP_CTL_FVC)
1333 return er->idx;
1334 return er->idx + (er->event >> NHMEX_M_PMON_CTL_INC_SEL_SHIFT) - 0xd;
1335}
1336
1337static int nhmex_mbox_hw_config(struct intel_uncore_box *box, struct perf_event *event)
1338{
1339 struct intel_uncore_type *type = box->pmu->type;
1340 struct hw_perf_event_extra *reg1 = &event->hw.extra_reg;
1341 struct hw_perf_event_extra *reg2 = &event->hw.branch_reg;
1342 struct extra_reg *er;
1343 unsigned msr;
1344 int reg_idx = 0;
1345
1346 if (WARN_ON_ONCE(reg1->idx != -1))
1347 return -EINVAL;
1348 /*
1349 * The mbox events may require 2 extra MSRs at the most. But only
1350 * the lower 32 bits in these MSRs are significant, so we can use
1351 * config1 to pass two MSRs' config.
1352 */
1353 for (er = nhmex_uncore_mbox_extra_regs; er->msr; er++) {
1354 if (er->event != (event->hw.config & er->config_mask))
1355 continue;
1356 if (event->attr.config1 & ~er->valid_mask)
1357 return -EINVAL;
1358 if (er->idx == __BITS_VALUE(reg1->idx, 0, 8) ||
1359 er->idx == __BITS_VALUE(reg1->idx, 1, 8))
1360 continue;
1361 if (WARN_ON_ONCE(reg_idx >= 2))
1362 return -EINVAL;
1363
1364 msr = er->msr + type->msr_offset * box->pmu->pmu_idx;
1365 if (WARN_ON_ONCE(msr >= 0xffff || er->idx >= 0xff))
1366 return -EINVAL;
1367
1368 /* always use the 32~63 bits to pass the PLD config */
1369 if (er->idx == EXTRA_REG_NHMEX_M_PLD)
1370 reg_idx = 1;
1371
1372 reg1->idx &= ~(0xff << (reg_idx * 8));
1373 reg1->reg &= ~(0xffff << (reg_idx * 16));
1374 reg1->idx |= nhmex_mbox_extra_reg_idx(er) << (reg_idx * 8);
1375 reg1->reg |= msr << (reg_idx * 16);
1376 reg1->config = event->attr.config1;
1377 reg_idx++;
1378 }
1379 /* use config2 to pass the filter config */
1380 reg2->idx = EXTRA_REG_NHMEX_M_FILTER;
1381 if (event->attr.config2 & NHMEX_M_PMON_MM_CFG_EN)
1382 reg2->config = event->attr.config2;
1383 else
1384 reg2->config = ~0ULL;
1385 if (box->pmu->pmu_idx == 0)
1386 reg2->reg = NHMEX_M0_MSR_PMU_MM_CFG;
1387 else
1388 reg2->reg = NHMEX_M1_MSR_PMU_MM_CFG;
1389
1390 return 0;
1391}
1392
1393static u64 nhmex_mbox_shared_reg_config(struct intel_uncore_box *box, int idx)
1394{
1395 struct intel_uncore_extra_reg *er;
1396 unsigned long flags;
1397 u64 config;
1398
1399 if (idx < EXTRA_REG_NHMEX_M_ZDP_CTL_FVC)
1400 return box->shared_regs[idx].config;
1401
1402 er = &box->shared_regs[EXTRA_REG_NHMEX_M_ZDP_CTL_FVC];
1403 raw_spin_lock_irqsave(&er->lock, flags);
1404 config = er->config;
1405 raw_spin_unlock_irqrestore(&er->lock, flags);
1406 return config;
1407}
1408
1409static void nhmex_mbox_msr_enable_event(struct intel_uncore_box *box, struct perf_event *event)
1410{
1411 struct hw_perf_event *hwc = &event->hw;
1412 struct hw_perf_event_extra *reg1 = &hwc->extra_reg;
1413 struct hw_perf_event_extra *reg2 = &hwc->branch_reg;
1414 int idx;
1415
1416 idx = __BITS_VALUE(reg1->idx, 0, 8);
1417 if (idx != 0xff)
1418 wrmsrl(__BITS_VALUE(reg1->reg, 0, 16),
1419 nhmex_mbox_shared_reg_config(box, idx));
1420 idx = __BITS_VALUE(reg1->idx, 1, 8);
1421 if (idx != 0xff)
1422 wrmsrl(__BITS_VALUE(reg1->reg, 1, 16),
1423 nhmex_mbox_shared_reg_config(box, idx));
1424
1425 wrmsrl(reg2->reg, 0);
1426 if (reg2->config != ~0ULL) {
1427 wrmsrl(reg2->reg + 1,
1428 reg2->config & NHMEX_M_PMON_ADDR_MATCH_MASK);
1429 wrmsrl(reg2->reg + 2, NHMEX_M_PMON_ADDR_MASK_MASK &
1430 (reg2->config >> NHMEX_M_PMON_ADDR_MASK_SHIFT));
1431 wrmsrl(reg2->reg, NHMEX_M_PMON_MM_CFG_EN);
1432 }
1433
1434 wrmsrl(hwc->config_base, hwc->config | NHMEX_PMON_CTL_EN_BIT0);
1435}
1436
1437DEFINE_UNCORE_FORMAT_ATTR(count_mode, count_mode, "config:2-3");
1438DEFINE_UNCORE_FORMAT_ATTR(storage_mode, storage_mode, "config:4-5");
1439DEFINE_UNCORE_FORMAT_ATTR(wrap_mode, wrap_mode, "config:6");
1440DEFINE_UNCORE_FORMAT_ATTR(flag_mode, flag_mode, "config:7");
1441DEFINE_UNCORE_FORMAT_ATTR(inc_sel, inc_sel, "config:9-13");
1442DEFINE_UNCORE_FORMAT_ATTR(set_flag_sel, set_flag_sel, "config:19-21");
1443DEFINE_UNCORE_FORMAT_ATTR(filter_cfg, filter_cfg, "config2:63");
1444DEFINE_UNCORE_FORMAT_ATTR(filter_match, filter_match, "config2:0-33");
1445DEFINE_UNCORE_FORMAT_ATTR(filter_mask, filter_mask, "config2:34-61");
1446DEFINE_UNCORE_FORMAT_ATTR(dsp, dsp, "config1:0-31");
1447DEFINE_UNCORE_FORMAT_ATTR(thr, thr, "config1:0-31");
1448DEFINE_UNCORE_FORMAT_ATTR(fvc, fvc, "config1:0-31");
1449DEFINE_UNCORE_FORMAT_ATTR(pgt, pgt, "config1:0-31");
1450DEFINE_UNCORE_FORMAT_ATTR(map, map, "config1:0-31");
1451DEFINE_UNCORE_FORMAT_ATTR(iss, iss, "config1:0-31");
1452DEFINE_UNCORE_FORMAT_ATTR(pld, pld, "config1:32-63");
1453
1454static struct attribute *nhmex_uncore_mbox_formats_attr[] = {
1455 &format_attr_count_mode.attr,
1456 &format_attr_storage_mode.attr,
1457 &format_attr_wrap_mode.attr,
1458 &format_attr_flag_mode.attr,
1459 &format_attr_inc_sel.attr,
1460 &format_attr_set_flag_sel.attr,
1461 &format_attr_filter_cfg.attr,
1462 &format_attr_filter_match.attr,
1463 &format_attr_filter_mask.attr,
1464 &format_attr_dsp.attr,
1465 &format_attr_thr.attr,
1466 &format_attr_fvc.attr,
1467 &format_attr_pgt.attr,
1468 &format_attr_map.attr,
1469 &format_attr_iss.attr,
1470 &format_attr_pld.attr,
1471 NULL,
1472};
1473
1474static struct attribute_group nhmex_uncore_mbox_format_group = {
1475 .name = "format",
1476 .attrs = nhmex_uncore_mbox_formats_attr,
1477};
1478
1479static struct uncore_event_desc nhmex_uncore_mbox_events[] = {
1480 INTEL_UNCORE_EVENT_DESC(bbox_cmds_read, "inc_sel=0xd,fvc=0x2800"),
1481 INTEL_UNCORE_EVENT_DESC(bbox_cmds_write, "inc_sel=0xd,fvc=0x2820"),
1482 { /* end: all zeroes */ },
1483};
1484
1485static struct intel_uncore_ops nhmex_uncore_mbox_ops = {
1486 NHMEX_UNCORE_OPS_COMMON_INIT(),
1487 .enable_event = nhmex_mbox_msr_enable_event,
1488 .hw_config = nhmex_mbox_hw_config,
1489 .get_constraint = nhmex_mbox_get_constraint,
1490 .put_constraint = nhmex_mbox_put_constraint,
1491};
1492
1493static struct intel_uncore_type nhmex_uncore_mbox = {
1494 .name = "mbox",
1495 .num_counters = 6,
1496 .num_boxes = 2,
1497 .perf_ctr_bits = 48,
1498 .event_ctl = NHMEX_M0_MSR_PMU_CTL0,
1499 .perf_ctr = NHMEX_M0_MSR_PMU_CNT0,
1500 .event_mask = NHMEX_M_PMON_RAW_EVENT_MASK,
1501 .box_ctl = NHMEX_M0_MSR_GLOBAL_CTL,
1502 .msr_offset = NHMEX_M_MSR_OFFSET,
1503 .pair_ctr_ctl = 1,
1504 .num_shared_regs = 8,
1505 .event_descs = nhmex_uncore_mbox_events,
1506 .ops = &nhmex_uncore_mbox_ops,
1507 .format_group = &nhmex_uncore_mbox_format_group,
1508};
1509
1510void nhmex_rbox_alter_er(struct intel_uncore_box *box, struct perf_event *event)
1511{
1512 struct hw_perf_event *hwc = &event->hw;
1513 struct hw_perf_event_extra *reg1 = &hwc->extra_reg;
1514 int port;
1515
1516 /* adjust the main event selector */
1517 if (reg1->idx % 2) {
1518 reg1->idx--;
1519 hwc->config -= 1 << NHMEX_R_PMON_CTL_EV_SEL_SHIFT;
1520 } else {
1521 reg1->idx++;
1522 hwc->config += 1 << NHMEX_R_PMON_CTL_EV_SEL_SHIFT;
1523 }
1524
1525 /* adjust address or config of extra register */
1526 port = reg1->idx / 6 + box->pmu->pmu_idx * 4;
1527 switch (reg1->idx % 6) {
1528 case 0:
1529 reg1->reg = NHMEX_R_MSR_PORTN_IPERF_CFG0(port);
1530 break;
1531 case 1:
1532 reg1->reg = NHMEX_R_MSR_PORTN_IPERF_CFG1(port);
1533 break;
1534 case 2:
1535 /* the 8~15 bits to the 0~7 bits */
1536 reg1->config >>= 8;
1537 break;
1538 case 3:
1539 /* the 0~7 bits to the 8~15 bits */
1540 reg1->config <<= 8;
1541 break;
1542 case 4:
1543 reg1->reg = NHMEX_R_MSR_PORTN_XBR_SET1_MM_CFG(port);
1544 break;
1545 case 5:
1546 reg1->reg = NHMEX_R_MSR_PORTN_XBR_SET2_MM_CFG(port);
1547 break;
1548 };
1549}
1550
1551/*
1552 * Each rbox has 4 event set which monitor PQI port 0~3 or 4~7.
1553 * An event set consists of 6 events, the 3rd and 4th events in
1554 * an event set use the same extra register. So an event set uses
1555 * 5 extra registers.
1556 */
1557static struct event_constraint *
1558nhmex_rbox_get_constraint(struct intel_uncore_box *box, struct perf_event *event)
1559{
1560 struct hw_perf_event *hwc = &event->hw;
1561 struct hw_perf_event_extra *reg1 = &hwc->extra_reg;
1562 struct hw_perf_event_extra *reg2 = &hwc->branch_reg;
1563 struct intel_uncore_extra_reg *er;
1564 unsigned long flags;
1565 int idx, er_idx;
1566 u64 config1;
1567 bool ok = false;
1568
1569 if (!uncore_box_is_fake(box) && reg1->alloc)
1570 return NULL;
1571
1572 idx = reg1->idx % 6;
1573 config1 = reg1->config;
1574again:
1575 er_idx = idx;
1576 /* the 3rd and 4th events use the same extra register */
1577 if (er_idx > 2)
1578 er_idx--;
1579 er_idx += (reg1->idx / 6) * 5;
1580
1581 er = &box->shared_regs[er_idx];
1582 raw_spin_lock_irqsave(&er->lock, flags);
1583 if (idx < 2) {
1584 if (!atomic_read(&er->ref) || er->config == reg1->config) {
1585 atomic_inc(&er->ref);
1586 er->config = reg1->config;
1587 ok = true;
1588 }
1589 } else if (idx == 2 || idx == 3) {
1590 /*
1591 * these two events use different fields in a extra register,
1592 * the 0~7 bits and the 8~15 bits respectively.
1593 */
1594 u64 mask = 0xff << ((idx - 2) * 8);
1595 if (!__BITS_VALUE(atomic_read(&er->ref), idx - 2, 8) ||
1596 !((er->config ^ config1) & mask)) {
1597 atomic_add(1 << ((idx - 2) * 8), &er->ref);
1598 er->config &= ~mask;
1599 er->config |= config1 & mask;
1600 ok = true;
1601 }
1602 } else {
1603 if (!atomic_read(&er->ref) ||
1604 (er->config == (hwc->config >> 32) &&
1605 er->config1 == reg1->config &&
1606 er->config2 == reg2->config)) {
1607 atomic_inc(&er->ref);
1608 er->config = (hwc->config >> 32);
1609 er->config1 = reg1->config;
1610 er->config2 = reg2->config;
1611 ok = true;
1612 }
1613 }
1614 raw_spin_unlock_irqrestore(&er->lock, flags);
1615
1616 if (!ok) {
1617 /*
1618 * The Rbox events are always in pairs. The paired
1619 * events are functional identical, but use different
1620 * extra registers. If we failed to take an extra
1621 * register, try the alternative.
1622 */
1623 if (idx % 2)
1624 idx--;
1625 else
1626 idx++;
1627 if (idx != reg1->idx % 6) {
1628 if (idx == 2)
1629 config1 >>= 8;
1630 else if (idx == 3)
1631 config1 <<= 8;
1632 goto again;
1633 }
1634 } else {
1635 if (!uncore_box_is_fake(box)) {
1636 if (idx != reg1->idx % 6)
1637 nhmex_rbox_alter_er(box, event);
1638 reg1->alloc = 1;
1639 }
1640 return NULL;
1641 }
1642 return &constraint_empty;
1643}
1644
1645static void nhmex_rbox_put_constraint(struct intel_uncore_box *box, struct perf_event *event)
1646{
1647 struct intel_uncore_extra_reg *er;
1648 struct hw_perf_event_extra *reg1 = &event->hw.extra_reg;
1649 int idx, er_idx;
1650
1651 if (uncore_box_is_fake(box) || !reg1->alloc)
1652 return;
1653
1654 idx = reg1->idx % 6;
1655 er_idx = idx;
1656 if (er_idx > 2)
1657 er_idx--;
1658 er_idx += (reg1->idx / 6) * 5;
1659
1660 er = &box->shared_regs[er_idx];
1661 if (idx == 2 || idx == 3)
1662 atomic_sub(1 << ((idx - 2) * 8), &er->ref);
1663 else
1664 atomic_dec(&er->ref);
1665
1666 reg1->alloc = 0;
1667}
1668
1669static int nhmex_rbox_hw_config(struct intel_uncore_box *box, struct perf_event *event)
1670{
1671 struct hw_perf_event *hwc = &event->hw;
1672 struct hw_perf_event_extra *reg1 = &event->hw.extra_reg;
1673 struct hw_perf_event_extra *reg2 = &event->hw.branch_reg;
1674 int port, idx;
1675
1676 idx = (event->hw.config & NHMEX_R_PMON_CTL_EV_SEL_MASK) >>
1677 NHMEX_R_PMON_CTL_EV_SEL_SHIFT;
1678 if (idx >= 0x18)
1679 return -EINVAL;
1680
1681 reg1->idx = idx;
1682 reg1->config = event->attr.config1;
1683
1684 port = idx / 6 + box->pmu->pmu_idx * 4;
1685 idx %= 6;
1686 switch (idx) {
1687 case 0:
1688 reg1->reg = NHMEX_R_MSR_PORTN_IPERF_CFG0(port);
1689 break;
1690 case 1:
1691 reg1->reg = NHMEX_R_MSR_PORTN_IPERF_CFG1(port);
1692 break;
1693 case 2:
1694 case 3:
1695 reg1->reg = NHMEX_R_MSR_PORTN_QLX_CFG(port);
1696 break;
1697 case 4:
1698 case 5:
1699 if (idx == 4)
1700 reg1->reg = NHMEX_R_MSR_PORTN_XBR_SET1_MM_CFG(port);
1701 else
1702 reg1->reg = NHMEX_R_MSR_PORTN_XBR_SET2_MM_CFG(port);
1703 reg2->config = event->attr.config2;
1704 hwc->config |= event->attr.config & (~0ULL << 32);
1705 break;
1706 };
1707 return 0;
1708}
1709
1710static u64 nhmex_rbox_shared_reg_config(struct intel_uncore_box *box, int idx)
1711{
1712 struct intel_uncore_extra_reg *er;
1713 unsigned long flags;
1714 u64 config;
1715
1716 er = &box->shared_regs[idx];
1717
1718 raw_spin_lock_irqsave(&er->lock, flags);
1719 config = er->config;
1720 raw_spin_unlock_irqrestore(&er->lock, flags);
1721
1722 return config;
1723}
1724
1725static void nhmex_rbox_msr_enable_event(struct intel_uncore_box *box, struct perf_event *event)
1726{
1727 struct hw_perf_event *hwc = &event->hw;
1728 struct hw_perf_event_extra *reg1 = &hwc->extra_reg;
1729 struct hw_perf_event_extra *reg2 = &hwc->branch_reg;
1730 int idx, er_idx;
1731
1732 idx = reg1->idx % 6;
1733 er_idx = idx;
1734 if (er_idx > 2)
1735 er_idx--;
1736 er_idx += (reg1->idx / 6) * 5;
1737
1738 switch (idx) {
1739 case 0:
1740 case 1:
1741 wrmsrl(reg1->reg, reg1->config);
1742 break;
1743 case 2:
1744 case 3:
1745 wrmsrl(reg1->reg, nhmex_rbox_shared_reg_config(box, er_idx));
1746 break;
1747 case 4:
1748 case 5:
1749 wrmsrl(reg1->reg, reg1->config);
1750 wrmsrl(reg1->reg + 1, hwc->config >> 32);
1751 wrmsrl(reg1->reg + 2, reg2->config);
1752 break;
1753 };
1754
1755 wrmsrl(hwc->config_base, NHMEX_PMON_CTL_EN_BIT0 |
1756 (hwc->config & NHMEX_R_PMON_CTL_EV_SEL_MASK));
1757}
1758
1759DEFINE_UNCORE_FORMAT_ATTR(xbr_match, xbr_match, "config:32-63");
1760DEFINE_UNCORE_FORMAT_ATTR(xbr_mm_cfg, xbr_mm_cfg, "config1:0-63");
1761DEFINE_UNCORE_FORMAT_ATTR(xbr_mask, xbr_mask, "config2:0-63");
1762DEFINE_UNCORE_FORMAT_ATTR(qlx_cfg, qlx_cfg, "config1:0-15");
1763DEFINE_UNCORE_FORMAT_ATTR(iperf_cfg, iperf_cfg, "config1:0-31");
1764
1765static struct attribute *nhmex_uncore_rbox_formats_attr[] = {
1766 &format_attr_event5.attr,
1767 &format_attr_xbr_mm_cfg.attr,
1768 &format_attr_xbr_match.attr,
1769 &format_attr_xbr_mask.attr,
1770 &format_attr_qlx_cfg.attr,
1771 &format_attr_iperf_cfg.attr,
1772 NULL,
1773};
1774
1775static struct attribute_group nhmex_uncore_rbox_format_group = {
1776 .name = "format",
1777 .attrs = nhmex_uncore_rbox_formats_attr,
1778};
1779
1780static struct uncore_event_desc nhmex_uncore_rbox_events[] = {
1781 INTEL_UNCORE_EVENT_DESC(qpi0_flit_send, "event=0x0,iperf_cfg=0x80000000"),
1782 INTEL_UNCORE_EVENT_DESC(qpi1_filt_send, "event=0x6,iperf_cfg=0x80000000"),
1783 INTEL_UNCORE_EVENT_DESC(qpi0_idle_filt, "event=0x0,iperf_cfg=0x40000000"),
1784 INTEL_UNCORE_EVENT_DESC(qpi1_idle_filt, "event=0x6,iperf_cfg=0x40000000"),
1785 INTEL_UNCORE_EVENT_DESC(qpi0_date_response, "event=0x0,iperf_cfg=0xc4"),
1786 INTEL_UNCORE_EVENT_DESC(qpi1_date_response, "event=0x6,iperf_cfg=0xc4"),
1787 { /* end: all zeroes */ },
1788};
1789
1790static struct intel_uncore_ops nhmex_uncore_rbox_ops = {
1791 NHMEX_UNCORE_OPS_COMMON_INIT(),
1792 .enable_event = nhmex_rbox_msr_enable_event,
1793 .hw_config = nhmex_rbox_hw_config,
1794 .get_constraint = nhmex_rbox_get_constraint,
1795 .put_constraint = nhmex_rbox_put_constraint,
1796};
1797
1798static struct intel_uncore_type nhmex_uncore_rbox = {
1799 .name = "rbox",
1800 .num_counters = 8,
1801 .num_boxes = 2,
1802 .perf_ctr_bits = 48,
1803 .event_ctl = NHMEX_R_MSR_PMON_CTL0,
1804 .perf_ctr = NHMEX_R_MSR_PMON_CNT0,
1805 .event_mask = NHMEX_R_PMON_RAW_EVENT_MASK,
1806 .box_ctl = NHMEX_R_MSR_GLOBAL_CTL,
1807 .msr_offset = NHMEX_R_MSR_OFFSET,
1808 .pair_ctr_ctl = 1,
1809 .num_shared_regs = 20,
1810 .event_descs = nhmex_uncore_rbox_events,
1811 .ops = &nhmex_uncore_rbox_ops,
1812 .format_group = &nhmex_uncore_rbox_format_group
1813};
1814
1815static struct intel_uncore_type *nhmex_msr_uncores[] = {
1816 &nhmex_uncore_ubox,
1817 &nhmex_uncore_cbox,
1818 &nhmex_uncore_bbox,
1819 &nhmex_uncore_sbox,
1820 &nhmex_uncore_mbox,
1821 &nhmex_uncore_rbox,
1822 &nhmex_uncore_wbox,
1823 NULL,
1824};
1825/* end of Nehalem-EX uncore support */
1826
1827static void uncore_assign_hw_event(struct intel_uncore_box *box, struct perf_event *event, int idx)
774{ 1828{
775 struct hw_perf_event *hwc = &event->hw; 1829 struct hw_perf_event *hwc = &event->hw;
776 1830
@@ -787,8 +1841,7 @@ static void uncore_assign_hw_event(struct intel_uncore_box *box,
787 hwc->event_base = uncore_perf_ctr(box, hwc->idx); 1841 hwc->event_base = uncore_perf_ctr(box, hwc->idx);
788} 1842}
789 1843
790static void uncore_perf_event_update(struct intel_uncore_box *box, 1844static void uncore_perf_event_update(struct intel_uncore_box *box, struct perf_event *event)
791 struct perf_event *event)
792{ 1845{
793 u64 prev_count, new_count, delta; 1846 u64 prev_count, new_count, delta;
794 int shift; 1847 int shift;
@@ -858,14 +1911,12 @@ static void uncore_pmu_init_hrtimer(struct intel_uncore_box *box)
858 box->hrtimer.function = uncore_pmu_hrtimer; 1911 box->hrtimer.function = uncore_pmu_hrtimer;
859} 1912}
860 1913
861struct intel_uncore_box *uncore_alloc_box(struct intel_uncore_type *type, 1914struct intel_uncore_box *uncore_alloc_box(struct intel_uncore_type *type, int cpu)
862 int cpu)
863{ 1915{
864 struct intel_uncore_box *box; 1916 struct intel_uncore_box *box;
865 int i, size; 1917 int i, size;
866 1918
867 size = sizeof(*box) + type->num_shared_regs * 1919 size = sizeof(*box) + type->num_shared_regs * sizeof(struct intel_uncore_extra_reg);
868 sizeof(struct intel_uncore_extra_reg);
869 1920
870 box = kmalloc_node(size, GFP_KERNEL | __GFP_ZERO, cpu_to_node(cpu)); 1921 box = kmalloc_node(size, GFP_KERNEL | __GFP_ZERO, cpu_to_node(cpu));
871 if (!box) 1922 if (!box)
@@ -915,12 +1966,11 @@ static struct intel_uncore_box *uncore_event_to_box(struct perf_event *event)
915 * perf core schedules event on the basis of cpu, uncore events are 1966 * perf core schedules event on the basis of cpu, uncore events are
916 * collected by one of the cpus inside a physical package. 1967 * collected by one of the cpus inside a physical package.
917 */ 1968 */
918 return uncore_pmu_to_box(uncore_event_to_pmu(event), 1969 return uncore_pmu_to_box(uncore_event_to_pmu(event), smp_processor_id());
919 smp_processor_id());
920} 1970}
921 1971
922static int uncore_collect_events(struct intel_uncore_box *box, 1972static int
923 struct perf_event *leader, bool dogrp) 1973uncore_collect_events(struct intel_uncore_box *box, struct perf_event *leader, bool dogrp)
924{ 1974{
925 struct perf_event *event; 1975 struct perf_event *event;
926 int n, max_count; 1976 int n, max_count;
@@ -952,8 +2002,7 @@ static int uncore_collect_events(struct intel_uncore_box *box,
952} 2002}
953 2003
954static struct event_constraint * 2004static struct event_constraint *
955uncore_get_event_constraint(struct intel_uncore_box *box, 2005uncore_get_event_constraint(struct intel_uncore_box *box, struct perf_event *event)
956 struct perf_event *event)
957{ 2006{
958 struct intel_uncore_type *type = box->pmu->type; 2007 struct intel_uncore_type *type = box->pmu->type;
959 struct event_constraint *c; 2008 struct event_constraint *c;
@@ -977,15 +2026,13 @@ uncore_get_event_constraint(struct intel_uncore_box *box,
977 return &type->unconstrainted; 2026 return &type->unconstrainted;
978} 2027}
979 2028
980static void uncore_put_event_constraint(struct intel_uncore_box *box, 2029static void uncore_put_event_constraint(struct intel_uncore_box *box, struct perf_event *event)
981 struct perf_event *event)
982{ 2030{
983 if (box->pmu->type->ops->put_constraint) 2031 if (box->pmu->type->ops->put_constraint)
984 box->pmu->type->ops->put_constraint(box, event); 2032 box->pmu->type->ops->put_constraint(box, event);
985} 2033}
986 2034
987static int uncore_assign_events(struct intel_uncore_box *box, 2035static int uncore_assign_events(struct intel_uncore_box *box, int assign[], int n)
988 int assign[], int n)
989{ 2036{
990 unsigned long used_mask[BITS_TO_LONGS(UNCORE_PMC_IDX_MAX)]; 2037 unsigned long used_mask[BITS_TO_LONGS(UNCORE_PMC_IDX_MAX)];
991 struct event_constraint *c, *constraints[UNCORE_PMC_IDX_MAX]; 2038 struct event_constraint *c, *constraints[UNCORE_PMC_IDX_MAX];
@@ -1407,8 +2454,7 @@ static bool pcidrv_registered;
1407/* 2454/*
1408 * add a pci uncore device 2455 * add a pci uncore device
1409 */ 2456 */
1410static int __devinit uncore_pci_add(struct intel_uncore_type *type, 2457static int __devinit uncore_pci_add(struct intel_uncore_type *type, struct pci_dev *pdev)
1411 struct pci_dev *pdev)
1412{ 2458{
1413 struct intel_uncore_pmu *pmu; 2459 struct intel_uncore_pmu *pmu;
1414 struct intel_uncore_box *box; 2460 struct intel_uncore_box *box;
@@ -1485,6 +2531,7 @@ static int __devinit uncore_pci_probe(struct pci_dev *pdev,
1485 struct intel_uncore_type *type; 2531 struct intel_uncore_type *type;
1486 2532
1487 type = (struct intel_uncore_type *)id->driver_data; 2533 type = (struct intel_uncore_type *)id->driver_data;
2534
1488 return uncore_pci_add(type, pdev); 2535 return uncore_pci_add(type, pdev);
1489} 2536}
1490 2537
@@ -1612,8 +2659,8 @@ static int __cpuinit uncore_cpu_prepare(int cpu, int phys_id)
1612 return 0; 2659 return 0;
1613} 2660}
1614 2661
1615static void __cpuinit uncore_change_context(struct intel_uncore_type **uncores, 2662static void __cpuinit
1616 int old_cpu, int new_cpu) 2663uncore_change_context(struct intel_uncore_type **uncores, int old_cpu, int new_cpu)
1617{ 2664{
1618 struct intel_uncore_type *type; 2665 struct intel_uncore_type *type;
1619 struct intel_uncore_pmu *pmu; 2666 struct intel_uncore_pmu *pmu;
@@ -1694,8 +2741,8 @@ static void __cpuinit uncore_event_init_cpu(int cpu)
1694 uncore_change_context(pci_uncores, -1, cpu); 2741 uncore_change_context(pci_uncores, -1, cpu);
1695} 2742}
1696 2743
1697static int __cpuinit uncore_cpu_notifier(struct notifier_block *self, 2744static int
1698 unsigned long action, void *hcpu) 2745 __cpuinit uncore_cpu_notifier(struct notifier_block *self, unsigned long action, void *hcpu)
1699{ 2746{
1700 unsigned int cpu = (long)hcpu; 2747 unsigned int cpu = (long)hcpu;
1701 2748
@@ -1732,12 +2779,12 @@ static int __cpuinit uncore_cpu_notifier(struct notifier_block *self,
1732} 2779}
1733 2780
1734static struct notifier_block uncore_cpu_nb __cpuinitdata = { 2781static struct notifier_block uncore_cpu_nb __cpuinitdata = {
1735 .notifier_call = uncore_cpu_notifier, 2782 .notifier_call = uncore_cpu_notifier,
1736 /* 2783 /*
1737 * to migrate uncore events, our notifier should be executed 2784 * to migrate uncore events, our notifier should be executed
1738 * before perf core's notifier. 2785 * before perf core's notifier.
1739 */ 2786 */
1740 .priority = CPU_PRI_PERF + 1, 2787 .priority = CPU_PRI_PERF + 1,
1741}; 2788};
1742 2789
1743static void __init uncore_cpu_setup(void *dummy) 2790static void __init uncore_cpu_setup(void *dummy)
@@ -1767,6 +2814,9 @@ static int __init uncore_cpu_init(void)
1767 snbep_uncore_cbox.num_boxes = max_cores; 2814 snbep_uncore_cbox.num_boxes = max_cores;
1768 msr_uncores = snbep_msr_uncores; 2815 msr_uncores = snbep_msr_uncores;
1769 break; 2816 break;
2817 case 46:
2818 msr_uncores = nhmex_msr_uncores;
2819 break;
1770 default: 2820 default:
1771 return 0; 2821 return 0;
1772 } 2822 }
diff --git a/arch/x86/kernel/cpu/perf_event_intel_uncore.h b/arch/x86/kernel/cpu/perf_event_intel_uncore.h
index b13e9ea81def..f3851892e077 100644
--- a/arch/x86/kernel/cpu/perf_event_intel_uncore.h
+++ b/arch/x86/kernel/cpu/perf_event_intel_uncore.h
@@ -5,8 +5,6 @@
5#include "perf_event.h" 5#include "perf_event.h"
6 6
7#define UNCORE_PMU_NAME_LEN 32 7#define UNCORE_PMU_NAME_LEN 32
8#define UNCORE_BOX_HASH_SIZE 8
9
10#define UNCORE_PMU_HRTIMER_INTERVAL (60 * NSEC_PER_SEC) 8#define UNCORE_PMU_HRTIMER_INTERVAL (60 * NSEC_PER_SEC)
11 9
12#define UNCORE_FIXED_EVENT 0xff 10#define UNCORE_FIXED_EVENT 0xff
@@ -115,6 +113,10 @@
115 SNBEP_PCU_MSR_PMON_CTL_OCC_INVERT | \ 113 SNBEP_PCU_MSR_PMON_CTL_OCC_INVERT | \
116 SNBEP_PCU_MSR_PMON_CTL_OCC_EDGE_DET) 114 SNBEP_PCU_MSR_PMON_CTL_OCC_EDGE_DET)
117 115
116#define SNBEP_QPI_PCI_PMON_RAW_EVENT_MASK \
117 (SNBEP_PMON_RAW_EVENT_MASK | \
118 SNBEP_PMON_CTL_EV_SEL_EXT)
119
118/* SNB-EP pci control register */ 120/* SNB-EP pci control register */
119#define SNBEP_PCI_PMON_BOX_CTL 0xf4 121#define SNBEP_PCI_PMON_BOX_CTL 0xf4
120#define SNBEP_PCI_PMON_CTL0 0xd8 122#define SNBEP_PCI_PMON_CTL0 0xd8
@@ -158,6 +160,193 @@
158#define SNBEP_PCU_MSR_CORE_C3_CTR 0x3fc 160#define SNBEP_PCU_MSR_CORE_C3_CTR 0x3fc
159#define SNBEP_PCU_MSR_CORE_C6_CTR 0x3fd 161#define SNBEP_PCU_MSR_CORE_C6_CTR 0x3fd
160 162
163/* NHM-EX event control */
164#define NHMEX_PMON_CTL_EV_SEL_MASK 0x000000ff
165#define NHMEX_PMON_CTL_UMASK_MASK 0x0000ff00
166#define NHMEX_PMON_CTL_EN_BIT0 (1 << 0)
167#define NHMEX_PMON_CTL_EDGE_DET (1 << 18)
168#define NHMEX_PMON_CTL_PMI_EN (1 << 20)
169#define NHMEX_PMON_CTL_EN_BIT22 (1 << 22)
170#define NHMEX_PMON_CTL_INVERT (1 << 23)
171#define NHMEX_PMON_CTL_TRESH_MASK 0xff000000
172#define NHMEX_PMON_RAW_EVENT_MASK (NHMEX_PMON_CTL_EV_SEL_MASK | \
173 NHMEX_PMON_CTL_UMASK_MASK | \
174 NHMEX_PMON_CTL_EDGE_DET | \
175 NHMEX_PMON_CTL_INVERT | \
176 NHMEX_PMON_CTL_TRESH_MASK)
177
178/* NHM-EX Ubox */
179#define NHMEX_U_MSR_PMON_GLOBAL_CTL 0xc00
180#define NHMEX_U_MSR_PMON_CTR 0xc11
181#define NHMEX_U_MSR_PMON_EV_SEL 0xc10
182
183#define NHMEX_U_PMON_GLOBAL_EN (1 << 0)
184#define NHMEX_U_PMON_GLOBAL_PMI_CORE_SEL 0x0000001e
185#define NHMEX_U_PMON_GLOBAL_EN_ALL (1 << 28)
186#define NHMEX_U_PMON_GLOBAL_RST_ALL (1 << 29)
187#define NHMEX_U_PMON_GLOBAL_FRZ_ALL (1 << 31)
188
189#define NHMEX_U_PMON_RAW_EVENT_MASK \
190 (NHMEX_PMON_CTL_EV_SEL_MASK | \
191 NHMEX_PMON_CTL_EDGE_DET)
192
193/* NHM-EX Cbox */
194#define NHMEX_C0_MSR_PMON_GLOBAL_CTL 0xd00
195#define NHMEX_C0_MSR_PMON_CTR0 0xd11
196#define NHMEX_C0_MSR_PMON_EV_SEL0 0xd10
197#define NHMEX_C_MSR_OFFSET 0x20
198
199/* NHM-EX Bbox */
200#define NHMEX_B0_MSR_PMON_GLOBAL_CTL 0xc20
201#define NHMEX_B0_MSR_PMON_CTR0 0xc31
202#define NHMEX_B0_MSR_PMON_CTL0 0xc30
203#define NHMEX_B_MSR_OFFSET 0x40
204#define NHMEX_B0_MSR_MATCH 0xe45
205#define NHMEX_B0_MSR_MASK 0xe46
206#define NHMEX_B1_MSR_MATCH 0xe4d
207#define NHMEX_B1_MSR_MASK 0xe4e
208
209#define NHMEX_B_PMON_CTL_EN (1 << 0)
210#define NHMEX_B_PMON_CTL_EV_SEL_SHIFT 1
211#define NHMEX_B_PMON_CTL_EV_SEL_MASK \
212 (0x1f << NHMEX_B_PMON_CTL_EV_SEL_SHIFT)
213#define NHMEX_B_PMON_CTR_SHIFT 6
214#define NHMEX_B_PMON_CTR_MASK \
215 (0x3 << NHMEX_B_PMON_CTR_SHIFT)
216#define NHMEX_B_PMON_RAW_EVENT_MASK \
217 (NHMEX_B_PMON_CTL_EV_SEL_MASK | \
218 NHMEX_B_PMON_CTR_MASK)
219
220/* NHM-EX Sbox */
221#define NHMEX_S0_MSR_PMON_GLOBAL_CTL 0xc40
222#define NHMEX_S0_MSR_PMON_CTR0 0xc51
223#define NHMEX_S0_MSR_PMON_CTL0 0xc50
224#define NHMEX_S_MSR_OFFSET 0x80
225#define NHMEX_S0_MSR_MM_CFG 0xe48
226#define NHMEX_S0_MSR_MATCH 0xe49
227#define NHMEX_S0_MSR_MASK 0xe4a
228#define NHMEX_S1_MSR_MM_CFG 0xe58
229#define NHMEX_S1_MSR_MATCH 0xe59
230#define NHMEX_S1_MSR_MASK 0xe5a
231
232#define NHMEX_S_PMON_MM_CFG_EN (0x1ULL << 63)
233
234/* NHM-EX Mbox */
235#define NHMEX_M0_MSR_GLOBAL_CTL 0xca0
236#define NHMEX_M0_MSR_PMU_DSP 0xca5
237#define NHMEX_M0_MSR_PMU_ISS 0xca6
238#define NHMEX_M0_MSR_PMU_MAP 0xca7
239#define NHMEX_M0_MSR_PMU_MSC_THR 0xca8
240#define NHMEX_M0_MSR_PMU_PGT 0xca9
241#define NHMEX_M0_MSR_PMU_PLD 0xcaa
242#define NHMEX_M0_MSR_PMU_ZDP_CTL_FVC 0xcab
243#define NHMEX_M0_MSR_PMU_CTL0 0xcb0
244#define NHMEX_M0_MSR_PMU_CNT0 0xcb1
245#define NHMEX_M_MSR_OFFSET 0x40
246#define NHMEX_M0_MSR_PMU_MM_CFG 0xe54
247#define NHMEX_M1_MSR_PMU_MM_CFG 0xe5c
248
249#define NHMEX_M_PMON_MM_CFG_EN (1ULL << 63)
250#define NHMEX_M_PMON_ADDR_MATCH_MASK 0x3ffffffffULL
251#define NHMEX_M_PMON_ADDR_MASK_MASK 0x7ffffffULL
252#define NHMEX_M_PMON_ADDR_MASK_SHIFT 34
253
254#define NHMEX_M_PMON_CTL_EN (1 << 0)
255#define NHMEX_M_PMON_CTL_PMI_EN (1 << 1)
256#define NHMEX_M_PMON_CTL_COUNT_MODE_SHIFT 2
257#define NHMEX_M_PMON_CTL_COUNT_MODE_MASK \
258 (0x3 << NHMEX_M_PMON_CTL_COUNT_MODE_SHIFT)
259#define NHMEX_M_PMON_CTL_STORAGE_MODE_SHIFT 4
260#define NHMEX_M_PMON_CTL_STORAGE_MODE_MASK \
261 (0x3 << NHMEX_M_PMON_CTL_STORAGE_MODE_SHIFT)
262#define NHMEX_M_PMON_CTL_WRAP_MODE (1 << 6)
263#define NHMEX_M_PMON_CTL_FLAG_MODE (1 << 7)
264#define NHMEX_M_PMON_CTL_INC_SEL_SHIFT 9
265#define NHMEX_M_PMON_CTL_INC_SEL_MASK \
266 (0x1f << NHMEX_M_PMON_CTL_INC_SEL_SHIFT)
267#define NHMEX_M_PMON_CTL_SET_FLAG_SEL_SHIFT 19
268#define NHMEX_M_PMON_CTL_SET_FLAG_SEL_MASK \
269 (0x7 << NHMEX_M_PMON_CTL_SET_FLAG_SEL_SHIFT)
270#define NHMEX_M_PMON_RAW_EVENT_MASK \
271 (NHMEX_M_PMON_CTL_COUNT_MODE_MASK | \
272 NHMEX_M_PMON_CTL_STORAGE_MODE_MASK | \
273 NHMEX_M_PMON_CTL_WRAP_MODE | \
274 NHMEX_M_PMON_CTL_FLAG_MODE | \
275 NHMEX_M_PMON_CTL_INC_SEL_MASK | \
276 NHMEX_M_PMON_CTL_SET_FLAG_SEL_MASK)
277
278
279#define NHMEX_M_PMON_ZDP_CTL_FVC_FVID_MASK 0x1f
280#define NHMEX_M_PMON_ZDP_CTL_FVC_BCMD_MASK (0x7 << 5)
281#define NHMEX_M_PMON_ZDP_CTL_FVC_RSP_MASK (0x7 << 8)
282#define NHMEX_M_PMON_ZDP_CTL_FVC_PBOX_INIT_ERR (1 << 23)
283#define NHMEX_M_PMON_ZDP_CTL_FVC_MASK \
284 (NHMEX_M_PMON_ZDP_CTL_FVC_FVID_MASK | \
285 NHMEX_M_PMON_ZDP_CTL_FVC_BCMD_MASK | \
286 NHMEX_M_PMON_ZDP_CTL_FVC_RSP_MASK | \
287 NHMEX_M_PMON_ZDP_CTL_FVC_PBOX_INIT_ERR)
288#define NHMEX_M_PMON_ZDP_CTL_FVC_EVENT_MASK(n) (0x7 << (11 + 3 * (n)))
289
290/*
291 * use the 9~13 bits to select event If the 7th bit is not set,
292 * otherwise use the 19~21 bits to select event.
293 */
294#define MBOX_INC_SEL(x) ((x) << NHMEX_M_PMON_CTL_INC_SEL_SHIFT)
295#define MBOX_SET_FLAG_SEL(x) (((x) << NHMEX_M_PMON_CTL_SET_FLAG_SEL_SHIFT) | \
296 NHMEX_M_PMON_CTL_FLAG_MODE)
297#define MBOX_INC_SEL_MASK (NHMEX_M_PMON_CTL_INC_SEL_MASK | \
298 NHMEX_M_PMON_CTL_FLAG_MODE)
299#define MBOX_SET_FLAG_SEL_MASK (NHMEX_M_PMON_CTL_SET_FLAG_SEL_MASK | \
300 NHMEX_M_PMON_CTL_FLAG_MODE)
301#define MBOX_INC_SEL_EXTAR_REG(c, r) \
302 EVENT_EXTRA_REG(MBOX_INC_SEL(c), NHMEX_M0_MSR_PMU_##r, \
303 MBOX_INC_SEL_MASK, (u64)-1, NHMEX_M_##r)
304#define MBOX_SET_FLAG_SEL_EXTRA_REG(c, r) \
305 EVENT_EXTRA_REG(MBOX_SET_FLAG_SEL(c), NHMEX_M0_MSR_PMU_##r, \
306 MBOX_SET_FLAG_SEL_MASK, \
307 (u64)-1, NHMEX_M_##r)
308
309/* NHM-EX Rbox */
310#define NHMEX_R_MSR_GLOBAL_CTL 0xe00
311#define NHMEX_R_MSR_PMON_CTL0 0xe10
312#define NHMEX_R_MSR_PMON_CNT0 0xe11
313#define NHMEX_R_MSR_OFFSET 0x20
314
315#define NHMEX_R_MSR_PORTN_QLX_CFG(n) \
316 ((n) < 4 ? (0xe0c + (n)) : (0xe2c + (n) - 4))
317#define NHMEX_R_MSR_PORTN_IPERF_CFG0(n) (0xe04 + (n))
318#define NHMEX_R_MSR_PORTN_IPERF_CFG1(n) (0xe24 + (n))
319#define NHMEX_R_MSR_PORTN_XBR_OFFSET(n) \
320 (((n) < 4 ? 0 : 0x10) + (n) * 4)
321#define NHMEX_R_MSR_PORTN_XBR_SET1_MM_CFG(n) \
322 (0xe60 + NHMEX_R_MSR_PORTN_XBR_OFFSET(n))
323#define NHMEX_R_MSR_PORTN_XBR_SET1_MATCH(n) \
324 (NHMEX_R_MSR_PORTN_XBR_SET1_MM_CFG(n) + 1)
325#define NHMEX_R_MSR_PORTN_XBR_SET1_MASK(n) \
326 (NHMEX_R_MSR_PORTN_XBR_SET1_MM_CFG(n) + 2)
327#define NHMEX_R_MSR_PORTN_XBR_SET2_MM_CFG(n) \
328 (0xe70 + NHMEX_R_MSR_PORTN_XBR_OFFSET(n))
329#define NHMEX_R_MSR_PORTN_XBR_SET2_MATCH(n) \
330 (NHMEX_R_MSR_PORTN_XBR_SET2_MM_CFG(n) + 1)
331#define NHMEX_R_MSR_PORTN_XBR_SET2_MASK(n) \
332 (NHMEX_R_MSR_PORTN_XBR_SET2_MM_CFG(n) + 2)
333
334#define NHMEX_R_PMON_CTL_EN (1 << 0)
335#define NHMEX_R_PMON_CTL_EV_SEL_SHIFT 1
336#define NHMEX_R_PMON_CTL_EV_SEL_MASK \
337 (0x1f << NHMEX_R_PMON_CTL_EV_SEL_SHIFT)
338#define NHMEX_R_PMON_CTL_PMI_EN (1 << 6)
339#define NHMEX_R_PMON_RAW_EVENT_MASK NHMEX_R_PMON_CTL_EV_SEL_MASK
340
341/* NHM-EX Wbox */
342#define NHMEX_W_MSR_GLOBAL_CTL 0xc80
343#define NHMEX_W_MSR_PMON_CNT0 0xc90
344#define NHMEX_W_MSR_PMON_EVT_SEL0 0xc91
345#define NHMEX_W_MSR_PMON_FIXED_CTR 0x394
346#define NHMEX_W_MSR_PMON_FIXED_CTL 0x395
347
348#define NHMEX_W_PMON_GLOBAL_FIXED_EN (1ULL << 31)
349
161struct intel_uncore_ops; 350struct intel_uncore_ops;
162struct intel_uncore_pmu; 351struct intel_uncore_pmu;
163struct intel_uncore_box; 352struct intel_uncore_box;
@@ -178,6 +367,7 @@ struct intel_uncore_type {
178 unsigned msr_offset; 367 unsigned msr_offset;
179 unsigned num_shared_regs:8; 368 unsigned num_shared_regs:8;
180 unsigned single_fixed:1; 369 unsigned single_fixed:1;
370 unsigned pair_ctr_ctl:1;
181 struct event_constraint unconstrainted; 371 struct event_constraint unconstrainted;
182 struct event_constraint *constraints; 372 struct event_constraint *constraints;
183 struct intel_uncore_pmu *pmus; 373 struct intel_uncore_pmu *pmus;
@@ -213,7 +403,7 @@ struct intel_uncore_pmu {
213 403
214struct intel_uncore_extra_reg { 404struct intel_uncore_extra_reg {
215 raw_spinlock_t lock; 405 raw_spinlock_t lock;
216 u64 config1; 406 u64 config, config1, config2;
217 atomic_t ref; 407 atomic_t ref;
218}; 408};
219 409
@@ -323,14 +513,16 @@ unsigned uncore_msr_fixed_ctr(struct intel_uncore_box *box)
323static inline 513static inline
324unsigned uncore_msr_event_ctl(struct intel_uncore_box *box, int idx) 514unsigned uncore_msr_event_ctl(struct intel_uncore_box *box, int idx)
325{ 515{
326 return idx + box->pmu->type->event_ctl + 516 return box->pmu->type->event_ctl +
517 (box->pmu->type->pair_ctr_ctl ? 2 * idx : idx) +
327 box->pmu->type->msr_offset * box->pmu->pmu_idx; 518 box->pmu->type->msr_offset * box->pmu->pmu_idx;
328} 519}
329 520
330static inline 521static inline
331unsigned uncore_msr_perf_ctr(struct intel_uncore_box *box, int idx) 522unsigned uncore_msr_perf_ctr(struct intel_uncore_box *box, int idx)
332{ 523{
333 return idx + box->pmu->type->perf_ctr + 524 return box->pmu->type->perf_ctr +
525 (box->pmu->type->pair_ctr_ctl ? 2 * idx : idx) +
334 box->pmu->type->msr_offset * box->pmu->pmu_idx; 526 box->pmu->type->msr_offset * box->pmu->pmu_idx;
335} 527}
336 528
@@ -422,3 +614,8 @@ static inline void uncore_box_init(struct intel_uncore_box *box)
422 box->pmu->type->ops->init_box(box); 614 box->pmu->type->ops->init_box(box);
423 } 615 }
424} 616}
617
618static inline bool uncore_box_is_fake(struct intel_uncore_box *box)
619{
620 return (box->phys_id < 0);
621}
diff --git a/kernel/events/uprobes.c b/kernel/events/uprobes.c
index f93532748bca..c08a22d02f72 100644
--- a/kernel/events/uprobes.c
+++ b/kernel/events/uprobes.c
@@ -32,6 +32,7 @@
32#include <linux/swap.h> /* try_to_free_swap */ 32#include <linux/swap.h> /* try_to_free_swap */
33#include <linux/ptrace.h> /* user_enable_single_step */ 33#include <linux/ptrace.h> /* user_enable_single_step */
34#include <linux/kdebug.h> /* notifier mechanism */ 34#include <linux/kdebug.h> /* notifier mechanism */
35#include "../../mm/internal.h" /* munlock_vma_page */
35 36
36#include <linux/uprobes.h> 37#include <linux/uprobes.h>
37 38
@@ -112,14 +113,14 @@ static bool valid_vma(struct vm_area_struct *vma, bool is_register)
112 return false; 113 return false;
113} 114}
114 115
115static loff_t vma_address(struct vm_area_struct *vma, loff_t offset) 116static unsigned long offset_to_vaddr(struct vm_area_struct *vma, loff_t offset)
116{ 117{
117 loff_t vaddr; 118 return vma->vm_start + offset - ((loff_t)vma->vm_pgoff << PAGE_SHIFT);
118 119}
119 vaddr = vma->vm_start + offset;
120 vaddr -= vma->vm_pgoff << PAGE_SHIFT;
121 120
122 return vaddr; 121static loff_t vaddr_to_offset(struct vm_area_struct *vma, unsigned long vaddr)
122{
123 return ((loff_t)vma->vm_pgoff << PAGE_SHIFT) + (vaddr - vma->vm_start);
123} 124}
124 125
125/** 126/**
@@ -127,25 +128,27 @@ static loff_t vma_address(struct vm_area_struct *vma, loff_t offset)
127 * based on replace_page in mm/ksm.c 128 * based on replace_page in mm/ksm.c
128 * 129 *
129 * @vma: vma that holds the pte pointing to page 130 * @vma: vma that holds the pte pointing to page
131 * @addr: address the old @page is mapped at
130 * @page: the cowed page we are replacing by kpage 132 * @page: the cowed page we are replacing by kpage
131 * @kpage: the modified page we replace page by 133 * @kpage: the modified page we replace page by
132 * 134 *
133 * Returns 0 on success, -EFAULT on failure. 135 * Returns 0 on success, -EFAULT on failure.
134 */ 136 */
135static int __replace_page(struct vm_area_struct *vma, struct page *page, struct page *kpage) 137static int __replace_page(struct vm_area_struct *vma, unsigned long addr,
138 struct page *page, struct page *kpage)
136{ 139{
137 struct mm_struct *mm = vma->vm_mm; 140 struct mm_struct *mm = vma->vm_mm;
138 unsigned long addr;
139 spinlock_t *ptl; 141 spinlock_t *ptl;
140 pte_t *ptep; 142 pte_t *ptep;
143 int err;
141 144
142 addr = page_address_in_vma(page, vma); 145 /* For try_to_free_swap() and munlock_vma_page() below */
143 if (addr == -EFAULT) 146 lock_page(page);
144 return -EFAULT;
145 147
148 err = -EAGAIN;
146 ptep = page_check_address(page, mm, addr, &ptl, 0); 149 ptep = page_check_address(page, mm, addr, &ptl, 0);
147 if (!ptep) 150 if (!ptep)
148 return -EAGAIN; 151 goto unlock;
149 152
150 get_page(kpage); 153 get_page(kpage);
151 page_add_new_anon_rmap(kpage, vma, addr); 154 page_add_new_anon_rmap(kpage, vma, addr);
@@ -162,10 +165,16 @@ static int __replace_page(struct vm_area_struct *vma, struct page *page, struct
162 page_remove_rmap(page); 165 page_remove_rmap(page);
163 if (!page_mapped(page)) 166 if (!page_mapped(page))
164 try_to_free_swap(page); 167 try_to_free_swap(page);
165 put_page(page);
166 pte_unmap_unlock(ptep, ptl); 168 pte_unmap_unlock(ptep, ptl);
167 169
168 return 0; 170 if (vma->vm_flags & VM_LOCKED)
171 munlock_vma_page(page);
172 put_page(page);
173
174 err = 0;
175 unlock:
176 unlock_page(page);
177 return err;
169} 178}
170 179
171/** 180/**
@@ -206,45 +215,23 @@ static int write_opcode(struct arch_uprobe *auprobe, struct mm_struct *mm,
206 unsigned long vaddr, uprobe_opcode_t opcode) 215 unsigned long vaddr, uprobe_opcode_t opcode)
207{ 216{
208 struct page *old_page, *new_page; 217 struct page *old_page, *new_page;
209 struct address_space *mapping;
210 void *vaddr_old, *vaddr_new; 218 void *vaddr_old, *vaddr_new;
211 struct vm_area_struct *vma; 219 struct vm_area_struct *vma;
212 struct uprobe *uprobe;
213 int ret; 220 int ret;
221
214retry: 222retry:
215 /* Read the page with vaddr into memory */ 223 /* Read the page with vaddr into memory */
216 ret = get_user_pages(NULL, mm, vaddr, 1, 0, 0, &old_page, &vma); 224 ret = get_user_pages(NULL, mm, vaddr, 1, 0, 0, &old_page, &vma);
217 if (ret <= 0) 225 if (ret <= 0)
218 return ret; 226 return ret;
219 227
220 ret = -EINVAL;
221
222 /*
223 * We are interested in text pages only. Our pages of interest
224 * should be mapped for read and execute only. We desist from
225 * adding probes in write mapped pages since the breakpoints
226 * might end up in the file copy.
227 */
228 if (!valid_vma(vma, is_swbp_insn(&opcode)))
229 goto put_out;
230
231 uprobe = container_of(auprobe, struct uprobe, arch);
232 mapping = uprobe->inode->i_mapping;
233 if (mapping != vma->vm_file->f_mapping)
234 goto put_out;
235
236 ret = -ENOMEM; 228 ret = -ENOMEM;
237 new_page = alloc_page_vma(GFP_HIGHUSER_MOVABLE, vma, vaddr); 229 new_page = alloc_page_vma(GFP_HIGHUSER_MOVABLE, vma, vaddr);
238 if (!new_page) 230 if (!new_page)
239 goto put_out; 231 goto put_old;
240 232
241 __SetPageUptodate(new_page); 233 __SetPageUptodate(new_page);
242 234
243 /*
244 * lock page will serialize against do_wp_page()'s
245 * PageAnon() handling
246 */
247 lock_page(old_page);
248 /* copy the page now that we've got it stable */ 235 /* copy the page now that we've got it stable */
249 vaddr_old = kmap_atomic(old_page); 236 vaddr_old = kmap_atomic(old_page);
250 vaddr_new = kmap_atomic(new_page); 237 vaddr_new = kmap_atomic(new_page);
@@ -257,17 +244,13 @@ retry:
257 244
258 ret = anon_vma_prepare(vma); 245 ret = anon_vma_prepare(vma);
259 if (ret) 246 if (ret)
260 goto unlock_out; 247 goto put_new;
261 248
262 lock_page(new_page); 249 ret = __replace_page(vma, vaddr, old_page, new_page);
263 ret = __replace_page(vma, old_page, new_page);
264 unlock_page(new_page);
265 250
266unlock_out: 251put_new:
267 unlock_page(old_page);
268 page_cache_release(new_page); 252 page_cache_release(new_page);
269 253put_old:
270put_out:
271 put_page(old_page); 254 put_page(old_page);
272 255
273 if (unlikely(ret == -EAGAIN)) 256 if (unlikely(ret == -EAGAIN))
@@ -791,7 +774,7 @@ build_map_info(struct address_space *mapping, loff_t offset, bool is_register)
791 curr = info; 774 curr = info;
792 775
793 info->mm = vma->vm_mm; 776 info->mm = vma->vm_mm;
794 info->vaddr = vma_address(vma, offset); 777 info->vaddr = offset_to_vaddr(vma, offset);
795 } 778 }
796 mutex_unlock(&mapping->i_mmap_mutex); 779 mutex_unlock(&mapping->i_mmap_mutex);
797 780
@@ -839,12 +822,13 @@ static int register_for_each_vma(struct uprobe *uprobe, bool is_register)
839 goto free; 822 goto free;
840 823
841 down_write(&mm->mmap_sem); 824 down_write(&mm->mmap_sem);
842 vma = find_vma(mm, (unsigned long)info->vaddr); 825 vma = find_vma(mm, info->vaddr);
843 if (!vma || !valid_vma(vma, is_register)) 826 if (!vma || !valid_vma(vma, is_register) ||
827 vma->vm_file->f_mapping->host != uprobe->inode)
844 goto unlock; 828 goto unlock;
845 829
846 if (vma->vm_file->f_mapping->host != uprobe->inode || 830 if (vma->vm_start > info->vaddr ||
847 vma_address(vma, uprobe->offset) != info->vaddr) 831 vaddr_to_offset(vma, info->vaddr) != uprobe->offset)
848 goto unlock; 832 goto unlock;
849 833
850 if (is_register) { 834 if (is_register) {
@@ -960,59 +944,66 @@ void uprobe_unregister(struct inode *inode, loff_t offset, struct uprobe_consume
960 put_uprobe(uprobe); 944 put_uprobe(uprobe);
961} 945}
962 946
963/* 947static struct rb_node *
964 * Of all the nodes that correspond to the given inode, return the node 948find_node_in_range(struct inode *inode, loff_t min, loff_t max)
965 * with the least offset.
966 */
967static struct rb_node *find_least_offset_node(struct inode *inode)
968{ 949{
969 struct uprobe u = { .inode = inode, .offset = 0};
970 struct rb_node *n = uprobes_tree.rb_node; 950 struct rb_node *n = uprobes_tree.rb_node;
971 struct rb_node *close_node = NULL;
972 struct uprobe *uprobe;
973 int match;
974 951
975 while (n) { 952 while (n) {
976 uprobe = rb_entry(n, struct uprobe, rb_node); 953 struct uprobe *u = rb_entry(n, struct uprobe, rb_node);
977 match = match_uprobe(&u, uprobe);
978
979 if (uprobe->inode == inode)
980 close_node = n;
981
982 if (!match)
983 return close_node;
984 954
985 if (match < 0) 955 if (inode < u->inode) {
986 n = n->rb_left; 956 n = n->rb_left;
987 else 957 } else if (inode > u->inode) {
988 n = n->rb_right; 958 n = n->rb_right;
959 } else {
960 if (max < u->offset)
961 n = n->rb_left;
962 else if (min > u->offset)
963 n = n->rb_right;
964 else
965 break;
966 }
989 } 967 }
990 968
991 return close_node; 969 return n;
992} 970}
993 971
994/* 972/*
995 * For a given inode, build a list of probes that need to be inserted. 973 * For a given range in vma, build a list of probes that need to be inserted.
996 */ 974 */
997static void build_probe_list(struct inode *inode, struct list_head *head) 975static void build_probe_list(struct inode *inode,
976 struct vm_area_struct *vma,
977 unsigned long start, unsigned long end,
978 struct list_head *head)
998{ 979{
999 struct uprobe *uprobe; 980 loff_t min, max;
1000 unsigned long flags; 981 unsigned long flags;
1001 struct rb_node *n; 982 struct rb_node *n, *t;
1002 983 struct uprobe *u;
1003 spin_lock_irqsave(&uprobes_treelock, flags);
1004
1005 n = find_least_offset_node(inode);
1006 984
1007 for (; n; n = rb_next(n)) { 985 INIT_LIST_HEAD(head);
1008 uprobe = rb_entry(n, struct uprobe, rb_node); 986 min = vaddr_to_offset(vma, start);
1009 if (uprobe->inode != inode) 987 max = min + (end - start) - 1;
1010 break;
1011 988
1012 list_add(&uprobe->pending_list, head); 989 spin_lock_irqsave(&uprobes_treelock, flags);
1013 atomic_inc(&uprobe->ref); 990 n = find_node_in_range(inode, min, max);
991 if (n) {
992 for (t = n; t; t = rb_prev(t)) {
993 u = rb_entry(t, struct uprobe, rb_node);
994 if (u->inode != inode || u->offset < min)
995 break;
996 list_add(&u->pending_list, head);
997 atomic_inc(&u->ref);
998 }
999 for (t = n; (t = rb_next(t)); ) {
1000 u = rb_entry(t, struct uprobe, rb_node);
1001 if (u->inode != inode || u->offset > max)
1002 break;
1003 list_add(&u->pending_list, head);
1004 atomic_inc(&u->ref);
1005 }
1014 } 1006 }
1015
1016 spin_unlock_irqrestore(&uprobes_treelock, flags); 1007 spin_unlock_irqrestore(&uprobes_treelock, flags);
1017} 1008}
1018 1009
@@ -1031,7 +1022,7 @@ static void build_probe_list(struct inode *inode, struct list_head *head)
1031int uprobe_mmap(struct vm_area_struct *vma) 1022int uprobe_mmap(struct vm_area_struct *vma)
1032{ 1023{
1033 struct list_head tmp_list; 1024 struct list_head tmp_list;
1034 struct uprobe *uprobe; 1025 struct uprobe *uprobe, *u;
1035 struct inode *inode; 1026 struct inode *inode;
1036 int ret, count; 1027 int ret, count;
1037 1028
@@ -1042,21 +1033,15 @@ int uprobe_mmap(struct vm_area_struct *vma)
1042 if (!inode) 1033 if (!inode)
1043 return 0; 1034 return 0;
1044 1035
1045 INIT_LIST_HEAD(&tmp_list);
1046 mutex_lock(uprobes_mmap_hash(inode)); 1036 mutex_lock(uprobes_mmap_hash(inode));
1047 build_probe_list(inode, &tmp_list); 1037 build_probe_list(inode, vma, vma->vm_start, vma->vm_end, &tmp_list);
1048 1038
1049 ret = 0; 1039 ret = 0;
1050 count = 0; 1040 count = 0;
1051 1041
1052 list_for_each_entry(uprobe, &tmp_list, pending_list) { 1042 list_for_each_entry_safe(uprobe, u, &tmp_list, pending_list) {
1053 if (!ret) { 1043 if (!ret) {
1054 loff_t vaddr = vma_address(vma, uprobe->offset); 1044 unsigned long vaddr = offset_to_vaddr(vma, uprobe->offset);
1055
1056 if (vaddr < vma->vm_start || vaddr >= vma->vm_end) {
1057 put_uprobe(uprobe);
1058 continue;
1059 }
1060 1045
1061 ret = install_breakpoint(uprobe, vma->vm_mm, vma, vaddr); 1046 ret = install_breakpoint(uprobe, vma->vm_mm, vma, vaddr);
1062 /* 1047 /*
@@ -1097,12 +1082,15 @@ int uprobe_mmap(struct vm_area_struct *vma)
1097void uprobe_munmap(struct vm_area_struct *vma, unsigned long start, unsigned long end) 1082void uprobe_munmap(struct vm_area_struct *vma, unsigned long start, unsigned long end)
1098{ 1083{
1099 struct list_head tmp_list; 1084 struct list_head tmp_list;
1100 struct uprobe *uprobe; 1085 struct uprobe *uprobe, *u;
1101 struct inode *inode; 1086 struct inode *inode;
1102 1087
1103 if (!atomic_read(&uprobe_events) || !valid_vma(vma, false)) 1088 if (!atomic_read(&uprobe_events) || !valid_vma(vma, false))
1104 return; 1089 return;
1105 1090
1091 if (!atomic_read(&vma->vm_mm->mm_users)) /* called by mmput() ? */
1092 return;
1093
1106 if (!atomic_read(&vma->vm_mm->uprobes_state.count)) 1094 if (!atomic_read(&vma->vm_mm->uprobes_state.count))
1107 return; 1095 return;
1108 1096
@@ -1110,21 +1098,17 @@ void uprobe_munmap(struct vm_area_struct *vma, unsigned long start, unsigned lon
1110 if (!inode) 1098 if (!inode)
1111 return; 1099 return;
1112 1100
1113 INIT_LIST_HEAD(&tmp_list);
1114 mutex_lock(uprobes_mmap_hash(inode)); 1101 mutex_lock(uprobes_mmap_hash(inode));
1115 build_probe_list(inode, &tmp_list); 1102 build_probe_list(inode, vma, start, end, &tmp_list);
1116 1103
1117 list_for_each_entry(uprobe, &tmp_list, pending_list) { 1104 list_for_each_entry_safe(uprobe, u, &tmp_list, pending_list) {
1118 loff_t vaddr = vma_address(vma, uprobe->offset); 1105 unsigned long vaddr = offset_to_vaddr(vma, uprobe->offset);
1119 1106 /*
1120 if (vaddr >= start && vaddr < end) { 1107 * An unregister could have removed the probe before
1121 /* 1108 * unmap. So check before we decrement the count.
1122 * An unregister could have removed the probe before 1109 */
1123 * unmap. So check before we decrement the count. 1110 if (is_swbp_at_addr(vma->vm_mm, vaddr) == 1)
1124 */ 1111 atomic_dec(&vma->vm_mm->uprobes_state.count);
1125 if (is_swbp_at_addr(vma->vm_mm, vaddr) == 1)
1126 atomic_dec(&vma->vm_mm->uprobes_state.count);
1127 }
1128 put_uprobe(uprobe); 1112 put_uprobe(uprobe);
1129 } 1113 }
1130 mutex_unlock(uprobes_mmap_hash(inode)); 1114 mutex_unlock(uprobes_mmap_hash(inode));
@@ -1463,12 +1447,9 @@ static struct uprobe *find_active_uprobe(unsigned long bp_vaddr, int *is_swbp)
1463 vma = find_vma(mm, bp_vaddr); 1447 vma = find_vma(mm, bp_vaddr);
1464 if (vma && vma->vm_start <= bp_vaddr) { 1448 if (vma && vma->vm_start <= bp_vaddr) {
1465 if (valid_vma(vma, false)) { 1449 if (valid_vma(vma, false)) {
1466 struct inode *inode; 1450 struct inode *inode = vma->vm_file->f_mapping->host;
1467 loff_t offset; 1451 loff_t offset = vaddr_to_offset(vma, bp_vaddr);
1468 1452
1469 inode = vma->vm_file->f_mapping->host;
1470 offset = bp_vaddr - vma->vm_start;
1471 offset += (vma->vm_pgoff << PAGE_SHIFT);
1472 uprobe = find_uprobe(inode, offset); 1453 uprobe = find_uprobe(inode, offset);
1473 } 1454 }
1474 1455
diff --git a/kernel/sched/core.c b/kernel/sched/core.c
index 5d011ef4c0df..d325c4b2dcbb 100644
--- a/kernel/sched/core.c
+++ b/kernel/sched/core.c
@@ -1910,12 +1910,12 @@ static inline void
1910prepare_task_switch(struct rq *rq, struct task_struct *prev, 1910prepare_task_switch(struct rq *rq, struct task_struct *prev,
1911 struct task_struct *next) 1911 struct task_struct *next)
1912{ 1912{
1913 trace_sched_switch(prev, next);
1913 sched_info_switch(prev, next); 1914 sched_info_switch(prev, next);
1914 perf_event_task_sched_out(prev, next); 1915 perf_event_task_sched_out(prev, next);
1915 fire_sched_out_preempt_notifiers(prev, next); 1916 fire_sched_out_preempt_notifiers(prev, next);
1916 prepare_lock_switch(rq, next); 1917 prepare_lock_switch(rq, next);
1917 prepare_arch_switch(next); 1918 prepare_arch_switch(next);
1918 trace_sched_switch(prev, next);
1919} 1919}
1920 1920
1921/** 1921/**
diff --git a/mm/mmap.c b/mm/mmap.c
index 3edfcdfa42d9..4fe2697339ed 100644
--- a/mm/mmap.c
+++ b/mm/mmap.c
@@ -2345,9 +2345,6 @@ int insert_vm_struct(struct mm_struct * mm, struct vm_area_struct * vma)
2345 security_vm_enough_memory_mm(mm, vma_pages(vma))) 2345 security_vm_enough_memory_mm(mm, vma_pages(vma)))
2346 return -ENOMEM; 2346 return -ENOMEM;
2347 2347
2348 if (vma->vm_file && uprobe_mmap(vma))
2349 return -EINVAL;
2350
2351 vma_link(mm, vma, prev, rb_link, rb_parent); 2348 vma_link(mm, vma, prev, rb_link, rb_parent);
2352 return 0; 2349 return 0;
2353} 2350}
@@ -2418,9 +2415,6 @@ struct vm_area_struct *copy_vma(struct vm_area_struct **vmap,
2418 if (new_vma->vm_file) { 2415 if (new_vma->vm_file) {
2419 get_file(new_vma->vm_file); 2416 get_file(new_vma->vm_file);
2420 2417
2421 if (uprobe_mmap(new_vma))
2422 goto out_free_mempol;
2423
2424 if (vma->vm_flags & VM_EXECUTABLE) 2418 if (vma->vm_flags & VM_EXECUTABLE)
2425 added_exe_file_vma(mm); 2419 added_exe_file_vma(mm);
2426 } 2420 }
diff --git a/tools/lib/traceevent/.gitignore b/tools/lib/traceevent/.gitignore
new file mode 100644
index 000000000000..35f56be5a4cd
--- /dev/null
+++ b/tools/lib/traceevent/.gitignore
@@ -0,0 +1 @@
TRACEEVENT-CFLAGS
diff --git a/tools/lib/traceevent/Makefile b/tools/lib/traceevent/Makefile
index 46c2f6b7b123..14131cb0522d 100644
--- a/tools/lib/traceevent/Makefile
+++ b/tools/lib/traceevent/Makefile
@@ -207,7 +207,7 @@ libtraceevent.so: $(PEVENT_LIB_OBJS)
207libtraceevent.a: $(PEVENT_LIB_OBJS) 207libtraceevent.a: $(PEVENT_LIB_OBJS)
208 $(Q)$(do_build_static_lib) 208 $(Q)$(do_build_static_lib)
209 209
210$(PEVENT_LIB_OBJS): %.o: $(src)/%.c 210$(PEVENT_LIB_OBJS): %.o: $(src)/%.c TRACEEVENT-CFLAGS
211 $(Q)$(do_fpic_compile) 211 $(Q)$(do_fpic_compile)
212 212
213define make_version.h 213define make_version.h
@@ -272,6 +272,16 @@ ifneq ($(dep_includes),)
272 include $(dep_includes) 272 include $(dep_includes)
273endif 273endif
274 274
275### Detect environment changes
276TRACK_CFLAGS = $(subst ','\'',$(CFLAGS)):$(ARCH):$(CROSS_COMPILE)
277
278TRACEEVENT-CFLAGS: force
279 @FLAGS='$(TRACK_CFLAGS)'; \
280 if test x"$$FLAGS" != x"`cat TRACEEVENT-CFLAGS 2>/dev/null`" ; then \
281 echo 1>&2 " * new build flags or cross compiler"; \
282 echo "$$FLAGS" >TRACEEVENT-CFLAGS; \
283 fi
284
275tags: force 285tags: force
276 $(RM) tags 286 $(RM) tags
277 find . -name '*.[ch]' | xargs ctags --extra=+f --c-kinds=+px \ 287 find . -name '*.[ch]' | xargs ctags --extra=+f --c-kinds=+px \
@@ -297,7 +307,7 @@ install: install_lib
297 307
298clean: 308clean:
299 $(RM) *.o *~ $(TARGETS) *.a *.so $(VERSION_FILES) .*.d 309 $(RM) *.o *~ $(TARGETS) *.a *.so $(VERSION_FILES) .*.d
300 $(RM) tags TAGS 310 $(RM) TRACEEVENT-CFLAGS tags TAGS
301 311
302endif # skip-makefile 312endif # skip-makefile
303 313
diff --git a/tools/perf/Makefile b/tools/perf/Makefile
index 75d74e5db8d5..77f124fe57ad 100644
--- a/tools/perf/Makefile
+++ b/tools/perf/Makefile
@@ -354,6 +354,7 @@ LIB_OBJS += $(OUTPUT)util/usage.o
354LIB_OBJS += $(OUTPUT)util/wrapper.o 354LIB_OBJS += $(OUTPUT)util/wrapper.o
355LIB_OBJS += $(OUTPUT)util/sigchain.o 355LIB_OBJS += $(OUTPUT)util/sigchain.o
356LIB_OBJS += $(OUTPUT)util/symbol.o 356LIB_OBJS += $(OUTPUT)util/symbol.o
357LIB_OBJS += $(OUTPUT)util/dso-test-data.o
357LIB_OBJS += $(OUTPUT)util/color.o 358LIB_OBJS += $(OUTPUT)util/color.o
358LIB_OBJS += $(OUTPUT)util/pager.o 359LIB_OBJS += $(OUTPUT)util/pager.o
359LIB_OBJS += $(OUTPUT)util/header.o 360LIB_OBJS += $(OUTPUT)util/header.o
@@ -803,6 +804,9 @@ $(OUTPUT)ui/browsers/map.o: ui/browsers/map.c $(OUTPUT)PERF-CFLAGS
803$(OUTPUT)util/rbtree.o: ../../lib/rbtree.c $(OUTPUT)PERF-CFLAGS 804$(OUTPUT)util/rbtree.o: ../../lib/rbtree.c $(OUTPUT)PERF-CFLAGS
804 $(QUIET_CC)$(CC) -o $@ -c $(ALL_CFLAGS) -DETC_PERFCONFIG='"$(ETC_PERFCONFIG_SQ)"' $< 805 $(QUIET_CC)$(CC) -o $@ -c $(ALL_CFLAGS) -DETC_PERFCONFIG='"$(ETC_PERFCONFIG_SQ)"' $<
805 806
807$(OUTPUT)util/parse-events.o: util/parse-events.c $(OUTPUT)PERF-CFLAGS
808 $(QUIET_CC)$(CC) -o $@ -c $(ALL_CFLAGS) -Wno-redundant-decls $<
809
806$(OUTPUT)util/scripting-engines/trace-event-perl.o: util/scripting-engines/trace-event-perl.c $(OUTPUT)PERF-CFLAGS 810$(OUTPUT)util/scripting-engines/trace-event-perl.o: util/scripting-engines/trace-event-perl.c $(OUTPUT)PERF-CFLAGS
807 $(QUIET_CC)$(CC) -o $@ -c $(ALL_CFLAGS) $(PERL_EMBED_CCOPTS) -Wno-redundant-decls -Wno-strict-prototypes -Wno-unused-parameter -Wno-shadow $< 811 $(QUIET_CC)$(CC) -o $@ -c $(ALL_CFLAGS) $(PERL_EMBED_CCOPTS) -Wno-redundant-decls -Wno-strict-prototypes -Wno-unused-parameter -Wno-shadow $<
808 812
diff --git a/tools/perf/builtin-test.c b/tools/perf/builtin-test.c
index 5ce30305462b..d909eb74a0eb 100644
--- a/tools/perf/builtin-test.c
+++ b/tools/perf/builtin-test.c
@@ -1142,6 +1142,10 @@ static struct test {
1142 .func = test__perf_pmu, 1142 .func = test__perf_pmu,
1143 }, 1143 },
1144 { 1144 {
1145 .desc = "Test dso data interface",
1146 .func = dso__test_data,
1147 },
1148 {
1145 .func = NULL, 1149 .func = NULL,
1146 }, 1150 },
1147}; 1151};
diff --git a/tools/perf/builtin-top.c b/tools/perf/builtin-top.c
index e3cab5f088f8..35e86c6df713 100644
--- a/tools/perf/builtin-top.c
+++ b/tools/perf/builtin-top.c
@@ -125,7 +125,7 @@ static int perf_top__parse_source(struct perf_top *top, struct hist_entry *he)
125 /* 125 /*
126 * We can't annotate with just /proc/kallsyms 126 * We can't annotate with just /proc/kallsyms
127 */ 127 */
128 if (map->dso->symtab_type == SYMTAB__KALLSYMS) { 128 if (map->dso->symtab_type == DSO_BINARY_TYPE__KALLSYMS) {
129 pr_err("Can't annotate %s: No vmlinux file was found in the " 129 pr_err("Can't annotate %s: No vmlinux file was found in the "
130 "path\n", sym->name); 130 "path\n", sym->name);
131 sleep(1); 131 sleep(1);
diff --git a/tools/perf/ui/browsers/hists.c b/tools/perf/ui/browsers/hists.c
index 482f0517b61e..413bd62eedb1 100644
--- a/tools/perf/ui/browsers/hists.c
+++ b/tools/perf/ui/browsers/hists.c
@@ -978,8 +978,8 @@ static int hist_browser__dump(struct hist_browser *browser)
978 fp = fopen(filename, "w"); 978 fp = fopen(filename, "w");
979 if (fp == NULL) { 979 if (fp == NULL) {
980 char bf[64]; 980 char bf[64];
981 strerror_r(errno, bf, sizeof(bf)); 981 const char *err = strerror_r(errno, bf, sizeof(bf));
982 ui_helpline__fpush("Couldn't write to %s: %s", filename, bf); 982 ui_helpline__fpush("Couldn't write to %s: %s", filename, err);
983 return -1; 983 return -1;
984 } 984 }
985 985
diff --git a/tools/perf/util/annotate.c b/tools/perf/util/annotate.c
index 8069dfb5ba77..3a282c0057d2 100644
--- a/tools/perf/util/annotate.c
+++ b/tools/perf/util/annotate.c
@@ -426,7 +426,18 @@ int symbol__alloc_hist(struct symbol *sym)
426{ 426{
427 struct annotation *notes = symbol__annotation(sym); 427 struct annotation *notes = symbol__annotation(sym);
428 const size_t size = symbol__size(sym); 428 const size_t size = symbol__size(sym);
429 size_t sizeof_sym_hist = (sizeof(struct sym_hist) + size * sizeof(u64)); 429 size_t sizeof_sym_hist;
430
431 /* Check for overflow when calculating sizeof_sym_hist */
432 if (size > (SIZE_MAX - sizeof(struct sym_hist)) / sizeof(u64))
433 return -1;
434
435 sizeof_sym_hist = (sizeof(struct sym_hist) + size * sizeof(u64));
436
437 /* Check for overflow in zalloc argument */
438 if (sizeof_sym_hist > (SIZE_MAX - sizeof(*notes->src))
439 / symbol_conf.nr_events)
440 return -1;
430 441
431 notes->src = zalloc(sizeof(*notes->src) + symbol_conf.nr_events * sizeof_sym_hist); 442 notes->src = zalloc(sizeof(*notes->src) + symbol_conf.nr_events * sizeof_sym_hist);
432 if (notes->src == NULL) 443 if (notes->src == NULL)
@@ -777,7 +788,7 @@ fallback:
777 free_filename = false; 788 free_filename = false;
778 } 789 }
779 790
780 if (dso->symtab_type == SYMTAB__KALLSYMS) { 791 if (dso->symtab_type == DSO_BINARY_TYPE__KALLSYMS) {
781 char bf[BUILD_ID_SIZE * 2 + 16] = " with build id "; 792 char bf[BUILD_ID_SIZE * 2 + 16] = " with build id ";
782 char *build_id_msg = NULL; 793 char *build_id_msg = NULL;
783 794
diff --git a/tools/perf/util/dso-test-data.c b/tools/perf/util/dso-test-data.c
new file mode 100644
index 000000000000..541cdc72c7df
--- /dev/null
+++ b/tools/perf/util/dso-test-data.c
@@ -0,0 +1,153 @@
1#include "util.h"
2
3#include <stdlib.h>
4#include <sys/types.h>
5#include <sys/stat.h>
6#include <fcntl.h>
7#include <string.h>
8
9#include "symbol.h"
10
11#define TEST_ASSERT_VAL(text, cond) \
12do { \
13 if (!(cond)) { \
14 pr_debug("FAILED %s:%d %s\n", __FILE__, __LINE__, text); \
15 return -1; \
16 } \
17} while (0)
18
19static char *test_file(int size)
20{
21 static char buf_templ[] = "/tmp/test-XXXXXX";
22 char *templ = buf_templ;
23 int fd, i;
24 unsigned char *buf;
25
26 fd = mkostemp(templ, O_CREAT|O_WRONLY|O_TRUNC);
27
28 buf = malloc(size);
29 if (!buf) {
30 close(fd);
31 return NULL;
32 }
33
34 for (i = 0; i < size; i++)
35 buf[i] = (unsigned char) ((int) i % 10);
36
37 if (size != write(fd, buf, size))
38 templ = NULL;
39
40 close(fd);
41 return templ;
42}
43
44#define TEST_FILE_SIZE (DSO__DATA_CACHE_SIZE * 20)
45
46struct test_data_offset {
47 off_t offset;
48 u8 data[10];
49 int size;
50};
51
52struct test_data_offset offsets[] = {
53 /* Fill first cache page. */
54 {
55 .offset = 10,
56 .data = { 0, 1, 2, 3, 4, 5, 6, 7, 8, 9 },
57 .size = 10,
58 },
59 /* Read first cache page. */
60 {
61 .offset = 10,
62 .data = { 0, 1, 2, 3, 4, 5, 6, 7, 8, 9 },
63 .size = 10,
64 },
65 /* Fill cache boundary pages. */
66 {
67 .offset = DSO__DATA_CACHE_SIZE - DSO__DATA_CACHE_SIZE % 10,
68 .data = { 0, 1, 2, 3, 4, 5, 6, 7, 8, 9 },
69 .size = 10,
70 },
71 /* Read cache boundary pages. */
72 {
73 .offset = DSO__DATA_CACHE_SIZE - DSO__DATA_CACHE_SIZE % 10,
74 .data = { 0, 1, 2, 3, 4, 5, 6, 7, 8, 9 },
75 .size = 10,
76 },
77 /* Fill final cache page. */
78 {
79 .offset = TEST_FILE_SIZE - 10,
80 .data = { 0, 1, 2, 3, 4, 5, 6, 7, 8, 9 },
81 .size = 10,
82 },
83 /* Read final cache page. */
84 {
85 .offset = TEST_FILE_SIZE - 10,
86 .data = { 0, 1, 2, 3, 4, 5, 6, 7, 8, 9 },
87 .size = 10,
88 },
89 /* Read final cache page. */
90 {
91 .offset = TEST_FILE_SIZE - 3,
92 .data = { 7, 8, 9, 0, 0, 0, 0, 0, 0, 0 },
93 .size = 3,
94 },
95};
96
97int dso__test_data(void)
98{
99 struct machine machine;
100 struct dso *dso;
101 char *file = test_file(TEST_FILE_SIZE);
102 size_t i;
103
104 TEST_ASSERT_VAL("No test file", file);
105
106 memset(&machine, 0, sizeof(machine));
107
108 dso = dso__new((const char *)file);
109
110 /* Basic 10 bytes tests. */
111 for (i = 0; i < ARRAY_SIZE(offsets); i++) {
112 struct test_data_offset *data = &offsets[i];
113 ssize_t size;
114 u8 buf[10];
115
116 memset(buf, 0, 10);
117 size = dso__data_read_offset(dso, &machine, data->offset,
118 buf, 10);
119
120 TEST_ASSERT_VAL("Wrong size", size == data->size);
121 TEST_ASSERT_VAL("Wrong data", !memcmp(buf, data->data, 10));
122 }
123
124 /* Read cross multiple cache pages. */
125 {
126 ssize_t size;
127 int c;
128 u8 *buf;
129
130 buf = malloc(TEST_FILE_SIZE);
131 TEST_ASSERT_VAL("ENOMEM\n", buf);
132
133 /* First iteration to fill caches, second one to read them. */
134 for (c = 0; c < 2; c++) {
135 memset(buf, 0, TEST_FILE_SIZE);
136 size = dso__data_read_offset(dso, &machine, 10,
137 buf, TEST_FILE_SIZE);
138
139 TEST_ASSERT_VAL("Wrong size",
140 size == (TEST_FILE_SIZE - 10));
141
142 for (i = 0; i < (size_t)size; i++)
143 TEST_ASSERT_VAL("Wrong data",
144 buf[i] == (i % 10));
145 }
146
147 free(buf);
148 }
149
150 dso__delete(dso);
151 unlink(file);
152 return 0;
153}
diff --git a/tools/perf/util/evlist.c b/tools/perf/util/evlist.c
index f74e9560350e..3edfd3483816 100644
--- a/tools/perf/util/evlist.c
+++ b/tools/perf/util/evlist.c
@@ -214,7 +214,7 @@ int perf_evlist__add_tracepoints(struct perf_evlist *evlist,
214 attrs[i].type = PERF_TYPE_TRACEPOINT; 214 attrs[i].type = PERF_TYPE_TRACEPOINT;
215 attrs[i].config = err; 215 attrs[i].config = err;
216 attrs[i].sample_type = (PERF_SAMPLE_RAW | PERF_SAMPLE_TIME | 216 attrs[i].sample_type = (PERF_SAMPLE_RAW | PERF_SAMPLE_TIME |
217 PERF_SAMPLE_CPU); 217 PERF_SAMPLE_CPU | PERF_SAMPLE_PERIOD);
218 attrs[i].sample_period = 1; 218 attrs[i].sample_period = 1;
219 } 219 }
220 220
diff --git a/tools/perf/util/header.c b/tools/perf/util/header.c
index 5a47aba46759..3a6d20443330 100644
--- a/tools/perf/util/header.c
+++ b/tools/perf/util/header.c
@@ -1212,6 +1212,12 @@ static void print_event_desc(struct perf_header *ph, int fd, FILE *fp)
1212 attr.exclude_user, 1212 attr.exclude_user,
1213 attr.exclude_kernel); 1213 attr.exclude_kernel);
1214 1214
1215 fprintf(fp, ", excl_host = %d, excl_guest = %d",
1216 attr.exclude_host,
1217 attr.exclude_guest);
1218
1219 fprintf(fp, ", precise_ip = %d", attr.precise_ip);
1220
1215 if (nr) 1221 if (nr)
1216 fprintf(fp, ", id = {"); 1222 fprintf(fp, ", id = {");
1217 1223
diff --git a/tools/perf/util/hist.c b/tools/perf/util/hist.c
index 514e2a4b367d..f247ef2789a4 100644
--- a/tools/perf/util/hist.c
+++ b/tools/perf/util/hist.c
@@ -708,7 +708,7 @@ static size_t callchain__fprintf_graph(FILE *fp, struct rb_root *root,
708 bool printed = false; 708 bool printed = false;
709 struct rb_node *node; 709 struct rb_node *node;
710 int i = 0; 710 int i = 0;
711 int ret; 711 int ret = 0;
712 712
713 /* 713 /*
714 * If have one single callchain root, don't bother printing 714 * If have one single callchain root, don't bother printing
@@ -747,8 +747,11 @@ static size_t callchain__fprintf_graph(FILE *fp, struct rb_root *root,
747 root = &cnode->rb_root; 747 root = &cnode->rb_root;
748 } 748 }
749 749
750 return __callchain__fprintf_graph(fp, root, total_samples, 750 ret += __callchain__fprintf_graph(fp, root, total_samples,
751 1, 1, left_margin); 751 1, 1, left_margin);
752 ret += fprintf(fp, "\n");
753
754 return ret;
752} 755}
753 756
754static size_t __callchain__fprintf_flat(FILE *fp, 757static size_t __callchain__fprintf_flat(FILE *fp,
diff --git a/tools/perf/util/map.c b/tools/perf/util/map.c
index a1f4e3669142..cc33486ad9e2 100644
--- a/tools/perf/util/map.c
+++ b/tools/perf/util/map.c
@@ -7,6 +7,8 @@
7#include <stdio.h> 7#include <stdio.h>
8#include <unistd.h> 8#include <unistd.h>
9#include "map.h" 9#include "map.h"
10#include "thread.h"
11#include "strlist.h"
10 12
11const char *map_type__name[MAP__NR_TYPES] = { 13const char *map_type__name[MAP__NR_TYPES] = {
12 [MAP__FUNCTION] = "Functions", 14 [MAP__FUNCTION] = "Functions",
@@ -585,7 +587,21 @@ int machine__init(struct machine *self, const char *root_dir, pid_t pid)
585 self->kmaps.machine = self; 587 self->kmaps.machine = self;
586 self->pid = pid; 588 self->pid = pid;
587 self->root_dir = strdup(root_dir); 589 self->root_dir = strdup(root_dir);
588 return self->root_dir == NULL ? -ENOMEM : 0; 590 if (self->root_dir == NULL)
591 return -ENOMEM;
592
593 if (pid != HOST_KERNEL_ID) {
594 struct thread *thread = machine__findnew_thread(self, pid);
595 char comm[64];
596
597 if (thread == NULL)
598 return -ENOMEM;
599
600 snprintf(comm, sizeof(comm), "[guest/%d]", pid);
601 thread__set_comm(thread, comm);
602 }
603
604 return 0;
589} 605}
590 606
591static void dsos__delete(struct list_head *self) 607static void dsos__delete(struct list_head *self)
@@ -680,7 +696,15 @@ struct machine *machines__findnew(struct rb_root *self, pid_t pid)
680 (symbol_conf.guestmount)) { 696 (symbol_conf.guestmount)) {
681 sprintf(path, "%s/%d", symbol_conf.guestmount, pid); 697 sprintf(path, "%s/%d", symbol_conf.guestmount, pid);
682 if (access(path, R_OK)) { 698 if (access(path, R_OK)) {
683 pr_err("Can't access file %s\n", path); 699 static struct strlist *seen;
700
701 if (!seen)
702 seen = strlist__new(true, NULL);
703
704 if (!strlist__has_entry(seen, path)) {
705 pr_err("Can't access file %s\n", path);
706 strlist__add(seen, path);
707 }
684 machine = NULL; 708 machine = NULL;
685 goto out; 709 goto out;
686 } 710 }
@@ -714,3 +738,16 @@ char *machine__mmap_name(struct machine *self, char *bf, size_t size)
714 738
715 return bf; 739 return bf;
716} 740}
741
742void machines__set_id_hdr_size(struct rb_root *machines, u16 id_hdr_size)
743{
744 struct rb_node *node;
745 struct machine *machine;
746
747 for (node = rb_first(machines); node; node = rb_next(node)) {
748 machine = rb_entry(node, struct machine, rb_node);
749 machine->id_hdr_size = id_hdr_size;
750 }
751
752 return;
753}
diff --git a/tools/perf/util/map.h b/tools/perf/util/map.h
index c14c665d9a25..03a1e9b08b21 100644
--- a/tools/perf/util/map.h
+++ b/tools/perf/util/map.h
@@ -151,6 +151,7 @@ struct machine *machines__add(struct rb_root *self, pid_t pid,
151struct machine *machines__find_host(struct rb_root *self); 151struct machine *machines__find_host(struct rb_root *self);
152struct machine *machines__find(struct rb_root *self, pid_t pid); 152struct machine *machines__find(struct rb_root *self, pid_t pid);
153struct machine *machines__findnew(struct rb_root *self, pid_t pid); 153struct machine *machines__findnew(struct rb_root *self, pid_t pid);
154void machines__set_id_hdr_size(struct rb_root *self, u16 id_hdr_size);
154char *machine__mmap_name(struct machine *self, char *bf, size_t size); 155char *machine__mmap_name(struct machine *self, char *bf, size_t size);
155int machine__init(struct machine *self, const char *root_dir, pid_t pid); 156int machine__init(struct machine *self, const char *root_dir, pid_t pid);
156void machine__exit(struct machine *self); 157void machine__exit(struct machine *self);
diff --git a/tools/perf/util/parse-events.c b/tools/perf/util/parse-events.c
index 1aa721d7c10f..74a5af4d33ec 100644
--- a/tools/perf/util/parse-events.c
+++ b/tools/perf/util/parse-events.c
@@ -377,6 +377,7 @@ static int add_tracepoint(struct list_head **list, int *idx,
377 attr.sample_type |= PERF_SAMPLE_RAW; 377 attr.sample_type |= PERF_SAMPLE_RAW;
378 attr.sample_type |= PERF_SAMPLE_TIME; 378 attr.sample_type |= PERF_SAMPLE_TIME;
379 attr.sample_type |= PERF_SAMPLE_CPU; 379 attr.sample_type |= PERF_SAMPLE_CPU;
380 attr.sample_type |= PERF_SAMPLE_PERIOD;
380 attr.sample_period = 1; 381 attr.sample_period = 1;
381 382
382 snprintf(name, MAX_NAME_LEN, "%s:%s", sys_name, evt_name); 383 snprintf(name, MAX_NAME_LEN, "%s:%s", sys_name, evt_name);
@@ -489,6 +490,7 @@ int parse_events_add_breakpoint(struct list_head **list, int *idx,
489 attr.bp_len = HW_BREAKPOINT_LEN_4; 490 attr.bp_len = HW_BREAKPOINT_LEN_4;
490 491
491 attr.type = PERF_TYPE_BREAKPOINT; 492 attr.type = PERF_TYPE_BREAKPOINT;
493 attr.sample_period = 1;
492 494
493 return add_event(list, idx, &attr, NULL); 495 return add_event(list, idx, &attr, NULL);
494} 496}
diff --git a/tools/perf/util/session.c b/tools/perf/util/session.c
index 8e485592ca20..8e4f0755d2aa 100644
--- a/tools/perf/util/session.c
+++ b/tools/perf/util/session.c
@@ -87,6 +87,7 @@ void perf_session__update_sample_type(struct perf_session *self)
87 self->sample_id_all = perf_evlist__sample_id_all(self->evlist); 87 self->sample_id_all = perf_evlist__sample_id_all(self->evlist);
88 self->id_hdr_size = perf_evlist__id_hdr_size(self->evlist); 88 self->id_hdr_size = perf_evlist__id_hdr_size(self->evlist);
89 self->host_machine.id_hdr_size = self->id_hdr_size; 89 self->host_machine.id_hdr_size = self->id_hdr_size;
90 machines__set_id_hdr_size(&self->machines, self->id_hdr_size);
90} 91}
91 92
92int perf_session__create_kernel_maps(struct perf_session *self) 93int perf_session__create_kernel_maps(struct perf_session *self)
@@ -918,7 +919,9 @@ static struct machine *
918{ 919{
919 const u8 cpumode = event->header.misc & PERF_RECORD_MISC_CPUMODE_MASK; 920 const u8 cpumode = event->header.misc & PERF_RECORD_MISC_CPUMODE_MASK;
920 921
921 if (cpumode == PERF_RECORD_MISC_GUEST_KERNEL && perf_guest) { 922 if (perf_guest &&
923 ((cpumode == PERF_RECORD_MISC_GUEST_KERNEL) ||
924 (cpumode == PERF_RECORD_MISC_GUEST_USER))) {
922 u32 pid; 925 u32 pid;
923 926
924 if (event->header.type == PERF_RECORD_MMAP) 927 if (event->header.type == PERF_RECORD_MMAP)
diff --git a/tools/perf/util/symbol.c b/tools/perf/util/symbol.c
index 50958bbeb26a..fdad4eeeb429 100644
--- a/tools/perf/util/symbol.c
+++ b/tools/perf/util/symbol.c
@@ -29,6 +29,7 @@
29#define NT_GNU_BUILD_ID 3 29#define NT_GNU_BUILD_ID 3
30#endif 30#endif
31 31
32static void dso_cache__free(struct rb_root *root);
32static bool dso__build_id_equal(const struct dso *dso, u8 *build_id); 33static bool dso__build_id_equal(const struct dso *dso, u8 *build_id);
33static int elf_read_build_id(Elf *elf, void *bf, size_t size); 34static int elf_read_build_id(Elf *elf, void *bf, size_t size);
34static void dsos__add(struct list_head *head, struct dso *dso); 35static void dsos__add(struct list_head *head, struct dso *dso);
@@ -48,6 +49,31 @@ struct symbol_conf symbol_conf = {
48 .symfs = "", 49 .symfs = "",
49}; 50};
50 51
52static enum dso_binary_type binary_type_symtab[] = {
53 DSO_BINARY_TYPE__KALLSYMS,
54 DSO_BINARY_TYPE__GUEST_KALLSYMS,
55 DSO_BINARY_TYPE__JAVA_JIT,
56 DSO_BINARY_TYPE__DEBUGLINK,
57 DSO_BINARY_TYPE__BUILD_ID_CACHE,
58 DSO_BINARY_TYPE__FEDORA_DEBUGINFO,
59 DSO_BINARY_TYPE__UBUNTU_DEBUGINFO,
60 DSO_BINARY_TYPE__BUILDID_DEBUGINFO,
61 DSO_BINARY_TYPE__SYSTEM_PATH_DSO,
62 DSO_BINARY_TYPE__GUEST_KMODULE,
63 DSO_BINARY_TYPE__SYSTEM_PATH_KMODULE,
64 DSO_BINARY_TYPE__NOT_FOUND,
65};
66
67#define DSO_BINARY_TYPE__SYMTAB_CNT sizeof(binary_type_symtab)
68
69static enum dso_binary_type binary_type_data[] = {
70 DSO_BINARY_TYPE__BUILD_ID_CACHE,
71 DSO_BINARY_TYPE__SYSTEM_PATH_DSO,
72 DSO_BINARY_TYPE__NOT_FOUND,
73};
74
75#define DSO_BINARY_TYPE__DATA_CNT sizeof(binary_type_data)
76
51int dso__name_len(const struct dso *dso) 77int dso__name_len(const struct dso *dso)
52{ 78{
53 if (!dso) 79 if (!dso)
@@ -318,7 +344,9 @@ struct dso *dso__new(const char *name)
318 dso__set_short_name(dso, dso->name); 344 dso__set_short_name(dso, dso->name);
319 for (i = 0; i < MAP__NR_TYPES; ++i) 345 for (i = 0; i < MAP__NR_TYPES; ++i)
320 dso->symbols[i] = dso->symbol_names[i] = RB_ROOT; 346 dso->symbols[i] = dso->symbol_names[i] = RB_ROOT;
321 dso->symtab_type = SYMTAB__NOT_FOUND; 347 dso->cache = RB_ROOT;
348 dso->symtab_type = DSO_BINARY_TYPE__NOT_FOUND;
349 dso->data_type = DSO_BINARY_TYPE__NOT_FOUND;
322 dso->loaded = 0; 350 dso->loaded = 0;
323 dso->sorted_by_name = 0; 351 dso->sorted_by_name = 0;
324 dso->has_build_id = 0; 352 dso->has_build_id = 0;
@@ -352,6 +380,7 @@ void dso__delete(struct dso *dso)
352 free((char *)dso->short_name); 380 free((char *)dso->short_name);
353 if (dso->lname_alloc) 381 if (dso->lname_alloc)
354 free(dso->long_name); 382 free(dso->long_name);
383 dso_cache__free(&dso->cache);
355 free(dso); 384 free(dso);
356} 385}
357 386
@@ -806,9 +835,9 @@ int dso__load_kallsyms(struct dso *dso, const char *filename,
806 symbols__fixup_end(&dso->symbols[map->type]); 835 symbols__fixup_end(&dso->symbols[map->type]);
807 836
808 if (dso->kernel == DSO_TYPE_GUEST_KERNEL) 837 if (dso->kernel == DSO_TYPE_GUEST_KERNEL)
809 dso->symtab_type = SYMTAB__GUEST_KALLSYMS; 838 dso->symtab_type = DSO_BINARY_TYPE__GUEST_KALLSYMS;
810 else 839 else
811 dso->symtab_type = SYMTAB__KALLSYMS; 840 dso->symtab_type = DSO_BINARY_TYPE__KALLSYMS;
812 841
813 return dso__split_kallsyms(dso, map, filter); 842 return dso__split_kallsyms(dso, map, filter);
814} 843}
@@ -1660,32 +1689,110 @@ out:
1660char dso__symtab_origin(const struct dso *dso) 1689char dso__symtab_origin(const struct dso *dso)
1661{ 1690{
1662 static const char origin[] = { 1691 static const char origin[] = {
1663 [SYMTAB__KALLSYMS] = 'k', 1692 [DSO_BINARY_TYPE__KALLSYMS] = 'k',
1664 [SYMTAB__JAVA_JIT] = 'j', 1693 [DSO_BINARY_TYPE__JAVA_JIT] = 'j',
1665 [SYMTAB__DEBUGLINK] = 'l', 1694 [DSO_BINARY_TYPE__DEBUGLINK] = 'l',
1666 [SYMTAB__BUILD_ID_CACHE] = 'B', 1695 [DSO_BINARY_TYPE__BUILD_ID_CACHE] = 'B',
1667 [SYMTAB__FEDORA_DEBUGINFO] = 'f', 1696 [DSO_BINARY_TYPE__FEDORA_DEBUGINFO] = 'f',
1668 [SYMTAB__UBUNTU_DEBUGINFO] = 'u', 1697 [DSO_BINARY_TYPE__UBUNTU_DEBUGINFO] = 'u',
1669 [SYMTAB__BUILDID_DEBUGINFO] = 'b', 1698 [DSO_BINARY_TYPE__BUILDID_DEBUGINFO] = 'b',
1670 [SYMTAB__SYSTEM_PATH_DSO] = 'd', 1699 [DSO_BINARY_TYPE__SYSTEM_PATH_DSO] = 'd',
1671 [SYMTAB__SYSTEM_PATH_KMODULE] = 'K', 1700 [DSO_BINARY_TYPE__SYSTEM_PATH_KMODULE] = 'K',
1672 [SYMTAB__GUEST_KALLSYMS] = 'g', 1701 [DSO_BINARY_TYPE__GUEST_KALLSYMS] = 'g',
1673 [SYMTAB__GUEST_KMODULE] = 'G', 1702 [DSO_BINARY_TYPE__GUEST_KMODULE] = 'G',
1674 }; 1703 };
1675 1704
1676 if (dso == NULL || dso->symtab_type == SYMTAB__NOT_FOUND) 1705 if (dso == NULL || dso->symtab_type == DSO_BINARY_TYPE__NOT_FOUND)
1677 return '!'; 1706 return '!';
1678 return origin[dso->symtab_type]; 1707 return origin[dso->symtab_type];
1679} 1708}
1680 1709
1710int dso__binary_type_file(struct dso *dso, enum dso_binary_type type,
1711 char *root_dir, char *file, size_t size)
1712{
1713 char build_id_hex[BUILD_ID_SIZE * 2 + 1];
1714 int ret = 0;
1715
1716 switch (type) {
1717 case DSO_BINARY_TYPE__DEBUGLINK: {
1718 char *debuglink;
1719
1720 strncpy(file, dso->long_name, size);
1721 debuglink = file + dso->long_name_len;
1722 while (debuglink != file && *debuglink != '/')
1723 debuglink--;
1724 if (*debuglink == '/')
1725 debuglink++;
1726 filename__read_debuglink(dso->long_name, debuglink,
1727 size - (debuglink - file));
1728 }
1729 break;
1730 case DSO_BINARY_TYPE__BUILD_ID_CACHE:
1731 /* skip the locally configured cache if a symfs is given */
1732 if (symbol_conf.symfs[0] ||
1733 (dso__build_id_filename(dso, file, size) == NULL))
1734 ret = -1;
1735 break;
1736
1737 case DSO_BINARY_TYPE__FEDORA_DEBUGINFO:
1738 snprintf(file, size, "%s/usr/lib/debug%s.debug",
1739 symbol_conf.symfs, dso->long_name);
1740 break;
1741
1742 case DSO_BINARY_TYPE__UBUNTU_DEBUGINFO:
1743 snprintf(file, size, "%s/usr/lib/debug%s",
1744 symbol_conf.symfs, dso->long_name);
1745 break;
1746
1747 case DSO_BINARY_TYPE__BUILDID_DEBUGINFO:
1748 if (!dso->has_build_id) {
1749 ret = -1;
1750 break;
1751 }
1752
1753 build_id__sprintf(dso->build_id,
1754 sizeof(dso->build_id),
1755 build_id_hex);
1756 snprintf(file, size,
1757 "%s/usr/lib/debug/.build-id/%.2s/%s.debug",
1758 symbol_conf.symfs, build_id_hex, build_id_hex + 2);
1759 break;
1760
1761 case DSO_BINARY_TYPE__SYSTEM_PATH_DSO:
1762 snprintf(file, size, "%s%s",
1763 symbol_conf.symfs, dso->long_name);
1764 break;
1765
1766 case DSO_BINARY_TYPE__GUEST_KMODULE:
1767 snprintf(file, size, "%s%s%s", symbol_conf.symfs,
1768 root_dir, dso->long_name);
1769 break;
1770
1771 case DSO_BINARY_TYPE__SYSTEM_PATH_KMODULE:
1772 snprintf(file, size, "%s%s", symbol_conf.symfs,
1773 dso->long_name);
1774 break;
1775
1776 default:
1777 case DSO_BINARY_TYPE__KALLSYMS:
1778 case DSO_BINARY_TYPE__GUEST_KALLSYMS:
1779 case DSO_BINARY_TYPE__JAVA_JIT:
1780 case DSO_BINARY_TYPE__NOT_FOUND:
1781 ret = -1;
1782 break;
1783 }
1784
1785 return ret;
1786}
1787
1681int dso__load(struct dso *dso, struct map *map, symbol_filter_t filter) 1788int dso__load(struct dso *dso, struct map *map, symbol_filter_t filter)
1682{ 1789{
1683 int size = PATH_MAX;
1684 char *name; 1790 char *name;
1685 int ret = -1; 1791 int ret = -1;
1686 int fd; 1792 int fd;
1793 u_int i;
1687 struct machine *machine; 1794 struct machine *machine;
1688 const char *root_dir; 1795 char *root_dir = (char *) "";
1689 int want_symtab; 1796 int want_symtab;
1690 1797
1691 dso__set_loaded(dso, map->type); 1798 dso__set_loaded(dso, map->type);
@@ -1700,7 +1807,7 @@ int dso__load(struct dso *dso, struct map *map, symbol_filter_t filter)
1700 else 1807 else
1701 machine = NULL; 1808 machine = NULL;
1702 1809
1703 name = malloc(size); 1810 name = malloc(PATH_MAX);
1704 if (!name) 1811 if (!name)
1705 return -1; 1812 return -1;
1706 1813
@@ -1719,81 +1826,27 @@ int dso__load(struct dso *dso, struct map *map, symbol_filter_t filter)
1719 } 1826 }
1720 1827
1721 ret = dso__load_perf_map(dso, map, filter); 1828 ret = dso__load_perf_map(dso, map, filter);
1722 dso->symtab_type = ret > 0 ? SYMTAB__JAVA_JIT : 1829 dso->symtab_type = ret > 0 ? DSO_BINARY_TYPE__JAVA_JIT :
1723 SYMTAB__NOT_FOUND; 1830 DSO_BINARY_TYPE__NOT_FOUND;
1724 return ret; 1831 return ret;
1725 } 1832 }
1726 1833
1834 if (machine)
1835 root_dir = machine->root_dir;
1836
1727 /* Iterate over candidate debug images. 1837 /* Iterate over candidate debug images.
1728 * On the first pass, only load images if they have a full symtab. 1838 * On the first pass, only load images if they have a full symtab.
1729 * Failing that, do a second pass where we accept .dynsym also 1839 * Failing that, do a second pass where we accept .dynsym also
1730 */ 1840 */
1731 want_symtab = 1; 1841 want_symtab = 1;
1732restart: 1842restart:
1733 for (dso->symtab_type = SYMTAB__DEBUGLINK; 1843 for (i = 0; i < DSO_BINARY_TYPE__SYMTAB_CNT; i++) {
1734 dso->symtab_type != SYMTAB__NOT_FOUND;
1735 dso->symtab_type++) {
1736 switch (dso->symtab_type) {
1737 case SYMTAB__DEBUGLINK: {
1738 char *debuglink;
1739 strncpy(name, dso->long_name, size);
1740 debuglink = name + dso->long_name_len;
1741 while (debuglink != name && *debuglink != '/')
1742 debuglink--;
1743 if (*debuglink == '/')
1744 debuglink++;
1745 filename__read_debuglink(dso->long_name, debuglink,
1746 size - (debuglink - name));
1747 }
1748 break;
1749 case SYMTAB__BUILD_ID_CACHE:
1750 /* skip the locally configured cache if a symfs is given */
1751 if (symbol_conf.symfs[0] ||
1752 (dso__build_id_filename(dso, name, size) == NULL)) {
1753 continue;
1754 }
1755 break;
1756 case SYMTAB__FEDORA_DEBUGINFO:
1757 snprintf(name, size, "%s/usr/lib/debug%s.debug",
1758 symbol_conf.symfs, dso->long_name);
1759 break;
1760 case SYMTAB__UBUNTU_DEBUGINFO:
1761 snprintf(name, size, "%s/usr/lib/debug%s",
1762 symbol_conf.symfs, dso->long_name);
1763 break;
1764 case SYMTAB__BUILDID_DEBUGINFO: {
1765 char build_id_hex[BUILD_ID_SIZE * 2 + 1];
1766 1844
1767 if (!dso->has_build_id) 1845 dso->symtab_type = binary_type_symtab[i];
1768 continue;
1769 1846
1770 build_id__sprintf(dso->build_id, 1847 if (dso__binary_type_file(dso, dso->symtab_type,
1771 sizeof(dso->build_id), 1848 root_dir, name, PATH_MAX))
1772 build_id_hex); 1849 continue;
1773 snprintf(name, size,
1774 "%s/usr/lib/debug/.build-id/%.2s/%s.debug",
1775 symbol_conf.symfs, build_id_hex, build_id_hex + 2);
1776 }
1777 break;
1778 case SYMTAB__SYSTEM_PATH_DSO:
1779 snprintf(name, size, "%s%s",
1780 symbol_conf.symfs, dso->long_name);
1781 break;
1782 case SYMTAB__GUEST_KMODULE:
1783 if (map->groups && machine)
1784 root_dir = machine->root_dir;
1785 else
1786 root_dir = "";
1787 snprintf(name, size, "%s%s%s", symbol_conf.symfs,
1788 root_dir, dso->long_name);
1789 break;
1790
1791 case SYMTAB__SYSTEM_PATH_KMODULE:
1792 snprintf(name, size, "%s%s", symbol_conf.symfs,
1793 dso->long_name);
1794 break;
1795 default:;
1796 }
1797 1850
1798 /* Name is now the name of the next image to try */ 1851 /* Name is now the name of the next image to try */
1799 fd = open(name, O_RDONLY); 1852 fd = open(name, O_RDONLY);
@@ -2010,9 +2063,9 @@ struct map *machine__new_module(struct machine *machine, u64 start,
2010 return NULL; 2063 return NULL;
2011 2064
2012 if (machine__is_host(machine)) 2065 if (machine__is_host(machine))
2013 dso->symtab_type = SYMTAB__SYSTEM_PATH_KMODULE; 2066 dso->symtab_type = DSO_BINARY_TYPE__SYSTEM_PATH_KMODULE;
2014 else 2067 else
2015 dso->symtab_type = SYMTAB__GUEST_KMODULE; 2068 dso->symtab_type = DSO_BINARY_TYPE__GUEST_KMODULE;
2016 map_groups__insert(&machine->kmaps, map); 2069 map_groups__insert(&machine->kmaps, map);
2017 return map; 2070 return map;
2018} 2071}
@@ -2564,8 +2617,15 @@ int machine__create_kernel_maps(struct machine *machine)
2564 __machine__create_kernel_maps(machine, kernel) < 0) 2617 __machine__create_kernel_maps(machine, kernel) < 0)
2565 return -1; 2618 return -1;
2566 2619
2567 if (symbol_conf.use_modules && machine__create_modules(machine) < 0) 2620 if (symbol_conf.use_modules && machine__create_modules(machine) < 0) {
2568 pr_debug("Problems creating module maps, continuing anyway...\n"); 2621 if (machine__is_host(machine))
2622 pr_debug("Problems creating module maps, "
2623 "continuing anyway...\n");
2624 else
2625 pr_debug("Problems creating module maps for guest %d, "
2626 "continuing anyway...\n", machine->pid);
2627 }
2628
2569 /* 2629 /*
2570 * Now that we have all the maps created, just set the ->end of them: 2630 * Now that we have all the maps created, just set the ->end of them:
2571 */ 2631 */
@@ -2905,3 +2965,218 @@ struct map *dso__new_map(const char *name)
2905 2965
2906 return map; 2966 return map;
2907} 2967}
2968
2969static int open_dso(struct dso *dso, struct machine *machine)
2970{
2971 char *root_dir = (char *) "";
2972 char *name;
2973 int fd;
2974
2975 name = malloc(PATH_MAX);
2976 if (!name)
2977 return -ENOMEM;
2978
2979 if (machine)
2980 root_dir = machine->root_dir;
2981
2982 if (dso__binary_type_file(dso, dso->data_type,
2983 root_dir, name, PATH_MAX)) {
2984 free(name);
2985 return -EINVAL;
2986 }
2987
2988 fd = open(name, O_RDONLY);
2989 free(name);
2990 return fd;
2991}
2992
2993int dso__data_fd(struct dso *dso, struct machine *machine)
2994{
2995 int i = 0;
2996
2997 if (dso->data_type != DSO_BINARY_TYPE__NOT_FOUND)
2998 return open_dso(dso, machine);
2999
3000 do {
3001 int fd;
3002
3003 dso->data_type = binary_type_data[i++];
3004
3005 fd = open_dso(dso, machine);
3006 if (fd >= 0)
3007 return fd;
3008
3009 } while (dso->data_type != DSO_BINARY_TYPE__NOT_FOUND);
3010
3011 return -EINVAL;
3012}
3013
3014static void
3015dso_cache__free(struct rb_root *root)
3016{
3017 struct rb_node *next = rb_first(root);
3018
3019 while (next) {
3020 struct dso_cache *cache;
3021
3022 cache = rb_entry(next, struct dso_cache, rb_node);
3023 next = rb_next(&cache->rb_node);
3024 rb_erase(&cache->rb_node, root);
3025 free(cache);
3026 }
3027}
3028
3029static struct dso_cache*
3030dso_cache__find(struct rb_root *root, u64 offset)
3031{
3032 struct rb_node **p = &root->rb_node;
3033 struct rb_node *parent = NULL;
3034 struct dso_cache *cache;
3035
3036 while (*p != NULL) {
3037 u64 end;
3038
3039 parent = *p;
3040 cache = rb_entry(parent, struct dso_cache, rb_node);
3041 end = cache->offset + DSO__DATA_CACHE_SIZE;
3042
3043 if (offset < cache->offset)
3044 p = &(*p)->rb_left;
3045 else if (offset >= end)
3046 p = &(*p)->rb_right;
3047 else
3048 return cache;
3049 }
3050 return NULL;
3051}
3052
3053static void
3054dso_cache__insert(struct rb_root *root, struct dso_cache *new)
3055{
3056 struct rb_node **p = &root->rb_node;
3057 struct rb_node *parent = NULL;
3058 struct dso_cache *cache;
3059 u64 offset = new->offset;
3060
3061 while (*p != NULL) {
3062 u64 end;
3063
3064 parent = *p;
3065 cache = rb_entry(parent, struct dso_cache, rb_node);
3066 end = cache->offset + DSO__DATA_CACHE_SIZE;
3067
3068 if (offset < cache->offset)
3069 p = &(*p)->rb_left;
3070 else if (offset >= end)
3071 p = &(*p)->rb_right;
3072 }
3073
3074 rb_link_node(&new->rb_node, parent, p);
3075 rb_insert_color(&new->rb_node, root);
3076}
3077
3078static ssize_t
3079dso_cache__memcpy(struct dso_cache *cache, u64 offset,
3080 u8 *data, u64 size)
3081{
3082 u64 cache_offset = offset - cache->offset;
3083 u64 cache_size = min(cache->size - cache_offset, size);
3084
3085 memcpy(data, cache->data + cache_offset, cache_size);
3086 return cache_size;
3087}
3088
3089static ssize_t
3090dso_cache__read(struct dso *dso, struct machine *machine,
3091 u64 offset, u8 *data, ssize_t size)
3092{
3093 struct dso_cache *cache;
3094 ssize_t ret;
3095 int fd;
3096
3097 fd = dso__data_fd(dso, machine);
3098 if (fd < 0)
3099 return -1;
3100
3101 do {
3102 u64 cache_offset;
3103
3104 ret = -ENOMEM;
3105
3106 cache = zalloc(sizeof(*cache) + DSO__DATA_CACHE_SIZE);
3107 if (!cache)
3108 break;
3109
3110 cache_offset = offset & DSO__DATA_CACHE_MASK;
3111 ret = -EINVAL;
3112
3113 if (-1 == lseek(fd, cache_offset, SEEK_SET))
3114 break;
3115
3116 ret = read(fd, cache->data, DSO__DATA_CACHE_SIZE);
3117 if (ret <= 0)
3118 break;
3119
3120 cache->offset = cache_offset;
3121 cache->size = ret;
3122 dso_cache__insert(&dso->cache, cache);
3123
3124 ret = dso_cache__memcpy(cache, offset, data, size);
3125
3126 } while (0);
3127
3128 if (ret <= 0)
3129 free(cache);
3130
3131 close(fd);
3132 return ret;
3133}
3134
3135static ssize_t dso_cache_read(struct dso *dso, struct machine *machine,
3136 u64 offset, u8 *data, ssize_t size)
3137{
3138 struct dso_cache *cache;
3139
3140 cache = dso_cache__find(&dso->cache, offset);
3141 if (cache)
3142 return dso_cache__memcpy(cache, offset, data, size);
3143 else
3144 return dso_cache__read(dso, machine, offset, data, size);
3145}
3146
3147ssize_t dso__data_read_offset(struct dso *dso, struct machine *machine,
3148 u64 offset, u8 *data, ssize_t size)
3149{
3150 ssize_t r = 0;
3151 u8 *p = data;
3152
3153 do {
3154 ssize_t ret;
3155
3156 ret = dso_cache_read(dso, machine, offset, p, size);
3157 if (ret < 0)
3158 return ret;
3159
3160 /* Reached EOF, return what we have. */
3161 if (!ret)
3162 break;
3163
3164 BUG_ON(ret > size);
3165
3166 r += ret;
3167 p += ret;
3168 offset += ret;
3169 size -= ret;
3170
3171 } while (size);
3172
3173 return r;
3174}
3175
3176ssize_t dso__data_read_addr(struct dso *dso, struct map *map,
3177 struct machine *machine, u64 addr,
3178 u8 *data, ssize_t size)
3179{
3180 u64 offset = map->map_ip(map, addr);
3181 return dso__data_read_offset(dso, machine, offset, data, size);
3182}
diff --git a/tools/perf/util/symbol.h b/tools/perf/util/symbol.h
index a884b99017f0..1fe733a1e21f 100644
--- a/tools/perf/util/symbol.h
+++ b/tools/perf/util/symbol.h
@@ -155,6 +155,21 @@ struct addr_location {
155 s32 cpu; 155 s32 cpu;
156}; 156};
157 157
158enum dso_binary_type {
159 DSO_BINARY_TYPE__KALLSYMS = 0,
160 DSO_BINARY_TYPE__GUEST_KALLSYMS,
161 DSO_BINARY_TYPE__JAVA_JIT,
162 DSO_BINARY_TYPE__DEBUGLINK,
163 DSO_BINARY_TYPE__BUILD_ID_CACHE,
164 DSO_BINARY_TYPE__FEDORA_DEBUGINFO,
165 DSO_BINARY_TYPE__UBUNTU_DEBUGINFO,
166 DSO_BINARY_TYPE__BUILDID_DEBUGINFO,
167 DSO_BINARY_TYPE__SYSTEM_PATH_DSO,
168 DSO_BINARY_TYPE__GUEST_KMODULE,
169 DSO_BINARY_TYPE__SYSTEM_PATH_KMODULE,
170 DSO_BINARY_TYPE__NOT_FOUND,
171};
172
158enum dso_kernel_type { 173enum dso_kernel_type {
159 DSO_TYPE_USER = 0, 174 DSO_TYPE_USER = 0,
160 DSO_TYPE_KERNEL, 175 DSO_TYPE_KERNEL,
@@ -167,19 +182,31 @@ enum dso_swap_type {
167 DSO_SWAP__YES, 182 DSO_SWAP__YES,
168}; 183};
169 184
185#define DSO__DATA_CACHE_SIZE 4096
186#define DSO__DATA_CACHE_MASK ~(DSO__DATA_CACHE_SIZE - 1)
187
188struct dso_cache {
189 struct rb_node rb_node;
190 u64 offset;
191 u64 size;
192 char data[0];
193};
194
170struct dso { 195struct dso {
171 struct list_head node; 196 struct list_head node;
172 struct rb_root symbols[MAP__NR_TYPES]; 197 struct rb_root symbols[MAP__NR_TYPES];
173 struct rb_root symbol_names[MAP__NR_TYPES]; 198 struct rb_root symbol_names[MAP__NR_TYPES];
199 struct rb_root cache;
174 enum dso_kernel_type kernel; 200 enum dso_kernel_type kernel;
175 enum dso_swap_type needs_swap; 201 enum dso_swap_type needs_swap;
202 enum dso_binary_type symtab_type;
203 enum dso_binary_type data_type;
176 u8 adjust_symbols:1; 204 u8 adjust_symbols:1;
177 u8 has_build_id:1; 205 u8 has_build_id:1;
178 u8 hit:1; 206 u8 hit:1;
179 u8 annotate_warned:1; 207 u8 annotate_warned:1;
180 u8 sname_alloc:1; 208 u8 sname_alloc:1;
181 u8 lname_alloc:1; 209 u8 lname_alloc:1;
182 unsigned char symtab_type;
183 u8 sorted_by_name; 210 u8 sorted_by_name;
184 u8 loaded; 211 u8 loaded;
185 u8 build_id[BUILD_ID_SIZE]; 212 u8 build_id[BUILD_ID_SIZE];
@@ -253,21 +280,6 @@ size_t dso__fprintf_symbols_by_name(struct dso *dso,
253 enum map_type type, FILE *fp); 280 enum map_type type, FILE *fp);
254size_t dso__fprintf(struct dso *dso, enum map_type type, FILE *fp); 281size_t dso__fprintf(struct dso *dso, enum map_type type, FILE *fp);
255 282
256enum symtab_type {
257 SYMTAB__KALLSYMS = 0,
258 SYMTAB__GUEST_KALLSYMS,
259 SYMTAB__JAVA_JIT,
260 SYMTAB__DEBUGLINK,
261 SYMTAB__BUILD_ID_CACHE,
262 SYMTAB__FEDORA_DEBUGINFO,
263 SYMTAB__UBUNTU_DEBUGINFO,
264 SYMTAB__BUILDID_DEBUGINFO,
265 SYMTAB__SYSTEM_PATH_DSO,
266 SYMTAB__GUEST_KMODULE,
267 SYMTAB__SYSTEM_PATH_KMODULE,
268 SYMTAB__NOT_FOUND,
269};
270
271char dso__symtab_origin(const struct dso *dso); 283char dso__symtab_origin(const struct dso *dso);
272void dso__set_long_name(struct dso *dso, char *name); 284void dso__set_long_name(struct dso *dso, char *name);
273void dso__set_build_id(struct dso *dso, void *build_id); 285void dso__set_build_id(struct dso *dso, void *build_id);
@@ -304,4 +316,14 @@ bool symbol_type__is_a(char symbol_type, enum map_type map_type);
304 316
305size_t machine__fprintf_vmlinux_path(struct machine *machine, FILE *fp); 317size_t machine__fprintf_vmlinux_path(struct machine *machine, FILE *fp);
306 318
319int dso__binary_type_file(struct dso *dso, enum dso_binary_type type,
320 char *root_dir, char *file, size_t size);
321
322int dso__data_fd(struct dso *dso, struct machine *machine);
323ssize_t dso__data_read_offset(struct dso *dso, struct machine *machine,
324 u64 offset, u8 *data, ssize_t size);
325ssize_t dso__data_read_addr(struct dso *dso, struct map *map,
326 struct machine *machine, u64 addr,
327 u8 *data, ssize_t size);
328int dso__test_data(void);
307#endif /* __PERF_SYMBOL */ 329#endif /* __PERF_SYMBOL */
diff --git a/tools/perf/util/target.c b/tools/perf/util/target.c
index 1064d5b148ad..3f59c496e64c 100644
--- a/tools/perf/util/target.c
+++ b/tools/perf/util/target.c
@@ -110,8 +110,17 @@ int perf_target__strerror(struct perf_target *target, int errnum,
110 int idx; 110 int idx;
111 const char *msg; 111 const char *msg;
112 112
113 BUG_ON(buflen > 0);
114
113 if (errnum >= 0) { 115 if (errnum >= 0) {
114 strerror_r(errnum, buf, buflen); 116 const char *err = strerror_r(errnum, buf, buflen);
117
118 if (err != buf) {
119 size_t len = strlen(err);
120 char *c = mempcpy(buf, err, min(buflen - 1, len));
121 *c = '\0';
122 }
123
115 return 0; 124 return 0;
116 } 125 }
117 126
generated by cgit v1.2.2 (git 2.25.0) at 2025-02-03 08:24:51 -0500