diff options
-rw-r--r-- | arch/Kconfig | 14 | ||||
-rw-r--r-- | arch/x86/oprofile/op_model_athlon.c | 33 | ||||
-rw-r--r-- | drivers/oprofile/buffer_sync.c | 6 | ||||
-rw-r--r-- | drivers/oprofile/cpu_buffer.c | 4 |
4 files changed, 47 insertions, 10 deletions
diff --git a/arch/Kconfig b/arch/Kconfig index b0fabfa864ff..2651af48b2e5 100644 --- a/arch/Kconfig +++ b/arch/Kconfig | |||
@@ -13,6 +13,20 @@ config OPROFILE | |||
13 | 13 | ||
14 | If unsure, say N. | 14 | If unsure, say N. |
15 | 15 | ||
16 | config OPROFILE_IBS | ||
17 | bool "OProfile AMD IBS support (EXPERIMENTAL)" | ||
18 | default n | ||
19 | depends on OPROFILE && SMP && X86 | ||
20 | help | ||
21 | Instruction-Based Sampling (IBS) is a new profiling | ||
22 | technique that provides rich, precise program performance | ||
23 | information. IBS is introduced by AMD Family10h processors | ||
24 | (AMD Opteron Quad-Core processor “Barcelona”) to overcome | ||
25 | the limitations of conventional performance counter | ||
26 | sampling. | ||
27 | |||
28 | If unsure, say N. | ||
29 | |||
16 | config HAVE_OPROFILE | 30 | config HAVE_OPROFILE |
17 | def_bool n | 31 | def_bool n |
18 | 32 | ||
diff --git a/arch/x86/oprofile/op_model_athlon.c b/arch/x86/oprofile/op_model_athlon.c index 9c8c8c583132..fb6015c1132b 100644 --- a/arch/x86/oprofile/op_model_athlon.c +++ b/arch/x86/oprofile/op_model_athlon.c | |||
@@ -47,6 +47,10 @@ | |||
47 | #define CTRL_SET_HOST_ONLY(val, h) (val |= ((h & 1) << 9)) | 47 | #define CTRL_SET_HOST_ONLY(val, h) (val |= ((h & 1) << 9)) |
48 | #define CTRL_SET_GUEST_ONLY(val, h) (val |= ((h & 1) << 8)) | 48 | #define CTRL_SET_GUEST_ONLY(val, h) (val |= ((h & 1) << 8)) |
49 | 49 | ||
50 | static unsigned long reset_value[NUM_COUNTERS]; | ||
51 | |||
52 | #ifdef CONFIG_OPROFILE_IBS | ||
53 | |||
50 | /* IbsFetchCtl bits/masks */ | 54 | /* IbsFetchCtl bits/masks */ |
51 | #define IBS_FETCH_HIGH_VALID_BIT (1UL << 17) /* bit 49 */ | 55 | #define IBS_FETCH_HIGH_VALID_BIT (1UL << 17) /* bit 49 */ |
52 | #define IBS_FETCH_HIGH_ENABLE (1UL << 16) /* bit 48 */ | 56 | #define IBS_FETCH_HIGH_ENABLE (1UL << 16) /* bit 48 */ |
@@ -104,7 +108,6 @@ struct ibs_op_sample { | |||
104 | */ | 108 | */ |
105 | static void clear_ibs_nmi(void); | 109 | static void clear_ibs_nmi(void); |
106 | 110 | ||
107 | static unsigned long reset_value[NUM_COUNTERS]; | ||
108 | static int ibs_allowed; /* AMD Family10h and later */ | 111 | static int ibs_allowed; /* AMD Family10h and later */ |
109 | 112 | ||
110 | struct op_ibs_config { | 113 | struct op_ibs_config { |
@@ -118,6 +121,8 @@ struct op_ibs_config { | |||
118 | 121 | ||
119 | static struct op_ibs_config ibs_config; | 122 | static struct op_ibs_config ibs_config; |
120 | 123 | ||
124 | #endif | ||
125 | |||
121 | /* functions for op_amd_spec */ | 126 | /* functions for op_amd_spec */ |
122 | 127 | ||
123 | static void op_amd_fill_in_addresses(struct op_msrs * const msrs) | 128 | static void op_amd_fill_in_addresses(struct op_msrs * const msrs) |
@@ -188,6 +193,8 @@ static void op_amd_setup_ctrs(struct op_msrs const * const msrs) | |||
188 | } | 193 | } |
189 | } | 194 | } |
190 | 195 | ||
196 | #ifdef CONFIG_OPROFILE_IBS | ||
197 | |||
191 | static inline int | 198 | static inline int |
192 | op_amd_handle_ibs(struct pt_regs * const regs, | 199 | op_amd_handle_ibs(struct pt_regs * const regs, |
193 | struct op_msrs const * const msrs) | 200 | struct op_msrs const * const msrs) |
@@ -261,6 +268,8 @@ op_amd_handle_ibs(struct pt_regs * const regs, | |||
261 | return 1; | 268 | return 1; |
262 | } | 269 | } |
263 | 270 | ||
271 | #endif | ||
272 | |||
264 | static int op_amd_check_ctrs(struct pt_regs * const regs, | 273 | static int op_amd_check_ctrs(struct pt_regs * const regs, |
265 | struct op_msrs const * const msrs) | 274 | struct op_msrs const * const msrs) |
266 | { | 275 | { |
@@ -277,7 +286,9 @@ static int op_amd_check_ctrs(struct pt_regs * const regs, | |||
277 | } | 286 | } |
278 | } | 287 | } |
279 | 288 | ||
289 | #ifdef CONFIG_OPROFILE_IBS | ||
280 | op_amd_handle_ibs(regs, msrs); | 290 | op_amd_handle_ibs(regs, msrs); |
291 | #endif | ||
281 | 292 | ||
282 | /* See op_model_ppro.c */ | 293 | /* See op_model_ppro.c */ |
283 | return 1; | 294 | return 1; |
@@ -294,6 +305,8 @@ static void op_amd_start(struct op_msrs const * const msrs) | |||
294 | CTRL_WRITE(low, high, msrs, i); | 305 | CTRL_WRITE(low, high, msrs, i); |
295 | } | 306 | } |
296 | } | 307 | } |
308 | |||
309 | #ifdef CONFIG_OPROFILE_IBS | ||
297 | if (ibs_allowed && ibs_config.fetch_enabled) { | 310 | if (ibs_allowed && ibs_config.fetch_enabled) { |
298 | low = (ibs_config.max_cnt_fetch >> 4) & 0xFFFF; | 311 | low = (ibs_config.max_cnt_fetch >> 4) & 0xFFFF; |
299 | high = IBS_FETCH_HIGH_ENABLE; | 312 | high = IBS_FETCH_HIGH_ENABLE; |
@@ -305,6 +318,7 @@ static void op_amd_start(struct op_msrs const * const msrs) | |||
305 | high = 0; | 318 | high = 0; |
306 | wrmsr(MSR_AMD64_IBSOPCTL, low, high); | 319 | wrmsr(MSR_AMD64_IBSOPCTL, low, high); |
307 | } | 320 | } |
321 | #endif | ||
308 | } | 322 | } |
309 | 323 | ||
310 | 324 | ||
@@ -323,6 +337,7 @@ static void op_amd_stop(struct op_msrs const * const msrs) | |||
323 | CTRL_WRITE(low, high, msrs, i); | 337 | CTRL_WRITE(low, high, msrs, i); |
324 | } | 338 | } |
325 | 339 | ||
340 | #ifdef CONFIG_OPROFILE_IBS | ||
326 | if (ibs_allowed && ibs_config.fetch_enabled) { | 341 | if (ibs_allowed && ibs_config.fetch_enabled) { |
327 | low = 0; /* clear max count and enable */ | 342 | low = 0; /* clear max count and enable */ |
328 | high = 0; | 343 | high = 0; |
@@ -334,6 +349,7 @@ static void op_amd_stop(struct op_msrs const * const msrs) | |||
334 | high = 0; | 349 | high = 0; |
335 | wrmsr(MSR_AMD64_IBSOPCTL, low, high); | 350 | wrmsr(MSR_AMD64_IBSOPCTL, low, high); |
336 | } | 351 | } |
352 | #endif | ||
337 | } | 353 | } |
338 | 354 | ||
339 | static void op_amd_shutdown(struct op_msrs const * const msrs) | 355 | static void op_amd_shutdown(struct op_msrs const * const msrs) |
@@ -350,17 +366,10 @@ static void op_amd_shutdown(struct op_msrs const * const msrs) | |||
350 | } | 366 | } |
351 | } | 367 | } |
352 | 368 | ||
353 | #ifndef CONFIG_SMP | 369 | #ifndef CONFIG_OPROFILE_IBS |
354 | 370 | ||
355 | /* no IBS support */ | 371 | /* no IBS support */ |
356 | 372 | ||
357 | static void setup_ibs(void) | ||
358 | { | ||
359 | ibs_allowed = 0; | ||
360 | } | ||
361 | |||
362 | static void clear_ibs_nmi(void) {} | ||
363 | |||
364 | static int op_amd_init(struct oprofile_operations *ops) | 373 | static int op_amd_init(struct oprofile_operations *ops) |
365 | { | 374 | { |
366 | return 0; | 375 | return 0; |
@@ -441,8 +450,12 @@ static void setup_ibs(void) | |||
441 | if (!ibs_allowed) | 450 | if (!ibs_allowed) |
442 | return; | 451 | return; |
443 | 452 | ||
444 | if (pfm_amd64_setup_eilvt()) | 453 | if (pfm_amd64_setup_eilvt()) { |
445 | ibs_allowed = 0; | 454 | ibs_allowed = 0; |
455 | return; | ||
456 | } | ||
457 | |||
458 | printk(KERN_INFO "oprofile: AMD IBS detected\n"); | ||
446 | } | 459 | } |
447 | 460 | ||
448 | 461 | ||
diff --git a/drivers/oprofile/buffer_sync.c b/drivers/oprofile/buffer_sync.c index e1782d2df09f..ed982273fb8b 100644 --- a/drivers/oprofile/buffer_sync.c +++ b/drivers/oprofile/buffer_sync.c | |||
@@ -328,6 +328,8 @@ static void add_trace_begin(void) | |||
328 | add_event_entry(TRACE_BEGIN_CODE); | 328 | add_event_entry(TRACE_BEGIN_CODE); |
329 | } | 329 | } |
330 | 330 | ||
331 | #ifdef CONFIG_OPROFILE_IBS | ||
332 | |||
331 | #define IBS_FETCH_CODE_SIZE 2 | 333 | #define IBS_FETCH_CODE_SIZE 2 |
332 | #define IBS_OP_CODE_SIZE 5 | 334 | #define IBS_OP_CODE_SIZE 5 |
333 | #define IBS_EIP(offset) \ | 335 | #define IBS_EIP(offset) \ |
@@ -390,6 +392,8 @@ static void add_ibs_begin(struct oprofile_cpu_buffer *cpu_buf, int code, | |||
390 | } | 392 | } |
391 | } | 393 | } |
392 | 394 | ||
395 | #endif | ||
396 | |||
393 | static void add_sample_entry(unsigned long offset, unsigned long event) | 397 | static void add_sample_entry(unsigned long offset, unsigned long event) |
394 | { | 398 | { |
395 | add_event_entry(offset); | 399 | add_event_entry(offset); |
@@ -586,6 +590,7 @@ void sync_buffer(int cpu) | |||
586 | } else if (s->event == CPU_TRACE_BEGIN) { | 590 | } else if (s->event == CPU_TRACE_BEGIN) { |
587 | state = sb_bt_start; | 591 | state = sb_bt_start; |
588 | add_trace_begin(); | 592 | add_trace_begin(); |
593 | #ifdef CONFIG_OPROFILE_IBS | ||
589 | } else if (s->event == IBS_FETCH_BEGIN) { | 594 | } else if (s->event == IBS_FETCH_BEGIN) { |
590 | state = sb_bt_start; | 595 | state = sb_bt_start; |
591 | add_ibs_begin(cpu_buf, | 596 | add_ibs_begin(cpu_buf, |
@@ -594,6 +599,7 @@ void sync_buffer(int cpu) | |||
594 | state = sb_bt_start; | 599 | state = sb_bt_start; |
595 | add_ibs_begin(cpu_buf, | 600 | add_ibs_begin(cpu_buf, |
596 | IBS_OP_CODE, in_kernel, mm); | 601 | IBS_OP_CODE, in_kernel, mm); |
602 | #endif | ||
597 | } else { | 603 | } else { |
598 | struct mm_struct *oldmm = mm; | 604 | struct mm_struct *oldmm = mm; |
599 | 605 | ||
diff --git a/drivers/oprofile/cpu_buffer.c b/drivers/oprofile/cpu_buffer.c index c9ac4e156918..aba905b3afb8 100644 --- a/drivers/oprofile/cpu_buffer.c +++ b/drivers/oprofile/cpu_buffer.c | |||
@@ -253,6 +253,8 @@ void oprofile_add_sample(struct pt_regs * const regs, unsigned long event) | |||
253 | oprofile_add_ext_sample(pc, regs, event, is_kernel); | 253 | oprofile_add_ext_sample(pc, regs, event, is_kernel); |
254 | } | 254 | } |
255 | 255 | ||
256 | #ifdef CONFIG_OPROFILE_IBS | ||
257 | |||
256 | #define MAX_IBS_SAMPLE_SIZE 14 | 258 | #define MAX_IBS_SAMPLE_SIZE 14 |
257 | static int log_ibs_sample(struct oprofile_cpu_buffer *cpu_buf, | 259 | static int log_ibs_sample(struct oprofile_cpu_buffer *cpu_buf, |
258 | unsigned long pc, int is_kernel, unsigned int *ibs, int ibs_code) | 260 | unsigned long pc, int is_kernel, unsigned int *ibs, int ibs_code) |
@@ -318,6 +320,8 @@ void oprofile_add_ibs_sample(struct pt_regs *const regs, | |||
318 | oprofile_ops.backtrace(regs, backtrace_depth); | 320 | oprofile_ops.backtrace(regs, backtrace_depth); |
319 | } | 321 | } |
320 | 322 | ||
323 | #endif | ||
324 | |||
321 | void oprofile_add_pc(unsigned long pc, int is_kernel, unsigned long event) | 325 | void oprofile_add_pc(unsigned long pc, int is_kernel, unsigned long event) |
322 | { | 326 | { |
323 | struct oprofile_cpu_buffer *cpu_buf = &__get_cpu_var(cpu_buffer); | 327 | struct oprofile_cpu_buffer *cpu_buf = &__get_cpu_var(cpu_buffer); |