Donate to e Foundation | Murena handsets with /e/OS | Own a part of Murena! Learn more

Commit 43315956 authored by Ingo Molnar's avatar Ingo Molnar
Browse files

Merge branch 'perf/core' into perf/probes 



Conflicts:
	tools/perf/Makefile

Merge reason:

 - fix the conflict
 - pick up the pr_*() infrastructure to queue up dependent patch

Signed-off-by: default avatarIngo Molnar <mingo@elte.hu>
parents 9bf4e7fb 6beba7ad

Documentation/kernel-parameters.txt

+7 −0
Original line number Diff line number Diff line
@@ -779,6 +779,13 @@ and is between 256 and 4096 characters. It is defined in the file 
			by the set_ftrace_notrace file in the debugfs

			tracing directory.



	ftrace_graph_filter=[function-list]

			[FTRACE] Limit the top level callers functions traced

			by the function graph tracer at boot up.

			function-list is a comma separated list of functions

			that can be changed at run time by the

			set_graph_function file in the debugfs tracing directory.



	gamecon.map[2|3]=

			[HW,JOY] Multisystem joystick and NES/SNES/PSX pad

			support via parallel port (up to 5 devices per port)

Documentation/trace/ftrace-design.txt

+11 −2
Original line number Diff line number Diff line
@@ -213,10 +213,19 @@ If you can't trace NMI functions, then skip this option. 
<details to be filled>





HAVE_FTRACE_SYSCALLS

HAVE_SYSCALL_TRACEPOINTS

---------------------



<details to be filled>

You need very few things to get the syscalls tracing in an arch.



- Have a NR_syscalls variable in <asm/unistd.h> that provides the number

  of syscalls supported by the arch.

- Implement arch_syscall_addr() that resolves a syscall address from a

  syscall number.

- Support the TIF_SYSCALL_TRACEPOINT thread flags

- Put the trace_sys_enter() and trace_sys_exit() tracepoints calls from ptrace

  in the ptrace syscalls tracing path.

- Tag this arch as HAVE_SYSCALL_TRACEPOINTS.





HAVE_FTRACE_MCOUNT_RECORD

arch/s390/kernel/ftrace.c

+2 −65
Original line number Diff line number Diff line
@@ -203,73 +203,10 @@ unsigned long prepare_ftrace_return(unsigned long ip, unsigned long parent) 


#ifdef CONFIG_FTRACE_SYSCALLS



extern unsigned long __start_syscalls_metadata[];

extern unsigned long __stop_syscalls_metadata[];

extern unsigned int sys_call_table[];



static struct syscall_metadata **syscalls_metadata;



struct syscall_metadata *syscall_nr_to_meta(int nr)

{

	if (!syscalls_metadata || nr >= NR_syscalls || nr < 0)

		return NULL;



	return syscalls_metadata[nr];

}



int syscall_name_to_nr(char *name)

{

	int i;



	if (!syscalls_metadata)

		return -1;

	for (i = 0; i < NR_syscalls; i++)

		if (syscalls_metadata[i])

			if (!strcmp(syscalls_metadata[i]->name, name))

				return i;

	return -1;

}



void set_syscall_enter_id(int num, int id)

{

	syscalls_metadata[num]->enter_id = id;

}



void set_syscall_exit_id(int num, int id)

unsigned long __init arch_syscall_addr(int nr)

{

	syscalls_metadata[num]->exit_id = id;

}



static struct syscall_metadata *find_syscall_meta(unsigned long syscall)

{

	struct syscall_metadata *start;

	struct syscall_metadata *stop;

	char str[KSYM_SYMBOL_LEN];



	start = (struct syscall_metadata *)__start_syscalls_metadata;

	stop = (struct syscall_metadata *)__stop_syscalls_metadata;

	kallsyms_lookup(syscall, NULL, NULL, NULL, str);



	for ( ; start < stop; start++) {

		if (start->name && !strcmp(start->name + 3, str + 3))

			return start;

	}

	return NULL;

}



static int __init arch_init_ftrace_syscalls(void)

{

	struct syscall_metadata *meta;

	int i;

	syscalls_metadata = kzalloc(sizeof(*syscalls_metadata) * NR_syscalls,

				    GFP_KERNEL);

	if (!syscalls_metadata)

		return -ENOMEM;

	for (i = 0; i < NR_syscalls; i++) {

		meta = find_syscall_meta((unsigned long)sys_call_table[i]);

		syscalls_metadata[i] = meta;

	}

	return 0;

	return (unsigned long)sys_call_table[nr];

}

arch_initcall(arch_init_ftrace_syscalls);

#endif

arch/x86/include/asm/perf_event.h

+12 −1
Original line number Diff line number Diff line
@@ -28,6 +28,17 @@ 
 */

#define ARCH_PERFMON_EVENT_MASK				    0xffff



/*

 * filter mask to validate fixed counter events.

 * the following filters disqualify for fixed counters:

 *  - inv

 *  - edge

 *  - cnt-mask

 *  The other filters are supported by fixed counters.

 *  The any-thread option is supported starting with v3.

 */

#define ARCH_PERFMON_EVENT_FILTER_MASK			0xff840000



#define ARCH_PERFMON_UNHALTED_CORE_CYCLES_SEL		      0x3c

#define ARCH_PERFMON_UNHALTED_CORE_CYCLES_UMASK		(0x00 << 8)

#define ARCH_PERFMON_UNHALTED_CORE_CYCLES_INDEX			 0

arch/x86/kernel/cpu/perf_event.c

+170 −15
Original line number Diff line number Diff line
@@ -77,6 +77,18 @@ struct cpu_hw_events { 
	struct debug_store	*ds;

};



struct event_constraint {

	unsigned long	idxmsk[BITS_TO_LONGS(X86_PMC_IDX_MAX)];

	int		code;

};



#define EVENT_CONSTRAINT(c, m) { .code = (c), .idxmsk[0] = (m) }

#define EVENT_CONSTRAINT_END  { .code = 0, .idxmsk[0] = 0 }



#define for_each_event_constraint(e, c) \

	for ((e) = (c); (e)->idxmsk[0]; (e)++)





/*

 * struct x86_pmu - generic x86 pmu

 */
@@ -102,6 +114,8 @@ struct x86_pmu { 
	u64		intel_ctrl;

	void		(*enable_bts)(u64 config);

	void		(*disable_bts)(void);

	int		(*get_event_idx)(struct cpu_hw_events *cpuc,

					 struct hw_perf_event *hwc);

};



static struct x86_pmu x86_pmu __read_mostly;
@@ -110,6 +124,8 @@ static DEFINE_PER_CPU(struct cpu_hw_events, cpu_hw_events) = { 
	.enabled = 1,

};



static const struct event_constraint *event_constraints;



/*

 * Not sure about some of these

 */
@@ -155,6 +171,16 @@ static u64 p6_pmu_raw_event(u64 hw_event) 
	return hw_event & P6_EVNTSEL_MASK;

}



static const struct event_constraint intel_p6_event_constraints[] =

{

	EVENT_CONSTRAINT(0xc1, 0x1),	/* FLOPS */

	EVENT_CONSTRAINT(0x10, 0x1),	/* FP_COMP_OPS_EXE */

	EVENT_CONSTRAINT(0x11, 0x1),	/* FP_ASSIST */

	EVENT_CONSTRAINT(0x12, 0x2),	/* MUL */

	EVENT_CONSTRAINT(0x13, 0x2),	/* DIV */

	EVENT_CONSTRAINT(0x14, 0x1),	/* CYCLES_DIV_BUSY */

	EVENT_CONSTRAINT_END

};



/*

 * Intel PerfMon v3. Used on Core2 and later.
@@ -170,6 +196,35 @@ static const u64 intel_perfmon_event_map[] = 
  [PERF_COUNT_HW_BUS_CYCLES]		= 0x013c,

};



static const struct event_constraint intel_core_event_constraints[] =

{

	EVENT_CONSTRAINT(0x10, 0x1),	/* FP_COMP_OPS_EXE */

	EVENT_CONSTRAINT(0x11, 0x2),	/* FP_ASSIST */

	EVENT_CONSTRAINT(0x12, 0x2),	/* MUL */

	EVENT_CONSTRAINT(0x13, 0x2),	/* DIV */

	EVENT_CONSTRAINT(0x14, 0x1),	/* CYCLES_DIV_BUSY */

	EVENT_CONSTRAINT(0x18, 0x1),	/* IDLE_DURING_DIV */

	EVENT_CONSTRAINT(0x19, 0x2),	/* DELAYED_BYPASS */

	EVENT_CONSTRAINT(0xa1, 0x1),	/* RS_UOPS_DISPATCH_CYCLES */

	EVENT_CONSTRAINT(0xcb, 0x1),	/* MEM_LOAD_RETIRED */

	EVENT_CONSTRAINT_END

};



static const struct event_constraint intel_nehalem_event_constraints[] =

{

	EVENT_CONSTRAINT(0x40, 0x3),	/* L1D_CACHE_LD */

	EVENT_CONSTRAINT(0x41, 0x3),	/* L1D_CACHE_ST */

	EVENT_CONSTRAINT(0x42, 0x3),	/* L1D_CACHE_LOCK */

	EVENT_CONSTRAINT(0x43, 0x3),	/* L1D_ALL_REF */

	EVENT_CONSTRAINT(0x4e, 0x3),	/* L1D_PREFETCH */

	EVENT_CONSTRAINT(0x4c, 0x3),	/* LOAD_HIT_PRE */

	EVENT_CONSTRAINT(0x51, 0x3),	/* L1D */

	EVENT_CONSTRAINT(0x52, 0x3),	/* L1D_CACHE_PREFETCH_LOCK_FB_HIT */

	EVENT_CONSTRAINT(0x53, 0x3),	/* L1D_CACHE_LOCK_FB_HIT */

	EVENT_CONSTRAINT(0xc5, 0x3),	/* CACHE_LOCK_CYCLES */

	EVENT_CONSTRAINT_END

};



static u64 intel_pmu_event_map(int hw_event)

{

	return intel_perfmon_event_map[hw_event];
@@ -932,6 +987,8 @@ static int __hw_perf_event_init(struct perf_event *event) 
	 */

	hwc->config = ARCH_PERFMON_EVENTSEL_INT;



	hwc->idx = -1;



	/*

	 * Count user and OS events unless requested not to.

	 */
@@ -1334,8 +1391,7 @@ static void amd_pmu_enable_event(struct hw_perf_event *hwc, int idx) 
		x86_pmu_enable_event(hwc, idx);

}



static int

fixed_mode_idx(struct perf_event *event, struct hw_perf_event *hwc)

static int fixed_mode_idx(struct hw_perf_event *hwc)

{

	unsigned int hw_event;
@@ -1349,6 +1405,12 @@ fixed_mode_idx(struct perf_event *event, struct hw_perf_event *hwc) 
	if (!x86_pmu.num_events_fixed)

		return -1;



	/*

	 * fixed counters do not take all possible filters

	 */

	if (hwc->config & ARCH_PERFMON_EVENT_FILTER_MASK)

		return -1;



	if (unlikely(hw_event == x86_pmu.event_map(PERF_COUNT_HW_INSTRUCTIONS)))

		return X86_PMC_IDX_FIXED_INSTRUCTIONS;

	if (unlikely(hw_event == x86_pmu.event_map(PERF_COUNT_HW_CPU_CYCLES)))
@@ -1360,15 +1422,50 @@ fixed_mode_idx(struct perf_event *event, struct hw_perf_event *hwc) 
}



/*

 * Find a PMC slot for the freshly enabled / scheduled in event:

 * generic counter allocator: get next free counter

 */

static int x86_pmu_enable(struct perf_event *event)

static int

gen_get_event_idx(struct cpu_hw_events *cpuc, struct hw_perf_event *hwc)

{

	int idx;



	idx = find_first_zero_bit(cpuc->used_mask, x86_pmu.num_events);

	return idx == x86_pmu.num_events ? -1 : idx;

}



/*

 * intel-specific counter allocator: check event constraints

 */

static int

intel_get_event_idx(struct cpu_hw_events *cpuc, struct hw_perf_event *hwc)

{

	const struct event_constraint *event_constraint;

	int i, code;



	if (!event_constraints)

		goto skip;



	code = hwc->config & CORE_EVNTSEL_EVENT_MASK;



	for_each_event_constraint(event_constraint, event_constraints) {

		if (code == event_constraint->code) {

			for_each_bit(i, event_constraint->idxmsk, X86_PMC_IDX_MAX) {

				if (!test_and_set_bit(i, cpuc->used_mask))

					return i;

			}

			return -1;

		}

	}

skip:

	return gen_get_event_idx(cpuc, hwc);

}



static int

x86_schedule_event(struct cpu_hw_events *cpuc, struct hw_perf_event *hwc)

{

	struct cpu_hw_events *cpuc = &__get_cpu_var(cpu_hw_events);

	struct hw_perf_event *hwc = &event->hw;

	int idx;



	idx = fixed_mode_idx(event, hwc);

	idx = fixed_mode_idx(hwc);

	if (idx == X86_PMC_IDX_FIXED_BTS) {

		/* BTS is already occupied. */

		if (test_and_set_bit(idx, cpuc->used_mask))
@@ -1396,11 +1493,10 @@ static int x86_pmu_enable(struct perf_event *event) 
	} else {

		idx = hwc->idx;

		/* Try to get the previous generic event again */

		if (test_and_set_bit(idx, cpuc->used_mask)) {

		if (idx == -1 || test_and_set_bit(idx, cpuc->used_mask)) {

try_generic:

			idx = find_first_zero_bit(cpuc->used_mask,

						  x86_pmu.num_events);

			if (idx == x86_pmu.num_events)

			idx = x86_pmu.get_event_idx(cpuc, hwc);

			if (idx == -1)

				return -EAGAIN;



			set_bit(idx, cpuc->used_mask);
@@ -1410,6 +1506,22 @@ static int x86_pmu_enable(struct perf_event *event) 
		hwc->event_base  = x86_pmu.perfctr;

	}



	return idx;

}



/*

 * Find a PMC slot for the freshly enabled / scheduled in event:

 */

static int x86_pmu_enable(struct perf_event *event)

{

	struct cpu_hw_events *cpuc = &__get_cpu_var(cpu_hw_events);

	struct hw_perf_event *hwc = &event->hw;

	int idx;



	idx = x86_schedule_event(cpuc, hwc);

	if (idx < 0)

		return idx;



	perf_events_lapic_init();



	x86_pmu.disable(hwc, idx);
@@ -1877,6 +1989,7 @@ static struct x86_pmu p6_pmu = { 
	 */

	.event_bits		= 32,

	.event_mask		= (1ULL << 32) - 1,

	.get_event_idx		= intel_get_event_idx,

};



static struct x86_pmu intel_pmu = {
@@ -1900,6 +2013,7 @@ static struct x86_pmu intel_pmu = { 
	.max_period		= (1ULL << 31) - 1,

	.enable_bts		= intel_pmu_enable_bts,

	.disable_bts		= intel_pmu_disable_bts,

	.get_event_idx		= intel_get_event_idx,

};



static struct x86_pmu amd_pmu = {
@@ -1920,6 +2034,7 @@ static struct x86_pmu amd_pmu = { 
	.apic			= 1,

	/* use highest bit to detect overflow */

	.max_period		= (1ULL << 47) - 1,

	.get_event_idx		= gen_get_event_idx,

};



static int p6_pmu_init(void)
@@ -1932,10 +2047,12 @@ static int p6_pmu_init(void) 
	case 7:

	case 8:

	case 11: /* Pentium III */

		event_constraints = intel_p6_event_constraints;

		break;

	case 9:

	case 13:

		/* Pentium M */

		event_constraints = intel_p6_event_constraints;

		break;

	default:

		pr_cont("unsupported p6 CPU model %d ",
@@ -2007,12 +2124,14 @@ static int intel_pmu_init(void) 
		       sizeof(hw_cache_event_ids));



		pr_cont("Core2 events, ");

		event_constraints = intel_core_event_constraints;

		break;

	default:

	case 26:

		memcpy(hw_cache_event_ids, nehalem_hw_cache_event_ids,

		       sizeof(hw_cache_event_ids));



		event_constraints = intel_nehalem_event_constraints;

		pr_cont("Nehalem/Corei7 events, ");

		break;

	case 28:
@@ -2105,11 +2224,47 @@ static const struct pmu pmu = { 
	.unthrottle	= x86_pmu_unthrottle,

};



static int

validate_event(struct cpu_hw_events *cpuc, struct perf_event *event)

{

	struct hw_perf_event fake_event = event->hw;



	if (event->pmu != &pmu)

		return 0;



	return x86_schedule_event(cpuc, &fake_event);

}



static int validate_group(struct perf_event *event)

{

	struct perf_event *sibling, *leader = event->group_leader;

	struct cpu_hw_events fake_pmu;



	memset(&fake_pmu, 0, sizeof(fake_pmu));



	if (!validate_event(&fake_pmu, leader))

		return -ENOSPC;



	list_for_each_entry(sibling, &leader->sibling_list, group_entry) {

		if (!validate_event(&fake_pmu, sibling))

			return -ENOSPC;

	}



	if (!validate_event(&fake_pmu, event))

		return -ENOSPC;



	return 0;

}



const struct pmu *hw_perf_event_init(struct perf_event *event)

{

	int err;



	err = __hw_perf_event_init(event);

	if (!err) {

		if (event->group_leader != event)

			err = validate_group(event);

	}

	if (err) {

		if (event->destroy)

			event->destroy(event);