[linux-2.6-block.git] / arch / powerpc / perf / hv-gpci.c

/*
 * Hypervisor supplied "gpci" ("get performance counter info") performance
 * counter support
 *
 * Author: Cody P Schafer <cody@linux.vnet.ibm.com>
 * Copyright 2014 IBM Corporation.
 *
 * This program is free software; you can redistribute it and/or
 * modify it under the terms of the GNU General Public License
 * as published by the Free Software Foundation; either version
 * 2 of the License, or (at your option) any later version.
 */

#define pr_fmt(fmt) "hv-gpci: " fmt

#include <linux/init.h>
#include <linux/perf_event.h>
#include <asm/firmware.h>
#include <asm/hvcall.h>
#include <asm/io.h>

#include "hv-gpci.h"
#include "hv-common.h"

/*
 * Example usage:
 *  perf stat -e 'hv_gpci/counter_info_version=3,offset=0,length=8,
 *		  secondary_index=0,starting_index=0xffffffff,request=0x10/' ...
 */

/* u32 */
EVENT_DEFINE_RANGE_FORMAT(request, config, 0, 31);
/* u32 */
/*
 * Note that starting_index, phys_processor_idx, sibling_part_id,
 * hw_chip_id, partition_id all refer to the same bit range. They
 * are basically aliases for the starting_index. The specific alias
 * used depends on the event. See REQUEST_IDX_KIND in hv-gpci-requests.h
 */
EVENT_DEFINE_RANGE_FORMAT(starting_index, config, 32, 63);
EVENT_DEFINE_RANGE_FORMAT_LITE(phys_processor_idx, config, 32, 63);
EVENT_DEFINE_RANGE_FORMAT_LITE(sibling_part_id, config, 32, 63);
EVENT_DEFINE_RANGE_FORMAT_LITE(hw_chip_id, config, 32, 63);
EVENT_DEFINE_RANGE_FORMAT_LITE(partition_id, config, 32, 63);

/* u16 */
EVENT_DEFINE_RANGE_FORMAT(secondary_index, config1, 0, 15);
/* u8 */
EVENT_DEFINE_RANGE_FORMAT(counter_info_version, config1, 16, 23);
/* u8, bytes of data (1-8) */
EVENT_DEFINE_RANGE_FORMAT(length, config1, 24, 31);
/* u32, byte offset */
EVENT_DEFINE_RANGE_FORMAT(offset, config1, 32, 63);

static struct attribute *format_attrs[] = {
	&format_attr_request.attr,
	&format_attr_starting_index.attr,
	&format_attr_phys_processor_idx.attr,
	&format_attr_sibling_part_id.attr,
	&format_attr_hw_chip_id.attr,
	&format_attr_partition_id.attr,
	&format_attr_secondary_index.attr,
	&format_attr_counter_info_version.attr,

	&format_attr_offset.attr,
	&format_attr_length.attr,
	NULL,
};

static struct attribute_group format_group = {
	.name = "format",
	.attrs = format_attrs,
};

static struct attribute_group event_group = {
	.name  = "events",
	.attrs = hv_gpci_event_attrs,
};

#define HV_CAPS_ATTR(_name, _format)				\
static ssize_t _name##_show(struct device *dev,			\
			    struct device_attribute *attr,	\
			    char *page)				\
{								\
	struct hv_perf_caps caps;				\
	unsigned long hret = hv_perf_caps_get(&caps);		\
	if (hret)						\
		return -EIO;					\
								\
	return sprintf(page, _format, caps._name);		\
}								\
static struct device_attribute hv_caps_attr_##_name = __ATTR_RO(_name)

static ssize_t kernel_version_show(struct device *dev,
				   struct device_attribute *attr,
				   char *page)
{
	return sprintf(page, "0x%x\n", COUNTER_INFO_VERSION_CURRENT);
}

static DEVICE_ATTR_RO(kernel_version);
HV_CAPS_ATTR(version, "0x%x\n");
HV_CAPS_ATTR(ga, "%d\n");
HV_CAPS_ATTR(expanded, "%d\n");
HV_CAPS_ATTR(lab, "%d\n");
HV_CAPS_ATTR(collect_privileged, "%d\n");

static struct attribute *interface_attrs[] = {
	&dev_attr_kernel_version.attr,
	&hv_caps_attr_version.attr,
	&hv_caps_attr_ga.attr,
	&hv_caps_attr_expanded.attr,
	&hv_caps_attr_lab.attr,
	&hv_caps_attr_collect_privileged.attr,
	NULL,
};

static struct attribute_group interface_group = {
	.name = "interface",
	.attrs = interface_attrs,
};

static const struct attribute_group *attr_groups[] = {
	&format_group,
	&event_group,
	&interface_group,
	NULL,
};

#define HGPCI_REQ_BUFFER_SIZE	4096
#define HGPCI_MAX_DATA_BYTES \
	(HGPCI_REQ_BUFFER_SIZE - sizeof(struct hv_get_perf_counter_info_params))

DEFINE_PER_CPU(char, hv_gpci_reqb[HGPCI_REQ_BUFFER_SIZE]) __aligned(sizeof(uint64_t));

struct hv_gpci_request_buffer {
	struct hv_get_perf_counter_info_params params;
	uint8_t bytes[HGPCI_MAX_DATA_BYTES];
} __packed;

static unsigned long single_gpci_request(u32 req, u32 starting_index,
		u16 secondary_index, u8 version_in, u32 offset, u8 length,
		u64 *value)
{
	unsigned long ret;
	size_t i;
	u64 count;
	struct hv_gpci_request_buffer *arg;

	arg = (void *)get_cpu_var(hv_gpci_reqb);
	memset(arg, 0, HGPCI_REQ_BUFFER_SIZE);

	arg->params.counter_request = cpu_to_be32(req);
	arg->params.starting_index = cpu_to_be32(starting_index);
	arg->params.secondary_index = cpu_to_be16(secondary_index);
	arg->params.counter_info_version_in = version_in;

	ret = plpar_hcall_norets(H_GET_PERF_COUNTER_INFO,
			virt_to_phys(arg), HGPCI_REQ_BUFFER_SIZE);
	if (ret) {
		pr_devel("hcall failed: 0x%lx\n", ret);
		goto out;
	}

	/*
	 * we verify offset and length are within the zeroed buffer at event
	 * init.
	 */
	count = 0;
	for (i = offset; i < offset + length; i++)
		count |= arg->bytes[i] << (i - offset);

	*value = count;
out:
	put_cpu_var(hv_gpci_reqb);
	return ret;
}

static u64 h_gpci_get_value(struct perf_event *event)
{
	u64 count;
	unsigned long ret = single_gpci_request(event_get_request(event),
					event_get_starting_index(event),
					event_get_secondary_index(event),
					event_get_counter_info_version(event),
					event_get_offset(event),
					event_get_length(event),
					&count);
	if (ret)
		return 0;
	return count;
}

static void h_gpci_event_update(struct perf_event *event)
{
	s64 prev;
	u64 now = h_gpci_get_value(event);
	prev = local64_xchg(&event->hw.prev_count, now);
	local64_add(now - prev, &event->count);
}

static void h_gpci_event_start(struct perf_event *event, int flags)
{
	local64_set(&event->hw.prev_count, h_gpci_get_value(event));
}

static void h_gpci_event_stop(struct perf_event *event, int flags)
{
	h_gpci_event_update(event);
}

static int h_gpci_event_add(struct perf_event *event, int flags)
{
	if (flags & PERF_EF_START)
		h_gpci_event_start(event, flags);

	return 0;
}

static int h_gpci_event_init(struct perf_event *event)
{
	u64 count;
	u8 length;

	/* Not our event */
	if (event->attr.type != event->pmu->type)
		return -ENOENT;

	/* config2 is unused */
	if (event->attr.config2) {
		pr_devel("config2 set when reserved\n");
		return -EINVAL;
	}

	/* unsupported modes and filters */
	if (event->attr.exclude_user   ||
	    event->attr.exclude_kernel ||
	    event->attr.exclude_hv     ||
	    event->attr.exclude_idle   ||
	    event->attr.exclude_host   ||
	    event->attr.exclude_guest)
		return -EINVAL;

	/* no branch sampling */
	if (has_branch_stack(event))
		return -EOPNOTSUPP;

	length = event_get_length(event);
	if (length < 1 || length > 8) {
		pr_devel("length invalid\n");
		return -EINVAL;
	}

	/* last byte within the buffer? */
	if ((event_get_offset(event) + length) > HGPCI_MAX_DATA_BYTES) {
		pr_devel("request outside of buffer: %zu > %zu\n",
				(size_t)event_get_offset(event) + length,
				HGPCI_MAX_DATA_BYTES);
		return -EINVAL;
	}

	/* check if the request works... */
	if (single_gpci_request(event_get_request(event),
				event_get_starting_index(event),
				event_get_secondary_index(event),
				event_get_counter_info_version(event),
				event_get_offset(event),
				length,
				&count)) {
		pr_devel("gpci hcall failed\n");
		return -EINVAL;
	}

	return 0;
}

static struct pmu h_gpci_pmu = {
	.task_ctx_nr = perf_invalid_context,

	.name = "hv_gpci",
	.attr_groups = attr_groups,
	.event_init  = h_gpci_event_init,
	.add         = h_gpci_event_add,
	.del         = h_gpci_event_stop,
	.start       = h_gpci_event_start,
	.stop        = h_gpci_event_stop,
	.read        = h_gpci_event_update,
};

static int hv_gpci_init(void)
{
	int r;
	unsigned long hret;
	struct hv_perf_caps caps;

	hv_gpci_assert_offsets_correct();

	if (!firmware_has_feature(FW_FEATURE_LPAR)) {
		pr_debug("not a virtualized system, not enabling\n");
		return -ENODEV;
	}

	hret = hv_perf_caps_get(&caps);
	if (hret) {
		pr_debug("could not obtain capabilities, not enabling, rc=%ld\n",
				hret);
		return -ENODEV;
	}

	/* sampling not supported */
	h_gpci_pmu.capabilities |= PERF_PMU_CAP_NO_INTERRUPT;

	r = perf_pmu_register(&h_gpci_pmu, h_gpci_pmu.name, -1);
	if (r)
		return r;

	return 0;
}

device_initcall(hv_gpci_init);
Commit	Line	Data
220a0c60 CS	1	/*
	2	* Hypervisor supplied "gpci" ("get performance counter info") performance
	3	* counter support
	4	*
	5	* Author: Cody P Schafer <cody@linux.vnet.ibm.com>
	6	* Copyright 2014 IBM Corporation.
	7	*
	8	* This program is free software; you can redistribute it and/or
	9	* modify it under the terms of the GNU General Public License
	10	* as published by the Free Software Foundation; either version
	11	* 2 of the License, or (at your option) any later version.
	12	*/
	13
	14	#define pr_fmt(fmt) "hv-gpci: " fmt
	15
	16	#include <linux/init.h>
	17	#include <linux/perf_event.h>
	18	#include <asm/firmware.h>
	19	#include <asm/hvcall.h>
	20	#include <asm/io.h>
	21
	22	#include "hv-gpci.h"
	23	#include "hv-common.h"
	24
	25	/*
	26	* Example usage:
	27	* perf stat -e 'hv_gpci/counter_info_version=3,offset=0,length=8,
	28	* secondary_index=0,starting_index=0xffffffff,request=0x10/' ...
	29	*/
	30
	31	/* u32 */
	32	EVENT_DEFINE_RANGE_FORMAT(request, config, 0, 31);
	33	/* u32 */
9e9f6010 CS	34	/*
	35	* Note that starting_index, phys_processor_idx, sibling_part_id,
	36	* hw_chip_id, partition_id all refer to the same bit range. They
	37	* are basically aliases for the starting_index. The specific alias
	38	* used depends on the event. See REQUEST_IDX_KIND in hv-gpci-requests.h
	39	*/
220a0c60	40	EVENT_DEFINE_RANGE_FORMAT(starting_index, config, 32, 63);
9e9f6010 CS	41	EVENT_DEFINE_RANGE_FORMAT_LITE(phys_processor_idx, config, 32, 63);
	42	EVENT_DEFINE_RANGE_FORMAT_LITE(sibling_part_id, config, 32, 63);
	43	EVENT_DEFINE_RANGE_FORMAT_LITE(hw_chip_id, config, 32, 63);
	44	EVENT_DEFINE_RANGE_FORMAT_LITE(partition_id, config, 32, 63);
	45
220a0c60 CS	46	/* u16 */
	47	EVENT_DEFINE_RANGE_FORMAT(secondary_index, config1, 0, 15);
	48	/* u8 */
	49	EVENT_DEFINE_RANGE_FORMAT(counter_info_version, config1, 16, 23);
	50	/* u8, bytes of data (1-8) */
	51	EVENT_DEFINE_RANGE_FORMAT(length, config1, 24, 31);
	52	/* u32, byte offset */
	53	EVENT_DEFINE_RANGE_FORMAT(offset, config1, 32, 63);
	54
	55	static struct attribute *format_attrs[] = {
	56	&format_attr_request.attr,
	57	&format_attr_starting_index.attr,
9e9f6010 CS	58	&format_attr_phys_processor_idx.attr,
	59	&format_attr_sibling_part_id.attr,
	60	&format_attr_hw_chip_id.attr,
	61	&format_attr_partition_id.attr,
220a0c60 CS	62	&format_attr_secondary_index.attr,
	63	&format_attr_counter_info_version.attr,
	64
	65	&format_attr_offset.attr,
	66	&format_attr_length.attr,
	67	NULL,
	68	};
	69
	70	static struct attribute_group format_group = {
	71	.name = "format",
	72	.attrs = format_attrs,
	73	};
	74
9e9f6010 CS	75	static struct attribute_group event_group = {
	76	.name = "events",
	77	.attrs = hv_gpci_event_attrs,
	78	};
	79
220a0c60 CS	80	#define HV_CAPS_ATTR(_name, _format) \
	81	static ssize_t _name##_show(struct device *dev, \
	82	struct device_attribute *attr, \
	83	char *page) \
	84	{ \
	85	struct hv_perf_caps caps; \
	86	unsigned long hret = hv_perf_caps_get(&caps); \
	87	if (hret) \
	88	return -EIO; \
	89	\
	90	return sprintf(page, _format, caps._name); \
	91	} \
	92	static struct device_attribute hv_caps_attr_##_name = __ATTR_RO(_name)
	93
	94	static ssize_t kernel_version_show(struct device *dev,
	95	struct device_attribute *attr,
	96	char *page)
	97	{
	98	return sprintf(page, "0x%x\n", COUNTER_INFO_VERSION_CURRENT);
	99	}
	100
58a685c2	101	static DEVICE_ATTR_RO(kernel_version);
220a0c60 CS	102	HV_CAPS_ATTR(version, "0x%x\n");
	103	HV_CAPS_ATTR(ga, "%d\n");
	104	HV_CAPS_ATTR(expanded, "%d\n");
	105	HV_CAPS_ATTR(lab, "%d\n");
	106	HV_CAPS_ATTR(collect_privileged, "%d\n");
	107
	108	static struct attribute *interface_attrs[] = {
	109	&dev_attr_kernel_version.attr,
	110	&hv_caps_attr_version.attr,
	111	&hv_caps_attr_ga.attr,
	112	&hv_caps_attr_expanded.attr,
	113	&hv_caps_attr_lab.attr,
	114	&hv_caps_attr_collect_privileged.attr,
	115	NULL,
	116	};
	117
	118	static struct attribute_group interface_group = {
	119	.name = "interface",
	120	.attrs = interface_attrs,
	121	};
	122
	123	static const struct attribute_group *attr_groups[] = {
	124	&format_group,
9e9f6010	125	&event_group,
220a0c60 CS	126	&interface_group,
	127	NULL,
	128	};
	129
e4f226b1 SB	130	#define HGPCI_REQ_BUFFER_SIZE 4096
	131	#define HGPCI_MAX_DATA_BYTES \
	132	(HGPCI_REQ_BUFFER_SIZE - sizeof(struct hv_get_perf_counter_info_params))
	133
	134	DEFINE_PER_CPU(char, hv_gpci_reqb[HGPCI_REQ_BUFFER_SIZE]) __aligned(sizeof(uint64_t));
	135
	136	struct hv_gpci_request_buffer {
	137	struct hv_get_perf_counter_info_params params;
	138	uint8_t bytes[HGPCI_MAX_DATA_BYTES];
	139	} __packed;
220a0c60 CS	140
	141	static unsigned long single_gpci_request(u32 req, u32 starting_index,
	142	u16 secondary_index, u8 version_in, u32 offset, u8 length,
	143	u64 *value)
	144	{
	145	unsigned long ret;
	146	size_t i;
	147	u64 count;
e4f226b1 SB	148	struct hv_gpci_request_buffer *arg;
	149
	150	arg = (void *)get_cpu_var(hv_gpci_reqb);
	151	memset(arg, 0, HGPCI_REQ_BUFFER_SIZE);
220a0c60	152
e4f226b1 SB	153	arg->params.counter_request = cpu_to_be32(req);
	154	arg->params.starting_index = cpu_to_be32(starting_index);
	155	arg->params.secondary_index = cpu_to_be16(secondary_index);
	156	arg->params.counter_info_version_in = version_in;
220a0c60 CS	157
220a0c60 CS	158	ret = plpar_hcall_norets(H_GET_PERF_COUNTER_INFO,
e4f226b1	159	virt_to_phys(arg), HGPCI_REQ_BUFFER_SIZE);
220a0c60 CS	160	if (ret) {
220a0c60 CS	161	pr_devel("hcall failed: 0x%lx\n", ret);
e4f226b1	162	goto out;
220a0c60 CS	163	}
	164
	165	/*
	166	* we verify offset and length are within the zeroed buffer at event
	167	* init.
	168	*/
	169	count = 0;
	170	for (i = offset; i < offset + length; i++)
e4f226b1	171	count \|= arg->bytes[i] << (i - offset);
220a0c60 CS	172
220a0c60 CS	173	*value = count;
e4f226b1 SB	174	out:
e4f226b1 SB	175	put_cpu_var(hv_gpci_reqb);
220a0c60 CS	176	return ret;
	177	}
	178
	179	static u64 h_gpci_get_value(struct perf_event *event)
	180	{
	181	u64 count;
	182	unsigned long ret = single_gpci_request(event_get_request(event),
	183	event_get_starting_index(event),
	184	event_get_secondary_index(event),
	185	event_get_counter_info_version(event),
	186	event_get_offset(event),
	187	event_get_length(event),
	188	&count);
	189	if (ret)
	190	return 0;
	191	return count;
	192	}
	193
	194	static void h_gpci_event_update(struct perf_event *event)
	195	{
	196	s64 prev;
	197	u64 now = h_gpci_get_value(event);
	198	prev = local64_xchg(&event->hw.prev_count, now);
	199	local64_add(now - prev, &event->count);
	200	}
	201
	202	static void h_gpci_event_start(struct perf_event *event, int flags)
	203	{
	204	local64_set(&event->hw.prev_count, h_gpci_get_value(event));
	205	}
	206
	207	static void h_gpci_event_stop(struct perf_event *event, int flags)
	208	{
	209	h_gpci_event_update(event);
	210	}
	211
	212	static int h_gpci_event_add(struct perf_event *event, int flags)
	213	{
	214	if (flags & PERF_EF_START)
	215	h_gpci_event_start(event, flags);
	216
	217	return 0;
	218	}
	219
	220	static int h_gpci_event_init(struct perf_event *event)
	221	{
	222	u64 count;
	223	u8 length;
	224
	225	/* Not our event */
	226	if (event->attr.type != event->pmu->type)
	227	return -ENOENT;
	228
	229	/* config2 is unused */
	230	if (event->attr.config2) {
	231	pr_devel("config2 set when reserved\n");
	232	return -EINVAL;
	233	}
	234
	235	/* unsupported modes and filters */
	236	if (event->attr.exclude_user \|\|
	237	event->attr.exclude_kernel \|\|
	238	event->attr.exclude_hv \|\|
	239	event->attr.exclude_idle \|\|
240	event->attr.exclude_host \|\|
cc56d673	241	event->attr.exclude_guest)
220a0c60 CS	242	return -EINVAL;
	243
	244	/* no branch sampling */
	245	if (has_branch_stack(event))
	246	return -EOPNOTSUPP;
	247
	248	length = event_get_length(event);
	249	if (length < 1 \|\| length > 8) {
	250	pr_devel("length invalid\n");
	251	return -EINVAL;
	252	}
	253
	254	/* last byte within the buffer? */
e4f226b1	255	if ((event_get_offset(event) + length) > HGPCI_MAX_DATA_BYTES) {
220a0c60 CS	256	pr_devel("request outside of buffer: %zu > %zu\n",
220a0c60 CS	257	(size_t)event_get_offset(event) + length,
e4f226b1	258	HGPCI_MAX_DATA_BYTES);
220a0c60 CS	259	return -EINVAL;
	260	}
	261
	262	/* check if the request works... */
	263	if (single_gpci_request(event_get_request(event),
	264	event_get_starting_index(event),
	265	event_get_secondary_index(event),
	266	event_get_counter_info_version(event),
	267	event_get_offset(event),
	268	length,
	269	&count)) {
	270	pr_devel("gpci hcall failed\n");
	271	return -EINVAL;
	272	}
	273
	274	return 0;
	275	}
	276
220a0c60 CS	277	static struct pmu h_gpci_pmu = {
	278	.task_ctx_nr = perf_invalid_context,
	279
	280	.name = "hv_gpci",
	281	.attr_groups = attr_groups,
	282	.event_init = h_gpci_event_init,
	283	.add = h_gpci_event_add,
	284	.del = h_gpci_event_stop,
	285	.start = h_gpci_event_start,
	286	.stop = h_gpci_event_stop,
	287	.read = h_gpci_event_update,
220a0c60 CS	288	};
	289
	290	static int hv_gpci_init(void)
	291	{
	292	int r;
	293	unsigned long hret;
	294	struct hv_perf_caps caps;
	295
9e9f6010 CS	296	hv_gpci_assert_offsets_correct();
9e9f6010 CS	297
220a0c60	298	if (!firmware_has_feature(FW_FEATURE_LPAR)) {
0a8cf9e2	299	pr_debug("not a virtualized system, not enabling\n");
220a0c60 CS	300	return -ENODEV;
	301	}
	302
	303	hret = hv_perf_caps_get(&caps);
	304	if (hret) {
0a8cf9e2	305	pr_debug("could not obtain capabilities, not enabling, rc=%ld\n",
220a0c60 CS	306	hret);
	307	return -ENODEV;
	308	}
	309
cc56d673 VW	310	/* sampling not supported */
	311	h_gpci_pmu.capabilities \|= PERF_PMU_CAP_NO_INTERRUPT;
	312
220a0c60 CS	313	r = perf_pmu_register(&h_gpci_pmu, h_gpci_pmu.name, -1);
	314	if (r)
	315	return r;
	316
	317	return 0;
	318	}
	319
	320	device_initcall(hv_gpci_init);