nmi_int.c 9.9 KB
Newer Older
L
Linus Torvalds 已提交
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
/**
 * @file nmi_int.c
 *
 * @remark Copyright 2002 OProfile authors
 * @remark Read the file COPYING
 *
 * @author John Levon <levon@movementarian.org>
 */

#include <linux/init.h>
#include <linux/notifier.h>
#include <linux/smp.h>
#include <linux/oprofile.h>
#include <linux/sysdev.h>
#include <linux/slab.h>
16
#include <linux/moduleparam.h>
17
#include <linux/kdebug.h>
L
Linus Torvalds 已提交
18 19 20 21 22 23
#include <asm/nmi.h>
#include <asm/msr.h>
#include <asm/apic.h>
 
#include "op_counter.h"
#include "op_x86_model.h"
24

L
Linus Torvalds 已提交
25 26 27
static struct op_x86_model_spec const * model;
static struct op_msrs cpu_msrs[NR_CPUS];
static unsigned long saved_lvtpc[NR_CPUS];
28

L
Linus Torvalds 已提交
29 30 31 32 33 34 35 36
static int nmi_start(void);
static void nmi_stop(void);

/* 0 == registered but off, 1 == registered and on */
static int nmi_enabled = 0;

#ifdef CONFIG_PM

37
static int nmi_suspend(struct sys_device *dev, pm_message_t state)
L
Linus Torvalds 已提交
38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53
{
	if (nmi_enabled == 1)
		nmi_stop();
	return 0;
}


static int nmi_resume(struct sys_device *dev)
{
	if (nmi_enabled == 1)
		nmi_start();
	return 0;
}


static struct sysdev_class oprofile_sysclass = {
54
	.name		= "oprofile",
L
Linus Torvalds 已提交
55 56 57 58 59 60 61 62 63 64 65
	.resume		= nmi_resume,
	.suspend	= nmi_suspend,
};


static struct sys_device device_oprofile = {
	.id	= 0,
	.cls	= &oprofile_sysclass,
};


66
static int __init init_sysfs(void)
L
Linus Torvalds 已提交
67 68 69 70 71 72 73 74
{
	int error;
	if (!(error = sysdev_class_register(&oprofile_sysclass)))
		error = sysdev_register(&device_oprofile);
	return error;
}


75
static void exit_sysfs(void)
L
Linus Torvalds 已提交
76 77 78 79 80 81
{
	sysdev_unregister(&device_oprofile);
	sysdev_class_unregister(&oprofile_sysclass);
}

#else
82 83
#define init_sysfs() do { } while (0)
#define exit_sysfs() do { } while (0)
L
Linus Torvalds 已提交
84 85
#endif /* CONFIG_PM */

86 87
static int profile_exceptions_notify(struct notifier_block *self,
				     unsigned long val, void *data)
L
Linus Torvalds 已提交
88
{
89 90 91 92 93 94 95 96 97 98 99 100 101
	struct die_args *args = (struct die_args *)data;
	int ret = NOTIFY_DONE;
	int cpu = smp_processor_id();

	switch(val) {
	case DIE_NMI:
		if (model->check_ctrs(args->regs, &cpu_msrs[cpu]))
			ret = NOTIFY_STOP;
		break;
	default:
		break;
	}
	return ret;
L
Linus Torvalds 已提交
102
}
103

L
Linus Torvalds 已提交
104 105 106 107 108 109 110 111 112
static void nmi_cpu_save_registers(struct op_msrs * msrs)
{
	unsigned int const nr_ctrs = model->num_counters;
	unsigned int const nr_ctrls = model->num_controls; 
	struct op_msr * counters = msrs->counters;
	struct op_msr * controls = msrs->controls;
	unsigned int i;

	for (i = 0; i < nr_ctrs; ++i) {
113 114 115 116 117
		if (counters[i].addr){
			rdmsr(counters[i].addr,
				counters[i].saved.low,
				counters[i].saved.high);
		}
L
Linus Torvalds 已提交
118 119 120
	}
 
	for (i = 0; i < nr_ctrls; ++i) {
121 122 123 124 125
		if (controls[i].addr){
			rdmsr(controls[i].addr,
				controls[i].saved.low,
				controls[i].saved.high);
		}
L
Linus Torvalds 已提交
126 127 128 129 130 131 132 133 134 135 136 137 138 139 140
	}
}


static void nmi_save_registers(void * dummy)
{
	int cpu = smp_processor_id();
	struct op_msrs * msrs = &cpu_msrs[cpu];
	nmi_cpu_save_registers(msrs);
}


static void free_msrs(void)
{
	int i;
141
	for_each_possible_cpu(i) {
L
Linus Torvalds 已提交
142 143 144 145 146 147 148 149 150 151 152 153 154 155 156
		kfree(cpu_msrs[i].counters);
		cpu_msrs[i].counters = NULL;
		kfree(cpu_msrs[i].controls);
		cpu_msrs[i].controls = NULL;
	}
}


static int allocate_msrs(void)
{
	int success = 1;
	size_t controls_size = sizeof(struct op_msr) * model->num_controls;
	size_t counters_size = sizeof(struct op_msr) * model->num_counters;

	int i;
C
Chris Wright 已提交
157
	for_each_possible_cpu(i) {
L
Linus Torvalds 已提交
158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187
		cpu_msrs[i].counters = kmalloc(counters_size, GFP_KERNEL);
		if (!cpu_msrs[i].counters) {
			success = 0;
			break;
		}
		cpu_msrs[i].controls = kmalloc(controls_size, GFP_KERNEL);
		if (!cpu_msrs[i].controls) {
			success = 0;
			break;
		}
	}

	if (!success)
		free_msrs();

	return success;
}


static void nmi_cpu_setup(void * dummy)
{
	int cpu = smp_processor_id();
	struct op_msrs * msrs = &cpu_msrs[cpu];
	spin_lock(&oprofilefs_lock);
	model->setup_ctrs(msrs);
	spin_unlock(&oprofilefs_lock);
	saved_lvtpc[cpu] = apic_read(APIC_LVTPC);
	apic_write(APIC_LVTPC, APIC_DM_NMI);
}

188 189 190 191 192
static struct notifier_block profile_exceptions_nb = {
	.notifier_call = profile_exceptions_notify,
	.next = NULL,
	.priority = 0
};
L
Linus Torvalds 已提交
193 194 195

static int nmi_setup(void)
{
196
	int err=0;
197
	int cpu;
198

L
Linus Torvalds 已提交
199 200 201
	if (!allocate_msrs())
		return -ENOMEM;

202
	if ((err = register_die_notifier(&profile_exceptions_nb))){
L
Linus Torvalds 已提交
203
		free_msrs();
204
		return err;
L
Linus Torvalds 已提交
205
	}
206

L
Linus Torvalds 已提交
207 208 209
	/* We need to serialize save and setup for HT because the subset
	 * of msrs are distinct for save and setup operations
	 */
210 211 212 213

	/* Assume saved/restored counters are the same on all CPUs */
	model->fill_in_addresses(&cpu_msrs[0]);
	for_each_possible_cpu (cpu) {
C
Chris Wright 已提交
214 215 216 217 218 219 220 221
		if (cpu != 0) {
			memcpy(cpu_msrs[cpu].counters, cpu_msrs[0].counters,
				sizeof(struct op_msr) * model->num_counters);

			memcpy(cpu_msrs[cpu].controls, cpu_msrs[0].controls,
				sizeof(struct op_msr) * model->num_controls);
		}

222
	}
L
Linus Torvalds 已提交
223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238
	on_each_cpu(nmi_save_registers, NULL, 0, 1);
	on_each_cpu(nmi_cpu_setup, NULL, 0, 1);
	nmi_enabled = 1;
	return 0;
}


static void nmi_restore_registers(struct op_msrs * msrs)
{
	unsigned int const nr_ctrs = model->num_counters;
	unsigned int const nr_ctrls = model->num_controls; 
	struct op_msr * counters = msrs->counters;
	struct op_msr * controls = msrs->controls;
	unsigned int i;

	for (i = 0; i < nr_ctrls; ++i) {
239 240 241 242 243
		if (controls[i].addr){
			wrmsr(controls[i].addr,
				controls[i].saved.low,
				controls[i].saved.high);
		}
L
Linus Torvalds 已提交
244 245 246
	}
 
	for (i = 0; i < nr_ctrs; ++i) {
247 248 249 250 251
		if (counters[i].addr){
			wrmsr(counters[i].addr,
				counters[i].saved.low,
				counters[i].saved.high);
		}
L
Linus Torvalds 已提交
252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278
	}
}
 

static void nmi_cpu_shutdown(void * dummy)
{
	unsigned int v;
	int cpu = smp_processor_id();
	struct op_msrs * msrs = &cpu_msrs[cpu];
 
	/* restoring APIC_LVTPC can trigger an apic error because the delivery
	 * mode and vector nr combination can be illegal. That's by design: on
	 * power on apic lvt contain a zero vector nr which are legal only for
	 * NMI delivery mode. So inhibit apic err before restoring lvtpc
	 */
	v = apic_read(APIC_LVTERR);
	apic_write(APIC_LVTERR, v | APIC_LVT_MASKED);
	apic_write(APIC_LVTPC, saved_lvtpc[cpu]);
	apic_write(APIC_LVTERR, v);
	nmi_restore_registers(msrs);
}

 
static void nmi_shutdown(void)
{
	nmi_enabled = 0;
	on_each_cpu(nmi_cpu_shutdown, NULL, 0, 1);
279
	unregister_die_notifier(&profile_exceptions_nb);
280
	model->shutdown(cpu_msrs);
L
Linus Torvalds 已提交
281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319
	free_msrs();
}

 
static void nmi_cpu_start(void * dummy)
{
	struct op_msrs const * msrs = &cpu_msrs[smp_processor_id()];
	model->start(msrs);
}
 

static int nmi_start(void)
{
	on_each_cpu(nmi_cpu_start, NULL, 0, 1);
	return 0;
}
 
 
static void nmi_cpu_stop(void * dummy)
{
	struct op_msrs const * msrs = &cpu_msrs[smp_processor_id()];
	model->stop(msrs);
}
 
 
static void nmi_stop(void)
{
	on_each_cpu(nmi_cpu_stop, NULL, 0, 1);
}


struct op_counter_config counter_config[OP_MAX_COUNTER];

static int nmi_create_files(struct super_block * sb, struct dentry * root)
{
	unsigned int i;

	for (i = 0; i < model->num_counters; ++i) {
		struct dentry * dir;
320
		char buf[4];
L
Linus Torvalds 已提交
321
 
322 323 324 325 326 327 328 329
 		/* quick little hack to _not_ expose a counter if it is not
		 * available for use.  This should protect userspace app.
		 * NOTE:  assumes 1:1 mapping here (that counters are organized
		 *        sequentially in their struct assignment).
		 */
		if (unlikely(!avail_to_resrv_perfctr_nmi_bit(i)))
			continue;

330
		snprintf(buf,  sizeof(buf), "%d", i);
L
Linus Torvalds 已提交
331 332 333 334 335 336 337 338 339 340 341 342
		dir = oprofilefs_mkdir(sb, root, buf);
		oprofilefs_create_ulong(sb, dir, "enabled", &counter_config[i].enabled); 
		oprofilefs_create_ulong(sb, dir, "event", &counter_config[i].event); 
		oprofilefs_create_ulong(sb, dir, "count", &counter_config[i].count); 
		oprofilefs_create_ulong(sb, dir, "unit_mask", &counter_config[i].unit_mask); 
		oprofilefs_create_ulong(sb, dir, "kernel", &counter_config[i].kernel); 
		oprofilefs_create_ulong(sb, dir, "user", &counter_config[i].user); 
	}

	return 0;
}
 
343 344
static int p4force;
module_param(p4force, int, 0);
L
Linus Torvalds 已提交
345 346 347 348 349
 
static int __init p4_init(char ** cpu_type)
{
	__u8 cpu_model = boot_cpu_data.x86_model;

350
	if (!p4force && (cpu_model > 6 || cpu_model == 5))
L
Linus Torvalds 已提交
351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380
		return 0;

#ifndef CONFIG_SMP
	*cpu_type = "i386/p4";
	model = &op_p4_spec;
	return 1;
#else
	switch (smp_num_siblings) {
		case 1:
			*cpu_type = "i386/p4";
			model = &op_p4_spec;
			return 1;

		case 2:
			*cpu_type = "i386/p4-ht";
			model = &op_p4_ht2_spec;
			return 1;
	}
#endif

	printk(KERN_INFO "oprofile: P4 HyperThreading detected with > 2 threads\n");
	printk(KERN_INFO "oprofile: Reverting to timer mode.\n");
	return 0;
}


static int __init ppro_init(char ** cpu_type)
{
	__u8 cpu_model = boot_cpu_data.x86_model;

381 382
	if (cpu_model == 14)
		*cpu_type = "i386/core";
383
	else if (cpu_model == 15 || cpu_model == 23)
384
		*cpu_type = "i386/core_2";
385
	else if (cpu_model > 0xd)
L
Linus Torvalds 已提交
386
		return 0;
387
	else if (cpu_model == 9) {
L
Linus Torvalds 已提交
388 389 390 391 392 393 394 395 396 397 398 399 400
		*cpu_type = "i386/p6_mobile";
	} else if (cpu_model > 5) {
		*cpu_type = "i386/piii";
	} else if (cpu_model > 2) {
		*cpu_type = "i386/pii";
	} else {
		*cpu_type = "i386/ppro";
	}

	model = &op_ppro_spec;
	return 1;
}

401
/* in order to get sysfs right */
L
Linus Torvalds 已提交
402 403
static int using_nmi;

404
int __init op_nmi_init(struct oprofile_operations *ops)
L
Linus Torvalds 已提交
405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429
{
	__u8 vendor = boot_cpu_data.x86_vendor;
	__u8 family = boot_cpu_data.x86;
	char *cpu_type;

	if (!cpu_has_apic)
		return -ENODEV;
 
	switch (vendor) {
		case X86_VENDOR_AMD:
			/* Needs to be at least an Athlon (or hammer in 32bit mode) */

			switch (family) {
			default:
				return -ENODEV;
			case 6:
				model = &op_athlon_spec;
				cpu_type = "i386/athlon";
				break;
			case 0xf:
				model = &op_athlon_spec;
				/* Actually it could be i386/hammer too, but give
				   user space an consistent name. */
				cpu_type = "x86-64/hammer";
				break;
430 431 432 433
			case 0x10:
				model = &op_athlon_spec;
				cpu_type = "x86-64/family10";
				break;
L
Linus Torvalds 已提交
434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459
			}
			break;
 
		case X86_VENDOR_INTEL:
			switch (family) {
				/* Pentium IV */
				case 0xf:
					if (!p4_init(&cpu_type))
						return -ENODEV;
					break;

				/* A P6-class processor */
				case 6:
					if (!ppro_init(&cpu_type))
						return -ENODEV;
					break;

				default:
					return -ENODEV;
			}
			break;

		default:
			return -ENODEV;
	}

460
	init_sysfs();
L
Linus Torvalds 已提交
461 462 463 464 465 466 467 468 469 470 471 472
	using_nmi = 1;
	ops->create_files = nmi_create_files;
	ops->setup = nmi_setup;
	ops->shutdown = nmi_shutdown;
	ops->start = nmi_start;
	ops->stop = nmi_stop;
	ops->cpu_type = cpu_type;
	printk(KERN_INFO "oprofile: using NMI interrupt.\n");
	return 0;
}


473
void op_nmi_exit(void)
L
Linus Torvalds 已提交
474 475
{
	if (using_nmi)
476
		exit_sysfs();
L
Linus Torvalds 已提交
477
}