spinlock.c 11.6 KB
Newer Older
1 2 3 4 5 6
/*
 * Split spinlock implementation out into its own file, so it can be
 * compiled in a FTRACE-compatible way.
 */
#include <linux/kernel_stat.h>
#include <linux/spinlock.h>
J
Jeremy Fitzhardinge 已提交
7 8
#include <linux/debugfs.h>
#include <linux/log2.h>
9
#include <linux/gfp.h>
10 11 12 13 14 15 16

#include <asm/paravirt.h>

#include <xen/interface/xen.h>
#include <xen/events.h>

#include "xen-ops.h"
J
Jeremy Fitzhardinge 已提交
17 18 19 20 21 22 23 24 25 26
#include "debugfs.h"

#ifdef CONFIG_XEN_DEBUG_FS
static struct xen_spinlock_stats
{
	u64 taken;
	u32 taken_slow;
	u32 taken_slow_nested;
	u32 taken_slow_pickup;
	u32 taken_slow_spurious;
27
	u32 taken_slow_irqenable;
J
Jeremy Fitzhardinge 已提交
28 29 30 31 32

	u64 released;
	u32 released_slow;
	u32 released_slow_kicked;

33 34 35 36 37 38 39 40
#define HISTO_BUCKETS	30
	u32 histo_spin_total[HISTO_BUCKETS+1];
	u32 histo_spin_spinning[HISTO_BUCKETS+1];
	u32 histo_spin_blocked[HISTO_BUCKETS+1];

	u64 time_total;
	u64 time_spinning;
	u64 time_blocked;
J
Jeremy Fitzhardinge 已提交
41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75
} spinlock_stats;

static u8 zero_stats;

static unsigned lock_timeout = 1 << 10;
#define TIMEOUT lock_timeout

static inline void check_zero(void)
{
	if (unlikely(zero_stats)) {
		memset(&spinlock_stats, 0, sizeof(spinlock_stats));
		zero_stats = 0;
	}
}

#define ADD_STATS(elem, val)			\
	do { check_zero(); spinlock_stats.elem += (val); } while(0)

static inline u64 spin_time_start(void)
{
	return xen_clocksource_read();
}

static void __spin_time_accum(u64 delta, u32 *array)
{
	unsigned index = ilog2(delta);

	check_zero();

	if (index < HISTO_BUCKETS)
		array[index]++;
	else
		array[HISTO_BUCKETS]++;
}

76 77 78 79 80 81 82 83 84
static inline void spin_time_accum_spinning(u64 start)
{
	u32 delta = xen_clocksource_read() - start;

	__spin_time_accum(delta, spinlock_stats.histo_spin_spinning);
	spinlock_stats.time_spinning += delta;
}

static inline void spin_time_accum_total(u64 start)
J
Jeremy Fitzhardinge 已提交
85 86 87
{
	u32 delta = xen_clocksource_read() - start;

88 89
	__spin_time_accum(delta, spinlock_stats.histo_spin_total);
	spinlock_stats.time_total += delta;
J
Jeremy Fitzhardinge 已提交
90 91
}

92
static inline void spin_time_accum_blocked(u64 start)
J
Jeremy Fitzhardinge 已提交
93 94 95
{
	u32 delta = xen_clocksource_read() - start;

96 97
	__spin_time_accum(delta, spinlock_stats.histo_spin_blocked);
	spinlock_stats.time_blocked += delta;
J
Jeremy Fitzhardinge 已提交
98 99 100 101 102 103 104 105 106 107
}
#else  /* !CONFIG_XEN_DEBUG_FS */
#define TIMEOUT			(1 << 10)
#define ADD_STATS(elem, val)	do { (void)(val); } while(0)

static inline u64 spin_time_start(void)
{
	return 0;
}

108 109 110 111
static inline void spin_time_accum_total(u64 start)
{
}
static inline void spin_time_accum_spinning(u64 start)
J
Jeremy Fitzhardinge 已提交
112 113
{
}
114
static inline void spin_time_accum_blocked(u64 start)
J
Jeremy Fitzhardinge 已提交
115 116 117
{
}
#endif  /* CONFIG_XEN_DEBUG_FS */
118

119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135
/*
 * Size struct xen_spinlock so it's the same as arch_spinlock_t.
 */
#if NR_CPUS < 256
typedef u8 xen_spinners_t;
# define inc_spinners(xl) \
	asm(LOCK_PREFIX " incb %0" : "+m" ((xl)->spinners) : : "memory");
# define dec_spinners(xl) \
	asm(LOCK_PREFIX " decb %0" : "+m" ((xl)->spinners) : : "memory");
#else
typedef u16 xen_spinners_t;
# define inc_spinners(xl) \
	asm(LOCK_PREFIX " incw %0" : "+m" ((xl)->spinners) : : "memory");
# define dec_spinners(xl) \
	asm(LOCK_PREFIX " decw %0" : "+m" ((xl)->spinners) : : "memory");
#endif

136 137
struct xen_spinlock {
	unsigned char lock;		/* 0 -> free; 1 -> locked */
138
	xen_spinners_t spinners;	/* count of waiting cpus */
139 140
};

141
static int xen_spin_is_locked(struct arch_spinlock *lock)
142 143 144 145 146 147
{
	struct xen_spinlock *xl = (struct xen_spinlock *)lock;

	return xl->lock != 0;
}

148
static int xen_spin_is_contended(struct arch_spinlock *lock)
149 150 151 152 153 154 155 156
{
	struct xen_spinlock *xl = (struct xen_spinlock *)lock;

	/* Not strictly true; this is only the count of contended
	   lock-takers entering the slow path. */
	return xl->spinners != 0;
}

157
static int xen_spin_trylock(struct arch_spinlock *lock)
158 159 160 161 162 163 164 165 166 167 168 169 170
{
	struct xen_spinlock *xl = (struct xen_spinlock *)lock;
	u8 old = 1;

	asm("xchgb %b0,%1"
	    : "+q" (old), "+m" (xl->lock) : : "memory");

	return old == 0;
}

static DEFINE_PER_CPU(int, lock_kicker_irq) = -1;
static DEFINE_PER_CPU(struct xen_spinlock *, lock_spinners);

171 172 173 174 175
/*
 * Mark a cpu as interested in a lock.  Returns the CPU's previous
 * lock of interest, in case we got preempted by an interrupt.
 */
static inline struct xen_spinlock *spinning_lock(struct xen_spinlock *xl)
176
{
177 178
	struct xen_spinlock *prev;

C
Christoph Lameter 已提交
179 180
	prev = __this_cpu_read(lock_spinners);
	__this_cpu_write(lock_spinners, xl);
181

182
	wmb();			/* set lock of interest before count */
183

184
	inc_spinners(xl);
185 186

	return prev;
187 188
}

189 190 191 192 193
/*
 * Mark a cpu as no longer interested in a lock.  Restores previous
 * lock of interest (NULL for none).
 */
static inline void unspinning_lock(struct xen_spinlock *xl, struct xen_spinlock *prev)
194
{
195
	dec_spinners(xl);
196
	wmb();			/* decrement count before restoring lock */
C
Christoph Lameter 已提交
197
	__this_cpu_write(lock_spinners, prev);
198 199
}

200
static noinline int xen_spin_lock_slow(struct arch_spinlock *lock, bool irq_enable)
201 202
{
	struct xen_spinlock *xl = (struct xen_spinlock *)lock;
203
	struct xen_spinlock *prev;
C
Christoph Lameter 已提交
204
	int irq = __this_cpu_read(lock_kicker_irq);
205
	int ret;
206
	u64 start;
207 208 209 210 211

	/* If kicker interrupts not initialized yet, just spin */
	if (irq == -1)
		return 0;

212 213
	start = spin_time_start();

214
	/* announce we're spinning */
215
	prev = spinning_lock(xl);
216

J
Jeremy Fitzhardinge 已提交
217 218 219
	ADD_STATS(taken_slow, 1);
	ADD_STATS(taken_slow_nested, prev != NULL);

220
	do {
221 222
		unsigned long flags;

223 224 225 226 227 228 229
		/* clear pending */
		xen_clear_irq_pending(irq);

		/* check again make sure it didn't become free while
		   we weren't looking  */
		ret = xen_spin_trylock(lock);
		if (ret) {
J
Jeremy Fitzhardinge 已提交
230 231
			ADD_STATS(taken_slow_pickup, 1);

232 233 234 235 236 237 238 239 240
			/*
			 * If we interrupted another spinlock while it
			 * was blocking, make sure it doesn't block
			 * without rechecking the lock.
			 */
			if (prev != NULL)
				xen_set_irq_pending(irq);
			goto out;
		}
241

D
David Howells 已提交
242
		flags = arch_local_save_flags();
243 244 245 246 247
		if (irq_enable) {
			ADD_STATS(taken_slow_irqenable, 1);
			raw_local_irq_enable();
		}

248 249 250 251 252 253 254 255 256 257
		/*
		 * Block until irq becomes pending.  If we're
		 * interrupted at this point (after the trylock but
		 * before entering the block), then the nested lock
		 * handler guarantees that the irq will be left
		 * pending if there's any chance the lock became free;
		 * xen_poll_irq() returns immediately if the irq is
		 * pending.
		 */
		xen_poll_irq(irq);
258 259 260

		raw_local_irq_restore(flags);

J
Jeremy Fitzhardinge 已提交
261
		ADD_STATS(taken_slow_spurious, !xen_test_irq_pending(irq));
262
	} while (!xen_test_irq_pending(irq)); /* check for spurious wakeups */
263

T
Thomas Gleixner 已提交
264
	kstat_incr_irqs_this_cpu(irq, irq_to_desc(irq));
265 266

out:
267
	unspinning_lock(xl, prev);
268 269
	spin_time_accum_blocked(start);

270 271 272
	return ret;
}

273
static inline void __xen_spin_lock(struct arch_spinlock *lock, bool irq_enable)
274 275
{
	struct xen_spinlock *xl = (struct xen_spinlock *)lock;
J
Jeremy Fitzhardinge 已提交
276
	unsigned timeout;
277
	u8 oldval;
J
Jeremy Fitzhardinge 已提交
278 279 280 281 282
	u64 start_spin;

	ADD_STATS(taken, 1);

	start_spin = spin_time_start();
283 284

	do {
J
Jeremy Fitzhardinge 已提交
285 286 287
		u64 start_spin_fast = spin_time_start();

		timeout = TIMEOUT;
288 289 290 291 292 293 294 295 296 297 298 299 300 301

		asm("1: xchgb %1,%0\n"
		    "   testb %1,%1\n"
		    "   jz 3f\n"
		    "2: rep;nop\n"
		    "   cmpb $0,%0\n"
		    "   je 1b\n"
		    "   dec %2\n"
		    "   jnz 2b\n"
		    "3:\n"
		    : "+m" (xl->lock), "=q" (oldval), "+r" (timeout)
		    : "1" (1)
		    : "memory");

302
		spin_time_accum_spinning(start_spin_fast);
303 304 305

	} while (unlikely(oldval != 0 &&
			  (TIMEOUT == ~0 || !xen_spin_lock_slow(lock, irq_enable))));
J
Jeremy Fitzhardinge 已提交
306

307
	spin_time_accum_total(start_spin);
308 309
}

310
static void xen_spin_lock(struct arch_spinlock *lock)
311 312 313 314
{
	__xen_spin_lock(lock, false);
}

315
static void xen_spin_lock_flags(struct arch_spinlock *lock, unsigned long flags)
316 317 318 319
{
	__xen_spin_lock(lock, !raw_irqs_disabled_flags(flags));
}

320 321 322 323
static noinline void xen_spin_unlock_slow(struct xen_spinlock *xl)
{
	int cpu;

J
Jeremy Fitzhardinge 已提交
324 325
	ADD_STATS(released_slow, 1);

326 327 328
	for_each_online_cpu(cpu) {
		/* XXX should mix up next cpu selection */
		if (per_cpu(lock_spinners, cpu) == xl) {
J
Jeremy Fitzhardinge 已提交
329
			ADD_STATS(released_slow_kicked, 1);
330 331 332 333 334
			xen_send_IPI_one(cpu, XEN_SPIN_UNLOCK_VECTOR);
		}
	}
}

335
static void xen_spin_unlock(struct arch_spinlock *lock)
336 337 338
{
	struct xen_spinlock *xl = (struct xen_spinlock *)lock;

J
Jeremy Fitzhardinge 已提交
339 340
	ADD_STATS(released, 1);

341 342 343
	smp_wmb();		/* make sure no writes get moved after unlock */
	xl->lock = 0;		/* release lock */

344 345 346 347 348 349 350
	/*
	 * Make sure unlock happens before checking for waiting
	 * spinners.  We need a strong barrier to enforce the
	 * write-read ordering to different memory locations, as the
	 * CPU makes no implied guarantees about their ordering.
	 */
	mb();
351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366

	if (unlikely(xl->spinners))
		xen_spin_unlock_slow(xl);
}

static irqreturn_t dummy_handler(int irq, void *dev_id)
{
	BUG();
	return IRQ_HANDLED;
}

void __cpuinit xen_init_lock_cpu(int cpu)
{
	int irq;
	const char *name;

367
	WARN(per_cpu(lock_kicker_irq, cpu) >= 0, "spinlock on CPU%d exists on IRQ%d!\n",
368 369
	     cpu, per_cpu(lock_kicker_irq, cpu));

370 371 372 373 374 375 376
	/*
	 * See git commit f10cd522c5fbfec9ae3cc01967868c9c2401ed23
	 * (xen: disable PV spinlocks on HVM)
	 */
	if (xen_hvm_domain())
		return;

377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392
	name = kasprintf(GFP_KERNEL, "spinlock%d", cpu);
	irq = bind_ipi_to_irqhandler(XEN_SPIN_UNLOCK_VECTOR,
				     cpu,
				     dummy_handler,
				     IRQF_DISABLED|IRQF_PERCPU|IRQF_NOBALANCING,
				     name,
				     NULL);

	if (irq >= 0) {
		disable_irq(irq); /* make sure it's never delivered */
		per_cpu(lock_kicker_irq, cpu) = irq;
	}

	printk("cpu %d spinlock event irq %d\n", cpu, irq);
}

A
Alex Nixon 已提交
393 394
void xen_uninit_lock_cpu(int cpu)
{
395 396 397 398 399 400 401
	/*
	 * See git commit f10cd522c5fbfec9ae3cc01967868c9c2401ed23
	 * (xen: disable PV spinlocks on HVM)
	 */
	if (xen_hvm_domain())
		return;

A
Alex Nixon 已提交
402
	unbind_from_irqhandler(per_cpu(lock_kicker_irq, cpu), NULL);
403
	per_cpu(lock_kicker_irq, cpu) = -1;
A
Alex Nixon 已提交
404 405
}

406 407
void __init xen_init_spinlocks(void)
{
408 409 410 411 412 413 414
	/*
	 * See git commit f10cd522c5fbfec9ae3cc01967868c9c2401ed23
	 * (xen: disable PV spinlocks on HVM)
	 */
	if (xen_hvm_domain())
		return;

415 416
	BUILD_BUG_ON(sizeof(struct xen_spinlock) > sizeof(arch_spinlock_t));

417 418 419
	pv_lock_ops.spin_is_locked = xen_spin_is_locked;
	pv_lock_ops.spin_is_contended = xen_spin_is_contended;
	pv_lock_ops.spin_lock = xen_spin_lock;
420
	pv_lock_ops.spin_lock_flags = xen_spin_lock_flags;
421 422 423
	pv_lock_ops.spin_trylock = xen_spin_trylock;
	pv_lock_ops.spin_unlock = xen_spin_unlock;
}
J
Jeremy Fitzhardinge 已提交
424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450

#ifdef CONFIG_XEN_DEBUG_FS

static struct dentry *d_spin_debug;

static int __init xen_spinlock_debugfs(void)
{
	struct dentry *d_xen = xen_init_debugfs();

	if (d_xen == NULL)
		return -ENOMEM;

	d_spin_debug = debugfs_create_dir("spinlocks", d_xen);

	debugfs_create_u8("zero_stats", 0644, d_spin_debug, &zero_stats);

	debugfs_create_u32("timeout", 0644, d_spin_debug, &lock_timeout);

	debugfs_create_u64("taken", 0444, d_spin_debug, &spinlock_stats.taken);
	debugfs_create_u32("taken_slow", 0444, d_spin_debug,
			   &spinlock_stats.taken_slow);
	debugfs_create_u32("taken_slow_nested", 0444, d_spin_debug,
			   &spinlock_stats.taken_slow_nested);
	debugfs_create_u32("taken_slow_pickup", 0444, d_spin_debug,
			   &spinlock_stats.taken_slow_pickup);
	debugfs_create_u32("taken_slow_spurious", 0444, d_spin_debug,
			   &spinlock_stats.taken_slow_spurious);
451 452
	debugfs_create_u32("taken_slow_irqenable", 0444, d_spin_debug,
			   &spinlock_stats.taken_slow_irqenable);
J
Jeremy Fitzhardinge 已提交
453 454 455 456 457 458 459 460

	debugfs_create_u64("released", 0444, d_spin_debug, &spinlock_stats.released);
	debugfs_create_u32("released_slow", 0444, d_spin_debug,
			   &spinlock_stats.released_slow);
	debugfs_create_u32("released_slow_kicked", 0444, d_spin_debug,
			   &spinlock_stats.released_slow_kicked);

	debugfs_create_u64("time_spinning", 0444, d_spin_debug,
461 462 463
			   &spinlock_stats.time_spinning);
	debugfs_create_u64("time_blocked", 0444, d_spin_debug,
			   &spinlock_stats.time_blocked);
J
Jeremy Fitzhardinge 已提交
464
	debugfs_create_u64("time_total", 0444, d_spin_debug,
465
			   &spinlock_stats.time_total);
J
Jeremy Fitzhardinge 已提交
466

467 468 469 470 471 472
	debugfs_create_u32_array("histo_total", 0444, d_spin_debug,
				spinlock_stats.histo_spin_total, HISTO_BUCKETS + 1);
	debugfs_create_u32_array("histo_spinning", 0444, d_spin_debug,
				spinlock_stats.histo_spin_spinning, HISTO_BUCKETS + 1);
	debugfs_create_u32_array("histo_blocked", 0444, d_spin_debug,
				spinlock_stats.histo_spin_blocked, HISTO_BUCKETS + 1);
J
Jeremy Fitzhardinge 已提交
473 474 475 476 477 478

	return 0;
}
fs_initcall(xen_spinlock_debugfs);

#endif	/* CONFIG_XEN_DEBUG_FS */