spinlock.c 10.5 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 10 11 12 13 14 15

#include <asm/paravirt.h>

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

#include "xen-ops.h"
J
Jeremy Fitzhardinge 已提交
16 17 18 19 20 21 22 23 24 25
#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;
26
	u32 taken_slow_irqenable;
J
Jeremy Fitzhardinge 已提交
27 28 29 30 31

	u64 released;
	u32 released_slow;
	u32 released_slow_kicked;

32 33 34 35 36 37 38 39
#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
40 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
} 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]++;
}

75 76 77 78 79 80 81 82 83
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
84 85 86
{
	u32 delta = xen_clocksource_read() - start;

87 88
	__spin_time_accum(delta, spinlock_stats.histo_spin_total);
	spinlock_stats.time_total += delta;
J
Jeremy Fitzhardinge 已提交
89 90
}

91
static inline void spin_time_accum_blocked(u64 start)
J
Jeremy Fitzhardinge 已提交
92 93 94
{
	u32 delta = xen_clocksource_read() - start;

95 96
	__spin_time_accum(delta, spinlock_stats.histo_spin_blocked);
	spinlock_stats.time_blocked += delta;
J
Jeremy Fitzhardinge 已提交
97 98 99 100 101 102 103 104 105 106
}
#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;
}

107 108 109 110
static inline void spin_time_accum_total(u64 start)
{
}
static inline void spin_time_accum_spinning(u64 start)
J
Jeremy Fitzhardinge 已提交
111 112
{
}
113
static inline void spin_time_accum_blocked(u64 start)
J
Jeremy Fitzhardinge 已提交
114 115 116
{
}
#endif  /* CONFIG_XEN_DEBUG_FS */
117 118 119 120 121 122

struct xen_spinlock {
	unsigned char lock;		/* 0 -> free; 1 -> locked */
	unsigned short spinners;	/* count of waiting cpus */
};

123
static int xen_spin_is_locked(struct arch_spinlock *lock)
124 125 126 127 128 129
{
	struct xen_spinlock *xl = (struct xen_spinlock *)lock;

	return xl->lock != 0;
}

130
static int xen_spin_is_contended(struct arch_spinlock *lock)
131 132 133 134 135 136 137 138
{
	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;
}

139
static int xen_spin_trylock(struct arch_spinlock *lock)
140 141 142 143 144 145 146 147 148 149 150 151 152
{
	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);

153 154 155 156 157
/*
 * 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)
158
{
159 160 161
	struct xen_spinlock *prev;

	prev = __get_cpu_var(lock_spinners);
162
	__get_cpu_var(lock_spinners) = xl;
163

164
	wmb();			/* set lock of interest before count */
165

166 167
	asm(LOCK_PREFIX " incw %0"
	    : "+m" (xl->spinners) : : "memory");
168 169

	return prev;
170 171
}

172 173 174 175 176
/*
 * 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)
177 178 179
{
	asm(LOCK_PREFIX " decw %0"
	    : "+m" (xl->spinners) : : "memory");
180 181
	wmb();			/* decrement count before restoring lock */
	__get_cpu_var(lock_spinners) = prev;
182 183
}

184
static noinline int xen_spin_lock_slow(struct arch_spinlock *lock, bool irq_enable)
185 186
{
	struct xen_spinlock *xl = (struct xen_spinlock *)lock;
187
	struct xen_spinlock *prev;
188 189
	int irq = __get_cpu_var(lock_kicker_irq);
	int ret;
190
	u64 start;
191 192 193 194 195

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

196 197
	start = spin_time_start();

198
	/* announce we're spinning */
199
	prev = spinning_lock(xl);
200

J
Jeremy Fitzhardinge 已提交
201 202 203
	ADD_STATS(taken_slow, 1);
	ADD_STATS(taken_slow_nested, prev != NULL);

204
	do {
205 206
		unsigned long flags;

207 208 209 210 211 212 213
		/* 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
214 215
			ADD_STATS(taken_slow_pickup, 1);

216 217 218 219 220 221 222 223 224
			/*
			 * 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;
		}
225

226 227 228 229 230 231
		flags = __raw_local_save_flags();
		if (irq_enable) {
			ADD_STATS(taken_slow_irqenable, 1);
			raw_local_irq_enable();
		}

232 233 234 235 236 237 238 239 240 241
		/*
		 * 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);
242 243 244

		raw_local_irq_restore(flags);

J
Jeremy Fitzhardinge 已提交
245
		ADD_STATS(taken_slow_spurious, !xen_test_irq_pending(irq));
246
	} while (!xen_test_irq_pending(irq)); /* check for spurious wakeups */
247

T
Thomas Gleixner 已提交
248
	kstat_incr_irqs_this_cpu(irq, irq_to_desc(irq));
249 250

out:
251
	unspinning_lock(xl, prev);
252 253
	spin_time_accum_blocked(start);

254 255 256
	return ret;
}

257
static inline void __xen_spin_lock(struct arch_spinlock *lock, bool irq_enable)
258 259
{
	struct xen_spinlock *xl = (struct xen_spinlock *)lock;
J
Jeremy Fitzhardinge 已提交
260
	unsigned timeout;
261
	u8 oldval;
J
Jeremy Fitzhardinge 已提交
262 263 264 265 266
	u64 start_spin;

	ADD_STATS(taken, 1);

	start_spin = spin_time_start();
267 268

	do {
J
Jeremy Fitzhardinge 已提交
269 270 271
		u64 start_spin_fast = spin_time_start();

		timeout = TIMEOUT;
272 273 274 275 276 277 278 279 280 281 282 283 284 285

		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");

286
		spin_time_accum_spinning(start_spin_fast);
287 288 289

	} while (unlikely(oldval != 0 &&
			  (TIMEOUT == ~0 || !xen_spin_lock_slow(lock, irq_enable))));
J
Jeremy Fitzhardinge 已提交
290

291
	spin_time_accum_total(start_spin);
292 293
}

294
static void xen_spin_lock(struct arch_spinlock *lock)
295 296 297 298
{
	__xen_spin_lock(lock, false);
}

299
static void xen_spin_lock_flags(struct arch_spinlock *lock, unsigned long flags)
300 301 302 303
{
	__xen_spin_lock(lock, !raw_irqs_disabled_flags(flags));
}

304 305 306 307
static noinline void xen_spin_unlock_slow(struct xen_spinlock *xl)
{
	int cpu;

J
Jeremy Fitzhardinge 已提交
308 309
	ADD_STATS(released_slow, 1);

310 311 312
	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
313
			ADD_STATS(released_slow_kicked, 1);
314 315 316 317 318 319
			xen_send_IPI_one(cpu, XEN_SPIN_UNLOCK_VECTOR);
			break;
		}
	}
}

320
static void xen_spin_unlock(struct arch_spinlock *lock)
321 322 323
{
	struct xen_spinlock *xl = (struct xen_spinlock *)lock;

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

326 327 328
	smp_wmb();		/* make sure no writes get moved after unlock */
	xl->lock = 0;		/* release lock */

329 330 331 332 333 334 335
	/*
	 * 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();
336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367

	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;

	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
368 369 370 371 372
void xen_uninit_lock_cpu(int cpu)
{
	unbind_from_irqhandler(per_cpu(lock_kicker_irq, cpu), NULL);
}

373 374 375 376 377
void __init xen_init_spinlocks(void)
{
	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;
378
	pv_lock_ops.spin_lock_flags = xen_spin_lock_flags;
379 380 381
	pv_lock_ops.spin_trylock = xen_spin_trylock;
	pv_lock_ops.spin_unlock = xen_spin_unlock;
}
J
Jeremy Fitzhardinge 已提交
382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408

#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);
409 410
	debugfs_create_u32("taken_slow_irqenable", 0444, d_spin_debug,
			   &spinlock_stats.taken_slow_irqenable);
J
Jeremy Fitzhardinge 已提交
411 412 413 414 415 416 417 418

	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,
419 420 421
			   &spinlock_stats.time_spinning);
	debugfs_create_u64("time_blocked", 0444, d_spin_debug,
			   &spinlock_stats.time_blocked);
J
Jeremy Fitzhardinge 已提交
422
	debugfs_create_u64("time_total", 0444, d_spin_debug,
423
			   &spinlock_stats.time_total);
J
Jeremy Fitzhardinge 已提交
424 425

	xen_debugfs_create_u32_array("histo_total", 0444, d_spin_debug,
426
				     spinlock_stats.histo_spin_total, HISTO_BUCKETS + 1);
J
Jeremy Fitzhardinge 已提交
427
	xen_debugfs_create_u32_array("histo_spinning", 0444, d_spin_debug,
428 429 430
				     spinlock_stats.histo_spin_spinning, HISTO_BUCKETS + 1);
	xen_debugfs_create_u32_array("histo_blocked", 0444, d_spin_debug,
				     spinlock_stats.histo_spin_blocked, HISTO_BUCKETS + 1);
J
Jeremy Fitzhardinge 已提交
431 432 433 434 435 436

	return 0;
}
fs_initcall(xen_spinlock_debugfs);

#endif	/* CONFIG_XEN_DEBUG_FS */