summaryrefslogtreecommitdiff
path: root/sys/kern/subr_percpu.c
blob: 04966595b8658817fa86ddab78e6d24c5b970ca1 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
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
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
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
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
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
279
280
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
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
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
368
369
370
371
/*	$NetBSD: subr_percpu.c,v 1.16 2012/01/27 19:48:40 para Exp $	*/

/*-
 * Copyright (c)2007,2008 YAMAMOTO Takashi,
 * All rights reserved.
 *
 * Redistribution and use in source and binary forms, with or without
 * modification, are permitted provided that the following conditions
 * are met:
 * 1. Redistributions of source code must retain the above copyright
 *    notice, this list of conditions and the following disclaimer.
 * 2. Redistributions in binary form must reproduce the above copyright
 *    notice, this list of conditions and the following disclaimer in the
 *    documentation and/or other materials provided with the distribution.
 *
 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
 * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
 * SUCH DAMAGE.
 */

/*
 * per-cpu storage.
 */

#include <sys/cdefs.h>
__KERNEL_RCSID(0, "$NetBSD: subr_percpu.c,v 1.16 2012/01/27 19:48:40 para Exp $");

#include <sys/param.h>
#include <sys/cpu.h>
#include <sys/kmem.h>
#include <sys/kernel.h>
#include <sys/mutex.h>
#include <sys/percpu.h>
#include <sys/rwlock.h>
#include <sys/vmem.h>
#include <sys/xcall.h>

#define	PERCPU_QUANTUM_SIZE	(ALIGNBYTES + 1)
#define	PERCPU_QCACHE_MAX	0
#define	PERCPU_IMPORT_SIZE	2048

#if defined(DIAGNOSTIC)
#define	MAGIC	0x50435055	/* "PCPU" */
#define	percpu_encrypt(pc)	((pc) ^ MAGIC)
#define	percpu_decrypt(pc)	((pc) ^ MAGIC)
#else /* defined(DIAGNOSTIC) */
#define	percpu_encrypt(pc)	(pc)
#define	percpu_decrypt(pc)	(pc)
#endif /* defined(DIAGNOSTIC) */

static krwlock_t	percpu_swap_lock	__cacheline_aligned;
static kmutex_t		percpu_allocation_lock	__cacheline_aligned;
static vmem_t *		percpu_offset_arena	__cacheline_aligned;
static unsigned int	percpu_nextoff		__cacheline_aligned;

static percpu_cpu_t *
cpu_percpu(struct cpu_info *ci)
{

	return &ci->ci_data.cpu_percpu;
}

static unsigned int
percpu_offset(percpu_t *pc)
{
	const unsigned int off = percpu_decrypt((uintptr_t)pc);

	KASSERT(off < percpu_nextoff);
	return off;
}

/*
 * percpu_cpu_swap: crosscall handler for percpu_cpu_enlarge
 */

static void
percpu_cpu_swap(void *p1, void *p2)
{
	struct cpu_info * const ci = p1;
	percpu_cpu_t * const newpcc = p2;
	percpu_cpu_t * const pcc = cpu_percpu(ci);

	KASSERT(ci == curcpu() || !mp_online);

	/*
	 * swap *pcc and *newpcc unless anyone has beaten us.
	 */
	rw_enter(&percpu_swap_lock, RW_WRITER);
	if (newpcc->pcc_size > pcc->pcc_size) {
		percpu_cpu_t tmp;
		int s;

		tmp = *pcc;

		/*
		 * block interrupts so that we don't lose their modifications.
		 */

		s = splhigh();

		/*
		 * copy data to new storage.
		 */

		memcpy(newpcc->pcc_data, pcc->pcc_data, pcc->pcc_size);

		/*
		 * this assignment needs to be atomic for percpu_getptr_remote.
		 */

		pcc->pcc_data = newpcc->pcc_data;

		splx(s);

		pcc->pcc_size = newpcc->pcc_size;
		*newpcc = tmp;
	}
	rw_exit(&percpu_swap_lock);
}

/*
 * percpu_cpu_enlarge: ensure that percpu_cpu_t of each cpus have enough space
 */

static void
percpu_cpu_enlarge(size_t size)
{
	CPU_INFO_ITERATOR cii;
	struct cpu_info *ci;

	for (CPU_INFO_FOREACH(cii, ci)) {
		percpu_cpu_t pcc;

		pcc.pcc_data = kmem_alloc(size, KM_SLEEP); /* XXX cacheline */
		pcc.pcc_size = size;
		if (!mp_online) {
			percpu_cpu_swap(ci, &pcc);
		} else {
			uint64_t where;

			where = xc_unicast(0, percpu_cpu_swap, ci, &pcc, ci);
			xc_wait(where);
		}
		KASSERT(pcc.pcc_size < size);
		if (pcc.pcc_data != NULL) {
			kmem_free(pcc.pcc_data, pcc.pcc_size);
		}
	}
}

/*
 * percpu_backend_alloc: vmem import callback for percpu_offset_arena
 */

static int
percpu_backend_alloc(vmem_t *dummy, vmem_size_t size, vmem_size_t *resultsize,
    vm_flag_t vmflags, vmem_addr_t *addrp)
{
	unsigned int offset;
	unsigned int nextoff;

	ASSERT_SLEEPABLE();
	KASSERT(dummy == NULL);

	if ((vmflags & VM_NOSLEEP) != 0)
		return ENOMEM;

	size = roundup(size, PERCPU_IMPORT_SIZE);
	mutex_enter(&percpu_allocation_lock);
	offset = percpu_nextoff;
	percpu_nextoff = nextoff = percpu_nextoff + size;
	mutex_exit(&percpu_allocation_lock);

	percpu_cpu_enlarge(nextoff);

	*resultsize = size;
	*addrp = (vmem_addr_t)offset;
	return 0;
}

static void
percpu_zero_cb(void *vp, void *vp2, struct cpu_info *ci)
{
	size_t sz = (uintptr_t)vp2;

	memset(vp, 0, sz);
}

/*
 * percpu_zero: initialize percpu storage with zero.
 */

static void
percpu_zero(percpu_t *pc, size_t sz)
{

	percpu_foreach(pc, percpu_zero_cb, (void *)(uintptr_t)sz);
}

/*
 * percpu_init: subsystem initialization
 */

void
percpu_init(void)
{

	ASSERT_SLEEPABLE();
	rw_init(&percpu_swap_lock);
	mutex_init(&percpu_allocation_lock, MUTEX_DEFAULT, IPL_NONE);
	percpu_nextoff = PERCPU_QUANTUM_SIZE;

	percpu_offset_arena = vmem_xcreate("percpu", 0, 0, PERCPU_QUANTUM_SIZE,
	    percpu_backend_alloc, NULL, NULL, PERCPU_QCACHE_MAX, VM_SLEEP,
	    IPL_NONE);
}

/*
 * percpu_init_cpu: cpu initialization
 *
 * => should be called before the cpu appears on the list for CPU_INFO_FOREACH.
 */

void
percpu_init_cpu(struct cpu_info *ci)
{
	percpu_cpu_t * const pcc = cpu_percpu(ci);
	size_t size = percpu_nextoff; /* XXX racy */

	ASSERT_SLEEPABLE();
	pcc->pcc_size = size;
	if (size) {
		pcc->pcc_data = kmem_zalloc(pcc->pcc_size, KM_SLEEP);
	}
}

/*
 * percpu_alloc: allocate percpu storage
 *
 * => called in thread context.
 * => considered as an expensive and rare operation.
 * => allocated storage is initialized with zeros.
 */

percpu_t *
percpu_alloc(size_t size)
{
	vmem_addr_t offset;
	percpu_t *pc;

	ASSERT_SLEEPABLE();
	if (vmem_alloc(percpu_offset_arena, size, VM_SLEEP | VM_BESTFIT,
	    &offset) != 0)
		return NULL;
	pc = (percpu_t *)percpu_encrypt((uintptr_t)offset);
	percpu_zero(pc, size);
	return pc;
}

/*
 * percpu_free: free percpu storage
 *
 * => called in thread context.
 * => considered as an expensive and rare operation.
 */

void
percpu_free(percpu_t *pc, size_t size)
{

	ASSERT_SLEEPABLE();
	vmem_free(percpu_offset_arena, (vmem_addr_t)percpu_offset(pc), size);
}

/*
 * percpu_getref:
 *
 * => safe to be used in either thread or interrupt context
 * => disables preemption; must be bracketed with a percpu_putref()
 */

void *
percpu_getref(percpu_t *pc)
{

	KPREEMPT_DISABLE(curlwp);
	return percpu_getptr_remote(pc, curcpu());
}

/*
 * percpu_putref:
 *
 * => drops the preemption-disabled count after caller is done with per-cpu
 *    data
 */

void
percpu_putref(percpu_t *pc)
{

	KPREEMPT_ENABLE(curlwp);
}

/*
 * percpu_traverse_enter, percpu_traverse_exit, percpu_getptr_remote:
 * helpers to access remote cpu's percpu data.
 *
 * => called in thread context.
 * => percpu_traverse_enter can block low-priority xcalls.
 * => typical usage would be:
 *
 *	sum = 0;
 *	percpu_traverse_enter();
 *	for (CPU_INFO_FOREACH(cii, ci)) {
 *		unsigned int *p = percpu_getptr_remote(pc, ci);
 *		sum += *p;
 *	}
 *	percpu_traverse_exit();
 */

void
percpu_traverse_enter(void)
{

	ASSERT_SLEEPABLE();
	rw_enter(&percpu_swap_lock, RW_READER);
}

void
percpu_traverse_exit(void)
{

	rw_exit(&percpu_swap_lock);
}

void *
percpu_getptr_remote(percpu_t *pc, struct cpu_info *ci)
{

	return &((char *)cpu_percpu(ci)->pcc_data)[percpu_offset(pc)];
}

/*
 * percpu_foreach: call the specified callback function for each cpus.
 *
 * => called in thread context.
 * => caller should not rely on the cpu iteration order.
 * => the callback function should be minimum because it is executed with
 *    holding a global lock, which can block low-priority xcalls.
 *    eg. it's illegal for a callback function to sleep for memory allocation.
 */
void
percpu_foreach(percpu_t *pc, percpu_callback_t cb, void *arg)
{
	CPU_INFO_ITERATOR cii;
	struct cpu_info *ci;

	percpu_traverse_enter();
	for (CPU_INFO_FOREACH(cii, ci)) {
		(*cb)(percpu_getptr_remote(pc, ci), arg, ci);
	}
	percpu_traverse_exit();
}