aboutsummaryrefslogtreecommitdiff
path: root/drivers/infiniband/core/fmr_pool.c
blob: 176c8f90f2bb359f0b3a933eae1e9ab696518f79 (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
372
373
374
375
376
377
378
379
380
381
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
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
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
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
/*
 * Copyright (c) 2004 Topspin Communications.  All rights reserved.
 * Copyright (c) 2005 Sun Microsystems, Inc. All rights reserved.
 *
 * This software is available to you under a choice of one of two
 * licenses.  You may choose to be licensed under the terms of the GNU
 * General Public License (GPL) Version 2, available from the file
 * COPYING in the main directory of this source tree, or the
 * OpenIB.org BSD license below:
 *
 *     Redistribution and use in source and binary forms, with or
 *     without modification, are permitted provided that the following
 *     conditions are met:
 *
 *      - Redistributions of source code must retain the above
 *        copyright notice, this list of conditions and the following
 *        disclaimer.
 *
 *      - 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.
 *
 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
 * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
 * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
 * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
 * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
 * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
 * SOFTWARE.
 */

#include <linux/errno.h>
#include <linux/spinlock.h>
#include <linux/export.h>
#include <linux/slab.h>
#include <linux/jhash.h>
#include <linux/kthread.h>

#include <rdma/ib_fmr_pool.h>

#include "core_priv.h"

#define PFX "fmr_pool: "

enum {
	IB_FMR_MAX_REMAPS = 32,

	IB_FMR_HASH_BITS  = 8,
	IB_FMR_HASH_SIZE  = 1 << IB_FMR_HASH_BITS,
	IB_FMR_HASH_MASK  = IB_FMR_HASH_SIZE - 1
};

/*
 * If an FMR is not in use, then the list member will point to either
 * its pool's free_list (if the FMR can be mapped again; that is,
 * remap_count < pool->max_remaps) or its pool's dirty_list (if the
 * FMR needs to be unmapped before being remapped).  In either of
 * these cases it is a bug if the ref_count is not 0.  In other words,
 * if ref_count is > 0, then the list member must not be linked into
 * either free_list or dirty_list.
 *
 * The cache_node member is used to link the FMR into a cache bucket
 * (if caching is enabled).  This is independent of the reference
 * count of the FMR.  When a valid FMR is released, its ref_count is
 * decremented, and if ref_count reaches 0, the FMR is placed in
 * either free_list or dirty_list as appropriate.  However, it is not
 * removed from the cache and may be "revived" if a call to
 * ib_fmr_register_physical() occurs before the FMR is remapped.  In
 * this case we just increment the ref_count and remove the FMR from
 * free_list/dirty_list.
 *
 * Before we remap an FMR from free_list, we remove it from the cache
 * (to prevent another user from obtaining a stale FMR).  When an FMR
 * is released, we add it to the tail of the free list, so that our
 * cache eviction policy is "least recently used."
 *
 * All manipulation of ref_count, list and cache_node is protected by
 * pool_lock to maintain consistency.
 */

struct ib_fmr_pool {
	spinlock_t                pool_lock;

	int                       pool_size;
	int                       max_pages;
	int			  max_remaps;
	int                       dirty_watermark;
	int                       dirty_len;
	struct list_head          free_list;
	struct list_head          dirty_list;
	struct hlist_head        *cache_bucket;

	void                     (*flush_function)(struct ib_fmr_pool *pool,
						   void *              arg);
	void                     *flush_arg;

	struct task_struct       *thread;

	atomic_t                  req_ser;
	atomic_t                  flush_ser;

	wait_queue_head_t         force_wait;
};

static inline u32 ib_fmr_hash(u64 first_page)
{
	return jhash_2words((u32) first_page, (u32) (first_page >> 32), 0) &
		(IB_FMR_HASH_SIZE - 1);
}

/* Caller must hold pool_lock */
static inline struct ib_pool_fmr *ib_fmr_cache_lookup(struct ib_fmr_pool *pool,
						      u64 *page_list,
						      int  page_list_len,
						      u64  io_virtual_address)
{
	struct hlist_head *bucket;
	struct ib_pool_fmr *fmr;
	struct hlist_node *pos;

	if (!pool->cache_bucket)
		return NULL;

	bucket = pool->cache_bucket + ib_fmr_hash(*page_list);

	hlist_for_each_entry(fmr, pos, bucket, cache_node)
		if (io_virtual_address == fmr->io_virtual_address &&
		    page_list_len      == fmr->page_list_len      &&
		    !memcmp(page_list, fmr->page_list,
			    page_list_len * sizeof *page_list))
			return fmr;

	return NULL;
}

static void ib_fmr_batch_release(struct ib_fmr_pool *pool)
{
	int                 ret;
	struct ib_pool_fmr *fmr;
	LIST_HEAD(unmap_list);
	LIST_HEAD(fmr_list);

	spin_lock_irq(&pool->pool_lock);

	list_for_each_entry(fmr, &pool->dirty_list, list) {
		hlist_del_init(&fmr->cache_node);
		fmr->remap_count = 0;
		list_add_tail(&fmr->fmr->list, &fmr_list);

#ifdef DEBUG
		if (fmr->ref_count !=0) {
			printk(KERN_WARNING PFX "Unmapping FMR 0x%08x with ref count %d\n",
			       fmr, fmr->ref_count);
		}
#endif
	}

	list_splice_init(&pool->dirty_list, &unmap_list);
	pool->dirty_len = 0;

	spin_unlock_irq(&pool->pool_lock);

	if (list_empty(&unmap_list)) {
		return;
	}

	ret = ib_unmap_fmr(&fmr_list);
	if (ret)
		printk(KERN_WARNING PFX "ib_unmap_fmr returned %d\n", ret);

	spin_lock_irq(&pool->pool_lock);
	list_splice(&unmap_list, &pool->free_list);
	spin_unlock_irq(&pool->pool_lock);
}

static int ib_fmr_cleanup_thread(void *pool_ptr)
{
	struct ib_fmr_pool *pool = pool_ptr;

	do {
		if (atomic_read(&pool->flush_ser) - atomic_read(&pool->req_ser) < 0) {
			ib_fmr_batch_release(pool);

			atomic_inc(&pool->flush_ser);
			wake_up_interruptible(&pool->force_wait);

			if (pool->flush_function)
				pool->flush_function(pool, pool->flush_arg);
		}

		set_current_state(TASK_INTERRUPTIBLE);
		if (atomic_read(&pool->flush_ser) - atomic_read(&pool->req_ser) >= 0 &&
		    !kthread_should_stop())
			schedule();
		__set_current_state(TASK_RUNNING);
	} while (!kthread_should_stop());

	return 0;
}

/**
 * ib_create_fmr_pool - Create an FMR pool
 * @pd:Protection domain for FMRs
 * @params:FMR pool parameters
 *
 * Create a pool of FMRs.  Return value is pointer to new pool or
 * error code if creation failed.
 */
struct ib_fmr_pool *ib_create_fmr_pool(struct ib_pd             *pd,
				       struct ib_fmr_pool_param *params)
{
	struct ib_device   *device;
	struct ib_fmr_pool *pool;
	struct ib_device_attr *attr;
	int i;
	int ret;
	int max_remaps;

	if (!params)
		return ERR_PTR(-EINVAL);

	device = pd->device;
	if (!device->alloc_fmr    || !device->dealloc_fmr  ||
	    !device->map_phys_fmr || !device->unmap_fmr) {
		printk(KERN_INFO PFX "Device %s does not support FMRs\n",
		       device->name);
		return ERR_PTR(-ENOSYS);
	}

	attr = kmalloc(sizeof *attr, GFP_KERNEL);
	if (!attr) {
		printk(KERN_WARNING PFX "couldn't allocate device attr struct\n");
		return ERR_PTR(-ENOMEM);
	}

	ret = ib_query_device(device, attr);
	if (ret) {
		printk(KERN_WARNING PFX "couldn't query device: %d\n", ret);
		kfree(attr);
		return ERR_PTR(ret);
	}

	if (!attr->max_map_per_fmr)
		max_remaps = IB_FMR_MAX_REMAPS;
	else
		max_remaps = attr->max_map_per_fmr;

	kfree(attr);

	pool = kmalloc(sizeof *pool, GFP_KERNEL);
	if (!pool) {
		printk(KERN_WARNING PFX "couldn't allocate pool struct\n");
		return ERR_PTR(-ENOMEM);
	}

	pool->cache_bucket   = NULL;

	pool->flush_function = params->flush_function;
	pool->flush_arg      = params->flush_arg;

	INIT_LIST_HEAD(&pool->free_list);
	INIT_LIST_HEAD(&pool->dirty_list);

	if (params->cache) {
		pool->cache_bucket =
			kmalloc(IB_FMR_HASH_SIZE * sizeof *pool->cache_bucket,
				GFP_KERNEL);
		if (!pool->cache_bucket) {
			printk(KERN_WARNING PFX "Failed to allocate cache in pool\n");
			ret = -ENOMEM;
			goto out_free_pool;
		}

		for (i = 0; i < IB_FMR_HASH_SIZE; ++i)
			INIT_HLIST_HEAD(pool->cache_bucket + i);
	}

	pool->pool_size       = 0;
	pool->max_pages       = params->max_pages_per_fmr;
	pool->max_remaps      = max_remaps;
	pool->dirty_watermark = params->dirty_watermark;
	pool->dirty_len       = 0;
	spin_lock_init(&pool->pool_lock);
	atomic_set(&pool->req_ser,   0);
	atomic_set(&pool->flush_ser, 0);
	init_waitqueue_head(&pool->force_wait);

	pool->thread = kthread_run(ib_fmr_cleanup_thread,
				   pool,
				   "ib_fmr(%s)",
				   device->name);
	if (IS_ERR(pool->thread)) {
		printk(KERN_WARNING PFX "couldn't start cleanup thread\n");
		ret = PTR_ERR(pool->thread);
		goto out_free_pool;
	}

	{
		struct ib_pool_fmr *fmr;
		struct ib_fmr_attr fmr_attr = {
			.max_pages  = params->max_pages_per_fmr,
			.max_maps   = pool->max_remaps,
			.page_shift = params->page_shift
		};
		int bytes_per_fmr = sizeof *fmr;

		if (pool->cache_bucket)
			bytes_per_fmr += params->max_pages_per_fmr * sizeof (u64);

		for (i = 0; i < params->pool_size; ++i) {
			fmr = kmalloc(bytes_per_fmr, GFP_KERNEL);
			if (!fmr) {
				printk(KERN_WARNING PFX "failed to allocate fmr "
				       "struct for FMR %d\n", i);
				goto out_fail;
			}

			fmr->pool             = pool;
			fmr->remap_count      = 0;
			fmr->ref_count        = 0;
			INIT_HLIST_NODE(&fmr->cache_node);

			fmr->fmr = ib_alloc_fmr(pd, params->access, &fmr_attr);
			if (IS_ERR(fmr->fmr)) {
				printk(KERN_WARNING PFX "fmr_create failed "
				       "for FMR %d\n", i);
				kfree(fmr);
				goto out_fail;
			}

			list_add_tail(&fmr->list, &pool->free_list);
			++pool->pool_size;
		}
	}

	return pool;

 out_free_pool:
	kfree(pool->cache_bucket);
	kfree(pool);

	return ERR_PTR(ret);

 out_fail:
	ib_destroy_fmr_pool(pool);

	return ERR_PTR(-ENOMEM);
}
EXPORT_SYMBOL(ib_create_fmr_pool);

/**
 * ib_destroy_fmr_pool - Free FMR pool
 * @pool:FMR pool to free
 *
 * Destroy an FMR pool and free all associated resources.
 */
void ib_destroy_fmr_pool(struct ib_fmr_pool *pool)
{
	struct ib_pool_fmr *fmr;
	struct ib_pool_fmr *tmp;
	LIST_HEAD(fmr_list);
	int                 i;

	kthread_stop(pool->thread);
	ib_fmr_batch_release(pool);

	i = 0;
	list_for_each_entry_safe(fmr, tmp, &pool->free_list, list) {
		if (fmr->remap_count) {
			INIT_LIST_HEAD(&fmr_list);
			list_add_tail(&fmr->fmr->list, &fmr_list);
			ib_unmap_fmr(&fmr_list);
		}
		ib_dealloc_fmr(fmr->fmr);
		list_del(&fmr->list);
		kfree(fmr);
		++i;
	}

	if (i < pool->pool_size)
		printk(KERN_WARNING PFX "pool still has %d regions registered\n",
		       pool->pool_size - i);

	kfree(pool->cache_bucket);
	kfree(pool);
}
EXPORT_SYMBOL(ib_destroy_fmr_pool);

/**
 * ib_flush_fmr_pool - Invalidate all unmapped FMRs
 * @pool:FMR pool to flush
 *
 * Ensure that all unmapped FMRs are fully invalidated.
 */
int ib_flush_fmr_pool(struct ib_fmr_pool *pool)
{
	int serial;
	struct ib_pool_fmr *fmr, *next;

	/*
	 * The free_list holds FMRs that may have been used
	 * but have not been remapped enough times to be dirty.
	 * Put them on the dirty list now so that the cleanup
	 * thread will reap them too.
	 */
	spin_lock_irq(&pool->pool_lock);
	list_for_each_entry_safe(fmr, next, &pool->free_list, list) {
		if (fmr->remap_count > 0)
			list_move(&fmr->list, &pool->dirty_list);
	}
	spin_unlock_irq(&pool->pool_lock);

	serial = atomic_inc_return(&pool->req_ser);
	wake_up_process(pool->thread);

	if (wait_event_interruptible(pool->force_wait,
				     atomic_read(&pool->flush_ser) - serial >= 0))
		return -EINTR;

	return 0;
}
EXPORT_SYMBOL(ib_flush_fmr_pool);

/**
 * ib_fmr_pool_map_phys -
 * @pool:FMR pool to allocate FMR from
 * @page_list:List of pages to map
 * @list_len:Number of pages in @page_list
 * @io_virtual_address:I/O virtual address for new FMR
 *
 * Map an FMR from an FMR pool.
 */
struct ib_pool_fmr *ib_fmr_pool_map_phys(struct ib_fmr_pool *pool_handle,
					 u64                *page_list,
					 int                 list_len,
					 u64                 io_virtual_address)
{
	struct ib_fmr_pool *pool = pool_handle;
	struct ib_pool_fmr *fmr;
	unsigned long       flags;
	int                 result;

	if (list_len < 1 || list_len > pool->max_pages)
		return ERR_PTR(-EINVAL);

	spin_lock_irqsave(&pool->pool_lock, flags);
	fmr = ib_fmr_cache_lookup(pool,
				  page_list,
				  list_len,
				  io_virtual_address);
	if (fmr) {
		/* found in cache */
		++fmr->ref_count;
		if (fmr->ref_count == 1) {
			list_del(&fmr->list);
		}

		spin_unlock_irqrestore(&pool->pool_lock, flags);

		return fmr;
	}

	if (list_empty(&pool->free_list)) {
		spin_unlock_irqrestore(&pool->pool_lock, flags);
		return ERR_PTR(-EAGAIN);
	}

	fmr = list_entry(pool->free_list.next, struct ib_pool_fmr, list);
	list_del(&fmr->list);
	hlist_del_init(&fmr->cache_node);
	spin_unlock_irqrestore(&pool->pool_lock, flags);

	result = ib_map_phys_fmr(fmr->fmr, page_list, list_len,
				 io_virtual_address);

	if (result) {
		spin_lock_irqsave(&pool->pool_lock, flags);
		list_add(&fmr->list, &pool->free_list);
		spin_unlock_irqrestore(&pool->pool_lock, flags);

		printk(KERN_WARNING PFX "fmr_map returns %d\n", result);

		return ERR_PTR(result);
	}

	++fmr->remap_count;
	fmr->ref_count = 1;

	if (pool->cache_bucket) {
		fmr->io_virtual_address = io_virtual_address;
		fmr->page_list_len      = list_len;
		memcpy(fmr->page_list, page_list, list_len * sizeof(*page_list));

		spin_lock_irqsave(&pool->pool_lock, flags);
		hlist_add_head(&fmr->cache_node,
			       pool->cache_bucket + ib_fmr_hash(fmr->page_list[0]));
		spin_unlock_irqrestore(&pool->pool_lock, flags);
	}

	return fmr;
}
EXPORT_SYMBOL(ib_fmr_pool_map_phys);

/**
 * ib_fmr_pool_unmap - Unmap FMR
 * @fmr:FMR to unmap
 *
 * Unmap an FMR.  The FMR mapping may remain valid until the FMR is
 * reused (or until ib_flush_fmr_pool() is called).
 */
int ib_fmr_pool_unmap(struct ib_pool_fmr *fmr)
{
	struct ib_fmr_pool *pool;
	unsigned long flags;

	pool = fmr->pool;

	spin_lock_irqsave(&pool->pool_lock, flags);

	--fmr->ref_count;
	if (!fmr->ref_count) {
		if (fmr->remap_count < pool->max_remaps) {
			list_add_tail(&fmr->list, &pool->free_list);
		} else {
			list_add_tail(&fmr->list, &pool->dirty_list);
			if (++pool->dirty_len >= pool->dirty_watermark) {
				atomic_inc(&pool->req_ser);
				wake_up_process(pool->thread);
			}
		}
	}

#ifdef DEBUG
	if (fmr->ref_count < 0)
		printk(KERN_WARNING PFX "FMR %p has ref count %d < 0\n",
		       fmr, fmr->ref_count);
#endif

	spin_unlock_irqrestore(&pool->pool_lock, flags);

	return 0;
}
EXPORT_SYMBOL(ib_fmr_pool_unmap);