mr.c 28.4 KB
Newer Older
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
/*
 * Copyright (c) 2013, Mellanox Technologies 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/kref.h>
#include <linux/random.h>
#include <linux/debugfs.h>
#include <linux/export.h>
E
Eli Cohen 已提交
38
#include <linux/delay.h>
39 40 41 42
#include <rdma/ib_umem.h>
#include "mlx5_ib.h"

enum {
E
Eli Cohen 已提交
43
	MAX_PENDING_REG_MR = 8,
44 45
};

46 47 48 49
enum {
	MLX5_UMR_ALIGN	= 2048
};

50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66
static __be64 *mr_align(__be64 *ptr, int align)
{
	unsigned long mask = align - 1;

	return (__be64 *)(((unsigned long)ptr + mask) & ~mask);
}

static int order2idx(struct mlx5_ib_dev *dev, int order)
{
	struct mlx5_mr_cache *cache = &dev->cache;

	if (order < cache->ent[0].order)
		return 0;
	else
		return order - cache->ent[0].order;
}

E
Eli Cohen 已提交
67 68 69 70 71 72 73 74 75
static void reg_mr_callback(int status, void *context)
{
	struct mlx5_ib_mr *mr = context;
	struct mlx5_ib_dev *dev = mr->dev;
	struct mlx5_mr_cache *cache = &dev->cache;
	int c = order2idx(dev, mr->order);
	struct mlx5_cache_ent *ent = &cache->ent[c];
	u8 key;
	unsigned long flags;
76
	struct mlx5_mr_table *table = &dev->mdev->priv.mr_table;
77
	int err;
E
Eli Cohen 已提交
78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99

	spin_lock_irqsave(&ent->lock, flags);
	ent->pending--;
	spin_unlock_irqrestore(&ent->lock, flags);
	if (status) {
		mlx5_ib_warn(dev, "async reg mr failed. status %d\n", status);
		kfree(mr);
		dev->fill_delay = 1;
		mod_timer(&dev->delay_timer, jiffies + HZ);
		return;
	}

	if (mr->out.hdr.status) {
		mlx5_ib_warn(dev, "failed - status %d, syndorme 0x%x\n",
			     mr->out.hdr.status,
			     be32_to_cpu(mr->out.hdr.syndrome));
		kfree(mr);
		dev->fill_delay = 1;
		mod_timer(&dev->delay_timer, jiffies + HZ);
		return;
	}

100 101 102
	spin_lock_irqsave(&dev->mdev->priv.mkey_lock, flags);
	key = dev->mdev->priv.mkey_key++;
	spin_unlock_irqrestore(&dev->mdev->priv.mkey_lock, flags);
E
Eli Cohen 已提交
103 104 105 106 107 108 109 110 111
	mr->mmr.key = mlx5_idx_to_mkey(be32_to_cpu(mr->out.mkey) & 0xffffff) | key;

	cache->last_add = jiffies;

	spin_lock_irqsave(&ent->lock, flags);
	list_add_tail(&mr->list, &ent->head);
	ent->cur++;
	ent->size++;
	spin_unlock_irqrestore(&ent->lock, flags);
112 113 114 115 116 117 118

	write_lock_irqsave(&table->lock, flags);
	err = radix_tree_insert(&table->tree, mlx5_base_mkey(mr->mmr.key),
				&mr->mmr);
	if (err)
		pr_err("Error inserting to mr tree. 0x%x\n", -err);
	write_unlock_irqrestore(&table->lock, flags);
E
Eli Cohen 已提交
119 120
}

121 122 123 124 125 126 127 128 129 130 131 132 133 134 135
static int add_keys(struct mlx5_ib_dev *dev, int c, int num)
{
	struct mlx5_mr_cache *cache = &dev->cache;
	struct mlx5_cache_ent *ent = &cache->ent[c];
	struct mlx5_create_mkey_mbox_in *in;
	struct mlx5_ib_mr *mr;
	int npages = 1 << ent->order;
	int err = 0;
	int i;

	in = kzalloc(sizeof(*in), GFP_KERNEL);
	if (!in)
		return -ENOMEM;

	for (i = 0; i < num; i++) {
E
Eli Cohen 已提交
136 137 138 139 140
		if (ent->pending >= MAX_PENDING_REG_MR) {
			err = -EAGAIN;
			break;
		}

141 142 143
		mr = kzalloc(sizeof(*mr), GFP_KERNEL);
		if (!mr) {
			err = -ENOMEM;
E
Eli Cohen 已提交
144
			break;
145 146 147
		}
		mr->order = ent->order;
		mr->umred = 1;
E
Eli Cohen 已提交
148
		mr->dev = dev;
149 150 151 152 153 154
		in->seg.status = 1 << 6;
		in->seg.xlt_oct_size = cpu_to_be32((npages + 1) / 2);
		in->seg.qpn_mkey7_0 = cpu_to_be32(0xffffff << 8);
		in->seg.flags = MLX5_ACCESS_MODE_MTT | MLX5_PERM_UMR_EN;
		in->seg.log2_page_size = 12;

E
Eli Cohen 已提交
155 156 157
		spin_lock_irq(&ent->lock);
		ent->pending++;
		spin_unlock_irq(&ent->lock);
158
		err = mlx5_core_create_mkey(dev->mdev, &mr->mmr, in,
E
Eli Cohen 已提交
159 160
					    sizeof(*in), reg_mr_callback,
					    mr, &mr->out);
161 162 163
		if (err) {
			mlx5_ib_warn(dev, "create mkey failed %d\n", err);
			kfree(mr);
E
Eli Cohen 已提交
164
			break;
165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180
		}
	}

	kfree(in);
	return err;
}

static void remove_keys(struct mlx5_ib_dev *dev, int c, int num)
{
	struct mlx5_mr_cache *cache = &dev->cache;
	struct mlx5_cache_ent *ent = &cache->ent[c];
	struct mlx5_ib_mr *mr;
	int err;
	int i;

	for (i = 0; i < num; i++) {
E
Eli Cohen 已提交
181
		spin_lock_irq(&ent->lock);
182
		if (list_empty(&ent->head)) {
E
Eli Cohen 已提交
183
			spin_unlock_irq(&ent->lock);
184 185 186 187 188 189
			return;
		}
		mr = list_first_entry(&ent->head, struct mlx5_ib_mr, list);
		list_del(&mr->list);
		ent->cur--;
		ent->size--;
E
Eli Cohen 已提交
190
		spin_unlock_irq(&ent->lock);
191
		err = mlx5_core_destroy_mkey(dev->mdev, &mr->mmr);
192
		if (err)
193
			mlx5_ib_warn(dev, "failed destroy mkey\n");
194
		else
195 196 197 198 199 200 201 202 203 204 205 206 207 208 209
			kfree(mr);
	}
}

static ssize_t size_write(struct file *filp, const char __user *buf,
			  size_t count, loff_t *pos)
{
	struct mlx5_cache_ent *ent = filp->private_data;
	struct mlx5_ib_dev *dev = ent->dev;
	char lbuf[20];
	u32 var;
	int err;
	int c;

	if (copy_from_user(lbuf, buf, sizeof(lbuf)))
210
		return -EFAULT;
211 212 213 214 215 216 217 218 219 220 221

	c = order2idx(dev, ent->order);
	lbuf[sizeof(lbuf) - 1] = 0;

	if (sscanf(lbuf, "%u", &var) != 1)
		return -EINVAL;

	if (var < ent->limit)
		return -EINVAL;

	if (var > ent->size) {
E
Eli Cohen 已提交
222 223 224 225 226 227 228
		do {
			err = add_keys(dev, c, var - ent->size);
			if (err && err != -EAGAIN)
				return err;

			usleep_range(3000, 5000);
		} while (err);
229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250
	} else if (var < ent->size) {
		remove_keys(dev, c, ent->size - var);
	}

	return count;
}

static ssize_t size_read(struct file *filp, char __user *buf, size_t count,
			 loff_t *pos)
{
	struct mlx5_cache_ent *ent = filp->private_data;
	char lbuf[20];
	int err;

	if (*pos)
		return 0;

	err = snprintf(lbuf, sizeof(lbuf), "%d\n", ent->size);
	if (err < 0)
		return err;

	if (copy_to_user(buf, lbuf, err))
251
		return -EFAULT;
252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275

	*pos += err;

	return err;
}

static const struct file_operations size_fops = {
	.owner	= THIS_MODULE,
	.open	= simple_open,
	.write	= size_write,
	.read	= size_read,
};

static ssize_t limit_write(struct file *filp, const char __user *buf,
			   size_t count, loff_t *pos)
{
	struct mlx5_cache_ent *ent = filp->private_data;
	struct mlx5_ib_dev *dev = ent->dev;
	char lbuf[20];
	u32 var;
	int err;
	int c;

	if (copy_from_user(lbuf, buf, sizeof(lbuf)))
276
		return -EFAULT;
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

	c = order2idx(dev, ent->order);
	lbuf[sizeof(lbuf) - 1] = 0;

	if (sscanf(lbuf, "%u", &var) != 1)
		return -EINVAL;

	if (var > ent->size)
		return -EINVAL;

	ent->limit = var;

	if (ent->cur < ent->limit) {
		err = add_keys(dev, c, 2 * ent->limit - ent->cur);
		if (err)
			return err;
	}

	return count;
}

static ssize_t limit_read(struct file *filp, char __user *buf, size_t count,
			  loff_t *pos)
{
	struct mlx5_cache_ent *ent = filp->private_data;
	char lbuf[20];
	int err;

	if (*pos)
		return 0;

	err = snprintf(lbuf, sizeof(lbuf), "%d\n", ent->limit);
	if (err < 0)
		return err;

	if (copy_to_user(buf, lbuf, err))
313
		return -EFAULT;
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

	*pos += err;

	return err;
}

static const struct file_operations limit_fops = {
	.owner	= THIS_MODULE,
	.open	= simple_open,
	.write	= limit_write,
	.read	= limit_read,
};

static int someone_adding(struct mlx5_mr_cache *cache)
{
	int i;

	for (i = 0; i < MAX_MR_CACHE_ENTRIES; i++) {
		if (cache->ent[i].cur < cache->ent[i].limit)
			return 1;
	}

	return 0;
}

static void __cache_work_func(struct mlx5_cache_ent *ent)
{
	struct mlx5_ib_dev *dev = ent->dev;
	struct mlx5_mr_cache *cache = &dev->cache;
	int i = order2idx(dev, ent->order);
E
Eli Cohen 已提交
344
	int err;
345 346 347 348 349

	if (cache->stopped)
		return;

	ent = &dev->cache.ent[i];
E
Eli Cohen 已提交
350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366
	if (ent->cur < 2 * ent->limit && !dev->fill_delay) {
		err = add_keys(dev, i, 1);
		if (ent->cur < 2 * ent->limit) {
			if (err == -EAGAIN) {
				mlx5_ib_dbg(dev, "returned eagain, order %d\n",
					    i + 2);
				queue_delayed_work(cache->wq, &ent->dwork,
						   msecs_to_jiffies(3));
			} else if (err) {
				mlx5_ib_warn(dev, "command failed order %d, err %d\n",
					     i + 2, err);
				queue_delayed_work(cache->wq, &ent->dwork,
						   msecs_to_jiffies(1000));
			} else {
				queue_work(cache->wq, &ent->work);
			}
		}
367 368
	} else if (ent->cur > 2 * ent->limit) {
		if (!someone_adding(cache) &&
E
Eli Cohen 已提交
369
		    time_after(jiffies, cache->last_add + 300 * HZ)) {
370 371 372 373
			remove_keys(dev, i, 1);
			if (ent->cur > ent->limit)
				queue_work(cache->wq, &ent->work);
		} else {
E
Eli Cohen 已提交
374
			queue_delayed_work(cache->wq, &ent->dwork, 300 * HZ);
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
		}
	}
}

static void delayed_cache_work_func(struct work_struct *work)
{
	struct mlx5_cache_ent *ent;

	ent = container_of(work, struct mlx5_cache_ent, dwork.work);
	__cache_work_func(ent);
}

static void cache_work_func(struct work_struct *work)
{
	struct mlx5_cache_ent *ent;

	ent = container_of(work, struct mlx5_cache_ent, work);
	__cache_work_func(ent);
}

static struct mlx5_ib_mr *alloc_cached_mr(struct mlx5_ib_dev *dev, int order)
{
	struct mlx5_mr_cache *cache = &dev->cache;
	struct mlx5_ib_mr *mr = NULL;
	struct mlx5_cache_ent *ent;
	int c;
	int i;

	c = order2idx(dev, order);
	if (c < 0 || c >= MAX_MR_CACHE_ENTRIES) {
		mlx5_ib_warn(dev, "order %d, cache index %d\n", order, c);
		return NULL;
	}

	for (i = c; i < MAX_MR_CACHE_ENTRIES; i++) {
		ent = &cache->ent[i];

		mlx5_ib_dbg(dev, "order %d, cache index %d\n", ent->order, i);

E
Eli Cohen 已提交
414
		spin_lock_irq(&ent->lock);
415 416 417 418 419
		if (!list_empty(&ent->head)) {
			mr = list_first_entry(&ent->head, struct mlx5_ib_mr,
					      list);
			list_del(&mr->list);
			ent->cur--;
E
Eli Cohen 已提交
420
			spin_unlock_irq(&ent->lock);
421 422 423 424
			if (ent->cur < ent->limit)
				queue_work(cache->wq, &ent->work);
			break;
		}
E
Eli Cohen 已提交
425
		spin_unlock_irq(&ent->lock);
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

		queue_work(cache->wq, &ent->work);

		if (mr)
			break;
	}

	if (!mr)
		cache->ent[c].miss++;

	return mr;
}

static void free_cached_mr(struct mlx5_ib_dev *dev, struct mlx5_ib_mr *mr)
{
	struct mlx5_mr_cache *cache = &dev->cache;
	struct mlx5_cache_ent *ent;
	int shrink = 0;
	int c;

	c = order2idx(dev, mr->order);
	if (c < 0 || c >= MAX_MR_CACHE_ENTRIES) {
		mlx5_ib_warn(dev, "order %d, cache index %d\n", mr->order, c);
		return;
	}
	ent = &cache->ent[c];
E
Eli Cohen 已提交
452
	spin_lock_irq(&ent->lock);
453 454 455 456
	list_add_tail(&mr->list, &ent->head);
	ent->cur++;
	if (ent->cur > 2 * ent->limit)
		shrink = 1;
E
Eli Cohen 已提交
457
	spin_unlock_irq(&ent->lock);
458 459 460 461 462 463 464 465 466 467 468 469

	if (shrink)
		queue_work(cache->wq, &ent->work);
}

static void clean_keys(struct mlx5_ib_dev *dev, int c)
{
	struct mlx5_mr_cache *cache = &dev->cache;
	struct mlx5_cache_ent *ent = &cache->ent[c];
	struct mlx5_ib_mr *mr;
	int err;

470
	cancel_delayed_work(&ent->dwork);
471
	while (1) {
E
Eli Cohen 已提交
472
		spin_lock_irq(&ent->lock);
473
		if (list_empty(&ent->head)) {
E
Eli Cohen 已提交
474
			spin_unlock_irq(&ent->lock);
475 476 477 478 479 480
			return;
		}
		mr = list_first_entry(&ent->head, struct mlx5_ib_mr, list);
		list_del(&mr->list);
		ent->cur--;
		ent->size--;
E
Eli Cohen 已提交
481
		spin_unlock_irq(&ent->lock);
482
		err = mlx5_core_destroy_mkey(dev->mdev, &mr->mmr);
483
		if (err)
484
			mlx5_ib_warn(dev, "failed destroy mkey\n");
485
		else
486 487 488 489 490 491 492 493 494 495 496 497 498
			kfree(mr);
	}
}

static int mlx5_mr_cache_debugfs_init(struct mlx5_ib_dev *dev)
{
	struct mlx5_mr_cache *cache = &dev->cache;
	struct mlx5_cache_ent *ent;
	int i;

	if (!mlx5_debugfs_root)
		return 0;

499
	cache->root = debugfs_create_dir("mr_cache", dev->mdev->priv.dbg_root);
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
	if (!cache->root)
		return -ENOMEM;

	for (i = 0; i < MAX_MR_CACHE_ENTRIES; i++) {
		ent = &cache->ent[i];
		sprintf(ent->name, "%d", ent->order);
		ent->dir = debugfs_create_dir(ent->name,  cache->root);
		if (!ent->dir)
			return -ENOMEM;

		ent->fsize = debugfs_create_file("size", 0600, ent->dir, ent,
						 &size_fops);
		if (!ent->fsize)
			return -ENOMEM;

		ent->flimit = debugfs_create_file("limit", 0600, ent->dir, ent,
						  &limit_fops);
		if (!ent->flimit)
			return -ENOMEM;

		ent->fcur = debugfs_create_u32("cur", 0400, ent->dir,
					       &ent->cur);
		if (!ent->fcur)
			return -ENOMEM;

		ent->fmiss = debugfs_create_u32("miss", 0600, ent->dir,
						&ent->miss);
		if (!ent->fmiss)
			return -ENOMEM;
	}

	return 0;
}

static void mlx5_mr_cache_debugfs_cleanup(struct mlx5_ib_dev *dev)
{
	if (!mlx5_debugfs_root)
		return;

	debugfs_remove_recursive(dev->cache.root);
}

E
Eli Cohen 已提交
542 543 544 545 546 547 548
static void delay_time_func(unsigned long ctx)
{
	struct mlx5_ib_dev *dev = (struct mlx5_ib_dev *)ctx;

	dev->fill_delay = 0;
}

549 550 551 552 553 554 555 556 557 558 559 560 561 562
int mlx5_mr_cache_init(struct mlx5_ib_dev *dev)
{
	struct mlx5_mr_cache *cache = &dev->cache;
	struct mlx5_cache_ent *ent;
	int limit;
	int err;
	int i;

	cache->wq = create_singlethread_workqueue("mkey_cache");
	if (!cache->wq) {
		mlx5_ib_warn(dev, "failed to create work queue\n");
		return -ENOMEM;
	}

E
Eli Cohen 已提交
563
	setup_timer(&dev->delay_timer, delay_time_func, (unsigned long)dev);
564 565 566 567 568 569 570 571 572 573
	for (i = 0; i < MAX_MR_CACHE_ENTRIES; i++) {
		INIT_LIST_HEAD(&cache->ent[i].head);
		spin_lock_init(&cache->ent[i].lock);

		ent = &cache->ent[i];
		INIT_LIST_HEAD(&ent->head);
		spin_lock_init(&ent->lock);
		ent->order = i + 2;
		ent->dev = dev;

574 575
		if (dev->mdev->profile->mask & MLX5_PROF_MASK_MR_CACHE)
			limit = dev->mdev->profile->mr_cache[i].limit;
E
Eli Cohen 已提交
576
		else
577
			limit = 0;
E
Eli Cohen 已提交
578

579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596
		INIT_WORK(&ent->work, cache_work_func);
		INIT_DELAYED_WORK(&ent->dwork, delayed_cache_work_func);
		ent->limit = limit;
		queue_work(cache->wq, &ent->work);
	}

	err = mlx5_mr_cache_debugfs_init(dev);
	if (err)
		mlx5_ib_warn(dev, "cache debugfs failure\n");

	return 0;
}

int mlx5_mr_cache_cleanup(struct mlx5_ib_dev *dev)
{
	int i;

	dev->cache.stopped = 1;
597
	flush_workqueue(dev->cache.wq);
598 599 600 601 602 603

	mlx5_mr_cache_debugfs_cleanup(dev);

	for (i = 0; i < MAX_MR_CACHE_ENTRIES; i++)
		clean_keys(dev, i);

604
	destroy_workqueue(dev->cache.wq);
E
Eli Cohen 已提交
605
	del_timer_sync(&dev->delay_timer);
606

607 608 609 610 611 612
	return 0;
}

struct ib_mr *mlx5_ib_get_dma_mr(struct ib_pd *pd, int acc)
{
	struct mlx5_ib_dev *dev = to_mdev(pd->device);
613
	struct mlx5_core_dev *mdev = dev->mdev;
614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634
	struct mlx5_create_mkey_mbox_in *in;
	struct mlx5_mkey_seg *seg;
	struct mlx5_ib_mr *mr;
	int err;

	mr = kzalloc(sizeof(*mr), GFP_KERNEL);
	if (!mr)
		return ERR_PTR(-ENOMEM);

	in = kzalloc(sizeof(*in), GFP_KERNEL);
	if (!in) {
		err = -ENOMEM;
		goto err_free;
	}

	seg = &in->seg;
	seg->flags = convert_access(acc) | MLX5_ACCESS_MODE_PA;
	seg->flags_pd = cpu_to_be32(to_mpd(pd)->pdn | MLX5_MKEY_LEN64);
	seg->qpn_mkey7_0 = cpu_to_be32(0xffffff << 8);
	seg->start_addr = 0;

E
Eli Cohen 已提交
635 636
	err = mlx5_core_create_mkey(mdev, &mr->mmr, in, sizeof(*in), NULL, NULL,
				    NULL);
637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710
	if (err)
		goto err_in;

	kfree(in);
	mr->ibmr.lkey = mr->mmr.key;
	mr->ibmr.rkey = mr->mmr.key;
	mr->umem = NULL;

	return &mr->ibmr;

err_in:
	kfree(in);

err_free:
	kfree(mr);

	return ERR_PTR(err);
}

static int get_octo_len(u64 addr, u64 len, int page_size)
{
	u64 offset;
	int npages;

	offset = addr & (page_size - 1);
	npages = ALIGN(len + offset, page_size) >> ilog2(page_size);
	return (npages + 1) / 2;
}

static int use_umr(int order)
{
	return order <= 17;
}

static void prep_umr_reg_wqe(struct ib_pd *pd, struct ib_send_wr *wr,
			     struct ib_sge *sg, u64 dma, int n, u32 key,
			     int page_shift, u64 virt_addr, u64 len,
			     int access_flags)
{
	struct mlx5_ib_dev *dev = to_mdev(pd->device);
	struct ib_mr *mr = dev->umrc.mr;

	sg->addr = dma;
	sg->length = ALIGN(sizeof(u64) * n, 64);
	sg->lkey = mr->lkey;

	wr->next = NULL;
	wr->send_flags = 0;
	wr->sg_list = sg;
	if (n)
		wr->num_sge = 1;
	else
		wr->num_sge = 0;

	wr->opcode = MLX5_IB_WR_UMR;
	wr->wr.fast_reg.page_list_len = n;
	wr->wr.fast_reg.page_shift = page_shift;
	wr->wr.fast_reg.rkey = key;
	wr->wr.fast_reg.iova_start = virt_addr;
	wr->wr.fast_reg.length = len;
	wr->wr.fast_reg.access_flags = access_flags;
	wr->wr.fast_reg.page_list = (struct ib_fast_reg_page_list *)pd;
}

static void prep_umr_unreg_wqe(struct mlx5_ib_dev *dev,
			       struct ib_send_wr *wr, u32 key)
{
	wr->send_flags = MLX5_IB_SEND_UMR_UNREG;
	wr->opcode = MLX5_IB_WR_UMR;
	wr->wr.fast_reg.rkey = key;
}

void mlx5_umr_cq_handler(struct ib_cq *cq, void *cq_context)
{
711
	struct mlx5_ib_umr_context *context;
712 713 714 715 716 717 718 719 720 721 722 723
	struct ib_wc wc;
	int err;

	while (1) {
		err = ib_poll_cq(cq, 1, &wc);
		if (err < 0) {
			pr_warn("poll cq error %d\n", err);
			return;
		}
		if (err == 0)
			break;

724
		context = (struct mlx5_ib_umr_context *) (unsigned long) wc.wr_id;
725 726
		context->status = wc.status;
		complete(&context->done);
727 728 729 730 731 732 733 734 735
	}
	ib_req_notify_cq(cq, IB_CQ_NEXT_COMP);
}

static struct mlx5_ib_mr *reg_umr(struct ib_pd *pd, struct ib_umem *umem,
				  u64 virt_addr, u64 len, int npages,
				  int page_shift, int order, int access_flags)
{
	struct mlx5_ib_dev *dev = to_mdev(pd->device);
736
	struct device *ddev = dev->ib_dev.dma_device;
737
	struct umr_common *umrc = &dev->umrc;
738
	struct mlx5_ib_umr_context umr_context;
739 740 741
	struct ib_send_wr wr, *bad;
	struct mlx5_ib_mr *mr;
	struct ib_sge sg;
742
	int size = sizeof(u64) * npages;
743
	int err = 0;
744 745
	int i;

E
Eli Cohen 已提交
746
	for (i = 0; i < 1; i++) {
747 748 749 750 751
		mr = alloc_cached_mr(dev, order);
		if (mr)
			break;

		err = add_keys(dev, order2idx(dev, order), 1);
E
Eli Cohen 已提交
752 753
		if (err && err != -EAGAIN) {
			mlx5_ib_warn(dev, "add_keys failed, err %d\n", err);
754 755 756 757 758 759 760
			break;
		}
	}

	if (!mr)
		return ERR_PTR(-EAGAIN);

761
	mr->pas = kmalloc(size + MLX5_UMR_ALIGN - 1, GFP_KERNEL);
762 763
	if (!mr->pas) {
		err = -ENOMEM;
764
		goto free_mr;
765
	}
766 767 768 769

	mlx5_ib_populate_pas(dev, umem, page_shift,
			     mr_align(mr->pas, MLX5_UMR_ALIGN), 1);

770
	mr->dma = dma_map_single(ddev, mr_align(mr->pas, MLX5_UMR_ALIGN), size,
771 772 773
				 DMA_TO_DEVICE);
	if (dma_mapping_error(ddev, mr->dma)) {
		err = -ENOMEM;
774
		goto free_pas;
775 776
	}

777
	memset(&wr, 0, sizeof(wr));
778
	wr.wr_id = (u64)(unsigned long)&umr_context;
779 780
	prep_umr_reg_wqe(pd, &wr, &sg, mr->dma, npages, mr->mmr.key, page_shift, virt_addr, len, access_flags);

781
	mlx5_ib_init_umr_context(&umr_context);
782 783 784 785
	down(&umrc->sem);
	err = ib_post_send(umrc->qp, &wr, &bad);
	if (err) {
		mlx5_ib_warn(dev, "post send failed, err %d\n", err);
786
		goto unmap_dma;
787 788 789 790 791 792
	} else {
		wait_for_completion(&umr_context.done);
		if (umr_context.status != IB_WC_SUCCESS) {
			mlx5_ib_warn(dev, "reg umr failed\n");
			err = -EFAULT;
		}
793
	}
794

795 796 797 798
	mr->mmr.iova = virt_addr;
	mr->mmr.size = len;
	mr->mmr.pd = to_mpd(pd)->pdn;

799 800
unmap_dma:
	up(&umrc->sem);
801
	dma_unmap_single(ddev, mr->dma, size, DMA_TO_DEVICE);
802 803

free_pas:
804 805
	kfree(mr->pas);

806 807 808 809
free_mr:
	if (err) {
		free_cached_mr(dev, mr);
		return ERR_PTR(err);
810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846
	}

	return mr;
}

static struct mlx5_ib_mr *reg_create(struct ib_pd *pd, u64 virt_addr,
				     u64 length, struct ib_umem *umem,
				     int npages, int page_shift,
				     int access_flags)
{
	struct mlx5_ib_dev *dev = to_mdev(pd->device);
	struct mlx5_create_mkey_mbox_in *in;
	struct mlx5_ib_mr *mr;
	int inlen;
	int err;

	mr = kzalloc(sizeof(*mr), GFP_KERNEL);
	if (!mr)
		return ERR_PTR(-ENOMEM);

	inlen = sizeof(*in) + sizeof(*in->pas) * ((npages + 1) / 2) * 2;
	in = mlx5_vzalloc(inlen);
	if (!in) {
		err = -ENOMEM;
		goto err_1;
	}
	mlx5_ib_populate_pas(dev, umem, page_shift, in->pas, 0);

	in->seg.flags = convert_access(access_flags) |
		MLX5_ACCESS_MODE_MTT;
	in->seg.flags_pd = cpu_to_be32(to_mpd(pd)->pdn);
	in->seg.start_addr = cpu_to_be64(virt_addr);
	in->seg.len = cpu_to_be64(length);
	in->seg.bsfs_octo_size = 0;
	in->seg.xlt_oct_size = cpu_to_be32(get_octo_len(virt_addr, length, 1 << page_shift));
	in->seg.log2_page_size = page_shift;
	in->seg.qpn_mkey7_0 = cpu_to_be32(0xffffff << 8);
E
Eli Cohen 已提交
847 848
	in->xlat_oct_act_size = cpu_to_be32(get_octo_len(virt_addr, length,
							 1 << page_shift));
849
	err = mlx5_core_create_mkey(dev->mdev, &mr->mmr, in, inlen, NULL,
E
Eli Cohen 已提交
850
				    NULL, NULL);
851 852 853 854 855
	if (err) {
		mlx5_ib_warn(dev, "create mkey failed\n");
		goto err_2;
	}
	mr->umem = umem;
A
Al Viro 已提交
856
	kvfree(in);
857 858 859 860 861 862

	mlx5_ib_dbg(dev, "mkey = 0x%x\n", mr->mmr.key);

	return mr;

err_2:
A
Al Viro 已提交
863
	kvfree(in);
864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883

err_1:
	kfree(mr);

	return ERR_PTR(err);
}

struct ib_mr *mlx5_ib_reg_user_mr(struct ib_pd *pd, u64 start, u64 length,
				  u64 virt_addr, int access_flags,
				  struct ib_udata *udata)
{
	struct mlx5_ib_dev *dev = to_mdev(pd->device);
	struct mlx5_ib_mr *mr = NULL;
	struct ib_umem *umem;
	int page_shift;
	int npages;
	int ncont;
	int order;
	int err;

884 885
	mlx5_ib_dbg(dev, "start 0x%llx, virt_addr 0x%llx, length 0x%llx, access_flags 0x%x\n",
		    start, virt_addr, length, access_flags);
886 887 888
	umem = ib_umem_get(pd->uobject->context, start, length, access_flags,
			   0);
	if (IS_ERR(umem)) {
889
		mlx5_ib_dbg(dev, "umem get failed (%ld)\n", PTR_ERR(umem));
890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925
		return (void *)umem;
	}

	mlx5_ib_cont_pages(umem, start, &npages, &page_shift, &ncont, &order);
	if (!npages) {
		mlx5_ib_warn(dev, "avoid zero region\n");
		err = -EINVAL;
		goto error;
	}

	mlx5_ib_dbg(dev, "npages %d, ncont %d, order %d, page_shift %d\n",
		    npages, ncont, order, page_shift);

	if (use_umr(order)) {
		mr = reg_umr(pd, umem, virt_addr, length, ncont, page_shift,
			     order, access_flags);
		if (PTR_ERR(mr) == -EAGAIN) {
			mlx5_ib_dbg(dev, "cache empty for order %d", order);
			mr = NULL;
		}
	}

	if (!mr)
		mr = reg_create(pd, virt_addr, length, umem, ncont, page_shift,
				access_flags);

	if (IS_ERR(mr)) {
		err = PTR_ERR(mr);
		goto error;
	}

	mlx5_ib_dbg(dev, "mkey 0x%x\n", mr->mmr.key);

	mr->umem = umem;
	mr->npages = npages;
	spin_lock(&dev->mr_lock);
926
	dev->mdev->priv.reg_pages += npages;
927 928 929 930 931 932 933 934 935 936 937 938 939 940
	spin_unlock(&dev->mr_lock);
	mr->ibmr.lkey = mr->mmr.key;
	mr->ibmr.rkey = mr->mmr.key;

	return &mr->ibmr;

error:
	ib_umem_release(umem);
	return ERR_PTR(err);
}

static int unreg_umr(struct mlx5_ib_dev *dev, struct mlx5_ib_mr *mr)
{
	struct umr_common *umrc = &dev->umrc;
941
	struct mlx5_ib_umr_context umr_context;
942 943 944 945
	struct ib_send_wr wr, *bad;
	int err;

	memset(&wr, 0, sizeof(wr));
946
	wr.wr_id = (u64)(unsigned long)&umr_context;
947 948
	prep_umr_unreg_wqe(dev, &wr, mr->mmr.key);

949
	mlx5_ib_init_umr_context(&umr_context);
950 951 952 953 954 955
	down(&umrc->sem);
	err = ib_post_send(umrc->qp, &wr, &bad);
	if (err) {
		up(&umrc->sem);
		mlx5_ib_dbg(dev, "err %d\n", err);
		goto error;
956 957 958
	} else {
		wait_for_completion(&umr_context.done);
		up(&umrc->sem);
959
	}
960
	if (umr_context.status != IB_WC_SUCCESS) {
961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980
		mlx5_ib_warn(dev, "unreg umr failed\n");
		err = -EFAULT;
		goto error;
	}
	return 0;

error:
	return err;
}

int mlx5_ib_dereg_mr(struct ib_mr *ibmr)
{
	struct mlx5_ib_dev *dev = to_mdev(ibmr->device);
	struct mlx5_ib_mr *mr = to_mmr(ibmr);
	struct ib_umem *umem = mr->umem;
	int npages = mr->npages;
	int umred = mr->umred;
	int err;

	if (!umred) {
981
		err = mlx5_core_destroy_mkey(dev->mdev, &mr->mmr);
982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998
		if (err) {
			mlx5_ib_warn(dev, "failed to destroy mkey 0x%x (%d)\n",
				     mr->mmr.key, err);
			return err;
		}
	} else {
		err = unreg_umr(dev, mr);
		if (err) {
			mlx5_ib_warn(dev, "failed unregister\n");
			return err;
		}
		free_cached_mr(dev, mr);
	}

	if (umem) {
		ib_umem_release(umem);
		spin_lock(&dev->mr_lock);
999
		dev->mdev->priv.reg_pages -= npages;
1000 1001 1002 1003 1004 1005 1006 1007 1008
		spin_unlock(&dev->mr_lock);
	}

	if (!umred)
		kfree(mr);

	return 0;
}

1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046
struct ib_mr *mlx5_ib_create_mr(struct ib_pd *pd,
				struct ib_mr_init_attr *mr_init_attr)
{
	struct mlx5_ib_dev *dev = to_mdev(pd->device);
	struct mlx5_create_mkey_mbox_in *in;
	struct mlx5_ib_mr *mr;
	int access_mode, err;
	int ndescs = roundup(mr_init_attr->max_reg_descriptors, 4);

	mr = kzalloc(sizeof(*mr), GFP_KERNEL);
	if (!mr)
		return ERR_PTR(-ENOMEM);

	in = kzalloc(sizeof(*in), GFP_KERNEL);
	if (!in) {
		err = -ENOMEM;
		goto err_free;
	}

	in->seg.status = 1 << 6; /* free */
	in->seg.xlt_oct_size = cpu_to_be32(ndescs);
	in->seg.qpn_mkey7_0 = cpu_to_be32(0xffffff << 8);
	in->seg.flags_pd = cpu_to_be32(to_mpd(pd)->pdn);
	access_mode = MLX5_ACCESS_MODE_MTT;

	if (mr_init_attr->flags & IB_MR_SIGNATURE_EN) {
		u32 psv_index[2];

		in->seg.flags_pd = cpu_to_be32(be32_to_cpu(in->seg.flags_pd) |
							   MLX5_MKEY_BSF_EN);
		in->seg.bsfs_octo_size = cpu_to_be32(MLX5_MKEY_BSF_OCTO_SIZE);
		mr->sig = kzalloc(sizeof(*mr->sig), GFP_KERNEL);
		if (!mr->sig) {
			err = -ENOMEM;
			goto err_free_in;
		}

		/* create mem & wire PSVs */
1047
		err = mlx5_core_create_psv(dev->mdev, to_mpd(pd)->pdn,
1048 1049 1050 1051 1052 1053 1054
					   2, psv_index);
		if (err)
			goto err_free_sig;

		access_mode = MLX5_ACCESS_MODE_KLM;
		mr->sig->psv_memory.psv_idx = psv_index[0];
		mr->sig->psv_wire.psv_idx = psv_index[1];
1055 1056 1057 1058 1059

		mr->sig->sig_status_checked = true;
		mr->sig->sig_err_exists = false;
		/* Next UMR, Arm SIGERR */
		++mr->sig->sigerr_count;
1060 1061 1062
	}

	in->seg.flags = MLX5_PERM_UMR_EN | access_mode;
1063
	err = mlx5_core_create_mkey(dev->mdev, &mr->mmr, in, sizeof(*in),
1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076
				    NULL, NULL, NULL);
	if (err)
		goto err_destroy_psv;

	mr->ibmr.lkey = mr->mmr.key;
	mr->ibmr.rkey = mr->mmr.key;
	mr->umem = NULL;
	kfree(in);

	return &mr->ibmr;

err_destroy_psv:
	if (mr->sig) {
1077
		if (mlx5_core_destroy_psv(dev->mdev,
1078 1079 1080
					  mr->sig->psv_memory.psv_idx))
			mlx5_ib_warn(dev, "failed to destroy mem psv %d\n",
				     mr->sig->psv_memory.psv_idx);
1081
		if (mlx5_core_destroy_psv(dev->mdev,
1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101
					  mr->sig->psv_wire.psv_idx))
			mlx5_ib_warn(dev, "failed to destroy wire psv %d\n",
				     mr->sig->psv_wire.psv_idx);
	}
err_free_sig:
	kfree(mr->sig);
err_free_in:
	kfree(in);
err_free:
	kfree(mr);
	return ERR_PTR(err);
}

int mlx5_ib_destroy_mr(struct ib_mr *ibmr)
{
	struct mlx5_ib_dev *dev = to_mdev(ibmr->device);
	struct mlx5_ib_mr *mr = to_mmr(ibmr);
	int err;

	if (mr->sig) {
1102
		if (mlx5_core_destroy_psv(dev->mdev,
1103 1104 1105
					  mr->sig->psv_memory.psv_idx))
			mlx5_ib_warn(dev, "failed to destroy mem psv %d\n",
				     mr->sig->psv_memory.psv_idx);
1106
		if (mlx5_core_destroy_psv(dev->mdev,
1107 1108 1109 1110 1111 1112
					  mr->sig->psv_wire.psv_idx))
			mlx5_ib_warn(dev, "failed to destroy wire psv %d\n",
				     mr->sig->psv_wire.psv_idx);
		kfree(mr->sig);
	}

1113
	err = mlx5_core_destroy_mkey(dev->mdev, &mr->mmr);
1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124
	if (err) {
		mlx5_ib_warn(dev, "failed to destroy mkey 0x%x (%d)\n",
			     mr->mmr.key, err);
		return err;
	}

	kfree(mr);

	return err;
}

1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151
struct ib_mr *mlx5_ib_alloc_fast_reg_mr(struct ib_pd *pd,
					int max_page_list_len)
{
	struct mlx5_ib_dev *dev = to_mdev(pd->device);
	struct mlx5_create_mkey_mbox_in *in;
	struct mlx5_ib_mr *mr;
	int err;

	mr = kzalloc(sizeof(*mr), GFP_KERNEL);
	if (!mr)
		return ERR_PTR(-ENOMEM);

	in = kzalloc(sizeof(*in), GFP_KERNEL);
	if (!in) {
		err = -ENOMEM;
		goto err_free;
	}

	in->seg.status = 1 << 6; /* free */
	in->seg.xlt_oct_size = cpu_to_be32((max_page_list_len + 1) / 2);
	in->seg.qpn_mkey7_0 = cpu_to_be32(0xffffff << 8);
	in->seg.flags = MLX5_PERM_UMR_EN | MLX5_ACCESS_MODE_MTT;
	in->seg.flags_pd = cpu_to_be32(to_mpd(pd)->pdn);
	/*
	 * TBD not needed - issue 197292 */
	in->seg.log2_page_size = PAGE_SHIFT;

1152
	err = mlx5_core_create_mkey(dev->mdev, &mr->mmr, in, sizeof(*in), NULL,
E
Eli Cohen 已提交
1153
				    NULL, NULL);
1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204
	kfree(in);
	if (err)
		goto err_free;

	mr->ibmr.lkey = mr->mmr.key;
	mr->ibmr.rkey = mr->mmr.key;
	mr->umem = NULL;

	return &mr->ibmr;

err_free:
	kfree(mr);
	return ERR_PTR(err);
}

struct ib_fast_reg_page_list *mlx5_ib_alloc_fast_reg_page_list(struct ib_device *ibdev,
							       int page_list_len)
{
	struct mlx5_ib_fast_reg_page_list *mfrpl;
	int size = page_list_len * sizeof(u64);

	mfrpl = kmalloc(sizeof(*mfrpl), GFP_KERNEL);
	if (!mfrpl)
		return ERR_PTR(-ENOMEM);

	mfrpl->ibfrpl.page_list = kmalloc(size, GFP_KERNEL);
	if (!mfrpl->ibfrpl.page_list)
		goto err_free;

	mfrpl->mapped_page_list = dma_alloc_coherent(ibdev->dma_device,
						     size, &mfrpl->map,
						     GFP_KERNEL);
	if (!mfrpl->mapped_page_list)
		goto err_free;

	WARN_ON(mfrpl->map & 0x3f);

	return &mfrpl->ibfrpl;

err_free:
	kfree(mfrpl->ibfrpl.page_list);
	kfree(mfrpl);
	return ERR_PTR(-ENOMEM);
}

void mlx5_ib_free_fast_reg_page_list(struct ib_fast_reg_page_list *page_list)
{
	struct mlx5_ib_fast_reg_page_list *mfrpl = to_mfrpl(page_list);
	struct mlx5_ib_dev *dev = to_mdev(page_list->device);
	int size = page_list->max_page_list_len * sizeof(u64);

1205
	dma_free_coherent(&dev->mdev->pdev->dev, size, mfrpl->mapped_page_list,
1206 1207 1208 1209
			  mfrpl->map);
	kfree(mfrpl->ibfrpl.page_list);
	kfree(mfrpl);
}
1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250

int mlx5_ib_check_mr_status(struct ib_mr *ibmr, u32 check_mask,
			    struct ib_mr_status *mr_status)
{
	struct mlx5_ib_mr *mmr = to_mmr(ibmr);
	int ret = 0;

	if (check_mask & ~IB_MR_CHECK_SIG_STATUS) {
		pr_err("Invalid status check mask\n");
		ret = -EINVAL;
		goto done;
	}

	mr_status->fail_status = 0;
	if (check_mask & IB_MR_CHECK_SIG_STATUS) {
		if (!mmr->sig) {
			ret = -EINVAL;
			pr_err("signature status check requested on a non-signature enabled MR\n");
			goto done;
		}

		mmr->sig->sig_status_checked = true;
		if (!mmr->sig->sig_err_exists)
			goto done;

		if (ibmr->lkey == mmr->sig->err_item.key)
			memcpy(&mr_status->sig_err, &mmr->sig->err_item,
			       sizeof(mr_status->sig_err));
		else {
			mr_status->sig_err.err_type = IB_SIG_BAD_GUARD;
			mr_status->sig_err.sig_err_offset = 0;
			mr_status->sig_err.key = mmr->sig->err_item.key;
		}

		mmr->sig->sig_err_exists = false;
		mr_status->fail_status |= IB_MR_CHECK_SIG_STATUS;
	}

done:
	return ret;
}