hns_roce_mr.c 34.8 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
/*
 * Copyright (c) 2016 Hisilicon Limited.
 * Copyright (c) 2007, 2008 Mellanox Technologies. 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/platform_device.h>
35
#include <linux/vmalloc.h>
36 37 38 39 40 41 42 43 44 45
#include <rdma/ib_umem.h>
#include "hns_roce_device.h"
#include "hns_roce_cmd.h"
#include "hns_roce_hem.h"

static u32 hw_index_to_key(unsigned long ind)
{
	return (u32)(ind >> 24) | (ind << 8);
}

S
Shaobo Xu 已提交
46
unsigned long key_to_hw_index(u32 key)
47 48 49
{
	return (key << 24) | (key >> 8);
}
50
EXPORT_SYMBOL_GPL(key_to_hw_index);
51 52 53 54 55 56 57

static int hns_roce_sw2hw_mpt(struct hns_roce_dev *hr_dev,
			      struct hns_roce_cmd_mailbox *mailbox,
			      unsigned long mpt_index)
{
	return hns_roce_cmd_mbox(hr_dev, mailbox->dma, 0, mpt_index, 0,
				 HNS_ROCE_CMD_SW2HW_MPT,
58
				 HNS_ROCE_CMD_TIMEOUT_MSECS);
59 60
}

S
Shaobo Xu 已提交
61
int hns_roce_hw2sw_mpt(struct hns_roce_dev *hr_dev,
62 63 64 65 66
			      struct hns_roce_cmd_mailbox *mailbox,
			      unsigned long mpt_index)
{
	return hns_roce_cmd_mbox(hr_dev, 0, mailbox ? mailbox->dma : 0,
				 mpt_index, !mailbox, HNS_ROCE_CMD_HW2SW_MPT,
67
				 HNS_ROCE_CMD_TIMEOUT_MSECS);
68
}
69
EXPORT_SYMBOL_GPL(hns_roce_hw2sw_mpt);
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

static int hns_roce_buddy_alloc(struct hns_roce_buddy *buddy, int order,
				unsigned long *seg)
{
	int o;
	u32 m;

	spin_lock(&buddy->lock);

	for (o = order; o <= buddy->max_order; ++o) {
		if (buddy->num_free[o]) {
			m = 1 << (buddy->max_order - o);
			*seg = find_first_bit(buddy->bits[o], m);
			if (*seg < m)
				goto found;
		}
	}
	spin_unlock(&buddy->lock);
	return -1;

 found:
	clear_bit(*seg, buddy->bits[o]);
	--buddy->num_free[o];

	while (o > order) {
		--o;
		*seg <<= 1;
		set_bit(*seg ^ 1, buddy->bits[o]);
		++buddy->num_free[o];
	}

	spin_unlock(&buddy->lock);

	*seg <<= order;
	return 0;
}

static void hns_roce_buddy_free(struct hns_roce_buddy *buddy, unsigned long seg,
				int order)
{
	seg >>= order;

	spin_lock(&buddy->lock);

	while (test_bit(seg ^ 1, buddy->bits[order])) {
		clear_bit(seg ^ 1, buddy->bits[order]);
		--buddy->num_free[order];
		seg >>= 1;
		++order;
	}

	set_bit(seg, buddy->bits[order]);
	++buddy->num_free[order];

	spin_unlock(&buddy->lock);
}

static int hns_roce_buddy_init(struct hns_roce_buddy *buddy, int max_order)
{
	int i, s;

	buddy->max_order = max_order;
	spin_lock_init(&buddy->lock);
133 134 135 136 137 138
	buddy->bits = kcalloc(buddy->max_order + 1,
			      sizeof(*buddy->bits),
			      GFP_KERNEL);
	buddy->num_free = kcalloc(buddy->max_order + 1,
				  sizeof(*buddy->num_free),
				  GFP_KERNEL);
139 140 141 142 143
	if (!buddy->bits || !buddy->num_free)
		goto err_out;

	for (i = 0; i <= buddy->max_order; ++i) {
		s = BITS_TO_LONGS(1 << (buddy->max_order - i));
144 145 146
		buddy->bits[i] = kcalloc(s, sizeof(long), GFP_KERNEL |
					 __GFP_NOWARN);
		if (!buddy->bits[i]) {
147
			buddy->bits[i] = vzalloc(array_size(s, sizeof(long)));
148 149 150
			if (!buddy->bits[i])
				goto err_out_free;
		}
151 152 153 154 155 156 157 158 159
	}

	set_bit(0, buddy->bits[buddy->max_order]);
	buddy->num_free[buddy->max_order] = 1;

	return 0;

err_out_free:
	for (i = 0; i <= buddy->max_order; ++i)
160
		kvfree(buddy->bits[i]);
161 162 163 164 165 166 167 168 169 170 171 172

err_out:
	kfree(buddy->bits);
	kfree(buddy->num_free);
	return -ENOMEM;
}

static void hns_roce_buddy_cleanup(struct hns_roce_buddy *buddy)
{
	int i;

	for (i = 0; i <= buddy->max_order; ++i)
173
		kvfree(buddy->bits[i]);
174 175 176 177 178 179

	kfree(buddy->bits);
	kfree(buddy->num_free);
}

static int hns_roce_alloc_mtt_range(struct hns_roce_dev *hr_dev, int order,
180
				    unsigned long *seg, u32 mtt_type)
181 182
{
	struct hns_roce_mr_table *mr_table = &hr_dev->mr_table;
183 184 185 186
	struct hns_roce_hem_table *table;
	struct hns_roce_buddy *buddy;
	int ret;

187 188
	switch (mtt_type) {
	case MTT_TYPE_WQE:
189 190
		buddy = &mr_table->mtt_buddy;
		table = &mr_table->mtt_table;
191 192
		break;
	case MTT_TYPE_CQE:
193 194
		buddy = &mr_table->mtt_cqe_buddy;
		table = &mr_table->mtt_cqe_table;
195 196 197 198 199 200 201 202 203 204 205 206 207
		break;
	case MTT_TYPE_SRQWQE:
		buddy = &mr_table->mtt_srqwqe_buddy;
		table = &mr_table->mtt_srqwqe_table;
		break;
	case MTT_TYPE_IDX:
		buddy = &mr_table->mtt_idx_buddy;
		table = &mr_table->mtt_idx_table;
		break;
	default:
		dev_err(hr_dev->dev, "Unsupport MTT table type: %d\n",
			mtt_type);
		return -EINVAL;
208
	}
209

210
	ret = hns_roce_buddy_alloc(buddy, order, seg);
211 212 213
	if (ret == -1)
		return -1;

214
	if (hns_roce_table_get_range(hr_dev, table, *seg,
215
				     *seg + (1 << order) - 1)) {
216
		hns_roce_buddy_free(buddy, *seg, order);
217 218 219 220 221 222 223 224 225
		return -1;
	}

	return 0;
}

int hns_roce_mtt_init(struct hns_roce_dev *hr_dev, int npages, int page_shift,
		      struct hns_roce_mtt *mtt)
{
226
	int ret;
227 228 229 230 231 232 233 234 235
	int i;

	/* Page num is zero, correspond to DMA memory register */
	if (!npages) {
		mtt->order = -1;
		mtt->page_shift = HNS_ROCE_HEM_PAGE_SHIFT;
		return 0;
	}

236
	/* Note: if page_shift is zero, FAST memory register */
237 238 239 240 241 242 243 244
	mtt->page_shift = page_shift;

	/* Compute MTT entry necessary */
	for (mtt->order = 0, i = HNS_ROCE_MTT_ENTRY_PER_SEG; i < npages;
	     i <<= 1)
		++mtt->order;

	/* Allocate MTT entry */
245 246
	ret = hns_roce_alloc_mtt_range(hr_dev, mtt->order, &mtt->first_seg,
				       mtt->mtt_type);
247 248 249 250 251 252 253 254 255 256 257 258 259
	if (ret == -1)
		return -ENOMEM;

	return 0;
}

void hns_roce_mtt_cleanup(struct hns_roce_dev *hr_dev, struct hns_roce_mtt *mtt)
{
	struct hns_roce_mr_table *mr_table = &hr_dev->mr_table;

	if (mtt->order < 0)
		return;

260 261
	switch (mtt->mtt_type) {
	case MTT_TYPE_WQE:
262 263 264 265 266
		hns_roce_buddy_free(&mr_table->mtt_buddy, mtt->first_seg,
				    mtt->order);
		hns_roce_table_put_range(hr_dev, &mr_table->mtt_table,
					mtt->first_seg,
					mtt->first_seg + (1 << mtt->order) - 1);
267 268
		break;
	case MTT_TYPE_CQE:
269 270 271 272 273
		hns_roce_buddy_free(&mr_table->mtt_cqe_buddy, mtt->first_seg,
				    mtt->order);
		hns_roce_table_put_range(hr_dev, &mr_table->mtt_cqe_table,
					mtt->first_seg,
					mtt->first_seg + (1 << mtt->order) - 1);
274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293
		break;
	case MTT_TYPE_SRQWQE:
		hns_roce_buddy_free(&mr_table->mtt_srqwqe_buddy, mtt->first_seg,
				    mtt->order);
		hns_roce_table_put_range(hr_dev, &mr_table->mtt_srqwqe_table,
					mtt->first_seg,
					mtt->first_seg + (1 << mtt->order) - 1);
		break;
	case MTT_TYPE_IDX:
		hns_roce_buddy_free(&mr_table->mtt_idx_buddy, mtt->first_seg,
				    mtt->order);
		hns_roce_table_put_range(hr_dev, &mr_table->mtt_idx_table,
					mtt->first_seg,
					mtt->first_seg + (1 << mtt->order) - 1);
		break;
	default:
		dev_err(hr_dev->dev,
			"Unsupport mtt type %d, clean mtt failed\n",
			mtt->mtt_type);
		break;
294
	}
295
}
296
EXPORT_SYMBOL_GPL(hns_roce_mtt_cleanup);
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
static void hns_roce_loop_free(struct hns_roce_dev *hr_dev,
			       struct hns_roce_mr *mr, int err_loop_index,
			       int loop_i, int loop_j)
{
	struct device *dev = hr_dev->dev;
	u32 mhop_num;
	u32 pbl_bt_sz;
	u64 bt_idx;
	int i, j;

	pbl_bt_sz = 1 << (hr_dev->caps.pbl_ba_pg_sz + PAGE_SHIFT);
	mhop_num = hr_dev->caps.pbl_hop_num;

	i = loop_i;
	if (mhop_num == 3 && err_loop_index == 2) {
		for (; i >= 0; i--) {
			dma_free_coherent(dev, pbl_bt_sz, mr->pbl_bt_l1[i],
					  mr->pbl_l1_dma_addr[i]);

			for (j = 0; j < pbl_bt_sz / 8; j++) {
				if (i == loop_i && j >= loop_j)
					break;

				bt_idx = i * pbl_bt_sz / 8 + j;
				dma_free_coherent(dev, pbl_bt_sz,
						  mr->pbl_bt_l2[bt_idx],
						  mr->pbl_l2_dma_addr[bt_idx]);
			}
		}
	} else if (mhop_num == 3 && err_loop_index == 1) {
		for (i -= 1; i >= 0; i--) {
			dma_free_coherent(dev, pbl_bt_sz, mr->pbl_bt_l1[i],
					  mr->pbl_l1_dma_addr[i]);

			for (j = 0; j < pbl_bt_sz / 8; j++) {
				bt_idx = i * pbl_bt_sz / 8 + j;
				dma_free_coherent(dev, pbl_bt_sz,
						  mr->pbl_bt_l2[bt_idx],
						  mr->pbl_l2_dma_addr[bt_idx]);
			}
		}
	} else if (mhop_num == 2 && err_loop_index == 1) {
		for (i -= 1; i >= 0; i--)
			dma_free_coherent(dev, pbl_bt_sz, mr->pbl_bt_l1[i],
					  mr->pbl_l1_dma_addr[i]);
	} else {
		dev_warn(dev, "not support: mhop_num=%d, err_loop_index=%d.",
			 mhop_num, err_loop_index);
		return;
	}

	dma_free_coherent(dev, pbl_bt_sz, mr->pbl_bt_l0, mr->pbl_l0_dma_addr);
	mr->pbl_bt_l0 = NULL;
	mr->pbl_l0_dma_addr = 0;
}

/* PBL multi hop addressing */
static int hns_roce_mhop_alloc(struct hns_roce_dev *hr_dev, int npages,
			       struct hns_roce_mr *mr)
{
	struct device *dev = hr_dev->dev;
	int mr_alloc_done = 0;
	int npages_allocated;
	int i = 0, j = 0;
	u32 pbl_bt_sz;
	u32 mhop_num;
	u64 pbl_last_bt_num;
	u64 pbl_bt_cnt = 0;
	u64 bt_idx;
	u64 size;

Y
Yixian Liu 已提交
369
	mhop_num = (mr->type == MR_TYPE_FRMR ? 1 : hr_dev->caps.pbl_hop_num);
370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390
	pbl_bt_sz = 1 << (hr_dev->caps.pbl_ba_pg_sz + PAGE_SHIFT);
	pbl_last_bt_num = (npages + pbl_bt_sz / 8 - 1) / (pbl_bt_sz / 8);

	if (mhop_num == HNS_ROCE_HOP_NUM_0)
		return 0;

	/* hop_num = 1 */
	if (mhop_num == 1) {
		if (npages > pbl_bt_sz / 8) {
			dev_err(dev, "npages %d is larger than buf_pg_sz!",
				npages);
			return -EINVAL;
		}
		mr->pbl_buf = dma_alloc_coherent(dev, npages * 8,
						 &(mr->pbl_dma_addr),
						 GFP_KERNEL);
		if (!mr->pbl_buf)
			return -ENOMEM;

		mr->pbl_size = npages;
		mr->pbl_ba = mr->pbl_dma_addr;
Y
Yixian Liu 已提交
391
		mr->pbl_hop_num = mhop_num;
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
		mr->pbl_ba_pg_sz = hr_dev->caps.pbl_ba_pg_sz;
		mr->pbl_buf_pg_sz = hr_dev->caps.pbl_buf_pg_sz;
		return 0;
	}

	mr->pbl_l1_dma_addr = kcalloc(pbl_bt_sz / 8,
				      sizeof(*mr->pbl_l1_dma_addr),
				      GFP_KERNEL);
	if (!mr->pbl_l1_dma_addr)
		return -ENOMEM;

	mr->pbl_bt_l1 = kcalloc(pbl_bt_sz / 8, sizeof(*mr->pbl_bt_l1),
				GFP_KERNEL);
	if (!mr->pbl_bt_l1)
		goto err_kcalloc_bt_l1;

	if (mhop_num == 3) {
		mr->pbl_l2_dma_addr = kcalloc(pbl_last_bt_num,
					      sizeof(*mr->pbl_l2_dma_addr),
					      GFP_KERNEL);
		if (!mr->pbl_l2_dma_addr)
			goto err_kcalloc_l2_dma;

		mr->pbl_bt_l2 = kcalloc(pbl_last_bt_num,
					sizeof(*mr->pbl_bt_l2),
					GFP_KERNEL);
		if (!mr->pbl_bt_l2)
			goto err_kcalloc_bt_l2;
	}

	/* alloc L0 BT */
	mr->pbl_bt_l0 = dma_alloc_coherent(dev, pbl_bt_sz,
					   &(mr->pbl_l0_dma_addr),
					   GFP_KERNEL);
	if (!mr->pbl_bt_l0)
		goto err_dma_alloc_l0;

	if (mhop_num == 2) {
		/* alloc L1 BT */
		for (i = 0; i < pbl_bt_sz / 8; i++) {
			if (pbl_bt_cnt + 1 < pbl_last_bt_num) {
				size = pbl_bt_sz;
			} else {
				npages_allocated = i * (pbl_bt_sz / 8);
				size = (npages - npages_allocated) * 8;
			}
			mr->pbl_bt_l1[i] = dma_alloc_coherent(dev, size,
						    &(mr->pbl_l1_dma_addr[i]),
						    GFP_KERNEL);
			if (!mr->pbl_bt_l1[i]) {
				hns_roce_loop_free(hr_dev, mr, 1, i, 0);
				goto err_dma_alloc_l0;
			}

			*(mr->pbl_bt_l0 + i) = mr->pbl_l1_dma_addr[i];

			pbl_bt_cnt++;
			if (pbl_bt_cnt >= pbl_last_bt_num)
				break;
		}
	} else if (mhop_num == 3) {
		/* alloc L1, L2 BT */
		for (i = 0; i < pbl_bt_sz / 8; i++) {
			mr->pbl_bt_l1[i] = dma_alloc_coherent(dev, pbl_bt_sz,
						    &(mr->pbl_l1_dma_addr[i]),
						    GFP_KERNEL);
			if (!mr->pbl_bt_l1[i]) {
				hns_roce_loop_free(hr_dev, mr, 1, i, 0);
				goto err_dma_alloc_l0;
			}

			*(mr->pbl_bt_l0 + i) = mr->pbl_l1_dma_addr[i];

			for (j = 0; j < pbl_bt_sz / 8; j++) {
				bt_idx = i * pbl_bt_sz / 8 + j;

				if (pbl_bt_cnt + 1 < pbl_last_bt_num) {
					size = pbl_bt_sz;
				} else {
					npages_allocated = bt_idx *
							   (pbl_bt_sz / 8);
					size = (npages - npages_allocated) * 8;
				}
				mr->pbl_bt_l2[bt_idx] = dma_alloc_coherent(
					      dev, size,
					      &(mr->pbl_l2_dma_addr[bt_idx]),
					      GFP_KERNEL);
				if (!mr->pbl_bt_l2[bt_idx]) {
					hns_roce_loop_free(hr_dev, mr, 2, i, j);
					goto err_dma_alloc_l0;
				}

				*(mr->pbl_bt_l1[i] + j) =
						mr->pbl_l2_dma_addr[bt_idx];

				pbl_bt_cnt++;
				if (pbl_bt_cnt >= pbl_last_bt_num) {
					mr_alloc_done = 1;
					break;
				}
			}

			if (mr_alloc_done)
				break;
		}
	}

	mr->l0_chunk_last_num = i + 1;
	if (mhop_num == 3)
		mr->l1_chunk_last_num = j + 1;

	mr->pbl_size = npages;
	mr->pbl_ba = mr->pbl_l0_dma_addr;
	mr->pbl_hop_num = hr_dev->caps.pbl_hop_num;
	mr->pbl_ba_pg_sz = hr_dev->caps.pbl_ba_pg_sz;
	mr->pbl_buf_pg_sz = hr_dev->caps.pbl_buf_pg_sz;

	return 0;

err_dma_alloc_l0:
	kfree(mr->pbl_bt_l2);
	mr->pbl_bt_l2 = NULL;

err_kcalloc_bt_l2:
	kfree(mr->pbl_l2_dma_addr);
	mr->pbl_l2_dma_addr = NULL;

err_kcalloc_l2_dma:
	kfree(mr->pbl_bt_l1);
	mr->pbl_bt_l1 = NULL;

err_kcalloc_bt_l1:
	kfree(mr->pbl_l1_dma_addr);
	mr->pbl_l1_dma_addr = NULL;

	return -ENOMEM;
}

530 531 532 533
static int hns_roce_mr_alloc(struct hns_roce_dev *hr_dev, u32 pd, u64 iova,
			     u64 size, u32 access, int npages,
			     struct hns_roce_mr *mr)
{
534
	struct device *dev = hr_dev->dev;
535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552
	unsigned long index = 0;
	int ret = 0;

	/* Allocate a key for mr from mr_table */
	ret = hns_roce_bitmap_alloc(&hr_dev->mr_table.mtpt_bitmap, &index);
	if (ret == -1)
		return -ENOMEM;

	mr->iova = iova;			/* MR va starting addr */
	mr->size = size;			/* MR addr range */
	mr->pd = pd;				/* MR num */
	mr->access = access;			/* MR access permit */
	mr->enabled = 0;			/* MR active status */
	mr->key = hw_index_to_key(index);	/* MR key */

	if (size == ~0ull) {
		mr->pbl_buf = NULL;
		mr->pbl_dma_addr = 0;
553 554 555 556 557 558 559
		/* PBL multi-hop addressing parameters */
		mr->pbl_bt_l2 = NULL;
		mr->pbl_bt_l1 = NULL;
		mr->pbl_bt_l0 = NULL;
		mr->pbl_l2_dma_addr = NULL;
		mr->pbl_l1_dma_addr = NULL;
		mr->pbl_l0_dma_addr = 0;
560
	} else {
561 562 563 564 565 566 567 568 569
		if (!hr_dev->caps.pbl_hop_num) {
			mr->pbl_buf = dma_alloc_coherent(dev, npages * 8,
							 &(mr->pbl_dma_addr),
							 GFP_KERNEL);
			if (!mr->pbl_buf)
				return -ENOMEM;
		} else {
			ret = hns_roce_mhop_alloc(hr_dev, npages, mr);
		}
570 571
	}

572 573 574 575 576 577 578 579 580 581 582 583 584 585
	return ret;
}

static void hns_roce_mhop_free(struct hns_roce_dev *hr_dev,
			       struct hns_roce_mr *mr)
{
	struct device *dev = hr_dev->dev;
	int npages_allocated;
	int npages;
	int i, j;
	u32 pbl_bt_sz;
	u32 mhop_num;
	u64 bt_idx;

Y
Yixian Liu 已提交
586
	npages = mr->pbl_size;
587
	pbl_bt_sz = 1 << (hr_dev->caps.pbl_ba_pg_sz + PAGE_SHIFT);
Y
Yixian Liu 已提交
588
	mhop_num = (mr->type == MR_TYPE_FRMR) ? 1 : hr_dev->caps.pbl_hop_num;
589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656

	if (mhop_num == HNS_ROCE_HOP_NUM_0)
		return;

	/* hop_num = 1 */
	if (mhop_num == 1) {
		dma_free_coherent(dev, (unsigned int)(npages * 8),
				  mr->pbl_buf, mr->pbl_dma_addr);
		return;
	}

	dma_free_coherent(dev, pbl_bt_sz, mr->pbl_bt_l0,
			  mr->pbl_l0_dma_addr);

	if (mhop_num == 2) {
		for (i = 0; i < mr->l0_chunk_last_num; i++) {
			if (i == mr->l0_chunk_last_num - 1) {
				npages_allocated = i * (pbl_bt_sz / 8);

				dma_free_coherent(dev,
					      (npages - npages_allocated) * 8,
					      mr->pbl_bt_l1[i],
					      mr->pbl_l1_dma_addr[i]);

				break;
			}

			dma_free_coherent(dev, pbl_bt_sz, mr->pbl_bt_l1[i],
					  mr->pbl_l1_dma_addr[i]);
		}
	} else if (mhop_num == 3) {
		for (i = 0; i < mr->l0_chunk_last_num; i++) {
			dma_free_coherent(dev, pbl_bt_sz, mr->pbl_bt_l1[i],
					  mr->pbl_l1_dma_addr[i]);

			for (j = 0; j < pbl_bt_sz / 8; j++) {
				bt_idx = i * (pbl_bt_sz / 8) + j;

				if ((i == mr->l0_chunk_last_num - 1)
				    && j == mr->l1_chunk_last_num - 1) {
					npages_allocated = bt_idx *
							   (pbl_bt_sz / 8);

					dma_free_coherent(dev,
					      (npages - npages_allocated) * 8,
					      mr->pbl_bt_l2[bt_idx],
					      mr->pbl_l2_dma_addr[bt_idx]);

					break;
				}

				dma_free_coherent(dev, pbl_bt_sz,
						mr->pbl_bt_l2[bt_idx],
						mr->pbl_l2_dma_addr[bt_idx]);
			}
		}
	}

	kfree(mr->pbl_bt_l1);
	kfree(mr->pbl_l1_dma_addr);
	mr->pbl_bt_l1 = NULL;
	mr->pbl_l1_dma_addr = NULL;
	if (mhop_num == 3) {
		kfree(mr->pbl_bt_l2);
		kfree(mr->pbl_l2_dma_addr);
		mr->pbl_bt_l2 = NULL;
		mr->pbl_l2_dma_addr = NULL;
	}
657 658 659 660 661
}

static void hns_roce_mr_free(struct hns_roce_dev *hr_dev,
			     struct hns_roce_mr *mr)
{
662
	struct device *dev = hr_dev->dev;
663 664 665 666 667 668 669 670 671 672 673
	int npages = 0;
	int ret;

	if (mr->enabled) {
		ret = hns_roce_hw2sw_mpt(hr_dev, NULL, key_to_hw_index(mr->key)
					 & (hr_dev->caps.num_mtpts - 1));
		if (ret)
			dev_warn(dev, "HW2SW_MPT failed (%d)\n", ret);
	}

	if (mr->size != ~0ULL) {
Y
Yixian Liu 已提交
674 675
		if (mr->type == MR_TYPE_MR)
			npages = ib_umem_page_count(mr->umem);
676 677 678 679 680 681

		if (!hr_dev->caps.pbl_hop_num)
			dma_free_coherent(dev, (unsigned int)(npages * 8),
					  mr->pbl_buf, mr->pbl_dma_addr);
		else
			hns_roce_mhop_free(hr_dev, mr);
682 683
	}

684 685 686 687
	if (mr->enabled)
		hns_roce_table_put(hr_dev, &hr_dev->mr_table.mtpt_table,
				   key_to_hw_index(mr->key));

688
	hns_roce_bitmap_free(&hr_dev->mr_table.mtpt_bitmap,
689
			     key_to_hw_index(mr->key), BITMAP_NO_RR);
690 691 692 693 694 695 696
}

static int hns_roce_mr_enable(struct hns_roce_dev *hr_dev,
			      struct hns_roce_mr *mr)
{
	int ret;
	unsigned long mtpt_idx = key_to_hw_index(mr->key);
697
	struct device *dev = hr_dev->dev;
698 699 700 701 702 703 704 705 706 707 708 709 710 711 712
	struct hns_roce_cmd_mailbox *mailbox;
	struct hns_roce_mr_table *mr_table = &hr_dev->mr_table;

	/* Prepare HEM entry memory */
	ret = hns_roce_table_get(hr_dev, &mr_table->mtpt_table, mtpt_idx);
	if (ret)
		return ret;

	/* Allocate mailbox memory */
	mailbox = hns_roce_alloc_cmd_mailbox(hr_dev);
	if (IS_ERR(mailbox)) {
		ret = PTR_ERR(mailbox);
		goto err_table;
	}

Y
Yixian Liu 已提交
713 714 715 716
	if (mr->type != MR_TYPE_FRMR)
		ret = hr_dev->hw->write_mtpt(mailbox->buf, mr, mtpt_idx);
	else
		ret = hr_dev->hw->frmr_write_mtpt(mailbox->buf, mr);
717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745
	if (ret) {
		dev_err(dev, "Write mtpt fail!\n");
		goto err_page;
	}

	ret = hns_roce_sw2hw_mpt(hr_dev, mailbox,
				 mtpt_idx & (hr_dev->caps.num_mtpts - 1));
	if (ret) {
		dev_err(dev, "SW2HW_MPT failed (%d)\n", ret);
		goto err_page;
	}

	mr->enabled = 1;
	hns_roce_free_cmd_mailbox(hr_dev, mailbox);

	return 0;

err_page:
	hns_roce_free_cmd_mailbox(hr_dev, mailbox);

err_table:
	hns_roce_table_put(hr_dev, &mr_table->mtpt_table, mtpt_idx);
	return ret;
}

static int hns_roce_write_mtt_chunk(struct hns_roce_dev *hr_dev,
				    struct hns_roce_mtt *mtt, u32 start_index,
				    u32 npages, u64 *page_list)
{
746
	struct hns_roce_hem_table *table;
747
	dma_addr_t dma_handle;
748
	__le64 *mtts;
749
	u32 s = start_index * sizeof(u64);
750
	u32 bt_page_size;
751
	u32 i;
752

753 754 755
	switch (mtt->mtt_type) {
	case MTT_TYPE_WQE:
		table = &hr_dev->mr_table.mtt_table;
756
		bt_page_size = 1 << (hr_dev->caps.mtt_ba_pg_sz + PAGE_SHIFT);
757 758 759
		break;
	case MTT_TYPE_CQE:
		table = &hr_dev->mr_table.mtt_cqe_table;
760
		bt_page_size = 1 << (hr_dev->caps.cqe_ba_pg_sz + PAGE_SHIFT);
761 762 763 764 765 766 767 768 769 770 771 772
		break;
	case MTT_TYPE_SRQWQE:
		table = &hr_dev->mr_table.mtt_srqwqe_table;
		bt_page_size = 1 << (hr_dev->caps.srqwqe_ba_pg_sz + PAGE_SHIFT);
		break;
	case MTT_TYPE_IDX:
		table = &hr_dev->mr_table.mtt_idx_table;
		bt_page_size = 1 << (hr_dev->caps.idx_ba_pg_sz + PAGE_SHIFT);
		break;
	default:
		return -EINVAL;
	}
773

774
	/* All MTTs must fit in the same page */
775 776
	if (start_index / (bt_page_size / sizeof(u64)) !=
		(start_index + npages - 1) / (bt_page_size / sizeof(u64)))
777 778 779 780 781
		return -EINVAL;

	if (start_index & (HNS_ROCE_MTT_ENTRY_PER_SEG - 1))
		return -EINVAL;

782
	mtts = hns_roce_table_find(hr_dev, table,
783 784 785 786 787 788
				mtt->first_seg + s / hr_dev->caps.mtt_entry_sz,
				&dma_handle);
	if (!mtts)
		return -ENOMEM;

	/* Save page addr, low 12 bits : 0 */
789 790 791 792 793 794
	for (i = 0; i < npages; ++i) {
		if (!hr_dev->caps.mtt_hop_num)
			mtts[i] = cpu_to_le64(page_list[i] >> PAGE_ADDR_SHIFT);
		else
			mtts[i] = cpu_to_le64(page_list[i]);
	}
795 796 797 798 799 800 801 802 803 804

	return 0;
}

static int hns_roce_write_mtt(struct hns_roce_dev *hr_dev,
			      struct hns_roce_mtt *mtt, u32 start_index,
			      u32 npages, u64 *page_list)
{
	int chunk;
	int ret;
805
	u32 bt_page_size;
806 807 808 809

	if (mtt->order < 0)
		return -EINVAL;

810 811
	switch (mtt->mtt_type) {
	case MTT_TYPE_WQE:
812
		bt_page_size = 1 << (hr_dev->caps.mtt_ba_pg_sz + PAGE_SHIFT);
813 814
		break;
	case MTT_TYPE_CQE:
815
		bt_page_size = 1 << (hr_dev->caps.cqe_ba_pg_sz + PAGE_SHIFT);
816 817 818 819 820 821 822 823 824 825 826 827 828
		break;
	case MTT_TYPE_SRQWQE:
		bt_page_size = 1 << (hr_dev->caps.srqwqe_ba_pg_sz + PAGE_SHIFT);
		break;
	case MTT_TYPE_IDX:
		bt_page_size = 1 << (hr_dev->caps.idx_ba_pg_sz + PAGE_SHIFT);
		break;
	default:
		dev_err(hr_dev->dev,
			"Unsupport mtt type %d, write mtt failed\n",
			mtt->mtt_type);
		return -EINVAL;
	}
829

830
	while (npages > 0) {
831
		chunk = min_t(int, bt_page_size / sizeof(u64), npages);
832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848

		ret = hns_roce_write_mtt_chunk(hr_dev, mtt, start_index, chunk,
					       page_list);
		if (ret)
			return ret;

		npages -= chunk;
		start_index += chunk;
		page_list += chunk;
	}

	return 0;
}

int hns_roce_buf_write_mtt(struct hns_roce_dev *hr_dev,
			   struct hns_roce_mtt *mtt, struct hns_roce_buf *buf)
{
849 850 851
	u64 *page_list;
	int ret;
	u32 i;
852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873

	page_list = kmalloc_array(buf->npages, sizeof(*page_list), GFP_KERNEL);
	if (!page_list)
		return -ENOMEM;

	for (i = 0; i < buf->npages; ++i) {
		if (buf->nbufs == 1)
			page_list[i] = buf->direct.map + (i << buf->page_shift);
		else
			page_list[i] = buf->page_list[i].map;

	}
	ret = hns_roce_write_mtt(hr_dev, mtt, 0, buf->npages, page_list);

	kfree(page_list);

	return ret;
}

int hns_roce_init_mr_table(struct hns_roce_dev *hr_dev)
{
	struct hns_roce_mr_table *mr_table = &hr_dev->mr_table;
874
	int ret;
875 876 877 878 879 880 881 882 883 884 885 886 887

	ret = hns_roce_bitmap_init(&mr_table->mtpt_bitmap,
				   hr_dev->caps.num_mtpts,
				   hr_dev->caps.num_mtpts - 1,
				   hr_dev->caps.reserved_mrws, 0);
	if (ret)
		return ret;

	ret = hns_roce_buddy_init(&mr_table->mtt_buddy,
				  ilog2(hr_dev->caps.num_mtt_segs));
	if (ret)
		goto err_buddy;

888 889 890 891 892 893
	if (hns_roce_check_whether_mhop(hr_dev, HEM_TYPE_CQE)) {
		ret = hns_roce_buddy_init(&mr_table->mtt_cqe_buddy,
					  ilog2(hr_dev->caps.num_cqe_segs));
		if (ret)
			goto err_buddy_cqe;
	}
894 895 896 897 898 899 900 901 902 903 904 905 906 907 908

	if (hr_dev->caps.num_srqwqe_segs) {
		ret = hns_roce_buddy_init(&mr_table->mtt_srqwqe_buddy,
					  ilog2(hr_dev->caps.num_srqwqe_segs));
		if (ret)
			goto err_buddy_srqwqe;
	}

	if (hr_dev->caps.num_idx_segs) {
		ret = hns_roce_buddy_init(&mr_table->mtt_idx_buddy,
					  ilog2(hr_dev->caps.num_idx_segs));
		if (ret)
			goto err_buddy_idx;
	}

909 910
	return 0;

911 912 913 914 915 916 917 918
err_buddy_idx:
	if (hr_dev->caps.num_srqwqe_segs)
		hns_roce_buddy_cleanup(&mr_table->mtt_srqwqe_buddy);

err_buddy_srqwqe:
	if (hns_roce_check_whether_mhop(hr_dev, HEM_TYPE_CQE))
		hns_roce_buddy_cleanup(&mr_table->mtt_cqe_buddy);

919 920 921
err_buddy_cqe:
	hns_roce_buddy_cleanup(&mr_table->mtt_buddy);

922 923 924 925 926 927 928 929 930
err_buddy:
	hns_roce_bitmap_cleanup(&mr_table->mtpt_bitmap);
	return ret;
}

void hns_roce_cleanup_mr_table(struct hns_roce_dev *hr_dev)
{
	struct hns_roce_mr_table *mr_table = &hr_dev->mr_table;

931 932 933 934
	if (hr_dev->caps.num_idx_segs)
		hns_roce_buddy_cleanup(&mr_table->mtt_idx_buddy);
	if (hr_dev->caps.num_srqwqe_segs)
		hns_roce_buddy_cleanup(&mr_table->mtt_srqwqe_buddy);
935
	hns_roce_buddy_cleanup(&mr_table->mtt_buddy);
936 937
	if (hns_roce_check_whether_mhop(hr_dev, HEM_TYPE_CQE))
		hns_roce_buddy_cleanup(&mr_table->mtt_cqe_buddy);
938 939 940 941 942
	hns_roce_bitmap_cleanup(&mr_table->mtpt_bitmap);
}

struct ib_mr *hns_roce_get_dma_mr(struct ib_pd *pd, int acc)
{
943 944
	struct hns_roce_mr *mr;
	int ret;
945 946 947 948 949

	mr = kmalloc(sizeof(*mr), GFP_KERNEL);
	if (mr == NULL)
		return  ERR_PTR(-ENOMEM);

Y
Yixian Liu 已提交
950 951
	mr->type = MR_TYPE_DMA;

952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977
	/* Allocate memory region key */
	ret = hns_roce_mr_alloc(to_hr_dev(pd->device), to_hr_pd(pd)->pdn, 0,
				~0ULL, acc, 0, mr);
	if (ret)
		goto err_free;

	ret = hns_roce_mr_enable(to_hr_dev(pd->device), mr);
	if (ret)
		goto err_mr;

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

	return &mr->ibmr;

err_mr:
	hns_roce_mr_free(to_hr_dev(pd->device), mr);

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

int hns_roce_ib_umem_write_mtt(struct hns_roce_dev *hr_dev,
			       struct hns_roce_mtt *mtt, struct ib_umem *umem)
{
978
	struct device *dev = hr_dev->dev;
979
	struct scatterlist *sg;
980
	unsigned int order;
981
	int i, k, entry;
982
	int npage = 0;
983
	int ret = 0;
984 985
	int len;
	u64 page_addr;
986
	u64 *pages;
987
	u32 bt_page_size;
988 989
	u32 n;

990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008
	switch (mtt->mtt_type) {
	case MTT_TYPE_WQE:
		order = hr_dev->caps.mtt_ba_pg_sz;
		break;
	case MTT_TYPE_CQE:
		order = hr_dev->caps.cqe_ba_pg_sz;
		break;
	case MTT_TYPE_SRQWQE:
		order = hr_dev->caps.srqwqe_ba_pg_sz;
		break;
	case MTT_TYPE_IDX:
		order = hr_dev->caps.idx_ba_pg_sz;
		break;
	default:
		dev_err(dev, "Unsupport mtt type %d, write mtt failed\n",
			mtt->mtt_type);
		return -EINVAL;
	}

1009 1010 1011
	bt_page_size = 1 << (order + PAGE_SHIFT);

	pages = (u64 *) __get_free_pages(GFP_KERNEL, order);
1012 1013 1014 1015 1016 1017
	if (!pages)
		return -ENOMEM;

	i = n = 0;

	for_each_sg(umem->sg_head.sgl, sg, umem->nmap, entry) {
1018
		len = sg_dma_len(sg) >> PAGE_SHIFT;
1019
		for (k = 0; k < len; ++k) {
1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032
			page_addr =
				sg_dma_address(sg) + (k << umem->page_shift);
			if (!(npage % (1 << (mtt->page_shift - PAGE_SHIFT)))) {
				if (page_addr & ((1 << mtt->page_shift) - 1)) {
					dev_err(dev, "page_addr 0x%llx is not page_shift %d alignment!\n",
						page_addr, mtt->page_shift);
					ret = -EINVAL;
					goto out;
				}
				pages[i++] = page_addr;
			}
			npage++;
			if (i == bt_page_size / sizeof(u64)) {
1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046
				ret = hns_roce_write_mtt(hr_dev, mtt, n, i,
							 pages);
				if (ret)
					goto out;
				n += i;
				i = 0;
			}
		}
	}

	if (i)
		ret = hns_roce_write_mtt(hr_dev, mtt, n, i, pages);

out:
1047
	free_pages((unsigned long) pages, order);
1048 1049 1050
	return ret;
}

1051 1052
static int hns_roce_ib_umem_write_mr(struct hns_roce_dev *hr_dev,
				     struct hns_roce_mr *mr,
1053 1054 1055
				     struct ib_umem *umem)
{
	struct scatterlist *sg;
1056
	int i = 0, j = 0, k;
1057
	int entry;
1058 1059 1060
	int len;
	u64 page_addr;
	u32 pbl_bt_sz;
1061 1062 1063

	if (hr_dev->caps.pbl_hop_num == HNS_ROCE_HOP_NUM_0)
		return 0;
1064

1065
	pbl_bt_sz = 1 << (hr_dev->caps.pbl_ba_pg_sz + PAGE_SHIFT);
1066
	for_each_sg(umem->sg_head.sgl, sg, umem->nmap, entry) {
1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086
		len = sg_dma_len(sg) >> PAGE_SHIFT;
		for (k = 0; k < len; ++k) {
			page_addr = sg_dma_address(sg) +
				    (k << umem->page_shift);

			if (!hr_dev->caps.pbl_hop_num) {
				mr->pbl_buf[i++] = page_addr >> 12;
			} else if (hr_dev->caps.pbl_hop_num == 1) {
				mr->pbl_buf[i++] = page_addr;
			} else {
				if (hr_dev->caps.pbl_hop_num == 2)
					mr->pbl_bt_l1[i][j] = page_addr;
				else if (hr_dev->caps.pbl_hop_num == 3)
					mr->pbl_bt_l2[i][j] = page_addr;

				j++;
				if (j >= (pbl_bt_sz / 8)) {
					i++;
					j = 0;
				}
1087 1088
			}
		}
1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101
	}

	/* Memory barrier */
	mb();

	return 0;
}

struct ib_mr *hns_roce_reg_user_mr(struct ib_pd *pd, u64 start, u64 length,
				   u64 virt_addr, int access_flags,
				   struct ib_udata *udata)
{
	struct hns_roce_dev *hr_dev = to_hr_dev(pd->device);
1102
	struct device *dev = hr_dev->dev;
1103 1104 1105 1106 1107
	struct hns_roce_mr *mr;
	int bt_size;
	int ret;
	int n;
	int i;
1108 1109 1110 1111 1112

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

1113
	mr->umem = ib_umem_get(udata, start, length, access_flags, 0);
1114 1115 1116 1117 1118 1119 1120
	if (IS_ERR(mr->umem)) {
		ret = PTR_ERR(mr->umem);
		goto err_free;
	}

	n = ib_umem_page_count(mr->umem);

1121 1122 1123 1124 1125 1126 1127 1128 1129
	if (!hr_dev->caps.pbl_hop_num) {
		if (n > HNS_ROCE_MAX_MTPT_PBL_NUM) {
			dev_err(dev,
			     " MR len %lld err. MR is limited to 4G at most!\n",
			     length);
			ret = -EINVAL;
			goto err_umem;
		}
	} else {
1130
		u64 pbl_size = 1;
1131

1132
		bt_size = (1 << (hr_dev->caps.pbl_ba_pg_sz + PAGE_SHIFT)) / 8;
1133 1134 1135 1136
		for (i = 0; i < hr_dev->caps.pbl_hop_num; i++)
			pbl_size *= bt_size;
		if (n > pbl_size) {
			dev_err(dev,
1137
			    " MR len %lld err. MR page num is limited to %lld!\n",
1138 1139 1140 1141
			    length, pbl_size);
			ret = -EINVAL;
			goto err_umem;
		}
1142 1143
	}

Y
Yixian Liu 已提交
1144 1145
	mr->type = MR_TYPE_MR;

1146 1147 1148 1149 1150
	ret = hns_roce_mr_alloc(hr_dev, to_hr_pd(pd)->pdn, virt_addr, length,
				access_flags, n, mr);
	if (ret)
		goto err_umem;

1151
	ret = hns_roce_ib_umem_write_mr(hr_dev, mr, mr->umem);
1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173
	if (ret)
		goto err_mr;

	ret = hns_roce_mr_enable(hr_dev, mr);
	if (ret)
		goto err_mr;

	mr->ibmr.rkey = mr->ibmr.lkey = mr->key;

	return &mr->ibmr;

err_mr:
	hns_roce_mr_free(hr_dev, mr);

err_umem:
	ib_umem_release(mr->umem);

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

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 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221
int hns_roce_rereg_user_mr(struct ib_mr *ibmr, int flags, u64 start, u64 length,
			   u64 virt_addr, int mr_access_flags, struct ib_pd *pd,
			   struct ib_udata *udata)
{
	struct hns_roce_dev *hr_dev = to_hr_dev(ibmr->device);
	struct hns_roce_mr *mr = to_hr_mr(ibmr);
	struct hns_roce_cmd_mailbox *mailbox;
	struct device *dev = hr_dev->dev;
	unsigned long mtpt_idx;
	u32 pdn = 0;
	int npages;
	int ret;

	if (!mr->enabled)
		return -EINVAL;

	mailbox = hns_roce_alloc_cmd_mailbox(hr_dev);
	if (IS_ERR(mailbox))
		return PTR_ERR(mailbox);

	mtpt_idx = key_to_hw_index(mr->key) & (hr_dev->caps.num_mtpts - 1);
	ret = hns_roce_cmd_mbox(hr_dev, 0, mailbox->dma, mtpt_idx, 0,
				HNS_ROCE_CMD_QUERY_MPT,
				HNS_ROCE_CMD_TIMEOUT_MSECS);
	if (ret)
		goto free_cmd_mbox;

	ret = hns_roce_hw2sw_mpt(hr_dev, NULL, mtpt_idx);
	if (ret)
		dev_warn(dev, "HW2SW_MPT failed (%d)\n", ret);

	mr->enabled = 0;

	if (flags & IB_MR_REREG_PD)
		pdn = to_hr_pd(pd)->pdn;

	if (flags & IB_MR_REREG_TRANS) {
		if (mr->size != ~0ULL) {
			npages = ib_umem_page_count(mr->umem);

			if (hr_dev->caps.pbl_hop_num)
				hns_roce_mhop_free(hr_dev, mr);
			else
				dma_free_coherent(dev, npages * 8, mr->pbl_buf,
						  mr->pbl_dma_addr);
		}
		ib_umem_release(mr->umem);

1222 1223
		mr->umem =
			ib_umem_get(udata, start, length, mr_access_flags, 0);
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 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296
		if (IS_ERR(mr->umem)) {
			ret = PTR_ERR(mr->umem);
			mr->umem = NULL;
			goto free_cmd_mbox;
		}
		npages = ib_umem_page_count(mr->umem);

		if (hr_dev->caps.pbl_hop_num) {
			ret = hns_roce_mhop_alloc(hr_dev, npages, mr);
			if (ret)
				goto release_umem;
		} else {
			mr->pbl_buf = dma_alloc_coherent(dev, npages * 8,
							 &(mr->pbl_dma_addr),
							 GFP_KERNEL);
			if (!mr->pbl_buf) {
				ret = -ENOMEM;
				goto release_umem;
			}
		}
	}

	ret = hr_dev->hw->rereg_write_mtpt(hr_dev, mr, flags, pdn,
					   mr_access_flags, virt_addr,
					   length, mailbox->buf);
	if (ret) {
		if (flags & IB_MR_REREG_TRANS)
			goto release_umem;
		else
			goto free_cmd_mbox;
	}

	if (flags & IB_MR_REREG_TRANS) {
		ret = hns_roce_ib_umem_write_mr(hr_dev, mr, mr->umem);
		if (ret) {
			if (mr->size != ~0ULL) {
				npages = ib_umem_page_count(mr->umem);

				if (hr_dev->caps.pbl_hop_num)
					hns_roce_mhop_free(hr_dev, mr);
				else
					dma_free_coherent(dev, npages * 8,
							  mr->pbl_buf,
							  mr->pbl_dma_addr);
			}

			goto release_umem;
		}
	}

	ret = hns_roce_sw2hw_mpt(hr_dev, mailbox, mtpt_idx);
	if (ret) {
		dev_err(dev, "SW2HW_MPT failed (%d)\n", ret);
		goto release_umem;
	}

	mr->enabled = 1;
	if (flags & IB_MR_REREG_ACCESS)
		mr->access = mr_access_flags;

	hns_roce_free_cmd_mailbox(hr_dev, mailbox);

	return 0;

release_umem:
	ib_umem_release(mr->umem);

free_cmd_mbox:
	hns_roce_free_cmd_mailbox(hr_dev, mailbox);

	return ret;
}

1297 1298
int hns_roce_dereg_mr(struct ib_mr *ibmr)
{
S
Shaobo Xu 已提交
1299
	struct hns_roce_dev *hr_dev = to_hr_dev(ibmr->device);
1300
	struct hns_roce_mr *mr = to_hr_mr(ibmr);
S
Shaobo Xu 已提交
1301
	int ret = 0;
1302

S
Shaobo Xu 已提交
1303 1304 1305 1306
	if (hr_dev->hw->dereg_mr) {
		ret = hr_dev->hw->dereg_mr(hr_dev, mr);
	} else {
		hns_roce_mr_free(hr_dev, mr);
1307

S
Shaobo Xu 已提交
1308 1309
		if (mr->umem)
			ib_umem_release(mr->umem);
1310

S
Shaobo Xu 已提交
1311 1312 1313 1314
		kfree(mr);
	}

	return ret;
1315
}
Y
Yixian Liu 已提交
1316

Y
Yixian Liu 已提交
1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357 1358 1359 1360 1361 1362 1363 1364 1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386
struct ib_mr *hns_roce_alloc_mr(struct ib_pd *pd, enum ib_mr_type mr_type,
				u32 max_num_sg)
{
	struct hns_roce_dev *hr_dev = to_hr_dev(pd->device);
	struct device *dev = hr_dev->dev;
	struct hns_roce_mr *mr;
	u64 length;
	u32 page_size;
	int ret;

	page_size = 1 << (hr_dev->caps.pbl_buf_pg_sz + PAGE_SHIFT);
	length = max_num_sg * page_size;

	if (mr_type != IB_MR_TYPE_MEM_REG)
		return ERR_PTR(-EINVAL);

	if (max_num_sg > HNS_ROCE_FRMR_MAX_PA) {
		dev_err(dev, "max_num_sg larger than %d\n",
			HNS_ROCE_FRMR_MAX_PA);
		return ERR_PTR(-EINVAL);
	}

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

	mr->type = MR_TYPE_FRMR;

	/* Allocate memory region key */
	ret = hns_roce_mr_alloc(hr_dev, to_hr_pd(pd)->pdn, 0, length,
				0, max_num_sg, mr);
	if (ret)
		goto err_free;

	ret = hns_roce_mr_enable(hr_dev, mr);
	if (ret)
		goto err_mr;

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

	return &mr->ibmr;

err_mr:
	hns_roce_mr_free(to_hr_dev(pd->device), mr);

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

static int hns_roce_set_page(struct ib_mr *ibmr, u64 addr)
{
	struct hns_roce_mr *mr = to_hr_mr(ibmr);

	mr->pbl_buf[mr->npages++] = cpu_to_le64(addr);

	return 0;
}

int hns_roce_map_mr_sg(struct ib_mr *ibmr, struct scatterlist *sg, int sg_nents,
		       unsigned int *sg_offset)
{
	struct hns_roce_mr *mr = to_hr_mr(ibmr);

	mr->npages = 0;

	return ib_sg_to_pages(ibmr, sg, sg_nents, sg_offset, hns_roce_set_page);
}

Y
Yixian Liu 已提交
1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398 1399 1400 1401 1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463 1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505
static void hns_roce_mw_free(struct hns_roce_dev *hr_dev,
			     struct hns_roce_mw *mw)
{
	struct device *dev = hr_dev->dev;
	int ret;

	if (mw->enabled) {
		ret = hns_roce_hw2sw_mpt(hr_dev, NULL, key_to_hw_index(mw->rkey)
					 & (hr_dev->caps.num_mtpts - 1));
		if (ret)
			dev_warn(dev, "MW HW2SW_MPT failed (%d)\n", ret);

		hns_roce_table_put(hr_dev, &hr_dev->mr_table.mtpt_table,
				   key_to_hw_index(mw->rkey));
	}

	hns_roce_bitmap_free(&hr_dev->mr_table.mtpt_bitmap,
			     key_to_hw_index(mw->rkey), BITMAP_NO_RR);
}

static int hns_roce_mw_enable(struct hns_roce_dev *hr_dev,
			      struct hns_roce_mw *mw)
{
	struct hns_roce_mr_table *mr_table = &hr_dev->mr_table;
	struct hns_roce_cmd_mailbox *mailbox;
	struct device *dev = hr_dev->dev;
	unsigned long mtpt_idx = key_to_hw_index(mw->rkey);
	int ret;

	/* prepare HEM entry memory */
	ret = hns_roce_table_get(hr_dev, &mr_table->mtpt_table, mtpt_idx);
	if (ret)
		return ret;

	mailbox = hns_roce_alloc_cmd_mailbox(hr_dev);
	if (IS_ERR(mailbox)) {
		ret = PTR_ERR(mailbox);
		goto err_table;
	}

	ret = hr_dev->hw->mw_write_mtpt(mailbox->buf, mw);
	if (ret) {
		dev_err(dev, "MW write mtpt fail!\n");
		goto err_page;
	}

	ret = hns_roce_sw2hw_mpt(hr_dev, mailbox,
				 mtpt_idx & (hr_dev->caps.num_mtpts - 1));
	if (ret) {
		dev_err(dev, "MW sw2hw_mpt failed (%d)\n", ret);
		goto err_page;
	}

	mw->enabled = 1;

	hns_roce_free_cmd_mailbox(hr_dev, mailbox);

	return 0;

err_page:
	hns_roce_free_cmd_mailbox(hr_dev, mailbox);

err_table:
	hns_roce_table_put(hr_dev, &mr_table->mtpt_table, mtpt_idx);

	return ret;
}

struct ib_mw *hns_roce_alloc_mw(struct ib_pd *ib_pd, enum ib_mw_type type,
				struct ib_udata *udata)
{
	struct hns_roce_dev *hr_dev = to_hr_dev(ib_pd->device);
	struct hns_roce_mw *mw;
	unsigned long index = 0;
	int ret;

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

	/* Allocate a key for mw from bitmap */
	ret = hns_roce_bitmap_alloc(&hr_dev->mr_table.mtpt_bitmap, &index);
	if (ret)
		goto err_bitmap;

	mw->rkey = hw_index_to_key(index);

	mw->ibmw.rkey = mw->rkey;
	mw->ibmw.type = type;
	mw->pdn = to_hr_pd(ib_pd)->pdn;
	mw->pbl_hop_num = hr_dev->caps.pbl_hop_num;
	mw->pbl_ba_pg_sz = hr_dev->caps.pbl_ba_pg_sz;
	mw->pbl_buf_pg_sz = hr_dev->caps.pbl_buf_pg_sz;

	ret = hns_roce_mw_enable(hr_dev, mw);
	if (ret)
		goto err_mw;

	return &mw->ibmw;

err_mw:
	hns_roce_mw_free(hr_dev, mw);

err_bitmap:
	kfree(mw);

	return ERR_PTR(ret);
}

int hns_roce_dealloc_mw(struct ib_mw *ibmw)
{
	struct hns_roce_dev *hr_dev = to_hr_dev(ibmw->device);
	struct hns_roce_mw *mw = to_hr_mw(ibmw);

	hns_roce_mw_free(hr_dev, mw);
	kfree(mw);

	return 0;
}