iser_memory.c 14.3 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) 2004, 2005, 2006 Voltaire, 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.
 *
 * $Id: iser_memory.c 6964 2006-05-07 11:11:43Z ogerlitz $
 */
#include <linux/module.h>
#include <linux/kernel.h>
#include <linux/slab.h>
#include <linux/mm.h>
A
Al Viro 已提交
38
#include <linux/highmem.h>
39 40 41 42 43 44 45
#include <asm/io.h>
#include <asm/scatterlist.h>
#include <linux/scatterlist.h>

#include "iscsi_iser.h"

#define ISER_KMALLOC_THRESHOLD 0x20000 /* 128K - kmalloc limit */
46

47 48 49 50 51 52 53 54 55 56 57 58 59
/**
 * Decrements the reference count for the
 * registered buffer & releases it
 *
 * returns 0 if released, 1 if deferred
 */
int iser_regd_buff_release(struct iser_regd_buf *regd_buf)
{
	struct device *dma_device;

	if ((atomic_read(&regd_buf->ref_count) == 0) ||
	    atomic_dec_and_test(&regd_buf->ref_count)) {
		/* if we used the dma mr, unreg is just NOP */
60
		if (regd_buf->reg.is_fmr)
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
			iser_unreg_mem(&regd_buf->reg);

		if (regd_buf->dma_addr) {
			dma_device = regd_buf->device->ib_device->dma_device;
			dma_unmap_single(dma_device,
					 regd_buf->dma_addr,
					 regd_buf->data_size,
					 regd_buf->direction);
		}
		/* else this regd buf is associated with task which we */
		/* dma_unmap_single/sg later */
		return 0;
	} else {
		iser_dbg("Release deferred, regd.buff: 0x%p\n", regd_buf);
		return 1;
	}
}

/**
 * iser_reg_single - fills registered buffer descriptor with
 *		     registration information
 */
void iser_reg_single(struct iser_device *device,
		     struct iser_regd_buf *regd_buf,
		     enum dma_data_direction direction)
{
	dma_addr_t dma_addr;

	dma_addr  = dma_map_single(device->ib_device->dma_device,
				   regd_buf->virt_addr,
				   regd_buf->data_size, direction);
	BUG_ON(dma_mapping_error(dma_addr));

	regd_buf->reg.lkey = device->mr->lkey;
	regd_buf->reg.len  = regd_buf->data_size;
	regd_buf->reg.va   = dma_addr;
97
	regd_buf->reg.is_fmr = 0;
98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116

	regd_buf->dma_addr  = dma_addr;
	regd_buf->direction = direction;
}

/**
 * iser_start_rdma_unaligned_sg
 */
int iser_start_rdma_unaligned_sg(struct iscsi_iser_cmd_task  *iser_ctask,
				 enum iser_data_dir cmd_dir)
{
	int dma_nents;
	struct device *dma_device;
	char *mem = NULL;
	struct iser_data_buf *data = &iser_ctask->data[cmd_dir];
	unsigned long  cmd_data_len = data->data_len;

	if (cmd_data_len > ISER_KMALLOC_THRESHOLD)
		mem = (void *)__get_free_pages(GFP_NOIO,
117
		      ilog2(roundup_pow_of_two(cmd_data_len)) - PAGE_SHIFT);
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
	else
		mem = kmalloc(cmd_data_len, GFP_NOIO);

	if (mem == NULL) {
		iser_err("Failed to allocate mem size %d %d for copying sglist\n",
			 data->size,(int)cmd_data_len);
		return -ENOMEM;
	}

	if (cmd_dir == ISER_DIR_OUT) {
		/* copy the unaligned sg the buffer which is used for RDMA */
		struct scatterlist *sg = (struct scatterlist *)data->buf;
		int i;
		char *p, *from;

		for (p = mem, i = 0; i < data->size; i++) {
			from = kmap_atomic(sg[i].page, KM_USER0);
			memcpy(p,
			       from + sg[i].offset,
			       sg[i].length);
			kunmap_atomic(from, KM_USER0);
			p += sg[i].length;
		}
	}

	sg_init_one(&iser_ctask->data_copy[cmd_dir].sg_single, mem, cmd_data_len);
	iser_ctask->data_copy[cmd_dir].buf  =
		&iser_ctask->data_copy[cmd_dir].sg_single;
	iser_ctask->data_copy[cmd_dir].size = 1;

	iser_ctask->data_copy[cmd_dir].copy_buf  = mem;

	dma_device = iser_ctask->iser_conn->ib_conn->device->ib_device->dma_device;

	if (cmd_dir == ISER_DIR_OUT)
		dma_nents = dma_map_sg(dma_device,
				       &iser_ctask->data_copy[cmd_dir].sg_single,
				       1, DMA_TO_DEVICE);
	else
		dma_nents = dma_map_sg(dma_device,
				       &iser_ctask->data_copy[cmd_dir].sg_single,
				       1, DMA_FROM_DEVICE);

	BUG_ON(dma_nents == 0);

	iser_ctask->data_copy[cmd_dir].dma_nents = dma_nents;
	return 0;
}

/**
 * iser_finalize_rdma_unaligned_sg
 */
void iser_finalize_rdma_unaligned_sg(struct iscsi_iser_cmd_task *iser_ctask,
				     enum iser_data_dir         cmd_dir)
{
	struct device *dma_device;
	struct iser_data_buf *mem_copy;
	unsigned long  cmd_data_len;

	dma_device = iser_ctask->iser_conn->ib_conn->device->ib_device->dma_device;
	mem_copy   = &iser_ctask->data_copy[cmd_dir];

	if (cmd_dir == ISER_DIR_OUT)
		dma_unmap_sg(dma_device, &mem_copy->sg_single, 1,
			     DMA_TO_DEVICE);
	else
		dma_unmap_sg(dma_device, &mem_copy->sg_single, 1,
			     DMA_FROM_DEVICE);

	if (cmd_dir == ISER_DIR_IN) {
		char *mem;
		struct scatterlist *sg;
		unsigned char *p, *to;
		unsigned int sg_size;
		int i;

		/* copy back read RDMA to unaligned sg */
		mem	= mem_copy->copy_buf;

		sg	= (struct scatterlist *)iser_ctask->data[ISER_DIR_IN].buf;
		sg_size = iser_ctask->data[ISER_DIR_IN].size;

		for (p = mem, i = 0; i < sg_size; i++){
			to = kmap_atomic(sg[i].page, KM_SOFTIRQ0);
			memcpy(to + sg[i].offset,
			       p,
			       sg[i].length);
			kunmap_atomic(to, KM_SOFTIRQ0);
			p += sg[i].length;
		}
	}

	cmd_data_len = iser_ctask->data[cmd_dir].data_len;

	if (cmd_data_len > ISER_KMALLOC_THRESHOLD)
		free_pages((unsigned long)mem_copy->copy_buf,
214
			   ilog2(roundup_pow_of_two(cmd_data_len)) - PAGE_SHIFT);
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
	else
		kfree(mem_copy->copy_buf);

	mem_copy->copy_buf = NULL;
}

/**
 * iser_sg_to_page_vec - Translates scatterlist entries to physical addresses
 * and returns the length of resulting physical address array (may be less than
 * the original due to possible compaction).
 *
 * we build a "page vec" under the assumption that the SG meets the RDMA
 * alignment requirements. Other then the first and last SG elements, all
 * the "internal" elements can be compacted into a list whose elements are
 * dma addresses of physical pages. The code supports also the weird case
 * where --few fragments of the same page-- are present in the SG as
 * consecutive elements. Also, it handles one entry SG.
 */
static int iser_sg_to_page_vec(struct iser_data_buf *data,
			       struct iser_page_vec *page_vec)
{
	struct scatterlist *sg = (struct scatterlist *)data->buf;
	dma_addr_t first_addr, last_addr, page;
	int start_aligned, end_aligned;
	unsigned int cur_page = 0;
	unsigned long total_sz = 0;
	int i;

	/* compute the offset of first element */
244
	page_vec->offset = (u64) sg[0].offset & ~MASK_4K;
245 246 247 248 249 250 251

	for (i = 0; i < data->dma_nents; i++) {
		total_sz += sg_dma_len(&sg[i]);

		first_addr = sg_dma_address(&sg[i]);
		last_addr  = first_addr + sg_dma_len(&sg[i]);

252 253
		start_aligned = !(first_addr & ~MASK_4K);
		end_aligned   = !(last_addr  & ~MASK_4K);
254 255 256 257 258 259

		/* continue to collect page fragments till aligned or SG ends */
		while (!end_aligned && (i + 1 < data->dma_nents)) {
			i++;
			total_sz += sg_dma_len(&sg[i]);
			last_addr = sg_dma_address(&sg[i]) + sg_dma_len(&sg[i]);
260
			end_aligned = !(last_addr  & ~MASK_4K);
261 262
		}

263 264 265 266 267 268 269 270 271 272 273 274 275
		/* handle the 1st page in the 1st DMA element */
		if (cur_page == 0) {
			page = first_addr & MASK_4K;
			page_vec->pages[cur_page] = page;
			cur_page++;
			page += SIZE_4K;
		} else
			page = first_addr;

		for (; page < last_addr; page += SIZE_4K) {
			page_vec->pages[cur_page] = page;
			cur_page++;
		}
276 277 278 279 280 281 282

	}
	page_vec->data_size = total_sz;
	iser_dbg("page_vec->data_size:%d cur_page %d\n", page_vec->data_size,cur_page);
	return cur_page;
}

283
#define IS_4K_ALIGNED(addr)	((((unsigned long)addr) & ~MASK_4K) == 0)
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

/**
 * iser_data_buf_aligned_len - Tries to determine the maximal correctly aligned
 * for RDMA sub-list of a scatter-gather list of memory buffers, and  returns
 * the number of entries which are aligned correctly. Supports the case where
 * consecutive SG elements are actually fragments of the same physcial page.
 */
static unsigned int iser_data_buf_aligned_len(struct iser_data_buf *data)
{
	struct scatterlist *sg;
	dma_addr_t end_addr, next_addr;
	int i, cnt;
	unsigned int ret_len = 0;

	sg = (struct scatterlist *)data->buf;

	for (cnt = 0, i = 0; i < data->dma_nents; i++, cnt++) {
		/* iser_dbg("Checking sg iobuf [%d]: phys=0x%08lX "
		   "offset: %ld sz: %ld\n", i,
		   (unsigned long)page_to_phys(sg[i].page),
		   (unsigned long)sg[i].offset,
		   (unsigned long)sg[i].length); */
		end_addr = sg_dma_address(&sg[i]) +
			   sg_dma_len(&sg[i]);
		/* iser_dbg("Checking sg iobuf end address "
		       "0x%08lX\n", end_addr); */
		if (i + 1 < data->dma_nents) {
			next_addr = sg_dma_address(&sg[i+1]);
			/* are i, i+1 fragments of the same page? */
			if (end_addr == next_addr)
				continue;
			else if (!IS_4K_ALIGNED(end_addr)) {
				ret_len = cnt + 1;
				break;
			}
		}
	}
	if (i == data->dma_nents)
		ret_len = cnt;	/* loop ended */
	iser_dbg("Found %d aligned entries out of %d in sg:0x%p\n",
		 ret_len, data->dma_nents, data);
	return ret_len;
}

static void iser_data_buf_dump(struct iser_data_buf *data)
{
	struct scatterlist *sg = (struct scatterlist *)data->buf;
	int i;

E
Erez Zilber 已提交
333
	for (i = 0; i < data->dma_nents; i++)
334
		iser_err("sg[%d] dma_addr:0x%lX page:0x%p "
E
Erez Zilber 已提交
335
			 "off:0x%x sz:0x%x dma_len:0x%x\n",
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
			 i, (unsigned long)sg_dma_address(&sg[i]),
			 sg[i].page, sg[i].offset,
			 sg[i].length,sg_dma_len(&sg[i]));
}

static void iser_dump_page_vec(struct iser_page_vec *page_vec)
{
	int i;

	iser_err("page vec length %d data size %d\n",
		 page_vec->length, page_vec->data_size);
	for (i = 0; i < page_vec->length; i++)
		iser_err("%d %lx\n",i,(unsigned long)page_vec->pages[i]);
}

static void iser_page_vec_build(struct iser_data_buf *data,
				struct iser_page_vec *page_vec)
{
	int page_vec_len = 0;

	page_vec->length = 0;
	page_vec->offset = 0;

	iser_dbg("Translating sg sz: %d\n", data->dma_nents);
	page_vec_len = iser_sg_to_page_vec(data,page_vec);
	iser_dbg("sg len %d page_vec_len %d\n", data->dma_nents,page_vec_len);

	page_vec->length = page_vec_len;

365
	if (page_vec_len * SIZE_4K < page_vec->data_size) {
366 367 368 369 370 371 372
		iser_err("page_vec too short to hold this SG\n");
		iser_data_buf_dump(data);
		iser_dump_page_vec(page_vec);
		BUG();
	}
}

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
int iser_dma_map_task_data(struct iscsi_iser_cmd_task *iser_ctask,
			    struct iser_data_buf       *data,
			    enum   iser_data_dir       iser_dir,
			    enum   dma_data_direction  dma_dir)
{
	struct device *dma_device;

	iser_ctask->dir[iser_dir] = 1;
	dma_device =
		iser_ctask->iser_conn->ib_conn->device->ib_device->dma_device;

	data->dma_nents = dma_map_sg(dma_device, data->buf, data->size, dma_dir);
	if (data->dma_nents == 0) {
		iser_err("dma_map_sg failed!!!\n");
		return -EINVAL;
	}
	return 0;
}

void iser_dma_unmap_task_data(struct iscsi_iser_cmd_task *iser_ctask)
{
	struct device  *dma_device;
	struct iser_data_buf *data;

	dma_device =
		iser_ctask->iser_conn->ib_conn->device->ib_device->dma_device;

	if (iser_ctask->dir[ISER_DIR_IN]) {
		data = &iser_ctask->data[ISER_DIR_IN];
		dma_unmap_sg(dma_device, data->buf, data->size, DMA_FROM_DEVICE);
	}

	if (iser_ctask->dir[ISER_DIR_OUT]) {
		data = &iser_ctask->data[ISER_DIR_OUT];
		dma_unmap_sg(dma_device, data->buf, data->size, DMA_TO_DEVICE);
	}
}

411 412 413 414 415 416 417 418 419 420
/**
 * iser_reg_rdma_mem - Registers memory intended for RDMA,
 * obtaining rkey and va
 *
 * returns 0 on success, errno code on failure
 */
int iser_reg_rdma_mem(struct iscsi_iser_cmd_task *iser_ctask,
		      enum   iser_data_dir        cmd_dir)
{
	struct iser_conn     *ib_conn = iser_ctask->iser_conn->ib_conn;
421
	struct iser_device   *device = ib_conn->device;
422 423 424 425
	struct iser_data_buf *mem = &iser_ctask->data[cmd_dir];
	struct iser_regd_buf *regd_buf;
	int aligned_len;
	int err;
E
Erez Zilber 已提交
426
	int i;
427
	struct scatterlist *sg;
428 429 430 431

	regd_buf = &iser_ctask->rdma_regd[cmd_dir];

	aligned_len = iser_data_buf_aligned_len(mem);
432
	if (aligned_len != mem->dma_nents) {
433 434 435
		iser_err("rdma alignment violation %d/%d aligned\n",
			 aligned_len, mem->size);
		iser_data_buf_dump(mem);
436 437 438 439

		/* unmap the command data before accessing it */
		iser_dma_unmap_task_data(iser_ctask);

440 441 442 443 444 445 446
		/* allocate copy buf, if we are writing, copy the */
		/* unaligned scatterlist, dma map the copy        */
		if (iser_start_rdma_unaligned_sg(iser_ctask, cmd_dir) != 0)
				return -ENOMEM;
		mem = &iser_ctask->data_copy[cmd_dir];
	}

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
	/* if there a single dma entry, FMR is not needed */
	if (mem->dma_nents == 1) {
		sg = (struct scatterlist *)mem->buf;

		regd_buf->reg.lkey = device->mr->lkey;
		regd_buf->reg.rkey = device->mr->rkey;
		regd_buf->reg.len  = sg_dma_len(&sg[0]);
		regd_buf->reg.va   = sg_dma_address(&sg[0]);
		regd_buf->reg.is_fmr = 0;

		iser_dbg("PHYSICAL Mem.register: lkey: 0x%08X rkey: 0x%08X  "
			 "va: 0x%08lX sz: %ld]\n",
			 (unsigned int)regd_buf->reg.lkey,
			 (unsigned int)regd_buf->reg.rkey,
			 (unsigned long)regd_buf->reg.va,
			 (unsigned long)regd_buf->reg.len);
	} else { /* use FMR for multiple dma entries */
		iser_page_vec_build(mem, ib_conn->page_vec);
		err = iser_reg_page_vec(ib_conn, ib_conn->page_vec, &regd_buf->reg);
		if (err) {
			iser_data_buf_dump(mem);
			iser_err("mem->dma_nents = %d (dlength = 0x%x)\n", mem->dma_nents,
				 ntoh24(iser_ctask->desc.iscsi_header.dlength));
			iser_err("page_vec: data_size = 0x%x, length = %d, offset = 0x%x\n",
				 ib_conn->page_vec->data_size, ib_conn->page_vec->length,
				 ib_conn->page_vec->offset);
			for (i=0 ; i<ib_conn->page_vec->length ; i++)
				iser_err("page_vec[%d] = 0x%llx\n", i,
					 (unsigned long long) ib_conn->page_vec->pages[i]);
			return err;
E
Erez Zilber 已提交
477 478
		}
	}
479 480 481 482 483 484

	/* take a reference on this regd buf such that it will not be released *
	 * (eg in send dto completion) before we get the scsi response         */
	atomic_inc(&regd_buf->ref_count);
	return 0;
}