aops.c 61.2 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
/* -*- mode: c; c-basic-offset: 8; -*-
 * vim: noexpandtab sw=8 ts=8 sts=0:
 *
 * Copyright (C) 2002, 2004 Oracle.  All rights reserved.
 *
 * This program is free software; you can redistribute it and/or
 * modify it under the terms of the GNU General Public
 * License as published by the Free Software Foundation; either
 * version 2 of the License, or (at your option) any later version.
 *
 * This program is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * General Public License for more details.
 *
 * You should have received a copy of the GNU General Public
 * License along with this program; if not, write to the
 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
 * Boston, MA 021110-1307, USA.
 */

#include <linux/fs.h>
#include <linux/slab.h>
#include <linux/highmem.h>
#include <linux/pagemap.h>
#include <asm/byteorder.h>
27
#include <linux/swap.h>
M
Mark Fasheh 已提交
28
#include <linux/pipe_fs_i.h>
M
Mark Fasheh 已提交
29
#include <linux/mpage.h>
30
#include <linux/quotaops.h>
31
#include <linux/blkdev.h>
32
#include <linux/uio.h>
33 34 35 36 37 38 39 40 41 42 43 44

#include <cluster/masklog.h>

#include "ocfs2.h"

#include "alloc.h"
#include "aops.h"
#include "dlmglue.h"
#include "extent_map.h"
#include "file.h"
#include "inode.h"
#include "journal.h"
45
#include "suballoc.h"
46 47
#include "super.h"
#include "symlink.h"
T
Tao Ma 已提交
48
#include "refcounttree.h"
49
#include "ocfs2_trace.h"
50 51

#include "buffer_head_io.h"
52 53 54
#include "dir.h"
#include "namei.h"
#include "sysfile.h"
55 56 57 58 59 60 61 62 63 64 65 66

static int ocfs2_symlink_get_block(struct inode *inode, sector_t iblock,
				   struct buffer_head *bh_result, int create)
{
	int err = -EIO;
	int status;
	struct ocfs2_dinode *fe = NULL;
	struct buffer_head *bh = NULL;
	struct buffer_head *buffer_cache_bh = NULL;
	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
	void *kaddr;

67 68 69
	trace_ocfs2_symlink_get_block(
			(unsigned long long)OCFS2_I(inode)->ip_blkno,
			(unsigned long long)iblock, bh_result, create);
70 71 72 73 74 75 76 77 78

	BUG_ON(ocfs2_inode_is_fast_symlink(inode));

	if ((iblock << inode->i_sb->s_blocksize_bits) > PATH_MAX + 1) {
		mlog(ML_ERROR, "block offset > PATH_MAX: %llu",
		     (unsigned long long)iblock);
		goto bail;
	}

79
	status = ocfs2_read_inode_block(inode, &bh);
80 81 82 83 84 85 86 87
	if (status < 0) {
		mlog_errno(status);
		goto bail;
	}
	fe = (struct ocfs2_dinode *) bh->b_data;

	if ((u64)iblock >= ocfs2_clusters_to_blocks(inode->i_sb,
						    le32_to_cpu(fe->i_clusters))) {
88
		err = -ENOMEM;
89 90 91 92 93 94 95 96 97 98 99 100
		mlog(ML_ERROR, "block offset is outside the allocated size: "
		     "%llu\n", (unsigned long long)iblock);
		goto bail;
	}

	/* We don't use the page cache to create symlink data, so if
	 * need be, copy it over from the buffer cache. */
	if (!buffer_uptodate(bh_result) && ocfs2_inode_is_new(inode)) {
		u64 blkno = le64_to_cpu(fe->id2.i_list.l_recs[0].e_blkno) +
			    iblock;
		buffer_cache_bh = sb_getblk(osb->sb, blkno);
		if (!buffer_cache_bh) {
101
			err = -ENOMEM;
102 103 104 105 106 107 108 109 110 111
			mlog(ML_ERROR, "couldn't getblock for symlink!\n");
			goto bail;
		}

		/* we haven't locked out transactions, so a commit
		 * could've happened. Since we've got a reference on
		 * the bh, even if it commits while we're doing the
		 * copy, the data is still good. */
		if (buffer_jbd(buffer_cache_bh)
		    && ocfs2_inode_is_new(inode)) {
112
			kaddr = kmap_atomic(bh_result->b_page);
113 114 115 116 117 118 119
			if (!kaddr) {
				mlog(ML_ERROR, "couldn't kmap!\n");
				goto bail;
			}
			memcpy(kaddr + (bh_result->b_size * iblock),
			       buffer_cache_bh->b_data,
			       bh_result->b_size);
120
			kunmap_atomic(kaddr);
121 122 123 124 125 126 127 128 129 130 131
			set_buffer_uptodate(bh_result);
		}
		brelse(buffer_cache_bh);
	}

	map_bh(bh_result, inode->i_sb,
	       le64_to_cpu(fe->id2.i_list.l_recs[0].e_blkno) + iblock);

	err = 0;

bail:
132
	brelse(bh);
133 134 135 136

	return err;
}

T
Tao Ma 已提交
137 138
int ocfs2_get_block(struct inode *inode, sector_t iblock,
		    struct buffer_head *bh_result, int create)
139 140
{
	int err = 0;
141
	unsigned int ext_flags;
M
Mark Fasheh 已提交
142 143
	u64 max_blocks = bh_result->b_size >> inode->i_blkbits;
	u64 p_blkno, count, past_eof;
144
	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
145

146 147
	trace_ocfs2_get_block((unsigned long long)OCFS2_I(inode)->ip_blkno,
			      (unsigned long long)iblock, bh_result, create);
148 149 150 151 152 153 154 155 156 157 158

	if (OCFS2_I(inode)->ip_flags & OCFS2_INODE_SYSTEM_FILE)
		mlog(ML_NOTICE, "get_block on system inode 0x%p (%lu)\n",
		     inode, inode->i_ino);

	if (S_ISLNK(inode->i_mode)) {
		/* this always does I/O for some reason. */
		err = ocfs2_symlink_get_block(inode, iblock, bh_result, create);
		goto bail;
	}

M
Mark Fasheh 已提交
159
	err = ocfs2_extent_map_get_blocks(inode, iblock, &p_blkno, &count,
160
					  &ext_flags);
161 162
	if (err) {
		mlog(ML_ERROR, "Error %d from get_blocks(0x%p, %llu, 1, "
163 164
		     "%llu, NULL)\n", err, inode, (unsigned long long)iblock,
		     (unsigned long long)p_blkno);
165 166 167
		goto bail;
	}

M
Mark Fasheh 已提交
168 169 170
	if (max_blocks < count)
		count = max_blocks;

171 172 173 174
	/*
	 * ocfs2 never allocates in this function - the only time we
	 * need to use BH_New is when we're extending i_size on a file
	 * system which doesn't support holes, in which case BH_New
C
Christoph Hellwig 已提交
175
	 * allows __block_write_begin() to zero.
176 177 178 179 180
	 *
	 * If we see this on a sparse file system, then a truncate has
	 * raced us and removed the cluster. In this case, we clear
	 * the buffers dirty and uptodate bits and let the buffer code
	 * ignore it as a hole.
181
	 */
182 183 184 185 186
	if (create && p_blkno == 0 && ocfs2_sparse_alloc(osb)) {
		clear_buffer_dirty(bh_result);
		clear_buffer_uptodate(bh_result);
		goto bail;
	}
187

188 189
	/* Treat the unwritten extent as a hole for zeroing purposes. */
	if (p_blkno && !(ext_flags & OCFS2_EXT_UNWRITTEN))
190 191
		map_bh(bh_result, inode->i_sb, p_blkno);

M
Mark Fasheh 已提交
192 193
	bh_result->b_size = count << inode->i_blkbits;

194 195 196 197 198 199 200 201 202 203
	if (!ocfs2_sparse_alloc(osb)) {
		if (p_blkno == 0) {
			err = -EIO;
			mlog(ML_ERROR,
			     "iblock = %llu p_blkno = %llu blkno=(%llu)\n",
			     (unsigned long long)iblock,
			     (unsigned long long)p_blkno,
			     (unsigned long long)OCFS2_I(inode)->ip_blkno);
			mlog(ML_ERROR, "Size %llu, clusters %u\n", (unsigned long long)i_size_read(inode), OCFS2_I(inode)->ip_clusters);
			dump_stack();
204
			goto bail;
205 206
		}
	}
207

208
	past_eof = ocfs2_blocks_for_bytes(inode->i_sb, i_size_read(inode));
209 210 211

	trace_ocfs2_get_block_end((unsigned long long)OCFS2_I(inode)->ip_blkno,
				  (unsigned long long)past_eof);
212 213 214
	if (create && (iblock >= past_eof))
		set_buffer_new(bh_result);

215 216 217 218 219 220 221
bail:
	if (err < 0)
		err = -EIO;

	return err;
}

M
Mark Fasheh 已提交
222 223
int ocfs2_read_inline_data(struct inode *inode, struct page *page,
			   struct buffer_head *di_bh)
M
Mark Fasheh 已提交
224 225
{
	void *kaddr;
J
Jan Kara 已提交
226
	loff_t size;
M
Mark Fasheh 已提交
227 228 229
	struct ocfs2_dinode *di = (struct ocfs2_dinode *)di_bh->b_data;

	if (!(le16_to_cpu(di->i_dyn_features) & OCFS2_INLINE_DATA_FL)) {
230
		ocfs2_error(inode->i_sb, "Inode %llu lost inline data flag\n",
M
Mark Fasheh 已提交
231 232 233 234 235 236
			    (unsigned long long)OCFS2_I(inode)->ip_blkno);
		return -EROFS;
	}

	size = i_size_read(inode);

237
	if (size > PAGE_SIZE ||
238
	    size > ocfs2_max_inline_data_with_xattr(inode->i_sb, di)) {
M
Mark Fasheh 已提交
239
		ocfs2_error(inode->i_sb,
240
			    "Inode %llu has with inline data has bad size: %Lu\n",
J
Jan Kara 已提交
241 242
			    (unsigned long long)OCFS2_I(inode)->ip_blkno,
			    (unsigned long long)size);
M
Mark Fasheh 已提交
243 244 245
		return -EROFS;
	}

246
	kaddr = kmap_atomic(page);
M
Mark Fasheh 已提交
247 248 249
	if (size)
		memcpy(kaddr, di->id2.i_data.id_data, size);
	/* Clear the remaining part of the page */
250
	memset(kaddr + size, 0, PAGE_SIZE - size);
M
Mark Fasheh 已提交
251
	flush_dcache_page(page);
252
	kunmap_atomic(kaddr);
M
Mark Fasheh 已提交
253 254 255 256 257 258 259 260 261 262 263 264

	SetPageUptodate(page);

	return 0;
}

static int ocfs2_readpage_inline(struct inode *inode, struct page *page)
{
	int ret;
	struct buffer_head *di_bh = NULL;

	BUG_ON(!PageLocked(page));
265
	BUG_ON(!(OCFS2_I(inode)->ip_dyn_features & OCFS2_INLINE_DATA_FL));
M
Mark Fasheh 已提交
266

267
	ret = ocfs2_read_inode_block(inode, &di_bh);
M
Mark Fasheh 已提交
268 269 270 271 272 273 274 275 276 277 278 279 280
	if (ret) {
		mlog_errno(ret);
		goto out;
	}

	ret = ocfs2_read_inline_data(inode, page, di_bh);
out:
	unlock_page(page);

	brelse(di_bh);
	return ret;
}

281 282 283
static int ocfs2_readpage(struct file *file, struct page *page)
{
	struct inode *inode = page->mapping->host;
M
Mark Fasheh 已提交
284
	struct ocfs2_inode_info *oi = OCFS2_I(inode);
285
	loff_t start = (loff_t)page->index << PAGE_SHIFT;
286 287
	int ret, unlock = 1;

288 289
	trace_ocfs2_readpage((unsigned long long)oi->ip_blkno,
			     (page ? page->index : 0));
290

M
Mark Fasheh 已提交
291
	ret = ocfs2_inode_lock_with_page(inode, NULL, 0, page);
292 293 294 295 296 297 298
	if (ret != 0) {
		if (ret == AOP_TRUNCATED_PAGE)
			unlock = 0;
		mlog_errno(ret);
		goto out;
	}

M
Mark Fasheh 已提交
299
	if (down_read_trylock(&oi->ip_alloc_sem) == 0) {
300 301 302 303
		/*
		 * Unlock the page and cycle ip_alloc_sem so that we don't
		 * busyloop waiting for ip_alloc_sem to unlock
		 */
M
Mark Fasheh 已提交
304
		ret = AOP_TRUNCATED_PAGE;
305 306 307 308
		unlock_page(page);
		unlock = 0;
		down_read(&oi->ip_alloc_sem);
		up_read(&oi->ip_alloc_sem);
M
Mark Fasheh 已提交
309
		goto out_inode_unlock;
M
Mark Fasheh 已提交
310
	}
311 312 313 314 315 316

	/*
	 * i_size might have just been updated as we grabed the meta lock.  We
	 * might now be discovering a truncate that hit on another node.
	 * block_read_full_page->get_block freaks out if it is asked to read
	 * beyond the end of a file, so we check here.  Callers
317
	 * (generic_file_read, vm_ops->fault) are clever enough to check i_size
318 319 320 321 322
	 * and notice that the page they just read isn't needed.
	 *
	 * XXX sys_readahead() seems to get that wrong?
	 */
	if (start >= i_size_read(inode)) {
323
		zero_user(page, 0, PAGE_SIZE);
324 325 326 327 328
		SetPageUptodate(page);
		ret = 0;
		goto out_alloc;
	}

M
Mark Fasheh 已提交
329 330 331 332
	if (oi->ip_dyn_features & OCFS2_INLINE_DATA_FL)
		ret = ocfs2_readpage_inline(inode, page);
	else
		ret = block_read_full_page(page, ocfs2_get_block);
333 334 335 336
	unlock = 0;

out_alloc:
	up_read(&OCFS2_I(inode)->ip_alloc_sem);
M
Mark Fasheh 已提交
337 338
out_inode_unlock:
	ocfs2_inode_unlock(inode, 0);
339 340 341 342 343 344
out:
	if (unlock)
		unlock_page(page);
	return ret;
}

M
Mark Fasheh 已提交
345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387
/*
 * This is used only for read-ahead. Failures or difficult to handle
 * situations are safe to ignore.
 *
 * Right now, we don't bother with BH_Boundary - in-inode extent lists
 * are quite large (243 extents on 4k blocks), so most inodes don't
 * grow out to a tree. If need be, detecting boundary extents could
 * trivially be added in a future version of ocfs2_get_block().
 */
static int ocfs2_readpages(struct file *filp, struct address_space *mapping,
			   struct list_head *pages, unsigned nr_pages)
{
	int ret, err = -EIO;
	struct inode *inode = mapping->host;
	struct ocfs2_inode_info *oi = OCFS2_I(inode);
	loff_t start;
	struct page *last;

	/*
	 * Use the nonblocking flag for the dlm code to avoid page
	 * lock inversion, but don't bother with retrying.
	 */
	ret = ocfs2_inode_lock_full(inode, NULL, 0, OCFS2_LOCK_NONBLOCK);
	if (ret)
		return err;

	if (down_read_trylock(&oi->ip_alloc_sem) == 0) {
		ocfs2_inode_unlock(inode, 0);
		return err;
	}

	/*
	 * Don't bother with inline-data. There isn't anything
	 * to read-ahead in that case anyway...
	 */
	if (oi->ip_dyn_features & OCFS2_INLINE_DATA_FL)
		goto out_unlock;

	/*
	 * Check whether a remote node truncated this file - we just
	 * drop out in that case as it's not worth handling here.
	 */
	last = list_entry(pages->prev, struct page, lru);
388
	start = (loff_t)last->index << PAGE_SHIFT;
M
Mark Fasheh 已提交
389 390 391 392 393 394 395 396 397 398 399 400
	if (start >= i_size_read(inode))
		goto out_unlock;

	err = mpage_readpages(mapping, pages, nr_pages, ocfs2_get_block);

out_unlock:
	up_read(&oi->ip_alloc_sem);
	ocfs2_inode_unlock(inode, 0);

	return err;
}

401 402 403 404 405 406 407 408 409 410 411 412 413
/* Note: Because we don't support holes, our allocation has
 * already happened (allocation writes zeros to the file data)
 * so we don't have to worry about ordered writes in
 * ocfs2_writepage.
 *
 * ->writepage is called during the process of invalidating the page cache
 * during blocked lock processing.  It can't block on any cluster locks
 * to during block mapping.  It's relying on the fact that the block
 * mapping can't have disappeared under the dirty pages that it is
 * being asked to write back.
 */
static int ocfs2_writepage(struct page *page, struct writeback_control *wbc)
{
414 415 416
	trace_ocfs2_writepage(
		(unsigned long long)OCFS2_I(page->mapping->host)->ip_blkno,
		page->index);
417

418
	return block_write_full_page(page, ocfs2_get_block, wbc);
419 420 421 422 423 424
}

/* Taken from ext3. We don't necessarily need the full blown
 * functionality yet, but IMHO it's better to cut and paste the whole
 * thing so we can avoid introducing our own bugs (and easily pick up
 * their fixes when they happen) --Mark */
425 426 427 428 429 430 431
int walk_page_buffers(	handle_t *handle,
			struct buffer_head *head,
			unsigned from,
			unsigned to,
			int *partial,
			int (*fn)(	handle_t *handle,
					struct buffer_head *bh))
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
{
	struct buffer_head *bh;
	unsigned block_start, block_end;
	unsigned blocksize = head->b_size;
	int err, ret = 0;
	struct buffer_head *next;

	for (	bh = head, block_start = 0;
		ret == 0 && (bh != head || !block_start);
	    	block_start = block_end, bh = next)
	{
		next = bh->b_this_page;
		block_end = block_start + blocksize;
		if (block_end <= from || block_start >= to) {
			if (partial && !buffer_uptodate(bh))
				*partial = 1;
			continue;
		}
		err = (*fn)(handle, bh);
		if (!ret)
			ret = err;
	}
	return ret;
}

static sector_t ocfs2_bmap(struct address_space *mapping, sector_t block)
{
	sector_t status;
	u64 p_blkno = 0;
	int err = 0;
	struct inode *inode = mapping->host;

464 465
	trace_ocfs2_bmap((unsigned long long)OCFS2_I(inode)->ip_blkno,
			 (unsigned long long)block);
466 467 468 469 470

	/* We don't need to lock journal system files, since they aren't
	 * accessed concurrently from multiple nodes.
	 */
	if (!INODE_JOURNAL(inode)) {
M
Mark Fasheh 已提交
471
		err = ocfs2_inode_lock(inode, NULL, 0);
472 473 474 475 476 477 478 479
		if (err) {
			if (err != -ENOENT)
				mlog_errno(err);
			goto bail;
		}
		down_read(&OCFS2_I(inode)->ip_alloc_sem);
	}

M
Mark Fasheh 已提交
480 481 482
	if (!(OCFS2_I(inode)->ip_dyn_features & OCFS2_INLINE_DATA_FL))
		err = ocfs2_extent_map_get_blocks(inode, block, &p_blkno, NULL,
						  NULL);
483 484 485

	if (!INODE_JOURNAL(inode)) {
		up_read(&OCFS2_I(inode)->ip_alloc_sem);
M
Mark Fasheh 已提交
486
		ocfs2_inode_unlock(inode, 0);
487 488 489 490 491 492 493 494 495 496 497 498 499 500 501
	}

	if (err) {
		mlog(ML_ERROR, "get_blocks() failed, block = %llu\n",
		     (unsigned long long)block);
		mlog_errno(err);
		goto bail;
	}

bail:
	status = err ? 0 : p_blkno;

	return status;
}

502 503 504 505
static int ocfs2_releasepage(struct page *page, gfp_t wait)
{
	if (!page_has_buffers(page))
		return 0;
506
	return try_to_free_buffers(page);
507 508
}

509 510 511 512 513
static void ocfs2_figure_cluster_boundaries(struct ocfs2_super *osb,
					    u32 cpos,
					    unsigned int *start,
					    unsigned int *end)
{
514
	unsigned int cluster_start = 0, cluster_end = PAGE_SIZE;
515

516
	if (unlikely(PAGE_SHIFT > osb->s_clustersize_bits)) {
517 518
		unsigned int cpp;

519
		cpp = 1 << (PAGE_SHIFT - osb->s_clustersize_bits);
520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552

		cluster_start = cpos % cpp;
		cluster_start = cluster_start << osb->s_clustersize_bits;

		cluster_end = cluster_start + osb->s_clustersize;
	}

	BUG_ON(cluster_start > PAGE_SIZE);
	BUG_ON(cluster_end > PAGE_SIZE);

	if (start)
		*start = cluster_start;
	if (end)
		*end = cluster_end;
}

/*
 * 'from' and 'to' are the region in the page to avoid zeroing.
 *
 * If pagesize > clustersize, this function will avoid zeroing outside
 * of the cluster boundary.
 *
 * from == to == 0 is code for "zero the entire cluster region"
 */
static void ocfs2_clear_page_regions(struct page *page,
				     struct ocfs2_super *osb, u32 cpos,
				     unsigned from, unsigned to)
{
	void *kaddr;
	unsigned int cluster_start, cluster_end;

	ocfs2_figure_cluster_boundaries(osb, cpos, &cluster_start, &cluster_end);

553
	kaddr = kmap_atomic(page);
554 555 556 557 558 559 560 561 562 563

	if (from || to) {
		if (from > cluster_start)
			memset(kaddr + cluster_start, 0, from - cluster_start);
		if (to < cluster_end)
			memset(kaddr + to, 0, cluster_end - to);
	} else {
		memset(kaddr + cluster_start, 0, cluster_end - cluster_start);
	}

564
	kunmap_atomic(kaddr);
565 566
}

567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587
/*
 * Nonsparse file systems fully allocate before we get to the write
 * code. This prevents ocfs2_write() from tagging the write as an
 * allocating one, which means ocfs2_map_page_blocks() might try to
 * read-in the blocks at the tail of our file. Avoid reading them by
 * testing i_size against each block offset.
 */
static int ocfs2_should_read_blk(struct inode *inode, struct page *page,
				 unsigned int block_start)
{
	u64 offset = page_offset(page) + block_start;

	if (ocfs2_sparse_alloc(OCFS2_SB(inode->i_sb)))
		return 1;

	if (i_size_read(inode) > offset)
		return 1;

	return 0;
}

588
/*
C
Christoph Hellwig 已提交
589
 * Some of this taken from __block_write_begin(). We already have our
590 591 592 593 594
 * mapping by now though, and the entire write will be allocating or
 * it won't, so not much need to use BH_New.
 *
 * This will also skip zeroing, which is handled externally.
 */
595 596 597
int ocfs2_map_page_blocks(struct page *page, u64 *p_blkno,
			  struct inode *inode, unsigned int from,
			  unsigned int to, int new)
598 599 600 601 602 603 604 605 606 607 608 609 610 611
{
	int ret = 0;
	struct buffer_head *head, *bh, *wait[2], **wait_bh = wait;
	unsigned int block_end, block_start;
	unsigned int bsize = 1 << inode->i_blkbits;

	if (!page_has_buffers(page))
		create_empty_buffers(page, bsize, 0);

	head = page_buffers(page);
	for (bh = head, block_start = 0; bh != head || !block_start;
	     bh = bh->b_this_page, block_start += bsize) {
		block_end = block_start + bsize;

612 613
		clear_buffer_new(bh);

614 615 616 617
		/*
		 * Ignore blocks outside of our i/o range -
		 * they may belong to unallocated clusters.
		 */
618
		if (block_start >= to || block_end <= from) {
619 620 621 622 623 624 625 626 627
			if (PageUptodate(page))
				set_buffer_uptodate(bh);
			continue;
		}

		/*
		 * For an allocating write with cluster size >= page
		 * size, we always write the entire page.
		 */
628 629
		if (new)
			set_buffer_new(bh);
630 631 632 633 634 635 636 637 638 639

		if (!buffer_mapped(bh)) {
			map_bh(bh, inode->i_sb, *p_blkno);
			unmap_underlying_metadata(bh->b_bdev, bh->b_blocknr);
		}

		if (PageUptodate(page)) {
			if (!buffer_uptodate(bh))
				set_buffer_uptodate(bh);
		} else if (!buffer_uptodate(bh) && !buffer_delay(bh) &&
640
			   !buffer_new(bh) &&
641
			   ocfs2_should_read_blk(inode, page, block_start) &&
642
			   (block_start < from || block_end > to)) {
643
			ll_rw_block(REQ_OP_READ, 0, 1, &bh);
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
			*wait_bh++=bh;
		}

		*p_blkno = *p_blkno + 1;
	}

	/*
	 * If we issued read requests - let them complete.
	 */
	while(wait_bh > wait) {
		wait_on_buffer(*--wait_bh);
		if (!buffer_uptodate(*wait_bh))
			ret = -EIO;
	}

	if (ret == 0 || !new)
		return ret;

	/*
	 * If we get -EIO above, zero out any newly allocated blocks
	 * to avoid exposing stale data.
	 */
	bh = head;
	block_start = 0;
	do {
		block_end = block_start + bsize;
		if (block_end <= from)
			goto next_bh;
		if (block_start >= to)
			break;

675
		zero_user(page, block_start, bh->b_size);
676 677 678 679 680 681 682 683 684 685 686
		set_buffer_uptodate(bh);
		mark_buffer_dirty(bh);

next_bh:
		block_start = block_end;
		bh = bh->b_this_page;
	} while (bh != head);

	return ret;
}

687
#if (PAGE_SIZE >= OCFS2_MAX_CLUSTERSIZE)
688 689
#define OCFS2_MAX_CTXT_PAGES	1
#else
690
#define OCFS2_MAX_CTXT_PAGES	(OCFS2_MAX_CLUSTERSIZE / PAGE_SIZE)
691 692
#endif

693
#define OCFS2_MAX_CLUSTERS_PER_PAGE	(PAGE_SIZE / OCFS2_MIN_CLUSTERSIZE)
694

695 696 697 698 699 700 701
struct ocfs2_unwritten_extent {
	struct list_head	ue_node;
	struct list_head	ue_ip_node;
	u32			ue_cpos;
	u32			ue_phys;
};

M
Mark Fasheh 已提交
702
/*
703
 * Describe the state of a single cluster to be written to.
M
Mark Fasheh 已提交
704
 */
705 706 707 708 709 710 711 712
struct ocfs2_write_cluster_desc {
	u32		c_cpos;
	u32		c_phys;
	/*
	 * Give this a unique field because c_phys eventually gets
	 * filled.
	 */
	unsigned	c_new;
713
	unsigned	c_clear_unwritten;
714
	unsigned	c_needs_zero;
715
};
M
Mark Fasheh 已提交
716

717 718 719 720
struct ocfs2_write_ctxt {
	/* Logical cluster position / len of write */
	u32				w_cpos;
	u32				w_clen;
M
Mark Fasheh 已提交
721

722 723 724
	/* First cluster allocated in a nonsparse extend */
	u32				w_first_new_cpos;

725 726 727
	/* Type of caller. Must be one of buffer, mmap, direct.  */
	ocfs2_write_type_t		w_type;

728
	struct ocfs2_write_cluster_desc	w_desc[OCFS2_MAX_CLUSTERS_PER_PAGE];
M
Mark Fasheh 已提交
729

730 731 732 733 734 735 736
	/*
	 * This is true if page_size > cluster_size.
	 *
	 * It triggers a set of special cases during write which might
	 * have to deal with allocating writes to partial pages.
	 */
	unsigned int			w_large_pages;
M
Mark Fasheh 已提交
737

738 739 740 741 742 743 744 745 746 747 748 749 750
	/*
	 * Pages involved in this write.
	 *
	 * w_target_page is the page being written to by the user.
	 *
	 * w_pages is an array of pages which always contains
	 * w_target_page, and in the case of an allocating write with
	 * page_size < cluster size, it will contain zero'd and mapped
	 * pages adjacent to w_target_page which need to be written
	 * out in so that future reads from that region will get
	 * zero's.
	 */
	unsigned int			w_num_pages;
751
	struct page			*w_pages[OCFS2_MAX_CTXT_PAGES];
752
	struct page			*w_target_page;
753

W
Wengang Wang 已提交
754 755 756 757 758 759
	/*
	 * w_target_locked is used for page_mkwrite path indicating no unlocking
	 * against w_target_page in ocfs2_write_end_nolock.
	 */
	unsigned int			w_target_locked:1;

760 761 762 763 764 765 766 767 768 769 770 771 772 773
	/*
	 * ocfs2_write_end() uses this to know what the real range to
	 * write in the target should be.
	 */
	unsigned int			w_target_from;
	unsigned int			w_target_to;

	/*
	 * We could use journal_current_handle() but this is cleaner,
	 * IMHO -Mark
	 */
	handle_t			*w_handle;

	struct buffer_head		*w_di_bh;
774 775

	struct ocfs2_cached_dealloc_ctxt w_dealloc;
776 777

	struct list_head		w_unwritten_list;
778 779
};

780
void ocfs2_unlock_and_free_pages(struct page **pages, int num_pages)
781 782 783
{
	int i;

784 785 786 787
	for(i = 0; i < num_pages; i++) {
		if (pages[i]) {
			unlock_page(pages[i]);
			mark_page_accessed(pages[i]);
788
			put_page(pages[i]);
789
		}
M
Mark Fasheh 已提交
790
	}
791 792
}

J
Junxiao Bi 已提交
793
static void ocfs2_unlock_pages(struct ocfs2_write_ctxt *wc)
794
{
W
Wengang Wang 已提交
795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810
	int i;

	/*
	 * w_target_locked is only set to true in the page_mkwrite() case.
	 * The intent is to allow us to lock the target page from write_begin()
	 * to write_end(). The caller must hold a ref on w_target_page.
	 */
	if (wc->w_target_locked) {
		BUG_ON(!wc->w_target_page);
		for (i = 0; i < wc->w_num_pages; i++) {
			if (wc->w_target_page == wc->w_pages[i]) {
				wc->w_pages[i] = NULL;
				break;
			}
		}
		mark_page_accessed(wc->w_target_page);
811
		put_page(wc->w_target_page);
W
Wengang Wang 已提交
812
	}
813
	ocfs2_unlock_and_free_pages(wc->w_pages, wc->w_num_pages);
J
Junxiao Bi 已提交
814
}
M
Mark Fasheh 已提交
815

816 817 818 819
static void ocfs2_free_unwritten_list(struct inode *inode,
				 struct list_head *head)
{
	struct ocfs2_inode_info *oi = OCFS2_I(inode);
820
	struct ocfs2_unwritten_extent *ue = NULL, *tmp = NULL;
821

822 823
	list_for_each_entry_safe(ue, tmp, head, ue_node) {
		list_del(&ue->ue_node);
824
		spin_lock(&oi->ip_lock);
825
		list_del(&ue->ue_ip_node);
826
		spin_unlock(&oi->ip_lock);
827
		kfree(ue);
828 829 830 831 832
	}
}

static void ocfs2_free_write_ctxt(struct inode *inode,
				  struct ocfs2_write_ctxt *wc)
J
Junxiao Bi 已提交
833
{
834
	ocfs2_free_unwritten_list(inode, &wc->w_unwritten_list);
J
Junxiao Bi 已提交
835
	ocfs2_unlock_pages(wc);
836 837 838 839 840 841
	brelse(wc->w_di_bh);
	kfree(wc);
}

static int ocfs2_alloc_write_ctxt(struct ocfs2_write_ctxt **wcp,
				  struct ocfs2_super *osb, loff_t pos,
842 843
				  unsigned len, ocfs2_write_type_t type,
				  struct buffer_head *di_bh)
844
{
845
	u32 cend;
846 847 848 849 850
	struct ocfs2_write_ctxt *wc;

	wc = kzalloc(sizeof(struct ocfs2_write_ctxt), GFP_NOFS);
	if (!wc)
		return -ENOMEM;
M
Mark Fasheh 已提交
851

852
	wc->w_cpos = pos >> osb->s_clustersize_bits;
853
	wc->w_first_new_cpos = UINT_MAX;
854 855
	cend = (pos + len - 1) >> osb->s_clustersize_bits;
	wc->w_clen = cend - wc->w_cpos + 1;
856 857
	get_bh(di_bh);
	wc->w_di_bh = di_bh;
858
	wc->w_type = type;
M
Mark Fasheh 已提交
859

860
	if (unlikely(PAGE_SHIFT > osb->s_clustersize_bits))
861 862 863 864
		wc->w_large_pages = 1;
	else
		wc->w_large_pages = 0;

865
	ocfs2_init_dealloc_ctxt(&wc->w_dealloc);
866
	INIT_LIST_HEAD(&wc->w_unwritten_list);
867

868
	*wcp = wc;
M
Mark Fasheh 已提交
869

870
	return 0;
M
Mark Fasheh 已提交
871 872
}

873
/*
874 875 876
 * If a page has any new buffers, zero them out here, and mark them uptodate
 * and dirty so they'll be written out (in order to prevent uninitialised
 * block data from leaking). And clear the new bit.
877
 */
878
static void ocfs2_zero_new_buffers(struct page *page, unsigned from, unsigned to)
879
{
880 881
	unsigned int block_start, block_end;
	struct buffer_head *head, *bh;
882

883 884 885
	BUG_ON(!PageLocked(page));
	if (!page_has_buffers(page))
		return;
886

887 888 889 890 891 892 893 894 895 896 897 898 899
	bh = head = page_buffers(page);
	block_start = 0;
	do {
		block_end = block_start + bh->b_size;

		if (buffer_new(bh)) {
			if (block_end > from && block_start < to) {
				if (!PageUptodate(page)) {
					unsigned start, end;

					start = max(from, block_start);
					end = min(to, block_end);

900
					zero_user_segment(page, start, end);
901 902 903 904 905 906 907
					set_buffer_uptodate(bh);
				}

				clear_buffer_new(bh);
				mark_buffer_dirty(bh);
			}
		}
908

909 910 911 912 913 914 915 916 917 918 919 920 921 922
		block_start = block_end;
		bh = bh->b_this_page;
	} while (bh != head);
}

/*
 * Only called when we have a failure during allocating write to write
 * zero's to the newly allocated region.
 */
static void ocfs2_write_failure(struct inode *inode,
				struct ocfs2_write_ctxt *wc,
				loff_t user_pos, unsigned user_len)
{
	int i;
923
	unsigned from = user_pos & (PAGE_SIZE - 1),
924
		to = user_pos + user_len;
925 926
	struct page *tmppage;

927 928
	if (wc->w_target_page)
		ocfs2_zero_new_buffers(wc->w_target_page, from, to);
929

930 931
	for(i = 0; i < wc->w_num_pages; i++) {
		tmppage = wc->w_pages[i];
932

933
		if (tmppage && page_has_buffers(tmppage)) {
934
			if (ocfs2_should_order_data(inode))
J
Joel Becker 已提交
935
				ocfs2_jbd2_file_inode(wc->w_handle, inode);
936 937 938

			block_commit_write(tmppage, from, to);
		}
939 940 941
	}
}

942 943 944 945 946
static int ocfs2_prepare_page_for_write(struct inode *inode, u64 *p_blkno,
					struct ocfs2_write_ctxt *wc,
					struct page *page, u32 cpos,
					loff_t user_pos, unsigned user_len,
					int new)
947
{
948 949
	int ret;
	unsigned int map_from = 0, map_to = 0;
950
	unsigned int cluster_start, cluster_end;
951
	unsigned int user_data_from = 0, user_data_to = 0;
952

953
	ocfs2_figure_cluster_boundaries(OCFS2_SB(inode->i_sb), cpos,
954 955
					&cluster_start, &cluster_end);

956 957 958 959 960 961
	/* treat the write as new if the a hole/lseek spanned across
	 * the page boundary.
	 */
	new = new | ((i_size_read(inode) <= page_offset(page)) &&
			(page_offset(page) <= user_pos));

962
	if (page == wc->w_target_page) {
963
		map_from = user_pos & (PAGE_SIZE - 1);
964 965 966 967 968 969 970 971 972 973
		map_to = map_from + user_len;

		if (new)
			ret = ocfs2_map_page_blocks(page, p_blkno, inode,
						    cluster_start, cluster_end,
						    new);
		else
			ret = ocfs2_map_page_blocks(page, p_blkno, inode,
						    map_from, map_to, new);
		if (ret) {
974 975 976 977
			mlog_errno(ret);
			goto out;
		}

978 979
		user_data_from = map_from;
		user_data_to = map_to;
980
		if (new) {
981 982
			map_from = cluster_start;
			map_to = cluster_end;
983 984 985 986 987 988 989 990 991
		}
	} else {
		/*
		 * If we haven't allocated the new page yet, we
		 * shouldn't be writing it out without copying user
		 * data. This is likely a math error from the caller.
		 */
		BUG_ON(!new);

992 993
		map_from = cluster_start;
		map_to = cluster_end;
994 995

		ret = ocfs2_map_page_blocks(page, p_blkno, inode,
996
					    cluster_start, cluster_end, new);
997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014
		if (ret) {
			mlog_errno(ret);
			goto out;
		}
	}

	/*
	 * Parts of newly allocated pages need to be zero'd.
	 *
	 * Above, we have also rewritten 'to' and 'from' - as far as
	 * the rest of the function is concerned, the entire cluster
	 * range inside of a page needs to be written.
	 *
	 * We can skip this if the page is up to date - it's already
	 * been zero'd from being read in as a hole.
	 */
	if (new && !PageUptodate(page))
		ocfs2_clear_page_regions(page, OCFS2_SB(inode->i_sb),
1015
					 cpos, user_data_from, user_data_to);
1016 1017 1018 1019

	flush_dcache_page(page);

out:
1020
	return ret;
1021 1022 1023
}

/*
1024
 * This function will only grab one clusters worth of pages.
1025
 */
1026 1027
static int ocfs2_grab_pages_for_write(struct address_space *mapping,
				      struct ocfs2_write_ctxt *wc,
1028 1029
				      u32 cpos, loff_t user_pos,
				      unsigned user_len, int new,
M
Mark Fasheh 已提交
1030
				      struct page *mmap_page)
1031
{
1032
	int ret = 0, i;
1033
	unsigned long start, target_index, end_index, index;
1034
	struct inode *inode = mapping->host;
1035
	loff_t last_byte;
1036

1037
	target_index = user_pos >> PAGE_SHIFT;
1038 1039 1040

	/*
	 * Figure out how many pages we'll be manipulating here. For
1041
	 * non allocating write, we just change the one
1042 1043 1044
	 * page. Otherwise, we'll need a whole clusters worth.  If we're
	 * writing past i_size, we only need enough pages to cover the
	 * last page of the write.
1045 1046
	 */
	if (new) {
1047 1048
		wc->w_num_pages = ocfs2_pages_per_cluster(inode->i_sb);
		start = ocfs2_align_clusters_to_page_index(inode->i_sb, cpos);
1049 1050 1051 1052 1053 1054 1055
		/*
		 * We need the index *past* the last page we could possibly
		 * touch.  This is the page past the end of the write or
		 * i_size, whichever is greater.
		 */
		last_byte = max(user_pos + user_len, i_size_read(inode));
		BUG_ON(last_byte < 1);
1056
		end_index = ((last_byte - 1) >> PAGE_SHIFT) + 1;
1057 1058
		if ((start + wc->w_num_pages) > end_index)
			wc->w_num_pages = end_index - start;
1059
	} else {
1060 1061
		wc->w_num_pages = 1;
		start = target_index;
1062
	}
1063
	end_index = (user_pos + user_len - 1) >> PAGE_SHIFT;
1064

1065
	for(i = 0; i < wc->w_num_pages; i++) {
1066 1067
		index = start + i;

1068 1069
		if (index >= target_index && index <= end_index &&
		    wc->w_type == OCFS2_WRITE_MMAP) {
M
Mark Fasheh 已提交
1070 1071 1072 1073 1074 1075 1076
			/*
			 * ocfs2_pagemkwrite() is a little different
			 * and wants us to directly use the page
			 * passed in.
			 */
			lock_page(mmap_page);

W
Wengang Wang 已提交
1077
			/* Exit and let the caller retry */
M
Mark Fasheh 已提交
1078
			if (mmap_page->mapping != mapping) {
W
Wengang Wang 已提交
1079
				WARN_ON(mmap_page->mapping);
M
Mark Fasheh 已提交
1080
				unlock_page(mmap_page);
W
Wengang Wang 已提交
1081
				ret = -EAGAIN;
M
Mark Fasheh 已提交
1082 1083 1084
				goto out;
			}

1085
			get_page(mmap_page);
M
Mark Fasheh 已提交
1086
			wc->w_pages[i] = mmap_page;
W
Wengang Wang 已提交
1087
			wc->w_target_locked = true;
1088 1089 1090 1091 1092
		} else if (index >= target_index && index <= end_index &&
			   wc->w_type == OCFS2_WRITE_DIRECT) {
			/* Direct write has no mapping page. */
			wc->w_pages[i] = NULL;
			continue;
M
Mark Fasheh 已提交
1093 1094 1095 1096 1097 1098 1099 1100
		} else {
			wc->w_pages[i] = find_or_create_page(mapping, index,
							     GFP_NOFS);
			if (!wc->w_pages[i]) {
				ret = -ENOMEM;
				mlog_errno(ret);
				goto out;
			}
1101
		}
1102
		wait_for_stable_page(wc->w_pages[i]);
1103 1104 1105

		if (index == target_index)
			wc->w_target_page = wc->w_pages[i];
1106
	}
1107
out:
W
Wengang Wang 已提交
1108 1109
	if (ret)
		wc->w_target_locked = false;
1110 1111 1112 1113 1114 1115 1116
	return ret;
}

/*
 * Prepare a single cluster for write one cluster into the file.
 */
static int ocfs2_write_cluster(struct address_space *mapping,
1117
			       u32 *phys, unsigned int new,
1118
			       unsigned int clear_unwritten,
1119
			       unsigned int should_zero,
1120
			       struct ocfs2_alloc_context *data_ac,
1121 1122 1123 1124
			       struct ocfs2_alloc_context *meta_ac,
			       struct ocfs2_write_ctxt *wc, u32 cpos,
			       loff_t user_pos, unsigned user_len)
{
1125
	int ret, i;
1126
	u64 p_blkno;
1127
	struct inode *inode = mapping->host;
1128
	struct ocfs2_extent_tree et;
1129
	int bpc = ocfs2_clusters_to_blocks(inode->i_sb, 1);
1130

1131
	if (new) {
1132 1133
		u32 tmp_pos;

1134 1135 1136 1137
		/*
		 * This is safe to call with the page locks - it won't take
		 * any additional semaphores or cluster locks.
		 */
1138
		tmp_pos = cpos;
1139
		ret = ocfs2_add_inode_data(OCFS2_SB(inode->i_sb), inode,
1140 1141 1142
					   &tmp_pos, 1, !clear_unwritten,
					   wc->w_di_bh, wc->w_handle,
					   data_ac, meta_ac, NULL);
1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158
		/*
		 * This shouldn't happen because we must have already
		 * calculated the correct meta data allocation required. The
		 * internal tree allocation code should know how to increase
		 * transaction credits itself.
		 *
		 * If need be, we could handle -EAGAIN for a
		 * RESTART_TRANS here.
		 */
		mlog_bug_on_msg(ret == -EAGAIN,
				"Inode %llu: EAGAIN return during allocation.\n",
				(unsigned long long)OCFS2_I(inode)->ip_blkno);
		if (ret < 0) {
			mlog_errno(ret);
			goto out;
		}
1159
	} else if (clear_unwritten) {
1160 1161
		ocfs2_init_dinode_extent_tree(&et, INODE_CACHE(inode),
					      wc->w_di_bh);
1162
		ret = ocfs2_mark_extent_written(inode, &et,
1163
						wc->w_handle, cpos, 1, *phys,
1164
						meta_ac, &wc->w_dealloc);
1165 1166 1167 1168 1169
		if (ret < 0) {
			mlog_errno(ret);
			goto out;
		}
	}
1170 1171 1172 1173 1174

	/*
	 * The only reason this should fail is due to an inability to
	 * find the extent added.
	 */
1175
	ret = ocfs2_get_clusters(inode, cpos, phys, NULL, NULL);
1176
	if (ret < 0) {
1177
		mlog(ML_ERROR, "Get physical blkno failed for inode %llu, "
1178 1179
			    "at logical cluster %u",
			    (unsigned long long)OCFS2_I(inode)->ip_blkno, cpos);
1180 1181 1182
		goto out;
	}

1183 1184 1185 1186 1187
	BUG_ON(*phys == 0);

	p_blkno = ocfs2_clusters_to_blocks(inode->i_sb, *phys);
	if (!should_zero)
		p_blkno += (user_pos >> inode->i_sb->s_blocksize_bits) & (u64)(bpc - 1);
1188

1189 1190
	for(i = 0; i < wc->w_num_pages; i++) {
		int tmpret;
1191

1192 1193 1194 1195 1196 1197
		/* This is the direct io target page. */
		if (wc->w_pages[i] == NULL) {
			p_blkno++;
			continue;
		}

1198 1199
		tmpret = ocfs2_prepare_page_for_write(inode, &p_blkno, wc,
						      wc->w_pages[i], cpos,
1200 1201
						      user_pos, user_len,
						      should_zero);
1202 1203 1204
		if (tmpret) {
			mlog_errno(tmpret);
			if (ret == 0)
1205
				ret = tmpret;
1206
		}
1207 1208
	}

1209 1210 1211 1212 1213 1214
	/*
	 * We only have cleanup to do in case of allocating write.
	 */
	if (ret && new)
		ocfs2_write_failure(inode, wc, user_pos, user_len);

1215 1216
out:

1217
	return ret;
1218 1219
}

1220 1221 1222 1223 1224 1225 1226
static int ocfs2_write_cluster_by_desc(struct address_space *mapping,
				       struct ocfs2_alloc_context *data_ac,
				       struct ocfs2_alloc_context *meta_ac,
				       struct ocfs2_write_ctxt *wc,
				       loff_t pos, unsigned len)
{
	int ret, i;
1227 1228
	loff_t cluster_off;
	unsigned int local_len = len;
1229
	struct ocfs2_write_cluster_desc *desc;
1230
	struct ocfs2_super *osb = OCFS2_SB(mapping->host->i_sb);
1231 1232 1233 1234

	for (i = 0; i < wc->w_clen; i++) {
		desc = &wc->w_desc[i];

1235 1236 1237 1238 1239 1240 1241 1242 1243
		/*
		 * We have to make sure that the total write passed in
		 * doesn't extend past a single cluster.
		 */
		local_len = len;
		cluster_off = pos & (osb->s_clustersize - 1);
		if ((cluster_off + local_len) > osb->s_clustersize)
			local_len = osb->s_clustersize - cluster_off;

1244
		ret = ocfs2_write_cluster(mapping, &desc->c_phys,
1245 1246
					  desc->c_new,
					  desc->c_clear_unwritten,
1247 1248
					  desc->c_needs_zero,
					  data_ac, meta_ac,
1249
					  wc, desc->c_cpos, pos, local_len);
1250 1251 1252 1253
		if (ret) {
			mlog_errno(ret);
			goto out;
		}
1254 1255 1256

		len -= local_len;
		pos += local_len;
1257 1258 1259 1260 1261 1262 1263
	}

	ret = 0;
out:
	return ret;
}

1264 1265 1266 1267 1268 1269 1270 1271
/*
 * ocfs2_write_end() wants to know which parts of the target page it
 * should complete the write on. It's easiest to compute them ahead of
 * time when a more complete view of the write is available.
 */
static void ocfs2_set_target_boundaries(struct ocfs2_super *osb,
					struct ocfs2_write_ctxt *wc,
					loff_t pos, unsigned len, int alloc)
1272
{
1273
	struct ocfs2_write_cluster_desc *desc;
1274

1275
	wc->w_target_from = pos & (PAGE_SIZE - 1);
1276 1277 1278 1279 1280 1281 1282 1283 1284 1285 1286 1287 1288
	wc->w_target_to = wc->w_target_from + len;

	if (alloc == 0)
		return;

	/*
	 * Allocating write - we may have different boundaries based
	 * on page size and cluster size.
	 *
	 * NOTE: We can no longer compute one value from the other as
	 * the actual write length and user provided length may be
	 * different.
	 */
1289

1290 1291 1292
	if (wc->w_large_pages) {
		/*
		 * We only care about the 1st and last cluster within
1293
		 * our range and whether they should be zero'd or not. Either
1294 1295 1296 1297
		 * value may be extended out to the start/end of a
		 * newly allocated cluster.
		 */
		desc = &wc->w_desc[0];
1298
		if (desc->c_needs_zero)
1299 1300 1301 1302 1303 1304
			ocfs2_figure_cluster_boundaries(osb,
							desc->c_cpos,
							&wc->w_target_from,
							NULL);

		desc = &wc->w_desc[wc->w_clen - 1];
1305
		if (desc->c_needs_zero)
1306 1307 1308 1309 1310 1311
			ocfs2_figure_cluster_boundaries(osb,
							desc->c_cpos,
							NULL,
							&wc->w_target_to);
	} else {
		wc->w_target_from = 0;
1312
		wc->w_target_to = PAGE_SIZE;
1313
	}
1314 1315
}

1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327
/*
 * Check if this extent is marked UNWRITTEN by direct io. If so, we need not to
 * do the zero work. And should not to clear UNWRITTEN since it will be cleared
 * by the direct io procedure.
 * If this is a new extent that allocated by direct io, we should mark it in
 * the ip_unwritten_list.
 */
static int ocfs2_unwritten_check(struct inode *inode,
				 struct ocfs2_write_ctxt *wc,
				 struct ocfs2_write_cluster_desc *desc)
{
	struct ocfs2_inode_info *oi = OCFS2_I(inode);
1328
	struct ocfs2_unwritten_extent *ue = NULL, *new = NULL;
1329 1330 1331 1332 1333 1334 1335 1336 1337 1338
	int ret = 0;

	if (!desc->c_needs_zero)
		return 0;

retry:
	spin_lock(&oi->ip_lock);
	/* Needs not to zero no metter buffer or direct. The one who is zero
	 * the cluster is doing zero. And he will clear unwritten after all
	 * cluster io finished. */
1339 1340
	list_for_each_entry(ue, &oi->ip_unwritten_list, ue_ip_node) {
		if (desc->c_cpos == ue->ue_cpos) {
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
			BUG_ON(desc->c_new);
			desc->c_needs_zero = 0;
			desc->c_clear_unwritten = 0;
			goto unlock;
		}
	}

	if (wc->w_type != OCFS2_WRITE_DIRECT)
		goto unlock;

	if (new == NULL) {
		spin_unlock(&oi->ip_lock);
		new = kmalloc(sizeof(struct ocfs2_unwritten_extent),
			     GFP_NOFS);
		if (new == NULL) {
			ret = -ENOMEM;
			goto out;
		}
		goto retry;
	}
	/* This direct write will doing zero. */
	new->ue_cpos = desc->c_cpos;
	new->ue_phys = desc->c_phys;
	desc->c_clear_unwritten = 0;
	list_add_tail(&new->ue_ip_node, &oi->ip_unwritten_list);
	list_add_tail(&new->ue_node, &wc->w_unwritten_list);
	new = NULL;
unlock:
	spin_unlock(&oi->ip_lock);
out:
	if (new)
		kfree(new);
	return ret;
}

1376 1377 1378
/*
 * Populate each single-cluster write descriptor in the write context
 * with information about the i/o to be done.
1379 1380 1381 1382
 *
 * Returns the number of clusters that will have to be allocated, as
 * well as a worst case estimate of the number of extent records that
 * would have to be created during a write to an unwritten region.
1383 1384 1385
 */
static int ocfs2_populate_write_desc(struct inode *inode,
				     struct ocfs2_write_ctxt *wc,
1386 1387
				     unsigned int *clusters_to_alloc,
				     unsigned int *extents_to_split)
1388
{
1389
	int ret;
1390
	struct ocfs2_write_cluster_desc *desc;
1391
	unsigned int num_clusters = 0;
1392
	unsigned int ext_flags = 0;
1393 1394
	u32 phys = 0;
	int i;
1395

1396 1397 1398
	*clusters_to_alloc = 0;
	*extents_to_split = 0;

1399 1400 1401 1402 1403
	for (i = 0; i < wc->w_clen; i++) {
		desc = &wc->w_desc[i];
		desc->c_cpos = wc->w_cpos + i;

		if (num_clusters == 0) {
1404 1405 1406
			/*
			 * Need to look up the next extent record.
			 */
1407
			ret = ocfs2_get_clusters(inode, desc->c_cpos, &phys,
1408
						 &num_clusters, &ext_flags);
1409 1410
			if (ret) {
				mlog_errno(ret);
1411
				goto out;
1412
			}
1413

T
Tao Ma 已提交
1414 1415 1416
			/* We should already CoW the refcountd extent. */
			BUG_ON(ext_flags & OCFS2_EXT_REFCOUNTED);

1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427
			/*
			 * Assume worst case - that we're writing in
			 * the middle of the extent.
			 *
			 * We can assume that the write proceeds from
			 * left to right, in which case the extent
			 * insert code is smart enough to coalesce the
			 * next splits into the previous records created.
			 */
			if (ext_flags & OCFS2_EXT_UNWRITTEN)
				*extents_to_split = *extents_to_split + 2;
1428 1429 1430 1431 1432 1433 1434 1435
		} else if (phys) {
			/*
			 * Only increment phys if it doesn't describe
			 * a hole.
			 */
			phys++;
		}

1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446
		/*
		 * If w_first_new_cpos is < UINT_MAX, we have a non-sparse
		 * file that got extended.  w_first_new_cpos tells us
		 * where the newly allocated clusters are so we can
		 * zero them.
		 */
		if (desc->c_cpos >= wc->w_first_new_cpos) {
			BUG_ON(phys == 0);
			desc->c_needs_zero = 1;
		}

1447 1448 1449
		desc->c_phys = phys;
		if (phys == 0) {
			desc->c_new = 1;
1450
			desc->c_needs_zero = 1;
1451
			desc->c_clear_unwritten = 1;
1452
			*clusters_to_alloc = *clusters_to_alloc + 1;
1453
		}
1454 1455

		if (ext_flags & OCFS2_EXT_UNWRITTEN) {
1456
			desc->c_clear_unwritten = 1;
1457 1458
			desc->c_needs_zero = 1;
		}
1459

1460 1461 1462 1463 1464 1465
		ret = ocfs2_unwritten_check(inode, wc, desc);
		if (ret) {
			mlog_errno(ret);
			goto out;
		}

1466
		num_clusters--;
1467 1468
	}

1469 1470 1471 1472 1473
	ret = 0;
out:
	return ret;
}

M
Mark Fasheh 已提交
1474 1475 1476 1477 1478 1479 1480 1481 1482 1483
static int ocfs2_write_begin_inline(struct address_space *mapping,
				    struct inode *inode,
				    struct ocfs2_write_ctxt *wc)
{
	int ret;
	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
	struct page *page;
	handle_t *handle;
	struct ocfs2_dinode *di = (struct ocfs2_dinode *)wc->w_di_bh->b_data;

1484 1485 1486 1487 1488 1489 1490
	handle = ocfs2_start_trans(osb, OCFS2_INODE_UPDATE_CREDITS);
	if (IS_ERR(handle)) {
		ret = PTR_ERR(handle);
		mlog_errno(ret);
		goto out;
	}

M
Mark Fasheh 已提交
1491 1492
	page = find_or_create_page(mapping, 0, GFP_NOFS);
	if (!page) {
1493
		ocfs2_commit_trans(osb, handle);
M
Mark Fasheh 已提交
1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504
		ret = -ENOMEM;
		mlog_errno(ret);
		goto out;
	}
	/*
	 * If we don't set w_num_pages then this page won't get unlocked
	 * and freed on cleanup of the write context.
	 */
	wc->w_pages[0] = wc->w_target_page = page;
	wc->w_num_pages = 1;

1505
	ret = ocfs2_journal_access_di(handle, INODE_CACHE(inode), wc->w_di_bh,
1506
				      OCFS2_JOURNAL_ACCESS_WRITE);
M
Mark Fasheh 已提交
1507 1508 1509 1510 1511 1512 1513 1514 1515 1516 1517 1518 1519 1520 1521 1522 1523 1524 1525 1526 1527 1528 1529 1530 1531 1532 1533 1534
	if (ret) {
		ocfs2_commit_trans(osb, handle);

		mlog_errno(ret);
		goto out;
	}

	if (!(OCFS2_I(inode)->ip_dyn_features & OCFS2_INLINE_DATA_FL))
		ocfs2_set_inode_data_inline(inode, di);

	if (!PageUptodate(page)) {
		ret = ocfs2_read_inline_data(inode, page, wc->w_di_bh);
		if (ret) {
			ocfs2_commit_trans(osb, handle);

			goto out;
		}
	}

	wc->w_handle = handle;
out:
	return ret;
}

int ocfs2_size_fits_inline_data(struct buffer_head *di_bh, u64 new_size)
{
	struct ocfs2_dinode *di = (struct ocfs2_dinode *)di_bh->b_data;

1535
	if (new_size <= le16_to_cpu(di->id2.i_data.id_count))
M
Mark Fasheh 已提交
1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 1546 1547
		return 1;
	return 0;
}

static int ocfs2_try_to_write_inline_data(struct address_space *mapping,
					  struct inode *inode, loff_t pos,
					  unsigned len, struct page *mmap_page,
					  struct ocfs2_write_ctxt *wc)
{
	int ret, written = 0;
	loff_t end = pos + len;
	struct ocfs2_inode_info *oi = OCFS2_I(inode);
1548
	struct ocfs2_dinode *di = NULL;
M
Mark Fasheh 已提交
1549

1550 1551 1552
	trace_ocfs2_try_to_write_inline_data((unsigned long long)oi->ip_blkno,
					     len, (unsigned long long)pos,
					     oi->ip_dyn_features);
M
Mark Fasheh 已提交
1553 1554 1555 1556 1557 1558 1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579 1580

	/*
	 * Handle inodes which already have inline data 1st.
	 */
	if (oi->ip_dyn_features & OCFS2_INLINE_DATA_FL) {
		if (mmap_page == NULL &&
		    ocfs2_size_fits_inline_data(wc->w_di_bh, end))
			goto do_inline_write;

		/*
		 * The write won't fit - we have to give this inode an
		 * inline extent list now.
		 */
		ret = ocfs2_convert_inline_data_to_extents(inode, wc->w_di_bh);
		if (ret)
			mlog_errno(ret);
		goto out;
	}

	/*
	 * Check whether the inode can accept inline data.
	 */
	if (oi->ip_clusters != 0 || i_size_read(inode) != 0)
		return 0;

	/*
	 * Check whether the write can fit.
	 */
1581 1582 1583
	di = (struct ocfs2_dinode *)wc->w_di_bh->b_data;
	if (mmap_page ||
	    end > ocfs2_max_inline_data_with_xattr(inode->i_sb, di))
M
Mark Fasheh 已提交
1584 1585 1586 1587 1588 1589 1590 1591 1592 1593 1594 1595 1596 1597 1598 1599 1600 1601
		return 0;

do_inline_write:
	ret = ocfs2_write_begin_inline(mapping, inode, wc);
	if (ret) {
		mlog_errno(ret);
		goto out;
	}

	/*
	 * This signals to the caller that the data can be written
	 * inline.
	 */
	written = 1;
out:
	return written ? written : ret;
}

1602 1603 1604 1605 1606 1607 1608 1609 1610
/*
 * This function only does anything for file systems which can't
 * handle sparse files.
 *
 * What we want to do here is fill in any hole between the current end
 * of allocation and the end of our write. That way the rest of the
 * write path can treat it as an non-allocating write, which has no
 * special case code for sparse/nonsparse files.
 */
1611 1612 1613
static int ocfs2_expand_nonsparse_inode(struct inode *inode,
					struct buffer_head *di_bh,
					loff_t pos, unsigned len,
1614 1615 1616 1617 1618
					struct ocfs2_write_ctxt *wc)
{
	int ret;
	loff_t newsize = pos + len;

1619
	BUG_ON(ocfs2_sparse_alloc(OCFS2_SB(inode->i_sb)));
1620 1621 1622 1623

	if (newsize <= i_size_read(inode))
		return 0;

1624
	ret = ocfs2_extend_no_holes(inode, di_bh, newsize, pos);
1625 1626 1627
	if (ret)
		mlog_errno(ret);

1628 1629 1630 1631
	/* There is no wc if this is call from direct. */
	if (wc)
		wc->w_first_new_cpos =
			ocfs2_clusters_for_bytes(inode->i_sb, i_size_read(inode));
1632

1633 1634 1635
	return ret;
}

1636 1637 1638 1639 1640 1641 1642 1643 1644 1645 1646 1647
static int ocfs2_zero_tail(struct inode *inode, struct buffer_head *di_bh,
			   loff_t pos)
{
	int ret = 0;

	BUG_ON(!ocfs2_sparse_alloc(OCFS2_SB(inode->i_sb)));
	if (pos > i_size_read(inode))
		ret = ocfs2_zero_extend(inode, di_bh, pos);

	return ret;
}

1648 1649
int ocfs2_write_begin_nolock(struct address_space *mapping,
			     loff_t pos, unsigned len, ocfs2_write_type_t type,
1650 1651 1652
			     struct page **pagep, void **fsdata,
			     struct buffer_head *di_bh, struct page *mmap_page)
{
1653
	int ret, cluster_of_pages, credits = OCFS2_INODE_UPDATE_CREDITS;
1654
	unsigned int clusters_to_alloc, extents_to_split, clusters_need = 0;
1655 1656 1657 1658 1659 1660 1661
	struct ocfs2_write_ctxt *wc;
	struct inode *inode = mapping->host;
	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
	struct ocfs2_dinode *di;
	struct ocfs2_alloc_context *data_ac = NULL;
	struct ocfs2_alloc_context *meta_ac = NULL;
	handle_t *handle;
1662
	struct ocfs2_extent_tree et;
1663
	int try_free = 1, ret1;
1664

1665
try_again:
1666
	ret = ocfs2_alloc_write_ctxt(&wc, osb, pos, len, type, di_bh);
1667 1668 1669 1670 1671
	if (ret) {
		mlog_errno(ret);
		return ret;
	}

M
Mark Fasheh 已提交
1672 1673 1674 1675 1676 1677 1678 1679 1680 1681 1682 1683 1684
	if (ocfs2_supports_inline_data(osb)) {
		ret = ocfs2_try_to_write_inline_data(mapping, inode, pos, len,
						     mmap_page, wc);
		if (ret == 1) {
			ret = 0;
			goto success;
		}
		if (ret < 0) {
			mlog_errno(ret);
			goto out;
		}
	}

1685 1686 1687 1688 1689 1690 1691 1692 1693 1694 1695
	/* Direct io change i_size late, should not zero tail here. */
	if (type != OCFS2_WRITE_DIRECT) {
		if (ocfs2_sparse_alloc(osb))
			ret = ocfs2_zero_tail(inode, di_bh, pos);
		else
			ret = ocfs2_expand_nonsparse_inode(inode, di_bh, pos,
							   len, wc);
		if (ret) {
			mlog_errno(ret);
			goto out;
		}
1696 1697
	}

T
Tao Ma 已提交
1698 1699 1700 1701 1702
	ret = ocfs2_check_range_for_refcount(inode, pos, len);
	if (ret < 0) {
		mlog_errno(ret);
		goto out;
	} else if (ret == 1) {
1703
		clusters_need = wc->w_clen;
1704
		ret = ocfs2_refcount_cow(inode, di_bh,
1705
					 wc->w_cpos, wc->w_clen, UINT_MAX);
T
Tao Ma 已提交
1706 1707 1708 1709 1710 1711
		if (ret) {
			mlog_errno(ret);
			goto out;
		}
	}

1712 1713
	ret = ocfs2_populate_write_desc(inode, wc, &clusters_to_alloc,
					&extents_to_split);
1714 1715 1716 1717
	if (ret) {
		mlog_errno(ret);
		goto out;
	}
1718
	clusters_need += clusters_to_alloc;
1719 1720 1721

	di = (struct ocfs2_dinode *)wc->w_di_bh->b_data;

1722 1723 1724 1725
	trace_ocfs2_write_begin_nolock(
			(unsigned long long)OCFS2_I(inode)->ip_blkno,
			(long long)i_size_read(inode),
			le32_to_cpu(di->i_clusters),
1726
			pos, len, type, mmap_page,
1727 1728
			clusters_to_alloc, extents_to_split);

1729 1730 1731 1732 1733 1734
	/*
	 * We set w_target_from, w_target_to here so that
	 * ocfs2_write_end() knows which range in the target page to
	 * write out. An allocation requires that we write the entire
	 * cluster range.
	 */
1735
	if (clusters_to_alloc || extents_to_split) {
1736 1737
		/*
		 * XXX: We are stretching the limits of
1738
		 * ocfs2_lock_allocators(). It greatly over-estimates
1739 1740
		 * the work to be done.
		 */
1741 1742
		ocfs2_init_dinode_extent_tree(&et, INODE_CACHE(inode),
					      wc->w_di_bh);
1743
		ret = ocfs2_lock_allocators(inode, &et,
1744
					    clusters_to_alloc, extents_to_split,
1745
					    &data_ac, &meta_ac);
1746 1747
		if (ret) {
			mlog_errno(ret);
1748
			goto out;
1749 1750
		}

1751 1752 1753
		if (data_ac)
			data_ac->ac_resv = &OCFS2_I(inode)->ip_la_data_resv;

1754
		credits = ocfs2_calc_extend_credits(inode->i_sb,
1755
						    &di->id2.i_list);
1756 1757 1758
	} else if (type == OCFS2_WRITE_DIRECT)
		/* direct write needs not to start trans if no extents alloc. */
		goto success;
1759

1760 1761 1762 1763 1764
	/*
	 * We have to zero sparse allocated clusters, unwritten extent clusters,
	 * and non-sparse clusters we just extended.  For non-sparse writes,
	 * we know zeros will only be needed in the first and/or last cluster.
	 */
1765 1766
	if (wc->w_clen && (wc->w_desc[0].c_needs_zero ||
			   wc->w_desc[wc->w_clen - 1].c_needs_zero))
1767 1768 1769 1770 1771
		cluster_of_pages = 1;
	else
		cluster_of_pages = 0;

	ocfs2_set_target_boundaries(osb, wc, pos, len, cluster_of_pages);
1772

1773 1774 1775 1776
	handle = ocfs2_start_trans(osb, credits);
	if (IS_ERR(handle)) {
		ret = PTR_ERR(handle);
		mlog_errno(ret);
1777
		goto out;
1778 1779
	}

1780 1781
	wc->w_handle = handle;

1782 1783 1784 1785 1786
	if (clusters_to_alloc) {
		ret = dquot_alloc_space_nodirty(inode,
			ocfs2_clusters_to_bytes(osb->sb, clusters_to_alloc));
		if (ret)
			goto out_commit;
1787
	}
1788

1789
	ret = ocfs2_journal_access_di(handle, INODE_CACHE(inode), wc->w_di_bh,
1790
				      OCFS2_JOURNAL_ACCESS_WRITE);
1791
	if (ret) {
1792
		mlog_errno(ret);
1793
		goto out_quota;
1794 1795
	}

1796 1797 1798 1799 1800
	/*
	 * Fill our page array first. That way we've grabbed enough so
	 * that we can zero and flush if we error after adding the
	 * extent.
	 */
1801
	ret = ocfs2_grab_pages_for_write(mapping, wc, wc->w_cpos, pos, len,
1802
					 cluster_of_pages, mmap_page);
W
Wengang Wang 已提交
1803
	if (ret && ret != -EAGAIN) {
1804
		mlog_errno(ret);
1805
		goto out_quota;
1806 1807
	}

W
Wengang Wang 已提交
1808 1809 1810 1811 1812 1813 1814 1815 1816 1817 1818 1819
	/*
	 * ocfs2_grab_pages_for_write() returns -EAGAIN if it could not lock
	 * the target page. In this case, we exit with no error and no target
	 * page. This will trigger the caller, page_mkwrite(), to re-try
	 * the operation.
	 */
	if (ret == -EAGAIN) {
		BUG_ON(wc->w_target_page);
		ret = 0;
		goto out_quota;
	}

1820 1821 1822 1823
	ret = ocfs2_write_cluster_by_desc(mapping, data_ac, meta_ac, wc, pos,
					  len);
	if (ret) {
		mlog_errno(ret);
1824
		goto out_quota;
1825 1826
	}

1827 1828 1829 1830
	if (data_ac)
		ocfs2_free_alloc_context(data_ac);
	if (meta_ac)
		ocfs2_free_alloc_context(meta_ac);
1831

M
Mark Fasheh 已提交
1832
success:
1833 1834
	if (pagep)
		*pagep = wc->w_target_page;
1835 1836
	*fsdata = wc;
	return 0;
1837 1838
out_quota:
	if (clusters_to_alloc)
1839
		dquot_free_space(inode,
1840
			  ocfs2_clusters_to_bytes(osb->sb, clusters_to_alloc));
1841 1842 1843 1844
out_commit:
	ocfs2_commit_trans(osb, handle);

out:
1845 1846 1847 1848 1849 1850 1851 1852 1853 1854
	/*
	 * The mmapped page won't be unlocked in ocfs2_free_write_ctxt(),
	 * even in case of error here like ENOSPC and ENOMEM. So, we need
	 * to unlock the target page manually to prevent deadlocks when
	 * retrying again on ENOSPC, or when returning non-VM_FAULT_LOCKED
	 * to VM code.
	 */
	if (wc->w_target_locked)
		unlock_page(mmap_page);

1855
	ocfs2_free_write_ctxt(inode, wc);
1856

1857
	if (data_ac) {
1858
		ocfs2_free_alloc_context(data_ac);
1859 1860 1861
		data_ac = NULL;
	}
	if (meta_ac) {
1862
		ocfs2_free_alloc_context(meta_ac);
1863 1864
		meta_ac = NULL;
	}
1865 1866 1867 1868 1869 1870 1871 1872 1873 1874 1875 1876 1877 1878 1879 1880

	if (ret == -ENOSPC && try_free) {
		/*
		 * Try to free some truncate log so that we can have enough
		 * clusters to allocate.
		 */
		try_free = 0;

		ret1 = ocfs2_try_to_free_truncate_log(osb, clusters_need);
		if (ret1 == 1)
			goto try_again;

		if (ret1 < 0)
			mlog_errno(ret1);
	}

1881 1882 1883
	return ret;
}

N
Nick Piggin 已提交
1884 1885 1886
static int ocfs2_write_begin(struct file *file, struct address_space *mapping,
			     loff_t pos, unsigned len, unsigned flags,
			     struct page **pagep, void **fsdata)
1887 1888 1889 1890 1891
{
	int ret;
	struct buffer_head *di_bh = NULL;
	struct inode *inode = mapping->host;

M
Mark Fasheh 已提交
1892
	ret = ocfs2_inode_lock(inode, &di_bh, 1);
1893 1894 1895 1896 1897 1898 1899 1900 1901 1902 1903 1904 1905 1906
	if (ret) {
		mlog_errno(ret);
		return ret;
	}

	/*
	 * Take alloc sem here to prevent concurrent lookups. That way
	 * the mapping, zeroing and tree manipulation within
	 * ocfs2_write() will be safe against ->readpage(). This
	 * should also serve to lock out allocation from a shared
	 * writeable region.
	 */
	down_write(&OCFS2_I(inode)->ip_alloc_sem);

1907 1908
	ret = ocfs2_write_begin_nolock(mapping, pos, len, OCFS2_WRITE_BUFFER,
				       pagep, fsdata, di_bh, NULL);
1909 1910
	if (ret) {
		mlog_errno(ret);
M
Mark Fasheh 已提交
1911
		goto out_fail;
1912 1913 1914 1915 1916 1917 1918 1919 1920 1921
	}

	brelse(di_bh);

	return 0;

out_fail:
	up_write(&OCFS2_I(inode)->ip_alloc_sem);

	brelse(di_bh);
M
Mark Fasheh 已提交
1922
	ocfs2_inode_unlock(inode, 1);
1923 1924 1925 1926

	return ret;
}

M
Mark Fasheh 已提交
1927 1928 1929 1930 1931 1932 1933 1934 1935 1936 1937 1938 1939 1940
static void ocfs2_write_end_inline(struct inode *inode, loff_t pos,
				   unsigned len, unsigned *copied,
				   struct ocfs2_dinode *di,
				   struct ocfs2_write_ctxt *wc)
{
	void *kaddr;

	if (unlikely(*copied < len)) {
		if (!PageUptodate(wc->w_target_page)) {
			*copied = 0;
			return;
		}
	}

1941
	kaddr = kmap_atomic(wc->w_target_page);
M
Mark Fasheh 已提交
1942
	memcpy(di->id2.i_data.id_data + pos, kaddr + pos, *copied);
1943
	kunmap_atomic(kaddr);
M
Mark Fasheh 已提交
1944

1945 1946
	trace_ocfs2_write_end_inline(
	     (unsigned long long)OCFS2_I(inode)->ip_blkno,
M
Mark Fasheh 已提交
1947 1948 1949 1950 1951
	     (unsigned long long)pos, *copied,
	     le16_to_cpu(di->id2.i_data.id_count),
	     le16_to_cpu(di->i_dyn_features));
}

M
Mark Fasheh 已提交
1952 1953 1954
int ocfs2_write_end_nolock(struct address_space *mapping,
			   loff_t pos, unsigned len, unsigned copied,
			   struct page *page, void *fsdata)
1955
{
1956
	int i, ret;
1957
	unsigned from, to, start = pos & (PAGE_SIZE - 1);
1958 1959 1960 1961 1962 1963 1964
	struct inode *inode = mapping->host;
	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
	struct ocfs2_write_ctxt *wc = fsdata;
	struct ocfs2_dinode *di = (struct ocfs2_dinode *)wc->w_di_bh->b_data;
	handle_t *handle = wc->w_handle;
	struct page *tmppage;

1965 1966
	BUG_ON(!list_empty(&wc->w_unwritten_list));

1967 1968 1969 1970 1971 1972 1973 1974
	if (handle) {
		ret = ocfs2_journal_access_di(handle, INODE_CACHE(inode),
				wc->w_di_bh, OCFS2_JOURNAL_ACCESS_WRITE);
		if (ret) {
			copied = ret;
			mlog_errno(ret);
			goto out;
		}
1975 1976
	}

M
Mark Fasheh 已提交
1977 1978 1979 1980 1981
	if (OCFS2_I(inode)->ip_dyn_features & OCFS2_INLINE_DATA_FL) {
		ocfs2_write_end_inline(inode, pos, len, &copied, di, wc);
		goto out_write_size;
	}

1982
	if (unlikely(copied < len) && wc->w_target_page) {
1983 1984 1985 1986 1987 1988
		if (!PageUptodate(wc->w_target_page))
			copied = 0;

		ocfs2_zero_new_buffers(wc->w_target_page, start+copied,
				       start+len);
	}
1989 1990
	if (wc->w_target_page)
		flush_dcache_page(wc->w_target_page);
1991 1992 1993 1994

	for(i = 0; i < wc->w_num_pages; i++) {
		tmppage = wc->w_pages[i];

1995 1996 1997 1998
		/* This is the direct io target page. */
		if (tmppage == NULL)
			continue;

1999 2000 2001 2002
		if (tmppage == wc->w_target_page) {
			from = wc->w_target_from;
			to = wc->w_target_to;

2003 2004
			BUG_ON(from > PAGE_SIZE ||
			       to > PAGE_SIZE ||
2005 2006 2007 2008 2009 2010 2011 2012
			       to < from);
		} else {
			/*
			 * Pages adjacent to the target (if any) imply
			 * a hole-filling write in which case we want
			 * to flush their entire range.
			 */
			from = 0;
2013
			to = PAGE_SIZE;
2014 2015
		}

2016
		if (page_has_buffers(tmppage)) {
2017 2018
			if (handle && ocfs2_should_order_data(inode))
				ocfs2_jbd2_file_inode(handle, inode);
2019 2020
			block_commit_write(tmppage, from, to);
		}
2021 2022
	}

M
Mark Fasheh 已提交
2023
out_write_size:
2024 2025 2026 2027 2028 2029 2030 2031 2032
	/* Direct io do not update i_size here. */
	if (wc->w_type != OCFS2_WRITE_DIRECT) {
		pos += copied;
		if (pos > i_size_read(inode)) {
			i_size_write(inode, pos);
			mark_inode_dirty(inode);
		}
		inode->i_blocks = ocfs2_inode_sector_count(inode);
		di->i_size = cpu_to_le64((u64)i_size_read(inode));
2033
		inode->i_mtime = inode->i_ctime = current_time(inode);
2034 2035 2036 2037 2038 2039
		di->i_mtime = di->i_ctime = cpu_to_le64(inode->i_mtime.tv_sec);
		di->i_mtime_nsec = di->i_ctime_nsec = cpu_to_le32(inode->i_mtime.tv_nsec);
		ocfs2_update_inode_fsync_trans(handle, inode, 1);
	}
	if (handle)
		ocfs2_journal_dirty(handle, wc->w_di_bh);
2040

2041
out:
J
Junxiao Bi 已提交
2042 2043 2044 2045 2046 2047 2048
	/* unlock pages before dealloc since it needs acquiring j_trans_barrier
	 * lock, or it will cause a deadlock since journal commit threads holds
	 * this lock and will ask for the page lock when flushing the data.
	 * put it here to preserve the unlock order.
	 */
	ocfs2_unlock_pages(wc);

2049 2050
	if (handle)
		ocfs2_commit_trans(osb, handle);
2051

2052 2053
	ocfs2_run_deallocs(osb, &wc->w_dealloc);

J
Junxiao Bi 已提交
2054 2055
	brelse(wc->w_di_bh);
	kfree(wc);
2056 2057 2058 2059

	return copied;
}

N
Nick Piggin 已提交
2060 2061 2062
static int ocfs2_write_end(struct file *file, struct address_space *mapping,
			   loff_t pos, unsigned len, unsigned copied,
			   struct page *page, void *fsdata)
2063 2064 2065 2066 2067 2068
{
	int ret;
	struct inode *inode = mapping->host;

	ret = ocfs2_write_end_nolock(mapping, pos, len, copied, page, fsdata);

2069
	up_write(&OCFS2_I(inode)->ip_alloc_sem);
M
Mark Fasheh 已提交
2070
	ocfs2_inode_unlock(inode, 1);
2071

2072
	return ret;
2073 2074
}

2075 2076 2077 2078 2079 2080 2081 2082 2083 2084 2085 2086 2087 2088 2089 2090 2091 2092 2093 2094 2095 2096 2097 2098 2099 2100 2101 2102 2103 2104 2105 2106 2107 2108 2109 2110 2111 2112 2113 2114 2115 2116 2117 2118 2119 2120 2121 2122 2123 2124 2125 2126
struct ocfs2_dio_write_ctxt {
	struct list_head	dw_zero_list;
	unsigned		dw_zero_count;
	int			dw_orphaned;
	pid_t			dw_writer_pid;
};

static struct ocfs2_dio_write_ctxt *
ocfs2_dio_alloc_write_ctx(struct buffer_head *bh, int *alloc)
{
	struct ocfs2_dio_write_ctxt *dwc = NULL;

	if (bh->b_private)
		return bh->b_private;

	dwc = kmalloc(sizeof(struct ocfs2_dio_write_ctxt), GFP_NOFS);
	if (dwc == NULL)
		return NULL;
	INIT_LIST_HEAD(&dwc->dw_zero_list);
	dwc->dw_zero_count = 0;
	dwc->dw_orphaned = 0;
	dwc->dw_writer_pid = task_pid_nr(current);
	bh->b_private = dwc;
	*alloc = 1;

	return dwc;
}

static void ocfs2_dio_free_write_ctx(struct inode *inode,
				     struct ocfs2_dio_write_ctxt *dwc)
{
	ocfs2_free_unwritten_list(inode, &dwc->dw_zero_list);
	kfree(dwc);
}

/*
 * TODO: Make this into a generic get_blocks function.
 *
 * From do_direct_io in direct-io.c:
 *  "So what we do is to permit the ->get_blocks function to populate
 *   bh.b_size with the size of IO which is permitted at this offset and
 *   this i_blkbits."
 *
 * This function is called directly from get_more_blocks in direct-io.c.
 *
 * called like this: dio->get_blocks(dio->inode, fs_startblk,
 * 					fs_count, map_bh, dio->rw == WRITE);
 */
static int ocfs2_dio_get_block(struct inode *inode, sector_t iblock,
			       struct buffer_head *bh_result, int create)
{
	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
2127
	struct ocfs2_inode_info *oi = OCFS2_I(inode);
2128 2129 2130 2131 2132 2133 2134 2135 2136 2137 2138 2139 2140 2141 2142
	struct ocfs2_write_ctxt *wc;
	struct ocfs2_write_cluster_desc *desc = NULL;
	struct ocfs2_dio_write_ctxt *dwc = NULL;
	struct buffer_head *di_bh = NULL;
	u64 p_blkno;
	loff_t pos = iblock << inode->i_sb->s_blocksize_bits;
	unsigned len, total_len = bh_result->b_size;
	int ret = 0, first_get_block = 0;

	len = osb->s_clustersize - (pos & (osb->s_clustersize - 1));
	len = min(total_len, len);

	mlog(0, "get block of %lu at %llu:%u req %u\n",
			inode->i_ino, pos, len, total_len);

2143 2144 2145 2146 2147 2148 2149 2150 2151
	/*
	 * Because we need to change file size in ocfs2_dio_end_io_write(), or
	 * we may need to add it to orphan dir. So can not fall to fast path
	 * while file size will be changed.
	 */
	if (pos + total_len <= i_size_read(inode)) {
		down_read(&oi->ip_alloc_sem);
		/* This is the fast path for re-write. */
		ret = ocfs2_get_block(inode, iblock, bh_result, create);
2152

2153
		up_read(&oi->ip_alloc_sem);
2154

2155 2156 2157 2158
		if (buffer_mapped(bh_result) &&
		    !buffer_new(bh_result) &&
		    ret == 0)
			goto out;
2159

2160 2161 2162
		/* Clear state set by ocfs2_get_block. */
		bh_result->b_state = 0;
	}
2163 2164 2165 2166 2167 2168 2169 2170 2171 2172 2173 2174 2175 2176 2177 2178 2179 2180 2181 2182 2183 2184 2185 2186 2187 2188 2189 2190 2191 2192

	dwc = ocfs2_dio_alloc_write_ctx(bh_result, &first_get_block);
	if (unlikely(dwc == NULL)) {
		ret = -ENOMEM;
		mlog_errno(ret);
		goto out;
	}

	if (ocfs2_clusters_for_bytes(inode->i_sb, pos + total_len) >
	    ocfs2_clusters_for_bytes(inode->i_sb, i_size_read(inode)) &&
	    !dwc->dw_orphaned) {
		/*
		 * when we are going to alloc extents beyond file size, add the
		 * inode to orphan dir, so we can recall those spaces when
		 * system crashed during write.
		 */
		ret = ocfs2_add_inode_to_orphan(osb, inode);
		if (ret < 0) {
			mlog_errno(ret);
			goto out;
		}
		dwc->dw_orphaned = 1;
	}

	ret = ocfs2_inode_lock(inode, &di_bh, 1);
	if (ret) {
		mlog_errno(ret);
		goto out;
	}

2193 2194
	down_write(&oi->ip_alloc_sem);

2195 2196 2197 2198 2199 2200 2201 2202 2203 2204 2205 2206 2207 2208 2209 2210 2211 2212 2213 2214 2215 2216 2217 2218 2219 2220 2221 2222 2223 2224 2225 2226 2227 2228 2229 2230 2231 2232 2233 2234 2235 2236 2237 2238 2239 2240 2241 2242 2243 2244 2245 2246 2247
	if (first_get_block) {
		if (ocfs2_sparse_alloc(OCFS2_SB(inode->i_sb)))
			ret = ocfs2_zero_tail(inode, di_bh, pos);
		else
			ret = ocfs2_expand_nonsparse_inode(inode, di_bh, pos,
							   total_len, NULL);
		if (ret < 0) {
			mlog_errno(ret);
			goto unlock;
		}
	}

	ret = ocfs2_write_begin_nolock(inode->i_mapping, pos, len,
				       OCFS2_WRITE_DIRECT, NULL,
				       (void **)&wc, di_bh, NULL);
	if (ret) {
		mlog_errno(ret);
		goto unlock;
	}

	desc = &wc->w_desc[0];

	p_blkno = ocfs2_clusters_to_blocks(inode->i_sb, desc->c_phys);
	BUG_ON(p_blkno == 0);
	p_blkno += iblock & (u64)(ocfs2_clusters_to_blocks(inode->i_sb, 1) - 1);

	map_bh(bh_result, inode->i_sb, p_blkno);
	bh_result->b_size = len;
	if (desc->c_needs_zero)
		set_buffer_new(bh_result);

	/* May sleep in end_io. It should not happen in a irq context. So defer
	 * it to dio work queue. */
	set_buffer_defer_completion(bh_result);

	if (!list_empty(&wc->w_unwritten_list)) {
		struct ocfs2_unwritten_extent *ue = NULL;

		ue = list_first_entry(&wc->w_unwritten_list,
				      struct ocfs2_unwritten_extent,
				      ue_node);
		BUG_ON(ue->ue_cpos != desc->c_cpos);
		/* The physical address may be 0, fill it. */
		ue->ue_phys = desc->c_phys;

		list_splice_tail_init(&wc->w_unwritten_list, &dwc->dw_zero_list);
		dwc->dw_zero_count++;
	}

	ret = ocfs2_write_end_nolock(inode->i_mapping, pos, len, len, NULL, wc);
	BUG_ON(ret != len);
	ret = 0;
unlock:
2248
	up_write(&oi->ip_alloc_sem);
2249 2250 2251 2252 2253 2254 2255 2256 2257 2258 2259 2260 2261 2262 2263 2264
	ocfs2_inode_unlock(inode, 1);
	brelse(di_bh);
out:
	if (ret < 0)
		ret = -EIO;
	return ret;
}

static void ocfs2_dio_end_io_write(struct inode *inode,
				   struct ocfs2_dio_write_ctxt *dwc,
				   loff_t offset,
				   ssize_t bytes)
{
	struct ocfs2_cached_dealloc_ctxt dealloc;
	struct ocfs2_extent_tree et;
	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
2265
	struct ocfs2_inode_info *oi = OCFS2_I(inode);
2266 2267 2268 2269 2270 2271 2272 2273 2274 2275 2276 2277 2278 2279 2280 2281 2282 2283 2284 2285 2286
	struct ocfs2_unwritten_extent *ue = NULL;
	struct buffer_head *di_bh = NULL;
	struct ocfs2_dinode *di;
	struct ocfs2_alloc_context *data_ac = NULL;
	struct ocfs2_alloc_context *meta_ac = NULL;
	handle_t *handle = NULL;
	loff_t end = offset + bytes;
	int ret = 0, credits = 0, locked = 0;

	ocfs2_init_dealloc_ctxt(&dealloc);

	/* We do clear unwritten, delete orphan, change i_size here. If neither
	 * of these happen, we can skip all this. */
	if (list_empty(&dwc->dw_zero_list) &&
	    end <= i_size_read(inode) &&
	    !dwc->dw_orphaned)
		goto out;

	/* ocfs2_file_write_iter will get i_mutex, so we need not lock if we
	 * are in that context. */
	if (dwc->dw_writer_pid != task_pid_nr(current)) {
2287
		inode_lock(inode);
2288 2289 2290
		locked = 1;
	}

2291 2292 2293 2294 2295 2296 2297 2298
	ret = ocfs2_inode_lock(inode, &di_bh, 1);
	if (ret < 0) {
		mlog_errno(ret);
		goto out;
	}

	down_write(&oi->ip_alloc_sem);

2299 2300 2301 2302 2303 2304 2305 2306 2307 2308 2309 2310 2311 2312 2313 2314 2315 2316
	/* Delete orphan before acquire i_mutex. */
	if (dwc->dw_orphaned) {
		BUG_ON(dwc->dw_writer_pid != task_pid_nr(current));

		end = end > i_size_read(inode) ? end : 0;

		ret = ocfs2_del_inode_from_orphan(osb, inode, di_bh,
				!!end, end);
		if (ret < 0)
			mlog_errno(ret);
	}

	di = (struct ocfs2_dinode *)di_bh;

	ocfs2_init_dinode_extent_tree(&et, INODE_CACHE(inode), di_bh);

	ret = ocfs2_lock_allocators(inode, &et, 0, dwc->dw_zero_count*2,
				    &data_ac, &meta_ac);
2317 2318 2319 2320
	if (ret) {
		mlog_errno(ret);
		goto unlock;
	}
2321 2322 2323 2324 2325 2326 2327 2328 2329 2330 2331 2332 2333 2334 2335 2336 2337 2338 2339 2340 2341 2342 2343 2344 2345 2346 2347 2348 2349 2350 2351 2352 2353 2354 2355

	credits = ocfs2_calc_extend_credits(inode->i_sb, &di->id2.i_list);

	handle = ocfs2_start_trans(osb, credits);
	if (IS_ERR(handle)) {
		ret = PTR_ERR(handle);
		mlog_errno(ret);
		goto unlock;
	}
	ret = ocfs2_journal_access_di(handle, INODE_CACHE(inode), di_bh,
				      OCFS2_JOURNAL_ACCESS_WRITE);
	if (ret) {
		mlog_errno(ret);
		goto commit;
	}

	list_for_each_entry(ue, &dwc->dw_zero_list, ue_node) {
		ret = ocfs2_mark_extent_written(inode, &et, handle,
						ue->ue_cpos, 1,
						ue->ue_phys,
						meta_ac, &dealloc);
		if (ret < 0) {
			mlog_errno(ret);
			break;
		}
	}

	if (end > i_size_read(inode)) {
		ret = ocfs2_set_inode_size(handle, inode, di_bh, end);
		if (ret < 0)
			mlog_errno(ret);
	}
commit:
	ocfs2_commit_trans(osb, handle);
unlock:
2356
	up_write(&oi->ip_alloc_sem);
2357 2358 2359 2360 2361 2362 2363
	ocfs2_inode_unlock(inode, 1);
	brelse(di_bh);
out:
	if (data_ac)
		ocfs2_free_alloc_context(data_ac);
	if (meta_ac)
		ocfs2_free_alloc_context(meta_ac);
2364 2365
	ocfs2_run_deallocs(osb, &dealloc);
	if (locked)
2366
		inode_unlock(inode);
2367
	ocfs2_dio_free_write_ctx(inode, dwc);
2368 2369 2370 2371 2372 2373 2374 2375 2376 2377 2378 2379 2380 2381 2382 2383 2384 2385 2386 2387 2388 2389 2390 2391 2392 2393 2394 2395 2396 2397 2398
}

/*
 * ocfs2_dio_end_io is called by the dio core when a dio is finished.  We're
 * particularly interested in the aio/dio case.  We use the rw_lock DLM lock
 * to protect io on one node from truncation on another.
 */
static int ocfs2_dio_end_io(struct kiocb *iocb,
			    loff_t offset,
			    ssize_t bytes,
			    void *private)
{
	struct inode *inode = file_inode(iocb->ki_filp);
	int level;

	if (bytes <= 0)
		return 0;

	/* this io's submitter should not have unlocked this before we could */
	BUG_ON(!ocfs2_iocb_is_rw_locked(iocb));

	if (private)
		ocfs2_dio_end_io_write(inode, private, offset, bytes);

	ocfs2_iocb_clear_rw_locked(iocb);

	level = ocfs2_iocb_rw_locked_level(iocb);
	ocfs2_rw_unlock(inode, level);
	return 0;
}

2399
static ssize_t ocfs2_direct_IO(struct kiocb *iocb, struct iov_iter *iter)
2400 2401
{
	struct file *file = iocb->ki_filp;
2402
	struct inode *inode = file->f_mapping->host;
2403 2404 2405 2406 2407 2408 2409 2410 2411 2412 2413
	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
	get_block_t *get_block;

	/*
	 * Fallback to buffered I/O if we see an inode without
	 * extents.
	 */
	if (OCFS2_I(inode)->ip_dyn_features & OCFS2_INLINE_DATA_FL)
		return 0;

	/* Fallback to buffered I/O if we do not support append dio. */
2414 2415
	if (iocb->ki_pos + iter->count > i_size_read(inode) &&
	    !ocfs2_supports_append_dio(osb))
2416 2417 2418 2419 2420 2421 2422 2423
		return 0;

	if (iov_iter_rw(iter) == READ)
		get_block = ocfs2_get_block;
	else
		get_block = ocfs2_dio_get_block;

	return __blockdev_direct_IO(iocb, inode, inode->i_sb->s_bdev,
2424
				    iter, get_block,
2425 2426 2427
				    ocfs2_dio_end_io, NULL, 0);
}

2428
const struct address_space_operations ocfs2_aops = {
2429 2430 2431 2432 2433 2434 2435
	.readpage		= ocfs2_readpage,
	.readpages		= ocfs2_readpages,
	.writepage		= ocfs2_writepage,
	.write_begin		= ocfs2_write_begin,
	.write_end		= ocfs2_write_end,
	.bmap			= ocfs2_bmap,
	.direct_IO		= ocfs2_direct_IO,
2436
	.invalidatepage		= block_invalidatepage,
2437 2438 2439
	.releasepage		= ocfs2_releasepage,
	.migratepage		= buffer_migrate_page,
	.is_partially_uptodate	= block_is_partially_uptodate,
2440
	.error_remove_page	= generic_error_remove_page,
2441
};