file.c 25.4 KB
Newer Older
D
David Teigland 已提交
1 2
/*
 * Copyright (C) Sistina Software, Inc.  1997-2003 All rights reserved.
3
 * Copyright (C) 2004-2006 Red Hat, Inc.  All rights reserved.
D
David Teigland 已提交
4 5 6
 *
 * This copyrighted material is made available to anyone wishing to use,
 * modify, copy, or redistribute it subject to the terms and conditions
7
 * of the GNU General Public License version 2.
D
David Teigland 已提交
8 9 10 11 12 13 14 15 16 17
 */

#include <linux/slab.h>
#include <linux/spinlock.h>
#include <linux/completion.h>
#include <linux/buffer_head.h>
#include <linux/pagemap.h>
#include <linux/uio.h>
#include <linux/blkdev.h>
#include <linux/mm.h>
M
Miklos Szeredi 已提交
18
#include <linux/mount.h>
19
#include <linux/fs.h>
20
#include <linux/gfs2_ondisk.h>
21
#include <linux/ext2_fs.h>
22 23
#include <linux/falloc.h>
#include <linux/swap.h>
24
#include <linux/crc32.h>
25
#include <linux/writeback.h>
D
David Teigland 已提交
26
#include <asm/uaccess.h>
27 28
#include <linux/dlm.h>
#include <linux/dlm_plock.h>
D
David Teigland 已提交
29 30

#include "gfs2.h"
31
#include "incore.h"
D
David Teigland 已提交
32 33 34 35 36 37 38 39 40 41
#include "bmap.h"
#include "dir.h"
#include "glock.h"
#include "glops.h"
#include "inode.h"
#include "log.h"
#include "meta_io.h"
#include "quota.h"
#include "rgrp.h"
#include "trans.h"
42
#include "util.h"
D
David Teigland 已提交
43 44 45 46 47 48 49 50 51 52 53 54 55 56 57

/**
 * gfs2_llseek - seek to a location in a file
 * @file: the file
 * @offset: the offset
 * @origin: Where to seek from (SEEK_SET, SEEK_CUR, or SEEK_END)
 *
 * SEEK_END requires the glock for the file because it references the
 * file's size.
 *
 * Returns: The new offset, or errno
 */

static loff_t gfs2_llseek(struct file *file, loff_t offset, int origin)
{
58
	struct gfs2_inode *ip = GFS2_I(file->f_mapping->host);
D
David Teigland 已提交
59 60 61
	struct gfs2_holder i_gh;
	loff_t error;

62 63 64 65
	switch (origin) {
	case SEEK_END: /* These reference inode->i_size */
	case SEEK_DATA:
	case SEEK_HOLE:
D
David Teigland 已提交
66 67 68
		error = gfs2_glock_nq_init(ip->i_gl, LM_ST_SHARED, LM_FLAG_ANY,
					   &i_gh);
		if (!error) {
69
			error = generic_file_llseek(file, offset, origin);
D
David Teigland 已提交
70 71
			gfs2_glock_dq_uninit(&i_gh);
		}
72 73 74
		break;
	case SEEK_CUR:
	case SEEK_SET:
75
		error = generic_file_llseek(file, offset, origin);
76 77 78 79
		break;
	default:
		error = -EINVAL;
	}
D
David Teigland 已提交
80 81 82 83 84

	return error;
}

/**
85
 * gfs2_readdir - Read directory entries from a directory
D
David Teigland 已提交
86 87 88 89 90 91 92
 * @file: The directory to read from
 * @dirent: Buffer for dirents
 * @filldir: Function used to do the copying
 *
 * Returns: errno
 */

93
static int gfs2_readdir(struct file *file, void *dirent, filldir_t filldir)
D
David Teigland 已提交
94
{
95
	struct inode *dir = file->f_mapping->host;
96
	struct gfs2_inode *dip = GFS2_I(dir);
D
David Teigland 已提交
97
	struct gfs2_holder d_gh;
98
	u64 offset = file->f_pos;
D
David Teigland 已提交
99 100
	int error;

101 102
	gfs2_holder_init(dip->i_gl, LM_ST_SHARED, 0, &d_gh);
	error = gfs2_glock_nq(&d_gh);
D
David Teigland 已提交
103 104 105 106 107
	if (error) {
		gfs2_holder_uninit(&d_gh);
		return error;
	}

108
	error = gfs2_dir_read(dir, &offset, dirent, filldir);
D
David Teigland 已提交
109 110 111 112 113 114 115 116

	gfs2_glock_dq_uninit(&d_gh);

	file->f_pos = offset;

	return error;
}

117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137
/**
 * fsflags_cvt
 * @table: A table of 32 u32 flags
 * @val: a 32 bit value to convert
 *
 * This function can be used to convert between fsflags values and
 * GFS2's own flags values.
 *
 * Returns: the converted flags
 */
static u32 fsflags_cvt(const u32 *table, u32 val)
{
	u32 res = 0;
	while(val) {
		if (val & 1)
			res |= *table;
		table++;
		val >>= 1;
	}
	return res;
}
D
David Teigland 已提交
138

139 140 141 142 143 144
static const u32 fsflags_to_gfs2[32] = {
	[3] = GFS2_DIF_SYNC,
	[4] = GFS2_DIF_IMMUTABLE,
	[5] = GFS2_DIF_APPENDONLY,
	[7] = GFS2_DIF_NOATIME,
	[12] = GFS2_DIF_EXHASH,
145
	[14] = GFS2_DIF_INHERIT_JDATA,
146 147
};

148 149 150 151 152 153 154
static const u32 gfs2_to_fsflags[32] = {
	[gfs2fl_Sync] = FS_SYNC_FL,
	[gfs2fl_Immutable] = FS_IMMUTABLE_FL,
	[gfs2fl_AppendOnly] = FS_APPEND_FL,
	[gfs2fl_NoAtime] = FS_NOATIME_FL,
	[gfs2fl_ExHash] = FS_INDEX_FL,
	[gfs2fl_InheritJdata] = FS_JOURNAL_DATA_FL,
155
};
156

157
static int gfs2_get_flags(struct file *filp, u32 __user *ptr)
158
{
J
Josef Sipek 已提交
159
	struct inode *inode = filp->f_path.dentry->d_inode;
160
	struct gfs2_inode *ip = GFS2_I(inode);
161 162
	struct gfs2_holder gh;
	int error;
163
	u32 fsflags;
164

165 166
	gfs2_holder_init(ip->i_gl, LM_ST_SHARED, 0, &gh);
	error = gfs2_glock_nq(&gh);
167 168
	if (error)
		return error;
169

170 171
	fsflags = fsflags_cvt(gfs2_to_fsflags, ip->i_diskflags);
	if (!S_ISDIR(inode->i_mode) && ip->i_diskflags & GFS2_DIF_JDATA)
172
		fsflags |= FS_JOURNAL_DATA_FL;
173
	if (put_user(fsflags, ptr))
174 175
		error = -EFAULT;

176
	gfs2_glock_dq(&gh);
177 178 179 180
	gfs2_holder_uninit(&gh);
	return error;
}

181 182 183 184 185
void gfs2_set_inode_flags(struct inode *inode)
{
	struct gfs2_inode *ip = GFS2_I(inode);
	unsigned int flags = inode->i_flags;

S
Steven Whitehouse 已提交
186 187 188
	flags &= ~(S_SYNC|S_APPEND|S_IMMUTABLE|S_NOATIME|S_DIRSYNC|S_NOSEC);
	if ((ip->i_eattr == 0) && !is_sxid(inode->i_mode))
		inode->i_flags |= S_NOSEC;
189
	if (ip->i_diskflags & GFS2_DIF_IMMUTABLE)
190
		flags |= S_IMMUTABLE;
191
	if (ip->i_diskflags & GFS2_DIF_APPENDONLY)
192
		flags |= S_APPEND;
193
	if (ip->i_diskflags & GFS2_DIF_NOATIME)
194
		flags |= S_NOATIME;
195
	if (ip->i_diskflags & GFS2_DIF_SYNC)
196 197 198 199
		flags |= S_SYNC;
	inode->i_flags = flags;
}

200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215
/* Flags that can be set by user space */
#define GFS2_FLAGS_USER_SET (GFS2_DIF_JDATA|			\
			     GFS2_DIF_IMMUTABLE|		\
			     GFS2_DIF_APPENDONLY|		\
			     GFS2_DIF_NOATIME|			\
			     GFS2_DIF_SYNC|			\
			     GFS2_DIF_SYSTEM|			\
			     GFS2_DIF_INHERIT_JDATA)

/**
 * gfs2_set_flags - set flags on an inode
 * @inode: The inode
 * @flags: The flags to set
 * @mask: Indicates which flags are valid
 *
 */
216
static int do_gfs2_set_flags(struct file *filp, u32 reqflags, u32 mask)
217
{
J
Josef Sipek 已提交
218
	struct inode *inode = filp->f_path.dentry->d_inode;
219 220
	struct gfs2_inode *ip = GFS2_I(inode);
	struct gfs2_sbd *sdp = GFS2_SB(inode);
221 222 223
	struct buffer_head *bh;
	struct gfs2_holder gh;
	int error;
224
	u32 new_flags, flags;
225

226
	error = mnt_want_write_file(filp);
227
	if (error)
228 229
		return error;

M
Miklos Szeredi 已提交
230 231 232 233
	error = gfs2_glock_nq_init(ip->i_gl, LM_ST_EXCLUSIVE, 0, &gh);
	if (error)
		goto out_drop_write;

234
	error = -EACCES;
235
	if (!inode_owner_or_capable(inode))
236 237 238
		goto out;

	error = 0;
239
	flags = ip->i_diskflags;
240
	new_flags = (flags & ~mask) | (reqflags & mask);
241 242 243 244 245 246 247 248 249 250 251 252
	if ((new_flags ^ flags) == 0)
		goto out;

	error = -EINVAL;
	if ((new_flags ^ flags) & ~GFS2_FLAGS_USER_SET)
		goto out;

	error = -EPERM;
	if (IS_IMMUTABLE(inode) && (new_flags & GFS2_DIF_IMMUTABLE))
		goto out;
	if (IS_APPEND(inode) && (new_flags & GFS2_DIF_APPENDONLY))
		goto out;
253
	if (((new_flags ^ flags) & GFS2_DIF_IMMUTABLE) &&
254
	    !capable(CAP_LINUX_IMMUTABLE))
255
		goto out;
256
	if (!IS_IMMUTABLE(inode)) {
257
		error = gfs2_permission(inode, MAY_WRITE);
258 259 260
		if (error)
			goto out;
	}
261 262 263 264 265 266 267 268 269 270
	if ((flags ^ new_flags) & GFS2_DIF_JDATA) {
		if (flags & GFS2_DIF_JDATA)
			gfs2_log_flush(sdp, ip->i_gl);
		error = filemap_fdatawrite(inode->i_mapping);
		if (error)
			goto out;
		error = filemap_fdatawait(inode->i_mapping);
		if (error)
			goto out;
	}
271
	error = gfs2_trans_begin(sdp, RES_DINODE, 0);
272 273
	if (error)
		goto out;
274 275 276
	error = gfs2_meta_inode_buffer(ip, &bh);
	if (error)
		goto out_trans_end;
277
	gfs2_trans_add_bh(ip->i_gl, bh, 1);
278
	ip->i_diskflags = new_flags;
279
	gfs2_dinode_out(ip, bh->b_data);
280
	brelse(bh);
281
	gfs2_set_inode_flags(inode);
282
	gfs2_set_aops(inode);
283 284
out_trans_end:
	gfs2_trans_end(sdp);
285 286
out:
	gfs2_glock_dq_uninit(&gh);
M
Miklos Szeredi 已提交
287 288
out_drop_write:
	mnt_drop_write(filp->f_path.mnt);
289 290 291
	return error;
}

292
static int gfs2_set_flags(struct file *filp, u32 __user *ptr)
293
{
294
	struct inode *inode = filp->f_path.dentry->d_inode;
295
	u32 fsflags, gfsflags;
296

297
	if (get_user(fsflags, ptr))
298
		return -EFAULT;
299

300
	gfsflags = fsflags_cvt(fsflags_to_gfs2, fsflags);
301 302 303 304 305 306
	if (!S_ISDIR(inode->i_mode)) {
		if (gfsflags & GFS2_DIF_INHERIT_JDATA)
			gfsflags ^= (GFS2_DIF_JDATA | GFS2_DIF_INHERIT_JDATA);
		return do_gfs2_set_flags(filp, gfsflags, ~0);
	}
	return do_gfs2_set_flags(filp, gfsflags, ~GFS2_DIF_JDATA);
307 308
}

309
static long gfs2_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
310 311
{
	switch(cmd) {
312
	case FS_IOC_GETFLAGS:
313
		return gfs2_get_flags(filp, (u32 __user *)arg);
314
	case FS_IOC_SETFLAGS:
315
		return gfs2_set_flags(filp, (u32 __user *)arg);
316 317 318 319
	}
	return -ENOTTY;
}

320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339
/**
 * gfs2_allocate_page_backing - Use bmap to allocate blocks
 * @page: The (locked) page to allocate backing for
 *
 * We try to allocate all the blocks required for the page in
 * one go. This might fail for various reasons, so we keep
 * trying until all the blocks to back this page are allocated.
 * If some of the blocks are already allocated, thats ok too.
 */

static int gfs2_allocate_page_backing(struct page *page)
{
	struct inode *inode = page->mapping->host;
	struct buffer_head bh;
	unsigned long size = PAGE_CACHE_SIZE;
	u64 lblock = page->index << (PAGE_CACHE_SHIFT - inode->i_blkbits);

	do {
		bh.b_state = 0;
		bh.b_size = size;
340
		gfs2_block_map(inode, lblock, &bh, 1);
341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357
		if (!buffer_mapped(&bh))
			return -EIO;
		size -= bh.b_size;
		lblock += (bh.b_size >> inode->i_blkbits);
	} while(size > 0);
	return 0;
}

/**
 * gfs2_page_mkwrite - Make a shared, mmap()ed, page writable
 * @vma: The virtual memory area
 * @page: The page which is about to become writable
 *
 * When the page becomes writable, we need to ensure that we have
 * blocks allocated on disk to back that page.
 */

358
static int gfs2_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf)
359
{
360
	struct page *page = vmf->page;
361 362 363 364
	struct inode *inode = vma->vm_file->f_path.dentry->d_inode;
	struct gfs2_inode *ip = GFS2_I(inode);
	struct gfs2_sbd *sdp = GFS2_SB(inode);
	unsigned long last_index;
365
	u64 pos = page->index << PAGE_CACHE_SHIFT;
366 367 368
	unsigned int data_blocks, ind_blocks, rblocks;
	struct gfs2_holder gh;
	struct gfs2_alloc *al;
S
Steven Whitehouse 已提交
369
	loff_t size;
370 371
	int ret;

S
Steven Whitehouse 已提交
372 373 374 375 376 377
	/* Wait if fs is frozen. This is racy so we check again later on
	 * and retry if the fs has been frozen after the page lock has
	 * been acquired
	 */
	vfs_check_frozen(inode->i_sb, SB_FREEZE_WRITE);

378 379
	gfs2_holder_init(ip->i_gl, LM_ST_EXCLUSIVE, 0, &gh);
	ret = gfs2_glock_nq(&gh);
380 381 382
	if (ret)
		goto out;

383 384 385
	set_bit(GLF_DIRTY, &ip->i_gl->gl_flags);
	set_bit(GIF_SW_PAGED, &ip->i_flags);

S
Steven Whitehouse 已提交
386 387 388 389 390 391
	if (!gfs2_write_alloc_required(ip, pos, PAGE_CACHE_SIZE)) {
		lock_page(page);
		if (!PageUptodate(page) || page->mapping != inode->i_mapping) {
			ret = -EAGAIN;
			unlock_page(page);
		}
392
		goto out_unlock;
S
Steven Whitehouse 已提交
393 394
	}

395
	ret = -ENOMEM;
396
	al = gfs2_alloc_get(ip);
397 398 399
	if (al == NULL)
		goto out_unlock;

400
	ret = gfs2_quota_lock_check(ip);
401 402
	if (ret)
		goto out_alloc_put;
403
	gfs2_write_calc_reserv(ip, PAGE_CACHE_SIZE, &data_blocks, &ind_blocks);
404 405 406 407 408 409 410 411
	al->al_requested = data_blocks + ind_blocks;
	ret = gfs2_inplace_reserve(ip);
	if (ret)
		goto out_quota_unlock;

	rblocks = RES_DINODE + ind_blocks;
	if (gfs2_is_jdata(ip))
		rblocks += data_blocks ? data_blocks : 1;
412
	if (ind_blocks || data_blocks) {
413
		rblocks += RES_STATFS + RES_QUOTA;
414
		rblocks += gfs2_rg_blocks(ip);
415
	}
416 417 418 419 420 421
	ret = gfs2_trans_begin(sdp, rblocks, 0);
	if (ret)
		goto out_trans_fail;

	lock_page(page);
	ret = -EINVAL;
S
Steven Whitehouse 已提交
422 423 424 425 426 427 428 429 430 431 432 433 434 435
	size = i_size_read(inode);
	last_index = (size - 1) >> PAGE_CACHE_SHIFT;
	/* Check page index against inode size */
	if (size == 0 || (page->index > last_index))
		goto out_trans_end;

	ret = -EAGAIN;
	/* If truncated, we must retry the operation, we may have raced
	 * with the glock demotion code.
	 */
	if (!PageUptodate(page) || page->mapping != inode->i_mapping)
		goto out_trans_end;

	/* Unstuff, if required, and allocate backing blocks for page */
436
	ret = 0;
S
Steven Whitehouse 已提交
437
	if (gfs2_is_stuffed(ip))
438
		ret = gfs2_unstuff_dinode(ip, page);
S
Steven Whitehouse 已提交
439 440
	if (ret == 0)
		ret = gfs2_allocate_page_backing(page);
441

S
Steven Whitehouse 已提交
442 443 444
out_trans_end:
	if (ret)
		unlock_page(page);
445 446 447 448 449 450 451 452 453 454 455
	gfs2_trans_end(sdp);
out_trans_fail:
	gfs2_inplace_release(ip);
out_quota_unlock:
	gfs2_quota_unlock(ip);
out_alloc_put:
	gfs2_alloc_put(ip);
out_unlock:
	gfs2_glock_dq(&gh);
out:
	gfs2_holder_uninit(&gh);
S
Steven Whitehouse 已提交
456 457 458 459 460 461 462 463 464 465 466
	if (ret == 0) {
		set_page_dirty(page);
		/* This check must be post dropping of transaction lock */
		if (inode->i_sb->s_frozen == SB_UNFROZEN) {
			wait_on_page_writeback(page);
		} else {
			ret = -EAGAIN;
			unlock_page(page);
		}
	}
	return block_page_mkwrite_return(ret);
467 468
}

469
static const struct vm_operations_struct gfs2_vm_ops = {
470 471 472 473
	.fault = filemap_fault,
	.page_mkwrite = gfs2_page_mkwrite,
};

D
David Teigland 已提交
474 475 476 477 478
/**
 * gfs2_mmap -
 * @file: The file to map
 * @vma: The VMA which described the mapping
 *
479 480 481 482 483
 * There is no need to get a lock here unless we should be updating
 * atime. We ignore any locking errors since the only consequence is
 * a missed atime update (which will just be deferred until later).
 *
 * Returns: 0
D
David Teigland 已提交
484 485 486 487
 */

static int gfs2_mmap(struct file *file, struct vm_area_struct *vma)
{
488
	struct gfs2_inode *ip = GFS2_I(file->f_mapping->host);
D
David Teigland 已提交
489

490 491
	if (!(file->f_flags & O_NOATIME) &&
	    !IS_NOATIME(&ip->i_inode)) {
492 493
		struct gfs2_holder i_gh;
		int error;
D
David Teigland 已提交
494

495
		gfs2_holder_init(ip->i_gl, LM_ST_SHARED, LM_FLAG_ANY, &i_gh);
496
		error = gfs2_glock_nq(&i_gh);
497 498 499 500 501 502 503
		if (error == 0) {
			file_accessed(file);
			gfs2_glock_dq(&i_gh);
		}
		gfs2_holder_uninit(&i_gh);
		if (error)
			return error;
504
	}
505
	vma->vm_ops = &gfs2_vm_ops;
506
	vma->vm_flags |= VM_CAN_NONLINEAR;
D
David Teigland 已提交
507

508
	return 0;
D
David Teigland 已提交
509 510 511 512 513 514 515 516 517 518 519 520
}

/**
 * gfs2_open - open a file
 * @inode: the inode to open
 * @file: the struct file for this opening
 *
 * Returns: errno
 */

static int gfs2_open(struct inode *inode, struct file *file)
{
521
	struct gfs2_inode *ip = GFS2_I(inode);
D
David Teigland 已提交
522 523 524 525 526 527 528 529
	struct gfs2_holder i_gh;
	struct gfs2_file *fp;
	int error;

	fp = kzalloc(sizeof(struct gfs2_file), GFP_KERNEL);
	if (!fp)
		return -ENOMEM;

530
	mutex_init(&fp->f_fl_mutex);
D
David Teigland 已提交
531

532
	gfs2_assert_warn(GFS2_SB(inode), !file->private_data);
533
	file->private_data = fp;
D
David Teigland 已提交
534

535
	if (S_ISREG(ip->i_inode.i_mode)) {
D
David Teigland 已提交
536 537 538 539 540 541
		error = gfs2_glock_nq_init(ip->i_gl, LM_ST_SHARED, LM_FLAG_ANY,
					   &i_gh);
		if (error)
			goto fail;

		if (!(file->f_flags & O_LARGEFILE) &&
S
Steven Whitehouse 已提交
542
		    i_size_read(inode) > MAX_NON_LFS) {
543
			error = -EOVERFLOW;
D
David Teigland 已提交
544 545 546 547 548 549 550 551
			goto fail_gunlock;
		}

		gfs2_glock_dq_uninit(&i_gh);
	}

	return 0;

552
fail_gunlock:
D
David Teigland 已提交
553
	gfs2_glock_dq_uninit(&i_gh);
554
fail:
555
	file->private_data = NULL;
D
David Teigland 已提交
556 557 558 559 560 561 562 563 564 565 566 567 568 569
	kfree(fp);
	return error;
}

/**
 * gfs2_close - called to close a struct file
 * @inode: the inode the struct file belongs to
 * @file: the struct file being closed
 *
 * Returns: errno
 */

static int gfs2_close(struct inode *inode, struct file *file)
{
570
	struct gfs2_sbd *sdp = inode->i_sb->s_fs_info;
D
David Teigland 已提交
571 572
	struct gfs2_file *fp;

573 574
	fp = file->private_data;
	file->private_data = NULL;
D
David Teigland 已提交
575 576 577 578 579 580 581 582 583 584 585

	if (gfs2_assert_warn(sdp, fp))
		return -EIO;

	kfree(fp);

	return 0;
}

/**
 * gfs2_fsync - sync the dirty data for a file (across the cluster)
586 587 588
 * @file: the file that points to the dentry
 * @start: the start position in the file to sync
 * @end: the end position in the file to sync
S
Steven Whitehouse 已提交
589
 * @datasync: set if we can ignore timestamp changes
D
David Teigland 已提交
590
 *
591 592 593 594 595 596 597 598 599 600
 * We split the data flushing here so that we don't wait for the data
 * until after we've also sent the metadata to disk. Note that for
 * data=ordered, we will write & wait for the data at the log flush
 * stage anyway, so this is unlikely to make much of a difference
 * except in the data=writeback case.
 *
 * If the fdatawrite fails due to any reason except -EIO, we will
 * continue the remainder of the fsync, although we'll still report
 * the error at the end. This is to match filemap_write_and_wait_range()
 * behaviour.
601
 *
D
David Teigland 已提交
602 603 604
 * Returns: errno
 */

605 606
static int gfs2_fsync(struct file *file, loff_t start, loff_t end,
		      int datasync)
D
David Teigland 已提交
607
{
608 609
	struct address_space *mapping = file->f_mapping;
	struct inode *inode = mapping->host;
610
	int sync_state = inode->i_state & (I_DIRTY_SYNC|I_DIRTY_DATASYNC);
S
Steven Whitehouse 已提交
611
	struct gfs2_inode *ip = GFS2_I(inode);
612
	int ret, ret1 = 0;
D
David Teigland 已提交
613

614 615 616 617 618
	if (mapping->nrpages) {
		ret1 = filemap_fdatawrite_range(mapping, start, end);
		if (ret1 == -EIO)
			return ret1;
	}
619

S
Steven Whitehouse 已提交
620 621
	if (datasync)
		sync_state &= ~I_DIRTY_SYNC;
D
David Teigland 已提交
622

S
Steven Whitehouse 已提交
623 624
	if (sync_state) {
		ret = sync_inode_metadata(inode, 1);
625
		if (ret)
S
Steven Whitehouse 已提交
626
			return ret;
627 628
		if (gfs2_is_jdata(ip))
			filemap_write_and_wait(mapping);
629
		gfs2_ail_flush(ip->i_gl, 1);
630 631
	}

632 633 634 635
	if (mapping->nrpages)
		ret = filemap_fdatawait_range(mapping, start, end);

	return ret ? ret : ret1;
D
David Teigland 已提交
636 637
}

638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671
/**
 * gfs2_file_aio_write - Perform a write to a file
 * @iocb: The io context
 * @iov: The data to write
 * @nr_segs: Number of @iov segments
 * @pos: The file position
 *
 * We have to do a lock/unlock here to refresh the inode size for
 * O_APPEND writes, otherwise we can land up writing at the wrong
 * offset. There is still a race, but provided the app is using its
 * own file locking, this will make O_APPEND work as expected.
 *
 */

static ssize_t gfs2_file_aio_write(struct kiocb *iocb, const struct iovec *iov,
				   unsigned long nr_segs, loff_t pos)
{
	struct file *file = iocb->ki_filp;

	if (file->f_flags & O_APPEND) {
		struct dentry *dentry = file->f_dentry;
		struct gfs2_inode *ip = GFS2_I(dentry->d_inode);
		struct gfs2_holder gh;
		int ret;

		ret = gfs2_glock_nq_init(ip->i_gl, LM_ST_SHARED, 0, &gh);
		if (ret)
			return ret;
		gfs2_glock_dq_uninit(&gh);
	}

	return generic_file_aio_write(iocb, iov, nr_segs, pos);
}

672 673 674 675 676 677
static int fallocate_chunk(struct inode *inode, loff_t offset, loff_t len,
			   int mode)
{
	struct gfs2_inode *ip = GFS2_I(inode);
	struct buffer_head *dibh;
	int error;
678 679
	unsigned int nr_blks;
	sector_t lblock = offset >> inode->i_blkbits;
680 681 682

	error = gfs2_meta_inode_buffer(ip, &dibh);
	if (unlikely(error))
683
		return error;
684 685 686 687 688 689 690 691 692

	gfs2_trans_add_bh(ip->i_gl, dibh, 1);

	if (gfs2_is_stuffed(ip)) {
		error = gfs2_unstuff_dinode(ip, NULL);
		if (unlikely(error))
			goto out;
	}

693 694 695 696
	while (len) {
		struct buffer_head bh_map = { .b_state = 0, .b_blocknr = 0 };
		bh_map.b_size = len;
		set_buffer_zeronew(&bh_map);
697

698 699 700 701 702 703 704 705 706 707
		error = gfs2_block_map(inode, lblock, &bh_map, 1);
		if (unlikely(error))
			goto out;
		len -= bh_map.b_size;
		nr_blks = bh_map.b_size >> inode->i_blkbits;
		lblock += nr_blks;
		if (!buffer_new(&bh_map))
			continue;
		if (unlikely(!buffer_zeronew(&bh_map))) {
			error = -EIO;
708
			goto out;
709
		}
710
	}
711 712
	if (offset + len > inode->i_size && !(mode & FALLOC_FL_KEEP_SIZE))
		i_size_write(inode, offset + len);
713 714 715 716

	mark_inode_dirty(inode);

out:
717
	brelse(dibh);
718 719 720 721 722 723 724
	return error;
}

static void calc_max_reserv(struct gfs2_inode *ip, loff_t max, loff_t *len,
			    unsigned int *data_blocks, unsigned int *ind_blocks)
{
	const struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode);
725
	unsigned int max_blocks = ip->i_rgd->rd_free_clone;
726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754
	unsigned int tmp, max_data = max_blocks - 3 * (sdp->sd_max_height - 1);

	for (tmp = max_data; tmp > sdp->sd_diptrs;) {
		tmp = DIV_ROUND_UP(tmp, sdp->sd_inptrs);
		max_data -= tmp;
	}
	/* This calculation isn't the exact reverse of gfs2_write_calc_reserve,
	   so it might end up with fewer data blocks */
	if (max_data <= *data_blocks)
		return;
	*data_blocks = max_data;
	*ind_blocks = max_blocks - max_data;
	*len = ((loff_t)max_data - 3) << sdp->sd_sb.sb_bsize_shift;
	if (*len > max) {
		*len = max;
		gfs2_write_calc_reserv(ip, max, data_blocks, ind_blocks);
	}
}

static long gfs2_fallocate(struct file *file, int mode, loff_t offset,
			   loff_t len)
{
	struct inode *inode = file->f_path.dentry->d_inode;
	struct gfs2_sbd *sdp = GFS2_SB(inode);
	struct gfs2_inode *ip = GFS2_I(inode);
	unsigned int data_blocks = 0, ind_blocks = 0, rblocks;
	loff_t bytes, max_bytes;
	struct gfs2_alloc *al;
	int error;
755
	loff_t bsize_mask = ~((loff_t)sdp->sd_sb.sb_bsize - 1);
756
	loff_t next = (offset + len - 1) >> sdp->sd_sb.sb_bsize_shift;
757
	loff_t max_chunk_size = UINT_MAX & bsize_mask;
758 759 760 761 762 763
	next = (next + 1) << sdp->sd_sb.sb_bsize_shift;

	/* We only support the FALLOC_FL_KEEP_SIZE mode */
	if (mode & ~FALLOC_FL_KEEP_SIZE)
		return -EOPNOTSUPP;

764
	offset &= bsize_mask;
765 766 767 768 769

	len = next - offset;
	bytes = sdp->sd_max_rg_data * sdp->sd_sb.sb_bsize / 2;
	if (!bytes)
		bytes = UINT_MAX;
770 771 772
	bytes &= bsize_mask;
	if (bytes == 0)
		bytes = sdp->sd_sb.sb_bsize;
773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802

	gfs2_holder_init(ip->i_gl, LM_ST_EXCLUSIVE, 0, &ip->i_gh);
	error = gfs2_glock_nq(&ip->i_gh);
	if (unlikely(error))
		goto out_uninit;

	if (!gfs2_write_alloc_required(ip, offset, len))
		goto out_unlock;

	while (len > 0) {
		if (len < bytes)
			bytes = len;
		al = gfs2_alloc_get(ip);
		if (!al) {
			error = -ENOMEM;
			goto out_unlock;
		}

		error = gfs2_quota_lock_check(ip);
		if (error)
			goto out_alloc_put;

retry:
		gfs2_write_calc_reserv(ip, bytes, &data_blocks, &ind_blocks);

		al->al_requested = data_blocks + ind_blocks;
		error = gfs2_inplace_reserve(ip);
		if (error) {
			if (error == -ENOSPC && bytes > sdp->sd_sb.sb_bsize) {
				bytes >>= 1;
803 804 805
				bytes &= bsize_mask;
				if (bytes == 0)
					bytes = sdp->sd_sb.sb_bsize;
806 807 808 809 810
				goto retry;
			}
			goto out_qunlock;
		}
		max_bytes = bytes;
811 812
		calc_max_reserv(ip, (len > max_chunk_size)? max_chunk_size: len,
				&max_bytes, &data_blocks, &ind_blocks);
813 814 815
		al->al_requested = data_blocks + ind_blocks;

		rblocks = RES_DINODE + ind_blocks + RES_STATFS + RES_QUOTA +
816
			  RES_RG_HDR + gfs2_rg_blocks(ip);
817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851
		if (gfs2_is_jdata(ip))
			rblocks += data_blocks ? data_blocks : 1;

		error = gfs2_trans_begin(sdp, rblocks,
					 PAGE_CACHE_SIZE/sdp->sd_sb.sb_bsize);
		if (error)
			goto out_trans_fail;

		error = fallocate_chunk(inode, offset, max_bytes, mode);
		gfs2_trans_end(sdp);

		if (error)
			goto out_trans_fail;

		len -= max_bytes;
		offset += max_bytes;
		gfs2_inplace_release(ip);
		gfs2_quota_unlock(ip);
		gfs2_alloc_put(ip);
	}
	goto out_unlock;

out_trans_fail:
	gfs2_inplace_release(ip);
out_qunlock:
	gfs2_quota_unlock(ip);
out_alloc_put:
	gfs2_alloc_put(ip);
out_unlock:
	gfs2_glock_dq(&ip->i_gh);
out_uninit:
	gfs2_holder_uninit(&ip->i_gh);
	return error;
}

852 853
#ifdef CONFIG_GFS2_FS_LOCKING_DLM

854 855 856 857 858 859
/**
 * gfs2_setlease - acquire/release a file lease
 * @file: the file pointer
 * @arg: lease type
 * @fl: file lock
 *
860 861 862 863
 * We don't currently have a way to enforce a lease across the whole
 * cluster; until we do, disable leases (by just returning -EINVAL),
 * unless the administrator has requested purely local locking.
 *
864 865
 * Locking: called under lock_flocks
 *
866 867 868 869 870
 * Returns: errno
 */

static int gfs2_setlease(struct file *file, long arg, struct file_lock **fl)
{
871
	return -EINVAL;
872 873
}

D
David Teigland 已提交
874 875 876 877 878 879 880 881 882 883 884
/**
 * gfs2_lock - acquire/release a posix lock on a file
 * @file: the file pointer
 * @cmd: either modify or retrieve lock state, possibly wait
 * @fl: type and range of lock
 *
 * Returns: errno
 */

static int gfs2_lock(struct file *file, int cmd, struct file_lock *fl)
{
885 886
	struct gfs2_inode *ip = GFS2_I(file->f_mapping->host);
	struct gfs2_sbd *sdp = GFS2_SB(file->f_mapping->host);
887
	struct lm_lockstruct *ls = &sdp->sd_lockstruct;
D
David Teigland 已提交
888 889 890

	if (!(fl->fl_flags & FL_POSIX))
		return -ENOLCK;
891
	if (__mandatory_lock(&ip->i_inode) && fl->fl_type != F_UNLCK)
D
David Teigland 已提交
892 893
		return -ENOLCK;

M
Marc Eshel 已提交
894 895 896 897 898
	if (cmd == F_CANCELLK) {
		/* Hack: */
		cmd = F_SETLK;
		fl->fl_type = F_UNLCK;
	}
899 900
	if (unlikely(test_bit(SDF_SHUTDOWN, &sdp->sd_flags)))
		return -EIO;
D
David Teigland 已提交
901
	if (IS_GETLK(cmd))
902
		return dlm_posix_get(ls->ls_dlm, ip->i_no_addr, file, fl);
D
David Teigland 已提交
903
	else if (fl->fl_type == F_UNLCK)
904
		return dlm_posix_unlock(ls->ls_dlm, ip->i_no_addr, file, fl);
D
David Teigland 已提交
905
	else
906
		return dlm_posix_lock(ls->ls_dlm, ip->i_no_addr, file, cmd, fl);
D
David Teigland 已提交
907 908 909 910
}

static int do_flock(struct file *file, int cmd, struct file_lock *fl)
{
911
	struct gfs2_file *fp = file->private_data;
D
David Teigland 已提交
912
	struct gfs2_holder *fl_gh = &fp->f_fl_gh;
J
Josef Sipek 已提交
913
	struct gfs2_inode *ip = GFS2_I(file->f_path.dentry->d_inode);
D
David Teigland 已提交
914 915 916 917 918 919
	struct gfs2_glock *gl;
	unsigned int state;
	int flags;
	int error = 0;

	state = (fl->fl_type == F_WRLCK) ? LM_ST_EXCLUSIVE : LM_ST_SHARED;
920
	flags = (IS_SETLKW(cmd) ? 0 : LM_FLAG_TRY) | GL_EXACT | GL_NOCACHE;
D
David Teigland 已提交
921

922
	mutex_lock(&fp->f_fl_mutex);
D
David Teigland 已提交
923 924 925 926 927 928

	gl = fl_gh->gh_gl;
	if (gl) {
		if (fl_gh->gh_state == state)
			goto out;
		flock_lock_file_wait(file,
929
				     &(struct file_lock){.fl_type = F_UNLCK});
930 931
		gfs2_glock_dq_wait(fl_gh);
		gfs2_holder_reinit(state, flags, fl_gh);
D
David Teigland 已提交
932
	} else {
933 934
		error = gfs2_glock_get(GFS2_SB(&ip->i_inode), ip->i_no_addr,
				       &gfs2_flock_glops, CREATE, &gl);
D
David Teigland 已提交
935 936
		if (error)
			goto out;
937 938
		gfs2_holder_init(gl, state, flags, fl_gh);
		gfs2_glock_put(gl);
D
David Teigland 已提交
939 940 941 942 943 944 945 946
	}
	error = gfs2_glock_nq(fl_gh);
	if (error) {
		gfs2_holder_uninit(fl_gh);
		if (error == GLR_TRYFAILED)
			error = -EAGAIN;
	} else {
		error = flock_lock_file_wait(file, fl);
947
		gfs2_assert_warn(GFS2_SB(&ip->i_inode), !error);
D
David Teigland 已提交
948 949
	}

950
out:
951
	mutex_unlock(&fp->f_fl_mutex);
D
David Teigland 已提交
952 953 954 955 956
	return error;
}

static void do_unflock(struct file *file, struct file_lock *fl)
{
957
	struct gfs2_file *fp = file->private_data;
D
David Teigland 已提交
958 959
	struct gfs2_holder *fl_gh = &fp->f_fl_gh;

960
	mutex_lock(&fp->f_fl_mutex);
D
David Teigland 已提交
961
	flock_lock_file_wait(file, fl);
962 963 964 965
	if (fl_gh->gh_gl) {
		gfs2_glock_dq_wait(fl_gh);
		gfs2_holder_uninit(fl_gh);
	}
966
	mutex_unlock(&fp->f_fl_mutex);
D
David Teigland 已提交
967 968 969 970 971 972 973 974 975 976 977 978 979 980 981
}

/**
 * gfs2_flock - acquire/release a flock lock on a file
 * @file: the file pointer
 * @cmd: either modify or retrieve lock state, possibly wait
 * @fl: type and range of lock
 *
 * Returns: errno
 */

static int gfs2_flock(struct file *file, int cmd, struct file_lock *fl)
{
	if (!(fl->fl_flags & FL_FLOCK))
		return -ENOLCK;
982 983
	if (fl->fl_type & LOCK_MAND)
		return -EOPNOTSUPP;
D
David Teigland 已提交
984 985 986 987

	if (fl->fl_type == F_UNLCK) {
		do_unflock(file, fl);
		return 0;
988
	} else {
D
David Teigland 已提交
989
		return do_flock(file, cmd, fl);
990
	}
D
David Teigland 已提交
991 992
}

993
const struct file_operations gfs2_file_fops = {
994
	.llseek		= gfs2_llseek,
995
	.read		= do_sync_read,
996
	.aio_read	= generic_file_aio_read,
997
	.write		= do_sync_write,
998
	.aio_write	= gfs2_file_aio_write,
999 1000 1001 1002 1003 1004 1005 1006 1007
	.unlocked_ioctl	= gfs2_ioctl,
	.mmap		= gfs2_mmap,
	.open		= gfs2_open,
	.release	= gfs2_close,
	.fsync		= gfs2_fsync,
	.lock		= gfs2_lock,
	.flock		= gfs2_flock,
	.splice_read	= generic_file_splice_read,
	.splice_write	= generic_file_splice_write,
1008
	.setlease	= gfs2_setlease,
1009
	.fallocate	= gfs2_fallocate,
D
David Teigland 已提交
1010 1011
};

1012
const struct file_operations gfs2_dir_fops = {
1013 1014 1015 1016 1017 1018 1019
	.readdir	= gfs2_readdir,
	.unlocked_ioctl	= gfs2_ioctl,
	.open		= gfs2_open,
	.release	= gfs2_close,
	.fsync		= gfs2_fsync,
	.lock		= gfs2_lock,
	.flock		= gfs2_flock,
1020
	.llseek		= default_llseek,
D
David Teigland 已提交
1021 1022
};

1023 1024
#endif /* CONFIG_GFS2_FS_LOCKING_DLM */

1025
const struct file_operations gfs2_file_fops_nolock = {
1026 1027 1028 1029
	.llseek		= gfs2_llseek,
	.read		= do_sync_read,
	.aio_read	= generic_file_aio_read,
	.write		= do_sync_write,
1030
	.aio_write	= gfs2_file_aio_write,
1031 1032 1033 1034 1035 1036 1037
	.unlocked_ioctl	= gfs2_ioctl,
	.mmap		= gfs2_mmap,
	.open		= gfs2_open,
	.release	= gfs2_close,
	.fsync		= gfs2_fsync,
	.splice_read	= generic_file_splice_read,
	.splice_write	= generic_file_splice_write,
1038
	.setlease	= generic_setlease,
1039
	.fallocate	= gfs2_fallocate,
1040 1041
};

1042
const struct file_operations gfs2_dir_fops_nolock = {
1043 1044 1045 1046 1047
	.readdir	= gfs2_readdir,
	.unlocked_ioctl	= gfs2_ioctl,
	.open		= gfs2_open,
	.release	= gfs2_close,
	.fsync		= gfs2_fsync,
1048
	.llseek		= default_llseek,
1049 1050
};