pagelist.c 28.6 KB
Newer Older
L
Linus Torvalds 已提交
1 2 3 4 5 6 7 8 9 10 11 12 13
/*
 * linux/fs/nfs/pagelist.c
 *
 * A set of helper functions for managing NFS read and write requests.
 * The main purpose of these routines is to provide support for the
 * coalescing of several requests into a single RPC call.
 *
 * Copyright 2000, 2001 (c) Trond Myklebust <trond.myklebust@fys.uio.no>
 *
 */

#include <linux/slab.h>
#include <linux/file.h>
A
Alexey Dobriyan 已提交
14
#include <linux/sched.h>
L
Linus Torvalds 已提交
15
#include <linux/sunrpc/clnt.h>
16
#include <linux/nfs.h>
L
Linus Torvalds 已提交
17 18 19 20 21
#include <linux/nfs3.h>
#include <linux/nfs4.h>
#include <linux/nfs_page.h>
#include <linux/nfs_fs.h>
#include <linux/nfs_mount.h>
22
#include <linux/export.h>
L
Linus Torvalds 已提交
23

24
#include "internal.h"
25
#include "pnfs.h"
26

27 28
#define NFSDBG_FACILITY		NFSDBG_PAGECACHE

29
static struct kmem_cache *nfs_page_cachep;
30
static const struct rpc_call_ops nfs_pgio_common_ops;
L
Linus Torvalds 已提交
31

32
static bool nfs_pgarray_set(struct nfs_page_array *p, unsigned int pagecount)
F
Fred Isaman 已提交
33 34 35 36 37 38 39 40 41 42 43 44
{
	p->npages = pagecount;
	if (pagecount <= ARRAY_SIZE(p->page_array))
		p->pagevec = p->page_array;
	else {
		p->pagevec = kcalloc(pagecount, sizeof(struct page *), GFP_KERNEL);
		if (!p->pagevec)
			p->npages = 0;
	}
	return p->pagevec != NULL;
}

45 46 47 48 49 50 51 52 53
void nfs_pgheader_init(struct nfs_pageio_descriptor *desc,
		       struct nfs_pgio_header *hdr,
		       void (*release)(struct nfs_pgio_header *hdr))
{
	hdr->req = nfs_list_entry(desc->pg_list.next);
	hdr->inode = desc->pg_inode;
	hdr->cred = hdr->req->wb_context->cred;
	hdr->io_start = req_offset(hdr->req);
	hdr->good_bytes = desc->pg_count;
54
	hdr->dreq = desc->pg_dreq;
55
	hdr->layout_private = desc->pg_layout_private;
56
	hdr->release = release;
57
	hdr->completion_ops = desc->pg_completion_ops;
58 59
	if (hdr->completion_ops->init_hdr)
		hdr->completion_ops->init_hdr(hdr);
60
}
B
Bryan Schumaker 已提交
61
EXPORT_SYMBOL_GPL(nfs_pgheader_init);
62 63 64 65 66 67 68 69 70 71 72 73 74

void nfs_set_pgio_error(struct nfs_pgio_header *hdr, int error, loff_t pos)
{
	spin_lock(&hdr->lock);
	if (pos < hdr->io_start + hdr->good_bytes) {
		set_bit(NFS_IOHDR_ERROR, &hdr->flags);
		clear_bit(NFS_IOHDR_EOF, &hdr->flags);
		hdr->good_bytes = pos - hdr->io_start;
		hdr->error = error;
	}
	spin_unlock(&hdr->lock);
}

L
Linus Torvalds 已提交
75 76 77
static inline struct nfs_page *
nfs_page_alloc(void)
{
78
	struct nfs_page	*p = kmem_cache_zalloc(nfs_page_cachep, GFP_NOIO);
79
	if (p)
L
Linus Torvalds 已提交
80 81 82 83 84 85 86 87 88 89
		INIT_LIST_HEAD(&p->wb_list);
	return p;
}

static inline void
nfs_page_free(struct nfs_page *p)
{
	kmem_cache_free(nfs_page_cachep, p);
}

90 91 92 93 94 95 96 97 98 99 100
static void
nfs_iocounter_inc(struct nfs_io_counter *c)
{
	atomic_inc(&c->io_count);
}

static void
nfs_iocounter_dec(struct nfs_io_counter *c)
{
	if (atomic_dec_and_test(&c->io_count)) {
		clear_bit(NFS_IO_INPROGRESS, &c->flags);
101
		smp_mb__after_atomic();
102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117
		wake_up_bit(&c->flags, NFS_IO_INPROGRESS);
	}
}

static int
__nfs_iocounter_wait(struct nfs_io_counter *c)
{
	wait_queue_head_t *wq = bit_waitqueue(&c->flags, NFS_IO_INPROGRESS);
	DEFINE_WAIT_BIT(q, &c->flags, NFS_IO_INPROGRESS);
	int ret = 0;

	do {
		prepare_to_wait(wq, &q.wait, TASK_KILLABLE);
		set_bit(NFS_IO_INPROGRESS, &c->flags);
		if (atomic_read(&c->io_count) == 0)
			break;
118
		ret = nfs_wait_bit_killable(&q.key);
119
	} while (atomic_read(&c->io_count) != 0 && !ret);
120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138
	finish_wait(wq, &q.wait);
	return ret;
}

/**
 * nfs_iocounter_wait - wait for i/o to complete
 * @c: nfs_io_counter to use
 *
 * returns -ERESTARTSYS if interrupted by a fatal signal.
 * Otherwise returns 0 once the io_count hits 0.
 */
int
nfs_iocounter_wait(struct nfs_io_counter *c)
{
	if (atomic_read(&c->io_count) == 0)
		return 0;
	return __nfs_iocounter_wait(c);
}

139 140 141
/*
 * nfs_page_group_lock - lock the head of the page group
 * @req - request in group that is to be locked
142
 * @nonblock - if true don't block waiting for lock
143 144
 *
 * this lock must be held if modifying the page group list
145
 *
146 147 148 149 150 151
 * return 0 on success, < 0 on error: -EDELAY if nonblocking or the
 * result from wait_on_bit_lock
 *
 * NOTE: calling with nonblock=false should always have set the
 *       lock bit (see fs/buffer.c and other uses of wait_on_bit_lock
 *       with TASK_UNINTERRUPTIBLE), so there is no need to check the result.
152
 */
153
int
154
nfs_page_group_lock(struct nfs_page *req, bool nonblock)
155 156 157 158 159
{
	struct nfs_page *head = req->wb_head;

	WARN_ON_ONCE(head != head->wb_head);

160 161
	if (!test_and_set_bit(PG_HEADLOCK, &head->wb_flags))
		return 0;
162

163 164 165 166 167
	if (!nonblock)
		return wait_on_bit_lock(&head->wb_flags, PG_HEADLOCK,
				TASK_UNINTERRUPTIBLE);

	return -EAGAIN;
168 169
}

170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186
/*
 * nfs_page_group_lock_wait - wait for the lock to clear, but don't grab it
 * @req - a request in the group
 *
 * This is a blocking call to wait for the group lock to be cleared.
 */
void
nfs_page_group_lock_wait(struct nfs_page *req)
{
	struct nfs_page *head = req->wb_head;

	WARN_ON_ONCE(head != head->wb_head);

	wait_on_bit(&head->wb_flags, PG_HEADLOCK,
		TASK_UNINTERRUPTIBLE);
}

187 188 189 190 191 192 193 194 195 196 197
/*
 * nfs_page_group_unlock - unlock the head of the page group
 * @req - request in group that is to be unlocked
 */
void
nfs_page_group_unlock(struct nfs_page *req)
{
	struct nfs_page *head = req->wb_head;

	WARN_ON_ONCE(head != head->wb_head);

198
	smp_mb__before_atomic();
199
	clear_bit(PG_HEADLOCK, &head->wb_flags);
200
	smp_mb__after_atomic();
201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244
	wake_up_bit(&head->wb_flags, PG_HEADLOCK);
}

/*
 * nfs_page_group_sync_on_bit_locked
 *
 * must be called with page group lock held
 */
static bool
nfs_page_group_sync_on_bit_locked(struct nfs_page *req, unsigned int bit)
{
	struct nfs_page *head = req->wb_head;
	struct nfs_page *tmp;

	WARN_ON_ONCE(!test_bit(PG_HEADLOCK, &head->wb_flags));
	WARN_ON_ONCE(test_and_set_bit(bit, &req->wb_flags));

	tmp = req->wb_this_page;
	while (tmp != req) {
		if (!test_bit(bit, &tmp->wb_flags))
			return false;
		tmp = tmp->wb_this_page;
	}

	/* true! reset all bits */
	tmp = req;
	do {
		clear_bit(bit, &tmp->wb_flags);
		tmp = tmp->wb_this_page;
	} while (tmp != req);

	return true;
}

/*
 * nfs_page_group_sync_on_bit - set bit on current request, but only
 *   return true if the bit is set for all requests in page group
 * @req - request in page group
 * @bit - PG_* bit that is used to sync page group
 */
bool nfs_page_group_sync_on_bit(struct nfs_page *req, unsigned int bit)
{
	bool ret;

245
	nfs_page_group_lock(req, false);
246 247 248 249 250 251 252 253 254 255 256 257 258 259 260
	ret = nfs_page_group_sync_on_bit_locked(req, bit);
	nfs_page_group_unlock(req);

	return ret;
}

/*
 * nfs_page_group_init - Initialize the page group linkage for @req
 * @req - a new nfs request
 * @prev - the previous request in page group, or NULL if @req is the first
 *         or only request in the group (the head).
 */
static inline void
nfs_page_group_init(struct nfs_page *req, struct nfs_page *prev)
{
261
	struct inode *inode;
262 263 264
	WARN_ON_ONCE(prev == req);

	if (!prev) {
265
		/* a head request */
266 267 268
		req->wb_head = req;
		req->wb_this_page = req;
	} else {
269
		/* a subrequest */
270 271 272 273 274 275
		WARN_ON_ONCE(prev->wb_this_page != prev->wb_head);
		WARN_ON_ONCE(!test_bit(PG_HEADLOCK, &prev->wb_head->wb_flags));
		req->wb_head = prev->wb_head;
		req->wb_this_page = prev->wb_this_page;
		prev->wb_this_page = req;

276 277 278 279
		/* All subrequests take a ref on the head request until
		 * nfs_page_group_destroy is called */
		kref_get(&req->wb_head->wb_kref);

280 281 282
		/* grab extra ref and bump the request count if head request
		 * has extra ref from the write/commit path to handle handoff
		 * between write and commit lists. */
283
		if (test_bit(PG_INODE_REF, &prev->wb_head->wb_flags)) {
284
			inode = page_file_mapping(req->wb_page)->host;
285
			set_bit(PG_INODE_REF, &req->wb_flags);
286
			kref_get(&req->wb_kref);
287 288 289
			spin_lock(&inode->i_lock);
			NFS_I(inode)->nrequests++;
			spin_unlock(&inode->i_lock);
290
		}
291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306
	}
}

/*
 * nfs_page_group_destroy - sync the destruction of page groups
 * @req - request that no longer needs the page group
 *
 * releases the page group reference from each member once all
 * members have called this function.
 */
static void
nfs_page_group_destroy(struct kref *kref)
{
	struct nfs_page *req = container_of(kref, struct nfs_page, wb_kref);
	struct nfs_page *tmp, *next;

307 308 309 310
	/* subrequests must release the ref on the head request */
	if (req->wb_head != req)
		nfs_release_request(req->wb_head);

311 312 313 314 315 316 317 318 319 320 321 322 323 324
	if (!nfs_page_group_sync_on_bit(req, PG_TEARDOWN))
		return;

	tmp = req;
	do {
		next = tmp->wb_this_page;
		/* unlink and free */
		tmp->wb_this_page = tmp;
		tmp->wb_head = tmp;
		nfs_free_request(tmp);
		tmp = next;
	} while (tmp != req);
}

L
Linus Torvalds 已提交
325 326
/**
 * nfs_create_request - Create an NFS read/write request.
327
 * @ctx: open context to use
L
Linus Torvalds 已提交
328
 * @page: page to write
329
 * @last: last nfs request created for this page group or NULL if head
L
Linus Torvalds 已提交
330 331 332 333
 * @offset: starting offset within the page for the write
 * @count: number of bytes to read/write
 *
 * The page must be locked by the caller. This makes sure we never
334
 * create two different requests for the same page.
L
Linus Torvalds 已提交
335 336 337
 * User should ensure it is safe to sleep in this function.
 */
struct nfs_page *
338
nfs_create_request(struct nfs_open_context *ctx, struct page *page,
339 340
		   struct nfs_page *last, unsigned int offset,
		   unsigned int count)
L
Linus Torvalds 已提交
341 342
{
	struct nfs_page		*req;
343
	struct nfs_lock_context *l_ctx;
L
Linus Torvalds 已提交
344

345 346
	if (test_bit(NFS_CONTEXT_BAD, &ctx->flags))
		return ERR_PTR(-EBADF);
347 348 349 350
	/* try to allocate the request struct */
	req = nfs_page_alloc();
	if (req == NULL)
		return ERR_PTR(-ENOMEM);
L
Linus Torvalds 已提交
351

352
	/* get lock context early so we can deal with alloc failures */
353 354
	l_ctx = nfs_get_lock_context(ctx);
	if (IS_ERR(l_ctx)) {
355
		nfs_page_free(req);
356
		return ERR_CAST(l_ctx);
357
	}
358
	req->wb_lock_context = l_ctx;
359
	nfs_iocounter_inc(&l_ctx->io_count);
360

L
Linus Torvalds 已提交
361 362 363 364
	/* Initialize the request struct. Initially, we assume a
	 * long write-back delay. This will be adjusted in
	 * update_nfs_request below if the region is not locked. */
	req->wb_page    = page;
365
	req->wb_index	= page_file_index(page);
L
Linus Torvalds 已提交
366 367 368 369 370
	page_cache_get(page);
	req->wb_offset  = offset;
	req->wb_pgbase	= offset;
	req->wb_bytes   = count;
	req->wb_context = get_nfs_open_context(ctx);
371
	kref_init(&req->wb_kref);
372
	nfs_page_group_init(req, last);
L
Linus Torvalds 已提交
373 374 375 376
	return req;
}

/**
377
 * nfs_unlock_request - Unlock request and wake up sleepers.
L
Linus Torvalds 已提交
378 379
 * @req:
 */
380
void nfs_unlock_request(struct nfs_page *req)
L
Linus Torvalds 已提交
381 382 383 384 385
{
	if (!NFS_WBACK_BUSY(req)) {
		printk(KERN_ERR "NFS: Invalid unlock attempted\n");
		BUG();
	}
386
	smp_mb__before_atomic();
L
Linus Torvalds 已提交
387
	clear_bit(PG_BUSY, &req->wb_flags);
388
	smp_mb__after_atomic();
389
	wake_up_bit(&req->wb_flags, PG_BUSY);
390 391 392
}

/**
393 394
 * nfs_unlock_and_release_request - Unlock request and release the nfs_page
 * @req:
395
 */
396
void nfs_unlock_and_release_request(struct nfs_page *req)
397
{
398
	nfs_unlock_request(req);
L
Linus Torvalds 已提交
399 400 401
	nfs_release_request(req);
}

402
/*
L
Linus Torvalds 已提交
403 404 405
 * nfs_clear_request - Free up all resources allocated to the request
 * @req:
 *
406 407
 * Release page and open context resources associated with a read/write
 * request after it has completed.
L
Linus Torvalds 已提交
408
 */
409
static void nfs_clear_request(struct nfs_page *req)
L
Linus Torvalds 已提交
410
{
411
	struct page *page = req->wb_page;
412
	struct nfs_open_context *ctx = req->wb_context;
413
	struct nfs_lock_context *l_ctx = req->wb_lock_context;
414

415 416
	if (page != NULL) {
		page_cache_release(page);
L
Linus Torvalds 已提交
417 418
		req->wb_page = NULL;
	}
419
	if (l_ctx != NULL) {
420
		nfs_iocounter_dec(&l_ctx->io_count);
421 422 423
		nfs_put_lock_context(l_ctx);
		req->wb_lock_context = NULL;
	}
424 425 426 427
	if (ctx != NULL) {
		put_nfs_open_context(ctx);
		req->wb_context = NULL;
	}
L
Linus Torvalds 已提交
428 429 430 431 432 433 434 435
}

/**
 * nfs_release_request - Release the count on an NFS read/write request
 * @req: request to release
 *
 * Note: Should never be called with the spinlock held!
 */
436
void nfs_free_request(struct nfs_page *req)
L
Linus Torvalds 已提交
437
{
438 439 440 441
	WARN_ON_ONCE(req->wb_this_page != req);

	/* extra debug: make sure no sync bits are still set */
	WARN_ON_ONCE(test_bit(PG_TEARDOWN, &req->wb_flags));
442 443
	WARN_ON_ONCE(test_bit(PG_UNLOCKPAGE, &req->wb_flags));
	WARN_ON_ONCE(test_bit(PG_UPTODATE, &req->wb_flags));
444 445
	WARN_ON_ONCE(test_bit(PG_WB_END, &req->wb_flags));
	WARN_ON_ONCE(test_bit(PG_REMOVE, &req->wb_flags));
L
Linus Torvalds 已提交
446

447
	/* Release struct file and open context */
L
Linus Torvalds 已提交
448 449 450 451
	nfs_clear_request(req);
	nfs_page_free(req);
}

452 453
void nfs_release_request(struct nfs_page *req)
{
454
	kref_put(&req->wb_kref, nfs_page_group_destroy);
455 456
}

L
Linus Torvalds 已提交
457 458 459 460
/**
 * nfs_wait_on_request - Wait for a request to complete.
 * @req: request to wait upon.
 *
461
 * Interruptible by fatal signals only.
L
Linus Torvalds 已提交
462 463 464 465 466
 * The user is responsible for holding a count on the request.
 */
int
nfs_wait_on_request(struct nfs_page *req)
{
467 468
	return wait_on_bit_io(&req->wb_flags, PG_BUSY,
			      TASK_UNINTERRUPTIBLE);
L
Linus Torvalds 已提交
469 470
}

471 472 473 474 475 476 477 478 479 480 481
/*
 * nfs_generic_pg_test - determine if requests can be coalesced
 * @desc: pointer to descriptor
 * @prev: previous request in desc, or NULL
 * @req: this request
 *
 * Returns zero if @req can be coalesced into @desc, otherwise it returns
 * the size of the request.
 */
size_t nfs_generic_pg_test(struct nfs_pageio_descriptor *desc,
			   struct nfs_page *prev, struct nfs_page *req)
482
{
483 484 485
	if (desc->pg_count > desc->pg_bsize) {
		/* should never happen */
		WARN_ON_ONCE(1);
486
		return 0;
487
	}
488

489 490 491 492 493 494 495 496
	/*
	 * Limit the request size so that we can still allocate a page array
	 * for it without upsetting the slab allocator.
	 */
	if (((desc->pg_count + req->wb_bytes) >> PAGE_SHIFT) *
			sizeof(struct page) > PAGE_SIZE)
		return 0;

497
	return min(desc->pg_bsize - desc->pg_count, (size_t)req->wb_bytes);
498
}
499
EXPORT_SYMBOL_GPL(nfs_generic_pg_test);
500

501
struct nfs_pgio_header *nfs_pgio_header_alloc(const struct nfs_rw_ops *ops)
502
{
503
	struct nfs_pgio_header *hdr = ops->rw_alloc_header();
504

505
	if (hdr) {
506 507 508 509
		INIT_LIST_HEAD(&hdr->pages);
		spin_lock_init(&hdr->lock);
		hdr->rw_ops = ops;
	}
510
	return hdr;
511
}
512
EXPORT_SYMBOL_GPL(nfs_pgio_header_alloc);
513 514

/*
515
 * nfs_pgio_header_free - Free a read or write header
516 517
 * @hdr: The header to free
 */
518
void nfs_pgio_header_free(struct nfs_pgio_header *hdr)
519
{
520
	hdr->rw_ops->rw_free_header(hdr);
521
}
522
EXPORT_SYMBOL_GPL(nfs_pgio_header_free);
523

524
/**
525 526 527 528 529 530
 * nfs_pgio_data_destroy - make @hdr suitable for reuse
 *
 * Frees memory and releases refs from nfs_generic_pgio, so that it may
 * be called again.
 *
 * @hdr: A header that has had nfs_generic_pgio called
531
 */
532
void nfs_pgio_data_destroy(struct nfs_pgio_header *hdr)
533
{
534 535
	if (hdr->args.context)
		put_nfs_open_context(hdr->args.context);
536 537
	if (hdr->page_array.pagevec != hdr->page_array.page_array)
		kfree(hdr->page_array.pagevec);
538
}
539
EXPORT_SYMBOL_GPL(nfs_pgio_data_destroy);
540

541 542
/**
 * nfs_pgio_rpcsetup - Set up arguments for a pageio call
543
 * @hdr: The pageio hdr
544 545 546 547 548
 * @count: Number of bytes to read
 * @offset: Initial offset
 * @how: How to commit data (writes only)
 * @cinfo: Commit information for the call (writes only)
 */
549
static void nfs_pgio_rpcsetup(struct nfs_pgio_header *hdr,
550 551 552
			      unsigned int count, unsigned int offset,
			      int how, struct nfs_commit_info *cinfo)
{
553
	struct nfs_page *req = hdr->req;
554 555

	/* Set up the RPC argument and reply structs
556
	 * NB: take care not to mess about with hdr->commit et al. */
557

558 559
	hdr->args.fh     = NFS_FH(hdr->inode);
	hdr->args.offset = req_offset(req) + offset;
560
	/* pnfs_set_layoutcommit needs this */
561 562 563 564 565 566 567
	hdr->mds_offset = hdr->args.offset;
	hdr->args.pgbase = req->wb_pgbase + offset;
	hdr->args.pages  = hdr->page_array.pagevec;
	hdr->args.count  = count;
	hdr->args.context = get_nfs_open_context(req->wb_context);
	hdr->args.lock_context = req->wb_lock_context;
	hdr->args.stable  = NFS_UNSTABLE;
568 569 570 571 572 573 574
	switch (how & (FLUSH_STABLE | FLUSH_COND_STABLE)) {
	case 0:
		break;
	case FLUSH_COND_STABLE:
		if (nfs_reqs_to_commit(cinfo))
			break;
	default:
575
		hdr->args.stable = NFS_FILE_SYNC;
576 577
	}

578 579 580
	hdr->res.fattr   = &hdr->fattr;
	hdr->res.count   = count;
	hdr->res.eof     = 0;
581
	hdr->res.verf    = &hdr->verf;
582
	nfs_fattr_init(&hdr->fattr);
583 584
}

585
/**
586
 * nfs_pgio_prepare - Prepare pageio hdr to go over the wire
587
 * @task: The current task
588
 * @calldata: pageio header to prepare
589
 */
590
static void nfs_pgio_prepare(struct rpc_task *task, void *calldata)
591
{
592
	struct nfs_pgio_header *hdr = calldata;
593
	int err;
594
	err = NFS_PROTO(hdr->inode)->pgio_rpc_prepare(task, hdr);
595 596 597 598
	if (err)
		rpc_exit(task, err);
}

599
int nfs_initiate_pgio(struct rpc_clnt *clnt, struct nfs_pgio_header *hdr,
600 601 602 603
		      const struct rpc_call_ops *call_ops, int how, int flags)
{
	struct rpc_task *task;
	struct rpc_message msg = {
604 605 606
		.rpc_argp = &hdr->args,
		.rpc_resp = &hdr->res,
		.rpc_cred = hdr->cred,
607 608 609
	};
	struct rpc_task_setup task_setup_data = {
		.rpc_client = clnt,
610
		.task = &hdr->task,
611 612
		.rpc_message = &msg,
		.callback_ops = call_ops,
613
		.callback_data = hdr,
614 615 616 617 618
		.workqueue = nfsiod_workqueue,
		.flags = RPC_TASK_ASYNC | flags,
	};
	int ret = 0;

619
	hdr->rw_ops->rw_initiate(hdr, &msg, &task_setup_data, how);
620 621 622

	dprintk("NFS: %5u initiated pgio call "
		"(req %s/%llu, %u bytes @ offset %llu)\n",
623
		hdr->task.tk_pid,
A
Anna Schumaker 已提交
624 625
		hdr->inode->i_sb->s_id,
		(unsigned long long)NFS_FILEID(hdr->inode),
626 627
		hdr->args.count,
		(unsigned long long)hdr->args.offset);
628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644

	task = rpc_run_task(&task_setup_data);
	if (IS_ERR(task)) {
		ret = PTR_ERR(task);
		goto out;
	}
	if (how & FLUSH_SYNC) {
		ret = rpc_wait_for_completion_task(task);
		if (ret == 0)
			ret = task->tk_status;
	}
	rpc_put_task(task);
out:
	return ret;
}
EXPORT_SYMBOL_GPL(nfs_initiate_pgio);

645 646 647 648 649
/**
 * nfs_pgio_error - Clean up from a pageio error
 * @desc: IO descriptor
 * @hdr: pageio header
 */
650
static int nfs_pgio_error(struct nfs_pageio_descriptor *desc,
651 652 653
			  struct nfs_pgio_header *hdr)
{
	set_bit(NFS_IOHDR_REDO, &hdr->flags);
654
	nfs_pgio_data_destroy(hdr);
655
	hdr->completion_ops->completion(hdr);
656 657 658 659
	desc->pg_completion_ops->error_cleanup(&desc->pg_list);
	return -ENOMEM;
}

660 661
/**
 * nfs_pgio_release - Release pageio data
662
 * @calldata: The pageio header to release
663
 */
664
static void nfs_pgio_release(void *calldata)
665
{
666 667 668 669
	struct nfs_pgio_header *hdr = calldata;
	if (hdr->rw_ops->rw_release)
		hdr->rw_ops->rw_release(hdr);
	nfs_pgio_data_destroy(hdr);
670
	hdr->completion_ops->completion(hdr);
671 672
}

L
Linus Torvalds 已提交
673
/**
674 675
 * nfs_pageio_init - initialise a page io descriptor
 * @desc: pointer to descriptor
676 677 678 679
 * @inode: pointer to inode
 * @doio: pointer to io function
 * @bsize: io block size
 * @io_flags: extra parameters for the io function
680
 */
681 682
void nfs_pageio_init(struct nfs_pageio_descriptor *desc,
		     struct inode *inode,
683
		     const struct nfs_pageio_ops *pg_ops,
684
		     const struct nfs_pgio_completion_ops *compl_ops,
685
		     const struct nfs_rw_ops *rw_ops,
686
		     size_t bsize,
687
		     int io_flags)
688 689
{
	INIT_LIST_HEAD(&desc->pg_list);
690
	desc->pg_bytes_written = 0;
691 692 693
	desc->pg_count = 0;
	desc->pg_bsize = bsize;
	desc->pg_base = 0;
694
	desc->pg_moreio = 0;
695
	desc->pg_recoalesce = 0;
696
	desc->pg_inode = inode;
697
	desc->pg_ops = pg_ops;
698
	desc->pg_completion_ops = compl_ops;
699
	desc->pg_rw_ops = rw_ops;
700 701
	desc->pg_ioflags = io_flags;
	desc->pg_error = 0;
702
	desc->pg_lseg = NULL;
703
	desc->pg_dreq = NULL;
704
	desc->pg_layout_private = NULL;
705
}
B
Bryan Schumaker 已提交
706
EXPORT_SYMBOL_GPL(nfs_pageio_init);
707

708 709 710
/**
 * nfs_pgio_result - Basic pageio error handling
 * @task: The task that ran
711
 * @calldata: Pageio header to check
712
 */
713
static void nfs_pgio_result(struct rpc_task *task, void *calldata)
714
{
715 716
	struct nfs_pgio_header *hdr = calldata;
	struct inode *inode = hdr->inode;
717 718 719 720

	dprintk("NFS: %s: %5u, (status %d)\n", __func__,
		task->tk_pid, task->tk_status);

721
	if (hdr->rw_ops->rw_done(task, hdr, inode) != 0)
722 723
		return;
	if (task->tk_status < 0)
724
		nfs_set_pgio_error(hdr, task->tk_status, hdr->args.offset);
725
	else
726
		hdr->rw_ops->rw_result(task, hdr);
727 728
}

729 730 731 732 733 734 735 736
/*
 * Create an RPC task for the given read or write request and kick it.
 * The page must have been locked by the caller.
 *
 * It may happen that the page we're passed is not marked dirty.
 * This is the case if nfs_updatepage detects a conflicting request
 * that has been written but not committed.
 */
737 738
int nfs_generic_pgio(struct nfs_pageio_descriptor *desc,
		     struct nfs_pgio_header *hdr)
739 740
{
	struct nfs_page		*req;
741 742
	struct page		**pages,
				*last_page;
743 744
	struct list_head *head = &desc->pg_list;
	struct nfs_commit_info cinfo;
745
	unsigned int pagecount, pageused;
746

747 748
	pagecount = nfs_page_array_len(desc->pg_base, desc->pg_count);
	if (!nfs_pgarray_set(&hdr->page_array, pagecount))
749 750 751
		return nfs_pgio_error(desc, hdr);

	nfs_init_cinfo(&cinfo, desc->pg_inode, desc->pg_dreq);
752
	pages = hdr->page_array.pagevec;
753 754
	last_page = NULL;
	pageused = 0;
755 756 757 758
	while (!list_empty(head)) {
		req = nfs_list_entry(head->next);
		nfs_list_remove_request(req);
		nfs_list_add_request(req, &hdr->pages);
759 760 761

		if (!last_page || last_page != req->wb_page) {
			pageused++;
762 763 764
			if (pageused > pagecount)
				break;
			*pages++ = last_page = req->wb_page;
765
		}
766
	}
767 768
	if (WARN_ON_ONCE(pageused != pagecount))
		return nfs_pgio_error(desc, hdr);
769 770 771 772 773 774

	if ((desc->pg_ioflags & FLUSH_COND_STABLE) &&
	    (desc->pg_moreio || nfs_reqs_to_commit(&cinfo)))
		desc->pg_ioflags &= ~FLUSH_COND_STABLE;

	/* Set up the argument struct */
775
	nfs_pgio_rpcsetup(hdr, desc->pg_count, 0, desc->pg_ioflags, &cinfo);
776 777 778
	desc->pg_rpc_callops = &nfs_pgio_common_ops;
	return 0;
}
779
EXPORT_SYMBOL_GPL(nfs_generic_pgio);
780

781
static int nfs_generic_pg_pgios(struct nfs_pageio_descriptor *desc)
782 783 784 785
{
	struct nfs_pgio_header *hdr;
	int ret;

786 787
	hdr = nfs_pgio_header_alloc(desc->pg_rw_ops);
	if (!hdr) {
788 789 790
		desc->pg_completion_ops->error_cleanup(&desc->pg_list);
		return -ENOMEM;
	}
791
	nfs_pgheader_init(desc, hdr, nfs_pgio_header_free);
792 793
	ret = nfs_generic_pgio(desc, hdr);
	if (ret == 0)
794
		ret = nfs_initiate_pgio(NFS_CLIENT(hdr->inode),
795
					hdr, desc->pg_rpc_callops,
796
					desc->pg_ioflags, 0);
797 798 799
	return ret;
}

800 801 802 803 804 805 806 807 808 809 810 811 812
static bool nfs_match_open_context(const struct nfs_open_context *ctx1,
		const struct nfs_open_context *ctx2)
{
	return ctx1->cred == ctx2->cred && ctx1->state == ctx2->state;
}

static bool nfs_match_lock_context(const struct nfs_lock_context *l1,
		const struct nfs_lock_context *l2)
{
	return l1->lockowner.l_owner == l2->lockowner.l_owner
		&& l1->lockowner.l_pid == l2->lockowner.l_pid;
}

813 814 815 816 817 818 819 820 821 822 823
/**
 * nfs_can_coalesce_requests - test two requests for compatibility
 * @prev: pointer to nfs_page
 * @req: pointer to nfs_page
 *
 * The nfs_page structures 'prev' and 'req' are compared to ensure that the
 * page data area they describe is contiguous, and that their RPC
 * credentials, NFSv4 open state, and lockowners are the same.
 *
 * Return 'true' if this is the case, else return 'false'.
 */
824 825 826
static bool nfs_can_coalesce_requests(struct nfs_page *prev,
				      struct nfs_page *req,
				      struct nfs_pageio_descriptor *pgio)
827
{
828
	size_t size;
829
	struct file_lock_context *flctx;
830

831 832 833 834 835 836 837
	if (prev) {
		if (!nfs_match_open_context(req->wb_context, prev->wb_context))
			return false;
		if (req->wb_context->dentry->d_inode->i_flock != NULL &&
		    !nfs_match_lock_context(req->wb_lock_context,
					    prev->wb_lock_context))
			return false;
838 839 840 841 842
		flctx = req->wb_context->dentry->d_inode->i_flctx;
		if (flctx != NULL && !list_empty_careful(&flctx->flc_flock) &&
		    !nfs_match_lock_context(req->wb_lock_context,
					    prev->wb_lock_context))
			return false;
843 844
		if (req_offset(req) != req_offset(prev) + prev->wb_bytes)
			return false;
845 846 847 848 849 850 851 852
		if (req->wb_page == prev->wb_page) {
			if (req->wb_pgbase != prev->wb_pgbase + prev->wb_bytes)
				return false;
		} else {
			if (req->wb_pgbase != 0 ||
			    prev->wb_pgbase + prev->wb_bytes != PAGE_CACHE_SIZE)
				return false;
		}
853
	}
854
	size = pgio->pg_ops->pg_test(pgio, prev, req);
855 856 857
	WARN_ON_ONCE(size > req->wb_bytes);
	if (size && size < req->wb_bytes)
		req->wb_bytes = size;
858
	return size > 0;
859 860 861
}

/**
862
 * nfs_pageio_do_add_request - Attempt to coalesce a request into a page list.
863 864 865 866 867 868
 * @desc: destination io descriptor
 * @req: request
 *
 * Returns true if the request 'req' was successfully coalesced into the
 * existing list of pages 'desc'.
 */
869 870
static int nfs_pageio_do_add_request(struct nfs_pageio_descriptor *desc,
				     struct nfs_page *req)
871
{
872
	struct nfs_page *prev = NULL;
873 874
	if (desc->pg_count != 0) {
		prev = nfs_list_entry(desc->pg_list.prev);
875
	} else {
876 877
		if (desc->pg_ops->pg_init)
			desc->pg_ops->pg_init(desc, req);
878
		desc->pg_base = req->wb_pgbase;
879
	}
880 881
	if (!nfs_can_coalesce_requests(prev, req, desc))
		return 0;
882 883
	nfs_list_remove_request(req);
	nfs_list_add_request(req, &desc->pg_list);
884
	desc->pg_count += req->wb_bytes;
885 886 887
	return 1;
}

888 889 890 891 892 893
/*
 * Helper for nfs_pageio_add_request and nfs_pageio_complete
 */
static void nfs_pageio_doio(struct nfs_pageio_descriptor *desc)
{
	if (!list_empty(&desc->pg_list)) {
894
		int error = desc->pg_ops->pg_doio(desc);
895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910
		if (error < 0)
			desc->pg_error = error;
		else
			desc->pg_bytes_written += desc->pg_count;
	}
	if (list_empty(&desc->pg_list)) {
		desc->pg_count = 0;
		desc->pg_base = 0;
	}
}

/**
 * nfs_pageio_add_request - Attempt to coalesce a request into a page list.
 * @desc: destination io descriptor
 * @req: request
 *
911 912 913
 * This may split a request into subrequests which are all part of the
 * same page group.
 *
914 915 916
 * Returns true if the request 'req' was successfully coalesced into the
 * existing list of pages 'desc'.
 */
917
static int __nfs_pageio_add_request(struct nfs_pageio_descriptor *desc,
918
			   struct nfs_page *req)
919
{
920 921 922 923
	struct nfs_page *subreq;
	unsigned int bytes_left = 0;
	unsigned int offset, pgbase;

924
	nfs_page_group_lock(req, false);
925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945

	subreq = req;
	bytes_left = subreq->wb_bytes;
	offset = subreq->wb_offset;
	pgbase = subreq->wb_pgbase;

	do {
		if (!nfs_pageio_do_add_request(desc, subreq)) {
			/* make sure pg_test call(s) did nothing */
			WARN_ON_ONCE(subreq->wb_bytes != bytes_left);
			WARN_ON_ONCE(subreq->wb_offset != offset);
			WARN_ON_ONCE(subreq->wb_pgbase != pgbase);

			nfs_page_group_unlock(req);
			desc->pg_moreio = 1;
			nfs_pageio_doio(desc);
			if (desc->pg_error < 0)
				return 0;
			if (desc->pg_recoalesce)
				return 0;
			/* retry add_request for this subreq */
946
			nfs_page_group_lock(req, false);
947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962
			continue;
		}

		/* check for buggy pg_test call(s) */
		WARN_ON_ONCE(subreq->wb_bytes + subreq->wb_pgbase > PAGE_SIZE);
		WARN_ON_ONCE(subreq->wb_bytes > bytes_left);
		WARN_ON_ONCE(subreq->wb_bytes == 0);

		bytes_left -= subreq->wb_bytes;
		offset += subreq->wb_bytes;
		pgbase += subreq->wb_bytes;

		if (bytes_left) {
			subreq = nfs_create_request(req->wb_context,
					req->wb_page,
					subreq, pgbase, bytes_left);
963 964
			if (IS_ERR(subreq))
				goto err_ptr;
965 966 967 968 969 970 971
			nfs_lock_request(subreq);
			subreq->wb_offset  = offset;
			subreq->wb_index = req->wb_index;
		}
	} while (bytes_left > 0);

	nfs_page_group_unlock(req);
972
	return 1;
973 974 975 976
err_ptr:
	desc->pg_error = PTR_ERR(subreq);
	nfs_page_group_unlock(req);
	return 0;
977 978
}

979 980 981 982 983 984 985 986 987 988
static int nfs_do_recoalesce(struct nfs_pageio_descriptor *desc)
{
	LIST_HEAD(head);

	do {
		list_splice_init(&desc->pg_list, &head);
		desc->pg_bytes_written -= desc->pg_count;
		desc->pg_count = 0;
		desc->pg_base = 0;
		desc->pg_recoalesce = 0;
989
		desc->pg_moreio = 0;
990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021

		while (!list_empty(&head)) {
			struct nfs_page *req;

			req = list_first_entry(&head, struct nfs_page, wb_list);
			nfs_list_remove_request(req);
			if (__nfs_pageio_add_request(desc, req))
				continue;
			if (desc->pg_error < 0)
				return 0;
			break;
		}
	} while (desc->pg_recoalesce);
	return 1;
}

int nfs_pageio_add_request(struct nfs_pageio_descriptor *desc,
		struct nfs_page *req)
{
	int ret;

	do {
		ret = __nfs_pageio_add_request(desc, req);
		if (ret)
			break;
		if (desc->pg_error < 0)
			break;
		ret = nfs_do_recoalesce(desc);
	} while (ret);
	return ret;
}

1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052
/*
 * nfs_pageio_resend - Transfer requests to new descriptor and resend
 * @hdr - the pgio header to move request from
 * @desc - the pageio descriptor to add requests to
 *
 * Try to move each request (nfs_page) from @hdr to @desc then attempt
 * to send them.
 *
 * Returns 0 on success and < 0 on error.
 */
int nfs_pageio_resend(struct nfs_pageio_descriptor *desc,
		      struct nfs_pgio_header *hdr)
{
	LIST_HEAD(failed);

	desc->pg_dreq = hdr->dreq;
	while (!list_empty(&hdr->pages)) {
		struct nfs_page *req = nfs_list_entry(hdr->pages.next);

		nfs_list_remove_request(req);
		if (!nfs_pageio_add_request(desc, req))
			nfs_list_add_request(req, &failed);
	}
	nfs_pageio_complete(desc);
	if (!list_empty(&failed)) {
		list_move(&failed, &hdr->pages);
		return -EIO;
	}
	return 0;
}
EXPORT_SYMBOL_GPL(nfs_pageio_resend);
1053

1054 1055 1056 1057 1058 1059
/**
 * nfs_pageio_complete - Complete I/O on an nfs_pageio_descriptor
 * @desc: pointer to io descriptor
 */
void nfs_pageio_complete(struct nfs_pageio_descriptor *desc)
{
1060 1061 1062 1063 1064 1065 1066
	for (;;) {
		nfs_pageio_doio(desc);
		if (!desc->pg_recoalesce)
			break;
		if (!nfs_do_recoalesce(desc))
			break;
	}
1067 1068
}

1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084
/**
 * nfs_pageio_cond_complete - Conditional I/O completion
 * @desc: pointer to io descriptor
 * @index: page index
 *
 * It is important to ensure that processes don't try to take locks
 * on non-contiguous ranges of pages as that might deadlock. This
 * function should be called before attempting to wait on a locked
 * nfs_page. It will complete the I/O if the page index 'index'
 * is not contiguous with the existing list of pages in 'desc'.
 */
void nfs_pageio_cond_complete(struct nfs_pageio_descriptor *desc, pgoff_t index)
{
	if (!list_empty(&desc->pg_list)) {
		struct nfs_page *prev = nfs_list_entry(desc->pg_list.prev);
		if (index != prev->wb_index + 1)
1085
			nfs_pageio_complete(desc);
1086 1087 1088
	}
}

D
David Howells 已提交
1089
int __init nfs_init_nfspagecache(void)
L
Linus Torvalds 已提交
1090 1091 1092 1093
{
	nfs_page_cachep = kmem_cache_create("nfs_page",
					    sizeof(struct nfs_page),
					    0, SLAB_HWCACHE_ALIGN,
1094
					    NULL);
L
Linus Torvalds 已提交
1095 1096 1097 1098 1099 1100
	if (nfs_page_cachep == NULL)
		return -ENOMEM;

	return 0;
}

1101
void nfs_destroy_nfspagecache(void)
L
Linus Torvalds 已提交
1102
{
1103
	kmem_cache_destroy(nfs_page_cachep);
L
Linus Torvalds 已提交
1104 1105
}

1106
static const struct rpc_call_ops nfs_pgio_common_ops = {
1107 1108 1109 1110
	.rpc_call_prepare = nfs_pgio_prepare,
	.rpc_call_done = nfs_pgio_result,
	.rpc_release = nfs_pgio_release,
};
1111 1112 1113 1114 1115

const struct nfs_pageio_ops nfs_pgio_rw_ops = {
	.pg_test = nfs_generic_pg_test,
	.pg_doio = nfs_generic_pg_pgios,
};