pagelist.c 27.4 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();
102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138
		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;
		ret = nfs_wait_bit_killable(&c->flags);
	} while (atomic_read(&c->io_count) != 0);
	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);
}

139 140 141 142 143 144
static int nfs_wait_bit_uninterruptible(void *word)
{
	io_schedule();
	return 0;
}

145 146 147 148 149 150 151 152 153 154 155 156 157
/*
 * nfs_page_group_lock - lock the head of the page group
 * @req - request in group that is to be locked
 *
 * this lock must be held if modifying the page group list
 */
void
nfs_page_group_lock(struct nfs_page *req)
{
	struct nfs_page *head = req->wb_head;

	WARN_ON_ONCE(head != head->wb_head);

158 159 160
	wait_on_bit_lock(&head->wb_flags, PG_HEADLOCK,
			nfs_wait_bit_uninterruptible,
			TASK_UNINTERRUPTIBLE);
161 162 163 164 165 166 167 168 169 170 171 172 173
}

/*
 * 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);

174
	smp_mb__before_atomic();
175
	clear_bit(PG_HEADLOCK, &head->wb_flags);
176
	smp_mb__after_atomic();
177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239
	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;

	nfs_page_group_lock(req);
	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)
{
	WARN_ON_ONCE(prev == req);

	if (!prev) {
240
		/* a head request */
241 242 243
		req->wb_head = req;
		req->wb_this_page = req;
	} else {
244
		/* a subrequest */
245 246 247 248 249 250
		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;

251 252 253 254
		/* All subrequests take a ref on the head request until
		 * nfs_page_group_destroy is called */
		kref_get(&req->wb_head->wb_kref);

255 256 257
		/* grab extra ref if head request has extra ref from
		 * the write/commit path to handle handoff between write
		 * and commit lists */
258 259
		if (test_bit(PG_INODE_REF, &prev->wb_head->wb_flags)) {
			set_bit(PG_INODE_REF, &req->wb_flags);
260
			kref_get(&req->wb_kref);
261
		}
262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277
	}
}

/*
 * 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;

278 279 280 281
	/* subrequests must release the ref on the head request */
	if (req->wb_head != req)
		nfs_release_request(req->wb_head);

282 283 284 285 286 287 288 289 290 291 292 293 294 295
	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
296 297
/**
 * nfs_create_request - Create an NFS read/write request.
298
 * @ctx: open context to use
L
Linus Torvalds 已提交
299
 * @page: page to write
300
 * @last: last nfs request created for this page group or NULL if head
L
Linus Torvalds 已提交
301 302 303 304
 * @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
305
 * create two different requests for the same page.
L
Linus Torvalds 已提交
306 307 308
 * User should ensure it is safe to sleep in this function.
 */
struct nfs_page *
309
nfs_create_request(struct nfs_open_context *ctx, struct page *page,
310 311
		   struct nfs_page *last, unsigned int offset,
		   unsigned int count)
L
Linus Torvalds 已提交
312 313
{
	struct nfs_page		*req;
314
	struct nfs_lock_context *l_ctx;
L
Linus Torvalds 已提交
315

316 317
	if (test_bit(NFS_CONTEXT_BAD, &ctx->flags))
		return ERR_PTR(-EBADF);
318 319 320 321
	/* try to allocate the request struct */
	req = nfs_page_alloc();
	if (req == NULL)
		return ERR_PTR(-ENOMEM);
L
Linus Torvalds 已提交
322

323
	/* get lock context early so we can deal with alloc failures */
324 325
	l_ctx = nfs_get_lock_context(ctx);
	if (IS_ERR(l_ctx)) {
326
		nfs_page_free(req);
327
		return ERR_CAST(l_ctx);
328
	}
329
	req->wb_lock_context = l_ctx;
330
	nfs_iocounter_inc(&l_ctx->io_count);
331

L
Linus Torvalds 已提交
332 333 334 335
	/* 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;
336
	req->wb_index	= page_file_index(page);
L
Linus Torvalds 已提交
337 338 339 340 341
	page_cache_get(page);
	req->wb_offset  = offset;
	req->wb_pgbase	= offset;
	req->wb_bytes   = count;
	req->wb_context = get_nfs_open_context(ctx);
342
	kref_init(&req->wb_kref);
343
	nfs_page_group_init(req, last);
L
Linus Torvalds 已提交
344 345 346 347
	return req;
}

/**
348
 * nfs_unlock_request - Unlock request and wake up sleepers.
L
Linus Torvalds 已提交
349 350
 * @req:
 */
351
void nfs_unlock_request(struct nfs_page *req)
L
Linus Torvalds 已提交
352 353 354 355 356
{
	if (!NFS_WBACK_BUSY(req)) {
		printk(KERN_ERR "NFS: Invalid unlock attempted\n");
		BUG();
	}
357
	smp_mb__before_atomic();
L
Linus Torvalds 已提交
358
	clear_bit(PG_BUSY, &req->wb_flags);
359
	smp_mb__after_atomic();
360
	wake_up_bit(&req->wb_flags, PG_BUSY);
361 362 363
}

/**
364 365
 * nfs_unlock_and_release_request - Unlock request and release the nfs_page
 * @req:
366
 */
367
void nfs_unlock_and_release_request(struct nfs_page *req)
368
{
369
	nfs_unlock_request(req);
L
Linus Torvalds 已提交
370 371 372
	nfs_release_request(req);
}

373
/*
L
Linus Torvalds 已提交
374 375 376
 * nfs_clear_request - Free up all resources allocated to the request
 * @req:
 *
377 378
 * Release page and open context resources associated with a read/write
 * request after it has completed.
L
Linus Torvalds 已提交
379
 */
380
static void nfs_clear_request(struct nfs_page *req)
L
Linus Torvalds 已提交
381
{
382
	struct page *page = req->wb_page;
383
	struct nfs_open_context *ctx = req->wb_context;
384
	struct nfs_lock_context *l_ctx = req->wb_lock_context;
385

386 387
	if (page != NULL) {
		page_cache_release(page);
L
Linus Torvalds 已提交
388 389
		req->wb_page = NULL;
	}
390
	if (l_ctx != NULL) {
391
		nfs_iocounter_dec(&l_ctx->io_count);
392 393 394
		nfs_put_lock_context(l_ctx);
		req->wb_lock_context = NULL;
	}
395 396 397 398
	if (ctx != NULL) {
		put_nfs_open_context(ctx);
		req->wb_context = NULL;
	}
L
Linus Torvalds 已提交
399 400 401 402 403 404 405 406
}

/**
 * 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!
 */
407
void nfs_free_request(struct nfs_page *req)
L
Linus Torvalds 已提交
408
{
409 410 411 412
	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));
413 414
	WARN_ON_ONCE(test_bit(PG_UNLOCKPAGE, &req->wb_flags));
	WARN_ON_ONCE(test_bit(PG_UPTODATE, &req->wb_flags));
415 416
	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
417

418
	/* Release struct file and open context */
L
Linus Torvalds 已提交
419 420 421 422
	nfs_clear_request(req);
	nfs_page_free(req);
}

423 424
void nfs_release_request(struct nfs_page *req)
{
425
	kref_put(&req->wb_kref, nfs_page_group_destroy);
426 427
}

L
Linus Torvalds 已提交
428 429 430 431
/**
 * nfs_wait_on_request - Wait for a request to complete.
 * @req: request to wait upon.
 *
432
 * Interruptible by fatal signals only.
L
Linus Torvalds 已提交
433 434 435 436 437
 * The user is responsible for holding a count on the request.
 */
int
nfs_wait_on_request(struct nfs_page *req)
{
438 439 440
	return wait_on_bit(&req->wb_flags, PG_BUSY,
			nfs_wait_bit_uninterruptible,
			TASK_UNINTERRUPTIBLE);
L
Linus Torvalds 已提交
441 442
}

443 444 445 446 447 448 449 450 451 452 453
/*
 * 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)
454
{
455 456 457
	if (desc->pg_count > desc->pg_bsize) {
		/* should never happen */
		WARN_ON_ONCE(1);
458
		return 0;
459
	}
460

461
	return min(desc->pg_bsize - desc->pg_count, (size_t)req->wb_bytes);
462
}
463
EXPORT_SYMBOL_GPL(nfs_generic_pg_test);
464

465 466 467 468 469
static inline struct nfs_rw_header *NFS_RW_HEADER(struct nfs_pgio_header *hdr)
{
	return container_of(hdr, struct nfs_rw_header, header);
}

470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499
/**
 * nfs_rw_header_alloc - Allocate a header for a read or write
 * @ops: Read or write function vector
 */
struct nfs_rw_header *nfs_rw_header_alloc(const struct nfs_rw_ops *ops)
{
	struct nfs_rw_header *header = ops->rw_alloc_header();

	if (header) {
		struct nfs_pgio_header *hdr = &header->header;

		INIT_LIST_HEAD(&hdr->pages);
		spin_lock_init(&hdr->lock);
		atomic_set(&hdr->refcnt, 0);
		hdr->rw_ops = ops;
	}
	return header;
}
EXPORT_SYMBOL_GPL(nfs_rw_header_alloc);

/*
 * nfs_rw_header_free - Free a read or write header
 * @hdr: The header to free
 */
void nfs_rw_header_free(struct nfs_pgio_header *hdr)
{
	hdr->rw_ops->rw_free_header(NFS_RW_HEADER(hdr));
}
EXPORT_SYMBOL_GPL(nfs_rw_header_free);

500 501 502 503 504
/**
 * nfs_pgio_data_alloc - Allocate pageio data
 * @hdr: The header making a request
 * @pagecount: Number of pages to create
 */
505 506
static struct nfs_pgio_data *nfs_pgio_data_alloc(struct nfs_pgio_header *hdr,
						 unsigned int pagecount)
507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554
{
	struct nfs_pgio_data *data, *prealloc;

	prealloc = &NFS_RW_HEADER(hdr)->rpc_data;
	if (prealloc->header == NULL)
		data = prealloc;
	else
		data = kzalloc(sizeof(*data), GFP_KERNEL);
	if (!data)
		goto out;

	if (nfs_pgarray_set(&data->pages, pagecount)) {
		data->header = hdr;
		atomic_inc(&hdr->refcnt);
	} else {
		if (data != prealloc)
			kfree(data);
		data = NULL;
	}
out:
	return data;
}

/**
 * nfs_pgio_data_release - Properly free pageio data
 * @data: The data to release
 */
void nfs_pgio_data_release(struct nfs_pgio_data *data)
{
	struct nfs_pgio_header *hdr = data->header;
	struct nfs_rw_header *pageio_header = NFS_RW_HEADER(hdr);

	put_nfs_open_context(data->args.context);
	if (data->pages.pagevec != data->pages.page_array)
		kfree(data->pages.pagevec);
	if (data == &pageio_header->rpc_data) {
		data->header = NULL;
		data = NULL;
	}
	if (atomic_dec_and_test(&hdr->refcnt))
		hdr->completion_ops->completion(hdr);
	/* Note: we only free the rpc_task after callbacks are done.
	 * See the comment in rpc_free_task() for why
	 */
	kfree(data);
}
EXPORT_SYMBOL_GPL(nfs_pgio_data_release);

555 556 557 558 559 560 561 562
/**
 * nfs_pgio_rpcsetup - Set up arguments for a pageio call
 * @data: The pageio data
 * @count: Number of bytes to read
 * @offset: Initial offset
 * @how: How to commit data (writes only)
 * @cinfo: Commit information for the call (writes only)
 */
563
static void nfs_pgio_rpcsetup(struct nfs_pgio_data *data,
564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598
			      unsigned int count, unsigned int offset,
			      int how, struct nfs_commit_info *cinfo)
{
	struct nfs_page *req = data->header->req;

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

	data->args.fh     = NFS_FH(data->header->inode);
	data->args.offset = req_offset(req) + offset;
	/* pnfs_set_layoutcommit needs this */
	data->mds_offset = data->args.offset;
	data->args.pgbase = req->wb_pgbase + offset;
	data->args.pages  = data->pages.pagevec;
	data->args.count  = count;
	data->args.context = get_nfs_open_context(req->wb_context);
	data->args.lock_context = req->wb_lock_context;
	data->args.stable  = NFS_UNSTABLE;
	switch (how & (FLUSH_STABLE | FLUSH_COND_STABLE)) {
	case 0:
		break;
	case FLUSH_COND_STABLE:
		if (nfs_reqs_to_commit(cinfo))
			break;
	default:
		data->args.stable = NFS_FILE_SYNC;
	}

	data->res.fattr   = &data->fattr;
	data->res.count   = count;
	data->res.eof     = 0;
	data->res.verf    = &data->verf;
	nfs_fattr_init(&data->fattr);
}

599 600 601 602 603
/**
 * nfs_pgio_prepare - Prepare pageio data to go over the wire
 * @task: The current task
 * @calldata: pageio data to prepare
 */
604
static void nfs_pgio_prepare(struct rpc_task *task, void *calldata)
605 606 607 608 609 610 611 612
{
	struct nfs_pgio_data *data = calldata;
	int err;
	err = NFS_PROTO(data->header->inode)->pgio_rpc_prepare(task, data);
	if (err)
		rpc_exit(task, err);
}

613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658
int nfs_initiate_pgio(struct rpc_clnt *clnt, struct nfs_pgio_data *data,
		      const struct rpc_call_ops *call_ops, int how, int flags)
{
	struct rpc_task *task;
	struct rpc_message msg = {
		.rpc_argp = &data->args,
		.rpc_resp = &data->res,
		.rpc_cred = data->header->cred,
	};
	struct rpc_task_setup task_setup_data = {
		.rpc_client = clnt,
		.task = &data->task,
		.rpc_message = &msg,
		.callback_ops = call_ops,
		.callback_data = data,
		.workqueue = nfsiod_workqueue,
		.flags = RPC_TASK_ASYNC | flags,
	};
	int ret = 0;

	data->header->rw_ops->rw_initiate(data, &msg, &task_setup_data, how);

	dprintk("NFS: %5u initiated pgio call "
		"(req %s/%llu, %u bytes @ offset %llu)\n",
		data->task.tk_pid,
		data->header->inode->i_sb->s_id,
		(unsigned long long)NFS_FILEID(data->header->inode),
		data->args.count,
		(unsigned long long)data->args.offset);

	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);

659 660 661 662 663
/**
 * nfs_pgio_error - Clean up from a pageio error
 * @desc: IO descriptor
 * @hdr: pageio header
 */
664
static int nfs_pgio_error(struct nfs_pageio_descriptor *desc,
665 666 667
			  struct nfs_pgio_header *hdr)
{
	set_bit(NFS_IOHDR_REDO, &hdr->flags);
668 669
	nfs_pgio_data_release(hdr->data);
	hdr->data = NULL;
670 671 672 673
	desc->pg_completion_ops->error_cleanup(&desc->pg_list);
	return -ENOMEM;
}

674 675 676 677
/**
 * nfs_pgio_release - Release pageio data
 * @calldata: The pageio data to release
 */
678
static void nfs_pgio_release(void *calldata)
679 680 681 682 683 684 685
{
	struct nfs_pgio_data *data = calldata;
	if (data->header->rw_ops->rw_release)
		data->header->rw_ops->rw_release(data);
	nfs_pgio_data_release(data);
}

L
Linus Torvalds 已提交
686
/**
687 688
 * nfs_pageio_init - initialise a page io descriptor
 * @desc: pointer to descriptor
689 690 691 692
 * @inode: pointer to inode
 * @doio: pointer to io function
 * @bsize: io block size
 * @io_flags: extra parameters for the io function
693
 */
694 695
void nfs_pageio_init(struct nfs_pageio_descriptor *desc,
		     struct inode *inode,
696
		     const struct nfs_pageio_ops *pg_ops,
697
		     const struct nfs_pgio_completion_ops *compl_ops,
698
		     const struct nfs_rw_ops *rw_ops,
699
		     size_t bsize,
700
		     int io_flags)
701 702
{
	INIT_LIST_HEAD(&desc->pg_list);
703
	desc->pg_bytes_written = 0;
704 705 706
	desc->pg_count = 0;
	desc->pg_bsize = bsize;
	desc->pg_base = 0;
707
	desc->pg_moreio = 0;
708
	desc->pg_recoalesce = 0;
709
	desc->pg_inode = inode;
710
	desc->pg_ops = pg_ops;
711
	desc->pg_completion_ops = compl_ops;
712
	desc->pg_rw_ops = rw_ops;
713 714
	desc->pg_ioflags = io_flags;
	desc->pg_error = 0;
715
	desc->pg_lseg = NULL;
716
	desc->pg_dreq = NULL;
717
	desc->pg_layout_private = NULL;
718
}
B
Bryan Schumaker 已提交
719
EXPORT_SYMBOL_GPL(nfs_pageio_init);
720

721 722 723 724 725
/**
 * nfs_pgio_result - Basic pageio error handling
 * @task: The task that ran
 * @calldata: Pageio data to check
 */
726
static void nfs_pgio_result(struct rpc_task *task, void *calldata)
727 728 729 730 731 732 733 734 735 736 737 738 739 740 741
{
	struct nfs_pgio_data *data = calldata;
	struct inode *inode = data->header->inode;

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

	if (data->header->rw_ops->rw_done(task, data, inode) != 0)
		return;
	if (task->tk_status < 0)
		nfs_set_pgio_error(data->header, task->tk_status, data->args.offset);
	else
		data->header->rw_ops->rw_result(task, data);
}

742 743 744 745 746 747 748 749
/*
 * 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.
 */
750 751
int nfs_generic_pgio(struct nfs_pageio_descriptor *desc,
		     struct nfs_pgio_header *hdr)
752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778
{
	struct nfs_page		*req;
	struct page		**pages;
	struct nfs_pgio_data	*data;
	struct list_head *head = &desc->pg_list;
	struct nfs_commit_info cinfo;

	data = nfs_pgio_data_alloc(hdr, nfs_page_array_len(desc->pg_base,
							   desc->pg_count));
	if (!data)
		return nfs_pgio_error(desc, hdr);

	nfs_init_cinfo(&cinfo, desc->pg_inode, desc->pg_dreq);
	pages = data->pages.pagevec;
	while (!list_empty(head)) {
		req = nfs_list_entry(head->next);
		nfs_list_remove_request(req);
		nfs_list_add_request(req, &hdr->pages);
		*pages++ = req->wb_page;
	}

	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 */
	nfs_pgio_rpcsetup(data, desc->pg_count, 0, desc->pg_ioflags, &cinfo);
779
	hdr->data = data;
780 781 782
	desc->pg_rpc_callops = &nfs_pgio_common_ops;
	return 0;
}
783
EXPORT_SYMBOL_GPL(nfs_generic_pgio);
784

785
static int nfs_generic_pg_pgios(struct nfs_pageio_descriptor *desc)
786 787 788 789 790 791 792 793 794 795 796 797 798 799 800
{
	struct nfs_rw_header *rw_hdr;
	struct nfs_pgio_header *hdr;
	int ret;

	rw_hdr = nfs_rw_header_alloc(desc->pg_rw_ops);
	if (!rw_hdr) {
		desc->pg_completion_ops->error_cleanup(&desc->pg_list);
		return -ENOMEM;
	}
	hdr = &rw_hdr->header;
	nfs_pgheader_init(desc, hdr, nfs_rw_header_free);
	atomic_inc(&hdr->refcnt);
	ret = nfs_generic_pgio(desc, hdr);
	if (ret == 0)
801 802 803
		ret = nfs_initiate_pgio(NFS_CLIENT(hdr->inode),
					hdr->data, desc->pg_rpc_callops,
					desc->pg_ioflags, 0);
804 805 806 807 808
	if (atomic_dec_and_test(&hdr->refcnt))
		hdr->completion_ops->completion(hdr);
	return ret;
}

809 810 811 812 813 814 815 816 817 818 819 820 821
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;
}

822 823 824 825 826 827 828 829 830 831 832
/**
 * 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'.
 */
833 834 835
static bool nfs_can_coalesce_requests(struct nfs_page *prev,
				      struct nfs_page *req,
				      struct nfs_pageio_descriptor *pgio)
836
{
837 838
	size_t size;

839 840 841 842 843 844 845 846 847 848
	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;
		if (req_offset(req) != req_offset(prev) + prev->wb_bytes)
			return false;
	}
849
	size = pgio->pg_ops->pg_test(pgio, prev, req);
850 851 852
	WARN_ON_ONCE(size > req->wb_bytes);
	if (size && size < req->wb_bytes)
		req->wb_bytes = size;
853
	return size > 0;
854 855 856
}

/**
857
 * nfs_pageio_do_add_request - Attempt to coalesce a request into a page list.
858 859 860 861 862 863
 * @desc: destination io descriptor
 * @req: request
 *
 * Returns true if the request 'req' was successfully coalesced into the
 * existing list of pages 'desc'.
 */
864 865
static int nfs_pageio_do_add_request(struct nfs_pageio_descriptor *desc,
				     struct nfs_page *req)
866
{
867
	struct nfs_page *prev = NULL;
868 869
	if (desc->pg_count != 0) {
		prev = nfs_list_entry(desc->pg_list.prev);
870
	} else {
871 872
		if (desc->pg_ops->pg_init)
			desc->pg_ops->pg_init(desc, req);
873
		desc->pg_base = req->wb_pgbase;
874
	}
875 876
	if (!nfs_can_coalesce_requests(prev, req, desc))
		return 0;
877 878
	nfs_list_remove_request(req);
	nfs_list_add_request(req, &desc->pg_list);
879
	desc->pg_count += req->wb_bytes;
880 881 882
	return 1;
}

883 884 885 886 887 888
/*
 * 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)) {
889
		int error = desc->pg_ops->pg_doio(desc);
890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905
		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
 *
906 907 908
 * This may split a request into subrequests which are all part of the
 * same page group.
 *
909 910 911
 * Returns true if the request 'req' was successfully coalesced into the
 * existing list of pages 'desc'.
 */
912
static int __nfs_pageio_add_request(struct nfs_pageio_descriptor *desc,
913
			   struct nfs_page *req)
914
{
915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957
	struct nfs_page *subreq;
	unsigned int bytes_left = 0;
	unsigned int offset, pgbase;

	nfs_page_group_lock(req);

	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 */
			nfs_page_group_lock(req);
			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);
958 959
			if (IS_ERR(subreq))
				goto err_ptr;
960 961 962 963 964 965 966
			nfs_lock_request(subreq);
			subreq->wb_offset  = offset;
			subreq->wb_index = req->wb_index;
		}
	} while (bytes_left > 0);

	nfs_page_group_unlock(req);
967
	return 1;
968 969 970 971
err_ptr:
	desc->pg_error = PTR_ERR(subreq);
	nfs_page_group_unlock(req);
	return 0;
972 973
}

974 975 976 977 978 979 980 981 982 983
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;
984
		desc->pg_moreio = 0;
985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015

		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;
}
B
Bryan Schumaker 已提交
1016
EXPORT_SYMBOL_GPL(nfs_pageio_add_request);
1017

1018 1019 1020 1021 1022 1023
/**
 * 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)
{
1024 1025 1026 1027 1028 1029 1030
	for (;;) {
		nfs_pageio_doio(desc);
		if (!desc->pg_recoalesce)
			break;
		if (!nfs_do_recoalesce(desc))
			break;
	}
1031
}
B
Bryan Schumaker 已提交
1032
EXPORT_SYMBOL_GPL(nfs_pageio_complete);
1033

1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049
/**
 * 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)
1050
			nfs_pageio_complete(desc);
1051 1052 1053
	}
}

D
David Howells 已提交
1054
int __init nfs_init_nfspagecache(void)
L
Linus Torvalds 已提交
1055 1056 1057 1058
{
	nfs_page_cachep = kmem_cache_create("nfs_page",
					    sizeof(struct nfs_page),
					    0, SLAB_HWCACHE_ALIGN,
1059
					    NULL);
L
Linus Torvalds 已提交
1060 1061 1062 1063 1064 1065
	if (nfs_page_cachep == NULL)
		return -ENOMEM;

	return 0;
}

1066
void nfs_destroy_nfspagecache(void)
L
Linus Torvalds 已提交
1067
{
1068
	kmem_cache_destroy(nfs_page_cachep);
L
Linus Torvalds 已提交
1069 1070
}

1071
static const struct rpc_call_ops nfs_pgio_common_ops = {
1072 1073 1074 1075
	.rpc_call_prepare = nfs_pgio_prepare,
	.rpc_call_done = nfs_pgio_result,
	.rpc_release = nfs_pgio_release,
};
1076 1077 1078 1079 1080

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