read.c 17.1 KB
Newer Older
L
Linus Torvalds 已提交
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
/*
 * linux/fs/nfs/read.c
 *
 * Block I/O for NFS
 *
 * Partial copy of Linus' read cache modifications to fs/nfs/file.c
 * modified for async RPC by okir@monad.swb.de
 */

#include <linux/time.h>
#include <linux/kernel.h>
#include <linux/errno.h>
#include <linux/fcntl.h>
#include <linux/stat.h>
#include <linux/mm.h>
#include <linux/slab.h>
#include <linux/pagemap.h>
#include <linux/sunrpc/clnt.h>
#include <linux/nfs_fs.h>
#include <linux/nfs_page.h>
A
Andy Adamson 已提交
21
#include <linux/module.h>
L
Linus Torvalds 已提交
22

23
#include "nfs4_fs.h"
24
#include "internal.h"
C
Chuck Lever 已提交
25
#include "iostat.h"
26
#include "fscache.h"
C
Chuck Lever 已提交
27

L
Linus Torvalds 已提交
28 29
#define NFSDBG_FACILITY		NFSDBG_PAGECACHE

30
static const struct nfs_pageio_ops nfs_pageio_read_ops;
31
static const struct rpc_call_ops nfs_read_common_ops;
32
static const struct nfs_pgio_completion_ops nfs_async_read_completion_ops;
L
Linus Torvalds 已提交
33

34
static struct kmem_cache *nfs_rdata_cachep;
L
Linus Torvalds 已提交
35

T
Trond Myklebust 已提交
36
struct nfs_read_header *nfs_readhdr_alloc(void)
37
{
38
	struct nfs_read_header *rhdr;
39

40 41 42
	rhdr = kmem_cache_zalloc(nfs_rdata_cachep, GFP_KERNEL);
	if (rhdr) {
		struct nfs_pgio_header *hdr = &rhdr->header;
43 44

		INIT_LIST_HEAD(&hdr->pages);
45 46 47 48 49 50 51
		INIT_LIST_HEAD(&hdr->rpc_list);
		spin_lock_init(&hdr->lock);
		atomic_set(&hdr->refcnt, 0);
	}
	return rhdr;
}

52 53
static struct nfs_read_data *nfs_readdata_alloc(struct nfs_pgio_header *hdr,
						unsigned int pagecount)
54 55 56 57 58 59 60 61 62 63 64 65
{
	struct nfs_read_data *data, *prealloc;

	prealloc = &container_of(hdr, struct nfs_read_header, header)->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)) {
66
		data->header = hdr;
67 68 69 70 71
		atomic_inc(&hdr->refcnt);
	} else {
		if (data != prealloc)
			kfree(data);
		data = NULL;
72
	}
73 74
out:
	return data;
75 76
}

77
void nfs_readhdr_free(struct nfs_pgio_header *hdr)
78
{
79 80 81
	struct nfs_read_header *rhdr = container_of(hdr, struct nfs_read_header, header);

	kmem_cache_free(nfs_rdata_cachep, rhdr);
82 83
}

84
void nfs_readdata_release(struct nfs_read_data *rdata)
L
Linus Torvalds 已提交
85
{
86 87 88
	struct nfs_pgio_header *hdr = rdata->header;
	struct nfs_read_header *read_header = container_of(hdr, struct nfs_read_header, header);

89
	put_nfs_open_context(rdata->args.context);
F
Fred Isaman 已提交
90 91
	if (rdata->pages.pagevec != rdata->pages.page_array)
		kfree(rdata->pages.pagevec);
92 93 94 95 96
	if (rdata != &read_header->rpc_data)
		kfree(rdata);
	else
		rdata->header = NULL;
	if (atomic_dec_and_test(&hdr->refcnt))
97
		hdr->completion_ops->completion(hdr);
L
Linus Torvalds 已提交
98 99 100 101 102
}

static
int nfs_return_empty_page(struct page *page)
{
103
	zero_user(page, 0, PAGE_CACHE_SIZE);
L
Linus Torvalds 已提交
104 105 106 107 108
	SetPageUptodate(page);
	unlock_page(page);
	return 0;
}

109
void nfs_pageio_init_read(struct nfs_pageio_descriptor *pgio,
110 111
			      struct inode *inode,
			      const struct nfs_pgio_completion_ops *compl_ops)
112
{
113
	nfs_pageio_init(pgio, inode, &nfs_pageio_read_ops, compl_ops,
114 115
			NFS_SERVER(inode)->rsize, 0);
}
B
Bryan Schumaker 已提交
116
EXPORT_SYMBOL_GPL(nfs_pageio_init_read);
117

118 119 120 121 122
void nfs_pageio_reset_read_mds(struct nfs_pageio_descriptor *pgio)
{
	pgio->pg_ops = &nfs_pageio_read_ops;
	pgio->pg_bsize = NFS_SERVER(pgio->pg_inode)->rsize;
}
123
EXPORT_SYMBOL_GPL(nfs_pageio_reset_read_mds);
124

125 126
int nfs_readpage_async(struct nfs_open_context *ctx, struct inode *inode,
		       struct page *page)
L
Linus Torvalds 已提交
127 128 129
{
	struct nfs_page	*new;
	unsigned int len;
F
Fred Isaman 已提交
130
	struct nfs_pageio_descriptor pgio;
L
Linus Torvalds 已提交
131

132
	len = nfs_page_length(page);
L
Linus Torvalds 已提交
133 134 135 136 137 138 139 140
	if (len == 0)
		return nfs_return_empty_page(page);
	new = nfs_create_request(ctx, inode, page, 0, len);
	if (IS_ERR(new)) {
		unlock_page(page);
		return PTR_ERR(new);
	}
	if (len < PAGE_CACHE_SIZE)
141
		zero_user_segment(page, len, PAGE_CACHE_SIZE);
L
Linus Torvalds 已提交
142

143
	NFS_PROTO(inode)->read_pageio_init(&pgio, inode, &nfs_async_read_completion_ops);
144
	nfs_pageio_add_request(&pgio, new);
145
	nfs_pageio_complete(&pgio);
146
	NFS_I(inode)->read_io += pgio.pg_bytes_written;
L
Linus Torvalds 已提交
147 148 149 150 151
	return 0;
}

static void nfs_readpage_release(struct nfs_page *req)
{
152
	struct inode *d_inode = req->wb_context->dentry->d_inode;
153 154 155 156

	if (PageUptodate(req->wb_page))
		nfs_readpage_to_fscache(d_inode, req->wb_page, 0);

L
Linus Torvalds 已提交
157 158 159
	unlock_page(req->wb_page);

	dprintk("NFS: read done (%s/%Ld %d@%Ld)\n",
160 161
			req->wb_context->dentry->d_inode->i_sb->s_id,
			(long long)NFS_FILEID(req->wb_context->dentry->d_inode),
L
Linus Torvalds 已提交
162 163
			req->wb_bytes,
			(long long)req_offset(req));
164
	nfs_release_request(req);
L
Linus Torvalds 已提交
165 166
}

167
/* Note io was page aligned */
168
static void nfs_read_completion(struct nfs_pgio_header *hdr)
169 170 171 172 173
{
	unsigned long bytes = 0;

	if (test_bit(NFS_IOHDR_REDO, &hdr->flags))
		goto out;
174 175 176 177 178 179 180 181 182 183 184
	while (!list_empty(&hdr->pages)) {
		struct nfs_page *req = nfs_list_entry(hdr->pages.next);
		struct page *page = req->wb_page;

		if (test_bit(NFS_IOHDR_EOF, &hdr->flags)) {
			if (bytes > hdr->good_bytes)
				zero_user(page, 0, PAGE_SIZE);
			else if (hdr->good_bytes - bytes < PAGE_SIZE)
				zero_user_segment(page,
					hdr->good_bytes & ~PAGE_MASK,
					PAGE_SIZE);
185
		}
186 187
		bytes += req->wb_bytes;
		if (test_bit(NFS_IOHDR_ERROR, &hdr->flags)) {
188
			if (bytes <= hdr->good_bytes)
189 190 191 192 193
				SetPageUptodate(page);
		} else
			SetPageUptodate(page);
		nfs_list_remove_request(req);
		nfs_readpage_release(req);
194 195 196 197 198
	}
out:
	hdr->release(hdr);
}

199 200
int nfs_initiate_read(struct rpc_clnt *clnt,
		      struct nfs_read_data *data,
201
		      const struct rpc_call_ops *call_ops, int flags)
L
Linus Torvalds 已提交
202
{
203
	struct inode *inode = data->header->inode;
204
	int swap_flags = IS_SWAPFILE(inode) ? NFS_RPC_SWAPFLAGS : 0;
205
	struct rpc_task *task;
206 207 208
	struct rpc_message msg = {
		.rpc_argp = &data->args,
		.rpc_resp = &data->res,
209
		.rpc_cred = data->header->cred,
210
	};
211
	struct rpc_task_setup task_setup_data = {
212
		.task = &data->task,
A
Andy Adamson 已提交
213
		.rpc_client = clnt,
214
		.rpc_message = &msg,
215 216
		.callback_ops = call_ops,
		.callback_data = data,
217
		.workqueue = nfsiod_workqueue,
218
		.flags = RPC_TASK_ASYNC | swap_flags | flags,
219
	};
L
Linus Torvalds 已提交
220

A
Andy Adamson 已提交
221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237
	/* Set up the initial task struct. */
	NFS_PROTO(inode)->read_setup(data, &msg);

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

	task = rpc_run_task(&task_setup_data);
	if (IS_ERR(task))
		return PTR_ERR(task);
	rpc_put_task(task);
	return 0;
}
A
Andy Adamson 已提交
238
EXPORT_SYMBOL_GPL(nfs_initiate_read);
A
Andy Adamson 已提交
239 240 241 242

/*
 * Set up the NFS read request struct
 */
243
static void nfs_read_rpcsetup(struct nfs_read_data *data,
244
		unsigned int count, unsigned int offset)
A
Andy Adamson 已提交
245
{
246
	struct nfs_page *req = data->header->req;
L
Linus Torvalds 已提交
247

248
	data->args.fh     = NFS_FH(data->header->inode);
L
Linus Torvalds 已提交
249 250
	data->args.offset = req_offset(req) + offset;
	data->args.pgbase = req->wb_pgbase + offset;
F
Fred Isaman 已提交
251
	data->args.pages  = data->pages.pagevec;
L
Linus Torvalds 已提交
252
	data->args.count  = count;
253
	data->args.context = get_nfs_open_context(req->wb_context);
254
	data->args.lock_context = req->wb_lock_context;
L
Linus Torvalds 已提交
255 256 257 258

	data->res.fattr   = &data->fattr;
	data->res.count   = count;
	data->res.eof     = 0;
259
	nfs_fattr_init(&data->fattr);
260
}
L
Linus Torvalds 已提交
261

262
static int nfs_do_read(struct nfs_read_data *data,
263
		const struct rpc_call_ops *call_ops)
264
{
265
	struct inode *inode = data->header->inode;
266

267
	return nfs_initiate_read(NFS_CLIENT(inode), data, call_ops, 0);
L
Linus Torvalds 已提交
268 269
}

270 271
static int
nfs_do_multiple_reads(struct list_head *head,
272
		const struct rpc_call_ops *call_ops)
273 274 275 276 277 278 279
{
	struct nfs_read_data *data;
	int ret = 0;

	while (!list_empty(head)) {
		int ret2;

280
		data = list_first_entry(head, struct nfs_read_data, list);
281 282
		list_del_init(&data->list);

283
		ret2 = nfs_do_read(data, call_ops);
284 285 286 287 288 289
		if (ret == 0)
			ret = ret2;
	}
	return ret;
}

290
static void
L
Linus Torvalds 已提交
291 292 293 294 295 296 297 298 299 300 301
nfs_async_read_error(struct list_head *head)
{
	struct nfs_page	*req;

	while (!list_empty(head)) {
		req = nfs_list_entry(head->next);
		nfs_list_remove_request(req);
		nfs_readpage_release(req);
	}
}

302 303 304 305 306
static const struct nfs_pgio_completion_ops nfs_async_read_completion_ops = {
	.error_cleanup = nfs_async_read_error,
	.completion = nfs_read_completion,
};

307 308 309 310 311 312 313 314 315 316 317 318 319
static void nfs_pagein_error(struct nfs_pageio_descriptor *desc,
		struct nfs_pgio_header *hdr)
{
	set_bit(NFS_IOHDR_REDO, &hdr->flags);
	while (!list_empty(&hdr->rpc_list)) {
		struct nfs_read_data *data = list_first_entry(&hdr->rpc_list,
				struct nfs_read_data, list);
		list_del(&data->list);
		nfs_readdata_release(data);
	}
	desc->pg_completion_ops->error_cleanup(&desc->pg_list);
}

L
Linus Torvalds 已提交
320 321 322 323 324 325 326 327 328 329 330 331 332
/*
 * Generate multiple requests to fill a single page.
 *
 * We optimize to reduce the number of read operations on the wire.  If we
 * detect that we're reading a page, or an area of a page, that is past the
 * end of file, we do not generate NFS read operations but just clear the
 * parts of the page that would have come back zero from the server anyway.
 *
 * We rely on the cached value of i_size to make this determination; another
 * client can fill pages on the server past our cached end-of-file, but we
 * won't see the new data until our attribute cache is updated.  This is more
 * or less conventional NFS client behavior.
 */
333 334
static int nfs_pagein_multi(struct nfs_pageio_descriptor *desc,
			    struct nfs_pgio_header *hdr)
L
Linus Torvalds 已提交
335
{
336
	struct nfs_page *req = hdr->req;
L
Linus Torvalds 已提交
337 338
	struct page *page = req->wb_page;
	struct nfs_read_data *data;
339
	size_t rsize = desc->pg_bsize, nbytes;
340
	unsigned int offset;
L
Linus Torvalds 已提交
341

342
	offset = 0;
F
Fred Isaman 已提交
343
	nbytes = desc->pg_count;
344 345 346
	do {
		size_t len = min(nbytes,rsize);

347
		data = nfs_readdata_alloc(hdr, 1);
348 349 350 351
		if (!data) {
			nfs_pagein_error(desc, hdr);
			return -ENOMEM;
		}
F
Fred Isaman 已提交
352
		data->pages.pagevec[0] = page;
353 354
		nfs_read_rpcsetup(data, len, offset);
		list_add(&data->list, &hdr->rpc_list);
355
		nbytes -= len;
356
		offset += len;
T
Trond Myklebust 已提交
357
	} while (nbytes != 0);
358 359 360

	nfs_list_remove_request(req);
	nfs_list_add_request(req, &hdr->pages);
361
	desc->pg_rpc_callops = &nfs_read_common_ops;
T
Trond Myklebust 已提交
362
	return 0;
L
Linus Torvalds 已提交
363 364
}

365 366
static int nfs_pagein_one(struct nfs_pageio_descriptor *desc,
			  struct nfs_pgio_header *hdr)
L
Linus Torvalds 已提交
367 368 369
{
	struct nfs_page		*req;
	struct page		**pages;
370
	struct nfs_read_data    *data;
F
Fred Isaman 已提交
371
	struct list_head *head = &desc->pg_list;
L
Linus Torvalds 已提交
372

373 374 375
	data = nfs_readdata_alloc(hdr, nfs_page_array_len(desc->pg_base,
							  desc->pg_count));
	if (!data) {
376
		nfs_pagein_error(desc, hdr);
T
Trond Myklebust 已提交
377
		return -ENOMEM;
378
	}
L
Linus Torvalds 已提交
379

F
Fred Isaman 已提交
380
	pages = data->pages.pagevec;
L
Linus Torvalds 已提交
381 382 383
	while (!list_empty(head)) {
		req = nfs_list_entry(head->next);
		nfs_list_remove_request(req);
384
		nfs_list_add_request(req, &hdr->pages);
L
Linus Torvalds 已提交
385 386 387
		*pages++ = req->wb_page;
	}

388 389 390
	nfs_read_rpcsetup(data, desc->pg_count, 0);
	list_add(&data->list, &hdr->rpc_list);
	desc->pg_rpc_callops = &nfs_read_common_ops;
T
Trond Myklebust 已提交
391
	return 0;
L
Linus Torvalds 已提交
392 393
}

394 395
int nfs_generic_pagein(struct nfs_pageio_descriptor *desc,
		       struct nfs_pgio_header *hdr)
396 397
{
	if (desc->pg_bsize < PAGE_CACHE_SIZE)
398 399
		return nfs_pagein_multi(desc, hdr);
	return nfs_pagein_one(desc, hdr);
400 401 402
}

static int nfs_generic_pg_readpages(struct nfs_pageio_descriptor *desc)
403
{
404 405
	struct nfs_read_header *rhdr;
	struct nfs_pgio_header *hdr;
406 407
	int ret;

408 409
	rhdr = nfs_readhdr_alloc();
	if (!rhdr) {
410
		desc->pg_completion_ops->error_cleanup(&desc->pg_list);
411 412 413 414 415 416
		return -ENOMEM;
	}
	hdr = &rhdr->header;
	nfs_pgheader_init(desc, hdr, nfs_readhdr_free);
	atomic_inc(&hdr->refcnt);
	ret = nfs_generic_pagein(desc, hdr);
417
	if (ret == 0)
418 419 420
		ret = nfs_do_multiple_reads(&hdr->rpc_list,
					    desc->pg_rpc_callops);
	if (atomic_dec_and_test(&hdr->refcnt))
421
		hdr->completion_ops->completion(hdr);
422
	return ret;
423 424 425 426 427 428 429
}

static const struct nfs_pageio_ops nfs_pageio_read_ops = {
	.pg_test = nfs_generic_pg_test,
	.pg_doio = nfs_generic_pg_readpages,
};

430 431 432 433 434 435
/*
 * This is the callback from RPC telling us whether a reply was
 * received or some error occurred (timeout or socket shutdown).
 */
int nfs_readpage_result(struct rpc_task *task, struct nfs_read_data *data)
{
436
	struct inode *inode = data->header->inode;
437 438
	int status;

439
	dprintk("NFS: %s: %5u, (status %d)\n", __func__, task->tk_pid,
440 441
			task->tk_status);

442
	status = NFS_PROTO(inode)->read_done(task, data);
443 444 445
	if (status != 0)
		return status;

446
	nfs_add_stats(inode, NFSIOS_SERVERREADBYTES, data->res.count);
447 448

	if (task->tk_status == -ESTALE) {
449 450
		set_bit(NFS_INO_STALE, &NFS_I(inode)->flags);
		nfs_mark_for_revalidate(inode);
451 452 453 454
	}
	return 0;
}

455
static void nfs_readpage_retry(struct rpc_task *task, struct nfs_read_data *data)
456 457 458 459 460
{
	struct nfs_readargs *argp = &data->args;
	struct nfs_readres *resp = &data->res;

	/* This is a short read! */
461
	nfs_inc_stats(data->header->inode, NFSIOS_SHORTREAD);
462
	/* Has the server at least made some progress? */
463 464
	if (resp->count == 0) {
		nfs_set_pgio_error(data->header, -EIO, argp->offset);
465
		return;
466
	}
467
	/* Yes, so retry the read at the end of the data */
468
	data->mds_offset += resp->count;
469 470 471
	argp->offset += resp->count;
	argp->pgbase += resp->count;
	argp->count -= resp->count;
472
	rpc_restart_call_prepare(task);
473 474
}

475
static void nfs_readpage_result_common(struct rpc_task *task, void *calldata)
L
Linus Torvalds 已提交
476
{
T
Trond Myklebust 已提交
477
	struct nfs_read_data *data = calldata;
478 479 480
	struct nfs_pgio_header *hdr = data->header;

	/* Note the only returns of nfs_readpage_result are 0 and -EAGAIN */
T
Trond Myklebust 已提交
481 482
	if (nfs_readpage_result(task, data) != 0)
		return;
483
	if (task->tk_status < 0)
484 485 486 487 488 489 490 491 492 493 494 495 496 497
		nfs_set_pgio_error(hdr, task->tk_status, data->args.offset);
	else if (data->res.eof) {
		loff_t bound;

		bound = data->args.offset + data->res.count;
		spin_lock(&hdr->lock);
		if (bound < hdr->io_start + hdr->good_bytes) {
			set_bit(NFS_IOHDR_EOF, &hdr->flags);
			clear_bit(NFS_IOHDR_ERROR, &hdr->flags);
			hdr->good_bytes = bound - hdr->io_start;
		}
		spin_unlock(&hdr->lock);
	} else if (data->res.count != data->args.count)
		nfs_readpage_retry(task, data);
498 499
}

500
static void nfs_readpage_release_common(void *calldata)
501
{
502
	nfs_readdata_release(calldata);
L
Linus Torvalds 已提交
503 504
}

505 506 507
void nfs_read_prepare(struct rpc_task *task, void *calldata)
{
	struct nfs_read_data *data = calldata;
508
	NFS_PROTO(data->header->inode)->read_rpc_prepare(task, data);
509 510
}

511
static const struct rpc_call_ops nfs_read_common_ops = {
512
	.rpc_call_prepare = nfs_read_prepare,
513 514
	.rpc_call_done = nfs_readpage_result_common,
	.rpc_release = nfs_readpage_release_common,
T
Trond Myklebust 已提交
515 516
};

L
Linus Torvalds 已提交
517 518 519 520 521 522 523 524 525 526 527 528 529 530
/*
 * Read a page over NFS.
 * We read the page synchronously in the following case:
 *  -	The error flag is set for this page. This happens only when a
 *	previous async read operation failed.
 */
int nfs_readpage(struct file *file, struct page *page)
{
	struct nfs_open_context *ctx;
	struct inode *inode = page->mapping->host;
	int		error;

	dprintk("NFS: nfs_readpage (%p %ld@%lu)\n",
		page, PAGE_CACHE_SIZE, page->index);
C
Chuck Lever 已提交
531 532 533
	nfs_inc_stats(inode, NFSIOS_VFSREADPAGE);
	nfs_add_stats(inode, NFSIOS_READPAGES, 1);

L
Linus Torvalds 已提交
534 535 536 537 538 539 540 541 542
	/*
	 * Try to flush any pending writes to the file..
	 *
	 * NOTE! Because we own the page lock, there cannot
	 * be any new pending writes generated at this point
	 * for this page (other pages can be written to).
	 */
	error = nfs_wb_page(inode, page);
	if (error)
543 544 545
		goto out_unlock;
	if (PageUptodate(page))
		goto out_unlock;
L
Linus Torvalds 已提交
546

547 548
	error = -ESTALE;
	if (NFS_STALE(inode))
549
		goto out_unlock;
550

L
Linus Torvalds 已提交
551
	if (file == NULL) {
552
		error = -EBADF;
553
		ctx = nfs_find_open_context(inode, NULL, FMODE_READ);
L
Linus Torvalds 已提交
554
		if (ctx == NULL)
555
			goto out_unlock;
L
Linus Torvalds 已提交
556
	} else
557
		ctx = get_nfs_open_context(nfs_file_open_context(file));
L
Linus Torvalds 已提交
558

559 560 561 562 563 564
	if (!IS_SYNC(inode)) {
		error = nfs_readpage_from_fscache(ctx, inode, page);
		if (error == 0)
			goto out;
	}

565 566
	error = nfs_readpage_async(ctx, inode, page);

567
out:
L
Linus Torvalds 已提交
568 569
	put_nfs_open_context(ctx);
	return error;
570
out_unlock:
L
Linus Torvalds 已提交
571 572 573 574 575
	unlock_page(page);
	return error;
}

struct nfs_readdesc {
576
	struct nfs_pageio_descriptor *pgio;
L
Linus Torvalds 已提交
577 578 579 580 581 582 583 584 585 586
	struct nfs_open_context *ctx;
};

static int
readpage_async_filler(void *data, struct page *page)
{
	struct nfs_readdesc *desc = (struct nfs_readdesc *)data;
	struct inode *inode = page->mapping->host;
	struct nfs_page *new;
	unsigned int len;
587 588
	int error;

589
	len = nfs_page_length(page);
L
Linus Torvalds 已提交
590 591
	if (len == 0)
		return nfs_return_empty_page(page);
592

L
Linus Torvalds 已提交
593
	new = nfs_create_request(desc->ctx, inode, page, 0, len);
594 595 596
	if (IS_ERR(new))
		goto out_error;

L
Linus Torvalds 已提交
597
	if (len < PAGE_CACHE_SIZE)
598
		zero_user_segment(page, len, PAGE_CACHE_SIZE);
599 600 601 602
	if (!nfs_pageio_add_request(desc->pgio, new)) {
		error = desc->pgio->pg_error;
		goto out_unlock;
	}
L
Linus Torvalds 已提交
603
	return 0;
604 605 606 607 608
out_error:
	error = PTR_ERR(new);
out_unlock:
	unlock_page(page);
	return error;
L
Linus Torvalds 已提交
609 610 611 612 613
}

int nfs_readpages(struct file *filp, struct address_space *mapping,
		struct list_head *pages, unsigned nr_pages)
{
614
	struct nfs_pageio_descriptor pgio;
L
Linus Torvalds 已提交
615
	struct nfs_readdesc desc = {
616
		.pgio = &pgio,
L
Linus Torvalds 已提交
617 618
	};
	struct inode *inode = mapping->host;
619
	unsigned long npages;
620
	int ret = -ESTALE;
L
Linus Torvalds 已提交
621 622 623 624 625

	dprintk("NFS: nfs_readpages (%s/%Ld %d)\n",
			inode->i_sb->s_id,
			(long long)NFS_FILEID(inode),
			nr_pages);
C
Chuck Lever 已提交
626
	nfs_inc_stats(inode, NFSIOS_VFSREADPAGES);
L
Linus Torvalds 已提交
627

628 629 630
	if (NFS_STALE(inode))
		goto out;

L
Linus Torvalds 已提交
631
	if (filp == NULL) {
632
		desc.ctx = nfs_find_open_context(inode, NULL, FMODE_READ);
L
Linus Torvalds 已提交
633 634 635
		if (desc.ctx == NULL)
			return -EBADF;
	} else
636
		desc.ctx = get_nfs_open_context(nfs_file_open_context(filp));
637 638 639 640 641 642 643 644 645

	/* attempt to read as many of the pages as possible from the cache
	 * - this returns -ENOBUFS immediately if the cookie is negative
	 */
	ret = nfs_readpages_from_fscache(desc.ctx, inode, mapping,
					 pages, &nr_pages);
	if (ret == 0)
		goto read_complete; /* all pages were read */

646
	NFS_PROTO(inode)->read_pageio_init(&pgio, inode, &nfs_async_read_completion_ops);
647

L
Linus Torvalds 已提交
648
	ret = read_cache_pages(mapping, pages, readpage_async_filler, &desc);
649 650

	nfs_pageio_complete(&pgio);
651
	NFS_I(inode)->read_io += pgio.pg_bytes_written;
652 653
	npages = (pgio.pg_bytes_written + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
	nfs_add_stats(inode, NFSIOS_READPAGES, npages);
654
read_complete:
L
Linus Torvalds 已提交
655
	put_nfs_open_context(desc.ctx);
656
out:
L
Linus Torvalds 已提交
657 658 659
	return ret;
}

D
David Howells 已提交
660
int __init nfs_init_readpagecache(void)
L
Linus Torvalds 已提交
661 662
{
	nfs_rdata_cachep = kmem_cache_create("nfs_read_data",
663
					     sizeof(struct nfs_read_header),
L
Linus Torvalds 已提交
664
					     0, SLAB_HWCACHE_ALIGN,
665
					     NULL);
L
Linus Torvalds 已提交
666 667 668 669 670 671
	if (nfs_rdata_cachep == NULL)
		return -ENOMEM;

	return 0;
}

672
void nfs_destroy_readpagecache(void)
L
Linus Torvalds 已提交
673
{
674
	kmem_cache_destroy(nfs_rdata_cachep);
L
Linus Torvalds 已提交
675
}