pagelist.c 11.4 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
static struct kmem_cache *nfs_page_cachep;
L
Linus Torvalds 已提交
28

F
Fred Isaman 已提交
29 30 31 32 33 34 35 36 37 38 39 40 41
bool nfs_pgarray_set(struct nfs_page_array *p, unsigned int pagecount)
{
	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;
}

42 43 44 45 46 47 48 49 50
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;
51
	hdr->dreq = desc->pg_dreq;
52
	hdr->release = release;
53
	hdr->completion_ops = desc->pg_completion_ops;
54 55
	if (hdr->completion_ops->init_hdr)
		hdr->completion_ops->init_hdr(hdr);
56 57 58 59 60 61 62 63 64 65 66 67 68 69
}

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
70 71 72
static inline struct nfs_page *
nfs_page_alloc(void)
{
73 74
	struct nfs_page	*p = kmem_cache_zalloc(nfs_page_cachep, GFP_KERNEL);
	if (p)
L
Linus Torvalds 已提交
75 76 77 78 79 80 81 82 83 84 85 86
		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);
}

/**
 * nfs_create_request - Create an NFS read/write request.
87
 * @ctx: open context to use
L
Linus Torvalds 已提交
88 89 90 91 92 93
 * @inode: inode to which the request is attached
 * @page: page to write
 * @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
94
 * create two different requests for the same page.
L
Linus Torvalds 已提交
95 96 97 98 99 100 101 102 103
 * User should ensure it is safe to sleep in this function.
 */
struct nfs_page *
nfs_create_request(struct nfs_open_context *ctx, struct inode *inode,
		   struct page *page,
		   unsigned int offset, unsigned int count)
{
	struct nfs_page		*req;

104 105 106 107
	/* try to allocate the request struct */
	req = nfs_page_alloc();
	if (req == NULL)
		return ERR_PTR(-ENOMEM);
L
Linus Torvalds 已提交
108

109 110 111 112 113 114 115
	/* get lock context early so we can deal with alloc failures */
	req->wb_lock_context = nfs_get_lock_context(ctx);
	if (req->wb_lock_context == NULL) {
		nfs_page_free(req);
		return ERR_PTR(-ENOMEM);
	}

L
Linus Torvalds 已提交
116 117 118 119 120 121 122 123 124 125
	/* 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;
	req->wb_index	= page->index;
	page_cache_get(page);
	req->wb_offset  = offset;
	req->wb_pgbase	= offset;
	req->wb_bytes   = count;
	req->wb_context = get_nfs_open_context(ctx);
126
	kref_init(&req->wb_kref);
L
Linus Torvalds 已提交
127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142
	return req;
}

/**
 * nfs_unlock_request - Unlock request and wake up sleepers.
 * @req:
 */
void nfs_unlock_request(struct nfs_page *req)
{
	if (!NFS_WBACK_BUSY(req)) {
		printk(KERN_ERR "NFS: Invalid unlock attempted\n");
		BUG();
	}
	smp_mb__before_clear_bit();
	clear_bit(PG_BUSY, &req->wb_flags);
	smp_mb__after_clear_bit();
143
	wake_up_bit(&req->wb_flags, PG_BUSY);
L
Linus Torvalds 已提交
144 145 146
	nfs_release_request(req);
}

147
/*
L
Linus Torvalds 已提交
148 149 150
 * nfs_clear_request - Free up all resources allocated to the request
 * @req:
 *
151 152
 * Release page and open context resources associated with a read/write
 * request after it has completed.
L
Linus Torvalds 已提交
153
 */
154
static void nfs_clear_request(struct nfs_page *req)
L
Linus Torvalds 已提交
155
{
156
	struct page *page = req->wb_page;
157
	struct nfs_open_context *ctx = req->wb_context;
158
	struct nfs_lock_context *l_ctx = req->wb_lock_context;
159

160 161
	if (page != NULL) {
		page_cache_release(page);
L
Linus Torvalds 已提交
162 163
		req->wb_page = NULL;
	}
164 165 166 167
	if (l_ctx != NULL) {
		nfs_put_lock_context(l_ctx);
		req->wb_lock_context = NULL;
	}
168 169 170 171
	if (ctx != NULL) {
		put_nfs_open_context(ctx);
		req->wb_context = NULL;
	}
L
Linus Torvalds 已提交
172 173 174 175 176 177 178 179 180
}


/**
 * 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!
 */
181
static void nfs_free_request(struct kref *kref)
L
Linus Torvalds 已提交
182
{
183
	struct nfs_page *req = container_of(kref, struct nfs_page, wb_kref);
L
Linus Torvalds 已提交
184

185
	/* Release struct file and open context */
L
Linus Torvalds 已提交
186 187 188 189
	nfs_clear_request(req);
	nfs_page_free(req);
}

190 191 192 193 194
void nfs_release_request(struct nfs_page *req)
{
	kref_put(&req->wb_kref, nfs_free_request);
}

195 196 197 198 199 200
static int nfs_wait_bit_uninterruptible(void *word)
{
	io_schedule();
	return 0;
}

L
Linus Torvalds 已提交
201 202 203 204
/**
 * nfs_wait_on_request - Wait for a request to complete.
 * @req: request to wait upon.
 *
205
 * Interruptible by fatal signals only.
L
Linus Torvalds 已提交
206 207 208 209 210
 * The user is responsible for holding a count on the request.
 */
int
nfs_wait_on_request(struct nfs_page *req)
{
211 212 213
	return wait_on_bit(&req->wb_flags, PG_BUSY,
			nfs_wait_bit_uninterruptible,
			TASK_UNINTERRUPTIBLE);
L
Linus Torvalds 已提交
214 215
}

216
bool nfs_generic_pg_test(struct nfs_pageio_descriptor *desc, struct nfs_page *prev, struct nfs_page *req)
217 218 219 220 221 222 223 224 225 226 227 228 229
{
	/*
	 * FIXME: ideally we should be able to coalesce all requests
	 * that are not block boundary aligned, but currently this
	 * is problematic for the case of bsize < PAGE_CACHE_SIZE,
	 * since nfs_flush_multi and nfs_pagein_multi assume you
	 * can have only one struct nfs_page.
	 */
	if (desc->pg_bsize < PAGE_SIZE)
		return 0;

	return desc->pg_count + req->wb_bytes <= desc->pg_bsize;
}
230
EXPORT_SYMBOL_GPL(nfs_generic_pg_test);
231

L
Linus Torvalds 已提交
232
/**
233 234
 * nfs_pageio_init - initialise a page io descriptor
 * @desc: pointer to descriptor
235 236 237 238
 * @inode: pointer to inode
 * @doio: pointer to io function
 * @bsize: io block size
 * @io_flags: extra parameters for the io function
239
 */
240 241
void nfs_pageio_init(struct nfs_pageio_descriptor *desc,
		     struct inode *inode,
242
		     const struct nfs_pageio_ops *pg_ops,
243
		     const struct nfs_pgio_completion_ops *compl_ops,
244
		     size_t bsize,
245
		     int io_flags)
246 247
{
	INIT_LIST_HEAD(&desc->pg_list);
248
	desc->pg_bytes_written = 0;
249 250 251
	desc->pg_count = 0;
	desc->pg_bsize = bsize;
	desc->pg_base = 0;
252
	desc->pg_moreio = 0;
253
	desc->pg_recoalesce = 0;
254
	desc->pg_inode = inode;
255
	desc->pg_ops = pg_ops;
256
	desc->pg_completion_ops = compl_ops;
257 258
	desc->pg_ioflags = io_flags;
	desc->pg_error = 0;
259
	desc->pg_lseg = NULL;
260
	desc->pg_dreq = NULL;
261 262 263 264 265 266 267 268 269 270 271 272 273
}

/**
 * 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'.
 */
274 275 276
static bool nfs_can_coalesce_requests(struct nfs_page *prev,
				      struct nfs_page *req,
				      struct nfs_pageio_descriptor *pgio)
277 278
{
	if (req->wb_context->cred != prev->wb_context->cred)
279
		return false;
280
	if (req->wb_lock_context->lockowner != prev->wb_lock_context->lockowner)
281
		return false;
282
	if (req->wb_context->state != prev->wb_context->state)
283
		return false;
284
	if (req->wb_pgbase != 0)
285
		return false;
286
	if (prev->wb_pgbase + prev->wb_bytes != PAGE_CACHE_SIZE)
287
		return false;
288 289
	if (req_offset(req) != req_offset(prev) + prev->wb_bytes)
		return false;
290
	return pgio->pg_ops->pg_test(pgio, prev, req);
291 292 293
}

/**
294
 * nfs_pageio_do_add_request - Attempt to coalesce a request into a page list.
295 296 297 298 299 300
 * @desc: destination io descriptor
 * @req: request
 *
 * Returns true if the request 'req' was successfully coalesced into the
 * existing list of pages 'desc'.
 */
301 302
static int nfs_pageio_do_add_request(struct nfs_pageio_descriptor *desc,
				     struct nfs_page *req)
303 304 305 306 307
{
	if (desc->pg_count != 0) {
		struct nfs_page *prev;

		prev = nfs_list_entry(desc->pg_list.prev);
308
		if (!nfs_can_coalesce_requests(prev, req, desc))
309
			return 0;
310
	} else {
311 312
		if (desc->pg_ops->pg_init)
			desc->pg_ops->pg_init(desc, req);
313
		desc->pg_base = req->wb_pgbase;
314
	}
315 316
	nfs_list_remove_request(req);
	nfs_list_add_request(req, &desc->pg_list);
317
	desc->pg_count += req->wb_bytes;
318 319 320
	return 1;
}

321 322 323 324 325 326
/*
 * 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)) {
327
		int error = desc->pg_ops->pg_doio(desc);
328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346
		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
 *
 * Returns true if the request 'req' was successfully coalesced into the
 * existing list of pages 'desc'.
 */
347
static int __nfs_pageio_add_request(struct nfs_pageio_descriptor *desc,
348
			   struct nfs_page *req)
349 350
{
	while (!nfs_pageio_do_add_request(desc, req)) {
351
		desc->pg_moreio = 1;
352 353 354
		nfs_pageio_doio(desc);
		if (desc->pg_error < 0)
			return 0;
355
		desc->pg_moreio = 0;
356 357
		if (desc->pg_recoalesce)
			return 0;
358 359 360 361
	}
	return 1;
}

362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403
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;

		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;
}

404 405 406 407 408 409
/**
 * 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)
{
410 411 412 413 414 415 416
	for (;;) {
		nfs_pageio_doio(desc);
		if (!desc->pg_recoalesce)
			break;
		if (!nfs_do_recoalesce(desc))
			break;
	}
417 418
}

419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434
/**
 * 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)
435
			nfs_pageio_complete(desc);
436 437 438
	}
}

D
David Howells 已提交
439
int __init nfs_init_nfspagecache(void)
L
Linus Torvalds 已提交
440 441 442 443
{
	nfs_page_cachep = kmem_cache_create("nfs_page",
					    sizeof(struct nfs_page),
					    0, SLAB_HWCACHE_ALIGN,
444
					    NULL);
L
Linus Torvalds 已提交
445 446 447 448 449 450
	if (nfs_page_cachep == NULL)
		return -ENOMEM;

	return 0;
}

451
void nfs_destroy_nfspagecache(void)
L
Linus Torvalds 已提交
452
{
453
	kmem_cache_destroy(nfs_page_cachep);
L
Linus Torvalds 已提交
454 455
}