pagelist.c 10.3 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 29 30 31

static inline struct nfs_page *
nfs_page_alloc(void)
{
32 33
	struct nfs_page	*p = kmem_cache_zalloc(nfs_page_cachep, GFP_KERNEL);
	if (p)
L
Linus Torvalds 已提交
34 35 36 37 38 39 40 41 42 43 44 45
		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.
46
 * @ctx: open context to use
L
Linus Torvalds 已提交
47 48 49 50 51 52
 * @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
53
 * create two different requests for the same page.
L
Linus Torvalds 已提交
54 55 56 57 58 59 60 61 62
 * 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;

63 64 65 66
	/* try to allocate the request struct */
	req = nfs_page_alloc();
	if (req == NULL)
		return ERR_PTR(-ENOMEM);
L
Linus Torvalds 已提交
67

68 69 70 71 72 73 74
	/* 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
75 76 77 78 79 80 81
	/* 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;
	atomic_set(&req->wb_complete, 0);
	req->wb_index	= page->index;
	page_cache_get(page);
82 83 84
	BUG_ON(PagePrivate(page));
	BUG_ON(!PageLocked(page));
	BUG_ON(page->mapping->host != inode);
L
Linus Torvalds 已提交
85 86 87 88
	req->wb_offset  = offset;
	req->wb_pgbase	= offset;
	req->wb_bytes   = count;
	req->wb_context = get_nfs_open_context(ctx);
89
	kref_init(&req->wb_kref);
L
Linus Torvalds 已提交
90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105
	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();
106
	wake_up_bit(&req->wb_flags, PG_BUSY);
L
Linus Torvalds 已提交
107 108 109
	nfs_release_request(req);
}

110
/*
L
Linus Torvalds 已提交
111 112 113
 * nfs_clear_request - Free up all resources allocated to the request
 * @req:
 *
114 115
 * Release page and open context resources associated with a read/write
 * request after it has completed.
L
Linus Torvalds 已提交
116
 */
117
static void nfs_clear_request(struct nfs_page *req)
L
Linus Torvalds 已提交
118
{
119
	struct page *page = req->wb_page;
120
	struct nfs_open_context *ctx = req->wb_context;
121
	struct nfs_lock_context *l_ctx = req->wb_lock_context;
122

123 124
	if (page != NULL) {
		page_cache_release(page);
L
Linus Torvalds 已提交
125 126
		req->wb_page = NULL;
	}
127 128 129 130
	if (l_ctx != NULL) {
		nfs_put_lock_context(l_ctx);
		req->wb_lock_context = NULL;
	}
131 132 133 134
	if (ctx != NULL) {
		put_nfs_open_context(ctx);
		req->wb_context = NULL;
	}
L
Linus Torvalds 已提交
135 136 137 138 139 140 141 142 143
}


/**
 * 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!
 */
144
static void nfs_free_request(struct kref *kref)
L
Linus Torvalds 已提交
145
{
146
	struct nfs_page *req = container_of(kref, struct nfs_page, wb_kref);
L
Linus Torvalds 已提交
147

148
	/* Release struct file and open context */
L
Linus Torvalds 已提交
149 150 151 152
	nfs_clear_request(req);
	nfs_page_free(req);
}

153 154 155 156 157
void nfs_release_request(struct nfs_page *req)
{
	kref_put(&req->wb_kref, nfs_free_request);
}

158 159 160 161 162 163
static int nfs_wait_bit_uninterruptible(void *word)
{
	io_schedule();
	return 0;
}

L
Linus Torvalds 已提交
164 165 166 167
/**
 * nfs_wait_on_request - Wait for a request to complete.
 * @req: request to wait upon.
 *
168
 * Interruptible by fatal signals only.
L
Linus Torvalds 已提交
169 170 171 172 173
 * The user is responsible for holding a count on the request.
 */
int
nfs_wait_on_request(struct nfs_page *req)
{
174 175 176
	return wait_on_bit(&req->wb_flags, PG_BUSY,
			nfs_wait_bit_uninterruptible,
			TASK_UNINTERRUPTIBLE);
L
Linus Torvalds 已提交
177 178
}

179
bool nfs_generic_pg_test(struct nfs_pageio_descriptor *desc, struct nfs_page *prev, struct nfs_page *req)
180 181 182 183 184 185 186 187 188 189 190 191 192
{
	/*
	 * 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;
}
193
EXPORT_SYMBOL_GPL(nfs_generic_pg_test);
194

L
Linus Torvalds 已提交
195
/**
196 197
 * nfs_pageio_init - initialise a page io descriptor
 * @desc: pointer to descriptor
198 199 200 201
 * @inode: pointer to inode
 * @doio: pointer to io function
 * @bsize: io block size
 * @io_flags: extra parameters for the io function
202
 */
203 204
void nfs_pageio_init(struct nfs_pageio_descriptor *desc,
		     struct inode *inode,
205
		     const struct nfs_pageio_ops *pg_ops,
206
		     size_t bsize,
207
		     int io_flags)
208 209
{
	INIT_LIST_HEAD(&desc->pg_list);
210
	desc->pg_bytes_written = 0;
211 212 213
	desc->pg_count = 0;
	desc->pg_bsize = bsize;
	desc->pg_base = 0;
214
	desc->pg_moreio = 0;
215
	desc->pg_recoalesce = 0;
216
	desc->pg_inode = inode;
217
	desc->pg_ops = pg_ops;
218 219
	desc->pg_ioflags = io_flags;
	desc->pg_error = 0;
220
	desc->pg_lseg = NULL;
221 222 223 224 225 226 227 228 229 230 231 232 233
}

/**
 * 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'.
 */
234 235 236
static bool nfs_can_coalesce_requests(struct nfs_page *prev,
				      struct nfs_page *req,
				      struct nfs_pageio_descriptor *pgio)
237 238
{
	if (req->wb_context->cred != prev->wb_context->cred)
239
		return false;
240
	if (req->wb_lock_context->lockowner != prev->wb_lock_context->lockowner)
241
		return false;
242
	if (req->wb_context->state != prev->wb_context->state)
243
		return false;
244
	if (req->wb_index != (prev->wb_index + 1))
245
		return false;
246
	if (req->wb_pgbase != 0)
247
		return false;
248
	if (prev->wb_pgbase + prev->wb_bytes != PAGE_CACHE_SIZE)
249
		return false;
250
	return pgio->pg_ops->pg_test(pgio, prev, req);
251 252 253
}

/**
254
 * nfs_pageio_do_add_request - Attempt to coalesce a request into a page list.
255 256 257 258 259 260
 * @desc: destination io descriptor
 * @req: request
 *
 * Returns true if the request 'req' was successfully coalesced into the
 * existing list of pages 'desc'.
 */
261 262
static int nfs_pageio_do_add_request(struct nfs_pageio_descriptor *desc,
				     struct nfs_page *req)
263 264 265 266 267
{
	if (desc->pg_count != 0) {
		struct nfs_page *prev;

		prev = nfs_list_entry(desc->pg_list.prev);
268
		if (!nfs_can_coalesce_requests(prev, req, desc))
269
			return 0;
270
	} else {
271 272
		if (desc->pg_ops->pg_init)
			desc->pg_ops->pg_init(desc, req);
273
		desc->pg_base = req->wb_pgbase;
274
	}
275 276
	nfs_list_remove_request(req);
	nfs_list_add_request(req, &desc->pg_list);
277
	desc->pg_count += req->wb_bytes;
278 279 280
	return 1;
}

281 282 283 284 285 286
/*
 * 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)) {
287
		int error = desc->pg_ops->pg_doio(desc);
288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306
		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'.
 */
307
static int __nfs_pageio_add_request(struct nfs_pageio_descriptor *desc,
308
			   struct nfs_page *req)
309 310
{
	while (!nfs_pageio_do_add_request(desc, req)) {
311
		desc->pg_moreio = 1;
312 313 314
		nfs_pageio_doio(desc);
		if (desc->pg_error < 0)
			return 0;
315
		desc->pg_moreio = 0;
316 317
		if (desc->pg_recoalesce)
			return 0;
318 319 320 321
	}
	return 1;
}

322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363
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;
}

364 365 366 367 368 369
/**
 * 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)
{
370 371 372 373 374 375 376
	for (;;) {
		nfs_pageio_doio(desc);
		if (!desc->pg_recoalesce)
			break;
		if (!nfs_do_recoalesce(desc))
			break;
	}
377 378
}

379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394
/**
 * 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)
395
			nfs_pageio_complete(desc);
396 397 398
	}
}

D
David Howells 已提交
399
int __init nfs_init_nfspagecache(void)
L
Linus Torvalds 已提交
400 401 402 403
{
	nfs_page_cachep = kmem_cache_create("nfs_page",
					    sizeof(struct nfs_page),
					    0, SLAB_HWCACHE_ALIGN,
404
					    NULL);
L
Linus Torvalds 已提交
405 406 407 408 409 410
	if (nfs_page_cachep == NULL)
		return -ENOMEM;

	return 0;
}

411
void nfs_destroy_nfspagecache(void)
L
Linus Torvalds 已提交
412
{
413
	kmem_cache_destroy(nfs_page_cachep);
L
Linus Torvalds 已提交
414 415
}