[net-next-2.6.git] / fs / nfs / pagelist.c

/*
 * 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>
#include <linux/sched.h>
#include <linux/sunrpc/clnt.h>
#include <linux/nfs3.h>
#include <linux/nfs4.h>
#include <linux/nfs_page.h>
#include <linux/nfs_fs.h>
#include <linux/nfs_mount.h>

#include "internal.h"

static struct kmem_cache *nfs_page_cachep;

static inline struct nfs_page *
nfs_page_alloc(void)
{
	struct nfs_page	*p;
	p = kmem_cache_alloc(nfs_page_cachep, GFP_KERNEL);
	if (p) {
		memset(p, 0, sizeof(*p));
		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.
 * @file: file descriptor to use
 * @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
 * create two different requests for the same page.
 * 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;

	/* try to allocate the request struct */
	req = nfs_page_alloc();
	if (req == NULL)
		return ERR_PTR(-ENOMEM);

	/* 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);
	BUG_ON(PagePrivate(page));
	BUG_ON(!PageLocked(page));
	BUG_ON(page->mapping->host != inode);
	req->wb_offset  = offset;
	req->wb_pgbase	= offset;
	req->wb_bytes   = count;
	req->wb_context = get_nfs_open_context(ctx);
	kref_init(&req->wb_kref);
	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();
	wake_up_bit(&req->wb_flags, PG_BUSY);
	nfs_release_request(req);
}

/**
 * nfs_set_page_tag_locked - Tag a request as locked
 * @req:
 */
int nfs_set_page_tag_locked(struct nfs_page *req)
{
	if (!nfs_lock_request_dontget(req))
		return 0;
	if (req->wb_page != NULL)
		radix_tree_tag_set(&NFS_I(req->wb_context->path.dentry->d_inode)->nfs_page_tree, req->wb_index, NFS_PAGE_TAG_LOCKED);
	return 1;
}

/**
 * nfs_clear_page_tag_locked - Clear request tag and wake up sleepers
 */
void nfs_clear_page_tag_locked(struct nfs_page *req)
{
	if (req->wb_page != NULL) {
		struct inode *inode = req->wb_context->path.dentry->d_inode;
		struct nfs_inode *nfsi = NFS_I(inode);

		spin_lock(&inode->i_lock);
		radix_tree_tag_clear(&nfsi->nfs_page_tree, req->wb_index, NFS_PAGE_TAG_LOCKED);
		nfs_unlock_request(req);
		spin_unlock(&inode->i_lock);
	} else
		nfs_unlock_request(req);
}

/**
 * nfs_clear_request - Free up all resources allocated to the request
 * @req:
 *
 * Release page and open context resources associated with a read/write
 * request after it has completed.
 */
void nfs_clear_request(struct nfs_page *req)
{
	struct page *page = req->wb_page;
	struct nfs_open_context *ctx = req->wb_context;

	if (page != NULL) {
		page_cache_release(page);
		req->wb_page = NULL;
	}
	if (ctx != NULL) {
		put_nfs_open_context(ctx);
		req->wb_context = NULL;
	}
}


/**
 * 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!
 */
static void nfs_free_request(struct kref *kref)
{
	struct nfs_page *req = container_of(kref, struct nfs_page, wb_kref);

	/* Release struct file and open context */
	nfs_clear_request(req);
	nfs_page_free(req);
}

void nfs_release_request(struct nfs_page *req)
{
	kref_put(&req->wb_kref, nfs_free_request);
}

static int nfs_wait_bit_uninterruptible(void *word)
{
	io_schedule();
	return 0;
}

/**
 * nfs_wait_on_request - Wait for a request to complete.
 * @req: request to wait upon.
 *
 * Interruptible by fatal signals only.
 * The user is responsible for holding a count on the request.
 */
int
nfs_wait_on_request(struct nfs_page *req)
{
	return wait_on_bit(&req->wb_flags, PG_BUSY,
			nfs_wait_bit_uninterruptible,
			TASK_UNINTERRUPTIBLE);
}

/**
 * nfs_pageio_init - initialise a page io descriptor
 * @desc: pointer to descriptor
 * @inode: pointer to inode
 * @doio: pointer to io function
 * @bsize: io block size
 * @io_flags: extra parameters for the io function
 */
void nfs_pageio_init(struct nfs_pageio_descriptor *desc,
		     struct inode *inode,
		     int (*doio)(struct inode *, struct list_head *, unsigned int, size_t, int),
		     size_t bsize,
		     int io_flags)
{
	INIT_LIST_HEAD(&desc->pg_list);
	desc->pg_bytes_written = 0;
	desc->pg_count = 0;
	desc->pg_bsize = bsize;
	desc->pg_base = 0;
	desc->pg_inode = inode;
	desc->pg_doio = doio;
	desc->pg_ioflags = io_flags;
	desc->pg_error = 0;
}

/**
 * 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'.
 */
static int nfs_can_coalesce_requests(struct nfs_page *prev,
				     struct nfs_page *req)
{
	if (req->wb_context->cred != prev->wb_context->cred)
		return 0;
	if (req->wb_context->lockowner != prev->wb_context->lockowner)
		return 0;
	if (req->wb_context->state != prev->wb_context->state)
		return 0;
	if (req->wb_index != (prev->wb_index + 1))
		return 0;
	if (req->wb_pgbase != 0)
		return 0;
	if (prev->wb_pgbase + prev->wb_bytes != PAGE_CACHE_SIZE)
		return 0;
	return 1;
}

/**
 * nfs_pageio_do_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'.
 */
static int nfs_pageio_do_add_request(struct nfs_pageio_descriptor *desc,
				     struct nfs_page *req)
{
	size_t newlen = req->wb_bytes;

	if (desc->pg_count != 0) {
		struct nfs_page *prev;

		/*
		 * 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;
		newlen += desc->pg_count;
		if (newlen > desc->pg_bsize)
			return 0;
		prev = nfs_list_entry(desc->pg_list.prev);
		if (!nfs_can_coalesce_requests(prev, req))
			return 0;
	} else
		desc->pg_base = req->wb_pgbase;
	nfs_list_remove_request(req);
	nfs_list_add_request(req, &desc->pg_list);
	desc->pg_count = newlen;
	return 1;
}

/*
 * 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)) {
		int error = desc->pg_doio(desc->pg_inode,
					  &desc->pg_list,
					  nfs_page_array_len(desc->pg_base,
							     desc->pg_count),
					  desc->pg_count,
					  desc->pg_ioflags);
		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'.
 */
int nfs_pageio_add_request(struct nfs_pageio_descriptor *desc,
			   struct nfs_page *req)
{
	while (!nfs_pageio_do_add_request(desc, req)) {
		nfs_pageio_doio(desc);
		if (desc->pg_error < 0)
			return 0;
	}
	return 1;
}

/**
 * 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)
{
	nfs_pageio_doio(desc);
}

/**
 * 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)
			nfs_pageio_doio(desc);
	}
}

#define NFS_SCAN_MAXENTRIES 16
/**
 * nfs_scan_list - Scan a list for matching requests
 * @nfsi: NFS inode
 * @dst: Destination list
 * @idx_start: lower bound of page->index to scan
 * @npages: idx_start + npages sets the upper bound to scan.
 * @tag: tag to scan for
 *
 * Moves elements from one of the inode request lists.
 * If the number of requests is set to 0, the entire address_space
 * starting at index idx_start, is scanned.
 * The requests are *not* checked to ensure that they form a contiguous set.
 * You must be holding the inode's i_lock when calling this function
 */
int nfs_scan_list(struct nfs_inode *nfsi,
		struct list_head *dst, pgoff_t idx_start,
		unsigned int npages, int tag)
{
	struct nfs_page *pgvec[NFS_SCAN_MAXENTRIES];
	struct nfs_page *req;
	pgoff_t idx_end;
	int found, i;
	int res;

	res = 0;
	if (npages == 0)
		idx_end = ~0;
	else
		idx_end = idx_start + npages - 1;

	for (;;) {
		found = radix_tree_gang_lookup_tag(&nfsi->nfs_page_tree,
				(void **)&pgvec[0], idx_start,
				NFS_SCAN_MAXENTRIES, tag);
		if (found <= 0)
			break;
		for (i = 0; i < found; i++) {
			req = pgvec[i];
			if (req->wb_index > idx_end)
				goto out;
			idx_start = req->wb_index + 1;
			if (nfs_set_page_tag_locked(req)) {
				kref_get(&req->wb_kref);
				nfs_list_remove_request(req);
				radix_tree_tag_clear(&nfsi->nfs_page_tree,
						req->wb_index, tag);
				nfs_list_add_request(req, dst);
				res++;
				if (res == INT_MAX)
					goto out;
			}
		}
		/* for latency reduction */
		cond_resched_lock(&nfsi->vfs_inode.i_lock);
	}
out:
	return res;
}

int __init nfs_init_nfspagecache(void)
{
	nfs_page_cachep = kmem_cache_create("nfs_page",
					    sizeof(struct nfs_page),
					    0, SLAB_HWCACHE_ALIGN,
					    NULL);
	if (nfs_page_cachep == NULL)
		return -ENOMEM;

	return 0;
}

void nfs_destroy_nfspagecache(void)
{
	kmem_cache_destroy(nfs_page_cachep);
}
Commit	Line	Data
1da177e4 LT	1	/*
	2	* linux/fs/nfs/pagelist.c
	3	*
	4	* A set of helper functions for managing NFS read and write requests.
	5	* The main purpose of these routines is to provide support for the
	6	* coalescing of several requests into a single RPC call.
	7	*
	8	* Copyright 2000, 2001 (c) Trond Myklebust <trond.myklebust@fys.uio.no>
	9	*
	10	*/
	11
1da177e4 LT	12	#include <linux/slab.h>
1da177e4 LT	13	#include <linux/file.h>
e8edc6e0	14	#include <linux/sched.h>
1da177e4 LT	15	#include <linux/sunrpc/clnt.h>
	16	#include <linux/nfs3.h>
	17	#include <linux/nfs4.h>
	18	#include <linux/nfs_page.h>
	19	#include <linux/nfs_fs.h>
	20	#include <linux/nfs_mount.h>
	21
8d5658c9 TM	22	#include "internal.h"
8d5658c9 TM	23
e18b890b	24	static struct kmem_cache *nfs_page_cachep;
1da177e4 LT	25
	26	static inline struct nfs_page *
	27	nfs_page_alloc(void)
	28	{
	29	struct nfs_page *p;
e94b1766	30	p = kmem_cache_alloc(nfs_page_cachep, GFP_KERNEL);
1da177e4 LT	31	if (p) {
	32	memset(p, 0, sizeof(*p));
	33	INIT_LIST_HEAD(&p->wb_list);
	34	}
	35	return p;
	36	}
	37
	38	static inline void
	39	nfs_page_free(struct nfs_page *p)
	40	{
	41	kmem_cache_free(nfs_page_cachep, p);
	42	}
	43
	44	/**
	45	* nfs_create_request - Create an NFS read/write request.
	46	* @file: file descriptor to use
	47	* @inode: inode to which the request is attached
	48	* @page: page to write
	49	* @offset: starting offset within the page for the write
	50	* @count: number of bytes to read/write
	51	*
	52	* The page must be locked by the caller. This makes sure we never
a19b89ca	53	* create two different requests for the same page.
1da177e4 LT	54	* User should ensure it is safe to sleep in this function.
	55	*/
	56	struct nfs_page *
	57	nfs_create_request(struct nfs_open_context ctx, struct inode inode,
	58	struct page *page,
	59	unsigned int offset, unsigned int count)
	60	{
1da177e4 LT	61	struct nfs_page *req;
1da177e4 LT	62
18eb8842 TM	63	/* try to allocate the request struct */
	64	req = nfs_page_alloc();
	65	if (req == NULL)
	66	return ERR_PTR(-ENOMEM);
1da177e4 LT	67
	68	/* Initialize the request struct. Initially, we assume a
	69	* long write-back delay. This will be adjusted in
	70	* update_nfs_request below if the region is not locked. */
	71	req->wb_page = page;
	72	atomic_set(&req->wb_complete, 0);
	73	req->wb_index = page->index;
	74	page_cache_get(page);
cd52ed35 TM	75	BUG_ON(PagePrivate(page));
	76	BUG_ON(!PageLocked(page));
	77	BUG_ON(page->mapping->host != inode);
1da177e4 LT	78	req->wb_offset = offset;
	79	req->wb_pgbase = offset;
	80	req->wb_bytes = count;
1da177e4	81	req->wb_context = get_nfs_open_context(ctx);
c03b4024	82	kref_init(&req->wb_kref);
1da177e4 LT	83	return req;
	84	}
	85
	86	/**
	87	* nfs_unlock_request - Unlock request and wake up sleepers.
	88	* @req:
	89	*/
	90	void nfs_unlock_request(struct nfs_page *req)
	91	{
	92	if (!NFS_WBACK_BUSY(req)) {
	93	printk(KERN_ERR "NFS: Invalid unlock attempted\n");
	94	BUG();
	95	}
	96	smp_mb__before_clear_bit();
	97	clear_bit(PG_BUSY, &req->wb_flags);
	98	smp_mb__after_clear_bit();
464a98bd	99	wake_up_bit(&req->wb_flags, PG_BUSY);
1da177e4 LT	100	nfs_release_request(req);
	101	}
	102
c6a556b8	103	/**
9fd367f0	104	* nfs_set_page_tag_locked - Tag a request as locked
c6a556b8 TM	105	* @req:
c6a556b8 TM	106	*/
acee478a	107	int nfs_set_page_tag_locked(struct nfs_page *req)
c6a556b8	108	{
acee478a	109	if (!nfs_lock_request_dontget(req))
c6a556b8	110	return 0;
acee478a	111	if (req->wb_page != NULL)
bb6fbc45	112	radix_tree_tag_set(&NFS_I(req->wb_context->path.dentry->d_inode)->nfs_page_tree, req->wb_index, NFS_PAGE_TAG_LOCKED);
c6a556b8 TM	113	return 1;
	114	}
	115
	116	/**
9fd367f0	117	* nfs_clear_page_tag_locked - Clear request tag and wake up sleepers
c6a556b8	118	*/
9fd367f0	119	void nfs_clear_page_tag_locked(struct nfs_page *req)
c6a556b8	120	{
deb7d638	121	if (req->wb_page != NULL) {
bb6fbc45 TM	122	struct inode *inode = req->wb_context->path.dentry->d_inode;
	123	struct nfs_inode *nfsi = NFS_I(inode);
	124
587142f8	125	spin_lock(&inode->i_lock);
9fd367f0	126	radix_tree_tag_clear(&nfsi->nfs_page_tree, req->wb_index, NFS_PAGE_TAG_LOCKED);
acee478a	127	nfs_unlock_request(req);
587142f8	128	spin_unlock(&inode->i_lock);
acee478a TM	129	} else
acee478a TM	130	nfs_unlock_request(req);
c6a556b8 TM	131	}
c6a556b8 TM	132
1da177e4 LT	133	/**
	134	* nfs_clear_request - Free up all resources allocated to the request
	135	* @req:
	136	*
bb6fbc45 TM	137	* Release page and open context resources associated with a read/write
bb6fbc45 TM	138	* request after it has completed.
1da177e4 LT	139	*/
	140	void nfs_clear_request(struct nfs_page *req)
	141	{
cd52ed35	142	struct page *page = req->wb_page;
bb6fbc45 TM	143	struct nfs_open_context *ctx = req->wb_context;
bb6fbc45 TM	144
cd52ed35	145	if (page != NULL) {
cd52ed35	146	page_cache_release(page);
1da177e4 LT	147	req->wb_page = NULL;
1da177e4 LT	148	}
bb6fbc45 TM	149	if (ctx != NULL) {
	150	put_nfs_open_context(ctx);
	151	req->wb_context = NULL;
	152	}
1da177e4 LT	153	}
	154
	155
	156	/**
	157	* nfs_release_request - Release the count on an NFS read/write request
	158	* @req: request to release
	159	*
	160	* Note: Should never be called with the spinlock held!
	161	*/
c03b4024	162	static void nfs_free_request(struct kref *kref)
1da177e4	163	{
c03b4024	164	struct nfs_page *req = container_of(kref, struct nfs_page, wb_kref);
1da177e4	165
bb6fbc45	166	/* Release struct file and open context */
1da177e4	167	nfs_clear_request(req);
1da177e4 LT	168	nfs_page_free(req);
	169	}
	170
c03b4024 TM	171	void nfs_release_request(struct nfs_page *req)
	172	{
	173	kref_put(&req->wb_kref, nfs_free_request);
	174	}
	175
9f557cd8 TM	176	static int nfs_wait_bit_uninterruptible(void *word)
	177	{
	178	io_schedule();
	179	return 0;
	180	}
	181
1da177e4 LT	182	/**
	183	* nfs_wait_on_request - Wait for a request to complete.
	184	* @req: request to wait upon.
	185	*
150030b7	186	* Interruptible by fatal signals only.
1da177e4 LT	187	* The user is responsible for holding a count on the request.
	188	*/
	189	int
	190	nfs_wait_on_request(struct nfs_page *req)
	191	{
9f557cd8 TM	192	return wait_on_bit(&req->wb_flags, PG_BUSY,
	193	nfs_wait_bit_uninterruptible,
	194	TASK_UNINTERRUPTIBLE);
1da177e4 LT	195	}
	196
	197	/**
d8a5ad75 TM	198	* nfs_pageio_init - initialise a page io descriptor
d8a5ad75 TM	199	* @desc: pointer to descriptor
bcb71bba TM	200	* @inode: pointer to inode
	201	* @doio: pointer to io function
	202	* @bsize: io block size
	203	* @io_flags: extra parameters for the io function
d8a5ad75	204	*/
bcb71bba TM	205	void nfs_pageio_init(struct nfs_pageio_descriptor *desc,
bcb71bba TM	206	struct inode *inode,
8d5658c9	207	int (doio)(struct inode , struct list_head *, unsigned int, size_t, int),
84dde76c	208	size_t bsize,
bcb71bba	209	int io_flags)
d8a5ad75 TM	210	{
d8a5ad75 TM	211	INIT_LIST_HEAD(&desc->pg_list);
bcb71bba	212	desc->pg_bytes_written = 0;
d8a5ad75 TM	213	desc->pg_count = 0;
	214	desc->pg_bsize = bsize;
	215	desc->pg_base = 0;
bcb71bba TM	216	desc->pg_inode = inode;
	217	desc->pg_doio = doio;
	218	desc->pg_ioflags = io_flags;
	219	desc->pg_error = 0;
d8a5ad75 TM	220	}
	221
	222	/**
	223	* nfs_can_coalesce_requests - test two requests for compatibility
	224	* @prev: pointer to nfs_page
	225	* @req: pointer to nfs_page
	226	*
	227	* The nfs_page structures 'prev' and 'req' are compared to ensure that the
	228	* page data area they describe is contiguous, and that their RPC
	229	* credentials, NFSv4 open state, and lockowners are the same.
	230	*
	231	* Return 'true' if this is the case, else return 'false'.
	232	*/
	233	static int nfs_can_coalesce_requests(struct nfs_page *prev,
	234	struct nfs_page *req)
	235	{
	236	if (req->wb_context->cred != prev->wb_context->cred)
	237	return 0;
	238	if (req->wb_context->lockowner != prev->wb_context->lockowner)
	239	return 0;
	240	if (req->wb_context->state != prev->wb_context->state)
	241	return 0;
	242	if (req->wb_index != (prev->wb_index + 1))
	243	return 0;
	244	if (req->wb_pgbase != 0)
	245	return 0;
	246	if (prev->wb_pgbase + prev->wb_bytes != PAGE_CACHE_SIZE)
	247	return 0;
	248	return 1;
	249	}
	250
	251	/**
bcb71bba	252	* nfs_pageio_do_add_request - Attempt to coalesce a request into a page list.
d8a5ad75 TM	253	* @desc: destination io descriptor
	254	* @req: request
	255	*
	256	* Returns true if the request 'req' was successfully coalesced into the
	257	* existing list of pages 'desc'.
	258	*/
bcb71bba TM	259	static int nfs_pageio_do_add_request(struct nfs_pageio_descriptor *desc,
bcb71bba TM	260	struct nfs_page *req)
d8a5ad75 TM	261	{
	262	size_t newlen = req->wb_bytes;
	263
	264	if (desc->pg_count != 0) {
	265	struct nfs_page *prev;
	266
	267	/*
	268	* FIXME: ideally we should be able to coalesce all requests
	269	* that are not block boundary aligned, but currently this
	270	* is problematic for the case of bsize < PAGE_CACHE_SIZE,
	271	* since nfs_flush_multi and nfs_pagein_multi assume you
	272	* can have only one struct nfs_page.
	273	*/
8d5658c9 TM	274	if (desc->pg_bsize < PAGE_SIZE)
8d5658c9 TM	275	return 0;
d8a5ad75	276	newlen += desc->pg_count;
8d5658c9	277	if (newlen > desc->pg_bsize)
d8a5ad75 TM	278	return 0;
	279	prev = nfs_list_entry(desc->pg_list.prev);
	280	if (!nfs_can_coalesce_requests(prev, req))
	281	return 0;
	282	} else
	283	desc->pg_base = req->wb_pgbase;
	284	nfs_list_remove_request(req);
	285	nfs_list_add_request(req, &desc->pg_list);
	286	desc->pg_count = newlen;
	287	return 1;
	288	}
	289
bcb71bba TM	290	/*
	291	* Helper for nfs_pageio_add_request and nfs_pageio_complete
	292	*/
	293	static void nfs_pageio_doio(struct nfs_pageio_descriptor *desc)
	294	{
	295	if (!list_empty(&desc->pg_list)) {
	296	int error = desc->pg_doio(desc->pg_inode,
	297	&desc->pg_list,
8d5658c9 TM	298	nfs_page_array_len(desc->pg_base,
8d5658c9 TM	299	desc->pg_count),
bcb71bba TM	300	desc->pg_count,
	301	desc->pg_ioflags);
	302	if (error < 0)
	303	desc->pg_error = error;
	304	else
	305	desc->pg_bytes_written += desc->pg_count;
	306	}
	307	if (list_empty(&desc->pg_list)) {
	308	desc->pg_count = 0;
	309	desc->pg_base = 0;
	310	}
	311	}
	312
	313	/**
	314	* nfs_pageio_add_request - Attempt to coalesce a request into a page list.
	315	* @desc: destination io descriptor
	316	* @req: request
	317	*
	318	* Returns true if the request 'req' was successfully coalesced into the
	319	* existing list of pages 'desc'.
	320	*/
8b09bee3 TM	321	int nfs_pageio_add_request(struct nfs_pageio_descriptor *desc,
8b09bee3 TM	322	struct nfs_page *req)
bcb71bba TM	323	{
	324	while (!nfs_pageio_do_add_request(desc, req)) {
	325	nfs_pageio_doio(desc);
	326	if (desc->pg_error < 0)
	327	return 0;
	328	}
	329	return 1;
	330	}
	331
bcb71bba TM	332	/**
	333	* nfs_pageio_complete - Complete I/O on an nfs_pageio_descriptor
	334	* @desc: pointer to io descriptor
	335	*/
	336	void nfs_pageio_complete(struct nfs_pageio_descriptor *desc)
	337	{
	338	nfs_pageio_doio(desc);
	339	}
	340
7fe7f848 TM	341	/**
	342	* nfs_pageio_cond_complete - Conditional I/O completion
	343	* @desc: pointer to io descriptor
	344	* @index: page index
	345	*
	346	* It is important to ensure that processes don't try to take locks
	347	* on non-contiguous ranges of pages as that might deadlock. This
	348	* function should be called before attempting to wait on a locked
	349	* nfs_page. It will complete the I/O if the page index 'index'
	350	* is not contiguous with the existing list of pages in 'desc'.
	351	*/
	352	void nfs_pageio_cond_complete(struct nfs_pageio_descriptor *desc, pgoff_t index)
	353	{
	354	if (!list_empty(&desc->pg_list)) {
	355	struct nfs_page *prev = nfs_list_entry(desc->pg_list.prev);
	356	if (index != prev->wb_index + 1)
	357	nfs_pageio_doio(desc);
	358	}
	359	}
	360
3da28eb1	361	#define NFS_SCAN_MAXENTRIES 16
1da177e4 LT	362	/**
1da177e4 LT	363	* nfs_scan_list - Scan a list for matching requests
d2ccddf0	364	* @nfsi: NFS inode
1da177e4 LT	365	* @dst: Destination list
	366	* @idx_start: lower bound of page->index to scan
	367	* @npages: idx_start + npages sets the upper bound to scan.
5c369683	368	* @tag: tag to scan for
1da177e4 LT	369	*
	370	* Moves elements from one of the inode request lists.
	371	* If the number of requests is set to 0, the entire address_space
	372	* starting at index idx_start, is scanned.
	373	* The requests are not checked to ensure that they form a contiguous set.
587142f8	374	* You must be holding the inode's i_lock when calling this function
1da177e4	375	*/
5c369683	376	int nfs_scan_list(struct nfs_inode *nfsi,
ca52fec1	377	struct list_head *dst, pgoff_t idx_start,
5c369683	378	unsigned int npages, int tag)
1da177e4	379	{
d2ccddf0 TM	380	struct nfs_page *pgvec[NFS_SCAN_MAXENTRIES];
d2ccddf0 TM	381	struct nfs_page *req;
ca52fec1	382	pgoff_t idx_end;
d2ccddf0 TM	383	int found, i;
d2ccddf0 TM	384	int res;
1da177e4 LT	385
	386	res = 0;
	387	if (npages == 0)
	388	idx_end = ~0;
	389	else
	390	idx_end = idx_start + npages - 1;
	391
d2ccddf0	392	for (;;) {
5c369683	393	found = radix_tree_gang_lookup_tag(&nfsi->nfs_page_tree,
d2ccddf0	394	(void **)&pgvec[0], idx_start,
5c369683	395	NFS_SCAN_MAXENTRIES, tag);
d2ccddf0	396	if (found <= 0)
1da177e4	397	break;
d2ccddf0 TM	398	for (i = 0; i < found; i++) {
	399	req = pgvec[i];
	400	if (req->wb_index > idx_end)
	401	goto out;
	402	idx_start = req->wb_index + 1;
9fd367f0	403	if (nfs_set_page_tag_locked(req)) {
acee478a	404	kref_get(&req->wb_kref);
d2ccddf0	405	nfs_list_remove_request(req);
5c369683 TM	406	radix_tree_tag_clear(&nfsi->nfs_page_tree,
5c369683 TM	407	req->wb_index, tag);
d2ccddf0 TM	408	nfs_list_add_request(req, dst);
d2ccddf0 TM	409	res++;
dce34ce2 TM	410	if (res == INT_MAX)
dce34ce2 TM	411	goto out;
d2ccddf0 TM	412	}
d2ccddf0 TM	413	}
edc05fc1	414	/* for latency reduction */
587142f8	415	cond_resched_lock(&nfsi->vfs_inode.i_lock);
1da177e4	416	}
d2ccddf0	417	out:
1da177e4 LT	418	return res;
	419	}
	420
f7b422b1	421	int __init nfs_init_nfspagecache(void)
1da177e4 LT	422	{
	423	nfs_page_cachep = kmem_cache_create("nfs_page",
	424	sizeof(struct nfs_page),
	425	0, SLAB_HWCACHE_ALIGN,
20c2df83	426	NULL);
1da177e4 LT	427	if (nfs_page_cachep == NULL)
	428	return -ENOMEM;
	429
	430	return 0;
	431	}
	432
266bee88	433	void nfs_destroy_nfspagecache(void)
1da177e4	434	{
1a1d92c1	435	kmem_cache_destroy(nfs_page_cachep);
1da177e4 LT	436	}
1da177e4 LT	437