[linux-2.6-block.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;

	for (;;) {
		/* try to allocate the request struct */
		req = nfs_page_alloc();
		if (req != NULL)
			break;

		if (fatal_signal_pending(current))
			return ERR_PTR(-ERESTARTSYS);
		yield();
	}

	/* 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)
{
	struct nfs_inode *nfsi = NFS_I(req->wb_context->path.dentry->d_inode);

	if (!nfs_lock_request_dontget(req))
		return 0;
	if (req->wb_page != NULL)
		radix_tree_tag_set(&nfsi->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)
{
	struct inode *inode = req->wb_context->path.dentry->d_inode;
	struct nfs_inode *nfsi = NFS_I(inode);

	if (req->wb_page != NULL) {
		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 resources associated with a write request after it
 * has completed.
 */
void nfs_clear_request(struct nfs_page *req)
{
	struct page *page = req->wb_page;
	if (page != NULL) {
		page_cache_release(page);
		req->wb_page = 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 or cached credential */
	nfs_clear_request(req);
	put_nfs_open_context(req->wb_context);
	nfs_page_free(req);
}

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

/**
 * 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)
{
	int ret = 0;

	if (!test_bit(PG_BUSY, &req->wb_flags))
		goto out;
	ret = out_of_line_wait_on_bit(&req->wb_flags, PG_BUSY,
			nfs_wait_bit_killable, TASK_KILLABLE);
out:
	return ret;
}

/**
 * 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
1da177e4 LT	63	for (;;) {
	64	/* try to allocate the request struct */
	65	req = nfs_page_alloc();
	66	if (req != NULL)
	67	break;
	68
150030b7	69	if (fatal_signal_pending(current))
1da177e4 LT	70	return ERR_PTR(-ERESTARTSYS);
	71	yield();
	72	}
	73
	74	/* Initialize the request struct. Initially, we assume a
	75	* long write-back delay. This will be adjusted in
	76	* update_nfs_request below if the region is not locked. */
	77	req->wb_page = page;
	78	atomic_set(&req->wb_complete, 0);
	79	req->wb_index = page->index;
	80	page_cache_get(page);
cd52ed35 TM	81	BUG_ON(PagePrivate(page));
	82	BUG_ON(!PageLocked(page));
	83	BUG_ON(page->mapping->host != inode);
1da177e4 LT	84	req->wb_offset = offset;
	85	req->wb_pgbase = offset;
	86	req->wb_bytes = count;
1da177e4	87	req->wb_context = get_nfs_open_context(ctx);
c03b4024	88	kref_init(&req->wb_kref);
1da177e4 LT	89	return req;
	90	}
	91
	92	/**
	93	* nfs_unlock_request - Unlock request and wake up sleepers.
	94	* @req:
	95	*/
	96	void nfs_unlock_request(struct nfs_page *req)
	97	{
	98	if (!NFS_WBACK_BUSY(req)) {
	99	printk(KERN_ERR "NFS: Invalid unlock attempted\n");
	100	BUG();
	101	}
	102	smp_mb__before_clear_bit();
	103	clear_bit(PG_BUSY, &req->wb_flags);
	104	smp_mb__after_clear_bit();
464a98bd	105	wake_up_bit(&req->wb_flags, PG_BUSY);
1da177e4 LT	106	nfs_release_request(req);
	107	}
	108
c6a556b8	109	/**
9fd367f0	110	* nfs_set_page_tag_locked - Tag a request as locked
c6a556b8 TM	111	* @req:
c6a556b8 TM	112	*/
acee478a	113	int nfs_set_page_tag_locked(struct nfs_page *req)
c6a556b8	114	{
88be9f99	115	struct nfs_inode *nfsi = NFS_I(req->wb_context->path.dentry->d_inode);
c6a556b8	116
acee478a	117	if (!nfs_lock_request_dontget(req))
c6a556b8	118	return 0;
acee478a TM	119	if (req->wb_page != NULL)
acee478a TM	120	radix_tree_tag_set(&nfsi->nfs_page_tree, req->wb_index, NFS_PAGE_TAG_LOCKED);
c6a556b8 TM	121	return 1;
	122	}
	123
	124	/**
9fd367f0	125	* nfs_clear_page_tag_locked - Clear request tag and wake up sleepers
c6a556b8	126	*/
9fd367f0	127	void nfs_clear_page_tag_locked(struct nfs_page *req)
c6a556b8	128	{
587142f8 TM	129	struct inode *inode = req->wb_context->path.dentry->d_inode;
587142f8 TM	130	struct nfs_inode *nfsi = NFS_I(inode);
c6a556b8	131
deb7d638	132	if (req->wb_page != NULL) {
587142f8	133	spin_lock(&inode->i_lock);
9fd367f0	134	radix_tree_tag_clear(&nfsi->nfs_page_tree, req->wb_index, NFS_PAGE_TAG_LOCKED);
acee478a	135	nfs_unlock_request(req);
587142f8	136	spin_unlock(&inode->i_lock);
acee478a TM	137	} else
acee478a TM	138	nfs_unlock_request(req);
c6a556b8 TM	139	}
c6a556b8 TM	140
1da177e4 LT	141	/**
	142	* nfs_clear_request - Free up all resources allocated to the request
	143	* @req:
	144	*
	145	* Release page resources associated with a write request after it
	146	* has completed.
	147	*/
	148	void nfs_clear_request(struct nfs_page *req)
	149	{
cd52ed35 TM	150	struct page *page = req->wb_page;
cd52ed35 TM	151	if (page != NULL) {
cd52ed35	152	page_cache_release(page);
1da177e4 LT	153	req->wb_page = NULL;
	154	}
	155	}
	156
	157
	158	/**
	159	* nfs_release_request - Release the count on an NFS read/write request
	160	* @req: request to release
	161	*
	162	* Note: Should never be called with the spinlock held!
	163	*/
c03b4024	164	static void nfs_free_request(struct kref *kref)
1da177e4	165	{
c03b4024	166	struct nfs_page *req = container_of(kref, struct nfs_page, wb_kref);
1da177e4	167
1da177e4 LT	168	/* Release struct file or cached credential */
	169	nfs_clear_request(req);
	170	put_nfs_open_context(req->wb_context);
	171	nfs_page_free(req);
	172	}
	173
c03b4024 TM	174	void nfs_release_request(struct nfs_page *req)
	175	{
	176	kref_put(&req->wb_kref, nfs_free_request);
	177	}
	178
1da177e4 LT	179	/**
	180	* nfs_wait_on_request - Wait for a request to complete.
	181	* @req: request to wait upon.
	182	*
150030b7	183	* Interruptible by fatal signals only.
1da177e4 LT	184	* The user is responsible for holding a count on the request.
	185	*/
	186	int
	187	nfs_wait_on_request(struct nfs_page *req)
	188	{
464a98bd TM	189	int ret = 0;
	190
	191	if (!test_bit(PG_BUSY, &req->wb_flags))
	192	goto out;
464a98bd	193	ret = out_of_line_wait_on_bit(&req->wb_flags, PG_BUSY,
150030b7	194	nfs_wait_bit_killable, TASK_KILLABLE);
464a98bd TM	195	out:
464a98bd TM	196	return ret;
1da177e4 LT	197	}
	198
	199	/**
d8a5ad75 TM	200	* nfs_pageio_init - initialise a page io descriptor
d8a5ad75 TM	201	* @desc: pointer to descriptor
bcb71bba TM	202	* @inode: pointer to inode
	203	* @doio: pointer to io function
	204	* @bsize: io block size
	205	* @io_flags: extra parameters for the io function
d8a5ad75	206	*/
bcb71bba TM	207	void nfs_pageio_init(struct nfs_pageio_descriptor *desc,
bcb71bba TM	208	struct inode *inode,
8d5658c9	209	int (doio)(struct inode , struct list_head *, unsigned int, size_t, int),
84dde76c	210	size_t bsize,
bcb71bba	211	int io_flags)
d8a5ad75 TM	212	{
d8a5ad75 TM	213	INIT_LIST_HEAD(&desc->pg_list);
bcb71bba	214	desc->pg_bytes_written = 0;
d8a5ad75 TM	215	desc->pg_count = 0;
	216	desc->pg_bsize = bsize;
	217	desc->pg_base = 0;
bcb71bba TM	218	desc->pg_inode = inode;
	219	desc->pg_doio = doio;
	220	desc->pg_ioflags = io_flags;
	221	desc->pg_error = 0;
d8a5ad75 TM	222	}
	223
	224	/**
	225	* nfs_can_coalesce_requests - test two requests for compatibility
	226	* @prev: pointer to nfs_page
	227	* @req: pointer to nfs_page
	228	*
	229	* The nfs_page structures 'prev' and 'req' are compared to ensure that the
	230	* page data area they describe is contiguous, and that their RPC
	231	* credentials, NFSv4 open state, and lockowners are the same.
	232	*
	233	* Return 'true' if this is the case, else return 'false'.
	234	*/
	235	static int nfs_can_coalesce_requests(struct nfs_page *prev,
	236	struct nfs_page *req)
	237	{
	238	if (req->wb_context->cred != prev->wb_context->cred)
	239	return 0;
	240	if (req->wb_context->lockowner != prev->wb_context->lockowner)
	241	return 0;
	242	if (req->wb_context->state != prev->wb_context->state)
	243	return 0;
	244	if (req->wb_index != (prev->wb_index + 1))
	245	return 0;
	246	if (req->wb_pgbase != 0)
	247	return 0;
	248	if (prev->wb_pgbase + prev->wb_bytes != PAGE_CACHE_SIZE)
	249	return 0;
	250	return 1;
	251	}
	252
	253	/**
bcb71bba	254	* nfs_pageio_do_add_request - Attempt to coalesce a request into a page list.
d8a5ad75 TM	255	* @desc: destination io descriptor
	256	* @req: request
	257	*
	258	* Returns true if the request 'req' was successfully coalesced into the
	259	* existing list of pages 'desc'.
	260	*/
bcb71bba TM	261	static int nfs_pageio_do_add_request(struct nfs_pageio_descriptor *desc,
bcb71bba TM	262	struct nfs_page *req)
d8a5ad75 TM	263	{
	264	size_t newlen = req->wb_bytes;
	265
	266	if (desc->pg_count != 0) {
	267	struct nfs_page *prev;
	268
	269	/*
	270	* FIXME: ideally we should be able to coalesce all requests
	271	* that are not block boundary aligned, but currently this
	272	* is problematic for the case of bsize < PAGE_CACHE_SIZE,
	273	* since nfs_flush_multi and nfs_pagein_multi assume you
	274	* can have only one struct nfs_page.
	275	*/
8d5658c9 TM	276	if (desc->pg_bsize < PAGE_SIZE)
8d5658c9 TM	277	return 0;
d8a5ad75	278	newlen += desc->pg_count;
8d5658c9	279	if (newlen > desc->pg_bsize)
d8a5ad75 TM	280	return 0;
	281	prev = nfs_list_entry(desc->pg_list.prev);
	282	if (!nfs_can_coalesce_requests(prev, req))
	283	return 0;
	284	} else
	285	desc->pg_base = req->wb_pgbase;
	286	nfs_list_remove_request(req);
	287	nfs_list_add_request(req, &desc->pg_list);
	288	desc->pg_count = newlen;
	289	return 1;
	290	}
	291
bcb71bba TM	292	/*
	293	* Helper for nfs_pageio_add_request and nfs_pageio_complete
	294	*/
	295	static void nfs_pageio_doio(struct nfs_pageio_descriptor *desc)
	296	{
	297	if (!list_empty(&desc->pg_list)) {
	298	int error = desc->pg_doio(desc->pg_inode,
	299	&desc->pg_list,
8d5658c9 TM	300	nfs_page_array_len(desc->pg_base,
8d5658c9 TM	301	desc->pg_count),
bcb71bba TM	302	desc->pg_count,
	303	desc->pg_ioflags);
	304	if (error < 0)
	305	desc->pg_error = error;
	306	else
	307	desc->pg_bytes_written += desc->pg_count;
	308	}
	309	if (list_empty(&desc->pg_list)) {
	310	desc->pg_count = 0;
	311	desc->pg_base = 0;
	312	}
	313	}
	314
	315	/**
	316	* nfs_pageio_add_request - Attempt to coalesce a request into a page list.
	317	* @desc: destination io descriptor
	318	* @req: request
	319	*
	320	* Returns true if the request 'req' was successfully coalesced into the
	321	* existing list of pages 'desc'.
	322	*/
8b09bee3 TM	323	int nfs_pageio_add_request(struct nfs_pageio_descriptor *desc,
8b09bee3 TM	324	struct nfs_page *req)
bcb71bba TM	325	{
	326	while (!nfs_pageio_do_add_request(desc, req)) {
	327	nfs_pageio_doio(desc);
	328	if (desc->pg_error < 0)
	329	return 0;
	330	}
	331	return 1;
	332	}
	333
bcb71bba TM	334	/**
	335	* nfs_pageio_complete - Complete I/O on an nfs_pageio_descriptor
	336	* @desc: pointer to io descriptor
	337	*/
	338	void nfs_pageio_complete(struct nfs_pageio_descriptor *desc)
	339	{
	340	nfs_pageio_doio(desc);
	341	}
	342
7fe7f848 TM	343	/**
	344	* nfs_pageio_cond_complete - Conditional I/O completion
	345	* @desc: pointer to io descriptor
	346	* @index: page index
	347	*
	348	* It is important to ensure that processes don't try to take locks
	349	* on non-contiguous ranges of pages as that might deadlock. This
	350	* function should be called before attempting to wait on a locked
	351	* nfs_page. It will complete the I/O if the page index 'index'
	352	* is not contiguous with the existing list of pages in 'desc'.
	353	*/
	354	void nfs_pageio_cond_complete(struct nfs_pageio_descriptor *desc, pgoff_t index)
	355	{
	356	if (!list_empty(&desc->pg_list)) {
	357	struct nfs_page *prev = nfs_list_entry(desc->pg_list.prev);
	358	if (index != prev->wb_index + 1)
	359	nfs_pageio_doio(desc);
	360	}
	361	}
	362
3da28eb1	363	#define NFS_SCAN_MAXENTRIES 16
1da177e4 LT	364	/**
1da177e4 LT	365	* nfs_scan_list - Scan a list for matching requests
d2ccddf0	366	* @nfsi: NFS inode
1da177e4 LT	367	* @dst: Destination list
	368	* @idx_start: lower bound of page->index to scan
	369	* @npages: idx_start + npages sets the upper bound to scan.
5c369683	370	* @tag: tag to scan for
1da177e4 LT	371	*
	372	* Moves elements from one of the inode request lists.
	373	* If the number of requests is set to 0, the entire address_space
	374	* starting at index idx_start, is scanned.
	375	* The requests are not checked to ensure that they form a contiguous set.
587142f8	376	* You must be holding the inode's i_lock when calling this function
1da177e4	377	*/
5c369683	378	int nfs_scan_list(struct nfs_inode *nfsi,
ca52fec1	379	struct list_head *dst, pgoff_t idx_start,
5c369683	380	unsigned int npages, int tag)
1da177e4	381	{
d2ccddf0 TM	382	struct nfs_page *pgvec[NFS_SCAN_MAXENTRIES];
d2ccddf0 TM	383	struct nfs_page *req;
ca52fec1	384	pgoff_t idx_end;
d2ccddf0 TM	385	int found, i;
d2ccddf0 TM	386	int res;
1da177e4 LT	387
	388	res = 0;
	389	if (npages == 0)
	390	idx_end = ~0;
	391	else
	392	idx_end = idx_start + npages - 1;
	393
d2ccddf0	394	for (;;) {
5c369683	395	found = radix_tree_gang_lookup_tag(&nfsi->nfs_page_tree,
d2ccddf0	396	(void **)&pgvec[0], idx_start,
5c369683	397	NFS_SCAN_MAXENTRIES, tag);
d2ccddf0	398	if (found <= 0)
1da177e4	399	break;
d2ccddf0 TM	400	for (i = 0; i < found; i++) {
	401	req = pgvec[i];
	402	if (req->wb_index > idx_end)
	403	goto out;
	404	idx_start = req->wb_index + 1;
9fd367f0	405	if (nfs_set_page_tag_locked(req)) {
acee478a	406	kref_get(&req->wb_kref);
d2ccddf0	407	nfs_list_remove_request(req);
5c369683 TM	408	radix_tree_tag_clear(&nfsi->nfs_page_tree,
5c369683 TM	409	req->wb_index, tag);
d2ccddf0 TM	410	nfs_list_add_request(req, dst);
d2ccddf0 TM	411	res++;
dce34ce2 TM	412	if (res == INT_MAX)
dce34ce2 TM	413	goto out;
d2ccddf0 TM	414	}
d2ccddf0 TM	415	}
edc05fc1	416	/* for latency reduction */
587142f8	417	cond_resched_lock(&nfsi->vfs_inode.i_lock);
1da177e4	418	}
d2ccddf0	419	out:
1da177e4 LT	420	return res;
	421	}
	422
f7b422b1	423	int __init nfs_init_nfspagecache(void)
1da177e4 LT	424	{
	425	nfs_page_cachep = kmem_cache_create("nfs_page",
	426	sizeof(struct nfs_page),
	427	0, SLAB_HWCACHE_ALIGN,
20c2df83	428	NULL);
1da177e4 LT	429	if (nfs_page_cachep == NULL)
	430	return -ENOMEM;
	431
	432	return 0;
	433	}
	434
266bee88	435	void nfs_destroy_nfspagecache(void)
1da177e4	436	{
1a1d92c1	437	kmem_cache_destroy(nfs_page_cachep);
1da177e4 LT	438	}
1da177e4 LT	439