[linux.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 <[email protected]>
 *
 */

#include <linux/slab.h>
#include <linux/file.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_server *server = NFS_SERVER(inode);
	struct nfs_page		*req;

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

		if (signalled() && (server->flags & NFS_MOUNT_INTR))
			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;
	atomic_set(&req->wb_count, 1);
	req->wb_context = get_nfs_open_context(ctx);

	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_writeback_locked - Lock a request for writeback
 * @req:
 */
int nfs_set_page_writeback_locked(struct nfs_page *req)
{
	struct nfs_inode *nfsi = NFS_I(req->wb_context->dentry->d_inode);

	if (!nfs_lock_request(req))
		return 0;
	radix_tree_tag_set(&nfsi->nfs_page_tree, req->wb_index, NFS_PAGE_TAG_WRITEBACK);
	return 1;
}

/**
 * nfs_clear_page_writeback - Unlock request and wake up sleepers
 */
void nfs_clear_page_writeback(struct nfs_page *req)
{
	struct nfs_inode *nfsi = NFS_I(req->wb_context->dentry->d_inode);

	if (req->wb_page != NULL) {
		spin_lock(&nfsi->req_lock);
		radix_tree_tag_clear(&nfsi->nfs_page_tree, req->wb_index, NFS_PAGE_TAG_WRITEBACK);
		spin_unlock(&nfsi->req_lock);
	}
	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!
 */
void
nfs_release_request(struct nfs_page *req)
{
	if (!atomic_dec_and_test(&req->wb_count))
		return;

	/* Release struct file or cached credential */
	nfs_clear_request(req);
	put_nfs_open_context(req->wb_context);
	nfs_page_free(req);
}

static int nfs_wait_bit_interruptible(void *word)
{
	int ret = 0;

	if (signal_pending(current))
		ret = -ERESTARTSYS;
	else
		schedule();
	return ret;
}

/**
 * nfs_wait_on_request - Wait for a request to complete.
 * @req: request to wait upon.
 *
 * Interruptible by signals only if mounted with intr flag.
 * The user is responsible for holding a count on the request.
 */
int
nfs_wait_on_request(struct nfs_page *req)
{
        struct rpc_clnt	*clnt = NFS_CLIENT(req->wb_context->dentry->d_inode);
	sigset_t oldmask;
	int ret = 0;

	if (!test_bit(PG_BUSY, &req->wb_flags))
		goto out;
	/*
	 * Note: the call to rpc_clnt_sigmask() suffices to ensure that we
	 *	 are not interrupted if intr flag is not set
	 */
	rpc_clnt_sigmask(clnt, &oldmask);
	ret = out_of_line_wait_on_bit(&req->wb_flags, PG_BUSY,
			nfs_wait_bit_interruptible, TASK_INTERRUPTIBLE);
	rpc_clnt_sigunmask(clnt, &oldmask);
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);
}

#define NFS_SCAN_MAXENTRIES 16
/**
 * nfs_scan_list - Scan a list for matching requests
 * @nfsi: NFS inode
 * @head: One of the NFS inode request lists
 * @dst: Destination list
 * @idx_start: lower bound of page->index to scan
 * @npages: idx_start + npages sets the upper bound to scan.
 *
 * 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 req_lock when calling this function
 */
int nfs_scan_list(struct nfs_inode *nfsi, struct list_head *head,
		struct list_head *dst, pgoff_t idx_start,
		unsigned int npages)
{
	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(&nfsi->nfs_page_tree,
				(void **)&pgvec[0], idx_start,
				NFS_SCAN_MAXENTRIES);
		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 (req->wb_list_head != head)
				continue;
			if (nfs_set_page_writeback_locked(req)) {
				nfs_list_remove_request(req);
				nfs_list_add_request(req, dst);
				res++;
			}
		}

	}
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, 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 <[email protected]>
	9	*
	10	*/
	11
1da177e4 LT	12	#include <linux/slab.h>
	13	#include <linux/file.h>
	14	#include <linux/sunrpc/clnt.h>
	15	#include <linux/nfs3.h>
	16	#include <linux/nfs4.h>
	17	#include <linux/nfs_page.h>
	18	#include <linux/nfs_fs.h>
	19	#include <linux/nfs_mount.h>
	20
8d5658c9 TM	21	#include "internal.h"
8d5658c9 TM	22
e18b890b	23	static struct kmem_cache *nfs_page_cachep;
1da177e4 LT	24
	25	static inline struct nfs_page *
	26	nfs_page_alloc(void)
	27	{
	28	struct nfs_page *p;
e94b1766	29	p = kmem_cache_alloc(nfs_page_cachep, GFP_KERNEL);
1da177e4 LT	30	if (p) {
	31	memset(p, 0, sizeof(*p));
	32	INIT_LIST_HEAD(&p->wb_list);
	33	}
	34	return p;
	35	}
	36
	37	static inline void
	38	nfs_page_free(struct nfs_page *p)
	39	{
	40	kmem_cache_free(nfs_page_cachep, p);
	41	}
	42
	43	/**
	44	* nfs_create_request - Create an NFS read/write request.
	45	* @file: file descriptor to use
	46	* @inode: inode to which the request is attached
	47	* @page: page to write
	48	* @offset: starting offset within the page for the write
	49	* @count: number of bytes to read/write
	50	*
	51	* The page must be locked by the caller. This makes sure we never
a19b89ca	52	* create two different requests for the same page.
1da177e4 LT	53	* User should ensure it is safe to sleep in this function.
	54	*/
	55	struct nfs_page *
	56	nfs_create_request(struct nfs_open_context ctx, struct inode inode,
	57	struct page *page,
	58	unsigned int offset, unsigned int count)
	59	{
	60	struct nfs_server *server = NFS_SERVER(inode);
	61	struct nfs_page *req;
	62
1da177e4 LT	63	for (;;) {
	64	/* try to allocate the request struct */
	65	req = nfs_page_alloc();
	66	if (req != NULL)
	67	break;
	68
1da177e4 LT	69	if (signalled() && (server->flags & NFS_MOUNT_INTR))
	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;
	87	atomic_set(&req->wb_count, 1);
	88	req->wb_context = get_nfs_open_context(ctx);
	89
	90	return req;
	91	}
	92
	93	/**
	94	* nfs_unlock_request - Unlock request and wake up sleepers.
	95	* @req:
	96	*/
	97	void nfs_unlock_request(struct nfs_page *req)
	98	{
	99	if (!NFS_WBACK_BUSY(req)) {
	100	printk(KERN_ERR "NFS: Invalid unlock attempted\n");
	101	BUG();
	102	}
	103	smp_mb__before_clear_bit();
	104	clear_bit(PG_BUSY, &req->wb_flags);
	105	smp_mb__after_clear_bit();
464a98bd	106	wake_up_bit(&req->wb_flags, PG_BUSY);
1da177e4 LT	107	nfs_release_request(req);
	108	}
	109
c6a556b8 TM	110	/**
	111	* nfs_set_page_writeback_locked - Lock a request for writeback
	112	* @req:
	113	*/
	114	int nfs_set_page_writeback_locked(struct nfs_page *req)
	115	{
	116	struct nfs_inode *nfsi = NFS_I(req->wb_context->dentry->d_inode);
	117
	118	if (!nfs_lock_request(req))
	119	return 0;
	120	radix_tree_tag_set(&nfsi->nfs_page_tree, req->wb_index, NFS_PAGE_TAG_WRITEBACK);
	121	return 1;
	122	}
	123
	124	/**
	125	* nfs_clear_page_writeback - Unlock request and wake up sleepers
	126	*/
	127	void nfs_clear_page_writeback(struct nfs_page *req)
	128	{
	129	struct nfs_inode *nfsi = NFS_I(req->wb_context->dentry->d_inode);
	130
deb7d638 TM	131	if (req->wb_page != NULL) {
	132	spin_lock(&nfsi->req_lock);
	133	radix_tree_tag_clear(&nfsi->nfs_page_tree, req->wb_index, NFS_PAGE_TAG_WRITEBACK);
	134	spin_unlock(&nfsi->req_lock);
	135	}
c6a556b8 TM	136	nfs_unlock_request(req);
	137	}
	138
1da177e4 LT	139	/**
	140	* nfs_clear_request - Free up all resources allocated to the request
	141	* @req:
	142	*
	143	* Release page resources associated with a write request after it
	144	* has completed.
	145	*/
	146	void nfs_clear_request(struct nfs_page *req)
	147	{
cd52ed35 TM	148	struct page *page = req->wb_page;
cd52ed35 TM	149	if (page != NULL) {
cd52ed35	150	page_cache_release(page);
1da177e4 LT	151	req->wb_page = NULL;
	152	}
	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	*/
	162	void
	163	nfs_release_request(struct nfs_page *req)
	164	{
	165	if (!atomic_dec_and_test(&req->wb_count))
	166	return;
	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
464a98bd TM	174	static int nfs_wait_bit_interruptible(void *word)
	175	{
	176	int ret = 0;
	177
	178	if (signal_pending(current))
	179	ret = -ERESTARTSYS;
	180	else
	181	schedule();
	182	return ret;
	183	}
	184
1da177e4 LT	185	/**
	186	* nfs_wait_on_request - Wait for a request to complete.
	187	* @req: request to wait upon.
	188	*
	189	* Interruptible by signals only if mounted with intr flag.
	190	* The user is responsible for holding a count on the request.
	191	*/
	192	int
	193	nfs_wait_on_request(struct nfs_page *req)
	194	{
464a98bd TM	195	struct rpc_clnt *clnt = NFS_CLIENT(req->wb_context->dentry->d_inode);
	196	sigset_t oldmask;
	197	int ret = 0;
	198
	199	if (!test_bit(PG_BUSY, &req->wb_flags))
	200	goto out;
	201	/*
	202	* Note: the call to rpc_clnt_sigmask() suffices to ensure that we
	203	* are not interrupted if intr flag is not set
	204	*/
	205	rpc_clnt_sigmask(clnt, &oldmask);
	206	ret = out_of_line_wait_on_bit(&req->wb_flags, PG_BUSY,
	207	nfs_wait_bit_interruptible, TASK_INTERRUPTIBLE);
	208	rpc_clnt_sigunmask(clnt, &oldmask);
	209	out:
	210	return ret;
1da177e4 LT	211	}
	212
	213	/**
d8a5ad75 TM	214	* nfs_pageio_init - initialise a page io descriptor
d8a5ad75 TM	215	* @desc: pointer to descriptor
bcb71bba TM	216	* @inode: pointer to inode
	217	* @doio: pointer to io function
	218	* @bsize: io block size
	219	* @io_flags: extra parameters for the io function
d8a5ad75	220	*/
bcb71bba TM	221	void nfs_pageio_init(struct nfs_pageio_descriptor *desc,
bcb71bba TM	222	struct inode *inode,
8d5658c9	223	int (doio)(struct inode , struct list_head *, unsigned int, size_t, int),
84dde76c	224	size_t bsize,
bcb71bba	225	int io_flags)
d8a5ad75 TM	226	{
d8a5ad75 TM	227	INIT_LIST_HEAD(&desc->pg_list);
bcb71bba	228	desc->pg_bytes_written = 0;
d8a5ad75 TM	229	desc->pg_count = 0;
	230	desc->pg_bsize = bsize;
	231	desc->pg_base = 0;
bcb71bba TM	232	desc->pg_inode = inode;
	233	desc->pg_doio = doio;
	234	desc->pg_ioflags = io_flags;
	235	desc->pg_error = 0;
d8a5ad75 TM	236	}
	237
	238	/**
	239	* nfs_can_coalesce_requests - test two requests for compatibility
	240	* @prev: pointer to nfs_page
	241	* @req: pointer to nfs_page
	242	*
	243	* The nfs_page structures 'prev' and 'req' are compared to ensure that the
	244	* page data area they describe is contiguous, and that their RPC
	245	* credentials, NFSv4 open state, and lockowners are the same.
	246	*
	247	* Return 'true' if this is the case, else return 'false'.
	248	*/
	249	static int nfs_can_coalesce_requests(struct nfs_page *prev,
	250	struct nfs_page *req)
	251	{
	252	if (req->wb_context->cred != prev->wb_context->cred)
	253	return 0;
	254	if (req->wb_context->lockowner != prev->wb_context->lockowner)
	255	return 0;
	256	if (req->wb_context->state != prev->wb_context->state)
	257	return 0;
	258	if (req->wb_index != (prev->wb_index + 1))
	259	return 0;
	260	if (req->wb_pgbase != 0)
	261	return 0;
	262	if (prev->wb_pgbase + prev->wb_bytes != PAGE_CACHE_SIZE)
	263	return 0;
	264	return 1;
	265	}
	266
	267	/**
bcb71bba	268	* nfs_pageio_do_add_request - Attempt to coalesce a request into a page list.
d8a5ad75 TM	269	* @desc: destination io descriptor
	270	* @req: request
	271	*
	272	* Returns true if the request 'req' was successfully coalesced into the
	273	* existing list of pages 'desc'.
	274	*/
bcb71bba TM	275	static int nfs_pageio_do_add_request(struct nfs_pageio_descriptor *desc,
bcb71bba TM	276	struct nfs_page *req)
d8a5ad75 TM	277	{
	278	size_t newlen = req->wb_bytes;
	279
	280	if (desc->pg_count != 0) {
	281	struct nfs_page *prev;
	282
	283	/*
	284	* FIXME: ideally we should be able to coalesce all requests
	285	* that are not block boundary aligned, but currently this
	286	* is problematic for the case of bsize < PAGE_CACHE_SIZE,
	287	* since nfs_flush_multi and nfs_pagein_multi assume you
	288	* can have only one struct nfs_page.
	289	*/
8d5658c9 TM	290	if (desc->pg_bsize < PAGE_SIZE)
8d5658c9 TM	291	return 0;
d8a5ad75	292	newlen += desc->pg_count;
8d5658c9	293	if (newlen > desc->pg_bsize)
d8a5ad75 TM	294	return 0;
	295	prev = nfs_list_entry(desc->pg_list.prev);
	296	if (!nfs_can_coalesce_requests(prev, req))
	297	return 0;
	298	} else
	299	desc->pg_base = req->wb_pgbase;
	300	nfs_list_remove_request(req);
	301	nfs_list_add_request(req, &desc->pg_list);
	302	desc->pg_count = newlen;
	303	return 1;
	304	}
	305
bcb71bba TM	306	/*
	307	* Helper for nfs_pageio_add_request and nfs_pageio_complete
	308	*/
	309	static void nfs_pageio_doio(struct nfs_pageio_descriptor *desc)
	310	{
	311	if (!list_empty(&desc->pg_list)) {
	312	int error = desc->pg_doio(desc->pg_inode,
	313	&desc->pg_list,
8d5658c9 TM	314	nfs_page_array_len(desc->pg_base,
8d5658c9 TM	315	desc->pg_count),
bcb71bba TM	316	desc->pg_count,
	317	desc->pg_ioflags);
	318	if (error < 0)
	319	desc->pg_error = error;
	320	else
	321	desc->pg_bytes_written += desc->pg_count;
	322	}
	323	if (list_empty(&desc->pg_list)) {
	324	desc->pg_count = 0;
	325	desc->pg_base = 0;
	326	}
	327	}
	328
	329	/**
	330	* nfs_pageio_add_request - Attempt to coalesce a request into a page list.
	331	* @desc: destination io descriptor
	332	* @req: request
	333	*
	334	* Returns true if the request 'req' was successfully coalesced into the
	335	* existing list of pages 'desc'.
	336	*/
8b09bee3 TM	337	int nfs_pageio_add_request(struct nfs_pageio_descriptor *desc,
8b09bee3 TM	338	struct nfs_page *req)
bcb71bba TM	339	{
	340	while (!nfs_pageio_do_add_request(desc, req)) {
	341	nfs_pageio_doio(desc);
	342	if (desc->pg_error < 0)
	343	return 0;
	344	}
	345	return 1;
	346	}
	347
bcb71bba TM	348	/**
	349	* nfs_pageio_complete - Complete I/O on an nfs_pageio_descriptor
	350	* @desc: pointer to io descriptor
	351	*/
	352	void nfs_pageio_complete(struct nfs_pageio_descriptor *desc)
	353	{
	354	nfs_pageio_doio(desc);
	355	}
	356
3da28eb1	357	#define NFS_SCAN_MAXENTRIES 16
1da177e4 LT	358	/**
1da177e4 LT	359	* nfs_scan_list - Scan a list for matching requests
d2ccddf0	360	* @nfsi: NFS inode
1da177e4 LT	361	* @head: One of the NFS inode request lists
	362	* @dst: Destination list
	363	* @idx_start: lower bound of page->index to scan
	364	* @npages: idx_start + npages sets the upper bound to scan.
	365	*
	366	* Moves elements from one of the inode request lists.
	367	* If the number of requests is set to 0, the entire address_space
	368	* starting at index idx_start, is scanned.
	369	* The requests are not checked to ensure that they form a contiguous set.
	370	* You must be holding the inode's req_lock when calling this function
	371	*/
d2ccddf0	372	int nfs_scan_list(struct nfs_inode nfsi, struct list_head head,
ca52fec1	373	struct list_head *dst, pgoff_t idx_start,
d2ccddf0	374	unsigned int npages)
1da177e4	375	{
d2ccddf0 TM	376	struct nfs_page *pgvec[NFS_SCAN_MAXENTRIES];
d2ccddf0 TM	377	struct nfs_page *req;
ca52fec1	378	pgoff_t idx_end;
d2ccddf0 TM	379	int found, i;
d2ccddf0 TM	380	int res;
1da177e4 LT	381
	382	res = 0;
	383	if (npages == 0)
	384	idx_end = ~0;
	385	else
	386	idx_end = idx_start + npages - 1;
	387
d2ccddf0 TM	388	for (;;) {
	389	found = radix_tree_gang_lookup(&nfsi->nfs_page_tree,
	390	(void **)&pgvec[0], idx_start,
	391	NFS_SCAN_MAXENTRIES);
	392	if (found <= 0)
1da177e4	393	break;
d2ccddf0 TM	394	for (i = 0; i < found; i++) {
	395	req = pgvec[i];
	396	if (req->wb_index > idx_end)
	397	goto out;
	398	idx_start = req->wb_index + 1;
	399	if (req->wb_list_head != head)
	400	continue;
	401	if (nfs_set_page_writeback_locked(req)) {
	402	nfs_list_remove_request(req);
	403	nfs_list_add_request(req, dst);
	404	res++;
	405	}
	406	}
1da177e4	407
1da177e4	408	}
d2ccddf0	409	out:
1da177e4 LT	410	return res;
	411	}
	412
f7b422b1	413	int __init nfs_init_nfspagecache(void)
1da177e4 LT	414	{
	415	nfs_page_cachep = kmem_cache_create("nfs_page",
	416	sizeof(struct nfs_page),
	417	0, SLAB_HWCACHE_ALIGN,
	418	NULL, NULL);
	419	if (nfs_page_cachep == NULL)
	420	return -ENOMEM;
	421
	422	return 0;
	423	}
	424
266bee88	425	void nfs_destroy_nfspagecache(void)
1da177e4	426	{
1a1d92c1	427	kmem_cache_destroy(nfs_page_cachep);
1da177e4 LT	428	}
1da177e4 LT	429