[linux-2.6-block.git] / fs / nfs / fscache.c

// SPDX-License-Identifier: GPL-2.0-or-later
/* NFS filesystem cache interface
 *
 * Copyright (C) 2008 Red Hat, Inc. All Rights Reserved.
 * Written by David Howells (dhowells@redhat.com)
 */

#include <linux/init.h>
#include <linux/kernel.h>
#include <linux/sched.h>
#include <linux/mm.h>
#include <linux/nfs_fs.h>
#include <linux/nfs_fs_sb.h>
#include <linux/in6.h>
#include <linux/seq_file.h>
#include <linux/slab.h>
#include <linux/iversion.h>

#include "internal.h"
#include "iostat.h"
#include "fscache.h"

#define NFSDBG_FACILITY		NFSDBG_FSCACHE

static struct rb_root nfs_fscache_keys = RB_ROOT;
static DEFINE_SPINLOCK(nfs_fscache_keys_lock);

/*
 * Layout of the key for an NFS server cache object.
 */
struct nfs_server_key {
	struct {
		uint16_t	nfsversion;		/* NFS protocol version */
		uint16_t	family;			/* address family */
		__be16		port;			/* IP port */
	} hdr;
	union {
		struct in_addr	ipv4_addr;	/* IPv4 address */
		struct in6_addr ipv6_addr;	/* IPv6 address */
	};
} __packed;

/*
 * Get the per-client index cookie for an NFS client if the appropriate mount
 * flag was set
 * - We always try and get an index cookie for the client, but get filehandle
 *   cookies on a per-superblock basis, depending on the mount flags
 */
void nfs_fscache_get_client_cookie(struct nfs_client *clp)
{
	const struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *) &clp->cl_addr;
	const struct sockaddr_in *sin = (struct sockaddr_in *) &clp->cl_addr;
	struct nfs_server_key key;
	uint16_t len = sizeof(key.hdr);

	memset(&key, 0, sizeof(key));
	key.hdr.nfsversion = clp->rpc_ops->version;
	key.hdr.family = clp->cl_addr.ss_family;

	switch (clp->cl_addr.ss_family) {
	case AF_INET:
		key.hdr.port = sin->sin_port;
		key.ipv4_addr = sin->sin_addr;
		len += sizeof(key.ipv4_addr);
		break;

	case AF_INET6:
		key.hdr.port = sin6->sin6_port;
		key.ipv6_addr = sin6->sin6_addr;
		len += sizeof(key.ipv6_addr);
		break;

	default:
		printk(KERN_WARNING "NFS: Unknown network family '%d'\n",
		       clp->cl_addr.ss_family);
		clp->fscache = NULL;
		return;
	}

	/* create a cache index for looking up filehandles */
	clp->fscache = fscache_acquire_cookie(nfs_fscache_netfs.primary_index,
					      &nfs_fscache_server_index_def,
					      &key, len,
					      NULL, 0,
					      clp, 0, true);
	dfprintk(FSCACHE, "NFS: get client cookie (0x%p/0x%p)\n",
		 clp, clp->fscache);
}

/*
 * Dispose of a per-client cookie
 */
void nfs_fscache_release_client_cookie(struct nfs_client *clp)
{
	dfprintk(FSCACHE, "NFS: releasing client cookie (0x%p/0x%p)\n",
		 clp, clp->fscache);

	fscache_relinquish_cookie(clp->fscache, NULL, false);
	clp->fscache = NULL;
}

/*
 * Get the cache cookie for an NFS superblock.  We have to handle
 * uniquification here because the cache doesn't do it for us.
 *
 * The default uniquifier is just an empty string, but it may be overridden
 * either by the 'fsc=xxx' option to mount, or by inheriting it from the parent
 * superblock across an automount point of some nature.
 */
void nfs_fscache_get_super_cookie(struct super_block *sb, const char *uniq, int ulen)
{
	struct nfs_fscache_key *key, *xkey;
	struct nfs_server *nfss = NFS_SB(sb);
	struct rb_node **p, *parent;
	int diff;

	nfss->fscache_key = NULL;
	nfss->fscache = NULL;
	if (!(nfss->options & NFS_OPTION_FSCACHE))
		return;
	if (!uniq) {
		uniq = "";
		ulen = 1;
	}

	key = kzalloc(sizeof(*key) + ulen, GFP_KERNEL);
	if (!key)
		return;

	key->nfs_client = nfss->nfs_client;
	key->key.super.s_flags = sb->s_flags & NFS_MS_MASK;
	key->key.nfs_server.flags = nfss->flags;
	key->key.nfs_server.rsize = nfss->rsize;
	key->key.nfs_server.wsize = nfss->wsize;
	key->key.nfs_server.acregmin = nfss->acregmin;
	key->key.nfs_server.acregmax = nfss->acregmax;
	key->key.nfs_server.acdirmin = nfss->acdirmin;
	key->key.nfs_server.acdirmax = nfss->acdirmax;
	key->key.nfs_server.fsid = nfss->fsid;
	key->key.rpc_auth.au_flavor = nfss->client->cl_auth->au_flavor;

	key->key.uniq_len = ulen;
	memcpy(key->key.uniquifier, uniq, ulen);

	spin_lock(&nfs_fscache_keys_lock);
	p = &nfs_fscache_keys.rb_node;
	parent = NULL;
	while (*p) {
		parent = *p;
		xkey = rb_entry(parent, struct nfs_fscache_key, node);

		if (key->nfs_client < xkey->nfs_client)
			goto go_left;
		if (key->nfs_client > xkey->nfs_client)
			goto go_right;

		diff = memcmp(&key->key, &xkey->key, sizeof(key->key));
		if (diff < 0)
			goto go_left;
		if (diff > 0)
			goto go_right;

		if (key->key.uniq_len == 0)
			goto non_unique;
		diff = memcmp(key->key.uniquifier,
			      xkey->key.uniquifier,
			      key->key.uniq_len);
		if (diff < 0)
			goto go_left;
		if (diff > 0)
			goto go_right;
		goto non_unique;

	go_left:
		p = &(*p)->rb_left;
		continue;
	go_right:
		p = &(*p)->rb_right;
	}

	rb_link_node(&key->node, parent, p);
	rb_insert_color(&key->node, &nfs_fscache_keys);
	spin_unlock(&nfs_fscache_keys_lock);
	nfss->fscache_key = key;

	/* create a cache index for looking up filehandles */
	nfss->fscache = fscache_acquire_cookie(nfss->nfs_client->fscache,
					       &nfs_fscache_super_index_def,
					       key, sizeof(*key) + ulen,
					       NULL, 0,
					       nfss, 0, true);
	dfprintk(FSCACHE, "NFS: get superblock cookie (0x%p/0x%p)\n",
		 nfss, nfss->fscache);
	return;

non_unique:
	spin_unlock(&nfs_fscache_keys_lock);
	kfree(key);
	nfss->fscache_key = NULL;
	nfss->fscache = NULL;
	printk(KERN_WARNING "NFS:"
	       " Cache request denied due to non-unique superblock keys\n");
}

/*
 * release a per-superblock cookie
 */
void nfs_fscache_release_super_cookie(struct super_block *sb)
{
	struct nfs_server *nfss = NFS_SB(sb);

	dfprintk(FSCACHE, "NFS: releasing superblock cookie (0x%p/0x%p)\n",
		 nfss, nfss->fscache);

	fscache_relinquish_cookie(nfss->fscache, NULL, false);
	nfss->fscache = NULL;

	if (nfss->fscache_key) {
		spin_lock(&nfs_fscache_keys_lock);
		rb_erase(&nfss->fscache_key->node, &nfs_fscache_keys);
		spin_unlock(&nfs_fscache_keys_lock);
		kfree(nfss->fscache_key);
		nfss->fscache_key = NULL;
	}
}

/*
 * Initialise the per-inode cache cookie pointer for an NFS inode.
 */
void nfs_fscache_init_inode(struct inode *inode)
{
	struct nfs_fscache_inode_auxdata auxdata;
	struct nfs_server *nfss = NFS_SERVER(inode);
	struct nfs_inode *nfsi = NFS_I(inode);

	nfsi->fscache = NULL;
	if (!(nfss->fscache && S_ISREG(inode->i_mode)))
		return;

	memset(&auxdata, 0, sizeof(auxdata));
	auxdata.mtime = timespec64_to_timespec(nfsi->vfs_inode.i_mtime);
	auxdata.ctime = timespec64_to_timespec(nfsi->vfs_inode.i_ctime);

	if (NFS_SERVER(&nfsi->vfs_inode)->nfs_client->rpc_ops->version == 4)
		auxdata.change_attr = inode_peek_iversion_raw(&nfsi->vfs_inode);

	nfsi->fscache = fscache_acquire_cookie(NFS_SB(inode->i_sb)->fscache,
					       &nfs_fscache_inode_object_def,
					       nfsi->fh.data, nfsi->fh.size,
					       &auxdata, sizeof(auxdata),
					       nfsi, nfsi->vfs_inode.i_size, false);
}

/*
 * Release a per-inode cookie.
 */
void nfs_fscache_clear_inode(struct inode *inode)
{
	struct nfs_fscache_inode_auxdata auxdata;
	struct nfs_inode *nfsi = NFS_I(inode);
	struct fscache_cookie *cookie = nfs_i_fscache(inode);

	dfprintk(FSCACHE, "NFS: clear cookie (0x%p/0x%p)\n", nfsi, cookie);

	memset(&auxdata, 0, sizeof(auxdata));
	auxdata.mtime = timespec64_to_timespec(nfsi->vfs_inode.i_mtime);
	auxdata.ctime = timespec64_to_timespec(nfsi->vfs_inode.i_ctime);
	fscache_relinquish_cookie(cookie, &auxdata, false);
	nfsi->fscache = NULL;
}

static bool nfs_fscache_can_enable(void *data)
{
	struct inode *inode = data;

	return !inode_is_open_for_write(inode);
}

/*
 * Enable or disable caching for a file that is being opened as appropriate.
 * The cookie is allocated when the inode is initialised, but is not enabled at
 * that time.  Enablement is deferred to file-open time to avoid stat() and
 * access() thrashing the cache.
 *
 * For now, with NFS, only regular files that are open read-only will be able
 * to use the cache.
 *
 * We enable the cache for an inode if we open it read-only and it isn't
 * currently open for writing.  We disable the cache if the inode is open
 * write-only.
 *
 * The caller uses the file struct to pin i_writecount on the inode before
 * calling us when a file is opened for writing, so we can make use of that.
 *
 * Note that this may be invoked multiple times in parallel by parallel
 * nfs_open() functions.
 */
void nfs_fscache_open_file(struct inode *inode, struct file *filp)
{
	struct nfs_fscache_inode_auxdata auxdata;
	struct nfs_inode *nfsi = NFS_I(inode);
	struct fscache_cookie *cookie = nfs_i_fscache(inode);

	if (!fscache_cookie_valid(cookie))
		return;

	memset(&auxdata, 0, sizeof(auxdata));
	auxdata.mtime = timespec64_to_timespec(nfsi->vfs_inode.i_mtime);
	auxdata.ctime = timespec64_to_timespec(nfsi->vfs_inode.i_ctime);

	if (inode_is_open_for_write(inode)) {
		dfprintk(FSCACHE, "NFS: nfsi 0x%p disabling cache\n", nfsi);
		clear_bit(NFS_INO_FSCACHE, &nfsi->flags);
		fscache_disable_cookie(cookie, &auxdata, true);
		fscache_uncache_all_inode_pages(cookie, inode);
	} else {
		dfprintk(FSCACHE, "NFS: nfsi 0x%p enabling cache\n", nfsi);
		fscache_enable_cookie(cookie, &auxdata, nfsi->vfs_inode.i_size,
				      nfs_fscache_can_enable, inode);
		if (fscache_cookie_enabled(cookie))
			set_bit(NFS_INO_FSCACHE, &NFS_I(inode)->flags);
	}
}
EXPORT_SYMBOL_GPL(nfs_fscache_open_file);

/*
 * Release the caching state associated with a page, if the page isn't busy
 * interacting with the cache.
 * - Returns true (can release page) or false (page busy).
 */
int nfs_fscache_release_page(struct page *page, gfp_t gfp)
{
	if (PageFsCache(page)) {
		struct fscache_cookie *cookie = nfs_i_fscache(page->mapping->host);

		BUG_ON(!cookie);
		dfprintk(FSCACHE, "NFS: fscache releasepage (0x%p/0x%p/0x%p)\n",
			 cookie, page, NFS_I(page->mapping->host));

		if (!fscache_maybe_release_page(cookie, page, gfp))
			return 0;

		nfs_inc_fscache_stats(page->mapping->host,
				      NFSIOS_FSCACHE_PAGES_UNCACHED);
	}

	return 1;
}

/*
 * Release the caching state associated with a page if undergoing complete page
 * invalidation.
 */
void __nfs_fscache_invalidate_page(struct page *page, struct inode *inode)
{
	struct fscache_cookie *cookie = nfs_i_fscache(inode);

	BUG_ON(!cookie);

	dfprintk(FSCACHE, "NFS: fscache invalidatepage (0x%p/0x%p/0x%p)\n",
		 cookie, page, NFS_I(inode));

	fscache_wait_on_page_write(cookie, page);

	BUG_ON(!PageLocked(page));
	fscache_uncache_page(cookie, page);
	nfs_inc_fscache_stats(page->mapping->host,
			      NFSIOS_FSCACHE_PAGES_UNCACHED);
}

/*
 * Handle completion of a page being read from the cache.
 * - Called in process (keventd) context.
 */
static void nfs_readpage_from_fscache_complete(struct page *page,
					       void *context,
					       int error)
{
	dfprintk(FSCACHE,
		 "NFS: readpage_from_fscache_complete (0x%p/0x%p/%d)\n",
		 page, context, error);

	/* if the read completes with an error, we just unlock the page and let
	 * the VM reissue the readpage */
	if (!error) {
		SetPageUptodate(page);
		unlock_page(page);
	} else {
		error = nfs_readpage_async(context, page->mapping->host, page);
		if (error)
			unlock_page(page);
	}
}

/*
 * Retrieve a page from fscache
 */
int __nfs_readpage_from_fscache(struct nfs_open_context *ctx,
				struct inode *inode, struct page *page)
{
	int ret;

	dfprintk(FSCACHE,
		 "NFS: readpage_from_fscache(fsc:%p/p:%p(i:%lx f:%lx)/0x%p)\n",
		 nfs_i_fscache(inode), page, page->index, page->flags, inode);

	ret = fscache_read_or_alloc_page(nfs_i_fscache(inode),
					 page,
					 nfs_readpage_from_fscache_complete,
					 ctx,
					 GFP_KERNEL);

	switch (ret) {
	case 0: /* read BIO submitted (page in fscache) */
		dfprintk(FSCACHE,
			 "NFS:    readpage_from_fscache: BIO submitted\n");
		nfs_inc_fscache_stats(inode, NFSIOS_FSCACHE_PAGES_READ_OK);
		return ret;

	case -ENOBUFS: /* inode not in cache */
	case -ENODATA: /* page not in cache */
		nfs_inc_fscache_stats(inode, NFSIOS_FSCACHE_PAGES_READ_FAIL);
		dfprintk(FSCACHE,
			 "NFS:    readpage_from_fscache %d\n", ret);
		return 1;

	default:
		dfprintk(FSCACHE, "NFS:    readpage_from_fscache %d\n", ret);
		nfs_inc_fscache_stats(inode, NFSIOS_FSCACHE_PAGES_READ_FAIL);
	}
	return ret;
}

/*
 * Retrieve a set of pages from fscache
 */
int __nfs_readpages_from_fscache(struct nfs_open_context *ctx,
				 struct inode *inode,
				 struct address_space *mapping,
				 struct list_head *pages,
				 unsigned *nr_pages)
{
	unsigned npages = *nr_pages;
	int ret;

	dfprintk(FSCACHE, "NFS: nfs_getpages_from_fscache (0x%p/%u/0x%p)\n",
		 nfs_i_fscache(inode), npages, inode);

	ret = fscache_read_or_alloc_pages(nfs_i_fscache(inode),
					  mapping, pages, nr_pages,
					  nfs_readpage_from_fscache_complete,
					  ctx,
					  mapping_gfp_mask(mapping));
	if (*nr_pages < npages)
		nfs_add_fscache_stats(inode, NFSIOS_FSCACHE_PAGES_READ_OK,
				      npages);
	if (*nr_pages > 0)
		nfs_add_fscache_stats(inode, NFSIOS_FSCACHE_PAGES_READ_FAIL,
				      *nr_pages);

	switch (ret) {
	case 0: /* read submitted to the cache for all pages */
		BUG_ON(!list_empty(pages));
		BUG_ON(*nr_pages != 0);
		dfprintk(FSCACHE,
			 "NFS: nfs_getpages_from_fscache: submitted\n");

		return ret;

	case -ENOBUFS: /* some pages aren't cached and can't be */
	case -ENODATA: /* some pages aren't cached */
		dfprintk(FSCACHE,
			 "NFS: nfs_getpages_from_fscache: no page: %d\n", ret);
		return 1;

	default:
		dfprintk(FSCACHE,
			 "NFS: nfs_getpages_from_fscache: ret  %d\n", ret);
	}

	return ret;
}

/*
 * Store a newly fetched page in fscache
 * - PG_fscache must be set on the page
 */
void __nfs_readpage_to_fscache(struct inode *inode, struct page *page, int sync)
{
	int ret;

	dfprintk(FSCACHE,
		 "NFS: readpage_to_fscache(fsc:%p/p:%p(i:%lx f:%lx)/%d)\n",
		 nfs_i_fscache(inode), page, page->index, page->flags, sync);

	ret = fscache_write_page(nfs_i_fscache(inode), page,
				 inode->i_size, GFP_KERNEL);
	dfprintk(FSCACHE,
		 "NFS:     readpage_to_fscache: p:%p(i:%lu f:%lx) ret %d\n",
		 page, page->index, page->flags, ret);

	if (ret != 0) {
		fscache_uncache_page(nfs_i_fscache(inode), page);
		nfs_inc_fscache_stats(inode,
				      NFSIOS_FSCACHE_PAGES_WRITTEN_FAIL);
		nfs_inc_fscache_stats(inode, NFSIOS_FSCACHE_PAGES_UNCACHED);
	} else {
		nfs_inc_fscache_stats(inode,
				      NFSIOS_FSCACHE_PAGES_WRITTEN_OK);
	}
}
Commit	Line	Data
b4d0d230	1	// SPDX-License-Identifier: GPL-2.0-or-later
14727281 DH	2	/* NFS filesystem cache interface
	3	*
	4	* Copyright (C) 2008 Red Hat, Inc. All Rights Reserved.
	5	* Written by David Howells (dhowells@redhat.com)
14727281 DH	6	*/
	7
	8	#include <linux/init.h>
	9	#include <linux/kernel.h>
	10	#include <linux/sched.h>
	11	#include <linux/mm.h>
	12	#include <linux/nfs_fs.h>
	13	#include <linux/nfs_fs_sb.h>
	14	#include <linux/in6.h>
	15	#include <linux/seq_file.h>
5a0e3ad6	16	#include <linux/slab.h>
402cb8dd	17	#include <linux/iversion.h>
14727281 DH	18
14727281 DH	19	#include "internal.h"
545db45f	20	#include "iostat.h"
14727281 DH	21	#include "fscache.h"
	22
	23	#define NFSDBG_FACILITY NFSDBG_FSCACHE
	24
08734048 DH	25	static struct rb_root nfs_fscache_keys = RB_ROOT;
	26	static DEFINE_SPINLOCK(nfs_fscache_keys_lock);
	27
402cb8dd DH	28	/*
	29	* Layout of the key for an NFS server cache object.
	30	*/
	31	struct nfs_server_key {
	32	struct {
	33	uint16_t nfsversion; /* NFS protocol version */
	34	uint16_t family; /* address family */
	35	__be16 port; /* IP port */
	36	} hdr;
	37	union {
	38	struct in_addr ipv4_addr; /* IPv4 address */
	39	struct in6_addr ipv6_addr; /* IPv6 address */
	40	};
	41	} __packed;
	42
14727281 DH	43	/*
	44	* Get the per-client index cookie for an NFS client if the appropriate mount
	45	* flag was set
	46	* - We always try and get an index cookie for the client, but get filehandle
	47	* cookies on a per-superblock basis, depending on the mount flags
	48	*/
	49	void nfs_fscache_get_client_cookie(struct nfs_client *clp)
	50	{
402cb8dd DH	51	const struct sockaddr_in6 sin6 = (struct sockaddr_in6 ) &clp->cl_addr;
	52	const struct sockaddr_in sin = (struct sockaddr_in ) &clp->cl_addr;
	53	struct nfs_server_key key;
	54	uint16_t len = sizeof(key.hdr);
	55
	56	memset(&key, 0, sizeof(key));
	57	key.hdr.nfsversion = clp->rpc_ops->version;
	58	key.hdr.family = clp->cl_addr.ss_family;
	59
	60	switch (clp->cl_addr.ss_family) {
	61	case AF_INET:
	62	key.hdr.port = sin->sin_port;
	63	key.ipv4_addr = sin->sin_addr;
	64	len += sizeof(key.ipv4_addr);
	65	break;
	66
	67	case AF_INET6:
	68	key.hdr.port = sin6->sin6_port;
	69	key.ipv6_addr = sin6->sin6_addr;
	70	len += sizeof(key.ipv6_addr);
	71	break;
	72
	73	default:
	74	printk(KERN_WARNING "NFS: Unknown network family '%d'\n",
	75	clp->cl_addr.ss_family);
	76	clp->fscache = NULL;
	77	return;
	78	}
	79
14727281 DH	80	/* create a cache index for looking up filehandles */
	81	clp->fscache = fscache_acquire_cookie(nfs_fscache_netfs.primary_index,
	82	&nfs_fscache_server_index_def,
402cb8dd DH	83	&key, len,
402cb8dd DH	84	NULL, 0,
ee1235a9	85	clp, 0, true);
14727281 DH	86	dfprintk(FSCACHE, "NFS: get client cookie (0x%p/0x%p)\n",
	87	clp, clp->fscache);
	88	}
	89
	90	/*
	91	* Dispose of a per-client cookie
	92	*/
	93	void nfs_fscache_release_client_cookie(struct nfs_client *clp)
	94	{
	95	dfprintk(FSCACHE, "NFS: releasing client cookie (0x%p/0x%p)\n",
	96	clp, clp->fscache);
	97
402cb8dd	98	fscache_relinquish_cookie(clp->fscache, NULL, false);
14727281 DH	99	clp->fscache = NULL;
14727281 DH	100	}
08734048 DH	101
	102	/*
	103	* Get the cache cookie for an NFS superblock. We have to handle
	104	* uniquification here because the cache doesn't do it for us.
2df54806 DH	105	*
	106	* The default uniquifier is just an empty string, but it may be overridden
	107	* either by the 'fsc=xxx' option to mount, or by inheriting it from the parent
	108	* superblock across an automount point of some nature.
08734048	109	*/
2311b943	110	void nfs_fscache_get_super_cookie(struct super_block sb, const char uniq, int ulen)
08734048 DH	111	{
	112	struct nfs_fscache_key key, xkey;
	113	struct nfs_server *nfss = NFS_SB(sb);
	114	struct rb_node *p, parent;
2311b943	115	int diff;
2df54806	116
dea1bb35 TM	117	nfss->fscache_key = NULL;
	118	nfss->fscache = NULL;
	119	if (!(nfss->options & NFS_OPTION_FSCACHE))
	120	return;
2df54806 DH	121	if (!uniq) {
	122	uniq = "";
	123	ulen = 1;
	124	}
	125
08734048 DH	126	key = kzalloc(sizeof(*key) + ulen, GFP_KERNEL);
	127	if (!key)
	128	return;
	129
	130	key->nfs_client = nfss->nfs_client;
	131	key->key.super.s_flags = sb->s_flags & NFS_MS_MASK;
	132	key->key.nfs_server.flags = nfss->flags;
	133	key->key.nfs_server.rsize = nfss->rsize;
	134	key->key.nfs_server.wsize = nfss->wsize;
	135	key->key.nfs_server.acregmin = nfss->acregmin;
	136	key->key.nfs_server.acregmax = nfss->acregmax;
	137	key->key.nfs_server.acdirmin = nfss->acdirmin;
	138	key->key.nfs_server.acdirmax = nfss->acdirmax;
	139	key->key.nfs_server.fsid = nfss->fsid;
	140	key->key.rpc_auth.au_flavor = nfss->client->cl_auth->au_flavor;
	141
	142	key->key.uniq_len = ulen;
	143	memcpy(key->key.uniquifier, uniq, ulen);
	144
	145	spin_lock(&nfs_fscache_keys_lock);
	146	p = &nfs_fscache_keys.rb_node;
	147	parent = NULL;
	148	while (*p) {
	149	parent = *p;
	150	xkey = rb_entry(parent, struct nfs_fscache_key, node);
	151
	152	if (key->nfs_client < xkey->nfs_client)
	153	goto go_left;
	154	if (key->nfs_client > xkey->nfs_client)
	155	goto go_right;
	156
	157	diff = memcmp(&key->key, &xkey->key, sizeof(key->key));
	158	if (diff < 0)
	159	goto go_left;
	160	if (diff > 0)
	161	goto go_right;
	162
	163	if (key->key.uniq_len == 0)
	164	goto non_unique;
	165	diff = memcmp(key->key.uniquifier,
	166	xkey->key.uniquifier,
	167	key->key.uniq_len);
	168	if (diff < 0)
	169	goto go_left;
	170	if (diff > 0)
	171	goto go_right;
	172	goto non_unique;
	173
	174	go_left:
	175	p = &(*p)->rb_left;
	176	continue;
	177	go_right:
	178	p = &(*p)->rb_right;
	179	}
	180
	181	rb_link_node(&key->node, parent, p);
	182	rb_insert_color(&key->node, &nfs_fscache_keys);
	183	spin_unlock(&nfs_fscache_keys_lock);
	184	nfss->fscache_key = key;
	185
	186	/* create a cache index for looking up filehandles */
	187	nfss->fscache = fscache_acquire_cookie(nfss->nfs_client->fscache,
	188	&nfs_fscache_super_index_def,
402cb8dd DH	189	key, sizeof(*key) + ulen,
402cb8dd DH	190	NULL, 0,
ee1235a9	191	nfss, 0, true);
08734048 DH	192	dfprintk(FSCACHE, "NFS: get superblock cookie (0x%p/0x%p)\n",
	193	nfss, nfss->fscache);
	194	return;
	195
	196	non_unique:
	197	spin_unlock(&nfs_fscache_keys_lock);
	198	kfree(key);
	199	nfss->fscache_key = NULL;
	200	nfss->fscache = NULL;
	201	printk(KERN_WARNING "NFS:"
	202	" Cache request denied due to non-unique superblock keys\n");
	203	}
	204
	205	/*
	206	* release a per-superblock cookie
	207	*/
	208	void nfs_fscache_release_super_cookie(struct super_block *sb)
	209	{
	210	struct nfs_server *nfss = NFS_SB(sb);
	211
	212	dfprintk(FSCACHE, "NFS: releasing superblock cookie (0x%p/0x%p)\n",
	213	nfss, nfss->fscache);
	214
402cb8dd	215	fscache_relinquish_cookie(nfss->fscache, NULL, false);
08734048 DH	216	nfss->fscache = NULL;
	217
	218	if (nfss->fscache_key) {
	219	spin_lock(&nfs_fscache_keys_lock);
	220	rb_erase(&nfss->fscache_key->node, &nfs_fscache_keys);
	221	spin_unlock(&nfs_fscache_keys_lock);
	222	kfree(nfss->fscache_key);
	223	nfss->fscache_key = NULL;
	224	}
	225	}
ef79c097 DH	226
	227	/*
	228	* Initialise the per-inode cache cookie pointer for an NFS inode.
	229	*/
f1fe29b4	230	void nfs_fscache_init_inode(struct inode *inode)
ef79c097	231	{
402cb8dd	232	struct nfs_fscache_inode_auxdata auxdata;
dea1bb35	233	struct nfs_server *nfss = NFS_SERVER(inode);
ef79c097 DH	234	struct nfs_inode *nfsi = NFS_I(inode);
ef79c097 DH	235
f1fe29b4	236	nfsi->fscache = NULL;
dea1bb35	237	if (!(nfss->fscache && S_ISREG(inode->i_mode)))
ef79c097	238	return;
402cb8dd DH	239
402cb8dd DH	240	memset(&auxdata, 0, sizeof(auxdata));
95582b00 DD	241	auxdata.mtime = timespec64_to_timespec(nfsi->vfs_inode.i_mtime);
95582b00 DD	242	auxdata.ctime = timespec64_to_timespec(nfsi->vfs_inode.i_ctime);
402cb8dd DH	243
	244	if (NFS_SERVER(&nfsi->vfs_inode)->nfs_client->rpc_ops->version == 4)
	245	auxdata.change_attr = inode_peek_iversion_raw(&nfsi->vfs_inode);
	246
f1fe29b4 DH	247	nfsi->fscache = fscache_acquire_cookie(NFS_SB(inode->i_sb)->fscache,
f1fe29b4 DH	248	&nfs_fscache_inode_object_def,
402cb8dd DH	249	nfsi->fh.data, nfsi->fh.size,
402cb8dd DH	250	&auxdata, sizeof(auxdata),
ee1235a9	251	nfsi, nfsi->vfs_inode.i_size, false);
ef79c097 DH	252	}
	253
	254	/*
	255	* Release a per-inode cookie.
	256	*/
f1fe29b4	257	void nfs_fscache_clear_inode(struct inode *inode)
ef79c097	258	{
402cb8dd	259	struct nfs_fscache_inode_auxdata auxdata;
ef79c097	260	struct nfs_inode *nfsi = NFS_I(inode);
f1fe29b4	261	struct fscache_cookie *cookie = nfs_i_fscache(inode);
ef79c097	262
f1fe29b4	263	dfprintk(FSCACHE, "NFS: clear cookie (0x%p/0x%p)\n", nfsi, cookie);
ef79c097	264
402cb8dd	265	memset(&auxdata, 0, sizeof(auxdata));
95582b00 DD	266	auxdata.mtime = timespec64_to_timespec(nfsi->vfs_inode.i_mtime);
95582b00 DD	267	auxdata.ctime = timespec64_to_timespec(nfsi->vfs_inode.i_ctime);
402cb8dd	268	fscache_relinquish_cookie(cookie, &auxdata, false);
ef79c097 DH	269	nfsi->fscache = NULL;
	270	}
	271
f1fe29b4	272	static bool nfs_fscache_can_enable(void *data)
ef79c097	273	{
f1fe29b4	274	struct inode *inode = data;
ef79c097	275
f1fe29b4	276	return !inode_is_open_for_write(inode);
ef79c097 DH	277	}
	278
	279	/*
f1fe29b4 DH	280	* Enable or disable caching for a file that is being opened as appropriate.
	281	* The cookie is allocated when the inode is initialised, but is not enabled at
	282	* that time. Enablement is deferred to file-open time to avoid stat() and
	283	* access() thrashing the cache.
	284	*
	285	* For now, with NFS, only regular files that are open read-only will be able
	286	* to use the cache.
	287	*
	288	* We enable the cache for an inode if we open it read-only and it isn't
	289	* currently open for writing. We disable the cache if the inode is open
	290	* write-only.
	291	*
	292	* The caller uses the file struct to pin i_writecount on the inode before
	293	* calling us when a file is opened for writing, so we can make use of that.
	294	*
	295	* Note that this may be invoked multiple times in parallel by parallel
	296	* nfs_open() functions.
ef79c097	297	*/
f1fe29b4	298	void nfs_fscache_open_file(struct inode inode, struct file filp)
ef79c097	299	{
402cb8dd	300	struct nfs_fscache_inode_auxdata auxdata;
ef79c097	301	struct nfs_inode *nfsi = NFS_I(inode);
f1fe29b4	302	struct fscache_cookie *cookie = nfs_i_fscache(inode);
ef79c097	303
f1fe29b4 DH	304	if (!fscache_cookie_valid(cookie))
f1fe29b4 DH	305	return;
ef79c097	306
402cb8dd	307	memset(&auxdata, 0, sizeof(auxdata));
95582b00 DD	308	auxdata.mtime = timespec64_to_timespec(nfsi->vfs_inode.i_mtime);
95582b00 DD	309	auxdata.ctime = timespec64_to_timespec(nfsi->vfs_inode.i_ctime);
402cb8dd	310
f1fe29b4 DH	311	if (inode_is_open_for_write(inode)) {
	312	dfprintk(FSCACHE, "NFS: nfsi 0x%p disabling cache\n", nfsi);
	313	clear_bit(NFS_INO_FSCACHE, &nfsi->flags);
402cb8dd	314	fscache_disable_cookie(cookie, &auxdata, true);
f1fe29b4 DH	315	fscache_uncache_all_inode_pages(cookie, inode);
	316	} else {
	317	dfprintk(FSCACHE, "NFS: nfsi 0x%p enabling cache\n", nfsi);
ee1235a9	318	fscache_enable_cookie(cookie, &auxdata, nfsi->vfs_inode.i_size,
402cb8dd	319	nfs_fscache_can_enable, inode);
f1fe29b4 DH	320	if (fscache_cookie_enabled(cookie))
f1fe29b4 DH	321	set_bit(NFS_INO_FSCACHE, &NFS_I(inode)->flags);
ef79c097	322	}
ef79c097	323	}
f1fe29b4	324	EXPORT_SYMBOL_GPL(nfs_fscache_open_file);
545db45f DH	325
	326	/*
	327	* Release the caching state associated with a page, if the page isn't busy
	328	* interacting with the cache.
	329	* - Returns true (can release page) or false (page busy).
	330	*/
	331	int nfs_fscache_release_page(struct page *page, gfp_t gfp)
	332	{
545db45f	333	if (PageFsCache(page)) {
f1fe29b4	334	struct fscache_cookie *cookie = nfs_i_fscache(page->mapping->host);
2c174009 TM	335
2c174009 TM	336	BUG_ON(!cookie);
545db45f	337	dfprintk(FSCACHE, "NFS: fscache releasepage (0x%p/0x%p/0x%p)\n",
f1fe29b4	338	cookie, page, NFS_I(page->mapping->host));
545db45f	339
201a1542 DH	340	if (!fscache_maybe_release_page(cookie, page, gfp))
	341	return 0;
	342
e9f456ca LR	343	nfs_inc_fscache_stats(page->mapping->host,
e9f456ca LR	344	NFSIOS_FSCACHE_PAGES_UNCACHED);
545db45f DH	345	}
	346
	347	return 1;
	348	}
	349
	350	/*
	351	* Release the caching state associated with a page if undergoing complete page
	352	* invalidation.
	353	*/
	354	void __nfs_fscache_invalidate_page(struct page page, struct inode inode)
	355	{
f1fe29b4	356	struct fscache_cookie *cookie = nfs_i_fscache(inode);
545db45f DH	357
	358	BUG_ON(!cookie);
	359
	360	dfprintk(FSCACHE, "NFS: fscache invalidatepage (0x%p/0x%p/0x%p)\n",
f1fe29b4	361	cookie, page, NFS_I(inode));
545db45f DH	362
	363	fscache_wait_on_page_write(cookie, page);
	364
	365	BUG_ON(!PageLocked(page));
	366	fscache_uncache_page(cookie, page);
e9f456ca LR	367	nfs_inc_fscache_stats(page->mapping->host,
e9f456ca LR	368	NFSIOS_FSCACHE_PAGES_UNCACHED);
545db45f	369	}
9a9fc1c0 DH	370
	371	/*
	372	* Handle completion of a page being read from the cache.
	373	* - Called in process (keventd) context.
	374	*/
	375	static void nfs_readpage_from_fscache_complete(struct page *page,
	376	void *context,
	377	int error)
	378	{
	379	dfprintk(FSCACHE,
	380	"NFS: readpage_from_fscache_complete (0x%p/0x%p/%d)\n",
	381	page, context, error);
	382
	383	/* if the read completes with an error, we just unlock the page and let
	384	* the VM reissue the readpage */
	385	if (!error) {
	386	SetPageUptodate(page);
	387	unlock_page(page);
	388	} else {
	389	error = nfs_readpage_async(context, page->mapping->host, page);
	390	if (error)
	391	unlock_page(page);
	392	}
	393	}
	394
	395	/*
	396	* Retrieve a page from fscache
	397	*/
	398	int __nfs_readpage_from_fscache(struct nfs_open_context *ctx,
	399	struct inode inode, struct page page)
	400	{
	401	int ret;
	402
	403	dfprintk(FSCACHE,
	404	"NFS: readpage_from_fscache(fsc:%p/p:%p(i:%lx f:%lx)/0x%p)\n",
f1fe29b4	405	nfs_i_fscache(inode), page, page->index, page->flags, inode);
9a9fc1c0	406
f1fe29b4	407	ret = fscache_read_or_alloc_page(nfs_i_fscache(inode),
9a9fc1c0 DH	408	page,
	409	nfs_readpage_from_fscache_complete,
	410	ctx,
	411	GFP_KERNEL);
	412
	413	switch (ret) {
	414	case 0: /* read BIO submitted (page in fscache) */
	415	dfprintk(FSCACHE,
	416	"NFS: readpage_from_fscache: BIO submitted\n");
e9f456ca	417	nfs_inc_fscache_stats(inode, NFSIOS_FSCACHE_PAGES_READ_OK);
9a9fc1c0 DH	418	return ret;
	419
	420	case -ENOBUFS: /* inode not in cache */
	421	case -ENODATA: /* page not in cache */
e9f456ca	422	nfs_inc_fscache_stats(inode, NFSIOS_FSCACHE_PAGES_READ_FAIL);
9a9fc1c0 DH	423	dfprintk(FSCACHE,
	424	"NFS: readpage_from_fscache %d\n", ret);
	425	return 1;
	426
	427	default:
	428	dfprintk(FSCACHE, "NFS: readpage_from_fscache %d\n", ret);
e9f456ca	429	nfs_inc_fscache_stats(inode, NFSIOS_FSCACHE_PAGES_READ_FAIL);
9a9fc1c0 DH	430	}
	431	return ret;
	432	}
	433
	434	/*
	435	* Retrieve a set of pages from fscache
	436	*/
	437	int __nfs_readpages_from_fscache(struct nfs_open_context *ctx,
	438	struct inode *inode,
	439	struct address_space *mapping,
	440	struct list_head *pages,
	441	unsigned *nr_pages)
	442	{
0f15c53d CL	443	unsigned npages = *nr_pages;
0f15c53d CL	444	int ret;
9a9fc1c0 DH	445
9a9fc1c0 DH	446	dfprintk(FSCACHE, "NFS: nfs_getpages_from_fscache (0x%p/%u/0x%p)\n",
f1fe29b4	447	nfs_i_fscache(inode), npages, inode);
9a9fc1c0	448
f1fe29b4	449	ret = fscache_read_or_alloc_pages(nfs_i_fscache(inode),
9a9fc1c0 DH	450	mapping, pages, nr_pages,
	451	nfs_readpage_from_fscache_complete,
	452	ctx,
	453	mapping_gfp_mask(mapping));
	454	if (*nr_pages < npages)
	455	nfs_add_fscache_stats(inode, NFSIOS_FSCACHE_PAGES_READ_OK,
	456	npages);
	457	if (*nr_pages > 0)
	458	nfs_add_fscache_stats(inode, NFSIOS_FSCACHE_PAGES_READ_FAIL,
	459	*nr_pages);
	460
	461	switch (ret) {
	462	case 0: /* read submitted to the cache for all pages */
	463	BUG_ON(!list_empty(pages));
	464	BUG_ON(*nr_pages != 0);
	465	dfprintk(FSCACHE,
	466	"NFS: nfs_getpages_from_fscache: submitted\n");
	467
	468	return ret;
	469
	470	case -ENOBUFS: /* some pages aren't cached and can't be */
	471	case -ENODATA: /* some pages aren't cached */
	472	dfprintk(FSCACHE,
	473	"NFS: nfs_getpages_from_fscache: no page: %d\n", ret);
	474	return 1;
	475
	476	default:
	477	dfprintk(FSCACHE,
	478	"NFS: nfs_getpages_from_fscache: ret %d\n", ret);
	479	}
	480
	481	return ret;
	482	}
7f8e05f6 DH	483
	484	/*
	485	* Store a newly fetched page in fscache
	486	* - PG_fscache must be set on the page
	487	*/
	488	void __nfs_readpage_to_fscache(struct inode inode, struct page page, int sync)
	489	{
	490	int ret;
	491
	492	dfprintk(FSCACHE,
	493	"NFS: readpage_to_fscache(fsc:%p/p:%p(i:%lx f:%lx)/%d)\n",
f1fe29b4	494	nfs_i_fscache(inode), page, page->index, page->flags, sync);
7f8e05f6	495
ee1235a9 DH	496	ret = fscache_write_page(nfs_i_fscache(inode), page,
ee1235a9 DH	497	inode->i_size, GFP_KERNEL);
7f8e05f6 DH	498	dfprintk(FSCACHE,
	499	"NFS: readpage_to_fscache: p:%p(i:%lu f:%lx) ret %d\n",
	500	page, page->index, page->flags, ret);
	501
	502	if (ret != 0) {
f1fe29b4	503	fscache_uncache_page(nfs_i_fscache(inode), page);
e9f456ca LR	504	nfs_inc_fscache_stats(inode,
	505	NFSIOS_FSCACHE_PAGES_WRITTEN_FAIL);
	506	nfs_inc_fscache_stats(inode, NFSIOS_FSCACHE_PAGES_UNCACHED);
7f8e05f6	507	} else {
e9f456ca LR	508	nfs_inc_fscache_stats(inode,
e9f456ca LR	509	NFSIOS_FSCACHE_PAGES_WRITTEN_OK);
7f8e05f6 DH	510	}
7f8e05f6 DH	511	}