[deliverable/linux.git] / fs / ceph / export.c

#include <linux/ceph/ceph_debug.h>

#include <linux/exportfs.h>
#include <linux/slab.h>
#include <asm/unaligned.h>

#include "super.h"
#include "mds_client.h"

/*
 * NFS export support
 *
 * NFS re-export of a ceph mount is, at present, only semireliable.
 * The basic issue is that the Ceph architectures doesn't lend itself
 * well to generating filehandles that will remain valid forever.
 *
 * So, we do our best.  If you're lucky, your inode will be in the
 * client's cache.  If it's not, and you have a connectable fh, then
 * the MDS server may be able to find it for you.  Otherwise, you get
 * ESTALE.
 *
 * There are ways to this more reliable, but in the non-connectable fh
 * case, we won't every work perfectly, and in the connectable case,
 * some changes are needed on the MDS side to work better.
 */

/*
 * Basic fh
 */
struct ceph_nfs_fh {
	u64 ino;
} __attribute__ ((packed));

/*
 * Larger 'connectable' fh that includes parent ino and name hash.
 * Use this whenever possible, as it works more reliably.
 */
struct ceph_nfs_confh {
	u64 ino, parent_ino;
	u32 parent_name_hash;
} __attribute__ ((packed));

/*
 * The presence of @parent_inode here tells us whether NFS wants a
 * connectable file handle.  However, we want to make a connectionable
 * file handle unconditionally so that the MDS gets as much of a hint
 * as possible.  That means we only use @parent_dentry to indicate
 * whether nfsd wants a connectable fh, and whether we should indicate
 * failure from a too-small @max_len.
 */
static int ceph_encode_fh(struct inode *inode, u32 *rawfh, int *max_len,
			  struct inode *parent_inode)
{
	int type;
	struct ceph_nfs_fh *fh = (void *)rawfh;
	struct ceph_nfs_confh *cfh = (void *)rawfh;
	int connected_handle_length = sizeof(*cfh)/4;
	int handle_length = sizeof(*fh)/4;
	struct dentry *dentry = d_find_alias(inode);
	struct dentry *parent;

	/* don't re-export snaps */
	if (ceph_snap(inode) != CEPH_NOSNAP)
		return -EINVAL;

	/* if we found an alias, generate a connectable fh */
	if (*max_len >= connected_handle_length && dentry) {
		dout("encode_fh %p connectable\n", dentry);
		spin_lock(&dentry->d_lock);
		parent = dentry->d_parent;
		cfh->ino = ceph_ino(inode);
		cfh->parent_ino = ceph_ino(parent->d_inode);
		cfh->parent_name_hash = ceph_dentry_hash(parent->d_inode,
							 dentry);
		*max_len = connected_handle_length;
		type = 2;
		spin_unlock(&dentry->d_lock);
	} else if (*max_len >= handle_length) {
		if (parent_inode) {
			/* nfsd wants connectable */
			*max_len = connected_handle_length;
			type = 255;
		} else {
			dout("encode_fh %p\n", dentry);
			fh->ino = ceph_ino(inode);
			*max_len = handle_length;
			type = 1;
		}
	} else {
		*max_len = handle_length;
		type = 255;
	}
	if (dentry)
		dput(dentry);
	return type;
}

/*
 * convert regular fh to dentry
 *
 * FIXME: we should try harder by querying the mds for the ino.
 */
static struct dentry *__fh_to_dentry(struct super_block *sb,
				     struct ceph_nfs_fh *fh, int fh_len)
{
	struct ceph_mds_client *mdsc = ceph_sb_to_client(sb)->mdsc;
	struct inode *inode;
	struct dentry *dentry;
	struct ceph_vino vino;
	int err;

	if (fh_len < sizeof(*fh) / 4)
		return ERR_PTR(-ESTALE);

	dout("__fh_to_dentry %llx\n", fh->ino);
	vino.ino = fh->ino;
	vino.snap = CEPH_NOSNAP;
	inode = ceph_find_inode(sb, vino);
	if (!inode) {
		struct ceph_mds_request *req;

		req = ceph_mdsc_create_request(mdsc, CEPH_MDS_OP_LOOKUPINO,
					       USE_ANY_MDS);
		if (IS_ERR(req))
			return ERR_CAST(req);

		req->r_ino1 = vino;
		req->r_num_caps = 1;
		err = ceph_mdsc_do_request(mdsc, NULL, req);
		inode = req->r_target_inode;
		if (inode)
			ihold(inode);
		ceph_mdsc_put_request(req);
		if (!inode)
			return ERR_PTR(-ESTALE);
	}

	dentry = d_obtain_alias(inode);
	if (IS_ERR(dentry)) {
		pr_err("fh_to_dentry %llx -- inode %p but ENOMEM\n",
		       fh->ino, inode);
		iput(inode);
		return dentry;
	}
	err = ceph_init_dentry(dentry);
	if (err < 0) {
		iput(inode);
		return ERR_PTR(err);
	}
	dout("__fh_to_dentry %llx %p dentry %p\n", fh->ino, inode, dentry);
	return dentry;
}

/*
 * convert connectable fh to dentry
 */
static struct dentry *__cfh_to_dentry(struct super_block *sb,
				      struct ceph_nfs_confh *cfh, int fh_len)
{
	struct ceph_mds_client *mdsc = ceph_sb_to_client(sb)->mdsc;
	struct inode *inode;
	struct dentry *dentry;
	struct ceph_vino vino;
	int err;

	if (fh_len < sizeof(*cfh) / 4)
		return ERR_PTR(-ESTALE);

	dout("__cfh_to_dentry %llx (%llx/%x)\n",
	     cfh->ino, cfh->parent_ino, cfh->parent_name_hash);

	vino.ino = cfh->ino;
	vino.snap = CEPH_NOSNAP;
	inode = ceph_find_inode(sb, vino);
	if (!inode) {
		struct ceph_mds_request *req;

		req = ceph_mdsc_create_request(mdsc, CEPH_MDS_OP_LOOKUPHASH,
					       USE_ANY_MDS);
		if (IS_ERR(req))
			return ERR_CAST(req);

		req->r_ino1 = vino;
		req->r_ino2.ino = cfh->parent_ino;
		req->r_ino2.snap = CEPH_NOSNAP;
		req->r_path2 = kmalloc(16, GFP_NOFS);
		snprintf(req->r_path2, 16, "%d", cfh->parent_name_hash);
		req->r_num_caps = 1;
		err = ceph_mdsc_do_request(mdsc, NULL, req);
		inode = req->r_target_inode;
		if (inode)
			ihold(inode);
		ceph_mdsc_put_request(req);
		if (!inode)
			return ERR_PTR(err ? err : -ESTALE);
	}

	dentry = d_obtain_alias(inode);
	if (IS_ERR(dentry)) {
		pr_err("cfh_to_dentry %llx -- inode %p but ENOMEM\n",
		       cfh->ino, inode);
		iput(inode);
		return dentry;
	}
	err = ceph_init_dentry(dentry);
	if (err < 0) {
		iput(inode);
		return ERR_PTR(err);
	}
	dout("__cfh_to_dentry %llx %p dentry %p\n", cfh->ino, inode, dentry);
	return dentry;
}

static struct dentry *ceph_fh_to_dentry(struct super_block *sb, struct fid *fid,
					int fh_len, int fh_type)
{
	if (fh_type == 1)
		return __fh_to_dentry(sb, (struct ceph_nfs_fh *)fid->raw,
								fh_len);
	else
		return __cfh_to_dentry(sb, (struct ceph_nfs_confh *)fid->raw,
								fh_len);
}

/*
 * get parent, if possible.
 *
 * FIXME: we could do better by querying the mds to discover the
 * parent.
 */
static struct dentry *ceph_fh_to_parent(struct super_block *sb,
					 struct fid *fid,
					int fh_len, int fh_type)
{
	struct ceph_nfs_confh *cfh = (void *)fid->raw;
	struct ceph_vino vino;
	struct inode *inode;
	struct dentry *dentry;
	int err;

	if (fh_type == 1)
		return ERR_PTR(-ESTALE);
	if (fh_len < sizeof(*cfh) / 4)
		return ERR_PTR(-ESTALE);

	pr_debug("fh_to_parent %llx/%d\n", cfh->parent_ino,
		 cfh->parent_name_hash);

	vino.ino = cfh->ino;
	vino.snap = CEPH_NOSNAP;
	inode = ceph_find_inode(sb, vino);
	if (!inode)
		return ERR_PTR(-ESTALE);

	dentry = d_obtain_alias(inode);
	if (IS_ERR(dentry)) {
		pr_err("fh_to_parent %llx -- inode %p but ENOMEM\n",
		       cfh->ino, inode);
		iput(inode);
		return dentry;
	}
	err = ceph_init_dentry(dentry);
	if (err < 0) {
		iput(inode);
		return ERR_PTR(err);
	}
	dout("fh_to_parent %llx %p dentry %p\n", cfh->ino, inode, dentry);
	return dentry;
}

const struct export_operations ceph_export_ops = {
	.encode_fh = ceph_encode_fh,
	.fh_to_dentry = ceph_fh_to_dentry,
	.fh_to_parent = ceph_fh_to_parent,
};
Commit	Line	Data
3d14c5d2	1	#include <linux/ceph/ceph_debug.h>
a8e63b7d SW	2
a8e63b7d SW	3	#include <linux/exportfs.h>
5a0e3ad6	4	#include <linux/slab.h>
a8e63b7d SW	5	#include <asm/unaligned.h>
	6
	7	#include "super.h"
3d14c5d2	8	#include "mds_client.h"
a8e63b7d SW	9
	10	/*
	11	* NFS export support
	12	*
	13	* NFS re-export of a ceph mount is, at present, only semireliable.
	14	* The basic issue is that the Ceph architectures doesn't lend itself
	15	* well to generating filehandles that will remain valid forever.
	16	*
	17	* So, we do our best. If you're lucky, your inode will be in the
	18	* client's cache. If it's not, and you have a connectable fh, then
	19	* the MDS server may be able to find it for you. Otherwise, you get
	20	* ESTALE.
	21	*
	22	* There are ways to this more reliable, but in the non-connectable fh
	23	* case, we won't every work perfectly, and in the connectable case,
	24	* some changes are needed on the MDS side to work better.
	25	*/
	26
	27	/*
	28	* Basic fh
	29	*/
	30	struct ceph_nfs_fh {
	31	u64 ino;
	32	} __attribute__ ((packed));
	33
	34	/*
	35	* Larger 'connectable' fh that includes parent ino and name hash.
	36	* Use this whenever possible, as it works more reliably.
	37	*/
	38	struct ceph_nfs_confh {
	39	u64 ino, parent_ino;
	40	u32 parent_name_hash;
	41	} __attribute__ ((packed));
	42
c862868b SW	43	/*
	44	* The presence of @parent_inode here tells us whether NFS wants a
	45	* connectable file handle. However, we want to make a connectionable
	46	* file handle unconditionally so that the MDS gets as much of a hint
	47	* as possible. That means we only use @parent_dentry to indicate
	48	* whether nfsd wants a connectable fh, and whether we should indicate
	49	* failure from a too-small @max_len.
	50	*/
	51	static int ceph_encode_fh(struct inode inode, u32 rawfh, int *max_len,
	52	struct inode *parent_inode)
a8e63b7d	53	{
92923dcb	54	int type;
a8e63b7d SW	55	struct ceph_nfs_fh fh = (void )rawfh;
a8e63b7d SW	56	struct ceph_nfs_confh cfh = (void )rawfh;
92923dcb AK	57	int connected_handle_length = sizeof(*cfh)/4;
92923dcb AK	58	int handle_length = sizeof(*fh)/4;
c862868b SW	59	struct dentry *dentry = d_find_alias(inode);
c862868b SW	60	struct dentry *parent;
a8e63b7d SW	61
	62	/* don't re-export snaps */
	63	if (ceph_snap(inode) != CEPH_NOSNAP)
	64	return -EINVAL;
	65
c862868b SW	66	/* if we found an alias, generate a connectable fh */
c862868b SW	67	if (*max_len >= connected_handle_length && dentry) {
a8e63b7d	68	dout("encode_fh %p connectable\n", dentry);
c862868b SW	69	spin_lock(&dentry->d_lock);
	70	parent = dentry->d_parent;
	71	cfh->ino = ceph_ino(inode);
a8e63b7d	72	cfh->parent_ino = ceph_ino(parent->d_inode);
e5f86dc3 SW	73	cfh->parent_name_hash = ceph_dentry_hash(parent->d_inode,
e5f86dc3 SW	74	dentry);
92923dcb	75	*max_len = connected_handle_length;
a8e63b7d	76	type = 2;
c862868b	77	spin_unlock(&dentry->d_lock);
92923dcb	78	} else if (*max_len >= handle_length) {
c862868b SW	79	if (parent_inode) {
c862868b SW	80	/* nfsd wants connectable */
bba0cd0e	81	*max_len = connected_handle_length;
e5f86dc3 SW	82	type = 255;
	83	} else {
	84	dout("encode_fh %p\n", dentry);
c862868b	85	fh->ino = ceph_ino(inode);
e5f86dc3 SW	86	*max_len = handle_length;
e5f86dc3 SW	87	type = 1;
bba0cd0e	88	}
a8e63b7d	89	} else {
bba0cd0e	90	*max_len = handle_length;
e5f86dc3	91	type = 255;
a8e63b7d	92	}
52eb5a90 DZ	93	if (dentry)
52eb5a90 DZ	94	dput(dentry);
a8e63b7d SW	95	return type;
	96	}
	97
	98	/*
	99	* convert regular fh to dentry
	100	*
	101	* FIXME: we should try harder by querying the mds for the ino.
	102	*/
	103	static struct dentry __fh_to_dentry(struct super_block sb,
35c2a7f4	104	struct ceph_nfs_fh *fh, int fh_len)
a8e63b7d	105	{
3c454cf2	106	struct ceph_mds_client *mdsc = ceph_sb_to_client(sb)->mdsc;
a8e63b7d SW	107	struct inode *inode;
	108	struct dentry *dentry;
	109	struct ceph_vino vino;
	110	int err;
	111
35c2a7f4 HD	112	if (fh_len < sizeof(*fh) / 4)
	113	return ERR_PTR(-ESTALE);
	114
a8e63b7d SW	115	dout("__fh_to_dentry %llx\n", fh->ino);
	116	vino.ino = fh->ino;
	117	vino.snap = CEPH_NOSNAP;
	118	inode = ceph_find_inode(sb, vino);
3c454cf2 SW	119	if (!inode) {
	120	struct ceph_mds_request *req;
	121
	122	req = ceph_mdsc_create_request(mdsc, CEPH_MDS_OP_LOOKUPINO,
	123	USE_ANY_MDS);
	124	if (IS_ERR(req))
	125	return ERR_CAST(req);
	126
	127	req->r_ino1 = vino;
	128	req->r_num_caps = 1;
	129	err = ceph_mdsc_do_request(mdsc, NULL, req);
45e3d3ee SW	130	inode = req->r_target_inode;
45e3d3ee SW	131	if (inode)
70b666c3	132	ihold(inode);
3c454cf2	133	ceph_mdsc_put_request(req);
3c454cf2 SW	134	if (!inode)
	135	return ERR_PTR(-ESTALE);
	136	}
a8e63b7d SW	137
a8e63b7d SW	138	dentry = d_obtain_alias(inode);
0d509c94	139	if (IS_ERR(dentry)) {
a8e63b7d SW	140	pr_err("fh_to_dentry %llx -- inode %p but ENOMEM\n",
	141	fh->ino, inode);
	142	iput(inode);
0d509c94	143	return dentry;
a8e63b7d SW	144	}
a8e63b7d SW	145	err = ceph_init_dentry(dentry);
a8e63b7d SW	146	if (err < 0) {
	147	iput(inode);
	148	return ERR_PTR(err);
	149	}
	150	dout("__fh_to_dentry %llx %p dentry %p\n", fh->ino, inode, dentry);
	151	return dentry;
	152	}
	153
	154	/*
	155	* convert connectable fh to dentry
	156	*/
	157	static struct dentry __cfh_to_dentry(struct super_block sb,
35c2a7f4	158	struct ceph_nfs_confh *cfh, int fh_len)
a8e63b7d	159	{
3d14c5d2	160	struct ceph_mds_client *mdsc = ceph_sb_to_client(sb)->mdsc;
a8e63b7d SW	161	struct inode *inode;
	162	struct dentry *dentry;
	163	struct ceph_vino vino;
	164	int err;
	165
35c2a7f4 HD	166	if (fh_len < sizeof(*cfh) / 4)
	167	return ERR_PTR(-ESTALE);
	168
a8e63b7d SW	169	dout("__cfh_to_dentry %llx (%llx/%x)\n",
	170	cfh->ino, cfh->parent_ino, cfh->parent_name_hash);
	171
	172	vino.ino = cfh->ino;
	173	vino.snap = CEPH_NOSNAP;
	174	inode = ceph_find_inode(sb, vino);
	175	if (!inode) {
	176	struct ceph_mds_request *req;
	177
	178	req = ceph_mdsc_create_request(mdsc, CEPH_MDS_OP_LOOKUPHASH,
	179	USE_ANY_MDS);
	180	if (IS_ERR(req))
7e34bc52	181	return ERR_CAST(req);
a8e63b7d SW	182
	183	req->r_ino1 = vino;
	184	req->r_ino2.ino = cfh->parent_ino;
	185	req->r_ino2.snap = CEPH_NOSNAP;
	186	req->r_path2 = kmalloc(16, GFP_NOFS);
	187	snprintf(req->r_path2, 16, "%d", cfh->parent_name_hash);
	188	req->r_num_caps = 1;
	189	err = ceph_mdsc_do_request(mdsc, NULL, req);
45e3d3ee SW	190	inode = req->r_target_inode;
45e3d3ee SW	191	if (inode)
70b666c3	192	ihold(inode);
a8e63b7d	193	ceph_mdsc_put_request(req);
a8e63b7d SW	194	if (!inode)
	195	return ERR_PTR(err ? err : -ESTALE);
	196	}
	197
	198	dentry = d_obtain_alias(inode);
0d509c94	199	if (IS_ERR(dentry)) {
a8e63b7d SW	200	pr_err("cfh_to_dentry %llx -- inode %p but ENOMEM\n",
	201	cfh->ino, inode);
	202	iput(inode);
0d509c94	203	return dentry;
a8e63b7d SW	204	}
	205	err = ceph_init_dentry(dentry);
	206	if (err < 0) {
	207	iput(inode);
	208	return ERR_PTR(err);
	209	}
	210	dout("__cfh_to_dentry %llx %p dentry %p\n", cfh->ino, inode, dentry);
	211	return dentry;
	212	}
	213
	214	static struct dentry ceph_fh_to_dentry(struct super_block sb, struct fid *fid,
	215	int fh_len, int fh_type)
	216	{
	217	if (fh_type == 1)
35c2a7f4 HD	218	return __fh_to_dentry(sb, (struct ceph_nfs_fh *)fid->raw,
35c2a7f4 HD	219	fh_len);
a8e63b7d	220	else
35c2a7f4 HD	221	return __cfh_to_dentry(sb, (struct ceph_nfs_confh *)fid->raw,
35c2a7f4 HD	222	fh_len);
a8e63b7d SW	223	}
	224
	225	/*
	226	* get parent, if possible.
	227	*
	228	* FIXME: we could do better by querying the mds to discover the
	229	* parent.
	230	*/
	231	static struct dentry ceph_fh_to_parent(struct super_block sb,
	232	struct fid *fid,
	233	int fh_len, int fh_type)
	234	{
	235	struct ceph_nfs_confh cfh = (void )fid->raw;
	236	struct ceph_vino vino;
	237	struct inode *inode;
	238	struct dentry *dentry;
	239	int err;
	240
	241	if (fh_type == 1)
	242	return ERR_PTR(-ESTALE);
35c2a7f4 HD	243	if (fh_len < sizeof(*cfh) / 4)
35c2a7f4 HD	244	return ERR_PTR(-ESTALE);
a8e63b7d SW	245
	246	pr_debug("fh_to_parent %llx/%d\n", cfh->parent_ino,
	247	cfh->parent_name_hash);
	248
	249	vino.ino = cfh->ino;
	250	vino.snap = CEPH_NOSNAP;
	251	inode = ceph_find_inode(sb, vino);
	252	if (!inode)
	253	return ERR_PTR(-ESTALE);
	254
	255	dentry = d_obtain_alias(inode);
0d509c94	256	if (IS_ERR(dentry)) {
a8e63b7d SW	257	pr_err("fh_to_parent %llx -- inode %p but ENOMEM\n",
	258	cfh->ino, inode);
	259	iput(inode);
0d509c94	260	return dentry;
a8e63b7d SW	261	}
	262	err = ceph_init_dentry(dentry);
	263	if (err < 0) {
	264	iput(inode);
	265	return ERR_PTR(err);
	266	}
	267	dout("fh_to_parent %llx %p dentry %p\n", cfh->ino, inode, dentry);
	268	return dentry;
	269	}
	270
	271	const struct export_operations ceph_export_ops = {
	272	.encode_fh = ceph_encode_fh,
	273	.fh_to_dentry = ceph_fh_to_dentry,
	274	.fh_to_parent = ceph_fh_to_parent,
	275	};