[linux.git] / fs / xfs / xfs_attr_list.c

// SPDX-License-Identifier: GPL-2.0
/*
 * Copyright (c) 2000-2005 Silicon Graphics, Inc.
 * Copyright (c) 2013 Red Hat, Inc.
 * All Rights Reserved.
 */
#include "xfs.h"
#include "xfs_fs.h"
#include "xfs_format.h"
#include "xfs_log_format.h"
#include "xfs_trans_resv.h"
#include "xfs_bit.h"
#include "xfs_mount.h"
#include "xfs_da_format.h"
#include "xfs_da_btree.h"
#include "xfs_inode.h"
#include "xfs_trans.h"
#include "xfs_inode_item.h"
#include "xfs_bmap.h"
#include "xfs_attr.h"
#include "xfs_attr_sf.h"
#include "xfs_attr_remote.h"
#include "xfs_attr_leaf.h"
#include "xfs_error.h"
#include "xfs_trace.h"
#include "xfs_buf_item.h"
#include "xfs_cksum.h"
#include "xfs_dir2.h"

STATIC int
xfs_attr_shortform_compare(const void *a, const void *b)
{
	xfs_attr_sf_sort_t *sa, *sb;

	sa = (xfs_attr_sf_sort_t *)a;
	sb = (xfs_attr_sf_sort_t *)b;
	if (sa->hash < sb->hash) {
		return -1;
	} else if (sa->hash > sb->hash) {
		return 1;
	} else {
		return sa->entno - sb->entno;
	}
}

#define XFS_ISRESET_CURSOR(cursor) \
	(!((cursor)->initted) && !((cursor)->hashval) && \
	 !((cursor)->blkno) && !((cursor)->offset))
/*
 * Copy out entries of shortform attribute lists for attr_list().
 * Shortform attribute lists are not stored in hashval sorted order.
 * If the output buffer is not large enough to hold them all, then we
 * we have to calculate each entries' hashvalue and sort them before
 * we can begin returning them to the user.
 */
static int
xfs_attr_shortform_list(xfs_attr_list_context_t *context)
{
	attrlist_cursor_kern_t *cursor;
	xfs_attr_sf_sort_t *sbuf, *sbp;
	xfs_attr_shortform_t *sf;
	xfs_attr_sf_entry_t *sfe;
	xfs_inode_t *dp;
	int sbsize, nsbuf, count, i;

	ASSERT(context != NULL);
	dp = context->dp;
	ASSERT(dp != NULL);
	ASSERT(dp->i_afp != NULL);
	sf = (xfs_attr_shortform_t *)dp->i_afp->if_u1.if_data;
	ASSERT(sf != NULL);
	if (!sf->hdr.count)
		return 0;
	cursor = context->cursor;
	ASSERT(cursor != NULL);

	trace_xfs_attr_list_sf(context);

	/*
	 * If the buffer is large enough and the cursor is at the start,
	 * do not bother with sorting since we will return everything in
	 * one buffer and another call using the cursor won't need to be
	 * made.
	 * Note the generous fudge factor of 16 overhead bytes per entry.
	 * If bufsize is zero then put_listent must be a search function
	 * and can just scan through what we have.
	 */
	if (context->bufsize == 0 ||
	    (XFS_ISRESET_CURSOR(cursor) &&
	     (dp->i_afp->if_bytes + sf->hdr.count * 16) < context->bufsize)) {
		for (i = 0, sfe = &sf->list[0]; i < sf->hdr.count; i++) {
			context->put_listent(context,
					     sfe->flags,
					     sfe->nameval,
					     (int)sfe->namelen,
					     (int)sfe->valuelen);
			/*
			 * Either search callback finished early or
			 * didn't fit it all in the buffer after all.
			 */
			if (context->seen_enough)
				break;
			sfe = XFS_ATTR_SF_NEXTENTRY(sfe);
		}
		trace_xfs_attr_list_sf_all(context);
		return 0;
	}

	/* do no more for a search callback */
	if (context->bufsize == 0)
		return 0;

	/*
	 * It didn't all fit, so we have to sort everything on hashval.
	 */
	sbsize = sf->hdr.count * sizeof(*sbuf);
	sbp = sbuf = kmem_alloc(sbsize, KM_SLEEP | KM_NOFS);

	/*
	 * Scan the attribute list for the rest of the entries, storing
	 * the relevant info from only those that match into a buffer.
	 */
	nsbuf = 0;
	for (i = 0, sfe = &sf->list[0]; i < sf->hdr.count; i++) {
		if (unlikely(
		    ((char *)sfe < (char *)sf) ||
		    ((char *)sfe >= ((char *)sf + dp->i_afp->if_bytes)))) {
			XFS_CORRUPTION_ERROR("xfs_attr_shortform_list",
					     XFS_ERRLEVEL_LOW,
					     context->dp->i_mount, sfe,
					     sizeof(*sfe));
			kmem_free(sbuf);
			return -EFSCORRUPTED;
		}

		sbp->entno = i;
		sbp->hash = xfs_da_hashname(sfe->nameval, sfe->namelen);
		sbp->name = sfe->nameval;
		sbp->namelen = sfe->namelen;
		/* These are bytes, and both on-disk, don't endian-flip */
		sbp->valuelen = sfe->valuelen;
		sbp->flags = sfe->flags;
		sfe = XFS_ATTR_SF_NEXTENTRY(sfe);
		sbp++;
		nsbuf++;
	}

	/*
	 * Sort the entries on hash then entno.
	 */
	xfs_sort(sbuf, nsbuf, sizeof(*sbuf), xfs_attr_shortform_compare);

	/*
	 * Re-find our place IN THE SORTED LIST.
	 */
	count = 0;
	cursor->initted = 1;
	cursor->blkno = 0;
	for (sbp = sbuf, i = 0; i < nsbuf; i++, sbp++) {
		if (sbp->hash == cursor->hashval) {
			if (cursor->offset == count) {
				break;
			}
			count++;
		} else if (sbp->hash > cursor->hashval) {
			break;
		}
	}
	if (i == nsbuf) {
		kmem_free(sbuf);
		return 0;
	}

	/*
	 * Loop putting entries into the user buffer.
	 */
	for ( ; i < nsbuf; i++, sbp++) {
		if (cursor->hashval != sbp->hash) {
			cursor->hashval = sbp->hash;
			cursor->offset = 0;
		}
		context->put_listent(context,
				     sbp->flags,
				     sbp->name,
				     sbp->namelen,
				     sbp->valuelen);
		if (context->seen_enough)
			break;
		cursor->offset++;
	}

	kmem_free(sbuf);
	return 0;
}

/*
 * We didn't find the block & hash mentioned in the cursor state, so
 * walk down the attr btree looking for the hash.
 */
STATIC int
xfs_attr_node_list_lookup(
	struct xfs_attr_list_context	*context,
	struct attrlist_cursor_kern	*cursor,
	struct xfs_buf			**pbp)
{
	struct xfs_da3_icnode_hdr	nodehdr;
	struct xfs_da_intnode		*node;
	struct xfs_da_node_entry	*btree;
	struct xfs_inode		*dp = context->dp;
	struct xfs_mount		*mp = dp->i_mount;
	struct xfs_trans		*tp = context->tp;
	struct xfs_buf			*bp;
	int				i;
	int				error = 0;
	unsigned int			expected_level = 0;
	uint16_t			magic;

	ASSERT(*pbp == NULL);
	cursor->blkno = 0;
	for (;;) {
		error = xfs_da3_node_read(tp, dp, cursor->blkno, -1, &bp,
				XFS_ATTR_FORK);
		if (error)
			return error;
		node = bp->b_addr;
		magic = be16_to_cpu(node->hdr.info.magic);
		if (magic == XFS_ATTR_LEAF_MAGIC ||
		    magic == XFS_ATTR3_LEAF_MAGIC)
			break;
		if (magic != XFS_DA_NODE_MAGIC &&
		    magic != XFS_DA3_NODE_MAGIC) {
			XFS_CORRUPTION_ERROR(__func__, XFS_ERRLEVEL_LOW, mp,
					node, sizeof(*node));
			goto out_corruptbuf;
		}

		dp->d_ops->node_hdr_from_disk(&nodehdr, node);

		/* Tree taller than we can handle; bail out! */
		if (nodehdr.level >= XFS_DA_NODE_MAXDEPTH)
			goto out_corruptbuf;

		/* Check the level from the root node. */
		if (cursor->blkno == 0)
			expected_level = nodehdr.level - 1;
		else if (expected_level != nodehdr.level)
			goto out_corruptbuf;
		else
			expected_level--;

		btree = dp->d_ops->node_tree_p(node);
		for (i = 0; i < nodehdr.count; btree++, i++) {
			if (cursor->hashval <= be32_to_cpu(btree->hashval)) {
				cursor->blkno = be32_to_cpu(btree->before);
				trace_xfs_attr_list_node_descend(context,
						btree);
				break;
			}
		}
		xfs_trans_brelse(tp, bp);

		if (i == nodehdr.count)
			return 0;

		/* We can't point back to the root. */
		if (cursor->blkno == 0)
			return -EFSCORRUPTED;
	}

	if (expected_level != 0)
		goto out_corruptbuf;

	*pbp = bp;
	return 0;

out_corruptbuf:
	xfs_trans_brelse(tp, bp);
	return -EFSCORRUPTED;
}

STATIC int
xfs_attr_node_list(
	struct xfs_attr_list_context	*context)
{
	struct xfs_attr3_icleaf_hdr	leafhdr;
	struct attrlist_cursor_kern	*cursor;
	struct xfs_attr_leafblock	*leaf;
	struct xfs_da_intnode		*node;
	struct xfs_buf			*bp;
	struct xfs_inode		*dp = context->dp;
	struct xfs_mount		*mp = dp->i_mount;
	int				error;

	trace_xfs_attr_node_list(context);

	cursor = context->cursor;
	cursor->initted = 1;

	/*
	 * Do all sorts of validation on the passed-in cursor structure.
	 * If anything is amiss, ignore the cursor and look up the hashval
	 * starting from the btree root.
	 */
	bp = NULL;
	if (cursor->blkno > 0) {
		error = xfs_da3_node_read(context->tp, dp, cursor->blkno, -1,
					      &bp, XFS_ATTR_FORK);
		if ((error != 0) && (error != -EFSCORRUPTED))
			return error;
		if (bp) {
			struct xfs_attr_leaf_entry *entries;

			node = bp->b_addr;
			switch (be16_to_cpu(node->hdr.info.magic)) {
			case XFS_DA_NODE_MAGIC:
			case XFS_DA3_NODE_MAGIC:
				trace_xfs_attr_list_wrong_blk(context);
				xfs_trans_brelse(context->tp, bp);
				bp = NULL;
				break;
			case XFS_ATTR_LEAF_MAGIC:
			case XFS_ATTR3_LEAF_MAGIC:
				leaf = bp->b_addr;
				xfs_attr3_leaf_hdr_from_disk(mp->m_attr_geo,
							     &leafhdr, leaf);
				entries = xfs_attr3_leaf_entryp(leaf);
				if (cursor->hashval > be32_to_cpu(
						entries[leafhdr.count - 1].hashval)) {
					trace_xfs_attr_list_wrong_blk(context);
					xfs_trans_brelse(context->tp, bp);
					bp = NULL;
				} else if (cursor->hashval <= be32_to_cpu(
						entries[0].hashval)) {
					trace_xfs_attr_list_wrong_blk(context);
					xfs_trans_brelse(context->tp, bp);
					bp = NULL;
				}
				break;
			default:
				trace_xfs_attr_list_wrong_blk(context);
				xfs_trans_brelse(context->tp, bp);
				bp = NULL;
			}
		}
	}

	/*
	 * We did not find what we expected given the cursor's contents,
	 * so we start from the top and work down based on the hash value.
	 * Note that start of node block is same as start of leaf block.
	 */
	if (bp == NULL) {
		error = xfs_attr_node_list_lookup(context, cursor, &bp);
		if (error || !bp)
			return error;
	}
	ASSERT(bp != NULL);

	/*
	 * Roll upward through the blocks, processing each leaf block in
	 * order.  As long as there is space in the result buffer, keep
	 * adding the information.
	 */
	for (;;) {
		leaf = bp->b_addr;
		xfs_attr3_leaf_list_int(bp, context);
		xfs_attr3_leaf_hdr_from_disk(mp->m_attr_geo, &leafhdr, leaf);
		if (context->seen_enough || leafhdr.forw == 0)
			break;
		cursor->blkno = leafhdr.forw;
		xfs_trans_brelse(context->tp, bp);
		error = xfs_attr3_leaf_read(context->tp, dp, cursor->blkno, -1, &bp);
		if (error)
			return error;
	}
	xfs_trans_brelse(context->tp, bp);
	return 0;
}

/*
 * Copy out attribute list entries for attr_list(), for leaf attribute lists.
 */
void
xfs_attr3_leaf_list_int(
	struct xfs_buf			*bp,
	struct xfs_attr_list_context	*context)
{
	struct attrlist_cursor_kern	*cursor;
	struct xfs_attr_leafblock	*leaf;
	struct xfs_attr3_icleaf_hdr	ichdr;
	struct xfs_attr_leaf_entry	*entries;
	struct xfs_attr_leaf_entry	*entry;
	int				i;
	struct xfs_mount		*mp = context->dp->i_mount;

	trace_xfs_attr_list_leaf(context);

	leaf = bp->b_addr;
	xfs_attr3_leaf_hdr_from_disk(mp->m_attr_geo, &ichdr, leaf);
	entries = xfs_attr3_leaf_entryp(leaf);

	cursor = context->cursor;
	cursor->initted = 1;

	/*
	 * Re-find our place in the leaf block if this is a new syscall.
	 */
	if (context->resynch) {
		entry = &entries[0];
		for (i = 0; i < ichdr.count; entry++, i++) {
			if (be32_to_cpu(entry->hashval) == cursor->hashval) {
				if (cursor->offset == context->dupcnt) {
					context->dupcnt = 0;
					break;
				}
				context->dupcnt++;
			} else if (be32_to_cpu(entry->hashval) >
					cursor->hashval) {
				context->dupcnt = 0;
				break;
			}
		}
		if (i == ichdr.count) {
			trace_xfs_attr_list_notfound(context);
			return;
		}
	} else {
		entry = &entries[0];
		i = 0;
	}
	context->resynch = 0;

	/*
	 * We have found our place, start copying out the new attributes.
	 */
	for (; i < ichdr.count; entry++, i++) {
		char *name;
		int namelen, valuelen;

		if (be32_to_cpu(entry->hashval) != cursor->hashval) {
			cursor->hashval = be32_to_cpu(entry->hashval);
			cursor->offset = 0;
		}

		if ((entry->flags & XFS_ATTR_INCOMPLETE) &&
		    !(context->flags & ATTR_INCOMPLETE))
			continue;		/* skip incomplete entries */

		if (entry->flags & XFS_ATTR_LOCAL) {
			xfs_attr_leaf_name_local_t *name_loc;

			name_loc = xfs_attr3_leaf_name_local(leaf, i);
			name = name_loc->nameval;
			namelen = name_loc->namelen;
			valuelen = be16_to_cpu(name_loc->valuelen);
		} else {
			xfs_attr_leaf_name_remote_t *name_rmt;

			name_rmt = xfs_attr3_leaf_name_remote(leaf, i);
			name = name_rmt->name;
			namelen = name_rmt->namelen;
			valuelen = be32_to_cpu(name_rmt->valuelen);
		}

		context->put_listent(context, entry->flags,
					      name, namelen, valuelen);
		if (context->seen_enough)
			break;
		cursor->offset++;
	}
	trace_xfs_attr_list_leaf_end(context);
	return;
}

/*
 * Copy out attribute entries for attr_list(), for leaf attribute lists.
 */
STATIC int
xfs_attr_leaf_list(xfs_attr_list_context_t *context)
{
	int error;
	struct xfs_buf *bp;

	trace_xfs_attr_leaf_list(context);

	context->cursor->blkno = 0;
	error = xfs_attr3_leaf_read(context->tp, context->dp, 0, -1, &bp);
	if (error)
		return error;

	xfs_attr3_leaf_list_int(bp, context);
	xfs_trans_brelse(context->tp, bp);
	return 0;
}

int
xfs_attr_list_int_ilocked(
	struct xfs_attr_list_context	*context)
{
	struct xfs_inode		*dp = context->dp;

	ASSERT(xfs_isilocked(dp, XFS_ILOCK_SHARED | XFS_ILOCK_EXCL));

	/*
	 * Decide on what work routines to call based on the inode size.
	 */
	if (!xfs_inode_hasattr(dp))
		return 0;
	else if (dp->i_d.di_aformat == XFS_DINODE_FMT_LOCAL)
		return xfs_attr_shortform_list(context);
	else if (xfs_bmap_one_block(dp, XFS_ATTR_FORK))
		return xfs_attr_leaf_list(context);
	return xfs_attr_node_list(context);
}

int
xfs_attr_list_int(
	xfs_attr_list_context_t *context)
{
	int error;
	xfs_inode_t *dp = context->dp;
	uint		lock_mode;

	XFS_STATS_INC(dp->i_mount, xs_attr_list);

	if (XFS_FORCED_SHUTDOWN(dp->i_mount))
		return -EIO;

	lock_mode = xfs_ilock_attr_map_shared(dp);
	error = xfs_attr_list_int_ilocked(context);
	xfs_iunlock(dp, lock_mode);
	return error;
}

#define	ATTR_ENTBASESIZE		/* minimum bytes used by an attr */ \
	(((struct attrlist_ent *) 0)->a_name - (char *) 0)
#define	ATTR_ENTSIZE(namelen)		/* actual bytes used by an attr */ \
	((ATTR_ENTBASESIZE + (namelen) + 1 + sizeof(uint32_t)-1) \
	 & ~(sizeof(uint32_t)-1))

/*
 * Format an attribute and copy it out to the user's buffer.
 * Take care to check values and protect against them changing later,
 * we may be reading them directly out of a user buffer.
 */
STATIC void
xfs_attr_put_listent(
	xfs_attr_list_context_t *context,
	int		flags,
	unsigned char	*name,
	int		namelen,
	int		valuelen)
{
	struct attrlist *alist = (struct attrlist *)context->alist;
	attrlist_ent_t *aep;
	int arraytop;

	ASSERT(!context->seen_enough);
	ASSERT(!(context->flags & ATTR_KERNOVAL));
	ASSERT(context->count >= 0);
	ASSERT(context->count < (ATTR_MAX_VALUELEN/8));
	ASSERT(context->firstu >= sizeof(*alist));
	ASSERT(context->firstu <= context->bufsize);

	/*
	 * Only list entries in the right namespace.
	 */
	if (((context->flags & ATTR_SECURE) == 0) !=
	    ((flags & XFS_ATTR_SECURE) == 0))
		return;
	if (((context->flags & ATTR_ROOT) == 0) !=
	    ((flags & XFS_ATTR_ROOT) == 0))
		return;

	arraytop = sizeof(*alist) +
			context->count * sizeof(alist->al_offset[0]);
	context->firstu -= ATTR_ENTSIZE(namelen);
	if (context->firstu < arraytop) {
		trace_xfs_attr_list_full(context);
		alist->al_more = 1;
		context->seen_enough = 1;
		return;
	}

	aep = (attrlist_ent_t *)&context->alist[context->firstu];
	aep->a_valuelen = valuelen;
	memcpy(aep->a_name, name, namelen);
	aep->a_name[namelen] = 0;
	alist->al_offset[context->count++] = context->firstu;
	alist->al_count = context->count;
	trace_xfs_attr_list_add(context);
	return;
}

/*
 * Generate a list of extended attribute names and optionally
 * also value lengths.  Positive return value follows the XFS
 * convention of being an error, zero or negative return code
 * is the length of the buffer returned (negated), indicating
 * success.
 */
int
xfs_attr_list(
	xfs_inode_t	*dp,
	char		*buffer,
	int		bufsize,
	int		flags,
	attrlist_cursor_kern_t *cursor)
{
	xfs_attr_list_context_t context;
	struct attrlist *alist;
	int error;

	/*
	 * Validate the cursor.
	 */
	if (cursor->pad1 || cursor->pad2)
		return -EINVAL;
	if ((cursor->initted == 0) &&
	    (cursor->hashval || cursor->blkno || cursor->offset))
		return -EINVAL;

	/* Only internal consumers can retrieve incomplete attrs. */
	if (flags & ATTR_INCOMPLETE)
		return -EINVAL;

	/*
	 * Check for a properly aligned buffer.
	 */
	if (((long)buffer) & (sizeof(int)-1))
		return -EFAULT;
	if (flags & ATTR_KERNOVAL)
		bufsize = 0;

	/*
	 * Initialize the output buffer.
	 */
	memset(&context, 0, sizeof(context));
	context.dp = dp;
	context.cursor = cursor;
	context.resynch = 1;
	context.flags = flags;
	context.alist = buffer;
	context.bufsize = (bufsize & ~(sizeof(int)-1));  /* align */
	context.firstu = context.bufsize;
	context.put_listent = xfs_attr_put_listent;

	alist = (struct attrlist *)context.alist;
	alist->al_count = 0;
	alist->al_more = 0;
	alist->al_offset[0] = context.bufsize;

	error = xfs_attr_list_int(&context);
	ASSERT(error <= 0);
	return error;
}
Commit	Line	Data
0b61f8a4	1	// SPDX-License-Identifier: GPL-2.0
abec5f2b DC	2	/*
	3	* Copyright (c) 2000-2005 Silicon Graphics, Inc.
	4	* Copyright (c) 2013 Red Hat, Inc.
	5	* All Rights Reserved.
abec5f2b DC	6	*/
	7	#include "xfs.h"
	8	#include "xfs_fs.h"
a4fbe6ab	9	#include "xfs_format.h"
239880ef DC	10	#include "xfs_log_format.h"
239880ef DC	11	#include "xfs_trans_resv.h"
abec5f2b	12	#include "xfs_bit.h"
abec5f2b	13	#include "xfs_mount.h"
57062787	14	#include "xfs_da_format.h"
abec5f2b	15	#include "xfs_da_btree.h"
abec5f2b	16	#include "xfs_inode.h"
239880ef	17	#include "xfs_trans.h"
abec5f2b DC	18	#include "xfs_inode_item.h"
	19	#include "xfs_bmap.h"
	20	#include "xfs_attr.h"
a4fbe6ab DC	21	#include "xfs_attr_sf.h"
a4fbe6ab DC	22	#include "xfs_attr_remote.h"
abec5f2b DC	23	#include "xfs_attr_leaf.h"
	24	#include "xfs_error.h"
	25	#include "xfs_trace.h"
	26	#include "xfs_buf_item.h"
	27	#include "xfs_cksum.h"
4bceb18f	28	#include "xfs_dir2.h"
abec5f2b DC	29
	30	STATIC int
	31	xfs_attr_shortform_compare(const void a, const void b)
	32	{
	33	xfs_attr_sf_sort_t sa, sb;
	34
	35	sa = (xfs_attr_sf_sort_t *)a;
	36	sb = (xfs_attr_sf_sort_t *)b;
	37	if (sa->hash < sb->hash) {
d99831ff	38	return -1;
abec5f2b	39	} else if (sa->hash > sb->hash) {
d99831ff	40	return 1;
abec5f2b	41	} else {
d99831ff	42	return sa->entno - sb->entno;
abec5f2b DC	43	}
	44	}
	45
	46	#define XFS_ISRESET_CURSOR(cursor) \
	47	(!((cursor)->initted) && !((cursor)->hashval) && \
	48	!((cursor)->blkno) && !((cursor)->offset))
	49	/*
	50	* Copy out entries of shortform attribute lists for attr_list().
	51	* Shortform attribute lists are not stored in hashval sorted order.
	52	* If the output buffer is not large enough to hold them all, then we
	53	* we have to calculate each entries' hashvalue and sort them before
	54	* we can begin returning them to the user.
	55	*/
0d5a75e9	56	static int
abec5f2b DC	57	xfs_attr_shortform_list(xfs_attr_list_context_t *context)
	58	{
	59	attrlist_cursor_kern_t *cursor;
	60	xfs_attr_sf_sort_t sbuf, sbp;
	61	xfs_attr_shortform_t *sf;
	62	xfs_attr_sf_entry_t *sfe;
	63	xfs_inode_t *dp;
	64	int sbsize, nsbuf, count, i;
abec5f2b DC	65
	66	ASSERT(context != NULL);
	67	dp = context->dp;
	68	ASSERT(dp != NULL);
	69	ASSERT(dp->i_afp != NULL);
	70	sf = (xfs_attr_shortform_t *)dp->i_afp->if_u1.if_data;
	71	ASSERT(sf != NULL);
	72	if (!sf->hdr.count)
d99831ff	73	return 0;
abec5f2b DC	74	cursor = context->cursor;
	75	ASSERT(cursor != NULL);
	76
	77	trace_xfs_attr_list_sf(context);
	78
	79	/*
	80	* If the buffer is large enough and the cursor is at the start,
	81	* do not bother with sorting since we will return everything in
	82	* one buffer and another call using the cursor won't need to be
	83	* made.
	84	* Note the generous fudge factor of 16 overhead bytes per entry.
	85	* If bufsize is zero then put_listent must be a search function
	86	* and can just scan through what we have.
	87	*/
	88	if (context->bufsize == 0 \|\|
	89	(XFS_ISRESET_CURSOR(cursor) &&
c2b6e159	90	(dp->i_afp->if_bytes + sf->hdr.count * 16) < context->bufsize)) {
abec5f2b	91	for (i = 0, sfe = &sf->list[0]; i < sf->hdr.count; i++) {
f7a136ae ES	92	context->put_listent(context,
	93	sfe->flags,
	94	sfe->nameval,
	95	(int)sfe->namelen,
	96	(int)sfe->valuelen);
abec5f2b DC	97	/*
	98	* Either search callback finished early or
	99	* didn't fit it all in the buffer after all.
	100	*/
	101	if (context->seen_enough)
	102	break;
abec5f2b DC	103	sfe = XFS_ATTR_SF_NEXTENTRY(sfe);
	104	}
	105	trace_xfs_attr_list_sf_all(context);
d99831ff	106	return 0;
abec5f2b DC	107	}
	108
	109	/* do no more for a search callback */
	110	if (context->bufsize == 0)
	111	return 0;
	112
	113	/*
	114	* It didn't all fit, so we have to sort everything on hashval.
	115	*/
	116	sbsize = sf->hdr.count * sizeof(*sbuf);
	117	sbp = sbuf = kmem_alloc(sbsize, KM_SLEEP \| KM_NOFS);
	118
	119	/*
	120	* Scan the attribute list for the rest of the entries, storing
	121	* the relevant info from only those that match into a buffer.
	122	*/
	123	nsbuf = 0;
	124	for (i = 0, sfe = &sf->list[0]; i < sf->hdr.count; i++) {
	125	if (unlikely(
	126	((char )sfe < (char )sf) \|\|
	127	((char )sfe >= ((char )sf + dp->i_afp->if_bytes)))) {
	128	XFS_CORRUPTION_ERROR("xfs_attr_shortform_list",
	129	XFS_ERRLEVEL_LOW,
2551a530 DW	130	context->dp->i_mount, sfe,
2551a530 DW	131	sizeof(*sfe));
abec5f2b	132	kmem_free(sbuf);
2451337d	133	return -EFSCORRUPTED;
abec5f2b DC	134	}
	135
	136	sbp->entno = i;
	137	sbp->hash = xfs_da_hashname(sfe->nameval, sfe->namelen);
	138	sbp->name = sfe->nameval;
	139	sbp->namelen = sfe->namelen;
	140	/* These are bytes, and both on-disk, don't endian-flip */
	141	sbp->valuelen = sfe->valuelen;
	142	sbp->flags = sfe->flags;
	143	sfe = XFS_ATTR_SF_NEXTENTRY(sfe);
	144	sbp++;
	145	nsbuf++;
	146	}
	147
	148	/*
	149	* Sort the entries on hash then entno.
	150	*/
	151	xfs_sort(sbuf, nsbuf, sizeof(*sbuf), xfs_attr_shortform_compare);
	152
	153	/*
	154	* Re-find our place IN THE SORTED LIST.
	155	*/
	156	count = 0;
	157	cursor->initted = 1;
	158	cursor->blkno = 0;
	159	for (sbp = sbuf, i = 0; i < nsbuf; i++, sbp++) {
	160	if (sbp->hash == cursor->hashval) {
	161	if (cursor->offset == count) {
	162	break;
	163	}
	164	count++;
	165	} else if (sbp->hash > cursor->hashval) {
	166	break;
	167	}
	168	}
	169	if (i == nsbuf) {
	170	kmem_free(sbuf);
d99831ff	171	return 0;
abec5f2b DC	172	}
	173
	174	/*
	175	* Loop putting entries into the user buffer.
	176	*/
	177	for ( ; i < nsbuf; i++, sbp++) {
	178	if (cursor->hashval != sbp->hash) {
	179	cursor->hashval = sbp->hash;
	180	cursor->offset = 0;
	181	}
f7a136ae ES	182	context->put_listent(context,
	183	sbp->flags,
	184	sbp->name,
	185	sbp->namelen,
	186	sbp->valuelen);
abec5f2b DC	187	if (context->seen_enough)
	188	break;
	189	cursor->offset++;
	190	}
	191
	192	kmem_free(sbuf);
d99831ff	193	return 0;
abec5f2b DC	194	}
abec5f2b DC	195
bdaac93f DW	196	/*
	197	* We didn't find the block & hash mentioned in the cursor state, so
	198	* walk down the attr btree looking for the hash.
	199	*/
abec5f2b	200	STATIC int
bdaac93f DW	201	xfs_attr_node_list_lookup(
	202	struct xfs_attr_list_context *context,
	203	struct attrlist_cursor_kern *cursor,
	204	struct xfs_buf **pbp)
abec5f2b	205	{
bdaac93f DW	206	struct xfs_da3_icnode_hdr nodehdr;
	207	struct xfs_da_intnode *node;
	208	struct xfs_da_node_entry *btree;
	209	struct xfs_inode *dp = context->dp;
	210	struct xfs_mount *mp = dp->i_mount;
	211	struct xfs_trans *tp = context->tp;
	212	struct xfs_buf *bp;
	213	int i;
	214	int error = 0;
8210f4dd	215	unsigned int expected_level = 0;
bdaac93f DW	216	uint16_t magic;
	217
	218	ASSERT(*pbp == NULL);
	219	cursor->blkno = 0;
	220	for (;;) {
	221	error = xfs_da3_node_read(tp, dp, cursor->blkno, -1, &bp,
	222	XFS_ATTR_FORK);
	223	if (error)
	224	return error;
	225	node = bp->b_addr;
	226	magic = be16_to_cpu(node->hdr.info.magic);
	227	if (magic == XFS_ATTR_LEAF_MAGIC \|\|
	228	magic == XFS_ATTR3_LEAF_MAGIC)
	229	break;
	230	if (magic != XFS_DA_NODE_MAGIC &&
	231	magic != XFS_DA3_NODE_MAGIC) {
	232	XFS_CORRUPTION_ERROR(__func__, XFS_ERRLEVEL_LOW, mp,
2551a530	233	node, sizeof(*node));
bdaac93f DW	234	goto out_corruptbuf;
	235	}
	236
	237	dp->d_ops->node_hdr_from_disk(&nodehdr, node);
	238
8210f4dd DW	239	/* Tree taller than we can handle; bail out! */
	240	if (nodehdr.level >= XFS_DA_NODE_MAXDEPTH)
	241	goto out_corruptbuf;
	242
	243	/* Check the level from the root node. */
	244	if (cursor->blkno == 0)
	245	expected_level = nodehdr.level - 1;
	246	else if (expected_level != nodehdr.level)
	247	goto out_corruptbuf;
	248	else
	249	expected_level--;
	250
bdaac93f DW	251	btree = dp->d_ops->node_tree_p(node);
	252	for (i = 0; i < nodehdr.count; btree++, i++) {
	253	if (cursor->hashval <= be32_to_cpu(btree->hashval)) {
	254	cursor->blkno = be32_to_cpu(btree->before);
	255	trace_xfs_attr_list_node_descend(context,
	256	btree);
	257	break;
	258	}
	259	}
	260	xfs_trans_brelse(tp, bp);
	261
	262	if (i == nodehdr.count)
	263	return 0;
8210f4dd DW	264
	265	/* We can't point back to the root. */
	266	if (cursor->blkno == 0)
	267	return -EFSCORRUPTED;
bdaac93f DW	268	}
bdaac93f DW	269
8210f4dd DW	270	if (expected_level != 0)
	271	goto out_corruptbuf;
	272
bdaac93f DW	273	*pbp = bp;
	274	return 0;
	275
	276	out_corruptbuf:
	277	xfs_trans_brelse(tp, bp);
	278	return -EFSCORRUPTED;
	279	}
	280
	281	STATIC int
	282	xfs_attr_node_list(
	283	struct xfs_attr_list_context *context)
	284	{
	285	struct xfs_attr3_icleaf_hdr leafhdr;
	286	struct attrlist_cursor_kern *cursor;
	287	struct xfs_attr_leafblock *leaf;
	288	struct xfs_da_intnode *node;
	289	struct xfs_buf *bp;
	290	struct xfs_inode *dp = context->dp;
	291	struct xfs_mount *mp = dp->i_mount;
	292	int error;
abec5f2b DC	293
	294	trace_xfs_attr_node_list(context);
	295
	296	cursor = context->cursor;
	297	cursor->initted = 1;
	298
	299	/*
	300	* Do all sorts of validation on the passed-in cursor structure.
	301	* If anything is amiss, ignore the cursor and look up the hashval
	302	* starting from the btree root.
	303	*/
	304	bp = NULL;
	305	if (cursor->blkno > 0) {
ad017f65	306	error = xfs_da3_node_read(context->tp, dp, cursor->blkno, -1,
abec5f2b	307	&bp, XFS_ATTR_FORK);
2451337d	308	if ((error != 0) && (error != -EFSCORRUPTED))
d99831ff	309	return error;
abec5f2b DC	310	if (bp) {
	311	struct xfs_attr_leaf_entry *entries;
	312
	313	node = bp->b_addr;
	314	switch (be16_to_cpu(node->hdr.info.magic)) {
	315	case XFS_DA_NODE_MAGIC:
	316	case XFS_DA3_NODE_MAGIC:
	317	trace_xfs_attr_list_wrong_blk(context);
ad017f65	318	xfs_trans_brelse(context->tp, bp);
abec5f2b DC	319	bp = NULL;
	320	break;
	321	case XFS_ATTR_LEAF_MAGIC:
	322	case XFS_ATTR3_LEAF_MAGIC:
	323	leaf = bp->b_addr;
2f661241 BF	324	xfs_attr3_leaf_hdr_from_disk(mp->m_attr_geo,
2f661241 BF	325	&leafhdr, leaf);
abec5f2b DC	326	entries = xfs_attr3_leaf_entryp(leaf);
	327	if (cursor->hashval > be32_to_cpu(
	328	entries[leafhdr.count - 1].hashval)) {
	329	trace_xfs_attr_list_wrong_blk(context);
ad017f65	330	xfs_trans_brelse(context->tp, bp);
abec5f2b DC	331	bp = NULL;
	332	} else if (cursor->hashval <= be32_to_cpu(
	333	entries[0].hashval)) {
	334	trace_xfs_attr_list_wrong_blk(context);
ad017f65	335	xfs_trans_brelse(context->tp, bp);
abec5f2b DC	336	bp = NULL;
	337	}
	338	break;
	339	default:
	340	trace_xfs_attr_list_wrong_blk(context);
ad017f65	341	xfs_trans_brelse(context->tp, bp);
abec5f2b DC	342	bp = NULL;
	343	}
	344	}
	345	}
	346
	347	/*
	348	* We did not find what we expected given the cursor's contents,
	349	* so we start from the top and work down based on the hash value.
	350	* Note that start of node block is same as start of leaf block.
	351	*/
	352	if (bp == NULL) {
bdaac93f DW	353	error = xfs_attr_node_list_lookup(context, cursor, &bp);
	354	if (error \|\| !bp)
	355	return error;
abec5f2b DC	356	}
	357	ASSERT(bp != NULL);
	358
	359	/*
	360	* Roll upward through the blocks, processing each leaf block in
	361	* order. As long as there is space in the result buffer, keep
	362	* adding the information.
	363	*/
	364	for (;;) {
	365	leaf = bp->b_addr;
f7a136ae	366	xfs_attr3_leaf_list_int(bp, context);
2f661241	367	xfs_attr3_leaf_hdr_from_disk(mp->m_attr_geo, &leafhdr, leaf);
abec5f2b DC	368	if (context->seen_enough \|\| leafhdr.forw == 0)
	369	break;
	370	cursor->blkno = leafhdr.forw;
ad017f65 DW	371	xfs_trans_brelse(context->tp, bp);
ad017f65 DW	372	error = xfs_attr3_leaf_read(context->tp, dp, cursor->blkno, -1, &bp);
abec5f2b DC	373	if (error)
	374	return error;
	375	}
ad017f65	376	xfs_trans_brelse(context->tp, bp);
abec5f2b DC	377	return 0;
	378	}
	379
	380	/*
	381	* Copy out attribute list entries for attr_list(), for leaf attribute lists.
	382	*/
f7a136ae	383	void
abec5f2b DC	384	xfs_attr3_leaf_list_int(
	385	struct xfs_buf *bp,
	386	struct xfs_attr_list_context *context)
	387	{
	388	struct attrlist_cursor_kern *cursor;
	389	struct xfs_attr_leafblock *leaf;
	390	struct xfs_attr3_icleaf_hdr ichdr;
	391	struct xfs_attr_leaf_entry *entries;
	392	struct xfs_attr_leaf_entry *entry;
abec5f2b	393	int i;
2f661241	394	struct xfs_mount *mp = context->dp->i_mount;
abec5f2b DC	395
	396	trace_xfs_attr_list_leaf(context);
	397
	398	leaf = bp->b_addr;
2f661241	399	xfs_attr3_leaf_hdr_from_disk(mp->m_attr_geo, &ichdr, leaf);
abec5f2b DC	400	entries = xfs_attr3_leaf_entryp(leaf);
	401
	402	cursor = context->cursor;
	403	cursor->initted = 1;
	404
	405	/*
	406	* Re-find our place in the leaf block if this is a new syscall.
	407	*/
	408	if (context->resynch) {
	409	entry = &entries[0];
	410	for (i = 0; i < ichdr.count; entry++, i++) {
	411	if (be32_to_cpu(entry->hashval) == cursor->hashval) {
	412	if (cursor->offset == context->dupcnt) {
	413	context->dupcnt = 0;
	414	break;
	415	}
	416	context->dupcnt++;
	417	} else if (be32_to_cpu(entry->hashval) >
	418	cursor->hashval) {
	419	context->dupcnt = 0;
	420	break;
	421	}
	422	}
	423	if (i == ichdr.count) {
	424	trace_xfs_attr_list_notfound(context);
f7a136ae	425	return;
abec5f2b DC	426	}
	427	} else {
	428	entry = &entries[0];
	429	i = 0;
	430	}
	431	context->resynch = 0;
	432
	433	/*
	434	* We have found our place, start copying out the new attributes.
	435	*/
abec5f2b	436	for (; i < ichdr.count; entry++, i++) {
3ab3ffca ES	437	char *name;
	438	int namelen, valuelen;
	439
abec5f2b DC	440	if (be32_to_cpu(entry->hashval) != cursor->hashval) {
	441	cursor->hashval = be32_to_cpu(entry->hashval);
	442	cursor->offset = 0;
	443	}
	444
eec0482e DW	445	if ((entry->flags & XFS_ATTR_INCOMPLETE) &&
eec0482e DW	446	!(context->flags & ATTR_INCOMPLETE))
abec5f2b DC	447	continue; /* skip incomplete entries */
	448
	449	if (entry->flags & XFS_ATTR_LOCAL) {
3ab3ffca	450	xfs_attr_leaf_name_local_t *name_loc;
abec5f2b	451
3ab3ffca ES	452	name_loc = xfs_attr3_leaf_name_local(leaf, i);
	453	name = name_loc->nameval;
	454	namelen = name_loc->namelen;
	455	valuelen = be16_to_cpu(name_loc->valuelen);
	456	} else {
	457	xfs_attr_leaf_name_remote_t *name_rmt;
abec5f2b	458
3ab3ffca ES	459	name_rmt = xfs_attr3_leaf_name_remote(leaf, i);
	460	name = name_rmt->name;
	461	namelen = name_rmt->namelen;
	462	valuelen = be32_to_cpu(name_rmt->valuelen);
abec5f2b	463	}
3ab3ffca	464
f7a136ae	465	context->put_listent(context, entry->flags,
3ab3ffca	466	name, namelen, valuelen);
abec5f2b DC	467	if (context->seen_enough)
	468	break;
	469	cursor->offset++;
	470	}
	471	trace_xfs_attr_list_leaf_end(context);
f7a136ae	472	return;
abec5f2b DC	473	}
	474
	475	/*
	476	* Copy out attribute entries for attr_list(), for leaf attribute lists.
	477	*/
	478	STATIC int
	479	xfs_attr_leaf_list(xfs_attr_list_context_t *context)
	480	{
	481	int error;
	482	struct xfs_buf *bp;
	483
	484	trace_xfs_attr_leaf_list(context);
	485
	486	context->cursor->blkno = 0;
ad017f65	487	error = xfs_attr3_leaf_read(context->tp, context->dp, 0, -1, &bp);
abec5f2b	488	if (error)
b474c7ae	489	return error;
abec5f2b	490
f7a136ae	491	xfs_attr3_leaf_list_int(bp, context);
ad017f65	492	xfs_trans_brelse(context->tp, bp);
f7a136ae	493	return 0;
abec5f2b DC	494	}
abec5f2b DC	495
ad017f65 DW	496	int
	497	xfs_attr_list_int_ilocked(
	498	struct xfs_attr_list_context *context)
	499	{
	500	struct xfs_inode *dp = context->dp;
	501
5af7777e CH	502	ASSERT(xfs_isilocked(dp, XFS_ILOCK_SHARED \| XFS_ILOCK_EXCL));
5af7777e CH	503
ad017f65 DW	504	/*
	505	* Decide on what work routines to call based on the inode size.
	506	*/
	507	if (!xfs_inode_hasattr(dp))
	508	return 0;
	509	else if (dp->i_d.di_aformat == XFS_DINODE_FMT_LOCAL)
	510	return xfs_attr_shortform_list(context);
	511	else if (xfs_bmap_one_block(dp, XFS_ATTR_FORK))
	512	return xfs_attr_leaf_list(context);
	513	return xfs_attr_node_list(context);
	514	}
	515
abec5f2b DC	516	int
	517	xfs_attr_list_int(
	518	xfs_attr_list_context_t *context)
	519	{
	520	int error;
	521	xfs_inode_t *dp = context->dp;
568d994e	522	uint lock_mode;
abec5f2b	523
ff6d6af2	524	XFS_STATS_INC(dp->i_mount, xs_attr_list);
abec5f2b DC	525
abec5f2b DC	526	if (XFS_FORCED_SHUTDOWN(dp->i_mount))
2451337d	527	return -EIO;
abec5f2b	528
568d994e	529	lock_mode = xfs_ilock_attr_map_shared(dp);
ad017f65	530	error = xfs_attr_list_int_ilocked(context);
568d994e	531	xfs_iunlock(dp, lock_mode);
abec5f2b DC	532	return error;
	533	}
	534
	535	#define ATTR_ENTBASESIZE /* minimum bytes used by an attr */ \
	536	(((struct attrlist_ent ) 0)->a_name - (char ) 0)
	537	#define ATTR_ENTSIZE(namelen) /* actual bytes used by an attr */ \
65a7935d DW	538	((ATTR_ENTBASESIZE + (namelen) + 1 + sizeof(uint32_t)-1) \
65a7935d DW	539	& ~(sizeof(uint32_t)-1))
abec5f2b DC	540
	541	/*
	542	* Format an attribute and copy it out to the user's buffer.
	543	* Take care to check values and protect against them changing later,
	544	* we may be reading them directly out of a user buffer.
	545	*/
f7a136ae	546	STATIC void
abec5f2b DC	547	xfs_attr_put_listent(
	548	xfs_attr_list_context_t *context,
	549	int flags,
	550	unsigned char *name,
	551	int namelen,
e5bd12bf	552	int valuelen)
abec5f2b DC	553	{
	554	struct attrlist alist = (struct attrlist )context->alist;
	555	attrlist_ent_t *aep;
	556	int arraytop;
	557
3b50086f	558	ASSERT(!context->seen_enough);
abec5f2b DC	559	ASSERT(!(context->flags & ATTR_KERNOVAL));
	560	ASSERT(context->count >= 0);
	561	ASSERT(context->count < (ATTR_MAX_VALUELEN/8));
	562	ASSERT(context->firstu >= sizeof(*alist));
	563	ASSERT(context->firstu <= context->bufsize);
	564
	565	/*
	566	* Only list entries in the right namespace.
	567	*/
	568	if (((context->flags & ATTR_SECURE) == 0) !=
	569	((flags & XFS_ATTR_SECURE) == 0))
f7a136ae	570	return;
abec5f2b DC	571	if (((context->flags & ATTR_ROOT) == 0) !=
abec5f2b DC	572	((flags & XFS_ATTR_ROOT) == 0))
f7a136ae	573	return;
abec5f2b DC	574
	575	arraytop = sizeof(*alist) +
	576	context->count * sizeof(alist->al_offset[0]);
	577	context->firstu -= ATTR_ENTSIZE(namelen);
	578	if (context->firstu < arraytop) {
	579	trace_xfs_attr_list_full(context);
	580	alist->al_more = 1;
	581	context->seen_enough = 1;
f7a136ae	582	return;
abec5f2b DC	583	}
	584
	585	aep = (attrlist_ent_t *)&context->alist[context->firstu];
	586	aep->a_valuelen = valuelen;
	587	memcpy(aep->a_name, name, namelen);
	588	aep->a_name[namelen] = 0;
	589	alist->al_offset[context->count++] = context->firstu;
	590	alist->al_count = context->count;
	591	trace_xfs_attr_list_add(context);
f7a136ae	592	return;
abec5f2b DC	593	}
	594
	595	/*
	596	* Generate a list of extended attribute names and optionally
	597	* also value lengths. Positive return value follows the XFS
	598	* convention of being an error, zero or negative return code
	599	* is the length of the buffer returned (negated), indicating
	600	* success.
	601	*/
	602	int
	603	xfs_attr_list(
	604	xfs_inode_t *dp,
	605	char *buffer,
	606	int bufsize,
	607	int flags,
	608	attrlist_cursor_kern_t *cursor)
	609	{
	610	xfs_attr_list_context_t context;
	611	struct attrlist *alist;
	612	int error;
	613
	614	/*
	615	* Validate the cursor.
	616	*/
	617	if (cursor->pad1 \|\| cursor->pad2)
2451337d	618	return -EINVAL;
abec5f2b DC	619	if ((cursor->initted == 0) &&
abec5f2b DC	620	(cursor->hashval \|\| cursor->blkno \|\| cursor->offset))
2451337d	621	return -EINVAL;
abec5f2b	622
eec0482e DW	623	/* Only internal consumers can retrieve incomplete attrs. */
	624	if (flags & ATTR_INCOMPLETE)
	625	return -EINVAL;
	626
abec5f2b DC	627	/*
	628	* Check for a properly aligned buffer.
	629	*/
	630	if (((long)buffer) & (sizeof(int)-1))
2451337d	631	return -EFAULT;
abec5f2b DC	632	if (flags & ATTR_KERNOVAL)
	633	bufsize = 0;
	634
	635	/*
	636	* Initialize the output buffer.
	637	*/
	638	memset(&context, 0, sizeof(context));
	639	context.dp = dp;
	640	context.cursor = cursor;
	641	context.resynch = 1;
	642	context.flags = flags;
	643	context.alist = buffer;
	644	context.bufsize = (bufsize & ~(sizeof(int)-1)); /* align */
	645	context.firstu = context.bufsize;
	646	context.put_listent = xfs_attr_put_listent;
	647
	648	alist = (struct attrlist *)context.alist;
	649	alist->al_count = 0;
	650	alist->al_more = 0;
	651	alist->al_offset[0] = context.bufsize;
	652
	653	error = xfs_attr_list_int(&context);
2451337d	654	ASSERT(error <= 0);
abec5f2b DC	655	return error;
abec5f2b DC	656	}