[linux-2.6-block.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_shared.h"
#include "xfs_format.h"
#include "xfs_log_format.h"
#include "xfs_trans_resv.h"
#include "xfs_mount.h"
#include "xfs_da_format.h"
#include "xfs_inode.h"
#include "xfs_trans.h"
#include "xfs_bmap.h"
#include "xfs_da_btree.h"
#include "xfs_attr.h"
#include "xfs_attr_sf.h"
#include "xfs_attr_leaf.h"
#include "xfs_error.h"
#include "xfs_trace.h"
#include "xfs_dir2.h"
#include "xfs_health.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 have to calculate each entries' hashvalue and sort them before
 * we can begin returning them to the user.
 */
static int
xfs_attr_shortform_list(
	struct xfs_attr_list_context	*context)
{
	struct xfs_attrlist_cursor_kern	*cursor = &context->cursor;
	struct xfs_inode		*dp = context->dp;
	struct xfs_attr_sf_sort		*sbuf, *sbp;
	struct xfs_attr_sf_hdr		*sf = dp->i_af.if_data;
	struct xfs_attr_sf_entry	*sfe;
	int				sbsize, nsbuf, count, i;
	int				error = 0;

	ASSERT(sf != NULL);
	if (!sf->count)
		return 0;

	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_af.if_bytes + sf->count * 16) < context->bufsize)) {
		for (i = 0, sfe = xfs_attr_sf_firstentry(sf); i < sf->count; i++) {
			if (XFS_IS_CORRUPT(context->dp->i_mount,
					   !xfs_attr_namecheck(sfe->flags,
							       sfe->nameval,
							       sfe->namelen))) {
				xfs_dirattr_mark_sick(context->dp, XFS_ATTR_FORK);
				return -EFSCORRUPTED;
			}
			context->put_listent(context,
					     sfe->flags,
					     sfe->nameval,
					     (int)sfe->namelen,
					     &sfe->nameval[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->count * sizeof(*sbuf);
	sbp = sbuf = kmalloc(sbsize,
			GFP_KERNEL | __GFP_NOLOCKDEP | __GFP_NOFAIL);

	/*
	 * 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 = xfs_attr_sf_firstentry(sf); i < sf->count; i++) {
		if (unlikely(
		    ((char *)sfe < (char *)sf) ||
		    ((char *)sfe >= ((char *)sf + dp->i_af.if_bytes)) ||
		    !xfs_attr_check_namespace(sfe->flags))) {
			XFS_CORRUPTION_ERROR("xfs_attr_shortform_list",
					     XFS_ERRLEVEL_LOW,
					     context->dp->i_mount, sfe,
					     sizeof(*sfe));
			kfree(sbuf);
			xfs_dirattr_mark_sick(dp, XFS_ATTR_FORK);
			return -EFSCORRUPTED;
		}

		sbp->entno = i;
		sbp->name = sfe->nameval;
		sbp->namelen = sfe->namelen;
		/* These are bytes, and both on-disk, don't endian-flip */
		sbp->value = &sfe->nameval[sfe->namelen];
		sbp->valuelen = sfe->valuelen;
		sbp->flags = sfe->flags;
		sbp->hash = xfs_attr_hashval(dp->i_mount, sfe->flags,
					     sfe->nameval, sfe->namelen,
					     sfe->nameval + sfe->namelen,
					     sfe->valuelen);
		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)
		goto out;

	/*
	 * Loop putting entries into the user buffer.
	 */
	for ( ; i < nsbuf; i++, sbp++) {
		if (cursor->hashval != sbp->hash) {
			cursor->hashval = sbp->hash;
			cursor->offset = 0;
		}
		if (XFS_IS_CORRUPT(context->dp->i_mount,
				   !xfs_attr_namecheck(sbp->flags, sbp->name,
						       sbp->namelen))) {
			xfs_dirattr_mark_sick(context->dp, XFS_ATTR_FORK);
			error = -EFSCORRUPTED;
			goto out;
		}
		context->put_listent(context,
				     sbp->flags,
				     sbp->name,
				     sbp->namelen,
				     sbp->value,
				     sbp->valuelen);
		if (context->seen_enough)
			break;
		cursor->offset++;
	}
out:
	kfree(sbuf);
	return error;
}

/*
 * 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 xfs_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;
	xfs_failaddr_t			fa;
	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, &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;
		}

		fa = xfs_da3_node_header_check(bp, dp->i_ino);
		if (fa)
			goto out_corruptbuf;

		xfs_da3_node_hdr_from_disk(mp, &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 = nodehdr.btree;
		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 (XFS_IS_CORRUPT(mp, cursor->blkno == 0)) {
			xfs_dirattr_mark_sick(dp, XFS_ATTR_FORK);
			return -EFSCORRUPTED;
		}
	}

	fa = xfs_attr3_leaf_header_check(bp, dp->i_ino);
	if (fa) {
		__xfs_buf_mark_corrupt(bp, fa);
		goto out_releasebuf;
	}

	if (expected_level != 0)
		goto out_corruptbuf;

	*pbp = bp;
	return 0;

out_corruptbuf:
	xfs_buf_mark_corrupt(bp);
out_releasebuf:
	xfs_trans_brelse(tp, bp);
	xfs_dirattr_mark_sick(dp, XFS_ATTR_FORK);
	return -EFSCORRUPTED;
}

STATIC int
xfs_attr_node_list(
	struct xfs_attr_list_context	*context)
{
	struct xfs_attrlist_cursor_kern	*cursor = &context->cursor;
	struct xfs_attr3_icleaf_hdr	leafhdr;
	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;
	xfs_failaddr_t			fa;
	int				error = 0;

	trace_xfs_attr_node_list(context);

	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) {
		struct xfs_attr_leaf_entry *entries;

		error = xfs_da3_node_read(context->tp, dp, cursor->blkno, &bp,
				XFS_ATTR_FORK);
		if (xfs_metadata_is_sick(error))
			xfs_dirattr_mark_sick(dp, XFS_ATTR_FORK);
		if (error != 0 && error != -EFSCORRUPTED)
			return error;
		if (!bp)
			goto need_lookup;

		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);
			fa = xfs_da3_node_header_check(bp, dp->i_ino);
			if (fa) {
				__xfs_buf_mark_corrupt(bp, fa);
				xfs_dirattr_mark_sick(dp, XFS_ATTR_FORK);
			}
			xfs_trans_brelse(context->tp, bp);
			bp = NULL;
			break;
		case XFS_ATTR_LEAF_MAGIC:
		case XFS_ATTR3_LEAF_MAGIC:
			leaf = bp->b_addr;
			fa = xfs_attr3_leaf_header_check(bp, dp->i_ino);
			if (fa) {
				__xfs_buf_mark_corrupt(bp, fa);
				xfs_trans_brelse(context->tp, bp);
				xfs_dirattr_mark_sick(dp, XFS_ATTR_FORK);
				bp = NULL;
				break;
			}
			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) {
need_lookup:
		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;
		error = xfs_attr3_leaf_list_int(bp, context);
		if (error)
			break;
		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, dp->i_ino,
				cursor->blkno, &bp);
		if (error)
			return error;
	}
	xfs_trans_brelse(context->tp, bp);
	return error;
}

/*
 * Copy out attribute list entries for attr_list(), for leaf attribute lists.
 */
int
xfs_attr3_leaf_list_int(
	struct xfs_buf			*bp,
	struct xfs_attr_list_context	*context)
{
	struct xfs_attrlist_cursor_kern	*cursor = &context->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->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 0;
		}
	} 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;
		void *value;
		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->allow_incomplete)
			continue;

		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;
			value = &name_loc->nameval[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;
			value = NULL;
			valuelen = be32_to_cpu(name_rmt->valuelen);
		}

		if (XFS_IS_CORRUPT(context->dp->i_mount,
				   !xfs_attr_namecheck(entry->flags, name,
						       namelen))) {
			xfs_dirattr_mark_sick(context->dp, XFS_ATTR_FORK);
			return -EFSCORRUPTED;
		}
		context->put_listent(context, entry->flags,
					      name, namelen, value, valuelen);
		if (context->seen_enough)
			break;
		cursor->offset++;
	}
	trace_xfs_attr_list_leaf_end(context);
	return 0;
}

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

	trace_xfs_attr_leaf_list(context);

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

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

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

	xfs_assert_ilocked(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;
	if (dp->i_af.if_format == XFS_DINODE_FMT_LOCAL)
		return xfs_attr_shortform_list(context);

	/* Prerequisite for xfs_attr_is_leaf */
	error = xfs_iread_extents(NULL, dp, XFS_ATTR_FORK);
	if (error)
		return error;

	if (xfs_attr_is_leaf(dp))
		return xfs_attr_leaf_list(context);
	return xfs_attr_node_list(context);
}

int
xfs_attr_list(
	struct xfs_attr_list_context	*context)
{
	struct xfs_inode		*dp = context->dp;
	uint				lock_mode;
	int				error;

	XFS_STATS_INC(dp->i_mount, xs_attr_list);

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

	lock_mode = xfs_ilock_attr_map_shared(dp);
	error = xfs_attr_list_ilocked(context);
	xfs_iunlock(dp, lock_mode);
	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"
5467b34b	9	#include "xfs_shared.h"
a4fbe6ab	10	#include "xfs_format.h"
239880ef DC	11	#include "xfs_log_format.h"
239880ef DC	12	#include "xfs_trans_resv.h"
abec5f2b	13	#include "xfs_mount.h"
57062787	14	#include "xfs_da_format.h"
abec5f2b	15	#include "xfs_inode.h"
239880ef	16	#include "xfs_trans.h"
abec5f2b	17	#include "xfs_bmap.h"
fd920008	18	#include "xfs_da_btree.h"
abec5f2b	19	#include "xfs_attr.h"
a4fbe6ab	20	#include "xfs_attr_sf.h"
abec5f2b DC	21	#include "xfs_attr_leaf.h"
	22	#include "xfs_error.h"
	23	#include "xfs_trace.h"
4bceb18f	24	#include "xfs_dir2.h"
ca14c096	25	#include "xfs_health.h"
abec5f2b DC	26
	27	STATIC int
	28	xfs_attr_shortform_compare(const void a, const void b)
	29	{
	30	xfs_attr_sf_sort_t sa, sb;
	31
	32	sa = (xfs_attr_sf_sort_t *)a;
	33	sb = (xfs_attr_sf_sort_t *)b;
	34	if (sa->hash < sb->hash) {
d99831ff	35	return -1;
abec5f2b	36	} else if (sa->hash > sb->hash) {
d99831ff	37	return 1;
abec5f2b	38	} else {
d99831ff	39	return sa->entno - sb->entno;
abec5f2b DC	40	}
	41	}
	42
	43	#define XFS_ISRESET_CURSOR(cursor) \
	44	(!((cursor)->initted) && !((cursor)->hashval) && \
	45	!((cursor)->blkno) && !((cursor)->offset))
	46	/*
	47	* Copy out entries of shortform attribute lists for attr_list().
	48	* Shortform attribute lists are not stored in hashval sorted order.
b63da6c8	49	* If the output buffer is not large enough to hold them all, then
abec5f2b DC	50	* we have to calculate each entries' hashvalue and sort them before
	51	* we can begin returning them to the user.
	52	*/
0d5a75e9	53	static int
16c6e92c DW	54	xfs_attr_shortform_list(
16c6e92c DW	55	struct xfs_attr_list_context *context)
abec5f2b	56	{
e3a19cde CH	57	struct xfs_attrlist_cursor_kern *cursor = &context->cursor;
e3a19cde CH	58	struct xfs_inode *dp = context->dp;
16c6e92c	59	struct xfs_attr_sf_sort sbuf, sbp;
41414722	60	struct xfs_attr_sf_hdr *sf = dp->i_af.if_data;
16c6e92c	61	struct xfs_attr_sf_entry *sfe;
16c6e92c DW	62	int sbsize, nsbuf, count, i;
16c6e92c DW	63	int error = 0;
abec5f2b	64
abec5f2b	65	ASSERT(sf != NULL);
41414722	66	if (!sf->count)
d99831ff	67	return 0;
abec5f2b DC	68
	69	trace_xfs_attr_list_sf(context);
	70
	71	/*
	72	* If the buffer is large enough and the cursor is at the start,
	73	* do not bother with sorting since we will return everything in
	74	* one buffer and another call using the cursor won't need to be
	75	* made.
	76	* Note the generous fudge factor of 16 overhead bytes per entry.
	77	* If bufsize is zero then put_listent must be a search function
	78	* and can just scan through what we have.
	79	*/
	80	if (context->bufsize == 0 \|\|
	81	(XFS_ISRESET_CURSOR(cursor) &&
41414722 CH	82	(dp->i_af.if_bytes + sf->count * 16) < context->bufsize)) {
41414722 CH	83	for (i = 0, sfe = xfs_attr_sf_firstentry(sf); i < sf->count; i++) {
a71895c5	84	if (XFS_IS_CORRUPT(context->dp->i_mount,
ea0b3e81 DW	85	!xfs_attr_namecheck(sfe->flags,
ea0b3e81 DW	86	sfe->nameval,
989d5ec3 DW	87	sfe->namelen))) {
989d5ec3 DW	88	xfs_dirattr_mark_sick(context->dp, XFS_ATTR_FORK);
16c6e92c	89	return -EFSCORRUPTED;
989d5ec3	90	}
f7a136ae ES	91	context->put_listent(context,
	92	sfe->flags,
	93	sfe->nameval,
	94	(int)sfe->namelen,
8f4b980e	95	&sfe->nameval[sfe->namelen],
f7a136ae	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;
e01b7eed	103	sfe = xfs_attr_sf_nextentry(sfe);
abec5f2b DC	104	}
abec5f2b DC	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	*/
41414722	116	sbsize = sf->count * sizeof(*sbuf);
45f69d09 LL	117	sbp = sbuf = kmalloc(sbsize,
45f69d09 LL	118	GFP_KERNEL \| __GFP_NOLOCKDEP \| __GFP_NOFAIL);
abec5f2b DC	119
	120	/*
	121	* Scan the attribute list for the rest of the entries, storing
	122	* the relevant info from only those that match into a buffer.
	123	*/
	124	nsbuf = 0;
41414722	125	for (i = 0, sfe = xfs_attr_sf_firstentry(sf); i < sf->count; i++) {
abec5f2b DC	126	if (unlikely(
abec5f2b DC	127	((char )sfe < (char )sf) \|\|
ea0b3e81 DW	128	((char )sfe >= ((char )sf + dp->i_af.if_bytes)) \|\|
ea0b3e81 DW	129	!xfs_attr_check_namespace(sfe->flags))) {
abec5f2b DC	130	XFS_CORRUPTION_ERROR("xfs_attr_shortform_list",
abec5f2b DC	131	XFS_ERRLEVEL_LOW,
2551a530 DW	132	context->dp->i_mount, sfe,
2551a530 DW	133	sizeof(*sfe));
d4c75a1b	134	kfree(sbuf);
ca14c096	135	xfs_dirattr_mark_sick(dp, XFS_ATTR_FORK);
2451337d	136	return -EFSCORRUPTED;
abec5f2b DC	137	}
	138
	139	sbp->entno = i;
abec5f2b DC	140	sbp->name = sfe->nameval;
	141	sbp->namelen = sfe->namelen;
	142	/* These are bytes, and both on-disk, don't endian-flip */
8c2263b9	143	sbp->value = &sfe->nameval[sfe->namelen];
abec5f2b DC	144	sbp->valuelen = sfe->valuelen;
abec5f2b DC	145	sbp->flags = sfe->flags;
a64e0134 DW	146	sbp->hash = xfs_attr_hashval(dp->i_mount, sfe->flags,
	147	sfe->nameval, sfe->namelen,
	148	sfe->nameval + sfe->namelen,
	149	sfe->valuelen);
e01b7eed	150	sfe = xfs_attr_sf_nextentry(sfe);
abec5f2b DC	151	sbp++;
	152	nsbuf++;
	153	}
	154
	155	/*
	156	* Sort the entries on hash then entno.
	157	*/
	158	xfs_sort(sbuf, nsbuf, sizeof(*sbuf), xfs_attr_shortform_compare);
	159
	160	/*
	161	* Re-find our place IN THE SORTED LIST.
	162	*/
	163	count = 0;
	164	cursor->initted = 1;
	165	cursor->blkno = 0;
	166	for (sbp = sbuf, i = 0; i < nsbuf; i++, sbp++) {
	167	if (sbp->hash == cursor->hashval) {
	168	if (cursor->offset == count) {
	169	break;
	170	}
	171	count++;
	172	} else if (sbp->hash > cursor->hashval) {
	173	break;
	174	}
	175	}
16c6e92c DW	176	if (i == nsbuf)
16c6e92c DW	177	goto out;
abec5f2b DC	178
	179	/*
	180	* Loop putting entries into the user buffer.
	181	*/
	182	for ( ; i < nsbuf; i++, sbp++) {
	183	if (cursor->hashval != sbp->hash) {
	184	cursor->hashval = sbp->hash;
	185	cursor->offset = 0;
	186	}
a71895c5	187	if (XFS_IS_CORRUPT(context->dp->i_mount,
ea0b3e81	188	!xfs_attr_namecheck(sbp->flags, sbp->name,
a71895c5	189	sbp->namelen))) {
989d5ec3	190	xfs_dirattr_mark_sick(context->dp, XFS_ATTR_FORK);
16c6e92c DW	191	error = -EFSCORRUPTED;
	192	goto out;
	193	}
f7a136ae ES	194	context->put_listent(context,
	195	sbp->flags,
	196	sbp->name,
	197	sbp->namelen,
8f4b980e	198	sbp->value,
f7a136ae	199	sbp->valuelen);
abec5f2b DC	200	if (context->seen_enough)
	201	break;
	202	cursor->offset++;
	203	}
16c6e92c	204	out:
d4c75a1b	205	kfree(sbuf);
16c6e92c	206	return error;
abec5f2b DC	207	}
abec5f2b DC	208
bdaac93f DW	209	/*
	210	* We didn't find the block & hash mentioned in the cursor state, so
	211	* walk down the attr btree looking for the hash.
	212	*/
abec5f2b	213	STATIC int
bdaac93f DW	214	xfs_attr_node_list_lookup(
bdaac93f DW	215	struct xfs_attr_list_context *context,
e3a19cde	216	struct xfs_attrlist_cursor_kern *cursor,
bdaac93f	217	struct xfs_buf **pbp)
abec5f2b	218	{
bdaac93f DW	219	struct xfs_da3_icnode_hdr nodehdr;
	220	struct xfs_da_intnode *node;
	221	struct xfs_da_node_entry *btree;
	222	struct xfs_inode *dp = context->dp;
	223	struct xfs_mount *mp = dp->i_mount;
	224	struct xfs_trans *tp = context->tp;
	225	struct xfs_buf *bp;
f4887fbc	226	xfs_failaddr_t fa;
bdaac93f DW	227	int i;
bdaac93f DW	228	int error = 0;
8210f4dd	229	unsigned int expected_level = 0;
bdaac93f DW	230	uint16_t magic;
	231
	232	ASSERT(*pbp == NULL);
	233	cursor->blkno = 0;
	234	for (;;) {
02c57f0a	235	error = xfs_da3_node_read(tp, dp, cursor->blkno, &bp,
bdaac93f DW	236	XFS_ATTR_FORK);
	237	if (error)
	238	return error;
	239	node = bp->b_addr;
	240	magic = be16_to_cpu(node->hdr.info.magic);
	241	if (magic == XFS_ATTR_LEAF_MAGIC \|\|
	242	magic == XFS_ATTR3_LEAF_MAGIC)
	243	break;
	244	if (magic != XFS_DA_NODE_MAGIC &&
	245	magic != XFS_DA3_NODE_MAGIC) {
	246	XFS_CORRUPTION_ERROR(__func__, XFS_ERRLEVEL_LOW, mp,
2551a530	247	node, sizeof(*node));
bdaac93f DW	248	goto out_corruptbuf;
	249	}
	250
d44bea9b DW	251	fa = xfs_da3_node_header_check(bp, dp->i_ino);
	252	if (fa)
	253	goto out_corruptbuf;
	254
f475dc4d	255	xfs_da3_node_hdr_from_disk(mp, &nodehdr, node);
bdaac93f	256
8210f4dd DW	257	/* Tree taller than we can handle; bail out! */
	258	if (nodehdr.level >= XFS_DA_NODE_MAXDEPTH)
	259	goto out_corruptbuf;
	260
	261	/* Check the level from the root node. */
	262	if (cursor->blkno == 0)
	263	expected_level = nodehdr.level - 1;
	264	else if (expected_level != nodehdr.level)
	265	goto out_corruptbuf;
	266	else
	267	expected_level--;
	268
51908ca7	269	btree = nodehdr.btree;
bdaac93f DW	270	for (i = 0; i < nodehdr.count; btree++, i++) {
	271	if (cursor->hashval <= be32_to_cpu(btree->hashval)) {
	272	cursor->blkno = be32_to_cpu(btree->before);
	273	trace_xfs_attr_list_node_descend(context,
	274	btree);
	275	break;
	276	}
	277	}
	278	xfs_trans_brelse(tp, bp);
	279
	280	if (i == nodehdr.count)
	281	return 0;
8210f4dd DW	282
8210f4dd DW	283	/* We can't point back to the root. */
ca14c096 DW	284	if (XFS_IS_CORRUPT(mp, cursor->blkno == 0)) {
ca14c096 DW	285	xfs_dirattr_mark_sick(dp, XFS_ATTR_FORK);
8210f4dd	286	return -EFSCORRUPTED;
ca14c096	287	}
bdaac93f DW	288	}
bdaac93f DW	289
f4887fbc DW	290	fa = xfs_attr3_leaf_header_check(bp, dp->i_ino);
	291	if (fa) {
	292	__xfs_buf_mark_corrupt(bp, fa);
	293	goto out_releasebuf;
	294	}
	295
8210f4dd DW	296	if (expected_level != 0)
	297	goto out_corruptbuf;
	298
bdaac93f DW	299	*pbp = bp;
	300	return 0;
	301
	302	out_corruptbuf:
8d57c216	303	xfs_buf_mark_corrupt(bp);
f4887fbc	304	out_releasebuf:
bdaac93f	305	xfs_trans_brelse(tp, bp);
ca14c096	306	xfs_dirattr_mark_sick(dp, XFS_ATTR_FORK);
bdaac93f DW	307	return -EFSCORRUPTED;
	308	}
	309
	310	STATIC int
	311	xfs_attr_node_list(
	312	struct xfs_attr_list_context *context)
	313	{
e3a19cde	314	struct xfs_attrlist_cursor_kern *cursor = &context->cursor;
bdaac93f	315	struct xfs_attr3_icleaf_hdr leafhdr;
bdaac93f DW	316	struct xfs_attr_leafblock *leaf;
	317	struct xfs_da_intnode *node;
	318	struct xfs_buf *bp;
	319	struct xfs_inode *dp = context->dp;
	320	struct xfs_mount *mp = dp->i_mount;
f4887fbc	321	xfs_failaddr_t fa;
16c6e92c	322	int error = 0;
abec5f2b DC	323
	324	trace_xfs_attr_node_list(context);
	325
abec5f2b DC	326	cursor->initted = 1;
	327
	328	/*
	329	* Do all sorts of validation on the passed-in cursor structure.
	330	* If anything is amiss, ignore the cursor and look up the hashval
	331	* starting from the btree root.
	332	*/
	333	bp = NULL;
	334	if (cursor->blkno > 0) {
33c028ff DW	335	struct xfs_attr_leaf_entry *entries;
33c028ff DW	336
02c57f0a CH	337	error = xfs_da3_node_read(context->tp, dp, cursor->blkno, &bp,
02c57f0a CH	338	XFS_ATTR_FORK);
ca14c096 DW	339	if (xfs_metadata_is_sick(error))
ca14c096 DW	340	xfs_dirattr_mark_sick(dp, XFS_ATTR_FORK);
33c028ff	341	if (error != 0 && error != -EFSCORRUPTED)
d99831ff	342	return error;
33c028ff DW	343	if (!bp)
33c028ff DW	344	goto need_lookup;
abec5f2b	345
33c028ff DW	346	node = bp->b_addr;
	347	switch (be16_to_cpu(node->hdr.info.magic)) {
	348	case XFS_DA_NODE_MAGIC:
	349	case XFS_DA3_NODE_MAGIC:
	350	trace_xfs_attr_list_wrong_blk(context);
d44bea9b DW	351	fa = xfs_da3_node_header_check(bp, dp->i_ino);
	352	if (fa) {
	353	__xfs_buf_mark_corrupt(bp, fa);
	354	xfs_dirattr_mark_sick(dp, XFS_ATTR_FORK);
	355	}
33c028ff DW	356	xfs_trans_brelse(context->tp, bp);
	357	bp = NULL;
	358	break;
	359	case XFS_ATTR_LEAF_MAGIC:
	360	case XFS_ATTR3_LEAF_MAGIC:
	361	leaf = bp->b_addr;
f4887fbc DW	362	fa = xfs_attr3_leaf_header_check(bp, dp->i_ino);
	363	if (fa) {
	364	__xfs_buf_mark_corrupt(bp, fa);
	365	xfs_trans_brelse(context->tp, bp);
	366	xfs_dirattr_mark_sick(dp, XFS_ATTR_FORK);
	367	bp = NULL;
	368	break;
	369	}
33c028ff DW	370	xfs_attr3_leaf_hdr_from_disk(mp->m_attr_geo,
	371	&leafhdr, leaf);
	372	entries = xfs_attr3_leaf_entryp(leaf);
	373	if (cursor->hashval > be32_to_cpu(
	374	entries[leafhdr.count - 1].hashval)) {
abec5f2b	375	trace_xfs_attr_list_wrong_blk(context);
ad017f65	376	xfs_trans_brelse(context->tp, bp);
abec5f2b	377	bp = NULL;
33c028ff DW	378	} else if (cursor->hashval <= be32_to_cpu(
33c028ff DW	379	entries[0].hashval)) {
abec5f2b	380	trace_xfs_attr_list_wrong_blk(context);
ad017f65	381	xfs_trans_brelse(context->tp, bp);
abec5f2b DC	382	bp = NULL;
abec5f2b DC	383	}
33c028ff DW	384	break;
	385	default:
	386	trace_xfs_attr_list_wrong_blk(context);
	387	xfs_trans_brelse(context->tp, bp);
	388	bp = NULL;
abec5f2b DC	389	}
	390	}
	391
	392	/*
	393	* We did not find what we expected given the cursor's contents,
	394	* so we start from the top and work down based on the hash value.
	395	* Note that start of node block is same as start of leaf block.
	396	*/
	397	if (bp == NULL) {
33c028ff	398	need_lookup:
bdaac93f DW	399	error = xfs_attr_node_list_lookup(context, cursor, &bp);
	400	if (error \|\| !bp)
	401	return error;
abec5f2b DC	402	}
	403	ASSERT(bp != NULL);
	404
	405	/*
	406	* Roll upward through the blocks, processing each leaf block in
	407	* order. As long as there is space in the result buffer, keep
	408	* adding the information.
	409	*/
	410	for (;;) {
	411	leaf = bp->b_addr;
16c6e92c DW	412	error = xfs_attr3_leaf_list_int(bp, context);
	413	if (error)
	414	break;
2f661241	415	xfs_attr3_leaf_hdr_from_disk(mp->m_attr_geo, &leafhdr, leaf);
abec5f2b DC	416	if (context->seen_enough \|\| leafhdr.forw == 0)
	417	break;
	418	cursor->blkno = leafhdr.forw;
ad017f65	419	xfs_trans_brelse(context->tp, bp);
f4887fbc DW	420	error = xfs_attr3_leaf_read(context->tp, dp, dp->i_ino,
f4887fbc DW	421	cursor->blkno, &bp);
abec5f2b DC	422	if (error)
	423	return error;
	424	}
ad017f65	425	xfs_trans_brelse(context->tp, bp);
16c6e92c	426	return error;
abec5f2b DC	427	}
	428
	429	/*
	430	* Copy out attribute list entries for attr_list(), for leaf attribute lists.
	431	*/
16c6e92c	432	int
abec5f2b DC	433	xfs_attr3_leaf_list_int(
	434	struct xfs_buf *bp,
	435	struct xfs_attr_list_context *context)
	436	{
e3a19cde	437	struct xfs_attrlist_cursor_kern *cursor = &context->cursor;
abec5f2b DC	438	struct xfs_attr_leafblock *leaf;
	439	struct xfs_attr3_icleaf_hdr ichdr;
	440	struct xfs_attr_leaf_entry *entries;
	441	struct xfs_attr_leaf_entry *entry;
abec5f2b	442	int i;
2f661241	443	struct xfs_mount *mp = context->dp->i_mount;
abec5f2b DC	444
	445	trace_xfs_attr_list_leaf(context);
	446
	447	leaf = bp->b_addr;
2f661241	448	xfs_attr3_leaf_hdr_from_disk(mp->m_attr_geo, &ichdr, leaf);
abec5f2b DC	449	entries = xfs_attr3_leaf_entryp(leaf);
abec5f2b DC	450
abec5f2b DC	451	cursor->initted = 1;
	452
	453	/*
	454	* Re-find our place in the leaf block if this is a new syscall.
	455	*/
	456	if (context->resynch) {
	457	entry = &entries[0];
	458	for (i = 0; i < ichdr.count; entry++, i++) {
	459	if (be32_to_cpu(entry->hashval) == cursor->hashval) {
	460	if (cursor->offset == context->dupcnt) {
	461	context->dupcnt = 0;
	462	break;
	463	}
	464	context->dupcnt++;
	465	} else if (be32_to_cpu(entry->hashval) >
	466	cursor->hashval) {
	467	context->dupcnt = 0;
	468	break;
	469	}
	470	}
	471	if (i == ichdr.count) {
	472	trace_xfs_attr_list_notfound(context);
16c6e92c	473	return 0;
abec5f2b DC	474	}
	475	} else {
	476	entry = &entries[0];
	477	i = 0;
	478	}
	479	context->resynch = 0;
	480
	481	/*
	482	* We have found our place, start copying out the new attributes.
	483	*/
abec5f2b	484	for (; i < ichdr.count; entry++, i++) {
3ab3ffca	485	char *name;
8f4b980e	486	void *value;
3ab3ffca ES	487	int namelen, valuelen;
3ab3ffca ES	488
abec5f2b DC	489	if (be32_to_cpu(entry->hashval) != cursor->hashval) {
	490	cursor->hashval = be32_to_cpu(entry->hashval);
	491	cursor->offset = 0;
	492	}
	493
eec0482e	494	if ((entry->flags & XFS_ATTR_INCOMPLETE) &&
5e813574 CH	495	!context->allow_incomplete)
5e813574 CH	496	continue;
abec5f2b DC	497
abec5f2b DC	498	if (entry->flags & XFS_ATTR_LOCAL) {
3ab3ffca	499	xfs_attr_leaf_name_local_t *name_loc;
abec5f2b	500
3ab3ffca ES	501	name_loc = xfs_attr3_leaf_name_local(leaf, i);
	502	name = name_loc->nameval;
	503	namelen = name_loc->namelen;
8f4b980e	504	value = &name_loc->nameval[name_loc->namelen];
3ab3ffca ES	505	valuelen = be16_to_cpu(name_loc->valuelen);
	506	} else {
	507	xfs_attr_leaf_name_remote_t *name_rmt;
abec5f2b	508
3ab3ffca ES	509	name_rmt = xfs_attr3_leaf_name_remote(leaf, i);
	510	name = name_rmt->name;
	511	namelen = name_rmt->namelen;
8f4b980e	512	value = NULL;
3ab3ffca	513	valuelen = be32_to_cpu(name_rmt->valuelen);
abec5f2b	514	}
3ab3ffca	515
a71895c5	516	if (XFS_IS_CORRUPT(context->dp->i_mount,
ea0b3e81 DW	517	!xfs_attr_namecheck(entry->flags, name,
ea0b3e81 DW	518	namelen))) {
989d5ec3	519	xfs_dirattr_mark_sick(context->dp, XFS_ATTR_FORK);
16c6e92c	520	return -EFSCORRUPTED;
989d5ec3	521	}
f7a136ae	522	context->put_listent(context, entry->flags,
8f4b980e	523	name, namelen, value, valuelen);
abec5f2b DC	524	if (context->seen_enough)
	525	break;
	526	cursor->offset++;
	527	}
	528	trace_xfs_attr_list_leaf_end(context);
16c6e92c	529	return 0;
abec5f2b DC	530	}
	531
	532	/*
	533	* Copy out attribute entries for attr_list(), for leaf attribute lists.
	534	*/
	535	STATIC int
a9c8c69b CH	536	xfs_attr_leaf_list(
a9c8c69b CH	537	struct xfs_attr_list_context *context)
abec5f2b	538	{
a9c8c69b CH	539	struct xfs_buf *bp;
a9c8c69b CH	540	int error;
abec5f2b DC	541
	542	trace_xfs_attr_leaf_list(context);
	543
e3a19cde	544	context->cursor.blkno = 0;
f4887fbc DW	545	error = xfs_attr3_leaf_read(context->tp, context->dp,
f4887fbc DW	546	context->dp->i_ino, 0, &bp);
abec5f2b	547	if (error)
b474c7ae	548	return error;
abec5f2b	549
16c6e92c	550	error = xfs_attr3_leaf_list_int(bp, context);
ad017f65	551	xfs_trans_brelse(context->tp, bp);
16c6e92c	552	return error;
abec5f2b DC	553	}
abec5f2b DC	554
ad017f65	555	int
17e1dd83	556	xfs_attr_list_ilocked(
ad017f65 DW	557	struct xfs_attr_list_context *context)
	558	{
	559	struct xfs_inode *dp = context->dp;
ef80de94	560	int error;
ad017f65	561
3fed24ff	562	xfs_assert_ilocked(dp, XFS_ILOCK_SHARED \| XFS_ILOCK_EXCL);
5af7777e	563
ad017f65 DW	564	/*
	565	* Decide on what work routines to call based on the inode size.
	566	*/
	567	if (!xfs_inode_hasattr(dp))
	568	return 0;
2ed5b09b	569	if (dp->i_af.if_format == XFS_DINODE_FMT_LOCAL)
ad017f65	570	return xfs_attr_shortform_list(context);
ef80de94 DW	571
	572	/* Prerequisite for xfs_attr_is_leaf */
	573	error = xfs_iread_extents(NULL, dp, XFS_ATTR_FORK);
	574	if (error)
	575	return error;
	576
2ac131df	577	if (xfs_attr_is_leaf(dp))
ad017f65 DW	578	return xfs_attr_leaf_list(context);
	579	return xfs_attr_node_list(context);
	580	}
	581
abec5f2b	582	int
17e1dd83	583	xfs_attr_list(
a9c8c69b	584	struct xfs_attr_list_context *context)
abec5f2b	585	{
a9c8c69b CH	586	struct xfs_inode *dp = context->dp;
	587	uint lock_mode;
	588	int error;
abec5f2b	589
ff6d6af2	590	XFS_STATS_INC(dp->i_mount, xs_attr_list);
abec5f2b	591
75c8c50f	592	if (xfs_is_shutdown(dp->i_mount))
2451337d	593	return -EIO;
abec5f2b	594
568d994e	595	lock_mode = xfs_ilock_attr_map_shared(dp);
17e1dd83	596	error = xfs_attr_list_ilocked(context);
568d994e	597	xfs_iunlock(dp, lock_mode);
abec5f2b DC	598	return error;
abec5f2b DC	599	}