1/*
2 * Copyright (C) 2010 Red Hat, Inc.
3 * All Rights Reserved.
4 *
5 * This program is free software; you can redistribute it and/or
6 * modify it under the terms of the GNU General Public License as
7 * published by the Free Software Foundation.
8 *
9 * This program is distributed in the hope that it would be useful,
10 * but WITHOUT ANY WARRANTY; without even the implied warranty of
11 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
12 * GNU General Public License for more details.
13 *
14 * You should have received a copy of the GNU General Public License
15 * along with this program; if not, write the Free Software Foundation,
16 * Inc.,  51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA
17 */
18#include "xfs.h"
19#include "xfs_format.h"
20#include "xfs_log_format.h"
21#include "xfs_trans_resv.h"
22#include "xfs_sb.h"
23#include "xfs_mount.h"
24#include "xfs_quota.h"
25#include "xfs_inode.h"
26#include "xfs_btree.h"
27#include "xfs_alloc_btree.h"
28#include "xfs_alloc.h"
29#include "xfs_error.h"
30#include "xfs_extent_busy.h"
31#include "xfs_discard.h"
32#include "xfs_trace.h"
33#include "xfs_log.h"
34
35STATIC int
36xfs_trim_extents(
37	struct xfs_mount	*mp,
38	xfs_agnumber_t		agno,
39	xfs_daddr_t		start,
40	xfs_daddr_t		end,
41	xfs_daddr_t		minlen,
42	__uint64_t		*blocks_trimmed)
43{
44	struct block_device	*bdev = mp->m_ddev_targp->bt_bdev;
45	struct xfs_btree_cur	*cur;
46	struct xfs_buf		*agbp;
47	struct xfs_perag	*pag;
48	int			error;
49	int			i;
50
51	pag = xfs_perag_get(mp, agno);
52
53	error = xfs_alloc_read_agf(mp, NULL, agno, 0, &agbp);
54	if (error || !agbp)
55		goto out_put_perag;
56
57	cur = xfs_allocbt_init_cursor(mp, NULL, agbp, agno, XFS_BTNUM_CNT);
58
59	/*
60	 * Force out the log.  This means any transactions that might have freed
61	 * space before we took the AGF buffer lock are now on disk, and the
62	 * volatile disk cache is flushed.
63	 */
64	xfs_log_force(mp, XFS_LOG_SYNC);
65
66	/*
67	 * Look up the longest btree in the AGF and start with it.
68	 */
69	error = xfs_alloc_lookup_ge(cur, 0,
70			    be32_to_cpu(XFS_BUF_TO_AGF(agbp)->agf_longest), &i);
71	if (error)
72		goto out_del_cursor;
73
74	/*
75	 * Loop until we are done with all extents that are large
76	 * enough to be worth discarding.
77	 */
78	while (i) {
79		xfs_agblock_t	fbno;
80		xfs_extlen_t	flen;
81		xfs_daddr_t	dbno;
82		xfs_extlen_t	dlen;
83
84		error = xfs_alloc_get_rec(cur, &fbno, &flen, &i);
85		if (error)
86			goto out_del_cursor;
87		XFS_WANT_CORRUPTED_GOTO(mp, i == 1, out_del_cursor);
88		ASSERT(flen <= be32_to_cpu(XFS_BUF_TO_AGF(agbp)->agf_longest));
89
90		/*
91		 * use daddr format for all range/len calculations as that is
92		 * the format the range/len variables are supplied in by
93		 * userspace.
94		 */
95		dbno = XFS_AGB_TO_DADDR(mp, agno, fbno);
96		dlen = XFS_FSB_TO_BB(mp, flen);
97
98		/*
99		 * Too small?  Give up.
100		 */
101		if (dlen < minlen) {
102			trace_xfs_discard_toosmall(mp, agno, fbno, flen);
103			goto out_del_cursor;
104		}
105
106		/*
107		 * If the extent is entirely outside of the range we are
108		 * supposed to discard skip it.  Do not bother to trim
109		 * down partially overlapping ranges for now.
110		 */
111		if (dbno + dlen < start || dbno > end) {
112			trace_xfs_discard_exclude(mp, agno, fbno, flen);
113			goto next_extent;
114		}
115
116		/*
117		 * If any blocks in the range are still busy, skip the
118		 * discard and try again the next time.
119		 */
120		if (xfs_extent_busy_search(mp, agno, fbno, flen)) {
121			trace_xfs_discard_busy(mp, agno, fbno, flen);
122			goto next_extent;
123		}
124
125		trace_xfs_discard_extent(mp, agno, fbno, flen);
126		error = blkdev_issue_discard(bdev, dbno, dlen, GFP_NOFS, 0);
127		if (error)
128			goto out_del_cursor;
129		*blocks_trimmed += flen;
130
131next_extent:
132		error = xfs_btree_decrement(cur, 0, &i);
133		if (error)
134			goto out_del_cursor;
135	}
136
137out_del_cursor:
138	xfs_btree_del_cursor(cur, error ? XFS_BTREE_ERROR : XFS_BTREE_NOERROR);
139	xfs_buf_relse(agbp);
140out_put_perag:
141	xfs_perag_put(pag);
142	return error;
143}
144
145/*
146 * trim a range of the filesystem.
147 *
148 * Note: the parameters passed from userspace are byte ranges into the
149 * filesystem which does not match to the format we use for filesystem block
150 * addressing. FSB addressing is sparse (AGNO|AGBNO), while the incoming format
151 * is a linear address range. Hence we need to use DADDR based conversions and
152 * comparisons for determining the correct offset and regions to trim.
153 */
154int
155xfs_ioc_trim(
156	struct xfs_mount		*mp,
157	struct fstrim_range __user	*urange)
158{
159	struct request_queue	*q = bdev_get_queue(mp->m_ddev_targp->bt_bdev);
160	unsigned int		granularity = q->limits.discard_granularity;
161	struct fstrim_range	range;
162	xfs_daddr_t		start, end, minlen;
163	xfs_agnumber_t		start_agno, end_agno, agno;
164	__uint64_t		blocks_trimmed = 0;
165	int			error, last_error = 0;
166
167	if (!capable(CAP_SYS_ADMIN))
168		return -EPERM;
169	if (!blk_queue_discard(q))
170		return -EOPNOTSUPP;
171	if (copy_from_user(&range, urange, sizeof(range)))
172		return -EFAULT;
173
174	/*
175	 * Truncating down the len isn't actually quite correct, but using
176	 * BBTOB would mean we trivially get overflows for values
177	 * of ULLONG_MAX or slightly lower.  And ULLONG_MAX is the default
178	 * used by the fstrim application.  In the end it really doesn't
179	 * matter as trimming blocks is an advisory interface.
180	 */
181	if (range.start >= XFS_FSB_TO_B(mp, mp->m_sb.sb_dblocks) ||
182	    range.minlen > XFS_FSB_TO_B(mp, XFS_ALLOC_AG_MAX_USABLE(mp)) ||
183	    range.len < mp->m_sb.sb_blocksize)
184		return -EINVAL;
185
186	start = BTOBB(range.start);
187	end = start + BTOBBT(range.len) - 1;
188	minlen = BTOBB(max_t(u64, granularity, range.minlen));
189
190	if (end > XFS_FSB_TO_BB(mp, mp->m_sb.sb_dblocks) - 1)
191		end = XFS_FSB_TO_BB(mp, mp->m_sb.sb_dblocks)- 1;
192
193	start_agno = xfs_daddr_to_agno(mp, start);
194	end_agno = xfs_daddr_to_agno(mp, end);
195
196	for (agno = start_agno; agno <= end_agno; agno++) {
197		error = xfs_trim_extents(mp, agno, start, end, minlen,
198					  &blocks_trimmed);
199		if (error)
200			last_error = error;
201	}
202
203	if (last_error)
204		return last_error;
205
206	range.len = XFS_FSB_TO_B(mp, blocks_trimmed);
207	if (copy_to_user(urange, &range, sizeof(range)))
208		return -EFAULT;
209	return 0;
210}
211
212int
213xfs_discard_extents(
214	struct xfs_mount	*mp,
215	struct list_head	*list)
216{
217	struct xfs_extent_busy	*busyp;
218	int			error = 0;
219
220	list_for_each_entry(busyp, list, list) {
221		trace_xfs_discard_extent(mp, busyp->agno, busyp->bno,
222					 busyp->length);
223
224		error = blkdev_issue_discard(mp->m_ddev_targp->bt_bdev,
225				XFS_AGB_TO_DADDR(mp, busyp->agno, busyp->bno),
226				XFS_FSB_TO_BB(mp, busyp->length),
227				GFP_NOFS, 0);
228		if (error && error != -EOPNOTSUPP) {
229			xfs_info(mp,
230	 "discard failed for extent [0x%llu,%u], error %d",
231				 (unsigned long long)busyp->bno,
232				 busyp->length,
233				 error);
234			return error;
235		}
236	}
237
238	return 0;
239}
240