2 * Copyright (C) 2010 Red Hat, Inc.
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.
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.
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
19 #include "xfs_format.h"
21 #include "xfs_trans.h"
24 #include "xfs_mount.h"
25 #include "xfs_quota.h"
26 #include "xfs_alloc_btree.h"
27 #include "xfs_bmap_btree.h"
28 #include "xfs_ialloc_btree.h"
29 #include "xfs_btree.h"
30 #include "xfs_inode.h"
31 #include "xfs_alloc.h"
32 #include "xfs_error.h"
33 #include "xfs_extent_busy.h"
34 #include "xfs_discard.h"
35 #include "xfs_trace.h"
44 __uint64_t *blocks_trimmed)
46 struct block_device *bdev = mp->m_ddev_targp->bt_bdev;
47 struct xfs_btree_cur *cur;
49 struct xfs_perag *pag;
53 pag = xfs_perag_get(mp, agno);
55 error = xfs_alloc_read_agf(mp, NULL, agno, 0, &agbp);
59 cur = xfs_allocbt_init_cursor(mp, NULL, agbp, agno, XFS_BTNUM_CNT);
62 * Force out the log. This means any transactions that might have freed
63 * space before we took the AGF buffer lock are now on disk, and the
64 * volatile disk cache is flushed.
66 xfs_log_force(mp, XFS_LOG_SYNC);
69 * Look up the longest btree in the AGF and start with it.
71 error = xfs_alloc_lookup_ge(cur, 0,
72 be32_to_cpu(XFS_BUF_TO_AGF(agbp)->agf_longest), &i);
77 * Loop until we are done with all extents that are large
78 * enough to be worth discarding.
86 error = xfs_alloc_get_rec(cur, &fbno, &flen, &i);
89 XFS_WANT_CORRUPTED_GOTO(i == 1, out_del_cursor);
90 ASSERT(flen <= be32_to_cpu(XFS_BUF_TO_AGF(agbp)->agf_longest));
93 * use daddr format for all range/len calculations as that is
94 * the format the range/len variables are supplied in by
97 dbno = XFS_AGB_TO_DADDR(mp, agno, fbno);
98 dlen = XFS_FSB_TO_BB(mp, flen);
101 * Too small? Give up.
104 trace_xfs_discard_toosmall(mp, agno, fbno, flen);
109 * If the extent is entirely outside of the range we are
110 * supposed to discard skip it. Do not bother to trim
111 * down partially overlapping ranges for now.
113 if (dbno + dlen < start || dbno > end) {
114 trace_xfs_discard_exclude(mp, agno, fbno, flen);
119 * If any blocks in the range are still busy, skip the
120 * discard and try again the next time.
122 if (xfs_extent_busy_search(mp, agno, fbno, flen)) {
123 trace_xfs_discard_busy(mp, agno, fbno, flen);
127 trace_xfs_discard_extent(mp, agno, fbno, flen);
128 error = -blkdev_issue_discard(bdev, dbno, dlen, GFP_NOFS, 0);
131 *blocks_trimmed += flen;
134 error = xfs_btree_decrement(cur, 0, &i);
140 xfs_btree_del_cursor(cur, error ? XFS_BTREE_ERROR : XFS_BTREE_NOERROR);
148 * trim a range of the filesystem.
150 * Note: the parameters passed from userspace are byte ranges into the
151 * filesystem which does not match to the format we use for filesystem block
152 * addressing. FSB addressing is sparse (AGNO|AGBNO), while the incoming format
153 * is a linear address range. Hence we need to use DADDR based conversions and
154 * comparisons for determining the correct offset and regions to trim.
158 struct xfs_mount *mp,
159 struct fstrim_range __user *urange)
161 struct request_queue *q = mp->m_ddev_targp->bt_bdev->bd_disk->queue;
162 unsigned int granularity = q->limits.discard_granularity;
163 struct fstrim_range range;
164 xfs_daddr_t start, end, minlen;
165 xfs_agnumber_t start_agno, end_agno, agno;
166 __uint64_t blocks_trimmed = 0;
167 int error, last_error = 0;
169 if (!capable(CAP_SYS_ADMIN))
170 return -XFS_ERROR(EPERM);
171 if (!blk_queue_discard(q))
172 return -XFS_ERROR(EOPNOTSUPP);
173 if (copy_from_user(&range, urange, sizeof(range)))
174 return -XFS_ERROR(EFAULT);
177 * Truncating down the len isn't actually quite correct, but using
178 * BBTOB would mean we trivially get overflows for values
179 * of ULLONG_MAX or slightly lower. And ULLONG_MAX is the default
180 * used by the fstrim application. In the end it really doesn't
181 * matter as trimming blocks is an advisory interface.
183 if (range.start >= XFS_FSB_TO_B(mp, mp->m_sb.sb_dblocks) ||
184 range.minlen > XFS_FSB_TO_B(mp, XFS_ALLOC_AG_MAX_USABLE(mp)))
185 return -XFS_ERROR(EINVAL);
187 start = BTOBB(range.start);
188 end = start + BTOBBT(range.len) - 1;
189 minlen = BTOBB(max_t(u64, granularity, range.minlen));
191 if (end > XFS_FSB_TO_BB(mp, mp->m_sb.sb_dblocks) - 1)
192 end = XFS_FSB_TO_BB(mp, mp->m_sb.sb_dblocks)- 1;
194 start_agno = xfs_daddr_to_agno(mp, start);
195 end_agno = xfs_daddr_to_agno(mp, end);
197 for (agno = start_agno; agno <= end_agno; agno++) {
198 error = -xfs_trim_extents(mp, agno, start, end, minlen,
207 range.len = XFS_FSB_TO_B(mp, blocks_trimmed);
208 if (copy_to_user(urange, &range, sizeof(range)))
209 return -XFS_ERROR(EFAULT);
215 struct xfs_mount *mp,
216 struct list_head *list)
218 struct xfs_extent_busy *busyp;
221 list_for_each_entry(busyp, list, list) {
222 trace_xfs_discard_extent(mp, busyp->agno, busyp->bno,
225 error = -blkdev_issue_discard(mp->m_ddev_targp->bt_bdev,
226 XFS_AGB_TO_DADDR(mp, busyp->agno, busyp->bno),
227 XFS_FSB_TO_BB(mp, busyp->length),
229 if (error && error != EOPNOTSUPP) {
231 "discard failed for extent [0x%llu,%u], error %d",
232 (unsigned long long)busyp->bno,