Commit | Line | Data |
---|---|---|
a46db608 CH |
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_sb.h" | |
20 | #include "xfs_inum.h" | |
21 | #include "xfs_log.h" | |
22 | #include "xfs_ag.h" | |
23 | #include "xfs_mount.h" | |
24 | #include "xfs_quota.h" | |
25 | #include "xfs_trans.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_discard.h" | |
34 | #include "xfs_trace.h" | |
35 | ||
36 | STATIC int | |
37 | xfs_trim_extents( | |
38 | struct xfs_mount *mp, | |
39 | xfs_agnumber_t agno, | |
40 | xfs_fsblock_t start, | |
c029a50d | 41 | xfs_fsblock_t end, |
a46db608 CH |
42 | xfs_fsblock_t minlen, |
43 | __uint64_t *blocks_trimmed) | |
44 | { | |
45 | struct block_device *bdev = mp->m_ddev_targp->bt_bdev; | |
46 | struct xfs_btree_cur *cur; | |
47 | struct xfs_buf *agbp; | |
48 | struct xfs_perag *pag; | |
49 | int error; | |
50 | int i; | |
51 | ||
52 | pag = xfs_perag_get(mp, agno); | |
53 | ||
54 | error = xfs_alloc_read_agf(mp, NULL, agno, 0, &agbp); | |
55 | if (error || !agbp) | |
56 | goto out_put_perag; | |
57 | ||
58 | cur = xfs_allocbt_init_cursor(mp, NULL, agbp, agno, XFS_BTNUM_CNT); | |
59 | ||
60 | /* | |
61 | * Force out the log. This means any transactions that might have freed | |
62 | * space before we took the AGF buffer lock are now on disk, and the | |
63 | * volatile disk cache is flushed. | |
64 | */ | |
65 | xfs_log_force(mp, XFS_LOG_SYNC); | |
66 | ||
67 | /* | |
68 | * Look up the longest btree in the AGF and start with it. | |
69 | */ | |
70 | error = xfs_alloc_lookup_le(cur, 0, | |
71 | XFS_BUF_TO_AGF(agbp)->agf_longest, &i); | |
72 | if (error) | |
73 | goto out_del_cursor; | |
74 | ||
75 | /* | |
76 | * Loop until we are done with all extents that are large | |
77 | * enough to be worth discarding. | |
78 | */ | |
79 | while (i) { | |
80 | xfs_agblock_t fbno; | |
81 | xfs_extlen_t flen; | |
82 | ||
83 | error = xfs_alloc_get_rec(cur, &fbno, &flen, &i); | |
84 | if (error) | |
85 | goto out_del_cursor; | |
86 | XFS_WANT_CORRUPTED_GOTO(i == 1, out_del_cursor); | |
87 | ASSERT(flen <= XFS_BUF_TO_AGF(agbp)->agf_longest); | |
88 | ||
89 | /* | |
90 | * Too small? Give up. | |
91 | */ | |
92 | if (flen < minlen) { | |
93 | trace_xfs_discard_toosmall(mp, agno, fbno, flen); | |
94 | goto out_del_cursor; | |
95 | } | |
96 | ||
97 | /* | |
98 | * If the extent is entirely outside of the range we are | |
99 | * supposed to discard skip it. Do not bother to trim | |
100 | * down partially overlapping ranges for now. | |
101 | */ | |
102 | if (XFS_AGB_TO_FSB(mp, agno, fbno) + flen < start || | |
c029a50d | 103 | XFS_AGB_TO_FSB(mp, agno, fbno) > end) { |
a46db608 CH |
104 | trace_xfs_discard_exclude(mp, agno, fbno, flen); |
105 | goto next_extent; | |
106 | } | |
107 | ||
108 | /* | |
109 | * If any blocks in the range are still busy, skip the | |
110 | * discard and try again the next time. | |
111 | */ | |
112 | if (xfs_alloc_busy_search(mp, agno, fbno, flen)) { | |
113 | trace_xfs_discard_busy(mp, agno, fbno, flen); | |
114 | goto next_extent; | |
115 | } | |
116 | ||
117 | trace_xfs_discard_extent(mp, agno, fbno, flen); | |
118 | error = -blkdev_issue_discard(bdev, | |
119 | XFS_AGB_TO_DADDR(mp, agno, fbno), | |
120 | XFS_FSB_TO_BB(mp, flen), | |
121 | GFP_NOFS, 0); | |
122 | if (error) | |
123 | goto out_del_cursor; | |
124 | *blocks_trimmed += flen; | |
125 | ||
126 | next_extent: | |
127 | error = xfs_btree_decrement(cur, 0, &i); | |
128 | if (error) | |
129 | goto out_del_cursor; | |
130 | } | |
131 | ||
132 | out_del_cursor: | |
133 | xfs_btree_del_cursor(cur, error ? XFS_BTREE_ERROR : XFS_BTREE_NOERROR); | |
134 | xfs_buf_relse(agbp); | |
135 | out_put_perag: | |
136 | xfs_perag_put(pag); | |
137 | return error; | |
138 | } | |
139 | ||
140 | int | |
141 | xfs_ioc_trim( | |
142 | struct xfs_mount *mp, | |
143 | struct fstrim_range __user *urange) | |
144 | { | |
145 | struct request_queue *q = mp->m_ddev_targp->bt_bdev->bd_disk->queue; | |
146 | unsigned int granularity = q->limits.discard_granularity; | |
147 | struct fstrim_range range; | |
c029a50d | 148 | xfs_fsblock_t start, end, minlen; |
a46db608 CH |
149 | xfs_agnumber_t start_agno, end_agno, agno; |
150 | __uint64_t blocks_trimmed = 0; | |
151 | int error, last_error = 0; | |
152 | ||
153 | if (!capable(CAP_SYS_ADMIN)) | |
154 | return -XFS_ERROR(EPERM); | |
be715140 LC |
155 | if (!blk_queue_discard(q)) |
156 | return -XFS_ERROR(EOPNOTSUPP); | |
a46db608 CH |
157 | if (copy_from_user(&range, urange, sizeof(range))) |
158 | return -XFS_ERROR(EFAULT); | |
159 | ||
160 | /* | |
161 | * Truncating down the len isn't actually quite correct, but using | |
162 | * XFS_B_TO_FSB would mean we trivially get overflows for values | |
163 | * of ULLONG_MAX or slightly lower. And ULLONG_MAX is the default | |
164 | * used by the fstrim application. In the end it really doesn't | |
165 | * matter as trimming blocks is an advisory interface. | |
166 | */ | |
167 | start = XFS_B_TO_FSBT(mp, range.start); | |
c029a50d | 168 | end = start + XFS_B_TO_FSBT(mp, range.len) - 1; |
a46db608 CH |
169 | minlen = XFS_B_TO_FSB(mp, max_t(u64, granularity, range.minlen)); |
170 | ||
c029a50d | 171 | if (start >= mp->m_sb.sb_dblocks) |
a46db608 | 172 | return -XFS_ERROR(EINVAL); |
c029a50d LC |
173 | if (end > mp->m_sb.sb_dblocks - 1) |
174 | end = mp->m_sb.sb_dblocks - 1; | |
a46db608 | 175 | |
c029a50d LC |
176 | start_agno = XFS_FSB_TO_AGNO(mp, start); |
177 | end_agno = XFS_FSB_TO_AGNO(mp, end); | |
a46db608 CH |
178 | |
179 | for (agno = start_agno; agno <= end_agno; agno++) { | |
c029a50d | 180 | error = -xfs_trim_extents(mp, agno, start, end, minlen, |
a46db608 CH |
181 | &blocks_trimmed); |
182 | if (error) | |
183 | last_error = error; | |
184 | } | |
185 | ||
186 | if (last_error) | |
187 | return last_error; | |
188 | ||
189 | range.len = XFS_FSB_TO_B(mp, blocks_trimmed); | |
190 | if (copy_to_user(urange, &range, sizeof(range))) | |
191 | return -XFS_ERROR(EFAULT); | |
192 | return 0; | |
193 | } | |
e84661aa CH |
194 | |
195 | int | |
196 | xfs_discard_extents( | |
197 | struct xfs_mount *mp, | |
198 | struct list_head *list) | |
199 | { | |
200 | struct xfs_busy_extent *busyp; | |
201 | int error = 0; | |
202 | ||
203 | list_for_each_entry(busyp, list, list) { | |
204 | trace_xfs_discard_extent(mp, busyp->agno, busyp->bno, | |
205 | busyp->length); | |
206 | ||
207 | error = -blkdev_issue_discard(mp->m_ddev_targp->bt_bdev, | |
208 | XFS_AGB_TO_DADDR(mp, busyp->agno, busyp->bno), | |
209 | XFS_FSB_TO_BB(mp, busyp->length), | |
210 | GFP_NOFS, 0); | |
211 | if (error && error != EOPNOTSUPP) { | |
212 | xfs_info(mp, | |
213 | "discard failed for extent [0x%llu,%u], error %d", | |
214 | (unsigned long long)busyp->bno, | |
215 | busyp->length, | |
216 | error); | |
217 | return error; | |
218 | } | |
219 | } | |
220 | ||
221 | return 0; | |
222 | } |