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" | |
6ca1c906 | 19 | #include "xfs_format.h" |
239880ef DC |
20 | #include "xfs_log_format.h" |
21 | #include "xfs_trans_resv.h" | |
7fd36c44 | 22 | #include "xfs_sb.h" |
a46db608 CH |
23 | #include "xfs_mount.h" |
24 | #include "xfs_quota.h" | |
a46db608 | 25 | #include "xfs_inode.h" |
a4fbe6ab DC |
26 | #include "xfs_btree.h" |
27 | #include "xfs_alloc_btree.h" | |
a46db608 CH |
28 | #include "xfs_alloc.h" |
29 | #include "xfs_error.h" | |
efc27b52 | 30 | #include "xfs_extent_busy.h" |
a46db608 CH |
31 | #include "xfs_discard.h" |
32 | #include "xfs_trace.h" | |
239880ef | 33 | #include "xfs_log.h" |
a46db608 CH |
34 | |
35 | STATIC int | |
36 | xfs_trim_extents( | |
37 | struct xfs_mount *mp, | |
38 | xfs_agnumber_t agno, | |
a66d6363 DC |
39 | xfs_daddr_t start, |
40 | xfs_daddr_t end, | |
41 | xfs_daddr_t minlen, | |
c8ce540d | 42 | uint64_t *blocks_trimmed) |
a46db608 CH |
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 | */ | |
a66d6363 | 69 | error = xfs_alloc_lookup_ge(cur, 0, |
b1c770c2 | 70 | be32_to_cpu(XFS_BUF_TO_AGF(agbp)->agf_longest), &i); |
a46db608 CH |
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) { | |
a66d6363 DC |
79 | xfs_agblock_t fbno; |
80 | xfs_extlen_t flen; | |
81 | xfs_daddr_t dbno; | |
82 | xfs_extlen_t dlen; | |
a46db608 CH |
83 | |
84 | error = xfs_alloc_get_rec(cur, &fbno, &flen, &i); | |
85 | if (error) | |
86 | goto out_del_cursor; | |
c29aad41 | 87 | XFS_WANT_CORRUPTED_GOTO(mp, i == 1, out_del_cursor); |
b1c770c2 | 88 | ASSERT(flen <= be32_to_cpu(XFS_BUF_TO_AGF(agbp)->agf_longest)); |
a46db608 | 89 | |
a66d6363 DC |
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 | ||
a46db608 CH |
98 | /* |
99 | * Too small? Give up. | |
100 | */ | |
a66d6363 | 101 | if (dlen < minlen) { |
a46db608 CH |
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 | */ | |
a66d6363 | 111 | if (dbno + dlen < start || dbno > end) { |
a46db608 CH |
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 | */ | |
4ecbfe63 | 120 | if (xfs_extent_busy_search(mp, agno, fbno, flen)) { |
a46db608 CH |
121 | trace_xfs_discard_busy(mp, agno, fbno, flen); |
122 | goto next_extent; | |
123 | } | |
124 | ||
125 | trace_xfs_discard_extent(mp, agno, fbno, flen); | |
2451337d | 126 | error = blkdev_issue_discard(bdev, dbno, dlen, GFP_NOFS, 0); |
a46db608 CH |
127 | if (error) |
128 | goto out_del_cursor; | |
129 | *blocks_trimmed += flen; | |
130 | ||
131 | next_extent: | |
132 | error = xfs_btree_decrement(cur, 0, &i); | |
133 | if (error) | |
134 | goto out_del_cursor; | |
3c378195 LC |
135 | |
136 | if (fatal_signal_pending(current)) { | |
137 | error = -ERESTARTSYS; | |
138 | goto out_del_cursor; | |
139 | } | |
a46db608 CH |
140 | } |
141 | ||
142 | out_del_cursor: | |
143 | xfs_btree_del_cursor(cur, error ? XFS_BTREE_ERROR : XFS_BTREE_NOERROR); | |
144 | xfs_buf_relse(agbp); | |
145 | out_put_perag: | |
146 | xfs_perag_put(pag); | |
147 | return error; | |
148 | } | |
149 | ||
a66d6363 DC |
150 | /* |
151 | * trim a range of the filesystem. | |
152 | * | |
153 | * Note: the parameters passed from userspace are byte ranges into the | |
154 | * filesystem which does not match to the format we use for filesystem block | |
155 | * addressing. FSB addressing is sparse (AGNO|AGBNO), while the incoming format | |
156 | * is a linear address range. Hence we need to use DADDR based conversions and | |
157 | * comparisons for determining the correct offset and regions to trim. | |
158 | */ | |
a46db608 CH |
159 | int |
160 | xfs_ioc_trim( | |
161 | struct xfs_mount *mp, | |
162 | struct fstrim_range __user *urange) | |
163 | { | |
2f42d612 | 164 | struct request_queue *q = bdev_get_queue(mp->m_ddev_targp->bt_bdev); |
a46db608 CH |
165 | unsigned int granularity = q->limits.discard_granularity; |
166 | struct fstrim_range range; | |
a66d6363 | 167 | xfs_daddr_t start, end, minlen; |
a46db608 | 168 | xfs_agnumber_t start_agno, end_agno, agno; |
c8ce540d | 169 | uint64_t blocks_trimmed = 0; |
a46db608 CH |
170 | int error, last_error = 0; |
171 | ||
172 | if (!capable(CAP_SYS_ADMIN)) | |
b474c7ae | 173 | return -EPERM; |
be715140 | 174 | if (!blk_queue_discard(q)) |
b474c7ae | 175 | return -EOPNOTSUPP; |
a46db608 | 176 | if (copy_from_user(&range, urange, sizeof(range))) |
b474c7ae | 177 | return -EFAULT; |
a46db608 CH |
178 | |
179 | /* | |
180 | * Truncating down the len isn't actually quite correct, but using | |
a66d6363 | 181 | * BBTOB would mean we trivially get overflows for values |
a46db608 CH |
182 | * of ULLONG_MAX or slightly lower. And ULLONG_MAX is the default |
183 | * used by the fstrim application. In the end it really doesn't | |
184 | * matter as trimming blocks is an advisory interface. | |
185 | */ | |
a672e1be | 186 | if (range.start >= XFS_FSB_TO_B(mp, mp->m_sb.sb_dblocks) || |
52548852 | 187 | range.minlen > XFS_FSB_TO_B(mp, mp->m_ag_max_usable) || |
2f42d612 | 188 | range.len < mp->m_sb.sb_blocksize) |
b474c7ae | 189 | return -EINVAL; |
a672e1be | 190 | |
a66d6363 DC |
191 | start = BTOBB(range.start); |
192 | end = start + BTOBBT(range.len) - 1; | |
193 | minlen = BTOBB(max_t(u64, granularity, range.minlen)); | |
a46db608 | 194 | |
a66d6363 DC |
195 | if (end > XFS_FSB_TO_BB(mp, mp->m_sb.sb_dblocks) - 1) |
196 | end = XFS_FSB_TO_BB(mp, mp->m_sb.sb_dblocks)- 1; | |
a46db608 | 197 | |
a66d6363 DC |
198 | start_agno = xfs_daddr_to_agno(mp, start); |
199 | end_agno = xfs_daddr_to_agno(mp, end); | |
a46db608 CH |
200 | |
201 | for (agno = start_agno; agno <= end_agno; agno++) { | |
2451337d | 202 | error = xfs_trim_extents(mp, agno, start, end, minlen, |
a46db608 | 203 | &blocks_trimmed); |
3c378195 | 204 | if (error) { |
a46db608 | 205 | last_error = error; |
3c378195 LC |
206 | if (error == -ERESTARTSYS) |
207 | break; | |
208 | } | |
a46db608 CH |
209 | } |
210 | ||
211 | if (last_error) | |
212 | return last_error; | |
213 | ||
214 | range.len = XFS_FSB_TO_B(mp, blocks_trimmed); | |
215 | if (copy_to_user(urange, &range, sizeof(range))) | |
b474c7ae | 216 | return -EFAULT; |
a46db608 CH |
217 | return 0; |
218 | } |