blob: 272c3f8b6f7d0f11a0564e40b00b0568a28683dd [file] [log] [blame]
Christoph Hellwiga46db602011-01-07 13:02:04 +00001/*
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"
Dave Chinner6ca1c902013-08-12 20:49:26 +100019#include "xfs_format.h"
Dave Chinner239880e2013-10-23 10:50:10 +110020#include "xfs_log_format.h"
21#include "xfs_trans_resv.h"
Dave Chinner7fd36c42013-08-12 20:49:32 +100022#include "xfs_sb.h"
Christoph Hellwiga46db602011-01-07 13:02:04 +000023#include "xfs_mount.h"
24#include "xfs_quota.h"
Christoph Hellwiga46db602011-01-07 13:02:04 +000025#include "xfs_inode.h"
Dave Chinnera4fbe6a2013-10-23 10:51:50 +110026#include "xfs_btree.h"
27#include "xfs_alloc_btree.h"
Christoph Hellwiga46db602011-01-07 13:02:04 +000028#include "xfs_alloc.h"
29#include "xfs_error.h"
Dave Chinnerefc27b52012-04-29 10:39:43 +000030#include "xfs_extent_busy.h"
Christoph Hellwiga46db602011-01-07 13:02:04 +000031#include "xfs_discard.h"
32#include "xfs_trace.h"
Dave Chinner239880e2013-10-23 10:50:10 +110033#include "xfs_log.h"
Christoph Hellwiga46db602011-01-07 13:02:04 +000034
35STATIC int
36xfs_trim_extents(
37 struct xfs_mount *mp,
38 xfs_agnumber_t agno,
Dave Chinnera66d6362012-03-22 05:15:12 +000039 xfs_daddr_t start,
40 xfs_daddr_t end,
41 xfs_daddr_t minlen,
Christoph Hellwiga46db602011-01-07 13:02:04 +000042 __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 */
Dave Chinnera66d6362012-03-22 05:15:12 +000069 error = xfs_alloc_lookup_ge(cur, 0,
Dave Chinnerb1c770c2011-12-21 00:07:42 +000070 be32_to_cpu(XFS_BUF_TO_AGF(agbp)->agf_longest), &i);
Christoph Hellwiga46db602011-01-07 13:02:04 +000071 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) {
Dave Chinnera66d6362012-03-22 05:15:12 +000079 xfs_agblock_t fbno;
80 xfs_extlen_t flen;
81 xfs_daddr_t dbno;
82 xfs_extlen_t dlen;
Christoph Hellwiga46db602011-01-07 13:02:04 +000083
84 error = xfs_alloc_get_rec(cur, &fbno, &flen, &i);
85 if (error)
86 goto out_del_cursor;
Eric Sandeenc29aad42015-02-23 22:39:08 +110087 XFS_WANT_CORRUPTED_GOTO(mp, i == 1, out_del_cursor);
Dave Chinnerb1c770c2011-12-21 00:07:42 +000088 ASSERT(flen <= be32_to_cpu(XFS_BUF_TO_AGF(agbp)->agf_longest));
Christoph Hellwiga46db602011-01-07 13:02:04 +000089
90 /*
Dave Chinnera66d6362012-03-22 05:15:12 +000091 * 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 /*
Christoph Hellwiga46db602011-01-07 13:02:04 +000099 * Too small? Give up.
100 */
Dave Chinnera66d6362012-03-22 05:15:12 +0000101 if (dlen < minlen) {
Christoph Hellwiga46db602011-01-07 13:02:04 +0000102 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 */
Dave Chinnera66d6362012-03-22 05:15:12 +0000111 if (dbno + dlen < start || dbno > end) {
Christoph Hellwiga46db602011-01-07 13:02:04 +0000112 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 */
Dave Chinner4ecbfe62012-04-29 10:41:10 +0000120 if (xfs_extent_busy_search(mp, agno, fbno, flen)) {
Christoph Hellwiga46db602011-01-07 13:02:04 +0000121 trace_xfs_discard_busy(mp, agno, fbno, flen);
122 goto next_extent;
123 }
124
125 trace_xfs_discard_extent(mp, agno, fbno, flen);
Dave Chinner24513372014-06-25 14:58:08 +1000126 error = blkdev_issue_discard(bdev, dbno, dlen, GFP_NOFS, 0);
Christoph Hellwiga46db602011-01-07 13:02:04 +0000127 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
Dave Chinnera66d6362012-03-22 05:15:12 +0000145/*
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 */
Christoph Hellwiga46db602011-01-07 13:02:04 +0000154int
155xfs_ioc_trim(
156 struct xfs_mount *mp,
157 struct fstrim_range __user *urange)
158{
Jie Liu2f42d612013-11-20 16:08:53 +0800159 struct request_queue *q = bdev_get_queue(mp->m_ddev_targp->bt_bdev);
Christoph Hellwiga46db602011-01-07 13:02:04 +0000160 unsigned int granularity = q->limits.discard_granularity;
161 struct fstrim_range range;
Dave Chinnera66d6362012-03-22 05:15:12 +0000162 xfs_daddr_t start, end, minlen;
Christoph Hellwiga46db602011-01-07 13:02:04 +0000163 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))
Eric Sandeenb474c7a2014-06-22 15:04:54 +1000168 return -EPERM;
Lukas Czernerbe715142011-02-15 17:07:36 +0000169 if (!blk_queue_discard(q))
Eric Sandeenb474c7a2014-06-22 15:04:54 +1000170 return -EOPNOTSUPP;
Christoph Hellwiga46db602011-01-07 13:02:04 +0000171 if (copy_from_user(&range, urange, sizeof(range)))
Eric Sandeenb474c7a2014-06-22 15:04:54 +1000172 return -EFAULT;
Christoph Hellwiga46db602011-01-07 13:02:04 +0000173
174 /*
175 * Truncating down the len isn't actually quite correct, but using
Dave Chinnera66d6362012-03-22 05:15:12 +0000176 * BBTOB would mean we trivially get overflows for values
Christoph Hellwiga46db602011-01-07 13:02:04 +0000177 * 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 */
Tomas Raceka672e1b2012-08-14 10:35:04 +0200181 if (range.start >= XFS_FSB_TO_B(mp, mp->m_sb.sb_dblocks) ||
Jie Liu2f42d612013-11-20 16:08:53 +0800182 range.minlen > XFS_FSB_TO_B(mp, XFS_ALLOC_AG_MAX_USABLE(mp)) ||
183 range.len < mp->m_sb.sb_blocksize)
Eric Sandeenb474c7a2014-06-22 15:04:54 +1000184 return -EINVAL;
Tomas Raceka672e1b2012-08-14 10:35:04 +0200185
Dave Chinnera66d6362012-03-22 05:15:12 +0000186 start = BTOBB(range.start);
187 end = start + BTOBBT(range.len) - 1;
188 minlen = BTOBB(max_t(u64, granularity, range.minlen));
Christoph Hellwiga46db602011-01-07 13:02:04 +0000189
Dave Chinnera66d6362012-03-22 05:15:12 +0000190 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;
Christoph Hellwiga46db602011-01-07 13:02:04 +0000192
Dave Chinnera66d6362012-03-22 05:15:12 +0000193 start_agno = xfs_daddr_to_agno(mp, start);
194 end_agno = xfs_daddr_to_agno(mp, end);
Christoph Hellwiga46db602011-01-07 13:02:04 +0000195
196 for (agno = start_agno; agno <= end_agno; agno++) {
Dave Chinner24513372014-06-25 14:58:08 +1000197 error = xfs_trim_extents(mp, agno, start, end, minlen,
Christoph Hellwiga46db602011-01-07 13:02:04 +0000198 &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)))
Eric Sandeenb474c7a2014-06-22 15:04:54 +1000208 return -EFAULT;
Christoph Hellwiga46db602011-01-07 13:02:04 +0000209 return 0;
210}
Christoph Hellwige84661a2011-05-20 13:45:32 +0000211
212int
213xfs_discard_extents(
214 struct xfs_mount *mp,
215 struct list_head *list)
216{
Dave Chinner4ecbfe62012-04-29 10:41:10 +0000217 struct xfs_extent_busy *busyp;
Christoph Hellwige84661a2011-05-20 13:45:32 +0000218 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
Dave Chinner24513372014-06-25 14:58:08 +1000224 error = blkdev_issue_discard(mp->m_ddev_targp->bt_bdev,
Christoph Hellwige84661a2011-05-20 13:45:32 +0000225 XFS_AGB_TO_DADDR(mp, busyp->agno, busyp->bno),
226 XFS_FSB_TO_BB(mp, busyp->length),
227 GFP_NOFS, 0);
Dave Chinner24513372014-06-25 14:58:08 +1000228 if (error && error != -EOPNOTSUPP) {
Christoph Hellwige84661a2011-05-20 13:45:32 +0000229 xfs_info(mp,
Colin Ian King5d518bd2016-03-02 09:57:04 +1100230 "discard failed for extent [0x%llx,%u], error %d",
Christoph Hellwige84661a2011-05-20 13:45:32 +0000231 (unsigned long long)busyp->bno,
232 busyp->length,
233 error);
234 return error;
235 }
236 }
237
238 return 0;
239}