Christoph Hellwig | a46db60 | 2011-01-07 13:02:04 +0000 | [diff] [blame] | 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, |
Dave Chinner | a66d636 | 2012-03-22 05:15:12 +0000 | [diff] [blame] | 40 | xfs_daddr_t start, |
| 41 | xfs_daddr_t end, |
| 42 | xfs_daddr_t minlen, |
Christoph Hellwig | a46db60 | 2011-01-07 13:02:04 +0000 | [diff] [blame] | 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 | */ |
Dave Chinner | a66d636 | 2012-03-22 05:15:12 +0000 | [diff] [blame] | 70 | error = xfs_alloc_lookup_ge(cur, 0, |
Dave Chinner | b1c770c | 2011-12-21 00:07:42 +0000 | [diff] [blame] | 71 | be32_to_cpu(XFS_BUF_TO_AGF(agbp)->agf_longest), &i); |
Christoph Hellwig | a46db60 | 2011-01-07 13:02:04 +0000 | [diff] [blame] | 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) { |
Dave Chinner | a66d636 | 2012-03-22 05:15:12 +0000 | [diff] [blame] | 80 | xfs_agblock_t fbno; |
| 81 | xfs_extlen_t flen; |
| 82 | xfs_daddr_t dbno; |
| 83 | xfs_extlen_t dlen; |
Christoph Hellwig | a46db60 | 2011-01-07 13:02:04 +0000 | [diff] [blame] | 84 | |
| 85 | error = xfs_alloc_get_rec(cur, &fbno, &flen, &i); |
| 86 | if (error) |
| 87 | goto out_del_cursor; |
| 88 | XFS_WANT_CORRUPTED_GOTO(i == 1, out_del_cursor); |
Dave Chinner | b1c770c | 2011-12-21 00:07:42 +0000 | [diff] [blame] | 89 | ASSERT(flen <= be32_to_cpu(XFS_BUF_TO_AGF(agbp)->agf_longest)); |
Christoph Hellwig | a46db60 | 2011-01-07 13:02:04 +0000 | [diff] [blame] | 90 | |
| 91 | /* |
Dave Chinner | a66d636 | 2012-03-22 05:15:12 +0000 | [diff] [blame] | 92 | * use daddr format for all range/len calculations as that is |
| 93 | * the format the range/len variables are supplied in by |
| 94 | * userspace. |
| 95 | */ |
| 96 | dbno = XFS_AGB_TO_DADDR(mp, agno, fbno); |
| 97 | dlen = XFS_FSB_TO_BB(mp, flen); |
| 98 | |
| 99 | /* |
Christoph Hellwig | a46db60 | 2011-01-07 13:02:04 +0000 | [diff] [blame] | 100 | * Too small? Give up. |
| 101 | */ |
Dave Chinner | a66d636 | 2012-03-22 05:15:12 +0000 | [diff] [blame] | 102 | if (dlen < minlen) { |
Christoph Hellwig | a46db60 | 2011-01-07 13:02:04 +0000 | [diff] [blame] | 103 | trace_xfs_discard_toosmall(mp, agno, fbno, flen); |
| 104 | goto out_del_cursor; |
| 105 | } |
| 106 | |
| 107 | /* |
| 108 | * If the extent is entirely outside of the range we are |
| 109 | * supposed to discard skip it. Do not bother to trim |
| 110 | * down partially overlapping ranges for now. |
| 111 | */ |
Dave Chinner | a66d636 | 2012-03-22 05:15:12 +0000 | [diff] [blame] | 112 | if (dbno + dlen < start || dbno > end) { |
Christoph Hellwig | a46db60 | 2011-01-07 13:02:04 +0000 | [diff] [blame] | 113 | trace_xfs_discard_exclude(mp, agno, fbno, flen); |
| 114 | goto next_extent; |
| 115 | } |
| 116 | |
| 117 | /* |
| 118 | * If any blocks in the range are still busy, skip the |
| 119 | * discard and try again the next time. |
| 120 | */ |
| 121 | if (xfs_alloc_busy_search(mp, agno, fbno, flen)) { |
| 122 | trace_xfs_discard_busy(mp, agno, fbno, flen); |
| 123 | goto next_extent; |
| 124 | } |
| 125 | |
| 126 | trace_xfs_discard_extent(mp, agno, fbno, flen); |
Dave Chinner | a66d636 | 2012-03-22 05:15:12 +0000 | [diff] [blame] | 127 | error = -blkdev_issue_discard(bdev, dbno, dlen, GFP_NOFS, 0); |
Christoph Hellwig | a46db60 | 2011-01-07 13:02:04 +0000 | [diff] [blame] | 128 | if (error) |
| 129 | goto out_del_cursor; |
| 130 | *blocks_trimmed += flen; |
| 131 | |
| 132 | next_extent: |
| 133 | error = xfs_btree_decrement(cur, 0, &i); |
| 134 | if (error) |
| 135 | goto out_del_cursor; |
| 136 | } |
| 137 | |
| 138 | out_del_cursor: |
| 139 | xfs_btree_del_cursor(cur, error ? XFS_BTREE_ERROR : XFS_BTREE_NOERROR); |
| 140 | xfs_buf_relse(agbp); |
| 141 | out_put_perag: |
| 142 | xfs_perag_put(pag); |
| 143 | return error; |
| 144 | } |
| 145 | |
Dave Chinner | a66d636 | 2012-03-22 05:15:12 +0000 | [diff] [blame] | 146 | /* |
| 147 | * trim a range of the filesystem. |
| 148 | * |
| 149 | * Note: the parameters passed from userspace are byte ranges into the |
| 150 | * filesystem which does not match to the format we use for filesystem block |
| 151 | * addressing. FSB addressing is sparse (AGNO|AGBNO), while the incoming format |
| 152 | * is a linear address range. Hence we need to use DADDR based conversions and |
| 153 | * comparisons for determining the correct offset and regions to trim. |
| 154 | */ |
Christoph Hellwig | a46db60 | 2011-01-07 13:02:04 +0000 | [diff] [blame] | 155 | int |
| 156 | xfs_ioc_trim( |
| 157 | struct xfs_mount *mp, |
| 158 | struct fstrim_range __user *urange) |
| 159 | { |
| 160 | struct request_queue *q = mp->m_ddev_targp->bt_bdev->bd_disk->queue; |
| 161 | unsigned int granularity = q->limits.discard_granularity; |
| 162 | struct fstrim_range range; |
Dave Chinner | a66d636 | 2012-03-22 05:15:12 +0000 | [diff] [blame] | 163 | xfs_daddr_t start, end, minlen; |
Christoph Hellwig | a46db60 | 2011-01-07 13:02:04 +0000 | [diff] [blame] | 164 | xfs_agnumber_t start_agno, end_agno, agno; |
| 165 | __uint64_t blocks_trimmed = 0; |
| 166 | int error, last_error = 0; |
| 167 | |
| 168 | if (!capable(CAP_SYS_ADMIN)) |
| 169 | return -XFS_ERROR(EPERM); |
Lukas Czerner | be71514 | 2011-02-15 17:07:36 +0000 | [diff] [blame] | 170 | if (!blk_queue_discard(q)) |
| 171 | return -XFS_ERROR(EOPNOTSUPP); |
Christoph Hellwig | a46db60 | 2011-01-07 13:02:04 +0000 | [diff] [blame] | 172 | if (copy_from_user(&range, urange, sizeof(range))) |
| 173 | return -XFS_ERROR(EFAULT); |
| 174 | |
| 175 | /* |
| 176 | * Truncating down the len isn't actually quite correct, but using |
Dave Chinner | a66d636 | 2012-03-22 05:15:12 +0000 | [diff] [blame] | 177 | * BBTOB would mean we trivially get overflows for values |
Christoph Hellwig | a46db60 | 2011-01-07 13:02:04 +0000 | [diff] [blame] | 178 | * of ULLONG_MAX or slightly lower. And ULLONG_MAX is the default |
| 179 | * used by the fstrim application. In the end it really doesn't |
| 180 | * matter as trimming blocks is an advisory interface. |
| 181 | */ |
Dave Chinner | a66d636 | 2012-03-22 05:15:12 +0000 | [diff] [blame] | 182 | start = BTOBB(range.start); |
| 183 | end = start + BTOBBT(range.len) - 1; |
| 184 | minlen = BTOBB(max_t(u64, granularity, range.minlen)); |
Christoph Hellwig | a46db60 | 2011-01-07 13:02:04 +0000 | [diff] [blame] | 185 | |
Dave Chinner | a66d636 | 2012-03-22 05:15:12 +0000 | [diff] [blame] | 186 | if (XFS_BB_TO_FSB(mp, start) >= mp->m_sb.sb_dblocks) |
Christoph Hellwig | a46db60 | 2011-01-07 13:02:04 +0000 | [diff] [blame] | 187 | return -XFS_ERROR(EINVAL); |
Dave Chinner | a66d636 | 2012-03-22 05:15:12 +0000 | [diff] [blame] | 188 | if (end > XFS_FSB_TO_BB(mp, mp->m_sb.sb_dblocks) - 1) |
| 189 | end = XFS_FSB_TO_BB(mp, mp->m_sb.sb_dblocks)- 1; |
Christoph Hellwig | a46db60 | 2011-01-07 13:02:04 +0000 | [diff] [blame] | 190 | |
Dave Chinner | a66d636 | 2012-03-22 05:15:12 +0000 | [diff] [blame] | 191 | start_agno = xfs_daddr_to_agno(mp, start); |
| 192 | end_agno = xfs_daddr_to_agno(mp, end); |
Christoph Hellwig | a46db60 | 2011-01-07 13:02:04 +0000 | [diff] [blame] | 193 | |
| 194 | for (agno = start_agno; agno <= end_agno; agno++) { |
Lukas Czerner | c029a50 | 2011-09-21 09:42:30 +0000 | [diff] [blame] | 195 | error = -xfs_trim_extents(mp, agno, start, end, minlen, |
Christoph Hellwig | a46db60 | 2011-01-07 13:02:04 +0000 | [diff] [blame] | 196 | &blocks_trimmed); |
| 197 | if (error) |
| 198 | last_error = error; |
| 199 | } |
| 200 | |
| 201 | if (last_error) |
| 202 | return last_error; |
| 203 | |
| 204 | range.len = XFS_FSB_TO_B(mp, blocks_trimmed); |
| 205 | if (copy_to_user(urange, &range, sizeof(range))) |
| 206 | return -XFS_ERROR(EFAULT); |
| 207 | return 0; |
| 208 | } |
Christoph Hellwig | e84661a | 2011-05-20 13:45:32 +0000 | [diff] [blame] | 209 | |
| 210 | int |
| 211 | xfs_discard_extents( |
| 212 | struct xfs_mount *mp, |
| 213 | struct list_head *list) |
| 214 | { |
| 215 | struct xfs_busy_extent *busyp; |
| 216 | int error = 0; |
| 217 | |
| 218 | list_for_each_entry(busyp, list, list) { |
| 219 | trace_xfs_discard_extent(mp, busyp->agno, busyp->bno, |
| 220 | busyp->length); |
| 221 | |
| 222 | error = -blkdev_issue_discard(mp->m_ddev_targp->bt_bdev, |
| 223 | XFS_AGB_TO_DADDR(mp, busyp->agno, busyp->bno), |
| 224 | XFS_FSB_TO_BB(mp, busyp->length), |
| 225 | GFP_NOFS, 0); |
| 226 | if (error && error != EOPNOTSUPP) { |
| 227 | xfs_info(mp, |
| 228 | "discard failed for extent [0x%llu,%u], error %d", |
| 229 | (unsigned long long)busyp->bno, |
| 230 | busyp->length, |
| 231 | error); |
| 232 | return error; |
| 233 | } |
| 234 | } |
| 235 | |
| 236 | return 0; |
| 237 | } |