fs/xfs/xfs_discard.c

   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_format.h"
  20 #include "xfs_log.h"
  21 #include "xfs_trans.h"
  22 #include "xfs_sb.h"
  23 #include "xfs_ag.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"
  36
  37 STATIC int
  38 xfs_trim_extents(
  39         struct xfs_mount        *mp,
  40         xfs_agnumber_t          agno,
  41         xfs_daddr_t             start,
  42         xfs_daddr_t             end,
  43         xfs_daddr_t             minlen,
  44         __uint64_t              *blocks_trimmed)
  45 {
  46         struct block_device     *bdev = mp->m_ddev_targp->bt_bdev;
  47         struct xfs_btree_cur    *cur;
  48         struct xfs_buf          *agbp;
  49         struct xfs_perag        *pag;
  50         int                     error;
  51         int                     i;
  52
  53         pag = xfs_perag_get(mp, agno);
  54
  55         error = xfs_alloc_read_agf(mp, NULL, agno, 0, &agbp);
  56         if (error || !agbp)
  57                 goto out_put_perag;
  58
  59         cur = xfs_allocbt_init_cursor(mp, NULL, agbp, agno, XFS_BTNUM_CNT);
  60
  61         /*
  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.
  65          */
  66         xfs_log_force(mp, XFS_LOG_SYNC);
  67
  68         /*
  69          * Look up the longest btree in the AGF and start with it.
  70          */
  71         error = xfs_alloc_lookup_ge(cur, 0,
  72                             be32_to_cpu(XFS_BUF_TO_AGF(agbp)->agf_longest), &i);
  73         if (error)
  74                 goto out_del_cursor;
  75
  76         /*
  77          * Loop until we are done with all extents that are large
  78          * enough to be worth discarding.
  79          */
  80         while (i) {
  81                 xfs_agblock_t   fbno;
  82                 xfs_extlen_t    flen;
  83                 xfs_daddr_t     dbno;
  84                 xfs_extlen_t    dlen;
  85
  86                 error = xfs_alloc_get_rec(cur, &fbno, &flen, &i);
  87                 if (error)
  88                         goto out_del_cursor;
  89                 XFS_WANT_CORRUPTED_GOTO(i == 1, out_del_cursor);
  90                 ASSERT(flen <= be32_to_cpu(XFS_BUF_TO_AGF(agbp)->agf_longest));
  91
  92                 /*
  93                  * use daddr format for all range/len calculations as that is
  94                  * the format the range/len variables are supplied in by
  95                  * userspace.
  96                  */
  97                 dbno = XFS_AGB_TO_DADDR(mp, agno, fbno);
  98                 dlen = XFS_FSB_TO_BB(mp, flen);
  99
 100                 /*
 101                  * Too small?  Give up.
 102                  */
 103                 if (dlen < minlen) {
 104                         trace_xfs_discard_toosmall(mp, agno, fbno, flen);
 105                         goto out_del_cursor;
 106                 }
 107
 108                 /*
 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.
 112                  */
 113                 if (dbno + dlen < start || dbno > end) {
 114                         trace_xfs_discard_exclude(mp, agno, fbno, flen);
 115                         goto next_extent;
 116                 }
 117
 118                 /*
 119                  * If any blocks in the range are still busy, skip the
 120                  * discard and try again the next time.
 121                  */
 122                 if (xfs_extent_busy_search(mp, agno, fbno, flen)) {
 123                         trace_xfs_discard_busy(mp, agno, fbno, flen);
 124                         goto next_extent;
 125                 }
 126
 127                 trace_xfs_discard_extent(mp, agno, fbno, flen);
 128                 error = -blkdev_issue_discard(bdev, dbno, dlen, GFP_NOFS, 0);
 129                 if (error)
 130                         goto out_del_cursor;
 131                 *blocks_trimmed += flen;
 132
 133 next_extent:
 134                 error = xfs_btree_decrement(cur, 0, &i);
 135                 if (error)
 136                         goto out_del_cursor;
 137         }
 138
 139 out_del_cursor:
 140         xfs_btree_del_cursor(cur, error ? XFS_BTREE_ERROR : XFS_BTREE_NOERROR);
 141         xfs_buf_relse(agbp);
 142 out_put_perag:
 143         xfs_perag_put(pag);
 144         return error;
 145 }
 146
 147 /*
 148  * trim a range of the filesystem.
 149  *
 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.
 155  */
 156 int
 157 xfs_ioc_trim(
 158         struct xfs_mount                *mp,
 159         struct fstrim_range __user      *urange)
 160 {
 161         struct request_queue    *q = bdev_get_queue(mp->m_ddev_targp->bt_bdev);
 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;
 168
 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);
 175
 176         /*
 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.
 182          */
 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             range.len < mp->m_sb.sb_blocksize)
 186                 return -XFS_ERROR(EINVAL);
 187
 188         start = BTOBB(range.start);
 189         end = start + BTOBBT(range.len) - 1;
 190         minlen = BTOBB(max_t(u64, granularity, range.minlen));
 191
 192         if (end > XFS_FSB_TO_BB(mp, mp->m_sb.sb_dblocks) - 1)
 193                 end = XFS_FSB_TO_BB(mp, mp->m_sb.sb_dblocks)- 1;
 194
 195         start_agno = xfs_daddr_to_agno(mp, start);
 196         end_agno = xfs_daddr_to_agno(mp, end);
 197
 198         for (agno = start_agno; agno <= end_agno; agno++) {
 199                 error = -xfs_trim_extents(mp, agno, start, end, minlen,
 200                                           &blocks_trimmed);
 201                 if (error)
 202                         last_error = error;
 203         }
 204
 205         if (last_error)
 206                 return last_error;
 207
 208         range.len = XFS_FSB_TO_B(mp, blocks_trimmed);
 209         if (copy_to_user(urange, &range, sizeof(range)))
 210                 return -XFS_ERROR(EFAULT);
 211         return 0;
 212 }
 213
 214 int
 215 xfs_discard_extents(
 216         struct xfs_mount        *mp,
 217         struct list_head        *list)
 218 {
 219         struct xfs_extent_busy  *busyp;
 220         int                     error = 0;
 221
 222         list_for_each_entry(busyp, list, list) {
 223                 trace_xfs_discard_extent(mp, busyp->agno, busyp->bno,
 224                                          busyp->length);
 225
 226                 error = -blkdev_issue_discard(mp->m_ddev_targp->bt_bdev,
 227                                 XFS_AGB_TO_DADDR(mp, busyp->agno, busyp->bno),
 228                                 XFS_FSB_TO_BB(mp, busyp->length),
 229                                 GFP_NOFS, 0);
 230                 if (error && error != EOPNOTSUPP) {
 231                         xfs_info(mp,
 232          "discard failed for extent [0x%llu,%u], error %d",
 233                                  (unsigned long long)busyp->bno,
 234                                  busyp->length,
 235                                  error);
 236                         return error;
 237                 }
 238         }
 239
 240         return 0;
 241 }