--- /dev/null
+/*
+ * Copyright (C) 2017 Oracle.  All Rights Reserved.
+ *
+ * Author: Darrick J. Wong <darrick.wong@oracle.com>
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License
+ * as published by the Free Software Foundation; either version 2
+ * of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it would be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write the Free Software Foundation,
+ * Inc.,  51 Franklin St, Fifth Floor, Boston, MA  02110-1301, USA.
+ */
+#include "xfs.h"
+#include "xfs_fs.h"
+#include "xfs_shared.h"
+#include "xfs_format.h"
+#include "xfs_log_format.h"
+#include "xfs_trans_resv.h"
+#include "xfs_sb.h"
+#include "xfs_mount.h"
+#include "xfs_defer.h"
+#include "xfs_inode.h"
+#include "xfs_trans.h"
+#include "xfs_error.h"
+#include "xfs_btree.h"
+#include "xfs_rmap_btree.h"
+#include "xfs_trace.h"
+#include "xfs_log.h"
+#include "xfs_rmap.h"
+#include "xfs_alloc.h"
+#include "xfs_bit.h"
+#include <linux/fsmap.h>
+#include "xfs_fsmap.h"
+#include "xfs_refcount.h"
+#include "xfs_refcount_btree.h"
+
+/* Convert an xfs_fsmap to an fsmap. */
+void
+xfs_fsmap_from_internal(
+       struct fsmap            *dest,
+       struct xfs_fsmap        *src)
+{
+       dest->fmr_device = src->fmr_device;
+       dest->fmr_flags = src->fmr_flags;
+       dest->fmr_physical = BBTOB(src->fmr_physical);
+       dest->fmr_owner = src->fmr_owner;
+       dest->fmr_offset = BBTOB(src->fmr_offset);
+       dest->fmr_length = BBTOB(src->fmr_length);
+       dest->fmr_reserved[0] = 0;
+       dest->fmr_reserved[1] = 0;
+       dest->fmr_reserved[2] = 0;
+}
+
+/* Convert an fsmap to an xfs_fsmap. */
+void
+xfs_fsmap_to_internal(
+       struct xfs_fsmap        *dest,
+       struct fsmap            *src)
+{
+       dest->fmr_device = src->fmr_device;
+       dest->fmr_flags = src->fmr_flags;
+       dest->fmr_physical = BTOBBT(src->fmr_physical);
+       dest->fmr_owner = src->fmr_owner;
+       dest->fmr_offset = BTOBBT(src->fmr_offset);
+       dest->fmr_length = BTOBBT(src->fmr_length);
+}
+
+/* Convert an fsmap owner into an rmapbt owner. */
+static int
+xfs_fsmap_owner_to_rmap(
+       struct xfs_rmap_irec    *dest,
+       struct xfs_fsmap        *src)
+{
+       if (!(src->fmr_flags & FMR_OF_SPECIAL_OWNER)) {
+               dest->rm_owner = src->fmr_owner;
+               return 0;
+       }
+
+       switch (src->fmr_owner) {
+       case 0:                 /* "lowest owner id possible" */
+       case -1ULL:             /* "highest owner id possible" */
+               dest->rm_owner = 0;
+               break;
+       case XFS_FMR_OWN_FREE:
+               dest->rm_owner = XFS_RMAP_OWN_NULL;
+               break;
+       case XFS_FMR_OWN_UNKNOWN:
+               dest->rm_owner = XFS_RMAP_OWN_UNKNOWN;
+               break;
+       case XFS_FMR_OWN_FS:
+               dest->rm_owner = XFS_RMAP_OWN_FS;
+               break;
+       case XFS_FMR_OWN_LOG:
+               dest->rm_owner = XFS_RMAP_OWN_LOG;
+               break;
+       case XFS_FMR_OWN_AG:
+               dest->rm_owner = XFS_RMAP_OWN_AG;
+               break;
+       case XFS_FMR_OWN_INOBT:
+               dest->rm_owner = XFS_RMAP_OWN_INOBT;
+               break;
+       case XFS_FMR_OWN_INODES:
+               dest->rm_owner = XFS_RMAP_OWN_INODES;
+               break;
+       case XFS_FMR_OWN_REFC:
+               dest->rm_owner = XFS_RMAP_OWN_REFC;
+               break;
+       case XFS_FMR_OWN_COW:
+               dest->rm_owner = XFS_RMAP_OWN_COW;
+               break;
+       case XFS_FMR_OWN_DEFECTIVE:     /* not implemented */
+               /* fall through */
+       default:
+               return -EINVAL;
+       }
+       return 0;
+}
+
+/* Convert an rmapbt owner into an fsmap owner. */
+static int
+xfs_fsmap_owner_from_rmap(
+       struct xfs_fsmap        *dest,
+       struct xfs_rmap_irec    *src)
+{
+       dest->fmr_flags = 0;
+       if (!XFS_RMAP_NON_INODE_OWNER(src->rm_owner)) {
+               dest->fmr_owner = src->rm_owner;
+               return 0;
+       }
+       dest->fmr_flags |= FMR_OF_SPECIAL_OWNER;
+
+       switch (src->rm_owner) {
+       case XFS_RMAP_OWN_FS:
+               dest->fmr_owner = XFS_FMR_OWN_FS;
+               break;
+       case XFS_RMAP_OWN_LOG:
+               dest->fmr_owner = XFS_FMR_OWN_LOG;
+               break;
+       case XFS_RMAP_OWN_AG:
+               dest->fmr_owner = XFS_FMR_OWN_AG;
+               break;
+       case XFS_RMAP_OWN_INOBT:
+               dest->fmr_owner = XFS_FMR_OWN_INOBT;
+               break;
+       case XFS_RMAP_OWN_INODES:
+               dest->fmr_owner = XFS_FMR_OWN_INODES;
+               break;
+       case XFS_RMAP_OWN_REFC:
+               dest->fmr_owner = XFS_FMR_OWN_REFC;
+               break;
+       case XFS_RMAP_OWN_COW:
+               dest->fmr_owner = XFS_FMR_OWN_COW;
+               break;
+       default:
+               return -EFSCORRUPTED;
+       }
+       return 0;
+}
+
+/* getfsmap query state */
+struct xfs_getfsmap_info {
+       struct xfs_fsmap_head   *head;
+       xfs_fsmap_format_t      formatter;      /* formatting fn */
+       void                    *format_arg;    /* format buffer */
+       struct xfs_buf          *agf_bp;        /* AGF, for refcount queries */
+       xfs_daddr_t             next_daddr;     /* next daddr we expect */
+       u64                     missing_owner;  /* owner of holes */
+       u32                     dev;            /* device id */
+       xfs_agnumber_t          agno;           /* AG number, if applicable */
+       struct xfs_rmap_irec    low;            /* low rmap key */
+       struct xfs_rmap_irec    high;           /* high rmap key */
+       bool                    last;           /* last extent? */
+};
+
+/* Associate a device with a getfsmap handler. */
+struct xfs_getfsmap_dev {
+       u32                     dev;
+       int                     (*fn)(struct xfs_trans *tp,
+                                     struct xfs_fsmap *keys,
+                                     struct xfs_getfsmap_info *info);
+};
+
+/* Compare two getfsmap device handlers. */
+static int
+xfs_getfsmap_dev_compare(
+       const void                      *p1,
+       const void                      *p2)
+{
+       const struct xfs_getfsmap_dev   *d1 = p1;
+       const struct xfs_getfsmap_dev   *d2 = p2;
+
+       return d1->dev - d2->dev;
+}
+
+/* Decide if this mapping is shared. */
+STATIC int
+xfs_getfsmap_is_shared(
+       struct xfs_trans                *tp,
+       struct xfs_getfsmap_info        *info,
+       struct xfs_rmap_irec            *rec,
+       bool                            *stat)
+{
+       struct xfs_mount                *mp = tp->t_mountp;
+       struct xfs_btree_cur            *cur;
+       xfs_agblock_t                   fbno;
+       xfs_extlen_t                    flen;
+       int                             error;
+
+       *stat = false;
+       if (!xfs_sb_version_hasreflink(&mp->m_sb))
+               return 0;
+       /* rt files will have agno set to NULLAGNUMBER */
+       if (info->agno == NULLAGNUMBER)
+               return 0;
+
+       /* Are there any shared blocks here? */
+       flen = 0;
+       cur = xfs_refcountbt_init_cursor(mp, tp, info->agf_bp,
+                       info->agno, NULL);
+
+       error = xfs_refcount_find_shared(cur, rec->rm_startblock,
+                       rec->rm_blockcount, &fbno, &flen, false);
+
+       xfs_btree_del_cursor(cur, error ? XFS_BTREE_ERROR : XFS_BTREE_NOERROR);
+       if (error)
+               return error;
+
+       *stat = flen > 0;
+       return 0;
+}
+
+/*
+ * Format a reverse mapping for getfsmap, having translated rm_startblock
+ * into the appropriate daddr units.
+ */
+STATIC int
+xfs_getfsmap_helper(
+       struct xfs_trans                *tp,
+       struct xfs_getfsmap_info        *info,
+       struct xfs_rmap_irec            *rec,
+       xfs_daddr_t                     rec_daddr)
+{
+       struct xfs_fsmap                fmr;
+       struct xfs_mount                *mp = tp->t_mountp;
+       bool                            shared;
+       int                             error;
+
+       if (fatal_signal_pending(current))
+               return -EINTR;
+
+       /*
+        * Filter out records that start before our startpoint, if the
+        * caller requested that.
+        */
+       if (xfs_rmap_compare(rec, &info->low) < 0) {
+               rec_daddr += XFS_FSB_TO_BB(mp, rec->rm_blockcount);
+               if (info->next_daddr < rec_daddr)
+                       info->next_daddr = rec_daddr;
+               return XFS_BTREE_QUERY_RANGE_CONTINUE;
+       }
+
+       /* Are we just counting mappings? */
+       if (info->head->fmh_count == 0) {
+               if (rec_daddr > info->next_daddr)
+                       info->head->fmh_entries++;
+
+               if (info->last)
+                       return XFS_BTREE_QUERY_RANGE_CONTINUE;
+
+               info->head->fmh_entries++;
+
+               rec_daddr += XFS_FSB_TO_BB(mp, rec->rm_blockcount);
+               if (info->next_daddr < rec_daddr)
+                       info->next_daddr = rec_daddr;
+               return XFS_BTREE_QUERY_RANGE_CONTINUE;
+       }
+
+       /*
+        * If the record starts past the last physical block we saw,
+        * then we've found a gap.  Report the gap as being owned by
+        * whatever the caller specified is the missing owner.
+        */
+       if (rec_daddr > info->next_daddr) {
+               if (info->head->fmh_entries >= info->head->fmh_count)
+                       return XFS_BTREE_QUERY_RANGE_ABORT;
+
+               fmr.fmr_device = info->dev;
+               fmr.fmr_physical = info->next_daddr;
+               fmr.fmr_owner = info->missing_owner;
+               fmr.fmr_offset = 0;
+               fmr.fmr_length = rec_daddr - info->next_daddr;
+               fmr.fmr_flags = FMR_OF_SPECIAL_OWNER;
+               error = info->formatter(&fmr, info->format_arg);
+               if (error)
+                       return error;
+               info->head->fmh_entries++;
+       }
+
+       if (info->last)
+               goto out;
+
+       /* Fill out the extent we found */
+       if (info->head->fmh_entries >= info->head->fmh_count)
+               return XFS_BTREE_QUERY_RANGE_ABORT;
+
+       trace_xfs_fsmap_mapping(mp, info->dev, info->agno, rec);
+
+       fmr.fmr_device = info->dev;
+       fmr.fmr_physical = rec_daddr;
+       error = xfs_fsmap_owner_from_rmap(&fmr, rec);
+       if (error)
+               return error;
+       fmr.fmr_offset = XFS_FSB_TO_BB(mp, rec->rm_offset);
+       fmr.fmr_length = XFS_FSB_TO_BB(mp, rec->rm_blockcount);
+       if (rec->rm_flags & XFS_RMAP_UNWRITTEN)
+               fmr.fmr_flags |= FMR_OF_PREALLOC;
+       if (rec->rm_flags & XFS_RMAP_ATTR_FORK)
+               fmr.fmr_flags |= FMR_OF_ATTR_FORK;
+       if (rec->rm_flags & XFS_RMAP_BMBT_BLOCK)
+               fmr.fmr_flags |= FMR_OF_EXTENT_MAP;
+       if (fmr.fmr_flags == 0) {
+               error = xfs_getfsmap_is_shared(tp, info, rec, &shared);
+               if (error)
+                       return error;
+               if (shared)
+                       fmr.fmr_flags |= FMR_OF_SHARED;
+       }
+       error = info->formatter(&fmr, info->format_arg);
+       if (error)
+               return error;
+       info->head->fmh_entries++;
+
+out:
+       rec_daddr += XFS_FSB_TO_BB(mp, rec->rm_blockcount);
+       if (info->next_daddr < rec_daddr)
+               info->next_daddr = rec_daddr;
+       return XFS_BTREE_QUERY_RANGE_CONTINUE;
+}
+
+/* Transform a rmapbt irec into a fsmap */
+STATIC int
+xfs_getfsmap_datadev_helper(
+       struct xfs_btree_cur            *cur,
+       struct xfs_rmap_irec            *rec,
+       void                            *priv)
+{
+       struct xfs_mount                *mp = cur->bc_mp;
+       struct xfs_getfsmap_info        *info = priv;
+       xfs_fsblock_t                   fsb;
+       xfs_daddr_t                     rec_daddr;
+
+       fsb = XFS_AGB_TO_FSB(mp, cur->bc_private.a.agno, rec->rm_startblock);
+       rec_daddr = XFS_FSB_TO_DADDR(mp, fsb);
+
+       return xfs_getfsmap_helper(cur->bc_tp, info, rec, rec_daddr);
+}
+
+/* Set rmap flags based on the getfsmap flags */
+static void
+xfs_getfsmap_set_irec_flags(
+       struct xfs_rmap_irec    *irec,
+       struct xfs_fsmap        *fmr)
+{
+       irec->rm_flags = 0;
+       if (fmr->fmr_flags & FMR_OF_ATTR_FORK)
+               irec->rm_flags |= XFS_RMAP_ATTR_FORK;
+       if (fmr->fmr_flags & FMR_OF_EXTENT_MAP)
+               irec->rm_flags |= XFS_RMAP_BMBT_BLOCK;
+       if (fmr->fmr_flags & FMR_OF_PREALLOC)
+               irec->rm_flags |= XFS_RMAP_UNWRITTEN;
+}
+
+/* Execute a getfsmap query against the log device. */
+STATIC int
+xfs_getfsmap_logdev(
+       struct xfs_trans                *tp,
+       struct xfs_fsmap                *keys,
+       struct xfs_getfsmap_info        *info)
+{
+       struct xfs_mount                *mp = tp->t_mountp;
+       struct xfs_rmap_irec            rmap;
+       int                             error;
+
+       /* Set up search keys */
+       info->low.rm_startblock = XFS_BB_TO_FSBT(mp, keys[0].fmr_physical);
+       info->low.rm_offset = XFS_BB_TO_FSBT(mp, keys[0].fmr_offset);
+       error = xfs_fsmap_owner_to_rmap(&info->low, keys);
+       if (error)
+               return error;
+       info->low.rm_blockcount = 0;
+       xfs_getfsmap_set_irec_flags(&info->low, &keys[0]);
+
+       error = xfs_fsmap_owner_to_rmap(&info->high, keys + 1);
+       if (error)
+               return error;
+       info->high.rm_startblock = -1U;
+       info->high.rm_owner = ULLONG_MAX;
+       info->high.rm_offset = ULLONG_MAX;
+       info->high.rm_blockcount = 0;
+       info->high.rm_flags = XFS_RMAP_KEY_FLAGS | XFS_RMAP_REC_FLAGS;
+       info->missing_owner = XFS_FMR_OWN_FREE;
+
+       trace_xfs_fsmap_low_key(mp, info->dev, info->agno, &info->low);
+       trace_xfs_fsmap_high_key(mp, info->dev, info->agno, &info->high);
+
+       if (keys[0].fmr_physical > 0)
+               return 0;
+
+       /* Fabricate an rmap entry for the external log device. */
+       rmap.rm_startblock = 0;
+       rmap.rm_blockcount = mp->m_sb.sb_logblocks;
+       rmap.rm_owner = XFS_RMAP_OWN_LOG;
+       rmap.rm_offset = 0;
+       rmap.rm_flags = 0;
+
+       return xfs_getfsmap_helper(tp, info, &rmap, 0);
+}
+
+/* Execute a getfsmap query against the regular data device. */
+STATIC int
+__xfs_getfsmap_datadev(
+       struct xfs_trans                *tp,
+       struct xfs_fsmap                *keys,
+       struct xfs_getfsmap_info        *info,
+       int                             (*query_fn)(struct xfs_trans *,
+                                                   struct xfs_getfsmap_info *,
+                                                   struct xfs_btree_cur **,
+                                                   void *),
+       void                            *priv)
+{
+       struct xfs_mount                *mp = tp->t_mountp;
+       struct xfs_btree_cur            *bt_cur = NULL;
+       xfs_fsblock_t                   start_fsb;
+       xfs_fsblock_t                   end_fsb;
+       xfs_agnumber_t                  start_ag;
+       xfs_agnumber_t                  end_ag;
+       xfs_daddr_t                     eofs;
+       int                             error = 0;
+
+       eofs = XFS_FSB_TO_BB(mp, mp->m_sb.sb_dblocks);
+       if (keys[0].fmr_physical >= eofs)
+               return 0;
+       if (keys[1].fmr_physical >= eofs)
+               keys[1].fmr_physical = eofs - 1;
+       start_fsb = XFS_DADDR_TO_FSB(mp, keys[0].fmr_physical);
+       end_fsb = XFS_DADDR_TO_FSB(mp, keys[1].fmr_physical);
+
+       /*
+        * Convert the fsmap low/high keys to AG based keys.  Initialize
+        * low to the fsmap low key and max out the high key to the end
+        * of the AG.
+        */
+       info->low.rm_startblock = XFS_FSB_TO_AGBNO(mp, start_fsb);
+       info->low.rm_offset = XFS_BB_TO_FSBT(mp, keys[0].fmr_offset);
+       error = xfs_fsmap_owner_to_rmap(&info->low, &keys[0]);
+       if (error)
+               return error;
+       info->low.rm_blockcount = 0;
+       xfs_getfsmap_set_irec_flags(&info->low, &keys[0]);
+
+       info->high.rm_startblock = -1U;
+       info->high.rm_owner = ULLONG_MAX;
+       info->high.rm_offset = ULLONG_MAX;
+       info->high.rm_blockcount = 0;
+       info->high.rm_flags = XFS_RMAP_KEY_FLAGS | XFS_RMAP_REC_FLAGS;
+
+       start_ag = XFS_FSB_TO_AGNO(mp, start_fsb);
+       end_ag = XFS_FSB_TO_AGNO(mp, end_fsb);
+
+       /* Query each AG */
+       for (info->agno = start_ag; info->agno <= end_ag; info->agno++) {
+               /*
+                * Set the AG high key from the fsmap high key if this
+                * is the last AG that we're querying.
+                */
+               if (info->agno == end_ag) {
+                       info->high.rm_startblock = XFS_FSB_TO_AGBNO(mp,
+                                       end_fsb);
+                       info->high.rm_offset = XFS_BB_TO_FSBT(mp,
+                                       keys[1].fmr_offset);
+                       error = xfs_fsmap_owner_to_rmap(&info->high, &keys[1]);
+                       if (error)
+                               goto err;
+                       xfs_getfsmap_set_irec_flags(&info->high, &keys[1]);
+               }
+
+               if (bt_cur) {
+                       xfs_btree_del_cursor(bt_cur, XFS_BTREE_NOERROR);
+                       bt_cur = NULL;
+                       xfs_trans_brelse(tp, info->agf_bp);
+                       info->agf_bp = NULL;
+               }
+
+               error = xfs_alloc_read_agf(mp, tp, info->agno, 0,
+                               &info->agf_bp);
+               if (error)
+                       goto err;
+
+               trace_xfs_fsmap_low_key(mp, info->dev, info->agno, &info->low);
+               trace_xfs_fsmap_high_key(mp, info->dev, info->agno,
+                               &info->high);
+
+               error = query_fn(tp, info, &bt_cur, priv);
+               if (error)
+                       goto err;
+
+               /*
+                * Set the AG low key to the start of the AG prior to
+                * moving on to the next AG.
+                */
+               if (info->agno == start_ag) {
+                       info->low.rm_startblock = 0;
+                       info->low.rm_owner = 0;
+                       info->low.rm_offset = 0;
+                       info->low.rm_flags = 0;
+               }
+       }
+
+       /* Report any gap at the end of the AG */
+       info->last = true;
+       error = query_fn(tp, info, &bt_cur, priv);
+       if (error)
+               goto err;
+
+err:
+       if (bt_cur)
+               xfs_btree_del_cursor(bt_cur, error < 0 ? XFS_BTREE_ERROR :
+                                                        XFS_BTREE_NOERROR);
+       if (info->agf_bp) {
+               xfs_trans_brelse(tp, info->agf_bp);
+               info->agf_bp = NULL;
+       }
+
+       return error;
+}
+
+/* Actually query the rmap btree. */
+STATIC int
+xfs_getfsmap_datadev_rmapbt_query(
+       struct xfs_trans                *tp,
+       struct xfs_getfsmap_info        *info,
+       struct xfs_btree_cur            **curpp,
+       void                            *priv)
+{
+       /* Report any gap at the end of the last AG. */
+       if (info->last)
+               return xfs_getfsmap_datadev_helper(*curpp, &info->high, info);
+
+       /* Allocate cursor for this AG and query_range it. */
+       *curpp = xfs_rmapbt_init_cursor(tp->t_mountp, tp, info->agf_bp,
+                       info->agno);
+       return xfs_rmap_query_range(*curpp, &info->low, &info->high,
+                       xfs_getfsmap_datadev_helper, info);
+}
+
+/* Execute a getfsmap query against the regular data device rmapbt. */
+STATIC int
+xfs_getfsmap_datadev_rmapbt(
+       struct xfs_trans                *tp,
+       struct xfs_fsmap                *keys,
+       struct xfs_getfsmap_info        *info)
+{
+       info->missing_owner = XFS_FMR_OWN_FREE;
+       return __xfs_getfsmap_datadev(tp, keys, info,
+                       xfs_getfsmap_datadev_rmapbt_query, NULL);
+}
+
+/* Do we recognize the device? */
+STATIC bool
+xfs_getfsmap_is_valid_device(
+       struct xfs_mount        *mp,
+       struct xfs_fsmap        *fm)
+{
+       if (fm->fmr_device == 0 || fm->fmr_device == UINT_MAX ||
+           fm->fmr_device == new_encode_dev(mp->m_ddev_targp->bt_dev))
+               return true;
+       if (mp->m_logdev_targp &&
+           fm->fmr_device == new_encode_dev(mp->m_logdev_targp->bt_dev))
+               return true;
+       return false;
+}
+
+/* Ensure that the low key is less than the high key. */
+STATIC bool
+xfs_getfsmap_check_keys(
+       struct xfs_fsmap                *low_key,
+       struct xfs_fsmap                *high_key)
+{
+       if (low_key->fmr_device > high_key->fmr_device)
+               return false;
+       if (low_key->fmr_device < high_key->fmr_device)
+               return true;
+
+       if (low_key->fmr_physical > high_key->fmr_physical)
+               return false;
+       if (low_key->fmr_physical < high_key->fmr_physical)
+               return true;
+
+       if (low_key->fmr_owner > high_key->fmr_owner)
+               return false;
+       if (low_key->fmr_owner < high_key->fmr_owner)
+               return true;
+
+       if (low_key->fmr_offset > high_key->fmr_offset)
+               return false;
+       if (low_key->fmr_offset < high_key->fmr_offset)
+               return true;
+
+       return false;
+}
+
+#define XFS_GETFSMAP_DEVS      2
+/*
+ * Get filesystem's extents as described in head, and format for
+ * output.  Calls formatter to fill the user's buffer until all
+ * extents are mapped, until the passed-in head->fmh_count slots have
+ * been filled, or until the formatter short-circuits the loop, if it
+ * is tracking filled-in extents on its own.
+ *
+ * Key to Confusion
+ * ----------------
+ * There are multiple levels of keys and counters at work here:
+ * xfs_fsmap_head.fmh_keys     -- low and high fsmap keys passed in;
+ *                                these reflect fs-wide sector addrs.
+ * dkeys                       -- fmh_keys used to query each device;
+ *                                these are fmh_keys but w/ the low key
+ *                                bumped up by fmr_length.
+ * xfs_getfsmap_info.next_daddr        -- next disk addr we expect to see; this
+ *                                is how we detect gaps in the fsmap
+                                  records and report them.
+ * xfs_getfsmap_info.low/high  -- per-AG low/high keys computed from
+ *                                dkeys; used to query the metadata.
+ */
+int
+xfs_getfsmap(
+       struct xfs_mount                *mp,
+       struct xfs_fsmap_head           *head,
+       xfs_fsmap_format_t              formatter,
+       void                            *arg)
+{
+       struct xfs_trans                *tp = NULL;
+       struct xfs_fsmap                dkeys[2];       /* per-dev keys */
+       struct xfs_getfsmap_dev         handlers[XFS_GETFSMAP_DEVS];
+       struct xfs_getfsmap_info        info = {0};
+       int                             i;
+       int                             error = 0;
+
+       if (!xfs_sb_version_hasrmapbt(&mp->m_sb))
+               return -EOPNOTSUPP;
+       if (head->fmh_iflags & ~FMH_IF_VALID)
+               return -EINVAL;
+       if (!xfs_getfsmap_is_valid_device(mp, &head->fmh_keys[0]) ||
+           !xfs_getfsmap_is_valid_device(mp, &head->fmh_keys[1]))
+               return -EINVAL;
+
+       head->fmh_entries = 0;
+
+       /* Set up our device handlers. */
+       memset(handlers, 0, sizeof(handlers));
+       handlers[0].dev = new_encode_dev(mp->m_ddev_targp->bt_dev);
+       handlers[0].fn = xfs_getfsmap_datadev_rmapbt;
+       if (mp->m_logdev_targp != mp->m_ddev_targp) {
+               handlers[1].dev = new_encode_dev(mp->m_logdev_targp->bt_dev);
+               handlers[1].fn = xfs_getfsmap_logdev;
+       }
+
+       xfs_sort(handlers, XFS_GETFSMAP_DEVS, sizeof(struct xfs_getfsmap_dev),
+                       xfs_getfsmap_dev_compare);
+
+       /*
+        * To continue where we left off, we allow userspace to use the
+        * last mapping from a previous call as the low key of the next.
+        * This is identified by a non-zero length in the low key. We
+        * have to increment the low key in this scenario to ensure we
+        * don't return the same mapping again, and instead return the
+        * very next mapping.
+        *
+        * If the low key mapping refers to file data, the same physical
+        * blocks could be mapped to several other files/offsets.
+        * According to rmapbt record ordering, the minimal next
+        * possible record for the block range is the next starting
+        * offset in the same inode. Therefore, bump the file offset to
+        * continue the search appropriately.  For all other low key
+        * mapping types (attr blocks, metadata), bump the physical
+        * offset as there can be no other mapping for the same physical
+        * block range.
+        */
+       dkeys[0] = head->fmh_keys[0];
+       if (dkeys[0].fmr_flags & (FMR_OF_SPECIAL_OWNER | FMR_OF_EXTENT_MAP)) {
+               dkeys[0].fmr_physical += dkeys[0].fmr_length;
+               dkeys[0].fmr_owner = 0;
+               if (dkeys[0].fmr_offset)
+                       return -EINVAL;
+       } else
+               dkeys[0].fmr_offset += dkeys[0].fmr_length;
+       dkeys[0].fmr_length = 0;
+       memset(&dkeys[1], 0xFF, sizeof(struct xfs_fsmap));
+
+       if (!xfs_getfsmap_check_keys(dkeys, &head->fmh_keys[1]))
+               return -EINVAL;
+
+       info.next_daddr = head->fmh_keys[0].fmr_physical +
+                         head->fmh_keys[0].fmr_length;
+       info.formatter = formatter;
+       info.format_arg = arg;
+       info.head = head;
+
+       /* For each device we support... */
+       for (i = 0; i < XFS_GETFSMAP_DEVS; i++) {
+               /* Is this device within the range the user asked for? */
+               if (!handlers[i].fn)
+                       continue;
+               if (head->fmh_keys[0].fmr_device > handlers[i].dev)
+                       continue;
+               if (head->fmh_keys[1].fmr_device < handlers[i].dev)
+                       break;
+
+               /*
+                * If this device number matches the high key, we have
+                * to pass the high key to the handler to limit the
+                * query results.  If the device number exceeds the
+                * low key, zero out the low key so that we get
+                * everything from the beginning.
+                */
+               if (handlers[i].dev == head->fmh_keys[1].fmr_device)
+                       dkeys[1] = head->fmh_keys[1];
+               if (handlers[i].dev > head->fmh_keys[0].fmr_device)
+                       memset(&dkeys[0], 0, sizeof(struct xfs_fsmap));
+
+               error = xfs_trans_alloc_empty(mp, &tp);
+               if (error)
+                       break;
+
+               info.dev = handlers[i].dev;
+               info.last = false;
+               info.agno = NULLAGNUMBER;
+               error = handlers[i].fn(tp, dkeys, &info);
+               if (error)
+                       break;
+               xfs_trans_cancel(tp);
+               tp = NULL;
+               info.next_daddr = 0;
+       }
+
+       if (tp)
+               xfs_trans_cancel(tp);
+       head->fmh_oflags = FMH_OF_DEV_T;
+       return error;
+}