From: Darrick J. Wong <djwong@xxxxxxxxxx>
Rebuild the realtime rmap btree file from the reverse mapping records we
gathered from walking the inodes.
Signed-off-by: "Darrick J. Wong" <djwong@xxxxxxxxxx>
---
libxfs/libxfs_api_defs.h | 8 +
repair/Makefile | 1
repair/bulkload.c | 41 +++++++
repair/bulkload.h | 2
repair/phase6.c | 21 ++++
repair/rmap.c | 26 +++++
repair/rmap.h | 1
repair/rtrmap_repair.c | 265 ++++++++++++++++++++++++++++++++++++++++++++++
repair/xfs_repair.c | 8 +
9 files changed, 371 insertions(+), 2 deletions(-)
create mode 100644 repair/rtrmap_repair.c
diff --git a/libxfs/libxfs_api_defs.h b/libxfs/libxfs_api_defs.h
index b62efad757470b..193b1eeaa7537e 100644
--- a/libxfs/libxfs_api_defs.h
+++ b/libxfs/libxfs_api_defs.h
@@ -278,6 +278,7 @@
#define xfs_rmap_irec_offset_unpack libxfs_rmap_irec_offset_unpack
#define xfs_rmap_lookup_le libxfs_rmap_lookup_le
#define xfs_rmap_lookup_le_range libxfs_rmap_lookup_le_range
+#define xfs_rmap_map_extent libxfs_rmap_map_extent
#define xfs_rmap_map_raw libxfs_rmap_map_raw
#define xfs_rmap_query_all libxfs_rmap_query_all
#define xfs_rmap_query_range libxfs_rmap_query_range
@@ -295,9 +296,12 @@
#define xfs_rtginode_name libxfs_rtginode_name
#define xfs_rtsummary_create libxfs_rtsummary_create
+#define xfs_rtginode_create libxfs_rtginode_create
#define xfs_rtginode_irele libxfs_rtginode_irele
#define xfs_rtginode_load libxfs_rtginode_load
#define xfs_rtginode_load_parent libxfs_rtginode_load_parent
+#define xfs_rtginode_mkdir_parent libxfs_rtginode_mkdir_parent
+#define xfs_rtginode_name libxfs_rtginode_name
#define xfs_rtgroup_alloc libxfs_rtgroup_alloc
#define xfs_rtgroup_extents libxfs_rtgroup_extents
#define xfs_rtgroup_grab libxfs_rtgroup_grab
@@ -314,12 +318,16 @@
#define xfs_rtgroup_get libxfs_rtgroup_get
#define xfs_rtgroup_put libxfs_rtgroup_put
#define xfs_rtrmapbt_calc_reserves libxfs_rtrmapbt_calc_reserves
+#define xfs_rtrmapbt_calc_size libxfs_rtrmapbt_calc_size
+#define xfs_rtrmapbt_commit_staged_btree libxfs_rtrmapbt_commit_staged_btree
+#define xfs_rtrmapbt_create libxfs_rtrmapbt_create
#define xfs_rtrmapbt_droot_maxrecs libxfs_rtrmapbt_droot_maxrecs
#define xfs_rtrmapbt_maxlevels_ondisk libxfs_rtrmapbt_maxlevels_ondisk
#define xfs_rtrmapbt_init_cursor libxfs_rtrmapbt_init_cursor
#define xfs_rtrmapbt_maxrecs libxfs_rtrmapbt_maxrecs
#define xfs_rtrmapbt_mem_init libxfs_rtrmapbt_mem_init
#define xfs_rtrmapbt_mem_cursor libxfs_rtrmapbt_mem_cursor
+#define xfs_rtrmapbt_stage_cursor libxfs_rtrmapbt_stage_cursor
#define xfs_sb_from_disk libxfs_sb_from_disk
#define xfs_sb_mount_rextsize libxfs_sb_mount_rextsize
diff --git a/repair/Makefile b/repair/Makefile
index a36a95e353a504..6f4ec3b3a9c4dc 100644
--- a/repair/Makefile
+++ b/repair/Makefile
@@ -73,6 +73,7 @@ CFILES = \
rcbag.c \
rmap.c \
rt.c \
+ rtrmap_repair.c \
sb.c \
scan.c \
slab.c \
diff --git a/repair/bulkload.c b/repair/bulkload.c
index aada5bbae579f8..a9e51de0a24c17 100644
--- a/repair/bulkload.c
+++ b/repair/bulkload.c
@@ -361,3 +361,44 @@ bulkload_estimate_ag_slack(
if (bload->node_slack < 0)
bload->node_slack = 2;
}
+
+/*
+ * Estimate proper slack values for a btree that's being reloaded.
+ *
+ * Under most circumstances, we'll take whatever default loading value the
+ * btree bulk loading code calculates for us. However, there are some
+ * exceptions to this rule:
+ *
+ * (1) If someone turned one of the debug knobs.
+ * (2) The FS has less than ~9% space free.
+ *
+ * Note that we actually use 3/32 for the comparison to avoid division.
+ */
+void
+bulkload_estimate_inode_slack(
+ struct xfs_mount *mp,
+ struct xfs_btree_bload *bload,
+ unsigned long long free)
+{
+ /*
+ * The global values are set to -1 (i.e. take the bload defaults)
+ * unless someone has set them otherwise, so we just pull the values
+ * here.
+ */
+ bload->leaf_slack = bload_leaf_slack;
+ bload->node_slack = bload_node_slack;
+
+ /* No further changes if there's more than 3/32ths space left. */
+ if (free >= ((mp->m_sb.sb_dblocks * 3) >> 5))
+ return;
+
+ /*
+ * We're low on space; load the btrees as tightly as possible. Leave
+ * a couple of open slots in each btree block so that we don't end up
+ * splitting the btrees like crazy right after mount.
+ */
+ if (bload->leaf_slack < 0)
+ bload->leaf_slack = 2;
+ if (bload->node_slack < 0)
+ bload->node_slack = 2;
+}
diff --git a/repair/bulkload.h b/repair/bulkload.h
index a88aafaa678a3a..842121b15190e7 100644
--- a/repair/bulkload.h
+++ b/repair/bulkload.h
@@ -78,5 +78,7 @@ void bulkload_cancel(struct bulkload *bkl);
int bulkload_commit(struct bulkload *bkl);
void bulkload_estimate_ag_slack(struct repair_ctx *sc,
struct xfs_btree_bload *bload, unsigned int free);
+void bulkload_estimate_inode_slack(struct xfs_mount *mp,
+ struct xfs_btree_bload *bload, unsigned long long free);
#endif /* __XFS_REPAIR_BULKLOAD_H__ */
diff --git a/repair/phase6.c b/repair/phase6.c
index 7d2e0554594265..cae9d970481840 100644
--- a/repair/phase6.c
+++ b/repair/phase6.c
@@ -21,6 +21,8 @@
#include "repair/pptr.h"
#include "repair/rt.h"
#include "repair/quotacheck.h"
+#include "repair/slab.h"
+#include "repair/rmap.h"
static xfs_ino_t orphanage_ino;
@@ -685,6 +687,15 @@ ensure_rtgroup_summary(
fill_rtsummary(rtg);
}
+static void
+ensure_rtgroup_rmapbt(
+ struct xfs_rtgroup *rtg,
+ xfs_filblks_t est_fdblocks)
+{
+ if (ensure_rtgroup_file(rtg, XFS_RTGI_RMAP))
+ populate_rtgroup_rmapbt(rtg, est_fdblocks);
+}
+
/* Initialize a root directory. */
static int
init_fs_root_dir(
@@ -3365,6 +3376,8 @@ reset_rt_metadir_inodes(
struct xfs_mount *mp)
{
struct xfs_rtgroup *rtg = NULL;
+ xfs_filblks_t metadata_blocks = 0;
+ xfs_filblks_t est_fdblocks = 0;
int error;
/*
@@ -3386,6 +3399,13 @@ reset_rt_metadir_inodes(
mark_ino_metadata(mp, mp->m_rtdirip->i_ino);
}
+ /* Estimate how much free space will be left after building btrees */
+ while ((rtg = xfs_rtgroup_next(mp, rtg)))
+ metadata_blocks += estimate_rtrmapbt_blocks(rtg);
+
+ if (mp->m_sb.sb_fdblocks > metadata_blocks)
+ est_fdblocks = mp->m_sb.sb_fdblocks - metadata_blocks;
+
/*
* This isn't the whole story, but it keeps the message that we've had
* for years and which is expected in xfstests and more.
@@ -3400,6 +3420,7 @@ _(" - resetting contents of realtime bitmap and summary inodes\n"));
while ((rtg = xfs_rtgroup_next(mp, rtg))) {
ensure_rtgroup_bitmap(rtg);
ensure_rtgroup_summary(rtg);
+ ensure_rtgroup_rmapbt(rtg, est_fdblocks);
}
}
diff --git a/repair/rmap.c b/repair/rmap.c
index a40851b4d0dc69..85a65048db9afc 100644
--- a/repair/rmap.c
+++ b/repair/rmap.c
@@ -1940,3 +1940,29 @@ estimate_refcountbt_blocks(
return libxfs_refcountbt_calc_size(mp,
slab_count(x->ar_refcount_items));
}
+
+/* Estimate the size of the ondisk rtrmapbt from the incore tree. */
+xfs_filblks_t
+estimate_rtrmapbt_blocks(
+ struct xfs_rtgroup *rtg)
+{
+ struct xfs_mount *mp = rtg_mount(rtg);
+ struct xfs_ag_rmap *x;
+ unsigned long long nr_recs;
+
+ if (!rmap_needs_work(mp) || !xfs_has_rtrmapbt(mp))
+ return 0;
+
+ /*
+ * Overestimate the amount of space needed by pretending that every
+ * byte in the incore tree is used to store rtrmapbt records. This
+ * means we can use SEEK_DATA/HOLE on the xfile, which is faster than
+ * walking the entire btree.
+ */
+ x = &rg_rmaps[rtg_rgno(rtg)];
+ if (!rmaps_has_observations(x))
+ return 0;
+
+ nr_recs = xmbuf_bytes(x->ar_xmbtp) / sizeof(struct xfs_rmap_rec);
+ return libxfs_rtrmapbt_calc_size(mp, nr_recs);
+}
diff --git a/repair/rmap.h b/repair/rmap.h
index ebda561e59bc8f..23859bf6c2ad42 100644
--- a/repair/rmap.h
+++ b/repair/rmap.h
@@ -60,5 +60,6 @@ int rmap_get_mem_rec(struct xfs_btree_cur *rmcur, struct xfs_rmap_irec *irec);
void populate_rtgroup_rmapbt(struct xfs_rtgroup *rtg,
xfs_filblks_t est_fdblocks);
+xfs_filblks_t estimate_rtrmapbt_blocks(struct xfs_rtgroup *rtg);
#endif /* RMAP_H_ */
diff --git a/repair/rtrmap_repair.c b/repair/rtrmap_repair.c
new file mode 100644
index 00000000000000..2b07e8943e591d
--- /dev/null
+++ b/repair/rtrmap_repair.c
@@ -0,0 +1,265 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/*
+ * Copyright (c) 2019-2025 Oracle. All Rights Reserved.
+ * Author: Darrick J. Wong <djwong@xxxxxxxxxx>
+ */
+#include <libxfs.h>
+#include "btree.h"
+#include "err_protos.h"
+#include "libxlog.h"
+#include "incore.h"
+#include "globals.h"
+#include "dinode.h"
+#include "slab.h"
+#include "rmap.h"
+#include "bulkload.h"
+
+/* Ported routines from fs/xfs/scrub/rtrmap_repair.c */
+
+/*
+ * Realtime Reverse Mapping (RTRMAPBT) Repair
+ * ==========================================
+ *
+ * Gather all the rmap records for the inode and fork we're fixing, reset the
+ * incore fork, then recreate the btree.
+ */
+struct xrep_rtrmap {
+ struct xfs_btree_cur *btree_cursor;
+
+ /* New fork. */
+ struct bulkload new_fork_info;
+ struct xfs_btree_bload rtrmap_bload;
+
+ struct repair_ctx *sc;
+ struct xfs_rtgroup *rtg;
+
+ /* Estimated free space after building all rt btrees */
+ xfs_filblks_t est_fdblocks;
+};
+
+/* Retrieve rtrmapbt data for bulk load. */
+STATIC int
+xrep_rtrmap_get_records(
+ struct xfs_btree_cur *cur,
+ unsigned int idx,
+ struct xfs_btree_block *block,
+ unsigned int nr_wanted,
+ void *priv)
+{
+ struct xrep_rtrmap *rr = priv;
+ union xfs_btree_rec *block_rec;
+ unsigned int loaded;
+ int ret;
+
+ for (loaded = 0; loaded < nr_wanted; loaded++, idx++) {
+ ret = rmap_get_mem_rec(rr->btree_cursor, &cur->bc_rec.r);
+ if (ret < 0)
+ return ret;
+ if (ret == 0)
+ do_error(
+ _("ran out of records while rebuilding rt rmap btree\n"));
+
+ block_rec = libxfs_btree_rec_addr(cur, idx, block);
+ cur->bc_ops->init_rec_from_cur(cur, block_rec);
+ }
+
+ return loaded;
+}
+
+/* Feed one of the new btree blocks to the bulk loader. */
+STATIC int
+xrep_rtrmap_claim_block(
+ struct xfs_btree_cur *cur,
+ union xfs_btree_ptr *ptr,
+ void *priv)
+{
+ struct xrep_rtrmap *rr = priv;
+
+ return bulkload_claim_block(cur, &rr->new_fork_info, ptr);
+}
+
+/* Figure out how much space we need to create the incore btree root block. */
+STATIC size_t
+xrep_rtrmap_iroot_size(
+ struct xfs_btree_cur *cur,
+ unsigned int level,
+ unsigned int nr_this_level,
+ void *priv)
+{
+ return xfs_rtrmap_broot_space_calc(cur->bc_mp, level, nr_this_level);
+}
+
+/* Reserve new btree blocks and bulk load all the rtrmap records. */
+STATIC int
+xrep_rtrmap_btree_load(
+ struct xrep_rtrmap *rr,
+ struct xfs_btree_cur *rtrmap_cur)
+{
+ struct repair_ctx *sc = rr->sc;
+ int error;
+
+ rr->rtrmap_bload.get_records = xrep_rtrmap_get_records;
+ rr->rtrmap_bload.claim_block = xrep_rtrmap_claim_block;
+ rr->rtrmap_bload.iroot_size = xrep_rtrmap_iroot_size;
+ bulkload_estimate_inode_slack(sc->mp, &rr->rtrmap_bload,
+ rr->est_fdblocks);
+
+ /* Compute how many blocks we'll need. */
+ error = -libxfs_btree_bload_compute_geometry(rtrmap_cur,
+ &rr->rtrmap_bload,
+ rmap_record_count(sc->mp, true, rtg_rgno(rr->rtg)));
+ if (error)
+ return error;
+
+ /*
+ * Guess how many blocks we're going to need to rebuild an entire rtrmap
+ * from the number of extents we found, and pump up our transaction to
+ * have sufficient block reservation.
+ */
+ error = -libxfs_trans_reserve_more(sc->tp, rr->rtrmap_bload.nr_blocks,
+ 0);
+ if (error)
+ return error;
+
+ /*
+ * Reserve the space we'll need for the new btree. Drop the cursor
+ * while we do this because that can roll the transaction and cursors
+ * can't handle that.
+ */
+ error = bulkload_alloc_file_blocks(&rr->new_fork_info,
+ rr->rtrmap_bload.nr_blocks);
+ if (error)
+ return error;
+
+ /* Add all observed rtrmap records. */
+ error = rmap_init_mem_cursor(rr->sc->mp, sc->tp, true,
+ rtg_rgno(rr->rtg), &rr->btree_cursor);
+ if (error)
+ return error;
+ error = -libxfs_btree_bload(rtrmap_cur, &rr->rtrmap_bload, rr);
+ libxfs_btree_del_cursor(rr->btree_cursor, error);
+ return error;
+}
+
+/* Update the inode counters. */
+STATIC int
+xrep_rtrmap_reset_counters(
+ struct xrep_rtrmap *rr)
+{
+ struct repair_ctx *sc = rr->sc;
+
+ /*
+ * Update the inode block counts to reflect the btree we just
+ * generated.
+ */
+ sc->ip->i_nblocks = rr->new_fork_info.ifake.if_blocks;
+ libxfs_trans_log_inode(sc->tp, sc->ip, XFS_ILOG_CORE);
+
+ /* Quotas don't exist so we're done. */
+ return 0;
+}
+
+/*
+ * Use the collected rmap information to stage a new rt rmap btree. If this is
+ * successful we'll return with the new btree root information logged to the
+ * repair transaction but not yet committed.
+ */
+static int
+xrep_rtrmap_build_new_tree(
+ struct xrep_rtrmap *rr)
+{
+ struct xfs_owner_info oinfo;
+ struct xfs_btree_cur *cur;
+ struct repair_ctx *sc = rr->sc;
+ struct xbtree_ifakeroot *ifake = &rr->new_fork_info.ifake;
+ int error;
+
+ /*
+ * Prepare to construct the new fork by initializing the new btree
+ * structure and creating a fake ifork in the ifakeroot structure.
+ */
+ libxfs_rmap_ino_bmbt_owner(&oinfo, sc->ip->i_ino, XFS_DATA_FORK);
+ bulkload_init_inode(&rr->new_fork_info, sc, XFS_DATA_FORK, &oinfo);
+ cur = libxfs_rtrmapbt_init_cursor(NULL, rr->rtg);
+ libxfs_btree_stage_ifakeroot(cur, ifake);
+
+ /*
+ * Figure out the size and format of the new fork, then fill it with
+ * all the rtrmap records we've found. Join the inode to the
+ * transaction so that we can roll the transaction while holding the
+ * inode locked.
+ */
+ libxfs_trans_ijoin(sc->tp, sc->ip, 0);
+ ifake->if_fork->if_format = XFS_DINODE_FMT_META_BTREE;
+ error = xrep_rtrmap_btree_load(rr, cur);
+ if (error)
+ goto err_cur;
+
+ /*
+ * Install the new fork in the inode. After this point the old mapping
+ * data are no longer accessible and the new tree is live. We delete
+ * the cursor immediately after committing the staged root because the
+ * staged fork might be in extents format.
+ */
+ libxfs_rtrmapbt_commit_staged_btree(cur, sc->tp);
+ libxfs_btree_del_cursor(cur, 0);
+
+ /* Reset the inode counters now that we've changed the fork. */
+ error = xrep_rtrmap_reset_counters(rr);
+ if (error)
+ goto err_newbt;
+
+ /* Dispose of any unused blocks and the accounting infomation. */
+ error = bulkload_commit(&rr->new_fork_info);
+ if (error)
+ return error;
+
+ return -libxfs_trans_roll_inode(&sc->tp, sc->ip);
+err_cur:
+ if (cur)
+ libxfs_btree_del_cursor(cur, error);
+err_newbt:
+ bulkload_cancel(&rr->new_fork_info);
+ return error;
+}
+
+/* Store the realtime reverse-mappings in the rtrmapbt. */
+void
+populate_rtgroup_rmapbt(
+ struct xfs_rtgroup *rtg,
+ xfs_filblks_t est_fdblocks)
+{
+ struct xfs_mount *mp = rtg_mount(rtg);
+ struct xfs_inode *ip = rtg_rmap(rtg);
+ struct repair_ctx sc = {
+ .mp = mp,
+ .ip = ip,
+ };
+ struct xrep_rtrmap rr = {
+ .sc = &sc,
+ .rtg = rtg,
+ .est_fdblocks = est_fdblocks,
+ };
+ int error;
+
+ if (!xfs_has_rtrmapbt(mp))
+ return;
+
+ error = -libxfs_trans_alloc(mp, &M_RES(mp)->tr_itruncate, 0, 0, 0,
+ &sc.tp);
+ if (error)
+ goto out;
+
+ error = xrep_rtrmap_build_new_tree(&rr);
+ if (error) {
+ libxfs_trans_cancel(sc.tp);
+ goto out;
+ }
+
+ error = -libxfs_trans_commit(sc.tp);
+out:
+ if (error)
+ do_error(
+ _("rtgroup %u rmap btree could not be rebuilt, error %d\n"),
+ rtg_rgno(rtg), error);
+}
diff --git a/repair/xfs_repair.c b/repair/xfs_repair.c
index 9509f04685c870..eeaaf643468941 100644
--- a/repair/xfs_repair.c
+++ b/repair/xfs_repair.c
@@ -1385,15 +1385,19 @@ main(int argc, char **argv)
rcbagbt_destroy_cur_cache();
/*
- * Done with the block usage maps, toss them...
+ * Done with the block usage maps, toss them. Realtime metadata aren't
+ * rebuilt until phase 6, so we have to keep them around.
*/
- rmaps_free(mp);
+ if (mp->m_sb.sb_rblocks == 0)
+ rmaps_free(mp);
free_bmaps(mp);
if (!bad_ino_btree) {
phase6(mp);
phase_end(mp, 6);
+ if (mp->m_sb.sb_rblocks != 0)
+ rmaps_free(mp);
free_rtgroup_inodes();
phase7(mp, phase2_threads);
