All of lore.kernel.org
 help / color / mirror / Atom feed
From: "Darrick J. Wong" <darrick.wong@oracle.com>
To: david@fromorbit.com, darrick.wong@oracle.com
Cc: xfs@oss.sgi.com
Subject: [PATCH 40/51] xfs_db: add support for checking the refcount btree
Date: Tue, 06 Oct 2015 22:09:33 -0700	[thread overview]
Message-ID: <20151007050933.1504.94823.stgit@birch.djwong.org> (raw)
In-Reply-To: <20151007050513.1504.28089.stgit@birch.djwong.org>

Do some basic checks of the refcount btree.  xfs_repair will have to
check that the reference counts match the various bmbt mappings.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 db/check.c |  136 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++----
 1 file changed, 128 insertions(+), 8 deletions(-)


diff --git a/db/check.c b/db/check.c
index 648e0d6..9b79f99 100644
--- a/db/check.c
+++ b/db/check.c
@@ -44,7 +44,8 @@ typedef enum {
 	DBM_FREE1,	DBM_FREE2,	DBM_FREELIST,	DBM_INODE,
 	DBM_LOG,	DBM_MISSING,	DBM_QUOTA,	DBM_RTBITMAP,
 	DBM_RTDATA,	DBM_RTFREE,	DBM_RTSUM,	DBM_SB,
-	DBM_SYMLINK,	DBM_BTFINO,	DBM_BTRMAP,
+	DBM_SYMLINK,	DBM_BTFINO,	DBM_BTRMAP,	DBM_BTREFC,
+	DBM_RLDATA,
 	DBM_NDBM
 } dbm_t;
 
@@ -52,7 +53,8 @@ typedef struct inodata {
 	struct inodata	*next;
 	nlink_t		link_set;
 	nlink_t		link_add;
-	char		isdir;
+	char		isdir:1;
+	char		isreflink:1;
 	char		security;
 	char		ilist;
 	xfs_ino_t	ino;
@@ -172,6 +174,8 @@ static const char	*typename[] = {
 	"symlink",
 	"btfino",
 	"btrmap",
+	"btrefcnt",
+	"rldata",
 	NULL
 };
 static int		verbose;
@@ -229,7 +233,8 @@ static int		blocktrash_f(int argc, char **argv);
 static int		blockuse_f(int argc, char **argv);
 static int		check_blist(xfs_fsblock_t bno);
 static void		check_dbmap(xfs_agnumber_t agno, xfs_agblock_t agbno,
-				    xfs_extlen_t len, dbm_t type);
+				    xfs_extlen_t len, dbm_t type,
+				    int ignore_reflink);
 static int		check_inomap(xfs_agnumber_t agno, xfs_agblock_t agbno,
 				     xfs_extlen_t len, xfs_ino_t c_ino);
 static void		check_linkcounts(xfs_agnumber_t agno);
@@ -353,6 +358,9 @@ static void		scanfunc_fino(struct xfs_btree_block *block, int level,
 static void		scanfunc_rmap(struct xfs_btree_block *block, int level,
 				     struct xfs_agf *agf, xfs_agblock_t bno,
 				     int isroot);
+static void		scanfunc_refcnt(struct xfs_btree_block *block, int level,
+				     struct xfs_agf *agf, xfs_agblock_t bno,
+				     int isroot);
 static void		set_dbmap(xfs_agnumber_t agno, xfs_agblock_t agbno,
 				  xfs_extlen_t len, dbm_t type,
 				  xfs_agnumber_t c_agno, xfs_agblock_t c_agbno);
@@ -1055,6 +1063,7 @@ blocktrash_f(
 		   (1 << DBM_SYMLINK) |
 		   (1 << DBM_BTFINO) |
 		   (1 << DBM_BTRMAP) |
+		   (1 << DBM_BTREFC) |
 		   (1 << DBM_SB);
 	while ((c = getopt(argc, argv, "0123n:o:s:t:x:y:z")) != EOF) {
 		switch (c) {
@@ -1292,18 +1301,25 @@ check_dbmap(
 	xfs_agnumber_t	agno,
 	xfs_agblock_t	agbno,
 	xfs_extlen_t	len,
-	dbm_t		type)
+	dbm_t		type,
+	int		ignore_reflink)
 {
 	xfs_extlen_t	i;
 	char		*p;
+	dbm_t		d;
 
 	for (i = 0, p = &dbmap[agno][agbno]; i < len; i++, p++) {
+		d = (dbm_t)*p;
+		if (ignore_reflink && (d == DBM_UNKNOWN || d == DBM_DATA ||
+				       d == DBM_RLDATA))
+			continue;
 		if ((dbm_t)*p != type) {
-			if (!sflag || CHECK_BLISTA(agno, agbno + i))
+			if (!sflag || CHECK_BLISTA(agno, agbno + i)) {
 				dbprintf(_("block %u/%u expected type %s got "
 					 "%s\n"),
 					agno, agbno + i, typename[type],
 					typename[(dbm_t)*p]);
+			}
 			error++;
 		}
 	}
@@ -1337,7 +1353,7 @@ check_inomap(
 		return 0;
 	}
 	for (i = 0, rval = 1, idp = &inomap[agno][agbno]; i < len; i++, idp++) {
-		if (*idp) {
+		if (*idp && !(*idp)->isreflink) {
 			if (!sflag || (*idp)->ilist ||
 			    CHECK_BLISTA(agno, agbno + i))
 				dbprintf(_("block %u/%u claimed by inode %lld, "
@@ -1543,6 +1559,26 @@ check_rrange(
 	return 1;
 }
 
+/*
+ * We don't check the accuracy of reference counts -- all we do is ensure
+ * that a data block never crosses with non-data blocks.  repair can check
+ * those kinds of things.
+ *
+ * So with that in mind, if we're setting a block to be data or rldata,
+ * don't complain so long as the block is currently unknown, data, or rldata.
+ * Don't let blocks downgrade from rldata -> data.
+ */
+static bool
+is_reflink(
+	dbm_t		type2)
+{
+	if (!xfs_sb_version_hasreflink(&mp->m_sb))
+		return false;
+	if (type2 == DBM_DATA || type2 == DBM_RLDATA)
+		return true;
+	return false;
+}
+
 static void
 check_set_dbmap(
 	xfs_agnumber_t	agno,
@@ -1562,10 +1598,15 @@ check_set_dbmap(
 			agbno, agbno + len - 1, c_agno, c_agbno);
 		return;
 	}
-	check_dbmap(agno, agbno, len, type1);
+	check_dbmap(agno, agbno, len, type1, is_reflink(type2));
 	mayprint = verbose | blist_size;
 	for (i = 0, p = &dbmap[agno][agbno]; i < len; i++, p++) {
-		*p = (char)type2;
+		if (*p == DBM_RLDATA && type2 == DBM_DATA)
+			;	/* do nothing */
+		if (*p == DBM_DATA && type2 == DBM_DATA)
+			*p = (char)DBM_RLDATA;
+		else
+			*p = (char)type2;
 		if (mayprint && (verbose || CHECK_BLISTA(agno, agbno + i)))
 			dbprintf(_("setting block %u/%u to %s\n"), agno, agbno + i,
 				typename[type2]);
@@ -2807,6 +2848,7 @@ process_inode(
 		type = DBM_UNKNOWN;
 		break;
 	}
+	id->isreflink = !!(idic.di_flags2 & XFS_DIFLAG2_REFLINK);
 	if (idic.di_version == 1)
 		setlink_inode(id, idic.di_onlink, type == DBM_DIR, security);
 	else {
@@ -3919,6 +3961,12 @@ scan_ag(
 			be32_to_cpu(agf->agf_levels[XFS_BTNUM_RMAP]),
 			1, scanfunc_rmap, TYP_RMAPBT);
 	}
+	if (agf->agf_refcount_root) {
+		scan_sbtree(agf,
+			be32_to_cpu(agf->agf_refcount_root),
+			be32_to_cpu(agf->agf_refcount_level),
+			1, scanfunc_refcnt, TYP_REFCBT);
+	}
 	scan_sbtree(agf,
 		be32_to_cpu(agi->agi_root),
 		be32_to_cpu(agi->agi_level),
@@ -4591,6 +4639,78 @@ scanfunc_rmap(
 }
 
 static void
+scanfunc_refcnt(
+	struct xfs_btree_block	*block,
+	int			level,
+	struct xfs_agf		*agf,
+	xfs_agblock_t		bno,
+	int			isroot)
+{
+	xfs_agnumber_t		seqno = be32_to_cpu(agf->agf_seqno);
+	int			i;
+	xfs_refcount_ptr_t	*pp;
+	struct xfs_refcount_rec	*rp;
+	xfs_agblock_t		lastblock;
+
+	if (be32_to_cpu(block->bb_magic) != XFS_REFC_CRC_MAGIC) {
+		dbprintf(_("bad magic # %#x in refcntbt block %u/%u\n"),
+			be32_to_cpu(block->bb_magic), seqno, bno);
+		serious_error++;
+		return;
+	}
+	if (be16_to_cpu(block->bb_level) != level) {
+		if (!sflag)
+			dbprintf(_("expected level %d got %d in refcntbt block "
+				 "%u/%u\n"),
+				level, be16_to_cpu(block->bb_level), seqno, bno);
+		error++;
+	}
+	set_dbmap(seqno, bno, 1, DBM_BTREFC, seqno, bno);
+	if (level == 0) {
+		if (be16_to_cpu(block->bb_numrecs) > mp->m_refc_mxr[0] ||
+		    (isroot == 0 && be16_to_cpu(block->bb_numrecs) < mp->m_refc_mnr[0])) {
+			dbprintf(_("bad btree nrecs (%u, min=%u, max=%u) in "
+				 "refcntbt block %u/%u\n"),
+				be16_to_cpu(block->bb_numrecs), mp->m_refc_mnr[0],
+				mp->m_refc_mxr[0], seqno, bno);
+			serious_error++;
+			return;
+		}
+		rp = XFS_REFCOUNT_REC_ADDR(block, 1);
+		lastblock = 0;
+		for (i = 0; i < be16_to_cpu(block->bb_numrecs); i++) {
+			set_dbmap(seqno, be32_to_cpu(rp[i].rc_startblock),
+				be32_to_cpu(rp[i].rc_blockcount), DBM_RLDATA,
+				seqno, bno);
+			if (be32_to_cpu(rp[i].rc_startblock) < lastblock) {
+				dbprintf(_(
+		"out-of-order refcnt btree record %d (%u %u) block %u/%u\n"),
+					 i, be32_to_cpu(rp[i].rc_startblock),
+					 be32_to_cpu(rp[i].rc_startblock),
+					 be32_to_cpu(agf->agf_seqno), bno);
+			} else {
+				lastblock = be32_to_cpu(rp[i].rc_startblock) +
+					    be32_to_cpu(rp[i].rc_blockcount);
+			}
+		}
+		return;
+	}
+	if (be16_to_cpu(block->bb_numrecs) > mp->m_refc_mxr[1] ||
+	    (isroot == 0 && be16_to_cpu(block->bb_numrecs) < mp->m_refc_mnr[1])) {
+		dbprintf(_("bad btree nrecs (%u, min=%u, max=%u) in refcntbt "
+			 "block %u/%u\n"),
+			be16_to_cpu(block->bb_numrecs), mp->m_refc_mnr[1],
+			mp->m_refc_mxr[1], seqno, bno);
+		serious_error++;
+		return;
+	}
+	pp = XFS_REFCOUNT_PTR_ADDR(block, 1, mp->m_refc_mxr[1]);
+	for (i = 0; i < be16_to_cpu(block->bb_numrecs); i++)
+		scan_sbtree(agf, be32_to_cpu(pp[i]), level, 0, scanfunc_refcnt,
+				TYP_REFCBT);
+}
+
+static void
 set_dbmap(
 	xfs_agnumber_t	agno,
 	xfs_agblock_t	agbno,

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

  parent reply	other threads:[~2015-10-07  5:09 UTC|newest]

Thread overview: 68+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2015-10-07  5:05 [RFCv3 00/51] xfsprogs: add reverse-mapping, reflink, and dedupe support Darrick J. Wong
2015-10-07  5:05 ` [PATCH 01/51] libxcmd: provide a common function to report command runtimes Darrick J. Wong
2015-10-13 17:48   ` Christoph Hellwig
2015-10-13 22:39     ` Darrick J. Wong
2015-10-14  5:35   ` [PATCH v2 " Darrick J. Wong
2015-10-14  7:31     ` Christoph Hellwig
2015-10-07  5:05 ` [PATCH 02/51] libxfs: add reflink and dedupe ioctls Darrick J. Wong
2015-10-07  5:05 ` [PATCH 03/51] xfs_io: support reflink and dedupe of file ranges Darrick J. Wong
2015-10-14  5:36   ` [PATCH v2 " Darrick J. Wong
2015-11-09  7:54     ` Christoph Hellwig
2015-11-09 18:33       ` Darrick J. Wong
2015-11-09 18:57         ` Darrick J. Wong
2015-11-09 21:35           ` Dave Chinner
2015-11-10  6:27             ` Darrick J. Wong
2015-10-07  5:05 ` [PATCH 04/51] xfs_io: unshare blocks via fallocate Darrick J. Wong
2015-10-07  5:05 ` [PATCH 05/51] xfs_db: enable blocktrash for checksummed filesystems Darrick J. Wong
2015-10-07  5:05 ` [PATCH 06/51] xfs_db: trash the block at the top of the cursor stack Darrick J. Wong
2015-10-07  5:05 ` [PATCH 07/51] xfs_db: enable blockget for v5 filesystems Darrick J. Wong
2015-10-14 17:08   ` Christoph Hellwig
2015-10-14 18:20     ` Darrick J. Wong
2015-10-14 18:23       ` Christoph Hellwig
2015-10-14 19:52         ` Darrick J. Wong
2015-10-14 21:26         ` Dave Chinner
2015-10-07  5:06 ` [PATCH 08/51] libxfs: reorder xfs_bmap_add_free args Darrick J. Wong
2015-10-07  5:06 ` [PATCH 09/51] libxfs: add the reverse-mapping btree Darrick J. Wong
2015-10-07  5:06 ` [PATCH 10/51] libxfs: resync xfs_prealloc_blocks with the kernel Darrick J. Wong
2015-10-07  5:06 ` [PATCH 11/51] xfs: rmap btree transaction reservations Darrick J. Wong
2015-10-07  5:06 ` [PATCH 12/51] xfs: rmap btree requires more reserved free space Darrick J. Wong
2015-10-07  5:06 ` [PATCH 13/51] libxfs: propagate a bunch of case changes to mkfs and repair Darrick J. Wong
2015-10-07  5:06 ` [PATCH 14/51] libxfs: fix min freelist length calculation Darrick J. Wong
2015-10-07  5:06 ` [PATCH 15/51] libxfs: add the RMAP CRC to the xfs_magics list Darrick J. Wong
2015-10-07  5:06 ` [PATCH 16/51] libxfs: enhance rmapbt definition to support reflink Darrick J. Wong
2015-10-07  5:07 ` [PATCH 17/51] libxfs: refactor short btree block verification Darrick J. Wong
2015-10-07  5:07 ` [PATCH 18/51] xfs: don't update rmapbt when fixing agfl Darrick J. Wong
2015-10-07  5:07 ` [PATCH 19/51] libxfs: implement XFS_IOC_SWAPEXT when rmap btree is enabled Darrick J. Wong
2015-10-07  5:07 ` [PATCH 20/51] xfs_db: display rmap btree contents Darrick J. Wong
2015-10-07  5:07 ` [PATCH 21/51] xfs_dump: display enhanced rmapbt fields Darrick J. Wong
2015-10-07  5:07 ` [PATCH 22/51] xfs_db: check rmapbt Darrick J. Wong
2015-10-07  5:07 ` [PATCH 23/51] xfs_db: copy the rmap btree Darrick J. Wong
2015-10-07  5:07 ` [PATCH 24/51] xfs_growfs: report rmapbt presence Darrick J. Wong
2015-10-07  5:07 ` [PATCH 25/51] xfs_repair: use rmap btree data to check block types Darrick J. Wong
2015-10-07  5:08 ` [PATCH 26/51] xfs_repair: mask off length appropriately Darrick J. Wong
2015-10-07  5:08 ` [PATCH 27/51] xfs_repair: fix fino_bno calculation when rmapbt is enabled Darrick J. Wong
2015-10-07  5:08 ` [PATCH 28/51] xfs_repair: create a slab API for allocating arrays in large chunks Darrick J. Wong
2015-10-07  5:08 ` [PATCH 29/51] xfs_repair: collect reverse-mapping data for refcount/rmap tree rebuilding Darrick J. Wong
2015-10-07  5:08 ` [PATCH 30/51] xfs_repair: record and merge raw rmap data Darrick J. Wong
2015-10-07  5:08 ` [PATCH 31/51] xfs_repair: add inode bmbt block rmaps Darrick J. Wong
2015-10-07  5:08 ` [PATCH 32/51] xfs_repair: add fixed-location per-AG rmaps Darrick J. Wong
2015-10-21 21:08   ` Darrick J. Wong
2015-10-07  5:08 ` [PATCH 33/51] xfs_repair: check existing rmapbt entries against observed rmaps Darrick J. Wong
2015-10-07  5:08 ` [PATCH 34/51] xfs_repair: rebuild reverse-mapping btree Darrick J. Wong
2015-10-07  5:09 ` [PATCH 35/51] xfs_repair: add per-AG btree blocks to rmap data and add to rmapbt Darrick J. Wong
2015-10-07  5:09 ` [PATCH 36/51] mkfs.xfs: Create rmapbt filesystems Darrick J. Wong
2015-10-07  5:09 ` [PATCH 37/51] xfs_mkfs: initialize extra fields during mkfs Darrick J. Wong
2015-10-07  5:09 ` [PATCH 38/51] libxfs: add support for refcount btrees Darrick J. Wong
2015-10-07  5:09 ` [PATCH 39/51] xfs_db: dump refcount btree data Darrick J. Wong
2015-10-07  5:09 ` Darrick J. Wong [this message]
2015-10-07  5:09 ` [PATCH 41/51] xfs_db: metadump should copy the refcount btree too Darrick J. Wong
2015-10-07  5:09 ` [PATCH 42/51] xfs_growfs: report the presence of the reflink feature Darrick J. Wong
2015-10-07  5:09 ` [PATCH 43/51] xfs_repair: check the existing refcount btree Darrick J. Wong
2015-10-07  5:09 ` [PATCH 44/51] xfs_repair: handle multiple owners of data blocks Darrick J. Wong
2015-10-07  5:10 ` [PATCH 45/51] xfs_repair: process reverse-mapping data into refcount data Darrick J. Wong
2015-10-07  5:10 ` [PATCH 46/51] xfs_repair: record reflink inode state Darrick J. Wong
2015-10-07  5:10 ` [PATCH 47/51] xfs_repair: fix inode reflink flags Darrick J. Wong
2015-10-07  5:10 ` [PATCH 48/51] xfs_repair: check the refcount btree against our observed reference counts when -n Darrick J. Wong
2015-10-07  5:10 ` [PATCH 49/51] xfs_repair: rebuild the refcount btree Darrick J. Wong
2015-10-07  5:10 ` [PATCH 50/51] mkfs.xfs: format reflink enabled filesystems Darrick J. Wong
2015-10-07  5:10 ` [PATCH 51/51] mkfs: hack around not having enough log blocks Darrick J. Wong

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20151007050933.1504.94823.stgit@birch.djwong.org \
    --to=darrick.wong@oracle.com \
    --cc=david@fromorbit.com \
    --cc=xfs@oss.sgi.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.