xfs: online repair of realtime summariesrepair-rtsummary_2020-10-26

Repair the realtime summary data by constructing a new rtsummary file in
the scrub temporary file, then atomically swapping the contents.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>

6 files changed, 283 insertions, 4 deletions

diff --git a/fs/xfs/Makefile b/fs/xfs/Makefile
index e46c622561ab..f47c9d6a43f4 100644
--- a/fs/xfs/Makefile
+++ b/fs/xfs/Makefile
@@ -185,6 +185,10 @@ xfs-y				+= $(addprefix scrub/, \
 				   symlink_repair.o \
 				   )
 
+xfs-$(CONFIG_XFS_RT)		+= $(addprefix scrub/, \
+				   rtsummary_repair.o \
+				   )
+
 xfs-$(CONFIG_XFS_QUOTA)		+= $(addprefix scrub/, \
 				   quota_repair.o \
 				   quotacheck_repair.o \
diff --git a/fs/xfs/scrub/repair.c b/fs/xfs/scrub/repair.c
index 222ab8746513..8a5d2ea26fdf 100644
--- a/fs/xfs/scrub/repair.c
+++ b/fs/xfs/scrub/repair.c
@@ -39,6 +39,7 @@
 #include "scrub/trace.h"
 #include "scrub/repair.h"
 #include "scrub/bitmap.h"
+#include "scrub/xfile.h"
 
 /*
  * Attempt to repair some metadata, if the metadata is corrupt and userspace
@@ -165,9 +166,21 @@ int
 xrep_roll_trans(
 	struct xfs_scrub	*sc)
 {
+	int			error;
+
 	if (!sc->ip)
 		return xrep_roll_ag_trans(sc);
-	return xfs_trans_roll_inode(&sc->tp, sc->ip);
+
+	/*
+	 * Roll the transaction with the inode we're fixing and the temp inode,
+	 * so that neither can pin the log.
+	 */
+	if (sc->tempip)
+		xfs_trans_log_inode(sc->tp, sc->tempip, XFS_ILOG_CORE);
+	error = xfs_trans_roll_inode(&sc->tp, sc->ip);
+	if (sc->tempip)
+		xfs_trans_ijoin(sc->tp, sc->tempip, 0);
+	return error;
 }
 
 /*
@@ -1618,3 +1631,165 @@ out_release_inode:
 
 	return error;
 }
+
+/*
+ * Make sure that the given range of the data fork of the temporary file is
+ * mapped to written blocks.  The caller must ensure that both inodes are
+ * joined to the transaction.
+ */
+int
+xrep_fallocate(
+	struct xfs_scrub	*sc,
+	xfs_fileoff_t		off,
+	xfs_filblks_t		len)
+{
+	struct xfs_bmbt_irec	map;
+	xfs_fileoff_t		end = off + len;
+	int			nmaps;
+	int			error = 0;
+
+	ASSERT(sc->tempip != NULL);
+	ASSERT(!XFS_NOT_DQATTACHED(sc->mp, sc->tempip));
+
+	while (off < len) {
+		/*
+		 * If we have a real extent mapping this block then we're
+		 * in ok shape.
+		 */
+		nmaps = 1;
+		error = xfs_bmapi_read(sc->tempip, off, end - off, &map, &nmaps,
+				XFS_DATA_FORK);
+		if (error)
+			break;
+
+		if (nmaps == 1 && xfs_bmap_is_written_extent(&map)) {
+			off += map.br_startblock;
+			continue;
+		}
+
+		/*
+		 * If we find a delalloc reservation then something is very
+		 * very wrong.  Bail out.
+		 */
+		if (map.br_startblock == DELAYSTARTBLOCK)
+			return -EFSCORRUPTED;
+
+		/*
+		 * Make sure this rtsum block has a real zeroed extent
+		 * allocated to it.
+		 */
+		nmaps = 1;
+		error = xfs_bmapi_write(sc->tp, sc->tempip, off, end - off,
+				XFS_BMAPI_CONVERT | XFS_BMAPI_ZERO, 0, &map,
+				&nmaps);
+		if (error)
+			break;
+
+		error = xrep_roll_trans(sc);
+		if (error)
+			break;
+		off += map.br_startblock;
+	}
+
+	return error;
+}
+
+/*
+ * Write a number of bytes from the xfile into the temp file.  The copybuf must
+ * be large enough to hold one filesystem block's worth of data.  The caller
+ * must join both inodes to the transaction.
+ */
+int
+xrep_set_file_contents(
+	struct xfs_scrub	*sc,
+	const struct xfs_buf_ops *ops,
+	enum xfs_blft		type,
+	xfs_fileoff_t		isize)
+{
+	LIST_HEAD(buffers_list);
+	struct xfs_bmbt_irec	map;
+	struct xfs_mount	*mp = sc->mp;
+	struct xfs_buf		*bp;
+	xfs_rtblock_t		off = 0;
+	loff_t			pos = 0;
+	unsigned int		nr_buffers = 0;
+	int			nmaps;
+	int			error = 0;
+
+	ASSERT(S_ISREG(VFS_I(sc->tempip)->i_mode));
+
+	for (; pos < isize; pos += mp->m_sb.sb_blocksize, off++) {
+		size_t		count;
+
+		/* Read block mapping for this file block. */
+		nmaps = 1;
+		error = xfs_bmapi_read(sc->tempip, off, 1, &map, &nmaps, 0);
+		if (error)
+			goto out;
+		if (nmaps == 0 || !xfs_bmap_is_written_extent(&map)) {
+			error = -EFSCORRUPTED;
+			goto out;
+		}
+
+		/* Get the metadata buffer for this offset in the file. */
+		error = xfs_trans_get_buf(sc->tp, mp->m_ddev_targp,
+				XFS_FSB_TO_DADDR(mp, map.br_startblock),
+				mp->m_bsize, 0, &bp);
+		if (error)
+			goto out;
+		bp->b_ops = ops;
+		xfs_trans_buf_set_type(sc->tp, bp, type);
+
+		/* Read in a block's worth of data from the xfile. */
+		count = min_t(loff_t, isize - pos, mp->m_sb.sb_blocksize);
+		error = xfile_pread(sc->xfile, bp->b_addr, count, pos);
+		if (error) {
+			xfs_trans_brelse(sc->tp, bp);
+			goto out;
+		}
+
+		/*
+		 * Put this buffer on the delwri list so we can write them all
+		 * out in batches.
+		 */
+		xfs_buf_delwri_queue(bp, &buffers_list);
+		xfs_trans_brelse(sc->tp, bp);
+		nr_buffers++;
+
+		/*
+		 * If we have more than 256K of data to write out, flush it to
+		 * disk so we don't use up too much memory.
+		 */
+		if (XFS_FSB_TO_B(mp, nr_buffers) > 262144) {
+			error = xfs_buf_delwri_submit(&buffers_list);
+			if (error)
+				goto out;
+			nr_buffers = 0;
+		}
+	}
+
+	/*
+	 * Write the new blocks to disk.  If the ordered list isn't empty after
+	 * that, then something went wrong and we have to fail.  This should
+	 * never happen, but we'll check anyway.
+	 */
+	error = xfs_buf_delwri_submit(&buffers_list);
+	if (error)
+		goto out;
+	if (!list_empty(&buffers_list)) {
+		ASSERT(list_empty(&buffers_list));
+		return -EIO;
+	}
+
+	/* Set the new inode size, if needed. */
+	if (sc->tempip->i_d.di_size != isize) {
+		sc->tempip->i_d.di_size = isize;
+		xfs_trans_log_inode(sc->tp, sc->tempip, XFS_ILOG_CORE);
+		return xrep_roll_trans(sc);
+	}
+
+	return 0;
+out:
+	xfs_buf_delwri_cancel(&buffers_list);
+	return error;
+}
diff --git a/fs/xfs/scrub/repair.h b/fs/xfs/scrub/repair.h
index 98399ebf7f08..ab45b85d3659 100644
--- a/fs/xfs/scrub/repair.h
+++ b/fs/xfs/scrub/repair.h
@@ -22,6 +22,9 @@ static inline int xrep_notsupported(struct xfs_scrub *sc)
 
 /* Repair helpers */
 
+enum xfs_blft;
+struct xbitmap;
+
 int xrep_attempt(struct xfs_inode *ip, struct xfs_scrub *sc);
 void xrep_failure(struct xfs_mount *mp);
 int xrep_roll_ag_trans(struct xfs_scrub *sc);
@@ -36,8 +39,13 @@ int xrep_init_btblock(struct xfs_scrub *sc, xfs_fsblock_t fsb,
 		struct xfs_buf **bpp, xfs_btnum_t btnum,
 		const struct xfs_buf_ops *ops);
 int xrep_setup_tempfile(struct xfs_scrub *sc, uint16_t mode);
+int xrep_fallocate(struct xfs_scrub *sc, xfs_fileoff_t off, xfs_filblks_t len);
 
-struct xbitmap;
+typedef int (*xrep_setfile_getbuf_fn)(struct xfs_scrub *sc,
+		xfs_fileoff_t off, struct xfs_buf **bpp);
+int xrep_set_file_contents(struct xfs_scrub *sc,
+		const struct xfs_buf_ops *ops, enum xfs_blft type,
+		xfs_fileoff_t isize);
 
 int xrep_fix_freelist(struct xfs_scrub *sc, int alloc_flags);
 int xrep_reap_extents(struct xfs_scrub *sc, struct xbitmap *exlist,
@@ -93,6 +101,12 @@ int xrep_quotacheck(struct xfs_scrub *sc);
 # define xrep_quotacheck		xrep_notsupported
 #endif /* CONFIG_XFS_QUOTA */
 
+#ifdef CONFIG_XFS_RT
+int xrep_rtsummary(struct xfs_scrub *sc);
+#else
+# define xrep_rtsummary			xrep_notsupported
+#endif /* CONFIG_XFS_RT */
+
 struct xrep_newbt_resv {
 	/* Link to list of extents that we've reserved. */
 	struct list_head	list;
@@ -209,6 +223,7 @@ xrep_rmapbt_setup(
 #define xrep_quota			xrep_notsupported
 #define xrep_quotacheck			xrep_notsupported
 #define xrep_fscounters			xrep_notsupported
+#define xrep_rtsummary			xrep_notsupported
 
 #endif /* CONFIG_XFS_ONLINE_REPAIR */
 
diff --git a/fs/xfs/scrub/rtsummary.c b/fs/xfs/scrub/rtsummary.c
index e6c7b52c2edf..149616f2d392 100644
--- a/fs/xfs/scrub/rtsummary.c
+++ b/fs/xfs/scrub/rtsummary.c
@@ -20,6 +20,7 @@
 #include "scrub/common.h"
 #include "scrub/trace.h"
 #include "scrub/xfile.h"
+#include "scrub/repair.h"
 
 /*
  * Realtime Summary
@@ -57,8 +58,13 @@ xchk_setup_rtsummary(
 	struct xfs_inode	*ip)
 {
 	struct xfs_mount	*mp = sc->mp;
+	unsigned long long	resblks = 0;
 	int			error;
 
+	error = xrep_setup_tempfile(sc, S_IFREG);
+	if (error)
+		return error;
+
 	/*
 	 * Create an xfile to construct a new rtsummary file.  The xfile allows
 	 * us to avoid pinning kernel memory for this purpose.
@@ -67,7 +73,21 @@ xchk_setup_rtsummary(
 	if (IS_ERR(sc->xfile))
 		return PTR_ERR(sc->xfile);
 
-	error = xchk_trans_alloc(sc, 0);
+	/*
+	 * If we're doing a repair, we reserve 2x the summary blocks: once for
+	 * the new summary contents and again for the bmbt blocks and the
+	 * remapping operation.  We cannot use xfs_swapext_estimate because we
+	 * have not yet constructed the replacement rtsummary and therefore do
+	 * not know how many extents it will use.  By the time we do, we will
+	 * have a dirty transaction (which we cannot drop because we cannot
+	 * drop the rtsummary ILOCK) and cannot ask for more reservation.
+	 */
+	if (sc->sm->sm_flags & XFS_SCRUB_IFLAG_REPAIR) {
+		resblks = XFS_B_TO_FSB(sc->mp, sc->mp->m_rsumsize) * 2;
+		if (resblks > UINT_MAX)
+			return -EOPNOTSUPP;
+	}
+	error = xchk_trans_alloc(sc, resblks);
 	if (error)
 		return error;
 
diff --git a/fs/xfs/scrub/rtsummary_repair.c b/fs/xfs/scrub/rtsummary_repair.c
new file mode 100644
index 000000000000..5f9ec44a0577
--- /dev/null
+++ b/fs/xfs/scrub/rtsummary_repair.c
@@ -0,0 +1,65 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/*
+ * Copyright (C) 2020 Oracle.  All Rights Reserved.
+ * Author: Darrick J. Wong <darrick.wong@oracle.com>
+ */
+#include "xfs.h"
+#include "xfs_fs.h"
+#include "xfs_shared.h"
+#include "xfs_format.h"
+#include "xfs_trans_resv.h"
+#include "xfs_mount.h"
+#include "xfs_btree.h"
+#include "xfs_log_format.h"
+#include "xfs_trans.h"
+#include "xfs_rtalloc.h"
+#include "xfs_inode.h"
+#include "xfs_bit.h"
+#include "xfs_bmap.h"
+#include "xfs_swapext.h"
+#include "scrub/scrub.h"
+#include "scrub/common.h"
+#include "scrub/trace.h"
+#include "scrub/repair.h"
+
+/* Repair the realtime summary. */
+int
+xrep_rtsummary(
+	struct xfs_scrub	*sc)
+{
+	struct xfs_swapext_req	req = { .flags = 0 };
+	int			error;
+
+	/* Make sure any problems with the fork are fixed. */
+	error = xrep_metadata_inode_forks(sc);
+	if (error)
+		return error;
+
+	/*
+	 * Trylock the temporary file.  We had better be the only ones holding
+	 * onto this inode...
+	 */
+	if (!xfs_ilock_nowait(sc->tempip, XFS_ILOCK_EXCL))
+		return -EAGAIN;
+	sc->temp_ilock_flags = XFS_ILOCK_EXCL;
+
+	/* Make sure we have space allocated for the entire summary file. */
+	xfs_trans_ijoin(sc->tp, sc->ip, 0);
+	xfs_trans_ijoin(sc->tp, sc->tempip, 0);
+	error = xrep_fallocate(sc, 0, XFS_B_TO_FSB(sc->mp, sc->mp->m_rsumsize));
+	if (error)
+		return error;
+
+	/* Copy the rtsummary file that we generated. */
+	error = xrep_set_file_contents(sc, &xfs_rtbuf_ops,
+			XFS_BLFT_RTSUMMARY_BUF, sc->mp->m_rsumsize);
+	if (error)
+		return error;
+
+	/* Now swap the extents. */
+	req.ip1 = sc->tempip;
+	req.ip2 = sc->ip;
+	req.whichfork = XFS_DATA_FORK;
+	req.blockcount = XFS_B_TO_FSB(sc->mp, sc->mp->m_rsumsize);
+	return xfs_swapext(&sc->tp, &req);
+}
diff --git a/fs/xfs/scrub/scrub.c b/fs/xfs/scrub/scrub.c
index fa46a9958c9f..3a54c6730c1b 100644
--- a/fs/xfs/scrub/scrub.c
+++ b/fs/xfs/scrub/scrub.c
@@ -342,7 +342,7 @@ static const struct xchk_meta_ops meta_scrub_ops[] = {
 		.setup	= xchk_setup_rtsummary,
 		.scrub	= xchk_rtsummary,
 		.has	= xfs_sb_version_hasrealtime,
-		.repair	= xrep_notsupported,
+		.repair	= xrep_rtsummary,
 	},
 	[XFS_SCRUB_TYPE_UQUOTA] = {	/* user quota */
 		.type	= ST_FS,