From 1804f4671b76293a8a792e5be4e514cefc0f012f Mon Sep 17 00:00:00 2001
From: "Darrick J. Wong" <djwong@kernel.org>
Date: Wed, 1 Sep 2021 11:25:04 -0700
Subject: xfs: experiment with dontcache when scanning inodes

Add some experimental flags to drop inodes from the cache after a scan.

Signed-off-by: Darrick J. Wong <djwong@kernel.org>
---
 fs/xfs/libxfs/xfs_fs.h       | 17 ++++++++++++++---
 fs/xfs/scrub/common.c        | 25 +++++++++++++++++++++----
 fs/xfs/scrub/common.h        |  3 +++
 fs/xfs/scrub/dir.c           |  2 +-
 fs/xfs/scrub/dir_repair.c    |  4 ++--
 fs/xfs/scrub/inode_repair.c  |  3 +--
 fs/xfs/scrub/iscan.c         |  4 ++--
 fs/xfs/scrub/parent.c        |  2 +-
 fs/xfs/scrub/parent_repair.c |  6 +++---
 fs/xfs/scrub/quotacheck.c    |  2 +-
 fs/xfs/scrub/rmap_repair.c   |  2 +-
 fs/xfs/scrub/rtrmap_repair.c |  2 +-
 fs/xfs/scrub/scrub.c         | 11 +++++------
 fs/xfs/scrub/trace.h         |  3 ++-
 fs/xfs/xfs_ioctl.c           |  3 +++
 fs/xfs/xfs_itable.c          |  8 +++++---
 fs/xfs/xfs_itable.h          |  3 +++
 fs/xfs/xfs_iwalk.h           |  6 +++++-
 18 files changed, 74 insertions(+), 32 deletions(-)

diff --git a/fs/xfs/libxfs/xfs_fs.h b/fs/xfs/libxfs/xfs_fs.h
index 9a4bff3187b8..0b03f198442e 100644
--- a/fs/xfs/libxfs/xfs_fs.h
+++ b/fs/xfs/libxfs/xfs_fs.h
@@ -502,9 +502,13 @@ struct xfs_bulk_ireq {
  */
 #define XFS_BULK_IREQ_METADIR	(1 << 2)
 
+/* Don't mark inodes DONTCACHE */
+#define XFS_BULK_IREQ_RETAIN_INODES	(1 << 3)
+
 #define XFS_BULK_IREQ_FLAGS_ALL	(XFS_BULK_IREQ_AGNO | \
 				 XFS_BULK_IREQ_SPECIAL | \
-				 XFS_BULK_IREQ_METADIR)
+				 XFS_BULK_IREQ_METADIR | \
+				 XFS_BULK_IREQ_RETAIN_INODES)
 
 /* Operate on the root directory inode. */
 #define XFS_BULK_IREQ_SPECIAL_ROOT	(1)
@@ -785,8 +789,12 @@ struct xfs_scrub_metadata {
 /* i: Allow scrub to freeze the filesystem to perform global scans. */
 #define XFS_SCRUB_IFLAG_FREEZE_OK	(1 << 8)
 
+/* i: Don't mark inodes DONTCACHE at the end. */
+#define XFS_SCRUB_IFLAG_RETAIN_INODES	(1 << 9)
+
 #define XFS_SCRUB_FLAGS_IN	(XFS_SCRUB_IFLAG_REPAIR | \
-				 XFS_SCRUB_IFLAG_FREEZE_OK)
+				 XFS_SCRUB_IFLAG_FREEZE_OK | \
+				 XFS_SCRUB_IFLAG_RETAIN_INODES)
 #define XFS_SCRUB_FLAGS_OUT	(XFS_SCRUB_OFLAG_CORRUPT | \
 				 XFS_SCRUB_OFLAG_PREEN | \
 				 XFS_SCRUB_OFLAG_XFAIL | \
@@ -815,7 +823,10 @@ struct xfs_scrub_vec_head {
 	struct xfs_scrub_vec svh_vecs[0];
 };
 
-#define XFS_SCRUB_VEC_FLAGS_ALL		(0)
+/* i: Don't mark inodes DONTCACHE at the end. */
+#define XFS_SCRUB_VEC_IFLAG_RETAIN_INODES	(1 << 0)
+
+#define XFS_SCRUB_VEC_FLAGS_ALL		(XFS_SCRUB_VEC_IFLAG_RETAIN_INODES)
 
 static inline size_t sizeof_xfs_scrub_vec(unsigned int nr)
 {
diff --git a/fs/xfs/scrub/common.c b/fs/xfs/scrub/common.c
index 0055299170e2..0ea9b6b299ae 100644
--- a/fs/xfs/scrub/common.c
+++ b/fs/xfs/scrub/common.c
@@ -863,8 +863,7 @@ xchk_get_inode(
 	/* Look up the inode, see if the generation number matches. */
 	if (xfs_internal_inum(mp, sc->sm->sm_ino))
 		return -ENOENT;
-	error = xfs_iget(mp, NULL, sc->sm->sm_ino,
-			XFS_IGET_UNTRUSTED | XFS_IGET_DONTCACHE, 0, &ip);
+	error = xfs_iget(mp, NULL, sc->sm->sm_ino, XFS_IGET_UNTRUSTED, 0, &ip);
 	switch (error) {
 	case -ENOENT:
 		/* Inode doesn't exist, just bail out. */
@@ -886,7 +885,7 @@ xchk_get_inode(
 		 * that it no longer exists.
 		 */
 		error = xfs_imap(sc->mp, sc->tp, sc->sm->sm_ino, &imap,
-				XFS_IGET_UNTRUSTED | XFS_IGET_DONTCACHE);
+				XFS_IGET_UNTRUSTED);
 		if (error)
 			return -ENOENT;
 		error = -EFSCORRUPTED;
@@ -911,7 +910,7 @@ xchk_get_inode(
 	 */
 	if (VFS_I(ip)->i_generation != sc->sm->sm_gen ||
 	    (xfs_is_metadata_inode(ip) && !S_ISDIR(VFS_I(ip)->i_mode))) {
-		xfs_irele(ip);
+		xchk_irele(sc, ip);
 		return -ENOENT;
 	}
 
@@ -919,6 +918,24 @@ xchk_get_inode(
 	return 0;
 }
 
+void
+__xchk_irele(
+	struct xfs_inode	*ip,
+	bool			set_dontcache)
+{
+	if (set_dontcache && atomic_read(&VFS_I(ip)->i_count) == 1)
+		d_mark_dontcache(VFS_I(ip));
+	xfs_irele(ip);
+}
+
+void
+xchk_irele(
+	struct xfs_scrub	*sc,
+	struct xfs_inode	*ip)
+{
+	__xchk_irele(ip, !(sc->sm->sm_flags & XFS_SCRUB_IFLAG_RETAIN_INODES));
+}
+
 /* Set us up to scrub a file's contents. */
 int
 xchk_setup_inode_contents(
diff --git a/fs/xfs/scrub/common.h b/fs/xfs/scrub/common.h
index ead7789fba5f..72c7d2636b56 100644
--- a/fs/xfs/scrub/common.h
+++ b/fs/xfs/scrub/common.h
@@ -177,6 +177,9 @@ void xchk_iunlock(struct xfs_scrub *sc, unsigned int ilock_flags);
 void xchk_buffer_recheck(struct xfs_scrub *sc, struct xfs_buf *bp);
 void xchk_whine(const struct xfs_mount *mp, const char *fmt, ...);
 
+void __xchk_irele(struct xfs_inode *ip, bool set_dontcache);
+void xchk_irele(struct xfs_scrub *sc, struct xfs_inode *ip);
+
 /*
  * Don't bother cross-referencing if we already found corruption or cross
  * referencing discrepancies.
diff --git a/fs/xfs/scrub/dir.c b/fs/xfs/scrub/dir.c
index 31e3626db100..7b8fafe4cffa 100644
--- a/fs/xfs/scrub/dir.c
+++ b/fs/xfs/scrub/dir.c
@@ -104,7 +104,7 @@ xchk_dir_check_ftype(
 	if (xfs_is_metadata_inode(ip) ^ xfs_is_metadata_inode(sdc->sc->ip))
 		xchk_fblock_set_corrupt(sdc->sc, XFS_DATA_FORK, 0);
 
-	xfs_irele(ip);
+	xchk_irele(sdc->sc, ip);
 out:
 	return error;
 }
diff --git a/fs/xfs/scrub/dir_repair.c b/fs/xfs/scrub/dir_repair.c
index b64be71d9884..0d1130ae455f 100644
--- a/fs/xfs/scrub/dir_repair.c
+++ b/fs/xfs/scrub/dir_repair.c
@@ -211,12 +211,12 @@ xrep_directory_salvage_entry(
 
 	/* Don't mix metadata and regular directory trees. */
 	if (xfs_is_metadata_inode(ip) ^ xfs_is_metadata_inode(rd->sc->ip)) {
-		xfs_irele(ip);
+		xchk_irele(rd->sc, ip);
 		return 0;
 	}
 
 	key.ftype = xfs_mode_to_ftype(VFS_I(ip)->i_mode);
-	xfs_irele(ip);
+	xchk_irele(rd->sc, ip);
 
 	/* Remember this for later. */
 	error = xfblob_store(rd->dir_names, &key.name_cookie, name,
diff --git a/fs/xfs/scrub/inode_repair.c b/fs/xfs/scrub/inode_repair.c
index c64adde40dce..97605313f097 100644
--- a/fs/xfs/scrub/inode_repair.c
+++ b/fs/xfs/scrub/inode_repair.c
@@ -1182,8 +1182,7 @@ xrep_dinode_core(
 		return error;
 
 	/* ...and reload it? */
-	error = xfs_iget(sc->mp, sc->tp, ino,
-			XFS_IGET_UNTRUSTED | XFS_IGET_DONTCACHE, 0, &sc->ip);
+	error = xfs_iget(sc->mp, sc->tp, ino, XFS_IGET_UNTRUSTED, 0, &sc->ip);
 	if (error)
 		return error;
 
diff --git a/fs/xfs/scrub/iscan.c b/fs/xfs/scrub/iscan.c
index ae056e6748f2..7b16d221b5e7 100644
--- a/fs/xfs/scrub/iscan.c
+++ b/fs/xfs/scrub/iscan.c
@@ -292,8 +292,8 @@ xchk_iscan_iget(
 	struct xfs_mount	*mp = sc->mp;
 	int			error;
 
-	error = xfs_iget(mp, sc->tp, iscan->cursor_ino,
-			XFS_IGET_DONTCACHE | XFS_IGET_UNTRUSTED, 0, ipp);
+	error = xfs_iget(mp, sc->tp, iscan->cursor_ino, XFS_IGET_UNTRUSTED, 0,
+			ipp);
 
 	trace_xchk_iscan_iget(mp, iscan, error);
 
diff --git a/fs/xfs/scrub/parent.c b/fs/xfs/scrub/parent.c
index e4de4cb6ea19..ec351a393b2d 100644
--- a/fs/xfs/scrub/parent.c
+++ b/fs/xfs/scrub/parent.c
@@ -270,7 +270,7 @@ xchk_parent_validate(
 out_unlock:
 	xfs_iunlock(dp, XFS_IOLOCK_SHARED);
 out_rele:
-	xfs_irele(dp);
+	xchk_irele(sc, dp);
 	return error;
 }
 
diff --git a/fs/xfs/scrub/parent_repair.c b/fs/xfs/scrub/parent_repair.c
index 28ffcee4f9e5..b9d244594913 100644
--- a/fs/xfs/scrub/parent_repair.c
+++ b/fs/xfs/scrub/parent_repair.c
@@ -285,7 +285,7 @@ xrep_parent_confirm(
 
 	*parent_ino = fpi.found_parent;
 out_rele:
-	xfs_irele(fpi.dp);
+	xchk_irele(sc, fpi.dp);
 	return error;
 }
 
@@ -319,7 +319,7 @@ xrep_parent_from_dcache(
 		ret = XFS_I(pip)->i_ino;
 	}
 
-	xfs_irele(XFS_I(pip));
+	xchk_irele(sc, XFS_I(pip));
 
 out_dput:
 	dput(dentry);
@@ -363,7 +363,7 @@ xrep_parent_scan(
 		if (S_ISDIR(VFS_I(fpi.dp)->i_mode))
 			ret = xrep_findparent_walk_directory(&fpi);
 		xchk_iscan_mark_visited(&iscan, fpi.dp);
-		xfs_irele(fpi.dp);
+		xchk_irele(sc, fpi.dp);
 		if (ret)
 			break;
 
diff --git a/fs/xfs/scrub/quotacheck.c b/fs/xfs/scrub/quotacheck.c
index 3c19bc157a03..d7cf1e1554ac 100644
--- a/fs/xfs/scrub/quotacheck.c
+++ b/fs/xfs/scrub/quotacheck.c
@@ -495,7 +495,7 @@ xqcheck_collect_counts(
 			break;
 
 		error = xqcheck_inode(xqc, ip);
-		xfs_irele(ip);
+		xchk_irele(sc, ip);
 		if (error)
 			break;
 
diff --git a/fs/xfs/scrub/rmap_repair.c b/fs/xfs/scrub/rmap_repair.c
index dfc0697809e7..876717816081 100644
--- a/fs/xfs/scrub/rmap_repair.c
+++ b/fs/xfs/scrub/rmap_repair.c
@@ -954,7 +954,7 @@ end_agscan:
 			break;
 
 		error = xrep_rmap_scan_inode(rr, ip);
-		xfs_irele(ip);
+		xchk_irele(sc, ip);
 		if (error)
 			break;
 
diff --git a/fs/xfs/scrub/rtrmap_repair.c b/fs/xfs/scrub/rtrmap_repair.c
index b91b1a378d15..f669f51f6f70 100644
--- a/fs/xfs/scrub/rtrmap_repair.c
+++ b/fs/xfs/scrub/rtrmap_repair.c
@@ -551,7 +551,7 @@ xrep_rtrmap_find_rmaps(
 			break;
 
 		error = xrep_rtrmap_scan_inode(rr, ip);
-		xfs_irele(ip);
+		xchk_irele(sc, ip);
 		if (error)
 			break;
 
diff --git a/fs/xfs/scrub/scrub.c b/fs/xfs/scrub/scrub.c
index 238dc7029070..f755c6d4ff5e 100644
--- a/fs/xfs/scrub/scrub.c
+++ b/fs/xfs/scrub/scrub.c
@@ -172,7 +172,7 @@ xchk_teardown(
 	if (sc->ip) {
 		if (sc->ilock_flags)
 			xchk_iunlock(sc, sc->ilock_flags);
-		xfs_irele(sc->ip);
+		xchk_irele(sc, sc->ip);
 		sc->ip = NULL;
 	}
 	if (sc->flags & XREP_ATOMIC_EXCHANGE) {
@@ -731,7 +731,8 @@ xfs_scrubv_metadata(
 		 * consider setting dontcache at the end.
 		 */
 		if (v->sv_type < XFS_SCRUB_TYPE_NR &&
-		    meta_scrub_ops[v->sv_type].type == ST_INODE)
+		    meta_scrub_ops[v->sv_type].type == ST_INODE &&
+		    !(vhead->svh_flags & XFS_SCRUB_VEC_IFLAG_RETAIN_INODES))
 			set_dontcache = true;
 
 		trace_xchk_scrubv_item(mp, vhead, v);
@@ -748,7 +749,7 @@ xfs_scrubv_metadata(
 		if (ip && (VFS_I(ip)->i_generation != vhead->svh_gen ||
 			   (xfs_is_metadata_inode(ip) &&
 			    !S_ISDIR(VFS_I(ip)->i_mode)))) {
-			xfs_irele(ip);
+			__xchk_irele(ip, set_dontcache);
 			ip = NULL;
 		}
 	}
@@ -803,8 +804,6 @@ xfs_scrubv_metadata(
 	 * If we're holding the only reference to this inode and the scan was
 	 * clean, mark it dontcache so that we don't pollute the cache.
 	 */
-	if (set_dontcache && atomic_read(&VFS_I(ip)->i_count) == 1)
-		d_mark_dontcache(VFS_I(ip));
-	xfs_irele(ip);
+	__xchk_irele(ip, set_dontcache);
 	return error;
 }
diff --git a/fs/xfs/scrub/trace.h b/fs/xfs/scrub/trace.h
index 04e3b3424b18..4b2b15d9ac1e 100644
--- a/fs/xfs/scrub/trace.h
+++ b/fs/xfs/scrub/trace.h
@@ -108,7 +108,8 @@ TRACE_DEFINE_ENUM(XFS_SCRUB_TYPE_BARRIER);
 	{ XFS_SCRUB_OFLAG_INCOMPLETE,		"incomplete" }, \
 	{ XFS_SCRUB_OFLAG_WARNING,		"warning" }, \
 	{ XFS_SCRUB_OFLAG_NO_REPAIR_NEEDED,	"norepair" }, \
-	{ XFS_SCRUB_IFLAG_FREEZE_OK,		"freeze" }
+	{ XFS_SCRUB_IFLAG_FREEZE_OK,		"freeze" }, \
+	{ XFS_SCRUB_IFLAG_RETAIN_INODES,	"icache" }
 
 DECLARE_EVENT_CLASS(xchk_class,
 	TP_PROTO(struct xfs_inode *ip, struct xfs_scrub_metadata *sm,
diff --git a/fs/xfs/xfs_ioctl.c b/fs/xfs/xfs_ioctl.c
index 97ccc69a3bf4..fb94301dbc85 100644
--- a/fs/xfs/xfs_ioctl.c
+++ b/fs/xfs/xfs_ioctl.c
@@ -897,6 +897,9 @@ xfs_bulk_ireq_setup(
 	if (hdr->flags & XFS_BULK_IREQ_METADIR)
 		breq->flags |= XFS_IWALK_METADIR;
 
+	if (hdr->flags & XFS_BULK_IREQ_RETAIN_INODES)
+		breq->flags |= XFS_IWALK_RETAIN_INODES;
+
 	return 0;
 }
 
diff --git a/fs/xfs/xfs_itable.c b/fs/xfs/xfs_itable.c
index f92057ad686b..528bf3e59995 100644
--- a/fs/xfs/xfs_itable.c
+++ b/fs/xfs/xfs_itable.c
@@ -74,11 +74,13 @@ xfs_bulkstat_one_int(
 	struct xfs_inode	*ip;		/* incore inode pointer */
 	struct inode		*inode;
 	struct xfs_bulkstat	*buf = bc->buf;
+	int			flags = XFS_IGET_UNTRUSTED;
 	int			error = -EINVAL;
 
-	error = xfs_iget(mp, tp, ino,
-			 (XFS_IGET_DONTCACHE | XFS_IGET_UNTRUSTED),
-			 XFS_ILOCK_SHARED, &ip);
+	if (!(bc->breq->flags & XFS_IBULK_RETAIN_INODES))
+		flags |= XFS_IGET_DONTCACHE;
+
+	error = xfs_iget(mp, tp, ino, flags, XFS_ILOCK_SHARED, &ip);
 	if (error == -ENOENT || error == -EINVAL)
 		goto out_advance;
 	if (error)
diff --git a/fs/xfs/xfs_itable.h b/fs/xfs/xfs_itable.h
index f5a13f69883a..7fc2e40d30bd 100644
--- a/fs/xfs/xfs_itable.h
+++ b/fs/xfs/xfs_itable.h
@@ -22,6 +22,9 @@ struct xfs_ibulk {
 /* Signal that we can return metadata directories. */
 #define XFS_IBULK_METADIR	(XFS_IWALK_METADIR)
 
+/* Don't drop inodes. */
+#define XFS_IBULK_RETAIN_INODES	(XFS_IWALK_RETAIN_INODES)
+
 /*
  * Advance the user buffer pointer by one record of the given size.  If the
  * buffer is now full, return the appropriate error code.
diff --git a/fs/xfs/xfs_iwalk.h b/fs/xfs/xfs_iwalk.h
index d7a082e45cbf..bdec747ff834 100644
--- a/fs/xfs/xfs_iwalk.h
+++ b/fs/xfs/xfs_iwalk.h
@@ -31,8 +31,12 @@ int xfs_iwalk_threaded(struct xfs_mount *mp, xfs_ino_t startino,
 /* Signal that we can return metadata directories. */
 #define XFS_IWALK_METADIR	(0x2)
 
+/* Don't drop inodes. */
+#define XFS_IWALK_RETAIN_INODES	(0x4)
+
 #define XFS_IWALK_FLAGS_ALL	(XFS_IWALK_SAME_AG | \
-				 XFS_IWALK_METADIR)
+				 XFS_IWALK_METADIR | \
+				 XFS_IWALK_RETAIN_INODES)
 
 /* Walk all inode btree records in the filesystem starting from @startino. */
 typedef int (*xfs_inobt_walk_fn)(struct xfs_mount *mp, struct xfs_trans *tp,
-- 
cgit v1.2.3