[PATCH v3 0/2] erofs: support metadata compression

[PATCH v3 0/2] erofs: support metadata compression

[Gao Xiang]

Hi,

This patchset implements metadata compression since many users are
interested in smaller image sizes (even at the cost of some I/O
latency).

In short, it uses a special "metabox" inode to gather all inode
metadata and then compresses it.  Since EROFS supports multiple
algorithms, you can select a faster algorithm (e.g., LZ4) from
the one used for data compression (e.g., LZMA).

Also see the detailed commit messages for more details.

Here are some preliminary numbers:

Command line: -zlzma,6 -Efragments,ztailpacking -C1048576

  ______________________________________________________________
 |         |_______ Vanilla _______|___ 2554769408 (2437MiB) ___|
 | Fedora  |_______ `-m4096` ______|___ 2524504064 (2408MiB) ___|
 |_________|_ `-m4096` (lz4hc,12) _|___ 2527326208 (2411MiB) ___|
 |         |_______ Vanilla _______|___  378634240 ( 362MiB) ___|
 |  AOSP   |_______ `-m4096` ______|___  377856000 ( 361MiB) ___|
 |_________|_ `-m4096` (lz4hc,12) _|___  377942016 ( 361MiB) ___|

Thanks,
Gao Xiang

v2: https://lore.kernel.org/r/20250711094004.2488-1-liubo03@inspur.com
Changes since v2:
 - refine the ondisk format and implementation.

Bo Liu (1):
  erofs: implement metadata compression

Gao Xiang (1):
  erofs: add on-disk definition for metadata compression

 fs/erofs/data.c         | 59 +++++++++++++++++++++++++----------------
 fs/erofs/decompressor.c |  2 +-
 fs/erofs/erofs_fs.h     | 15 ++++++++---
 fs/erofs/fileio.c       |  2 +-
 fs/erofs/inode.c        |  5 ++--
 fs/erofs/internal.h     | 19 ++++++++++---
 fs/erofs/super.c        | 22 +++++++++++++--
 fs/erofs/xattr.c        | 20 +++++++++-----
 fs/erofs/zdata.c        |  5 +++-
 fs/erofs/zmap.c         | 16 ++++++-----
 10 files changed, 115 insertions(+), 50 deletions(-)

-- 
2.43.5

[PATCH v3 1/2] erofs: add on-disk definition for metadata compression

[Gao Xiang]

Filesystem metadata has a high degree of redundancy, so it should
compress well in the general case.

Although metadata compression can increase overall I/O latency, many
users care more about minimized image sizes than extreme runtime
performance.  Let's implement metadata compression in response to user
requests [1].

Actually, it's quite simple to implement metadata compression: since
EROFS already supports per-inode compression, we can simply treat a
special inode (called `the metabox inode`) as a container for compressed
inode metadata.  Since EROFS supports multiple algorithms, users can
even specify LZ4 for metadata and LZMA for data.

To better support incremental builds, the MSB of NIDs indicates where
the inode metadata is located: if bit 63 is set, the inode itself should
be read from `the metabox inode`.

Optionally, shared xattrs can also be kept in `the metabox inode` if
COMPAT_SHARED_EA_IN_METABOX is set.

[1] https://issues.redhat.com/browse/RHEL-75783
Signed-off-by: Gao Xiang <hsiangkao@linux.alibaba.com>
---
 fs/erofs/erofs_fs.h | 13 ++++++++++---
 fs/erofs/internal.h |  2 ++
 2 files changed, 12 insertions(+), 3 deletions(-)

diff --git a/fs/erofs/erofs_fs.h b/fs/erofs/erofs_fs.h
index 767fb4acdc93..0c9047e4a295 100644
--- a/fs/erofs/erofs_fs.h
+++ b/fs/erofs/erofs_fs.h
@@ -15,6 +15,7 @@
 #define EROFS_FEATURE_COMPAT_SB_CHKSUM          0x00000001
 #define EROFS_FEATURE_COMPAT_MTIME              0x00000002
 #define EROFS_FEATURE_COMPAT_XATTR_FILTER	0x00000004
+#define EROFS_FEATURE_COMPAT_SHARED_EA_IN_METABOX	0x00000008
 
 /*
  * Any bits that aren't in EROFS_ALL_FEATURE_INCOMPAT should
@@ -31,6 +32,7 @@
 #define EROFS_FEATURE_INCOMPAT_DEDUPE		0x00000020
 #define EROFS_FEATURE_INCOMPAT_XATTR_PREFIXES	0x00000040
 #define EROFS_FEATURE_INCOMPAT_48BIT		0x00000080
+#define EROFS_FEATURE_INCOMPAT_METABOX		0x00000100
 #define EROFS_ALL_FEATURE_INCOMPAT		\
 	((EROFS_FEATURE_INCOMPAT_48BIT << 1) - 1)
 
@@ -46,7 +48,7 @@ struct erofs_deviceslot {
 };
 #define EROFS_DEVT_SLOT_SIZE	sizeof(struct erofs_deviceslot)
 
-/* erofs on-disk super block (currently 128 bytes) */
+/* erofs on-disk super block (currently 144 bytes at maximum) */
 struct erofs_super_block {
 	__le32 magic;           /* file system magic number */
 	__le32 checksum;        /* crc32c to avoid unexpected on-disk overlap */
@@ -82,7 +84,9 @@ struct erofs_super_block {
 	__u8 reserved[3];
 	__le32 build_time;	/* seconds added to epoch for mkfs time */
 	__le64 rootnid_8b;	/* (48BIT on) nid of root directory */
-	__u8 reserved2[8];
+	__le64 reserved2;
+	__le64 metabox_nid;     /* (METABOX on) nid of the metabox inode */
+	__le64 reserved3;	/* [align to extslot 1] */
 };
 
 /*
@@ -267,6 +271,9 @@ struct erofs_inode_chunk_index {
 	__le32 startblk_lo;	/* starting block number of this chunk */
 };
 
+#define EROFS_DIRENT_NID_METABOX_BIT	63
+#define EROFS_DIRENT_NID_MASK		(BIT(EROFS_DIRENT_NID_METABOX_BIT) - 1)
+
 /* dirent sorts in alphabet order, thus we can do binary search */
 struct erofs_dirent {
 	__le64 nid;     /* node number */
@@ -434,7 +441,7 @@ static inline void erofs_check_ondisk_layout_definitions(void)
 		.h_clusterbits = 1 << Z_EROFS_FRAGMENT_INODE_BIT
 	};
 
-	BUILD_BUG_ON(sizeof(struct erofs_super_block) != 128);
+	BUILD_BUG_ON(sizeof(struct erofs_super_block) != 144);
 	BUILD_BUG_ON(sizeof(struct erofs_inode_compact) != 32);
 	BUILD_BUG_ON(sizeof(struct erofs_inode_extended) != 64);
 	BUILD_BUG_ON(sizeof(struct erofs_xattr_ibody_header) != 12);
diff --git a/fs/erofs/internal.h b/fs/erofs/internal.h
index a7699114f6fe..ad932f670bb6 100644
--- a/fs/erofs/internal.h
+++ b/fs/erofs/internal.h
@@ -227,8 +227,10 @@ EROFS_FEATURE_FUNCS(fragments, incompat, INCOMPAT_FRAGMENTS)
 EROFS_FEATURE_FUNCS(dedupe, incompat, INCOMPAT_DEDUPE)
 EROFS_FEATURE_FUNCS(xattr_prefixes, incompat, INCOMPAT_XATTR_PREFIXES)
 EROFS_FEATURE_FUNCS(48bit, incompat, INCOMPAT_48BIT)
+EROFS_FEATURE_FUNCS(metabox, incompat, INCOMPAT_METABOX)
 EROFS_FEATURE_FUNCS(sb_chksum, compat, COMPAT_SB_CHKSUM)
 EROFS_FEATURE_FUNCS(xattr_filter, compat, COMPAT_XATTR_FILTER)
+EROFS_FEATURE_FUNCS(shared_ea_in_metabox, compat, COMPAT_SHARED_EA_IN_METABOX)
 
 /* atomic flag definitions */
 #define EROFS_I_EA_INITED_BIT	0
-- 
2.43.5

[PATCH v3 2/2] erofs: implement metadata compression

[Gao Xiang]

From: Bo Liu <liubo03@inspur.com>

Thanks to the meta buffer infrastructure, metadata-compressed inodes are
just read from the metabox inode instead of the blockdevice (or backing
file) inode.

The same is true for shared extended attributes.

Co-developed-by: Gao Xiang <hsiangkao@linux.alibaba.com>
Signed-off-by: Bo Liu <liubo03@inspur.com>
Signed-off-by: Gao Xiang <hsiangkao@linux.alibaba.com>
---
 fs/erofs/data.c         | 59 +++++++++++++++++++++++++----------------
 fs/erofs/decompressor.c |  2 +-
 fs/erofs/erofs_fs.h     |  2 +-
 fs/erofs/fileio.c       |  2 +-
 fs/erofs/inode.c        |  5 ++--
 fs/erofs/internal.h     | 17 +++++++++---
 fs/erofs/super.c        | 22 +++++++++++++--
 fs/erofs/xattr.c        | 20 +++++++++-----
 fs/erofs/zdata.c        |  5 +++-
 fs/erofs/zmap.c         | 16 ++++++-----
 10 files changed, 103 insertions(+), 47 deletions(-)

diff --git a/fs/erofs/data.c b/fs/erofs/data.c
index 383c1337e157..f46c47335b9c 100644
--- a/fs/erofs/data.c
+++ b/fs/erofs/data.c
@@ -49,11 +49,18 @@ void *erofs_bread(struct erofs_buf *buf, erofs_off_t offset, bool need_kmap)
 	return buf->base + (offset & ~PAGE_MASK);
 }
 
-void erofs_init_metabuf(struct erofs_buf *buf, struct super_block *sb)
+int erofs_init_metabuf(struct erofs_buf *buf, struct super_block *sb,
+		       bool in_metabox)
 {
 	struct erofs_sb_info *sbi = EROFS_SB(sb);
 
 	buf->file = NULL;
+	if (in_metabox) {
+		if (unlikely(!sbi->metabox_inode))
+			return -EFSCORRUPTED;
+		buf->mapping = sbi->metabox_inode->i_mapping;
+		return 0;
+	}
 	buf->off = sbi->dif0.fsoff;
 	if (erofs_is_fileio_mode(sbi)) {
 		buf->file = sbi->dif0.file;	/* some fs like FUSE needs it */
@@ -62,12 +69,17 @@ void erofs_init_metabuf(struct erofs_buf *buf, struct super_block *sb)
 		buf->mapping = sbi->dif0.fscache->inode->i_mapping;
 	else
 		buf->mapping = sb->s_bdev->bd_mapping;
+	return 0;
 }
 
 void *erofs_read_metabuf(struct erofs_buf *buf, struct super_block *sb,
-			 erofs_off_t offset)
+			 erofs_off_t offset, bool in_metabox)
 {
-	erofs_init_metabuf(buf, sb);
+	int err;
+
+	err = erofs_init_metabuf(buf, sb, in_metabox);
+	if (err)
+		return ERR_PTR(err);
 	return erofs_bread(buf, offset, true);
 }
 
@@ -118,7 +130,7 @@ int erofs_map_blocks(struct inode *inode, struct erofs_map_blocks *map)
 	pos = ALIGN(erofs_iloc(inode) + vi->inode_isize +
 		    vi->xattr_isize, unit) + unit * chunknr;
 
-	idx = erofs_read_metabuf(&buf, sb, pos);
+	idx = erofs_read_metabuf(&buf, sb, pos, erofs_inode_in_metabox(inode));
 	if (IS_ERR(idx)) {
 		err = PTR_ERR(idx);
 		goto out;
@@ -264,7 +276,6 @@ static int erofs_iomap_begin(struct inode *inode, loff_t offset, loff_t length,
 
 	map.m_la = offset;
 	map.m_llen = length;
-
 	ret = erofs_map_blocks(inode, &map);
 	if (ret < 0)
 		return ret;
@@ -273,35 +284,37 @@ static int erofs_iomap_begin(struct inode *inode, loff_t offset, loff_t length,
 	iomap->length = map.m_llen;
 	iomap->flags = 0;
 	iomap->private = NULL;
+	iomap->addr = IOMAP_NULL_ADDR;
 	if (!(map.m_flags & EROFS_MAP_MAPPED)) {
 		iomap->type = IOMAP_HOLE;
-		iomap->addr = IOMAP_NULL_ADDR;
 		return 0;
 	}
 
-	mdev = (struct erofs_map_dev) {
-		.m_deviceid = map.m_deviceid,
-		.m_pa = map.m_pa,
-	};
-	ret = erofs_map_dev(sb, &mdev);
-	if (ret)
-		return ret;
-
-	if (flags & IOMAP_DAX)
-		iomap->dax_dev = mdev.m_dif->dax_dev;
-	else
-		iomap->bdev = mdev.m_bdev;
-
-	iomap->addr = mdev.m_dif->fsoff + mdev.m_pa;
-	if (flags & IOMAP_DAX)
-		iomap->addr += mdev.m_dif->dax_part_off;
+	if (!(map.m_flags & EROFS_MAP_META) || !erofs_inode_in_metabox(inode)) {
+		mdev = (struct erofs_map_dev) {
+			.m_deviceid = map.m_deviceid,
+			.m_pa = map.m_pa,
+		};
+		ret = erofs_map_dev(sb, &mdev);
+		if (ret)
+			return ret;
+
+		if (flags & IOMAP_DAX)
+			iomap->dax_dev = mdev.m_dif->dax_dev;
+		else
+			iomap->bdev = mdev.m_bdev;
+		iomap->addr = mdev.m_dif->fsoff + mdev.m_pa;
+		if (flags & IOMAP_DAX)
+			iomap->addr += mdev.m_dif->dax_part_off;
+	}
 
 	if (map.m_flags & EROFS_MAP_META) {
 		void *ptr;
 		struct erofs_buf buf = __EROFS_BUF_INITIALIZER;
 
 		iomap->type = IOMAP_INLINE;
-		ptr = erofs_read_metabuf(&buf, sb, mdev.m_pa);
+		ptr = erofs_read_metabuf(&buf, sb, map.m_pa,
+					 erofs_inode_in_metabox(inode));
 		if (IS_ERR(ptr))
 			return PTR_ERR(ptr);
 		iomap->inline_data = ptr;
diff --git a/fs/erofs/decompressor.c b/fs/erofs/decompressor.c
index 358061d7b660..354762c9723f 100644
--- a/fs/erofs/decompressor.c
+++ b/fs/erofs/decompressor.c
@@ -467,7 +467,7 @@ int z_erofs_parse_cfgs(struct super_block *sb, struct erofs_super_block *dsb)
 		return -EOPNOTSUPP;
 	}
 
-	erofs_init_metabuf(&buf, sb);
+	(void)erofs_init_metabuf(&buf, sb, false);
 	offset = EROFS_SUPER_OFFSET + sbi->sb_size;
 	alg = 0;
 	for (algs = sbi->available_compr_algs; algs; algs >>= 1, ++alg) {
diff --git a/fs/erofs/erofs_fs.h b/fs/erofs/erofs_fs.h
index 0c9047e4a295..a61831a82a73 100644
--- a/fs/erofs/erofs_fs.h
+++ b/fs/erofs/erofs_fs.h
@@ -34,7 +34,7 @@
 #define EROFS_FEATURE_INCOMPAT_48BIT		0x00000080
 #define EROFS_FEATURE_INCOMPAT_METABOX		0x00000100
 #define EROFS_ALL_FEATURE_INCOMPAT		\
-	((EROFS_FEATURE_INCOMPAT_48BIT << 1) - 1)
+	((EROFS_FEATURE_INCOMPAT_METABOX << 1) - 1)
 
 #define EROFS_SB_EXTSLOT_SIZE	16
 
diff --git a/fs/erofs/fileio.c b/fs/erofs/fileio.c
index 3ee082476c8c..b7b3432a9882 100644
--- a/fs/erofs/fileio.c
+++ b/fs/erofs/fileio.c
@@ -115,7 +115,7 @@ static int erofs_fileio_scan_folio(struct erofs_fileio *io, struct folio *folio)
 			void *src;
 
 			src = erofs_read_metabuf(&buf, inode->i_sb,
-						 map->m_pa + ofs);
+				map->m_pa + ofs, erofs_inode_in_metabox(inode));
 			if (IS_ERR(src)) {
 				err = PTR_ERR(src);
 				break;
diff --git a/fs/erofs/inode.c b/fs/erofs/inode.c
index 47215c5e3385..045ccca6ab30 100644
--- a/fs/erofs/inode.c
+++ b/fs/erofs/inode.c
@@ -29,6 +29,7 @@ static int erofs_read_inode(struct inode *inode)
 	struct super_block *sb = inode->i_sb;
 	erofs_blk_t blkaddr = erofs_blknr(sb, erofs_iloc(inode));
 	unsigned int ofs = erofs_blkoff(sb, erofs_iloc(inode));
+	bool in_mbox = erofs_inode_in_metabox(inode);
 	struct erofs_buf buf = __EROFS_BUF_INITIALIZER;
 	struct erofs_sb_info *sbi = EROFS_SB(sb);
 	erofs_blk_t addrmask = BIT_ULL(48) - 1;
@@ -39,7 +40,7 @@ static int erofs_read_inode(struct inode *inode)
 	void *ptr;
 	int err = 0;
 
-	ptr = erofs_read_metabuf(&buf, sb, erofs_pos(sb, blkaddr));
+	ptr = erofs_read_metabuf(&buf, sb, erofs_pos(sb, blkaddr), in_mbox);
 	if (IS_ERR(ptr)) {
 		err = PTR_ERR(ptr);
 		erofs_err(sb, "failed to read inode meta block (nid: %llu): %d",
@@ -78,7 +79,7 @@ static int erofs_read_inode(struct inode *inode)
 
 			memcpy(&copied, dic, gotten);
 			ptr = erofs_read_metabuf(&buf, sb,
-					erofs_pos(sb, blkaddr + 1));
+					erofs_pos(sb, blkaddr + 1), in_mbox);
 			if (IS_ERR(ptr)) {
 				err = PTR_ERR(ptr);
 				erofs_err(sb, "failed to read inode payload block (nid: %llu): %d",
diff --git a/fs/erofs/internal.h b/fs/erofs/internal.h
index ad932f670bb6..a0e1b0b06d33 100644
--- a/fs/erofs/internal.h
+++ b/fs/erofs/internal.h
@@ -125,6 +125,7 @@ struct erofs_sb_info {
 	struct erofs_sb_lz4_info lz4;
 #endif	/* CONFIG_EROFS_FS_ZIP */
 	struct inode *packed_inode;
+	struct inode *metabox_inode;
 	struct erofs_dev_context *devs;
 	u64 total_blocks;
 
@@ -148,6 +149,7 @@ struct erofs_sb_info {
 	/* what we really care is nid, rather than ino.. */
 	erofs_nid_t root_nid;
 	erofs_nid_t packed_nid;
+	erofs_nid_t metabox_nid;
 	/* used for statfs, f_files - f_favail */
 	u64 inos;
 
@@ -281,12 +283,20 @@ struct erofs_inode {
 
 #define EROFS_I(ptr)	container_of(ptr, struct erofs_inode, vfs_inode)
 
+static inline bool erofs_inode_in_metabox(struct inode *inode)
+{
+	return EROFS_I(inode)->nid & BIT(EROFS_DIRENT_NID_METABOX_BIT);
+}
+
 static inline erofs_off_t erofs_iloc(struct inode *inode)
 {
 	struct erofs_sb_info *sbi = EROFS_I_SB(inode);
+	erofs_nid_t nid_lo = EROFS_I(inode)->nid & EROFS_DIRENT_NID_MASK;
 
+	if (erofs_inode_in_metabox(inode))
+		return nid_lo << sbi->islotbits;
 	return erofs_pos(inode->i_sb, sbi->meta_blkaddr) +
-		(EROFS_I(inode)->nid << sbi->islotbits);
+		(nid_lo << sbi->islotbits);
 }
 
 static inline unsigned int erofs_inode_version(unsigned int ifmt)
@@ -385,9 +395,10 @@ void *erofs_read_metadata(struct super_block *sb, struct erofs_buf *buf,
 void erofs_unmap_metabuf(struct erofs_buf *buf);
 void erofs_put_metabuf(struct erofs_buf *buf);
 void *erofs_bread(struct erofs_buf *buf, erofs_off_t offset, bool need_kmap);
-void erofs_init_metabuf(struct erofs_buf *buf, struct super_block *sb);
+int erofs_init_metabuf(struct erofs_buf *buf, struct super_block *sb,
+		       bool in_metabox);
 void *erofs_read_metabuf(struct erofs_buf *buf, struct super_block *sb,
-			 erofs_off_t offset);
+			 erofs_off_t offset, bool in_metabox);
 int erofs_map_dev(struct super_block *sb, struct erofs_map_dev *dev);
 int erofs_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
 		 u64 start, u64 len);
diff --git a/fs/erofs/super.c b/fs/erofs/super.c
index bc27fa3bd678..539551cf59db 100644
--- a/fs/erofs/super.c
+++ b/fs/erofs/super.c
@@ -141,7 +141,7 @@ static int erofs_init_device(struct erofs_buf *buf, struct super_block *sb,
 	struct erofs_deviceslot *dis;
 	struct file *file;
 
-	dis = erofs_read_metabuf(buf, sb, *pos);
+	dis = erofs_read_metabuf(buf, sb, *pos, false);
 	if (IS_ERR(dis))
 		return PTR_ERR(dis);
 
@@ -258,7 +258,7 @@ static int erofs_read_superblock(struct super_block *sb)
 	void *data;
 	int ret;
 
-	data = erofs_read_metabuf(&buf, sb, 0);
+	data = erofs_read_metabuf(&buf, sb, 0, false);
 	if (IS_ERR(data)) {
 		erofs_err(sb, "cannot read erofs superblock");
 		return PTR_ERR(data);
@@ -319,6 +319,14 @@ static int erofs_read_superblock(struct super_block *sb)
 		sbi->root_nid = le16_to_cpu(dsb->rb.rootnid_2b);
 	}
 	sbi->packed_nid = le64_to_cpu(dsb->packed_nid);
+	if (erofs_sb_has_metabox(sbi)) {
+		if (sbi->sb_size <= offsetof(struct erofs_super_block,
+					     metabox_nid))
+			return -EFSCORRUPTED;
+		sbi->metabox_nid = le64_to_cpu(dsb->metabox_nid);
+		if (sbi->metabox_nid & BIT(EROFS_DIRENT_NID_METABOX_BIT))
+			return -EFSCORRUPTED;	/* self-loop detection */
+	}
 	sbi->inos = le64_to_cpu(dsb->inos);
 
 	sbi->epoch = (s64)le64_to_cpu(dsb->epoch);
@@ -335,6 +343,8 @@ static int erofs_read_superblock(struct super_block *sb)
 
 	if (erofs_sb_has_48bit(sbi))
 		erofs_info(sb, "EXPERIMENTAL 48-bit layout support in use. Use at your own risk!");
+	if (erofs_sb_has_metabox(sbi))
+		erofs_info(sb, "EXPERIMENTAL metadata compression support in use. Use at your own risk!");
 	if (erofs_is_fscache_mode(sb))
 		erofs_info(sb, "[deprecated] fscache-based on-demand read feature in use. Use at your own risk!");
 out:
@@ -690,6 +700,12 @@ static int erofs_fc_fill_super(struct super_block *sb, struct fs_context *fc)
 			return PTR_ERR(inode);
 		sbi->packed_inode = inode;
 	}
+	if (erofs_sb_has_metabox(sbi)) {
+		inode = erofs_iget(sb, sbi->metabox_nid);
+		if (IS_ERR(inode))
+			return PTR_ERR(inode);
+		sbi->metabox_inode = inode;
+	}
 
 	inode = erofs_iget(sb, sbi->root_nid);
 	if (IS_ERR(inode))
@@ -845,6 +861,8 @@ static void erofs_drop_internal_inodes(struct erofs_sb_info *sbi)
 {
 	iput(sbi->packed_inode);
 	sbi->packed_inode = NULL;
+	iput(sbi->metabox_inode);
+	sbi->metabox_inode = NULL;
 #ifdef CONFIG_EROFS_FS_ZIP
 	iput(sbi->managed_cache);
 	sbi->managed_cache = NULL;
diff --git a/fs/erofs/xattr.c b/fs/erofs/xattr.c
index 9cf84717a92e..6d2da6ad2be2 100644
--- a/fs/erofs/xattr.c
+++ b/fs/erofs/xattr.c
@@ -77,7 +77,9 @@ static int erofs_init_inode_xattrs(struct inode *inode)
 	}
 
 	it.buf = __EROFS_BUF_INITIALIZER;
-	erofs_init_metabuf(&it.buf, sb);
+	ret = erofs_init_metabuf(&it.buf, sb, erofs_inode_in_metabox(inode));
+	if (ret)
+		goto out_unlock;
 	it.pos = erofs_iloc(inode) + vi->inode_isize;
 
 	/* read in shared xattr array (non-atomic, see kmalloc below) */
@@ -326,6 +328,9 @@ static int erofs_xattr_iter_inline(struct erofs_xattr_iter *it,
 		return -ENOATTR;
 	}
 
+	ret = erofs_init_metabuf(&it->buf, it->sb, erofs_inode_in_metabox(inode));
+	if (ret)
+		return ret;
 	remaining = vi->xattr_isize - xattr_header_sz;
 	it->pos = erofs_iloc(inode) + vi->inode_isize + xattr_header_sz;
 
@@ -362,7 +367,12 @@ static int erofs_xattr_iter_shared(struct erofs_xattr_iter *it,
 	struct super_block *const sb = it->sb;
 	struct erofs_sb_info *sbi = EROFS_SB(sb);
 	unsigned int i;
-	int ret = -ENOATTR;
+	int ret;
+
+	ret = erofs_init_metabuf(&it->buf, sb,
+				 erofs_sb_has_shared_ea_in_metabox(sbi));
+	if (ret)
+		return ret;
 
 	for (i = 0; i < vi->xattr_shared_count; ++i) {
 		it->pos = erofs_pos(sb, sbi->xattr_blkaddr) +
@@ -378,7 +388,7 @@ static int erofs_xattr_iter_shared(struct erofs_xattr_iter *it,
 		if ((getxattr && ret != -ENOATTR) || (!getxattr && ret))
 			break;
 	}
-	return ret;
+	return i ? ret : -ENOATTR;
 }
 
 int erofs_getxattr(struct inode *inode, int index, const char *name,
@@ -413,7 +423,6 @@ int erofs_getxattr(struct inode *inode, int index, const char *name,
 
 	it.sb = inode->i_sb;
 	it.buf = __EROFS_BUF_INITIALIZER;
-	erofs_init_metabuf(&it.buf, it.sb);
 	it.buffer = buffer;
 	it.buffer_size = buffer_size;
 	it.buffer_ofs = 0;
@@ -439,7 +448,6 @@ ssize_t erofs_listxattr(struct dentry *dentry, char *buffer, size_t buffer_size)
 
 	it.sb = dentry->d_sb;
 	it.buf = __EROFS_BUF_INITIALIZER;
-	erofs_init_metabuf(&it.buf, it.sb);
 	it.dentry = dentry;
 	it.buffer = buffer;
 	it.buffer_size = buffer_size;
@@ -485,7 +493,7 @@ int erofs_xattr_prefixes_init(struct super_block *sb)
 	if (sbi->packed_inode)
 		buf.mapping = sbi->packed_inode->i_mapping;
 	else
-		erofs_init_metabuf(&buf, sb);
+		(void)erofs_init_metabuf(&buf, sb, false);
 
 	for (i = 0; i < sbi->xattr_prefix_count; i++) {
 		void *ptr = erofs_read_metadata(sb, &buf, &pos, &len);
diff --git a/fs/erofs/zdata.c b/fs/erofs/zdata.c
index 0d1ddd9b15de..792f20888a8f 100644
--- a/fs/erofs/zdata.c
+++ b/fs/erofs/zdata.c
@@ -855,7 +855,10 @@ static int z_erofs_pcluster_begin(struct z_erofs_frontend *fe)
 		/* bind cache first when cached decompression is preferred */
 		z_erofs_bind_cache(fe);
 	} else {
-		erofs_init_metabuf(&map->buf, sb);
+		ret = erofs_init_metabuf(&map->buf, sb,
+					 erofs_inode_in_metabox(fe->inode));
+		if (ret)
+			return ret;
 		ptr = erofs_bread(&map->buf, map->m_pa, false);
 		if (IS_ERR(ptr)) {
 			ret = PTR_ERR(ptr);
diff --git a/fs/erofs/zmap.c b/fs/erofs/zmap.c
index b72a0e3f9362..a93efd95c555 100644
--- a/fs/erofs/zmap.c
+++ b/fs/erofs/zmap.c
@@ -17,7 +17,7 @@ struct z_erofs_maprecorder {
 	u16 delta[2];
 	erofs_blk_t pblk, compressedblks;
 	erofs_off_t nextpackoff;
-	bool partialref;
+	bool partialref, in_mbox;
 };
 
 static int z_erofs_load_full_lcluster(struct z_erofs_maprecorder *m,
@@ -31,7 +31,7 @@ static int z_erofs_load_full_lcluster(struct z_erofs_maprecorder *m,
 	struct z_erofs_lcluster_index *di;
 	unsigned int advise;
 
-	di = erofs_read_metabuf(&m->map->buf, inode->i_sb, pos);
+	di = erofs_read_metabuf(&m->map->buf, inode->i_sb, pos, m->in_mbox);
 	if (IS_ERR(di))
 		return PTR_ERR(di);
 	m->lcn = lcn;
@@ -146,7 +146,7 @@ static int z_erofs_load_compact_lcluster(struct z_erofs_maprecorder *m,
 	else
 		return -EOPNOTSUPP;
 
-	in = erofs_read_metabuf(&m->map->buf, m->inode->i_sb, pos);
+	in = erofs_read_metabuf(&m->map->buf, inode->i_sb, pos, m->in_mbox);
 	if (IS_ERR(in))
 		return PTR_ERR(in);
 
@@ -392,6 +392,7 @@ static int z_erofs_map_blocks_fo(struct inode *inode,
 	struct z_erofs_maprecorder m = {
 		.inode = inode,
 		.map = map,
+		.in_mbox = erofs_inode_in_metabox(inode),
 	};
 	int err = 0;
 	unsigned int endoff, afmt;
@@ -521,6 +522,7 @@ static int z_erofs_map_blocks_ext(struct inode *inode,
 	unsigned int recsz = z_erofs_extent_recsize(vi->z_advise);
 	erofs_off_t pos = round_up(Z_EROFS_MAP_HEADER_END(erofs_iloc(inode) +
 				   vi->inode_isize + vi->xattr_isize), recsz);
+	bool in_mbox = erofs_inode_in_metabox(inode);
 	erofs_off_t lend = inode->i_size;
 	erofs_off_t l, r, mid, pa, la, lstart;
 	struct z_erofs_extent *ext;
@@ -530,7 +532,7 @@ static int z_erofs_map_blocks_ext(struct inode *inode,
 	map->m_flags = 0;
 	if (recsz <= offsetof(struct z_erofs_extent, pstart_hi)) {
 		if (recsz <= offsetof(struct z_erofs_extent, pstart_lo)) {
-			ext = erofs_read_metabuf(&map->buf, sb, pos);
+			ext = erofs_read_metabuf(&map->buf, sb, pos, in_mbox);
 			if (IS_ERR(ext))
 				return PTR_ERR(ext);
 			pa = le64_to_cpu(*(__le64 *)ext);
@@ -543,7 +545,7 @@ static int z_erofs_map_blocks_ext(struct inode *inode,
 		}
 
 		for (; lstart <= map->m_la; lstart += 1 << vi->z_lclusterbits) {
-			ext = erofs_read_metabuf(&map->buf, sb, pos);
+			ext = erofs_read_metabuf(&map->buf, sb, pos, in_mbox);
 			if (IS_ERR(ext))
 				return PTR_ERR(ext);
 			map->m_plen = le32_to_cpu(ext->plen);
@@ -563,7 +565,7 @@ static int z_erofs_map_blocks_ext(struct inode *inode,
 		for (l = 0, r = vi->z_extents; l < r; ) {
 			mid = l + (r - l) / 2;
 			ext = erofs_read_metabuf(&map->buf, sb,
-						 pos + mid * recsz);
+						 pos + mid * recsz, in_mbox);
 			if (IS_ERR(ext))
 				return PTR_ERR(ext);
 
@@ -645,7 +647,7 @@ static int z_erofs_fill_inode(struct inode *inode, struct erofs_map_blocks *map)
 		goto out_unlock;
 
 	pos = ALIGN(erofs_iloc(inode) + vi->inode_isize + vi->xattr_isize, 8);
-	h = erofs_read_metabuf(&map->buf, sb, pos);
+	h = erofs_read_metabuf(&map->buf, sb, pos, erofs_inode_in_metabox(inode));
 	if (IS_ERR(h)) {
 		err = PTR_ERR(h);
 		goto out_unlock;
-- 
2.43.5

[PATCH v4 2/2] erofs: implement metadata compression

[Gao Xiang]

From: Bo Liu <liubo03@inspur.com>

Thanks to the meta buffer infrastructure, metadata-compressed inodes are
just read from the metabox inode instead of the blockdevice (or backing
file) inode.

The same is true for shared extended attributes.

Co-developed-by: Gao Xiang <hsiangkao@linux.alibaba.com>
Signed-off-by: Bo Liu <liubo03@inspur.com>
Signed-off-by: Gao Xiang <hsiangkao@linux.alibaba.com>
---
changes since v3:
 - add the missing sysfs part;

 Documentation/ABI/testing/sysfs-fs-erofs |  2 +-
 fs/erofs/data.c                          | 59 +++++++++++++++---------
 fs/erofs/decompressor.c                  |  2 +-
 fs/erofs/erofs_fs.h                      |  2 +-
 fs/erofs/fileio.c                        |  2 +-
 fs/erofs/inode.c                         |  5 +-
 fs/erofs/internal.h                      | 17 +++++--
 fs/erofs/super.c                         | 22 ++++++++-
 fs/erofs/sysfs.c                         |  2 +
 fs/erofs/xattr.c                         | 20 +++++---
 fs/erofs/zdata.c                         |  5 +-
 fs/erofs/zmap.c                          | 16 ++++---
 12 files changed, 106 insertions(+), 48 deletions(-)

diff --git a/Documentation/ABI/testing/sysfs-fs-erofs b/Documentation/ABI/testing/sysfs-fs-erofs
index bf3b6299c15e..ad6d1a3ccd4e 100644
--- a/Documentation/ABI/testing/sysfs-fs-erofs
+++ b/Documentation/ABI/testing/sysfs-fs-erofs
@@ -5,7 +5,7 @@ Description:	Shows all enabled kernel features.
 		Supported features:
 		zero_padding, compr_cfgs, big_pcluster, chunked_file,
 		device_table, compr_head2, sb_chksum, ztailpacking,
-		dedupe, fragments.
+		dedupe, fragments, 48bit, metabox.
 
 What:		/sys/fs/erofs/<disk>/sync_decompress
 Date:		November 2021
diff --git a/fs/erofs/data.c b/fs/erofs/data.c
index 383c1337e157..f46c47335b9c 100644
--- a/fs/erofs/data.c
+++ b/fs/erofs/data.c
@@ -49,11 +49,18 @@ void *erofs_bread(struct erofs_buf *buf, erofs_off_t offset, bool need_kmap)
 	return buf->base + (offset & ~PAGE_MASK);
 }
 
-void erofs_init_metabuf(struct erofs_buf *buf, struct super_block *sb)
+int erofs_init_metabuf(struct erofs_buf *buf, struct super_block *sb,
+		       bool in_metabox)
 {
 	struct erofs_sb_info *sbi = EROFS_SB(sb);
 
 	buf->file = NULL;
+	if (in_metabox) {
+		if (unlikely(!sbi->metabox_inode))
+			return -EFSCORRUPTED;
+		buf->mapping = sbi->metabox_inode->i_mapping;
+		return 0;
+	}
 	buf->off = sbi->dif0.fsoff;
 	if (erofs_is_fileio_mode(sbi)) {
 		buf->file = sbi->dif0.file;	/* some fs like FUSE needs it */
@@ -62,12 +69,17 @@ void erofs_init_metabuf(struct erofs_buf *buf, struct super_block *sb)
 		buf->mapping = sbi->dif0.fscache->inode->i_mapping;
 	else
 		buf->mapping = sb->s_bdev->bd_mapping;
+	return 0;
 }
 
 void *erofs_read_metabuf(struct erofs_buf *buf, struct super_block *sb,
-			 erofs_off_t offset)
+			 erofs_off_t offset, bool in_metabox)
 {
-	erofs_init_metabuf(buf, sb);
+	int err;
+
+	err = erofs_init_metabuf(buf, sb, in_metabox);
+	if (err)
+		return ERR_PTR(err);
 	return erofs_bread(buf, offset, true);
 }
 
@@ -118,7 +130,7 @@ int erofs_map_blocks(struct inode *inode, struct erofs_map_blocks *map)
 	pos = ALIGN(erofs_iloc(inode) + vi->inode_isize +
 		    vi->xattr_isize, unit) + unit * chunknr;
 
-	idx = erofs_read_metabuf(&buf, sb, pos);
+	idx = erofs_read_metabuf(&buf, sb, pos, erofs_inode_in_metabox(inode));
 	if (IS_ERR(idx)) {
 		err = PTR_ERR(idx);
 		goto out;
@@ -264,7 +276,6 @@ static int erofs_iomap_begin(struct inode *inode, loff_t offset, loff_t length,
 
 	map.m_la = offset;
 	map.m_llen = length;
-
 	ret = erofs_map_blocks(inode, &map);
 	if (ret < 0)
 		return ret;
@@ -273,35 +284,37 @@ static int erofs_iomap_begin(struct inode *inode, loff_t offset, loff_t length,
 	iomap->length = map.m_llen;
 	iomap->flags = 0;
 	iomap->private = NULL;
+	iomap->addr = IOMAP_NULL_ADDR;
 	if (!(map.m_flags & EROFS_MAP_MAPPED)) {
 		iomap->type = IOMAP_HOLE;
-		iomap->addr = IOMAP_NULL_ADDR;
 		return 0;
 	}
 
-	mdev = (struct erofs_map_dev) {
-		.m_deviceid = map.m_deviceid,
-		.m_pa = map.m_pa,
-	};
-	ret = erofs_map_dev(sb, &mdev);
-	if (ret)
-		return ret;
-
-	if (flags & IOMAP_DAX)
-		iomap->dax_dev = mdev.m_dif->dax_dev;
-	else
-		iomap->bdev = mdev.m_bdev;
-
-	iomap->addr = mdev.m_dif->fsoff + mdev.m_pa;
-	if (flags & IOMAP_DAX)
-		iomap->addr += mdev.m_dif->dax_part_off;
+	if (!(map.m_flags & EROFS_MAP_META) || !erofs_inode_in_metabox(inode)) {
+		mdev = (struct erofs_map_dev) {
+			.m_deviceid = map.m_deviceid,
+			.m_pa = map.m_pa,
+		};
+		ret = erofs_map_dev(sb, &mdev);
+		if (ret)
+			return ret;
+
+		if (flags & IOMAP_DAX)
+			iomap->dax_dev = mdev.m_dif->dax_dev;
+		else
+			iomap->bdev = mdev.m_bdev;
+		iomap->addr = mdev.m_dif->fsoff + mdev.m_pa;
+		if (flags & IOMAP_DAX)
+			iomap->addr += mdev.m_dif->dax_part_off;
+	}
 
 	if (map.m_flags & EROFS_MAP_META) {
 		void *ptr;
 		struct erofs_buf buf = __EROFS_BUF_INITIALIZER;
 
 		iomap->type = IOMAP_INLINE;
-		ptr = erofs_read_metabuf(&buf, sb, mdev.m_pa);
+		ptr = erofs_read_metabuf(&buf, sb, map.m_pa,
+					 erofs_inode_in_metabox(inode));
 		if (IS_ERR(ptr))
 			return PTR_ERR(ptr);
 		iomap->inline_data = ptr;
diff --git a/fs/erofs/decompressor.c b/fs/erofs/decompressor.c
index 358061d7b660..354762c9723f 100644
--- a/fs/erofs/decompressor.c
+++ b/fs/erofs/decompressor.c
@@ -467,7 +467,7 @@ int z_erofs_parse_cfgs(struct super_block *sb, struct erofs_super_block *dsb)
 		return -EOPNOTSUPP;
 	}
 
-	erofs_init_metabuf(&buf, sb);
+	(void)erofs_init_metabuf(&buf, sb, false);
 	offset = EROFS_SUPER_OFFSET + sbi->sb_size;
 	alg = 0;
 	for (algs = sbi->available_compr_algs; algs; algs >>= 1, ++alg) {
diff --git a/fs/erofs/erofs_fs.h b/fs/erofs/erofs_fs.h
index 0c9047e4a295..a61831a82a73 100644
--- a/fs/erofs/erofs_fs.h
+++ b/fs/erofs/erofs_fs.h
@@ -34,7 +34,7 @@
 #define EROFS_FEATURE_INCOMPAT_48BIT		0x00000080
 #define EROFS_FEATURE_INCOMPAT_METABOX		0x00000100
 #define EROFS_ALL_FEATURE_INCOMPAT		\
-	((EROFS_FEATURE_INCOMPAT_48BIT << 1) - 1)
+	((EROFS_FEATURE_INCOMPAT_METABOX << 1) - 1)
 
 #define EROFS_SB_EXTSLOT_SIZE	16
 
diff --git a/fs/erofs/fileio.c b/fs/erofs/fileio.c
index 3ee082476c8c..b7b3432a9882 100644
--- a/fs/erofs/fileio.c
+++ b/fs/erofs/fileio.c
@@ -115,7 +115,7 @@ static int erofs_fileio_scan_folio(struct erofs_fileio *io, struct folio *folio)
 			void *src;
 
 			src = erofs_read_metabuf(&buf, inode->i_sb,
-						 map->m_pa + ofs);
+				map->m_pa + ofs, erofs_inode_in_metabox(inode));
 			if (IS_ERR(src)) {
 				err = PTR_ERR(src);
 				break;
diff --git a/fs/erofs/inode.c b/fs/erofs/inode.c
index 47215c5e3385..045ccca6ab30 100644
--- a/fs/erofs/inode.c
+++ b/fs/erofs/inode.c
@@ -29,6 +29,7 @@ static int erofs_read_inode(struct inode *inode)
 	struct super_block *sb = inode->i_sb;
 	erofs_blk_t blkaddr = erofs_blknr(sb, erofs_iloc(inode));
 	unsigned int ofs = erofs_blkoff(sb, erofs_iloc(inode));
+	bool in_mbox = erofs_inode_in_metabox(inode);
 	struct erofs_buf buf = __EROFS_BUF_INITIALIZER;
 	struct erofs_sb_info *sbi = EROFS_SB(sb);
 	erofs_blk_t addrmask = BIT_ULL(48) - 1;
@@ -39,7 +40,7 @@ static int erofs_read_inode(struct inode *inode)
 	void *ptr;
 	int err = 0;
 
-	ptr = erofs_read_metabuf(&buf, sb, erofs_pos(sb, blkaddr));
+	ptr = erofs_read_metabuf(&buf, sb, erofs_pos(sb, blkaddr), in_mbox);
 	if (IS_ERR(ptr)) {
 		err = PTR_ERR(ptr);
 		erofs_err(sb, "failed to read inode meta block (nid: %llu): %d",
@@ -78,7 +79,7 @@ static int erofs_read_inode(struct inode *inode)
 
 			memcpy(&copied, dic, gotten);
 			ptr = erofs_read_metabuf(&buf, sb,
-					erofs_pos(sb, blkaddr + 1));
+					erofs_pos(sb, blkaddr + 1), in_mbox);
 			if (IS_ERR(ptr)) {
 				err = PTR_ERR(ptr);
 				erofs_err(sb, "failed to read inode payload block (nid: %llu): %d",
diff --git a/fs/erofs/internal.h b/fs/erofs/internal.h
index ad932f670bb6..a0e1b0b06d33 100644
--- a/fs/erofs/internal.h
+++ b/fs/erofs/internal.h
@@ -125,6 +125,7 @@ struct erofs_sb_info {
 	struct erofs_sb_lz4_info lz4;
 #endif	/* CONFIG_EROFS_FS_ZIP */
 	struct inode *packed_inode;
+	struct inode *metabox_inode;
 	struct erofs_dev_context *devs;
 	u64 total_blocks;
 
@@ -148,6 +149,7 @@ struct erofs_sb_info {
 	/* what we really care is nid, rather than ino.. */
 	erofs_nid_t root_nid;
 	erofs_nid_t packed_nid;
+	erofs_nid_t metabox_nid;
 	/* used for statfs, f_files - f_favail */
 	u64 inos;
 
@@ -281,12 +283,20 @@ struct erofs_inode {
 
 #define EROFS_I(ptr)	container_of(ptr, struct erofs_inode, vfs_inode)
 
+static inline bool erofs_inode_in_metabox(struct inode *inode)
+{
+	return EROFS_I(inode)->nid & BIT(EROFS_DIRENT_NID_METABOX_BIT);
+}
+
 static inline erofs_off_t erofs_iloc(struct inode *inode)
 {
 	struct erofs_sb_info *sbi = EROFS_I_SB(inode);
+	erofs_nid_t nid_lo = EROFS_I(inode)->nid & EROFS_DIRENT_NID_MASK;
 
+	if (erofs_inode_in_metabox(inode))
+		return nid_lo << sbi->islotbits;
 	return erofs_pos(inode->i_sb, sbi->meta_blkaddr) +
-		(EROFS_I(inode)->nid << sbi->islotbits);
+		(nid_lo << sbi->islotbits);
 }
 
 static inline unsigned int erofs_inode_version(unsigned int ifmt)
@@ -385,9 +395,10 @@ void *erofs_read_metadata(struct super_block *sb, struct erofs_buf *buf,
 void erofs_unmap_metabuf(struct erofs_buf *buf);
 void erofs_put_metabuf(struct erofs_buf *buf);
 void *erofs_bread(struct erofs_buf *buf, erofs_off_t offset, bool need_kmap);
-void erofs_init_metabuf(struct erofs_buf *buf, struct super_block *sb);
+int erofs_init_metabuf(struct erofs_buf *buf, struct super_block *sb,
+		       bool in_metabox);
 void *erofs_read_metabuf(struct erofs_buf *buf, struct super_block *sb,
-			 erofs_off_t offset);
+			 erofs_off_t offset, bool in_metabox);
 int erofs_map_dev(struct super_block *sb, struct erofs_map_dev *dev);
 int erofs_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
 		 u64 start, u64 len);
diff --git a/fs/erofs/super.c b/fs/erofs/super.c
index bc27fa3bd678..539551cf59db 100644
--- a/fs/erofs/super.c
+++ b/fs/erofs/super.c
@@ -141,7 +141,7 @@ static int erofs_init_device(struct erofs_buf *buf, struct super_block *sb,
 	struct erofs_deviceslot *dis;
 	struct file *file;
 
-	dis = erofs_read_metabuf(buf, sb, *pos);
+	dis = erofs_read_metabuf(buf, sb, *pos, false);
 	if (IS_ERR(dis))
 		return PTR_ERR(dis);
 
@@ -258,7 +258,7 @@ static int erofs_read_superblock(struct super_block *sb)
 	void *data;
 	int ret;
 
-	data = erofs_read_metabuf(&buf, sb, 0);
+	data = erofs_read_metabuf(&buf, sb, 0, false);
 	if (IS_ERR(data)) {
 		erofs_err(sb, "cannot read erofs superblock");
 		return PTR_ERR(data);
@@ -319,6 +319,14 @@ static int erofs_read_superblock(struct super_block *sb)
 		sbi->root_nid = le16_to_cpu(dsb->rb.rootnid_2b);
 	}
 	sbi->packed_nid = le64_to_cpu(dsb->packed_nid);
+	if (erofs_sb_has_metabox(sbi)) {
+		if (sbi->sb_size <= offsetof(struct erofs_super_block,
+					     metabox_nid))
+			return -EFSCORRUPTED;
+		sbi->metabox_nid = le64_to_cpu(dsb->metabox_nid);
+		if (sbi->metabox_nid & BIT(EROFS_DIRENT_NID_METABOX_BIT))
+			return -EFSCORRUPTED;	/* self-loop detection */
+	}
 	sbi->inos = le64_to_cpu(dsb->inos);
 
 	sbi->epoch = (s64)le64_to_cpu(dsb->epoch);
@@ -335,6 +343,8 @@ static int erofs_read_superblock(struct super_block *sb)
 
 	if (erofs_sb_has_48bit(sbi))
 		erofs_info(sb, "EXPERIMENTAL 48-bit layout support in use. Use at your own risk!");
+	if (erofs_sb_has_metabox(sbi))
+		erofs_info(sb, "EXPERIMENTAL metadata compression support in use. Use at your own risk!");
 	if (erofs_is_fscache_mode(sb))
 		erofs_info(sb, "[deprecated] fscache-based on-demand read feature in use. Use at your own risk!");
 out:
@@ -690,6 +700,12 @@ static int erofs_fc_fill_super(struct super_block *sb, struct fs_context *fc)
 			return PTR_ERR(inode);
 		sbi->packed_inode = inode;
 	}
+	if (erofs_sb_has_metabox(sbi)) {
+		inode = erofs_iget(sb, sbi->metabox_nid);
+		if (IS_ERR(inode))
+			return PTR_ERR(inode);
+		sbi->metabox_inode = inode;
+	}
 
 	inode = erofs_iget(sb, sbi->root_nid);
 	if (IS_ERR(inode))
@@ -845,6 +861,8 @@ static void erofs_drop_internal_inodes(struct erofs_sb_info *sbi)
 {
 	iput(sbi->packed_inode);
 	sbi->packed_inode = NULL;
+	iput(sbi->metabox_inode);
+	sbi->metabox_inode = NULL;
 #ifdef CONFIG_EROFS_FS_ZIP
 	iput(sbi->managed_cache);
 	sbi->managed_cache = NULL;
diff --git a/fs/erofs/sysfs.c b/fs/erofs/sysfs.c
index eed8797a193f..a9fe35ba0906 100644
--- a/fs/erofs/sysfs.c
+++ b/fs/erofs/sysfs.c
@@ -95,6 +95,7 @@ EROFS_ATTR_FEATURE(ztailpacking);
 EROFS_ATTR_FEATURE(fragments);
 EROFS_ATTR_FEATURE(dedupe);
 EROFS_ATTR_FEATURE(48bit);
+EROFS_ATTR_FEATURE(metabox);
 
 static struct attribute *erofs_feat_attrs[] = {
 	ATTR_LIST(zero_padding),
@@ -108,6 +109,7 @@ static struct attribute *erofs_feat_attrs[] = {
 	ATTR_LIST(fragments),
 	ATTR_LIST(dedupe),
 	ATTR_LIST(48bit),
+	ATTR_LIST(metabox),
 	NULL,
 };
 ATTRIBUTE_GROUPS(erofs_feat);
diff --git a/fs/erofs/xattr.c b/fs/erofs/xattr.c
index 9cf84717a92e..6d2da6ad2be2 100644
--- a/fs/erofs/xattr.c
+++ b/fs/erofs/xattr.c
@@ -77,7 +77,9 @@ static int erofs_init_inode_xattrs(struct inode *inode)
 	}
 
 	it.buf = __EROFS_BUF_INITIALIZER;
-	erofs_init_metabuf(&it.buf, sb);
+	ret = erofs_init_metabuf(&it.buf, sb, erofs_inode_in_metabox(inode));
+	if (ret)
+		goto out_unlock;
 	it.pos = erofs_iloc(inode) + vi->inode_isize;
 
 	/* read in shared xattr array (non-atomic, see kmalloc below) */
@@ -326,6 +328,9 @@ static int erofs_xattr_iter_inline(struct erofs_xattr_iter *it,
 		return -ENOATTR;
 	}
 
+	ret = erofs_init_metabuf(&it->buf, it->sb, erofs_inode_in_metabox(inode));
+	if (ret)
+		return ret;
 	remaining = vi->xattr_isize - xattr_header_sz;
 	it->pos = erofs_iloc(inode) + vi->inode_isize + xattr_header_sz;
 
@@ -362,7 +367,12 @@ static int erofs_xattr_iter_shared(struct erofs_xattr_iter *it,
 	struct super_block *const sb = it->sb;
 	struct erofs_sb_info *sbi = EROFS_SB(sb);
 	unsigned int i;
-	int ret = -ENOATTR;
+	int ret;
+
+	ret = erofs_init_metabuf(&it->buf, sb,
+				 erofs_sb_has_shared_ea_in_metabox(sbi));
+	if (ret)
+		return ret;
 
 	for (i = 0; i < vi->xattr_shared_count; ++i) {
 		it->pos = erofs_pos(sb, sbi->xattr_blkaddr) +
@@ -378,7 +388,7 @@ static int erofs_xattr_iter_shared(struct erofs_xattr_iter *it,
 		if ((getxattr && ret != -ENOATTR) || (!getxattr && ret))
 			break;
 	}
-	return ret;
+	return i ? ret : -ENOATTR;
 }
 
 int erofs_getxattr(struct inode *inode, int index, const char *name,
@@ -413,7 +423,6 @@ int erofs_getxattr(struct inode *inode, int index, const char *name,
 
 	it.sb = inode->i_sb;
 	it.buf = __EROFS_BUF_INITIALIZER;
-	erofs_init_metabuf(&it.buf, it.sb);
 	it.buffer = buffer;
 	it.buffer_size = buffer_size;
 	it.buffer_ofs = 0;
@@ -439,7 +448,6 @@ ssize_t erofs_listxattr(struct dentry *dentry, char *buffer, size_t buffer_size)
 
 	it.sb = dentry->d_sb;
 	it.buf = __EROFS_BUF_INITIALIZER;
-	erofs_init_metabuf(&it.buf, it.sb);
 	it.dentry = dentry;
 	it.buffer = buffer;
 	it.buffer_size = buffer_size;
@@ -485,7 +493,7 @@ int erofs_xattr_prefixes_init(struct super_block *sb)
 	if (sbi->packed_inode)
 		buf.mapping = sbi->packed_inode->i_mapping;
 	else
-		erofs_init_metabuf(&buf, sb);
+		(void)erofs_init_metabuf(&buf, sb, false);
 
 	for (i = 0; i < sbi->xattr_prefix_count; i++) {
 		void *ptr = erofs_read_metadata(sb, &buf, &pos, &len);
diff --git a/fs/erofs/zdata.c b/fs/erofs/zdata.c
index 0d1ddd9b15de..792f20888a8f 100644
--- a/fs/erofs/zdata.c
+++ b/fs/erofs/zdata.c
@@ -855,7 +855,10 @@ static int z_erofs_pcluster_begin(struct z_erofs_frontend *fe)
 		/* bind cache first when cached decompression is preferred */
 		z_erofs_bind_cache(fe);
 	} else {
-		erofs_init_metabuf(&map->buf, sb);
+		ret = erofs_init_metabuf(&map->buf, sb,
+					 erofs_inode_in_metabox(fe->inode));
+		if (ret)
+			return ret;
 		ptr = erofs_bread(&map->buf, map->m_pa, false);
 		if (IS_ERR(ptr)) {
 			ret = PTR_ERR(ptr);
diff --git a/fs/erofs/zmap.c b/fs/erofs/zmap.c
index b72a0e3f9362..a93efd95c555 100644
--- a/fs/erofs/zmap.c
+++ b/fs/erofs/zmap.c
@@ -17,7 +17,7 @@ struct z_erofs_maprecorder {
 	u16 delta[2];
 	erofs_blk_t pblk, compressedblks;
 	erofs_off_t nextpackoff;
-	bool partialref;
+	bool partialref, in_mbox;
 };
 
 static int z_erofs_load_full_lcluster(struct z_erofs_maprecorder *m,
@@ -31,7 +31,7 @@ static int z_erofs_load_full_lcluster(struct z_erofs_maprecorder *m,
 	struct z_erofs_lcluster_index *di;
 	unsigned int advise;
 
-	di = erofs_read_metabuf(&m->map->buf, inode->i_sb, pos);
+	di = erofs_read_metabuf(&m->map->buf, inode->i_sb, pos, m->in_mbox);
 	if (IS_ERR(di))
 		return PTR_ERR(di);
 	m->lcn = lcn;
@@ -146,7 +146,7 @@ static int z_erofs_load_compact_lcluster(struct z_erofs_maprecorder *m,
 	else
 		return -EOPNOTSUPP;
 
-	in = erofs_read_metabuf(&m->map->buf, m->inode->i_sb, pos);
+	in = erofs_read_metabuf(&m->map->buf, inode->i_sb, pos, m->in_mbox);
 	if (IS_ERR(in))
 		return PTR_ERR(in);
 
@@ -392,6 +392,7 @@ static int z_erofs_map_blocks_fo(struct inode *inode,
 	struct z_erofs_maprecorder m = {
 		.inode = inode,
 		.map = map,
+		.in_mbox = erofs_inode_in_metabox(inode),
 	};
 	int err = 0;
 	unsigned int endoff, afmt;
@@ -521,6 +522,7 @@ static int z_erofs_map_blocks_ext(struct inode *inode,
 	unsigned int recsz = z_erofs_extent_recsize(vi->z_advise);
 	erofs_off_t pos = round_up(Z_EROFS_MAP_HEADER_END(erofs_iloc(inode) +
 				   vi->inode_isize + vi->xattr_isize), recsz);
+	bool in_mbox = erofs_inode_in_metabox(inode);
 	erofs_off_t lend = inode->i_size;
 	erofs_off_t l, r, mid, pa, la, lstart;
 	struct z_erofs_extent *ext;
@@ -530,7 +532,7 @@ static int z_erofs_map_blocks_ext(struct inode *inode,
 	map->m_flags = 0;
 	if (recsz <= offsetof(struct z_erofs_extent, pstart_hi)) {
 		if (recsz <= offsetof(struct z_erofs_extent, pstart_lo)) {
-			ext = erofs_read_metabuf(&map->buf, sb, pos);
+			ext = erofs_read_metabuf(&map->buf, sb, pos, in_mbox);
 			if (IS_ERR(ext))
 				return PTR_ERR(ext);
 			pa = le64_to_cpu(*(__le64 *)ext);
@@ -543,7 +545,7 @@ static int z_erofs_map_blocks_ext(struct inode *inode,
 		}
 
 		for (; lstart <= map->m_la; lstart += 1 << vi->z_lclusterbits) {
-			ext = erofs_read_metabuf(&map->buf, sb, pos);
+			ext = erofs_read_metabuf(&map->buf, sb, pos, in_mbox);
 			if (IS_ERR(ext))
 				return PTR_ERR(ext);
 			map->m_plen = le32_to_cpu(ext->plen);
@@ -563,7 +565,7 @@ static int z_erofs_map_blocks_ext(struct inode *inode,
 		for (l = 0, r = vi->z_extents; l < r; ) {
 			mid = l + (r - l) / 2;
 			ext = erofs_read_metabuf(&map->buf, sb,
-						 pos + mid * recsz);
+						 pos + mid * recsz, in_mbox);
 			if (IS_ERR(ext))
 				return PTR_ERR(ext);
 
@@ -645,7 +647,7 @@ static int z_erofs_fill_inode(struct inode *inode, struct erofs_map_blocks *map)
 		goto out_unlock;
 
 	pos = ALIGN(erofs_iloc(inode) + vi->inode_isize + vi->xattr_isize, 8);
-	h = erofs_read_metabuf(&map->buf, sb, pos);
+	h = erofs_read_metabuf(&map->buf, sb, pos, erofs_inode_in_metabox(inode));
 	if (IS_ERR(h)) {
 		err = PTR_ERR(h);
 		goto out_unlock;
-- 
2.43.5

[PATCH v5 0/3] erofs: support metadata compression

[Gao Xiang]

Hi,

This patchset implements metadata compression since many users are
interested in smaller image sizes (even at the cost of some I/O
latency).

In short, it uses a special "metabox" inode to gather all inode
metadata and then compresses it.  Since EROFS supports multiple
algorithms, you can select a faster algorithm (e.g., LZ4) from
the one used for data compression (e.g., LZMA).

Also see the detailed commit messages for more details.

Here are some preliminary numbers:

Command line: -zlzma,6 -Efragments,ztailpacking -C1048576

  ______________________________________________________________
 |         |_______ Vanilla _______|___ 2554769408 (2437MiB) ___|
 | Fedora  |_______ `-m4096` ______|___ 2524504064 (2408MiB) ___|
 |_________|_ `-m4096` (lz4hc,12) _|___ 2527326208 (2411MiB) ___|
 |         |_______ Vanilla _______|____ 378634240 ( 362MiB) ___|
 |  AOSP   |_______ `-m4096` ______|____ 377856000 ( 361MiB) ___|
 |_________|_ `-m4096` (lz4hc,12) _|____ 377942016 ( 361MiB) ___|
 |         |_______ Vanilla _______|______ 4837376 (4724KiB) ___|
 | OpenWRT |_______ `-m4096` ______|______ 4734976 (4624KiB) ___|
 |_________|_ `-m4096` (lz4hc,12) _|______ 4747264 (4636KiB) ___|

[ Note: directory data is still left uncompressed by `mkfs.erofs` so
        the final image sizes can be further smaller.  Directory data
        is just like regular data, which only needs some userspace work
        so let's address it later. ]

Thanks,
Gao Xiang

v3/4: https://lore.kernel.org/r/20250716173314.308744-1-hsiangkao@linux.alibaba.com 
Changes since v4:
 - address build failure (https://lore.kernel.org/r/202507170548.rvm67YSU-lkp@intel.com);
 - address build failure (https://lore.kernel.org/r/202507170506.Wzz1lR5I-lkp@intel.com).

Bo Liu (1):
  erofs: implement metadata compression

Gao Xiang (1):
  erofs: add on-disk definition for metadata compression

 Documentation/ABI/testing/sysfs-fs-erofs |  2 +-
 fs/erofs/data.c                          | 59 +++++++++++++++---------
 fs/erofs/decompressor.c                  |  2 +-
 fs/erofs/erofs_fs.h                      | 15 ++++--
 fs/erofs/fileio.c                        |  2 +-
 fs/erofs/fscache.c                       |  3 +-
 fs/erofs/inode.c                         |  5 +-
 fs/erofs/internal.h                      | 19 ++++++--
 fs/erofs/super.c                         | 22 ++++++++-
 fs/erofs/sysfs.c                         |  2 +
 fs/erofs/xattr.c                         | 26 +++++++----
 fs/erofs/zdata.c                         |  5 +-
 fs/erofs/zmap.c                          | 16 ++++---
 13 files changed, 123 insertions(+), 55 deletions(-)

-- 
2.43.5

[PATCH v5 1/2] erofs: add on-disk definition for metadata compression

[Gao Xiang]

Filesystem metadata has a high degree of redundancy, so it should
compress well in the general case.

Although metadata compression can increase overall I/O latency, many
users care more about minimized image sizes than extreme runtime
performance.  Let's implement metadata compression in response to user
requests [1].

Actually, it's quite simple to implement metadata compression: since
EROFS already supports per-inode compression, we can simply treat a
special inode (called `the metabox inode`) as a container for compressed
inode metadata.  Since EROFS supports multiple algorithms, users can
even specify LZ4 for metadata and LZMA for data.

To better support incremental builds, the MSB of NIDs indicates where
the inode metadata is located: if bit 63 is set, the inode itself should
be read from `the metabox inode`.

Optionally, shared xattrs can also be kept in `the metabox inode` if
COMPAT_SHARED_EA_IN_METABOX is set.

[1] https://issues.redhat.com/browse/RHEL-75783
Signed-off-by: Gao Xiang <hsiangkao@linux.alibaba.com>
---
 fs/erofs/erofs_fs.h | 13 ++++++++++---
 fs/erofs/internal.h |  2 ++
 2 files changed, 12 insertions(+), 3 deletions(-)

diff --git a/fs/erofs/erofs_fs.h b/fs/erofs/erofs_fs.h
index 767fb4acdc93..7b150ac64742 100644
--- a/fs/erofs/erofs_fs.h
+++ b/fs/erofs/erofs_fs.h
@@ -15,6 +15,7 @@
 #define EROFS_FEATURE_COMPAT_SB_CHKSUM          0x00000001
 #define EROFS_FEATURE_COMPAT_MTIME              0x00000002
 #define EROFS_FEATURE_COMPAT_XATTR_FILTER	0x00000004
+#define EROFS_FEATURE_COMPAT_SHARED_EA_IN_METABOX	0x00000008
 
 /*
  * Any bits that aren't in EROFS_ALL_FEATURE_INCOMPAT should
@@ -31,6 +32,7 @@
 #define EROFS_FEATURE_INCOMPAT_DEDUPE		0x00000020
 #define EROFS_FEATURE_INCOMPAT_XATTR_PREFIXES	0x00000040
 #define EROFS_FEATURE_INCOMPAT_48BIT		0x00000080
+#define EROFS_FEATURE_INCOMPAT_METABOX		0x00000100
 #define EROFS_ALL_FEATURE_INCOMPAT		\
 	((EROFS_FEATURE_INCOMPAT_48BIT << 1) - 1)
 
@@ -46,7 +48,7 @@ struct erofs_deviceslot {
 };
 #define EROFS_DEVT_SLOT_SIZE	sizeof(struct erofs_deviceslot)
 
-/* erofs on-disk super block (currently 128 bytes) */
+/* erofs on-disk super block (currently 144 bytes at maximum) */
 struct erofs_super_block {
 	__le32 magic;           /* file system magic number */
 	__le32 checksum;        /* crc32c to avoid unexpected on-disk overlap */
@@ -82,7 +84,9 @@ struct erofs_super_block {
 	__u8 reserved[3];
 	__le32 build_time;	/* seconds added to epoch for mkfs time */
 	__le64 rootnid_8b;	/* (48BIT on) nid of root directory */
-	__u8 reserved2[8];
+	__le64 reserved2;
+	__le64 metabox_nid;     /* (METABOX on) nid of the metabox inode */
+	__le64 reserved3;	/* [align to extslot 1] */
 };
 
 /*
@@ -267,6 +271,9 @@ struct erofs_inode_chunk_index {
 	__le32 startblk_lo;	/* starting block number of this chunk */
 };
 
+#define EROFS_DIRENT_NID_METABOX_BIT	63
+#define EROFS_DIRENT_NID_MASK	(BIT_ULL(EROFS_DIRENT_NID_METABOX_BIT) - 1)
+
 /* dirent sorts in alphabet order, thus we can do binary search */
 struct erofs_dirent {
 	__le64 nid;     /* node number */
@@ -434,7 +441,7 @@ static inline void erofs_check_ondisk_layout_definitions(void)
 		.h_clusterbits = 1 << Z_EROFS_FRAGMENT_INODE_BIT
 	};
 
-	BUILD_BUG_ON(sizeof(struct erofs_super_block) != 128);
+	BUILD_BUG_ON(sizeof(struct erofs_super_block) != 144);
 	BUILD_BUG_ON(sizeof(struct erofs_inode_compact) != 32);
 	BUILD_BUG_ON(sizeof(struct erofs_inode_extended) != 64);
 	BUILD_BUG_ON(sizeof(struct erofs_xattr_ibody_header) != 12);
diff --git a/fs/erofs/internal.h b/fs/erofs/internal.h
index a7699114f6fe..ad932f670bb6 100644
--- a/fs/erofs/internal.h
+++ b/fs/erofs/internal.h
@@ -227,8 +227,10 @@ EROFS_FEATURE_FUNCS(fragments, incompat, INCOMPAT_FRAGMENTS)
 EROFS_FEATURE_FUNCS(dedupe, incompat, INCOMPAT_DEDUPE)
 EROFS_FEATURE_FUNCS(xattr_prefixes, incompat, INCOMPAT_XATTR_PREFIXES)
 EROFS_FEATURE_FUNCS(48bit, incompat, INCOMPAT_48BIT)
+EROFS_FEATURE_FUNCS(metabox, incompat, INCOMPAT_METABOX)
 EROFS_FEATURE_FUNCS(sb_chksum, compat, COMPAT_SB_CHKSUM)
 EROFS_FEATURE_FUNCS(xattr_filter, compat, COMPAT_XATTR_FILTER)
+EROFS_FEATURE_FUNCS(shared_ea_in_metabox, compat, COMPAT_SHARED_EA_IN_METABOX)
 
 /* atomic flag definitions */
 #define EROFS_I_EA_INITED_BIT	0
-- 
2.43.5

Re: [PATCH v5 1/2] erofs: add on-disk definition for metadata compression

[Chao Yu]

On 7/17/25 15:08, Gao Xiang wrote:
> Filesystem metadata has a high degree of redundancy, so it should
> compress well in the general case.
> 
> Although metadata compression can increase overall I/O latency, many
> users care more about minimized image sizes than extreme runtime
> performance.  Let's implement metadata compression in response to user
> requests [1].
> 
> Actually, it's quite simple to implement metadata compression: since
> EROFS already supports per-inode compression, we can simply treat a
> special inode (called `the metabox inode`) as a container for compressed
> inode metadata.  Since EROFS supports multiple algorithms, users can
> even specify LZ4 for metadata and LZMA for data.
> 
> To better support incremental builds, the MSB of NIDs indicates where
> the inode metadata is located: if bit 63 is set, the inode itself should
> be read from `the metabox inode`.
> 
> Optionally, shared xattrs can also be kept in `the metabox inode` if
> COMPAT_SHARED_EA_IN_METABOX is set.
> 
> [1] https://issues.redhat.com/browse/RHEL-75783
> Signed-off-by: Gao Xiang <hsiangkao@linux.alibaba.com>

Acked-by: Chao Yu <chao@kernel.org>

Thanks,

[PATCH v5 2/2] erofs: implement metadata compression

[Gao Xiang]

From: Bo Liu <liubo03@inspur.com>

Thanks to the meta buffer infrastructure, metadata-compressed inodes are
just read from the metabox inode instead of the blockdevice (or backing
file) inode.

The same is true for shared extended attributes.

Co-developed-by: Gao Xiang <hsiangkao@linux.alibaba.com>
Signed-off-by: Bo Liu <liubo03@inspur.com>
Signed-off-by: Gao Xiang <hsiangkao@linux.alibaba.com>
---
 Documentation/ABI/testing/sysfs-fs-erofs |  2 +-
 fs/erofs/data.c                          | 59 +++++++++++++++---------
 fs/erofs/decompressor.c                  |  2 +-
 fs/erofs/erofs_fs.h                      |  2 +-
 fs/erofs/fileio.c                        |  2 +-
 fs/erofs/fscache.c                       |  3 +-
 fs/erofs/inode.c                         |  5 +-
 fs/erofs/internal.h                      | 17 +++++--
 fs/erofs/super.c                         | 22 ++++++++-
 fs/erofs/sysfs.c                         |  2 +
 fs/erofs/xattr.c                         | 26 +++++++----
 fs/erofs/zdata.c                         |  5 +-
 fs/erofs/zmap.c                          | 16 ++++---
 13 files changed, 111 insertions(+), 52 deletions(-)

diff --git a/Documentation/ABI/testing/sysfs-fs-erofs b/Documentation/ABI/testing/sysfs-fs-erofs
index bf3b6299c15e..ad6d1a3ccd4e 100644
--- a/Documentation/ABI/testing/sysfs-fs-erofs
+++ b/Documentation/ABI/testing/sysfs-fs-erofs
@@ -5,7 +5,7 @@ Description:	Shows all enabled kernel features.
 		Supported features:
 		zero_padding, compr_cfgs, big_pcluster, chunked_file,
 		device_table, compr_head2, sb_chksum, ztailpacking,
-		dedupe, fragments.
+		dedupe, fragments, 48bit, metabox.
 
 What:		/sys/fs/erofs/<disk>/sync_decompress
 Date:		November 2021
diff --git a/fs/erofs/data.c b/fs/erofs/data.c
index 383c1337e157..f46c47335b9c 100644
--- a/fs/erofs/data.c
+++ b/fs/erofs/data.c
@@ -49,11 +49,18 @@ void *erofs_bread(struct erofs_buf *buf, erofs_off_t offset, bool need_kmap)
 	return buf->base + (offset & ~PAGE_MASK);
 }
 
-void erofs_init_metabuf(struct erofs_buf *buf, struct super_block *sb)
+int erofs_init_metabuf(struct erofs_buf *buf, struct super_block *sb,
+		       bool in_metabox)
 {
 	struct erofs_sb_info *sbi = EROFS_SB(sb);
 
 	buf->file = NULL;
+	if (in_metabox) {
+		if (unlikely(!sbi->metabox_inode))
+			return -EFSCORRUPTED;
+		buf->mapping = sbi->metabox_inode->i_mapping;
+		return 0;
+	}
 	buf->off = sbi->dif0.fsoff;
 	if (erofs_is_fileio_mode(sbi)) {
 		buf->file = sbi->dif0.file;	/* some fs like FUSE needs it */
@@ -62,12 +69,17 @@ void erofs_init_metabuf(struct erofs_buf *buf, struct super_block *sb)
 		buf->mapping = sbi->dif0.fscache->inode->i_mapping;
 	else
 		buf->mapping = sb->s_bdev->bd_mapping;
+	return 0;
 }
 
 void *erofs_read_metabuf(struct erofs_buf *buf, struct super_block *sb,
-			 erofs_off_t offset)
+			 erofs_off_t offset, bool in_metabox)
 {
-	erofs_init_metabuf(buf, sb);
+	int err;
+
+	err = erofs_init_metabuf(buf, sb, in_metabox);
+	if (err)
+		return ERR_PTR(err);
 	return erofs_bread(buf, offset, true);
 }
 
@@ -118,7 +130,7 @@ int erofs_map_blocks(struct inode *inode, struct erofs_map_blocks *map)
 	pos = ALIGN(erofs_iloc(inode) + vi->inode_isize +
 		    vi->xattr_isize, unit) + unit * chunknr;
 
-	idx = erofs_read_metabuf(&buf, sb, pos);
+	idx = erofs_read_metabuf(&buf, sb, pos, erofs_inode_in_metabox(inode));
 	if (IS_ERR(idx)) {
 		err = PTR_ERR(idx);
 		goto out;
@@ -264,7 +276,6 @@ static int erofs_iomap_begin(struct inode *inode, loff_t offset, loff_t length,
 
 	map.m_la = offset;
 	map.m_llen = length;
-
 	ret = erofs_map_blocks(inode, &map);
 	if (ret < 0)
 		return ret;
@@ -273,35 +284,37 @@ static int erofs_iomap_begin(struct inode *inode, loff_t offset, loff_t length,
 	iomap->length = map.m_llen;
 	iomap->flags = 0;
 	iomap->private = NULL;
+	iomap->addr = IOMAP_NULL_ADDR;
 	if (!(map.m_flags & EROFS_MAP_MAPPED)) {
 		iomap->type = IOMAP_HOLE;
-		iomap->addr = IOMAP_NULL_ADDR;
 		return 0;
 	}
 
-	mdev = (struct erofs_map_dev) {
-		.m_deviceid = map.m_deviceid,
-		.m_pa = map.m_pa,
-	};
-	ret = erofs_map_dev(sb, &mdev);
-	if (ret)
-		return ret;
-
-	if (flags & IOMAP_DAX)
-		iomap->dax_dev = mdev.m_dif->dax_dev;
-	else
-		iomap->bdev = mdev.m_bdev;
-
-	iomap->addr = mdev.m_dif->fsoff + mdev.m_pa;
-	if (flags & IOMAP_DAX)
-		iomap->addr += mdev.m_dif->dax_part_off;
+	if (!(map.m_flags & EROFS_MAP_META) || !erofs_inode_in_metabox(inode)) {
+		mdev = (struct erofs_map_dev) {
+			.m_deviceid = map.m_deviceid,
+			.m_pa = map.m_pa,
+		};
+		ret = erofs_map_dev(sb, &mdev);
+		if (ret)
+			return ret;
+
+		if (flags & IOMAP_DAX)
+			iomap->dax_dev = mdev.m_dif->dax_dev;
+		else
+			iomap->bdev = mdev.m_bdev;
+		iomap->addr = mdev.m_dif->fsoff + mdev.m_pa;
+		if (flags & IOMAP_DAX)
+			iomap->addr += mdev.m_dif->dax_part_off;
+	}
 
 	if (map.m_flags & EROFS_MAP_META) {
 		void *ptr;
 		struct erofs_buf buf = __EROFS_BUF_INITIALIZER;
 
 		iomap->type = IOMAP_INLINE;
-		ptr = erofs_read_metabuf(&buf, sb, mdev.m_pa);
+		ptr = erofs_read_metabuf(&buf, sb, map.m_pa,
+					 erofs_inode_in_metabox(inode));
 		if (IS_ERR(ptr))
 			return PTR_ERR(ptr);
 		iomap->inline_data = ptr;
diff --git a/fs/erofs/decompressor.c b/fs/erofs/decompressor.c
index 358061d7b660..354762c9723f 100644
--- a/fs/erofs/decompressor.c
+++ b/fs/erofs/decompressor.c
@@ -467,7 +467,7 @@ int z_erofs_parse_cfgs(struct super_block *sb, struct erofs_super_block *dsb)
 		return -EOPNOTSUPP;
 	}
 
-	erofs_init_metabuf(&buf, sb);
+	(void)erofs_init_metabuf(&buf, sb, false);
 	offset = EROFS_SUPER_OFFSET + sbi->sb_size;
 	alg = 0;
 	for (algs = sbi->available_compr_algs; algs; algs >>= 1, ++alg) {
diff --git a/fs/erofs/erofs_fs.h b/fs/erofs/erofs_fs.h
index 7b150ac64742..377ee12b8b96 100644
--- a/fs/erofs/erofs_fs.h
+++ b/fs/erofs/erofs_fs.h
@@ -34,7 +34,7 @@
 #define EROFS_FEATURE_INCOMPAT_48BIT		0x00000080
 #define EROFS_FEATURE_INCOMPAT_METABOX		0x00000100
 #define EROFS_ALL_FEATURE_INCOMPAT		\
-	((EROFS_FEATURE_INCOMPAT_48BIT << 1) - 1)
+	((EROFS_FEATURE_INCOMPAT_METABOX << 1) - 1)
 
 #define EROFS_SB_EXTSLOT_SIZE	16
 
diff --git a/fs/erofs/fileio.c b/fs/erofs/fileio.c
index 3ee082476c8c..b7b3432a9882 100644
--- a/fs/erofs/fileio.c
+++ b/fs/erofs/fileio.c
@@ -115,7 +115,7 @@ static int erofs_fileio_scan_folio(struct erofs_fileio *io, struct folio *folio)
 			void *src;
 
 			src = erofs_read_metabuf(&buf, inode->i_sb,
-						 map->m_pa + ofs);
+				map->m_pa + ofs, erofs_inode_in_metabox(inode));
 			if (IS_ERR(src)) {
 				err = PTR_ERR(src);
 				break;
diff --git a/fs/erofs/fscache.c b/fs/erofs/fscache.c
index 9a8ee646e51d..362acf828279 100644
--- a/fs/erofs/fscache.c
+++ b/fs/erofs/fscache.c
@@ -274,7 +274,8 @@ static int erofs_fscache_data_read_slice(struct erofs_fscache_rq *req)
 		size_t size = map.m_llen;
 		void *src;
 
-		src = erofs_read_metabuf(&buf, sb, map.m_pa);
+		src = erofs_read_metabuf(&buf, sb, map.m_pa,
+					 erofs_inode_in_metabox(inode));
 		if (IS_ERR(src))
 			return PTR_ERR(src);
 
diff --git a/fs/erofs/inode.c b/fs/erofs/inode.c
index 47215c5e3385..045ccca6ab30 100644
--- a/fs/erofs/inode.c
+++ b/fs/erofs/inode.c
@@ -29,6 +29,7 @@ static int erofs_read_inode(struct inode *inode)
 	struct super_block *sb = inode->i_sb;
 	erofs_blk_t blkaddr = erofs_blknr(sb, erofs_iloc(inode));
 	unsigned int ofs = erofs_blkoff(sb, erofs_iloc(inode));
+	bool in_mbox = erofs_inode_in_metabox(inode);
 	struct erofs_buf buf = __EROFS_BUF_INITIALIZER;
 	struct erofs_sb_info *sbi = EROFS_SB(sb);
 	erofs_blk_t addrmask = BIT_ULL(48) - 1;
@@ -39,7 +40,7 @@ static int erofs_read_inode(struct inode *inode)
 	void *ptr;
 	int err = 0;
 
-	ptr = erofs_read_metabuf(&buf, sb, erofs_pos(sb, blkaddr));
+	ptr = erofs_read_metabuf(&buf, sb, erofs_pos(sb, blkaddr), in_mbox);
 	if (IS_ERR(ptr)) {
 		err = PTR_ERR(ptr);
 		erofs_err(sb, "failed to read inode meta block (nid: %llu): %d",
@@ -78,7 +79,7 @@ static int erofs_read_inode(struct inode *inode)
 
 			memcpy(&copied, dic, gotten);
 			ptr = erofs_read_metabuf(&buf, sb,
-					erofs_pos(sb, blkaddr + 1));
+					erofs_pos(sb, blkaddr + 1), in_mbox);
 			if (IS_ERR(ptr)) {
 				err = PTR_ERR(ptr);
 				erofs_err(sb, "failed to read inode payload block (nid: %llu): %d",
diff --git a/fs/erofs/internal.h b/fs/erofs/internal.h
index ad932f670bb6..a0e1b0b06d33 100644
--- a/fs/erofs/internal.h
+++ b/fs/erofs/internal.h
@@ -125,6 +125,7 @@ struct erofs_sb_info {
 	struct erofs_sb_lz4_info lz4;
 #endif	/* CONFIG_EROFS_FS_ZIP */
 	struct inode *packed_inode;
+	struct inode *metabox_inode;
 	struct erofs_dev_context *devs;
 	u64 total_blocks;
 
@@ -148,6 +149,7 @@ struct erofs_sb_info {
 	/* what we really care is nid, rather than ino.. */
 	erofs_nid_t root_nid;
 	erofs_nid_t packed_nid;
+	erofs_nid_t metabox_nid;
 	/* used for statfs, f_files - f_favail */
 	u64 inos;
 
@@ -281,12 +283,20 @@ struct erofs_inode {
 
 #define EROFS_I(ptr)	container_of(ptr, struct erofs_inode, vfs_inode)
 
+static inline bool erofs_inode_in_metabox(struct inode *inode)
+{
+	return EROFS_I(inode)->nid & BIT(EROFS_DIRENT_NID_METABOX_BIT);
+}
+
 static inline erofs_off_t erofs_iloc(struct inode *inode)
 {
 	struct erofs_sb_info *sbi = EROFS_I_SB(inode);
+	erofs_nid_t nid_lo = EROFS_I(inode)->nid & EROFS_DIRENT_NID_MASK;
 
+	if (erofs_inode_in_metabox(inode))
+		return nid_lo << sbi->islotbits;
 	return erofs_pos(inode->i_sb, sbi->meta_blkaddr) +
-		(EROFS_I(inode)->nid << sbi->islotbits);
+		(nid_lo << sbi->islotbits);
 }
 
 static inline unsigned int erofs_inode_version(unsigned int ifmt)
@@ -385,9 +395,10 @@ void *erofs_read_metadata(struct super_block *sb, struct erofs_buf *buf,
 void erofs_unmap_metabuf(struct erofs_buf *buf);
 void erofs_put_metabuf(struct erofs_buf *buf);
 void *erofs_bread(struct erofs_buf *buf, erofs_off_t offset, bool need_kmap);
-void erofs_init_metabuf(struct erofs_buf *buf, struct super_block *sb);
+int erofs_init_metabuf(struct erofs_buf *buf, struct super_block *sb,
+		       bool in_metabox);
 void *erofs_read_metabuf(struct erofs_buf *buf, struct super_block *sb,
-			 erofs_off_t offset);
+			 erofs_off_t offset, bool in_metabox);
 int erofs_map_dev(struct super_block *sb, struct erofs_map_dev *dev);
 int erofs_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
 		 u64 start, u64 len);
diff --git a/fs/erofs/super.c b/fs/erofs/super.c
index bc27fa3bd678..539551cf59db 100644
--- a/fs/erofs/super.c
+++ b/fs/erofs/super.c
@@ -141,7 +141,7 @@ static int erofs_init_device(struct erofs_buf *buf, struct super_block *sb,
 	struct erofs_deviceslot *dis;
 	struct file *file;
 
-	dis = erofs_read_metabuf(buf, sb, *pos);
+	dis = erofs_read_metabuf(buf, sb, *pos, false);
 	if (IS_ERR(dis))
 		return PTR_ERR(dis);
 
@@ -258,7 +258,7 @@ static int erofs_read_superblock(struct super_block *sb)
 	void *data;
 	int ret;
 
-	data = erofs_read_metabuf(&buf, sb, 0);
+	data = erofs_read_metabuf(&buf, sb, 0, false);
 	if (IS_ERR(data)) {
 		erofs_err(sb, "cannot read erofs superblock");
 		return PTR_ERR(data);
@@ -319,6 +319,14 @@ static int erofs_read_superblock(struct super_block *sb)
 		sbi->root_nid = le16_to_cpu(dsb->rb.rootnid_2b);
 	}
 	sbi->packed_nid = le64_to_cpu(dsb->packed_nid);
+	if (erofs_sb_has_metabox(sbi)) {
+		if (sbi->sb_size <= offsetof(struct erofs_super_block,
+					     metabox_nid))
+			return -EFSCORRUPTED;
+		sbi->metabox_nid = le64_to_cpu(dsb->metabox_nid);
+		if (sbi->metabox_nid & BIT(EROFS_DIRENT_NID_METABOX_BIT))
+			return -EFSCORRUPTED;	/* self-loop detection */
+	}
 	sbi->inos = le64_to_cpu(dsb->inos);
 
 	sbi->epoch = (s64)le64_to_cpu(dsb->epoch);
@@ -335,6 +343,8 @@ static int erofs_read_superblock(struct super_block *sb)
 
 	if (erofs_sb_has_48bit(sbi))
 		erofs_info(sb, "EXPERIMENTAL 48-bit layout support in use. Use at your own risk!");
+	if (erofs_sb_has_metabox(sbi))
+		erofs_info(sb, "EXPERIMENTAL metadata compression support in use. Use at your own risk!");
 	if (erofs_is_fscache_mode(sb))
 		erofs_info(sb, "[deprecated] fscache-based on-demand read feature in use. Use at your own risk!");
 out:
@@ -690,6 +700,12 @@ static int erofs_fc_fill_super(struct super_block *sb, struct fs_context *fc)
 			return PTR_ERR(inode);
 		sbi->packed_inode = inode;
 	}
+	if (erofs_sb_has_metabox(sbi)) {
+		inode = erofs_iget(sb, sbi->metabox_nid);
+		if (IS_ERR(inode))
+			return PTR_ERR(inode);
+		sbi->metabox_inode = inode;
+	}
 
 	inode = erofs_iget(sb, sbi->root_nid);
 	if (IS_ERR(inode))
@@ -845,6 +861,8 @@ static void erofs_drop_internal_inodes(struct erofs_sb_info *sbi)
 {
 	iput(sbi->packed_inode);
 	sbi->packed_inode = NULL;
+	iput(sbi->metabox_inode);
+	sbi->metabox_inode = NULL;
 #ifdef CONFIG_EROFS_FS_ZIP
 	iput(sbi->managed_cache);
 	sbi->managed_cache = NULL;
diff --git a/fs/erofs/sysfs.c b/fs/erofs/sysfs.c
index eed8797a193f..a9fe35ba0906 100644
--- a/fs/erofs/sysfs.c
+++ b/fs/erofs/sysfs.c
@@ -95,6 +95,7 @@ EROFS_ATTR_FEATURE(ztailpacking);
 EROFS_ATTR_FEATURE(fragments);
 EROFS_ATTR_FEATURE(dedupe);
 EROFS_ATTR_FEATURE(48bit);
+EROFS_ATTR_FEATURE(metabox);
 
 static struct attribute *erofs_feat_attrs[] = {
 	ATTR_LIST(zero_padding),
@@ -108,6 +109,7 @@ static struct attribute *erofs_feat_attrs[] = {
 	ATTR_LIST(fragments),
 	ATTR_LIST(dedupe),
 	ATTR_LIST(48bit),
+	ATTR_LIST(metabox),
 	NULL,
 };
 ATTRIBUTE_GROUPS(erofs_feat);
diff --git a/fs/erofs/xattr.c b/fs/erofs/xattr.c
index d61110f511e0..eaa9efd766ee 100644
--- a/fs/erofs/xattr.c
+++ b/fs/erofs/xattr.c
@@ -77,7 +77,9 @@ static int erofs_init_inode_xattrs(struct inode *inode)
 	}
 
 	it.buf = __EROFS_BUF_INITIALIZER;
-	erofs_init_metabuf(&it.buf, sb);
+	ret = erofs_init_metabuf(&it.buf, sb, erofs_inode_in_metabox(inode));
+	if (ret)
+		goto out_unlock;
 	it.pos = erofs_iloc(inode) + vi->inode_isize;
 
 	/* read in shared xattr array (non-atomic, see kmalloc below) */
@@ -326,6 +328,9 @@ static int erofs_xattr_iter_inline(struct erofs_xattr_iter *it,
 		return -ENODATA;
 	}
 
+	ret = erofs_init_metabuf(&it->buf, it->sb, erofs_inode_in_metabox(inode));
+	if (ret)
+		return ret;
 	remaining = vi->xattr_isize - xattr_header_sz;
 	it->pos = erofs_iloc(inode) + vi->inode_isize + xattr_header_sz;
 
@@ -361,12 +366,17 @@ static int erofs_xattr_iter_shared(struct erofs_xattr_iter *it,
 	struct erofs_inode *const vi = EROFS_I(inode);
 	struct super_block *const sb = it->sb;
 	struct erofs_sb_info *sbi = EROFS_SB(sb);
-	unsigned int i;
-	int ret = -ENODATA;
+	unsigned int i = 0;
+	int ret;
 
-	for (i = 0; i < vi->xattr_shared_count; ++i) {
+	ret = erofs_init_metabuf(&it->buf, sb,
+				 erofs_sb_has_shared_ea_in_metabox(sbi));
+	if (ret)
+		return ret;
+
+	while (i < vi->xattr_shared_count) {
 		it->pos = erofs_pos(sb, sbi->xattr_blkaddr) +
-				vi->xattr_shared_xattrs[i] * sizeof(__le32);
+				vi->xattr_shared_xattrs[i++] * sizeof(__le32);
 		it->kaddr = erofs_bread(&it->buf, it->pos, true);
 		if (IS_ERR(it->kaddr))
 			return PTR_ERR(it->kaddr);
@@ -378,7 +388,7 @@ static int erofs_xattr_iter_shared(struct erofs_xattr_iter *it,
 		if ((getxattr && ret != -ENODATA) || (!getxattr && ret))
 			break;
 	}
-	return ret;
+	return i ? ret : -ENODATA;
 }
 
 int erofs_getxattr(struct inode *inode, int index, const char *name,
@@ -413,7 +423,6 @@ int erofs_getxattr(struct inode *inode, int index, const char *name,
 
 	it.sb = inode->i_sb;
 	it.buf = __EROFS_BUF_INITIALIZER;
-	erofs_init_metabuf(&it.buf, it.sb);
 	it.buffer = buffer;
 	it.buffer_size = buffer_size;
 	it.buffer_ofs = 0;
@@ -439,7 +448,6 @@ ssize_t erofs_listxattr(struct dentry *dentry, char *buffer, size_t buffer_size)
 
 	it.sb = dentry->d_sb;
 	it.buf = __EROFS_BUF_INITIALIZER;
-	erofs_init_metabuf(&it.buf, it.sb);
 	it.dentry = dentry;
 	it.buffer = buffer;
 	it.buffer_size = buffer_size;
@@ -485,7 +493,7 @@ int erofs_xattr_prefixes_init(struct super_block *sb)
 	if (sbi->packed_inode)
 		buf.mapping = sbi->packed_inode->i_mapping;
 	else
-		erofs_init_metabuf(&buf, sb);
+		(void)erofs_init_metabuf(&buf, sb, false);
 
 	for (i = 0; i < sbi->xattr_prefix_count; i++) {
 		void *ptr = erofs_read_metadata(sb, &buf, &pos, &len);
diff --git a/fs/erofs/zdata.c b/fs/erofs/zdata.c
index 0d1ddd9b15de..792f20888a8f 100644
--- a/fs/erofs/zdata.c
+++ b/fs/erofs/zdata.c
@@ -855,7 +855,10 @@ static int z_erofs_pcluster_begin(struct z_erofs_frontend *fe)
 		/* bind cache first when cached decompression is preferred */
 		z_erofs_bind_cache(fe);
 	} else {
-		erofs_init_metabuf(&map->buf, sb);
+		ret = erofs_init_metabuf(&map->buf, sb,
+					 erofs_inode_in_metabox(fe->inode));
+		if (ret)
+			return ret;
 		ptr = erofs_bread(&map->buf, map->m_pa, false);
 		if (IS_ERR(ptr)) {
 			ret = PTR_ERR(ptr);
diff --git a/fs/erofs/zmap.c b/fs/erofs/zmap.c
index b72a0e3f9362..a93efd95c555 100644
--- a/fs/erofs/zmap.c
+++ b/fs/erofs/zmap.c
@@ -17,7 +17,7 @@ struct z_erofs_maprecorder {
 	u16 delta[2];
 	erofs_blk_t pblk, compressedblks;
 	erofs_off_t nextpackoff;
-	bool partialref;
+	bool partialref, in_mbox;
 };
 
 static int z_erofs_load_full_lcluster(struct z_erofs_maprecorder *m,
@@ -31,7 +31,7 @@ static int z_erofs_load_full_lcluster(struct z_erofs_maprecorder *m,
 	struct z_erofs_lcluster_index *di;
 	unsigned int advise;
 
-	di = erofs_read_metabuf(&m->map->buf, inode->i_sb, pos);
+	di = erofs_read_metabuf(&m->map->buf, inode->i_sb, pos, m->in_mbox);
 	if (IS_ERR(di))
 		return PTR_ERR(di);
 	m->lcn = lcn;
@@ -146,7 +146,7 @@ static int z_erofs_load_compact_lcluster(struct z_erofs_maprecorder *m,
 	else
 		return -EOPNOTSUPP;
 
-	in = erofs_read_metabuf(&m->map->buf, m->inode->i_sb, pos);
+	in = erofs_read_metabuf(&m->map->buf, inode->i_sb, pos, m->in_mbox);
 	if (IS_ERR(in))
 		return PTR_ERR(in);
 
@@ -392,6 +392,7 @@ static int z_erofs_map_blocks_fo(struct inode *inode,
 	struct z_erofs_maprecorder m = {
 		.inode = inode,
 		.map = map,
+		.in_mbox = erofs_inode_in_metabox(inode),
 	};
 	int err = 0;
 	unsigned int endoff, afmt;
@@ -521,6 +522,7 @@ static int z_erofs_map_blocks_ext(struct inode *inode,
 	unsigned int recsz = z_erofs_extent_recsize(vi->z_advise);
 	erofs_off_t pos = round_up(Z_EROFS_MAP_HEADER_END(erofs_iloc(inode) +
 				   vi->inode_isize + vi->xattr_isize), recsz);
+	bool in_mbox = erofs_inode_in_metabox(inode);
 	erofs_off_t lend = inode->i_size;
 	erofs_off_t l, r, mid, pa, la, lstart;
 	struct z_erofs_extent *ext;
@@ -530,7 +532,7 @@ static int z_erofs_map_blocks_ext(struct inode *inode,
 	map->m_flags = 0;
 	if (recsz <= offsetof(struct z_erofs_extent, pstart_hi)) {
 		if (recsz <= offsetof(struct z_erofs_extent, pstart_lo)) {
-			ext = erofs_read_metabuf(&map->buf, sb, pos);
+			ext = erofs_read_metabuf(&map->buf, sb, pos, in_mbox);
 			if (IS_ERR(ext))
 				return PTR_ERR(ext);
 			pa = le64_to_cpu(*(__le64 *)ext);
@@ -543,7 +545,7 @@ static int z_erofs_map_blocks_ext(struct inode *inode,
 		}
 
 		for (; lstart <= map->m_la; lstart += 1 << vi->z_lclusterbits) {
-			ext = erofs_read_metabuf(&map->buf, sb, pos);
+			ext = erofs_read_metabuf(&map->buf, sb, pos, in_mbox);
 			if (IS_ERR(ext))
 				return PTR_ERR(ext);
 			map->m_plen = le32_to_cpu(ext->plen);
@@ -563,7 +565,7 @@ static int z_erofs_map_blocks_ext(struct inode *inode,
 		for (l = 0, r = vi->z_extents; l < r; ) {
 			mid = l + (r - l) / 2;
 			ext = erofs_read_metabuf(&map->buf, sb,
-						 pos + mid * recsz);
+						 pos + mid * recsz, in_mbox);
 			if (IS_ERR(ext))
 				return PTR_ERR(ext);
 
@@ -645,7 +647,7 @@ static int z_erofs_fill_inode(struct inode *inode, struct erofs_map_blocks *map)
 		goto out_unlock;
 
 	pos = ALIGN(erofs_iloc(inode) + vi->inode_isize + vi->xattr_isize, 8);
-	h = erofs_read_metabuf(&map->buf, sb, pos);
+	h = erofs_read_metabuf(&map->buf, sb, pos, erofs_inode_in_metabox(inode));
 	if (IS_ERR(h)) {
 		err = PTR_ERR(h);
 		goto out_unlock;
-- 
2.43.5

[PATCH v6 2/2] erofs: implement metadata compression

[Gao Xiang]

From: Bo Liu <liubo03@inspur.com>

Thanks to the meta buffer infrastructure, metadata-compressed inodes are
just read from the metabox inode instead of the blockdevice (or backing
file) inode.

The same is true for shared extended attributes.

When metadata compression is enabled, inode numbers are divided from
on-disk NIDs because of non-LTS 32-bit application compatibility.

Co-developed-by: Gao Xiang <hsiangkao@linux.alibaba.com>
Signed-off-by: Bo Liu <liubo03@inspur.com>
Signed-off-by: Gao Xiang <hsiangkao@linux.alibaba.com>
---
change since v5:
 - adjust inode numbers for metadata-compressed inodes in order to
   keep compatibility with non-LTS applications. 

 Documentation/ABI/testing/sysfs-fs-erofs |  2 +-
 fs/erofs/data.c                          | 59 +++++++++++++++---------
 fs/erofs/decompressor.c                  |  2 +-
 fs/erofs/dir.c                           |  3 +-
 fs/erofs/erofs_fs.h                      |  2 +-
 fs/erofs/fileio.c                        |  2 +-
 fs/erofs/fscache.c                       |  3 +-
 fs/erofs/inode.c                         | 17 +++----
 fs/erofs/internal.h                      | 34 ++++++++++++--
 fs/erofs/super.c                         | 22 ++++++++-
 fs/erofs/sysfs.c                         |  2 +
 fs/erofs/xattr.c                         | 26 +++++++----
 fs/erofs/zdata.c                         |  5 +-
 fs/erofs/zmap.c                          | 16 ++++---
 14 files changed, 136 insertions(+), 59 deletions(-)

diff --git a/Documentation/ABI/testing/sysfs-fs-erofs b/Documentation/ABI/testing/sysfs-fs-erofs
index bf3b6299c15e..ad6d1a3ccd4e 100644
--- a/Documentation/ABI/testing/sysfs-fs-erofs
+++ b/Documentation/ABI/testing/sysfs-fs-erofs
@@ -5,7 +5,7 @@ Description:	Shows all enabled kernel features.
 		Supported features:
 		zero_padding, compr_cfgs, big_pcluster, chunked_file,
 		device_table, compr_head2, sb_chksum, ztailpacking,
-		dedupe, fragments.
+		dedupe, fragments, 48bit, metabox.
 
 What:		/sys/fs/erofs/<disk>/sync_decompress
 Date:		November 2021
diff --git a/fs/erofs/data.c b/fs/erofs/data.c
index 383c1337e157..f46c47335b9c 100644
--- a/fs/erofs/data.c
+++ b/fs/erofs/data.c
@@ -49,11 +49,18 @@ void *erofs_bread(struct erofs_buf *buf, erofs_off_t offset, bool need_kmap)
 	return buf->base + (offset & ~PAGE_MASK);
 }
 
-void erofs_init_metabuf(struct erofs_buf *buf, struct super_block *sb)
+int erofs_init_metabuf(struct erofs_buf *buf, struct super_block *sb,
+		       bool in_metabox)
 {
 	struct erofs_sb_info *sbi = EROFS_SB(sb);
 
 	buf->file = NULL;
+	if (in_metabox) {
+		if (unlikely(!sbi->metabox_inode))
+			return -EFSCORRUPTED;
+		buf->mapping = sbi->metabox_inode->i_mapping;
+		return 0;
+	}
 	buf->off = sbi->dif0.fsoff;
 	if (erofs_is_fileio_mode(sbi)) {
 		buf->file = sbi->dif0.file;	/* some fs like FUSE needs it */
@@ -62,12 +69,17 @@ void erofs_init_metabuf(struct erofs_buf *buf, struct super_block *sb)
 		buf->mapping = sbi->dif0.fscache->inode->i_mapping;
 	else
 		buf->mapping = sb->s_bdev->bd_mapping;
+	return 0;
 }
 
 void *erofs_read_metabuf(struct erofs_buf *buf, struct super_block *sb,
-			 erofs_off_t offset)
+			 erofs_off_t offset, bool in_metabox)
 {
-	erofs_init_metabuf(buf, sb);
+	int err;
+
+	err = erofs_init_metabuf(buf, sb, in_metabox);
+	if (err)
+		return ERR_PTR(err);
 	return erofs_bread(buf, offset, true);
 }
 
@@ -118,7 +130,7 @@ int erofs_map_blocks(struct inode *inode, struct erofs_map_blocks *map)
 	pos = ALIGN(erofs_iloc(inode) + vi->inode_isize +
 		    vi->xattr_isize, unit) + unit * chunknr;
 
-	idx = erofs_read_metabuf(&buf, sb, pos);
+	idx = erofs_read_metabuf(&buf, sb, pos, erofs_inode_in_metabox(inode));
 	if (IS_ERR(idx)) {
 		err = PTR_ERR(idx);
 		goto out;
@@ -264,7 +276,6 @@ static int erofs_iomap_begin(struct inode *inode, loff_t offset, loff_t length,
 
 	map.m_la = offset;
 	map.m_llen = length;
-
 	ret = erofs_map_blocks(inode, &map);
 	if (ret < 0)
 		return ret;
@@ -273,35 +284,37 @@ static int erofs_iomap_begin(struct inode *inode, loff_t offset, loff_t length,
 	iomap->length = map.m_llen;
 	iomap->flags = 0;
 	iomap->private = NULL;
+	iomap->addr = IOMAP_NULL_ADDR;
 	if (!(map.m_flags & EROFS_MAP_MAPPED)) {
 		iomap->type = IOMAP_HOLE;
-		iomap->addr = IOMAP_NULL_ADDR;
 		return 0;
 	}
 
-	mdev = (struct erofs_map_dev) {
-		.m_deviceid = map.m_deviceid,
-		.m_pa = map.m_pa,
-	};
-	ret = erofs_map_dev(sb, &mdev);
-	if (ret)
-		return ret;
-
-	if (flags & IOMAP_DAX)
-		iomap->dax_dev = mdev.m_dif->dax_dev;
-	else
-		iomap->bdev = mdev.m_bdev;
-
-	iomap->addr = mdev.m_dif->fsoff + mdev.m_pa;
-	if (flags & IOMAP_DAX)
-		iomap->addr += mdev.m_dif->dax_part_off;
+	if (!(map.m_flags & EROFS_MAP_META) || !erofs_inode_in_metabox(inode)) {
+		mdev = (struct erofs_map_dev) {
+			.m_deviceid = map.m_deviceid,
+			.m_pa = map.m_pa,
+		};
+		ret = erofs_map_dev(sb, &mdev);
+		if (ret)
+			return ret;
+
+		if (flags & IOMAP_DAX)
+			iomap->dax_dev = mdev.m_dif->dax_dev;
+		else
+			iomap->bdev = mdev.m_bdev;
+		iomap->addr = mdev.m_dif->fsoff + mdev.m_pa;
+		if (flags & IOMAP_DAX)
+			iomap->addr += mdev.m_dif->dax_part_off;
+	}
 
 	if (map.m_flags & EROFS_MAP_META) {
 		void *ptr;
 		struct erofs_buf buf = __EROFS_BUF_INITIALIZER;
 
 		iomap->type = IOMAP_INLINE;
-		ptr = erofs_read_metabuf(&buf, sb, mdev.m_pa);
+		ptr = erofs_read_metabuf(&buf, sb, map.m_pa,
+					 erofs_inode_in_metabox(inode));
 		if (IS_ERR(ptr))
 			return PTR_ERR(ptr);
 		iomap->inline_data = ptr;
diff --git a/fs/erofs/decompressor.c b/fs/erofs/decompressor.c
index 358061d7b660..354762c9723f 100644
--- a/fs/erofs/decompressor.c
+++ b/fs/erofs/decompressor.c
@@ -467,7 +467,7 @@ int z_erofs_parse_cfgs(struct super_block *sb, struct erofs_super_block *dsb)
 		return -EOPNOTSUPP;
 	}
 
-	erofs_init_metabuf(&buf, sb);
+	(void)erofs_init_metabuf(&buf, sb, false);
 	offset = EROFS_SUPER_OFFSET + sbi->sb_size;
 	alg = 0;
 	for (algs = sbi->available_compr_algs; algs; algs >>= 1, ++alg) {
diff --git a/fs/erofs/dir.c b/fs/erofs/dir.c
index 3e4b38bec0aa..d1c6cd1a45e8 100644
--- a/fs/erofs/dir.c
+++ b/fs/erofs/dir.c
@@ -34,7 +34,8 @@ static int erofs_fill_dentries(struct inode *dir, struct dir_context *ctx,
 		}
 
 		if (!dir_emit(ctx, de_name, de_namelen,
-			      le64_to_cpu(de->nid), d_type))
+			      erofs_nid_to_ino64(EROFS_SB(dir->i_sb),
+						 le64_to_cpu(de->nid)), d_type))
 			return 1;
 		++de;
 		ctx->pos += sizeof(struct erofs_dirent);
diff --git a/fs/erofs/erofs_fs.h b/fs/erofs/erofs_fs.h
index 7b150ac64742..377ee12b8b96 100644
--- a/fs/erofs/erofs_fs.h
+++ b/fs/erofs/erofs_fs.h
@@ -34,7 +34,7 @@
 #define EROFS_FEATURE_INCOMPAT_48BIT		0x00000080
 #define EROFS_FEATURE_INCOMPAT_METABOX		0x00000100
 #define EROFS_ALL_FEATURE_INCOMPAT		\
-	((EROFS_FEATURE_INCOMPAT_48BIT << 1) - 1)
+	((EROFS_FEATURE_INCOMPAT_METABOX << 1) - 1)
 
 #define EROFS_SB_EXTSLOT_SIZE	16
 
diff --git a/fs/erofs/fileio.c b/fs/erofs/fileio.c
index 3ee082476c8c..b7b3432a9882 100644
--- a/fs/erofs/fileio.c
+++ b/fs/erofs/fileio.c
@@ -115,7 +115,7 @@ static int erofs_fileio_scan_folio(struct erofs_fileio *io, struct folio *folio)
 			void *src;
 
 			src = erofs_read_metabuf(&buf, inode->i_sb,
-						 map->m_pa + ofs);
+				map->m_pa + ofs, erofs_inode_in_metabox(inode));
 			if (IS_ERR(src)) {
 				err = PTR_ERR(src);
 				break;
diff --git a/fs/erofs/fscache.c b/fs/erofs/fscache.c
index 9a8ee646e51d..362acf828279 100644
--- a/fs/erofs/fscache.c
+++ b/fs/erofs/fscache.c
@@ -274,7 +274,8 @@ static int erofs_fscache_data_read_slice(struct erofs_fscache_rq *req)
 		size_t size = map.m_llen;
 		void *src;
 
-		src = erofs_read_metabuf(&buf, sb, map.m_pa);
+		src = erofs_read_metabuf(&buf, sb, map.m_pa,
+					 erofs_inode_in_metabox(inode));
 		if (IS_ERR(src))
 			return PTR_ERR(src);
 
diff --git a/fs/erofs/inode.c b/fs/erofs/inode.c
index 47215c5e3385..9a2f59721522 100644
--- a/fs/erofs/inode.c
+++ b/fs/erofs/inode.c
@@ -29,6 +29,7 @@ static int erofs_read_inode(struct inode *inode)
 	struct super_block *sb = inode->i_sb;
 	erofs_blk_t blkaddr = erofs_blknr(sb, erofs_iloc(inode));
 	unsigned int ofs = erofs_blkoff(sb, erofs_iloc(inode));
+	bool in_mbox = erofs_inode_in_metabox(inode);
 	struct erofs_buf buf = __EROFS_BUF_INITIALIZER;
 	struct erofs_sb_info *sbi = EROFS_SB(sb);
 	erofs_blk_t addrmask = BIT_ULL(48) - 1;
@@ -39,7 +40,7 @@ static int erofs_read_inode(struct inode *inode)
 	void *ptr;
 	int err = 0;
 
-	ptr = erofs_read_metabuf(&buf, sb, erofs_pos(sb, blkaddr));
+	ptr = erofs_read_metabuf(&buf, sb, erofs_pos(sb, blkaddr), in_mbox);
 	if (IS_ERR(ptr)) {
 		err = PTR_ERR(ptr);
 		erofs_err(sb, "failed to read inode meta block (nid: %llu): %d",
@@ -78,7 +79,7 @@ static int erofs_read_inode(struct inode *inode)
 
 			memcpy(&copied, dic, gotten);
 			ptr = erofs_read_metabuf(&buf, sb,
-					erofs_pos(sb, blkaddr + 1));
+					erofs_pos(sb, blkaddr + 1), in_mbox);
 			if (IS_ERR(ptr)) {
 				err = PTR_ERR(ptr);
 				erofs_err(sb, "failed to read inode payload block (nid: %llu): %d",
@@ -264,13 +265,13 @@ static int erofs_fill_inode(struct inode *inode)
  * ino_t is 32-bits on 32-bit arch. We have to squash the 64-bit value down
  * so that it will fit.
  */
-static ino_t erofs_squash_ino(erofs_nid_t nid)
+static ino_t erofs_squash_ino(struct super_block *sb, erofs_nid_t nid)
 {
-	ino_t ino = (ino_t)nid;
+	u64 ino64 = erofs_nid_to_ino64(EROFS_SB(sb), nid);
 
 	if (sizeof(ino_t) < sizeof(erofs_nid_t))
-		ino ^= nid >> (sizeof(erofs_nid_t) - sizeof(ino_t)) * 8;
-	return ino;
+		ino64 ^= ino64 >> (sizeof(erofs_nid_t) - sizeof(ino_t)) * 8;
+	return (ino_t)ino64;
 }
 
 static int erofs_iget5_eq(struct inode *inode, void *opaque)
@@ -282,7 +283,7 @@ static int erofs_iget5_set(struct inode *inode, void *opaque)
 {
 	const erofs_nid_t nid = *(erofs_nid_t *)opaque;
 
-	inode->i_ino = erofs_squash_ino(nid);
+	inode->i_ino = erofs_squash_ino(inode->i_sb, nid);
 	EROFS_I(inode)->nid = nid;
 	return 0;
 }
@@ -291,7 +292,7 @@ struct inode *erofs_iget(struct super_block *sb, erofs_nid_t nid)
 {
 	struct inode *inode;
 
-	inode = iget5_locked(sb, erofs_squash_ino(nid), erofs_iget5_eq,
+	inode = iget5_locked(sb, erofs_squash_ino(sb, nid), erofs_iget5_eq,
 			     erofs_iget5_set, &nid);
 	if (!inode)
 		return ERR_PTR(-ENOMEM);
diff --git a/fs/erofs/internal.h b/fs/erofs/internal.h
index ad932f670bb6..c24100ef7bc5 100644
--- a/fs/erofs/internal.h
+++ b/fs/erofs/internal.h
@@ -125,6 +125,7 @@ struct erofs_sb_info {
 	struct erofs_sb_lz4_info lz4;
 #endif	/* CONFIG_EROFS_FS_ZIP */
 	struct inode *packed_inode;
+	struct inode *metabox_inode;
 	struct erofs_dev_context *devs;
 	u64 total_blocks;
 
@@ -148,6 +149,7 @@ struct erofs_sb_info {
 	/* what we really care is nid, rather than ino.. */
 	erofs_nid_t root_nid;
 	erofs_nid_t packed_nid;
+	erofs_nid_t metabox_nid;
 	/* used for statfs, f_files - f_favail */
 	u64 inos;
 
@@ -232,6 +234,23 @@ EROFS_FEATURE_FUNCS(sb_chksum, compat, COMPAT_SB_CHKSUM)
 EROFS_FEATURE_FUNCS(xattr_filter, compat, COMPAT_XATTR_FILTER)
 EROFS_FEATURE_FUNCS(shared_ea_in_metabox, compat, COMPAT_SHARED_EA_IN_METABOX)
 
+static inline u64 erofs_nid_to_ino64(struct erofs_sb_info *sbi, erofs_nid_t nid)
+{
+	if (!erofs_sb_has_metabox(sbi))
+		return nid;
+
+	/*
+	 * When metadata compression is enabled, avoid generating excessively
+	 * large inode numbers for metadata-compressed inodes.  Shift NIDs in
+	 * the 31-62 bit range left by one and move the metabox flag to bit 31.
+	 *
+	 * Note: on-disk NIDs remain unchanged as they are primarily used for
+	 * compatibility with non-LFS 32-bit applications.
+	 */
+	return ((nid << 1) & GENMASK_ULL(63, 32)) | (nid & GENMASK(30, 0)) |
+		((nid >> EROFS_DIRENT_NID_METABOX_BIT) << 31);
+}
+
 /* atomic flag definitions */
 #define EROFS_I_EA_INITED_BIT	0
 #define EROFS_I_Z_INITED_BIT	1
@@ -281,12 +300,20 @@ struct erofs_inode {
 
 #define EROFS_I(ptr)	container_of(ptr, struct erofs_inode, vfs_inode)
 
+static inline bool erofs_inode_in_metabox(struct inode *inode)
+{
+	return EROFS_I(inode)->nid & BIT(EROFS_DIRENT_NID_METABOX_BIT);
+}
+
 static inline erofs_off_t erofs_iloc(struct inode *inode)
 {
 	struct erofs_sb_info *sbi = EROFS_I_SB(inode);
+	erofs_nid_t nid_lo = EROFS_I(inode)->nid & EROFS_DIRENT_NID_MASK;
 
+	if (erofs_inode_in_metabox(inode))
+		return nid_lo << sbi->islotbits;
 	return erofs_pos(inode->i_sb, sbi->meta_blkaddr) +
-		(EROFS_I(inode)->nid << sbi->islotbits);
+		(nid_lo << sbi->islotbits);
 }
 
 static inline unsigned int erofs_inode_version(unsigned int ifmt)
@@ -385,9 +412,10 @@ void *erofs_read_metadata(struct super_block *sb, struct erofs_buf *buf,
 void erofs_unmap_metabuf(struct erofs_buf *buf);
 void erofs_put_metabuf(struct erofs_buf *buf);
 void *erofs_bread(struct erofs_buf *buf, erofs_off_t offset, bool need_kmap);
-void erofs_init_metabuf(struct erofs_buf *buf, struct super_block *sb);
+int erofs_init_metabuf(struct erofs_buf *buf, struct super_block *sb,
+		       bool in_metabox);
 void *erofs_read_metabuf(struct erofs_buf *buf, struct super_block *sb,
-			 erofs_off_t offset);
+			 erofs_off_t offset, bool in_metabox);
 int erofs_map_dev(struct super_block *sb, struct erofs_map_dev *dev);
 int erofs_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
 		 u64 start, u64 len);
diff --git a/fs/erofs/super.c b/fs/erofs/super.c
index bc27fa3bd678..539551cf59db 100644
--- a/fs/erofs/super.c
+++ b/fs/erofs/super.c
@@ -141,7 +141,7 @@ static int erofs_init_device(struct erofs_buf *buf, struct super_block *sb,
 	struct erofs_deviceslot *dis;
 	struct file *file;
 
-	dis = erofs_read_metabuf(buf, sb, *pos);
+	dis = erofs_read_metabuf(buf, sb, *pos, false);
 	if (IS_ERR(dis))
 		return PTR_ERR(dis);
 
@@ -258,7 +258,7 @@ static int erofs_read_superblock(struct super_block *sb)
 	void *data;
 	int ret;
 
-	data = erofs_read_metabuf(&buf, sb, 0);
+	data = erofs_read_metabuf(&buf, sb, 0, false);
 	if (IS_ERR(data)) {
 		erofs_err(sb, "cannot read erofs superblock");
 		return PTR_ERR(data);
@@ -319,6 +319,14 @@ static int erofs_read_superblock(struct super_block *sb)
 		sbi->root_nid = le16_to_cpu(dsb->rb.rootnid_2b);
 	}
 	sbi->packed_nid = le64_to_cpu(dsb->packed_nid);
+	if (erofs_sb_has_metabox(sbi)) {
+		if (sbi->sb_size <= offsetof(struct erofs_super_block,
+					     metabox_nid))
+			return -EFSCORRUPTED;
+		sbi->metabox_nid = le64_to_cpu(dsb->metabox_nid);
+		if (sbi->metabox_nid & BIT(EROFS_DIRENT_NID_METABOX_BIT))
+			return -EFSCORRUPTED;	/* self-loop detection */
+	}
 	sbi->inos = le64_to_cpu(dsb->inos);
 
 	sbi->epoch = (s64)le64_to_cpu(dsb->epoch);
@@ -335,6 +343,8 @@ static int erofs_read_superblock(struct super_block *sb)
 
 	if (erofs_sb_has_48bit(sbi))
 		erofs_info(sb, "EXPERIMENTAL 48-bit layout support in use. Use at your own risk!");
+	if (erofs_sb_has_metabox(sbi))
+		erofs_info(sb, "EXPERIMENTAL metadata compression support in use. Use at your own risk!");
 	if (erofs_is_fscache_mode(sb))
 		erofs_info(sb, "[deprecated] fscache-based on-demand read feature in use. Use at your own risk!");
 out:
@@ -690,6 +700,12 @@ static int erofs_fc_fill_super(struct super_block *sb, struct fs_context *fc)
 			return PTR_ERR(inode);
 		sbi->packed_inode = inode;
 	}
+	if (erofs_sb_has_metabox(sbi)) {
+		inode = erofs_iget(sb, sbi->metabox_nid);
+		if (IS_ERR(inode))
+			return PTR_ERR(inode);
+		sbi->metabox_inode = inode;
+	}
 
 	inode = erofs_iget(sb, sbi->root_nid);
 	if (IS_ERR(inode))
@@ -845,6 +861,8 @@ static void erofs_drop_internal_inodes(struct erofs_sb_info *sbi)
 {
 	iput(sbi->packed_inode);
 	sbi->packed_inode = NULL;
+	iput(sbi->metabox_inode);
+	sbi->metabox_inode = NULL;
 #ifdef CONFIG_EROFS_FS_ZIP
 	iput(sbi->managed_cache);
 	sbi->managed_cache = NULL;
diff --git a/fs/erofs/sysfs.c b/fs/erofs/sysfs.c
index eed8797a193f..a9fe35ba0906 100644
--- a/fs/erofs/sysfs.c
+++ b/fs/erofs/sysfs.c
@@ -95,6 +95,7 @@ EROFS_ATTR_FEATURE(ztailpacking);
 EROFS_ATTR_FEATURE(fragments);
 EROFS_ATTR_FEATURE(dedupe);
 EROFS_ATTR_FEATURE(48bit);
+EROFS_ATTR_FEATURE(metabox);
 
 static struct attribute *erofs_feat_attrs[] = {
 	ATTR_LIST(zero_padding),
@@ -108,6 +109,7 @@ static struct attribute *erofs_feat_attrs[] = {
 	ATTR_LIST(fragments),
 	ATTR_LIST(dedupe),
 	ATTR_LIST(48bit),
+	ATTR_LIST(metabox),
 	NULL,
 };
 ATTRIBUTE_GROUPS(erofs_feat);
diff --git a/fs/erofs/xattr.c b/fs/erofs/xattr.c
index d61110f511e0..eaa9efd766ee 100644
--- a/fs/erofs/xattr.c
+++ b/fs/erofs/xattr.c
@@ -77,7 +77,9 @@ static int erofs_init_inode_xattrs(struct inode *inode)
 	}
 
 	it.buf = __EROFS_BUF_INITIALIZER;
-	erofs_init_metabuf(&it.buf, sb);
+	ret = erofs_init_metabuf(&it.buf, sb, erofs_inode_in_metabox(inode));
+	if (ret)
+		goto out_unlock;
 	it.pos = erofs_iloc(inode) + vi->inode_isize;
 
 	/* read in shared xattr array (non-atomic, see kmalloc below) */
@@ -326,6 +328,9 @@ static int erofs_xattr_iter_inline(struct erofs_xattr_iter *it,
 		return -ENODATA;
 	}
 
+	ret = erofs_init_metabuf(&it->buf, it->sb, erofs_inode_in_metabox(inode));
+	if (ret)
+		return ret;
 	remaining = vi->xattr_isize - xattr_header_sz;
 	it->pos = erofs_iloc(inode) + vi->inode_isize + xattr_header_sz;
 
@@ -361,12 +366,17 @@ static int erofs_xattr_iter_shared(struct erofs_xattr_iter *it,
 	struct erofs_inode *const vi = EROFS_I(inode);
 	struct super_block *const sb = it->sb;
 	struct erofs_sb_info *sbi = EROFS_SB(sb);
-	unsigned int i;
-	int ret = -ENODATA;
+	unsigned int i = 0;
+	int ret;
 
-	for (i = 0; i < vi->xattr_shared_count; ++i) {
+	ret = erofs_init_metabuf(&it->buf, sb,
+				 erofs_sb_has_shared_ea_in_metabox(sbi));
+	if (ret)
+		return ret;
+
+	while (i < vi->xattr_shared_count) {
 		it->pos = erofs_pos(sb, sbi->xattr_blkaddr) +
-				vi->xattr_shared_xattrs[i] * sizeof(__le32);
+				vi->xattr_shared_xattrs[i++] * sizeof(__le32);
 		it->kaddr = erofs_bread(&it->buf, it->pos, true);
 		if (IS_ERR(it->kaddr))
 			return PTR_ERR(it->kaddr);
@@ -378,7 +388,7 @@ static int erofs_xattr_iter_shared(struct erofs_xattr_iter *it,
 		if ((getxattr && ret != -ENODATA) || (!getxattr && ret))
 			break;
 	}
-	return ret;
+	return i ? ret : -ENODATA;
 }
 
 int erofs_getxattr(struct inode *inode, int index, const char *name,
@@ -413,7 +423,6 @@ int erofs_getxattr(struct inode *inode, int index, const char *name,
 
 	it.sb = inode->i_sb;
 	it.buf = __EROFS_BUF_INITIALIZER;
-	erofs_init_metabuf(&it.buf, it.sb);
 	it.buffer = buffer;
 	it.buffer_size = buffer_size;
 	it.buffer_ofs = 0;
@@ -439,7 +448,6 @@ ssize_t erofs_listxattr(struct dentry *dentry, char *buffer, size_t buffer_size)
 
 	it.sb = dentry->d_sb;
 	it.buf = __EROFS_BUF_INITIALIZER;
-	erofs_init_metabuf(&it.buf, it.sb);
 	it.dentry = dentry;
 	it.buffer = buffer;
 	it.buffer_size = buffer_size;
@@ -485,7 +493,7 @@ int erofs_xattr_prefixes_init(struct super_block *sb)
 	if (sbi->packed_inode)
 		buf.mapping = sbi->packed_inode->i_mapping;
 	else
-		erofs_init_metabuf(&buf, sb);
+		(void)erofs_init_metabuf(&buf, sb, false);
 
 	for (i = 0; i < sbi->xattr_prefix_count; i++) {
 		void *ptr = erofs_read_metadata(sb, &buf, &pos, &len);
diff --git a/fs/erofs/zdata.c b/fs/erofs/zdata.c
index 0d1ddd9b15de..792f20888a8f 100644
--- a/fs/erofs/zdata.c
+++ b/fs/erofs/zdata.c
@@ -855,7 +855,10 @@ static int z_erofs_pcluster_begin(struct z_erofs_frontend *fe)
 		/* bind cache first when cached decompression is preferred */
 		z_erofs_bind_cache(fe);
 	} else {
-		erofs_init_metabuf(&map->buf, sb);
+		ret = erofs_init_metabuf(&map->buf, sb,
+					 erofs_inode_in_metabox(fe->inode));
+		if (ret)
+			return ret;
 		ptr = erofs_bread(&map->buf, map->m_pa, false);
 		if (IS_ERR(ptr)) {
 			ret = PTR_ERR(ptr);
diff --git a/fs/erofs/zmap.c b/fs/erofs/zmap.c
index b72a0e3f9362..a93efd95c555 100644
--- a/fs/erofs/zmap.c
+++ b/fs/erofs/zmap.c
@@ -17,7 +17,7 @@ struct z_erofs_maprecorder {
 	u16 delta[2];
 	erofs_blk_t pblk, compressedblks;
 	erofs_off_t nextpackoff;
-	bool partialref;
+	bool partialref, in_mbox;
 };
 
 static int z_erofs_load_full_lcluster(struct z_erofs_maprecorder *m,
@@ -31,7 +31,7 @@ static int z_erofs_load_full_lcluster(struct z_erofs_maprecorder *m,
 	struct z_erofs_lcluster_index *di;
 	unsigned int advise;
 
-	di = erofs_read_metabuf(&m->map->buf, inode->i_sb, pos);
+	di = erofs_read_metabuf(&m->map->buf, inode->i_sb, pos, m->in_mbox);
 	if (IS_ERR(di))
 		return PTR_ERR(di);
 	m->lcn = lcn;
@@ -146,7 +146,7 @@ static int z_erofs_load_compact_lcluster(struct z_erofs_maprecorder *m,
 	else
 		return -EOPNOTSUPP;
 
-	in = erofs_read_metabuf(&m->map->buf, m->inode->i_sb, pos);
+	in = erofs_read_metabuf(&m->map->buf, inode->i_sb, pos, m->in_mbox);
 	if (IS_ERR(in))
 		return PTR_ERR(in);
 
@@ -392,6 +392,7 @@ static int z_erofs_map_blocks_fo(struct inode *inode,
 	struct z_erofs_maprecorder m = {
 		.inode = inode,
 		.map = map,
+		.in_mbox = erofs_inode_in_metabox(inode),
 	};
 	int err = 0;
 	unsigned int endoff, afmt;
@@ -521,6 +522,7 @@ static int z_erofs_map_blocks_ext(struct inode *inode,
 	unsigned int recsz = z_erofs_extent_recsize(vi->z_advise);
 	erofs_off_t pos = round_up(Z_EROFS_MAP_HEADER_END(erofs_iloc(inode) +
 				   vi->inode_isize + vi->xattr_isize), recsz);
+	bool in_mbox = erofs_inode_in_metabox(inode);
 	erofs_off_t lend = inode->i_size;
 	erofs_off_t l, r, mid, pa, la, lstart;
 	struct z_erofs_extent *ext;
@@ -530,7 +532,7 @@ static int z_erofs_map_blocks_ext(struct inode *inode,
 	map->m_flags = 0;
 	if (recsz <= offsetof(struct z_erofs_extent, pstart_hi)) {
 		if (recsz <= offsetof(struct z_erofs_extent, pstart_lo)) {
-			ext = erofs_read_metabuf(&map->buf, sb, pos);
+			ext = erofs_read_metabuf(&map->buf, sb, pos, in_mbox);
 			if (IS_ERR(ext))
 				return PTR_ERR(ext);
 			pa = le64_to_cpu(*(__le64 *)ext);
@@ -543,7 +545,7 @@ static int z_erofs_map_blocks_ext(struct inode *inode,
 		}
 
 		for (; lstart <= map->m_la; lstart += 1 << vi->z_lclusterbits) {
-			ext = erofs_read_metabuf(&map->buf, sb, pos);
+			ext = erofs_read_metabuf(&map->buf, sb, pos, in_mbox);
 			if (IS_ERR(ext))
 				return PTR_ERR(ext);
 			map->m_plen = le32_to_cpu(ext->plen);
@@ -563,7 +565,7 @@ static int z_erofs_map_blocks_ext(struct inode *inode,
 		for (l = 0, r = vi->z_extents; l < r; ) {
 			mid = l + (r - l) / 2;
 			ext = erofs_read_metabuf(&map->buf, sb,
-						 pos + mid * recsz);
+						 pos + mid * recsz, in_mbox);
 			if (IS_ERR(ext))
 				return PTR_ERR(ext);
 
@@ -645,7 +647,7 @@ static int z_erofs_fill_inode(struct inode *inode, struct erofs_map_blocks *map)
 		goto out_unlock;
 
 	pos = ALIGN(erofs_iloc(inode) + vi->inode_isize + vi->xattr_isize, 8);
-	h = erofs_read_metabuf(&map->buf, sb, pos);
+	h = erofs_read_metabuf(&map->buf, sb, pos, erofs_inode_in_metabox(inode));
 	if (IS_ERR(h)) {
 		err = PTR_ERR(h);
 		goto out_unlock;
-- 
2.43.5

Re: [PATCH v6 2/2] erofs: implement metadata compression

[Chao Yu]

On 7/18/25 11:19, Gao Xiang wrote:
> From: Bo Liu <liubo03@inspur.com>
> 
> Thanks to the meta buffer infrastructure, metadata-compressed inodes are
> just read from the metabox inode instead of the blockdevice (or backing
> file) inode.
> 
> The same is true for shared extended attributes.
> 
> When metadata compression is enabled, inode numbers are divided from
> on-disk NIDs because of non-LTS 32-bit application compatibility.
> 
> Co-developed-by: Gao Xiang <hsiangkao@linux.alibaba.com>
> Signed-off-by: Bo Liu <liubo03@inspur.com>
> Signed-off-by: Gao Xiang <hsiangkao@linux.alibaba.com>

Acked-by: Chao Yu <chao@kernel.org>

Thanks,

[PATCH v7 2/2] erofs: implement metadata compression

[Gao Xiang]

From: Bo Liu <liubo03@inspur.com>

Thanks to the meta buffer infrastructure, metadata-compressed inodes are
just read from the metabox inode instead of the blockdevice (or backing
file) inode.

The same is true for shared extended attributes.

When metadata compression is enabled, inode numbers are divided from
on-disk NIDs because of non-LTS 32-bit application compatibility.

Co-developed-by: Gao Xiang <hsiangkao@linux.alibaba.com>
Signed-off-by: Bo Liu <liubo03@inspur.com>
Acked-by: Chao Yu <chao@kernel.org>
Signed-off-by: Gao Xiang <hsiangkao@linux.alibaba.com>
---
change since v6:
 - fix i386 build warning:
     https://lore.kernel.org/r/b555f01c-4e9e-4b42-aa5a-2d35ef9c1c50@infradead.org
 Documentation/ABI/testing/sysfs-fs-erofs |  2 +-
 fs/erofs/data.c                          | 59 +++++++++++++++---------
 fs/erofs/decompressor.c                  |  2 +-
 fs/erofs/dir.c                           |  3 +-
 fs/erofs/erofs_fs.h                      |  2 +-
 fs/erofs/fileio.c                        |  2 +-
 fs/erofs/fscache.c                       |  3 +-
 fs/erofs/inode.c                         | 17 +++----
 fs/erofs/internal.h                      | 34 ++++++++++++--
 fs/erofs/super.c                         | 22 ++++++++-
 fs/erofs/sysfs.c                         |  2 +
 fs/erofs/xattr.c                         | 26 +++++++----
 fs/erofs/zdata.c                         |  5 +-
 fs/erofs/zmap.c                          | 16 ++++---
 14 files changed, 136 insertions(+), 59 deletions(-)

diff --git a/Documentation/ABI/testing/sysfs-fs-erofs b/Documentation/ABI/testing/sysfs-fs-erofs
index bf3b6299c15e..ad6d1a3ccd4e 100644
--- a/Documentation/ABI/testing/sysfs-fs-erofs
+++ b/Documentation/ABI/testing/sysfs-fs-erofs
@@ -5,7 +5,7 @@ Description:	Shows all enabled kernel features.
 		Supported features:
 		zero_padding, compr_cfgs, big_pcluster, chunked_file,
 		device_table, compr_head2, sb_chksum, ztailpacking,
-		dedupe, fragments.
+		dedupe, fragments, 48bit, metabox.
 
 What:		/sys/fs/erofs/<disk>/sync_decompress
 Date:		November 2021
diff --git a/fs/erofs/data.c b/fs/erofs/data.c
index 383c1337e157..f46c47335b9c 100644
--- a/fs/erofs/data.c
+++ b/fs/erofs/data.c
@@ -49,11 +49,18 @@ void *erofs_bread(struct erofs_buf *buf, erofs_off_t offset, bool need_kmap)
 	return buf->base + (offset & ~PAGE_MASK);
 }
 
-void erofs_init_metabuf(struct erofs_buf *buf, struct super_block *sb)
+int erofs_init_metabuf(struct erofs_buf *buf, struct super_block *sb,
+		       bool in_metabox)
 {
 	struct erofs_sb_info *sbi = EROFS_SB(sb);
 
 	buf->file = NULL;
+	if (in_metabox) {
+		if (unlikely(!sbi->metabox_inode))
+			return -EFSCORRUPTED;
+		buf->mapping = sbi->metabox_inode->i_mapping;
+		return 0;
+	}
 	buf->off = sbi->dif0.fsoff;
 	if (erofs_is_fileio_mode(sbi)) {
 		buf->file = sbi->dif0.file;	/* some fs like FUSE needs it */
@@ -62,12 +69,17 @@ void erofs_init_metabuf(struct erofs_buf *buf, struct super_block *sb)
 		buf->mapping = sbi->dif0.fscache->inode->i_mapping;
 	else
 		buf->mapping = sb->s_bdev->bd_mapping;
+	return 0;
 }
 
 void *erofs_read_metabuf(struct erofs_buf *buf, struct super_block *sb,
-			 erofs_off_t offset)
+			 erofs_off_t offset, bool in_metabox)
 {
-	erofs_init_metabuf(buf, sb);
+	int err;
+
+	err = erofs_init_metabuf(buf, sb, in_metabox);
+	if (err)
+		return ERR_PTR(err);
 	return erofs_bread(buf, offset, true);
 }
 
@@ -118,7 +130,7 @@ int erofs_map_blocks(struct inode *inode, struct erofs_map_blocks *map)
 	pos = ALIGN(erofs_iloc(inode) + vi->inode_isize +
 		    vi->xattr_isize, unit) + unit * chunknr;
 
-	idx = erofs_read_metabuf(&buf, sb, pos);
+	idx = erofs_read_metabuf(&buf, sb, pos, erofs_inode_in_metabox(inode));
 	if (IS_ERR(idx)) {
 		err = PTR_ERR(idx);
 		goto out;
@@ -264,7 +276,6 @@ static int erofs_iomap_begin(struct inode *inode, loff_t offset, loff_t length,
 
 	map.m_la = offset;
 	map.m_llen = length;
-
 	ret = erofs_map_blocks(inode, &map);
 	if (ret < 0)
 		return ret;
@@ -273,35 +284,37 @@ static int erofs_iomap_begin(struct inode *inode, loff_t offset, loff_t length,
 	iomap->length = map.m_llen;
 	iomap->flags = 0;
 	iomap->private = NULL;
+	iomap->addr = IOMAP_NULL_ADDR;
 	if (!(map.m_flags & EROFS_MAP_MAPPED)) {
 		iomap->type = IOMAP_HOLE;
-		iomap->addr = IOMAP_NULL_ADDR;
 		return 0;
 	}
 
-	mdev = (struct erofs_map_dev) {
-		.m_deviceid = map.m_deviceid,
-		.m_pa = map.m_pa,
-	};
-	ret = erofs_map_dev(sb, &mdev);
-	if (ret)
-		return ret;
-
-	if (flags & IOMAP_DAX)
-		iomap->dax_dev = mdev.m_dif->dax_dev;
-	else
-		iomap->bdev = mdev.m_bdev;
-
-	iomap->addr = mdev.m_dif->fsoff + mdev.m_pa;
-	if (flags & IOMAP_DAX)
-		iomap->addr += mdev.m_dif->dax_part_off;
+	if (!(map.m_flags & EROFS_MAP_META) || !erofs_inode_in_metabox(inode)) {
+		mdev = (struct erofs_map_dev) {
+			.m_deviceid = map.m_deviceid,
+			.m_pa = map.m_pa,
+		};
+		ret = erofs_map_dev(sb, &mdev);
+		if (ret)
+			return ret;
+
+		if (flags & IOMAP_DAX)
+			iomap->dax_dev = mdev.m_dif->dax_dev;
+		else
+			iomap->bdev = mdev.m_bdev;
+		iomap->addr = mdev.m_dif->fsoff + mdev.m_pa;
+		if (flags & IOMAP_DAX)
+			iomap->addr += mdev.m_dif->dax_part_off;
+	}
 
 	if (map.m_flags & EROFS_MAP_META) {
 		void *ptr;
 		struct erofs_buf buf = __EROFS_BUF_INITIALIZER;
 
 		iomap->type = IOMAP_INLINE;
-		ptr = erofs_read_metabuf(&buf, sb, mdev.m_pa);
+		ptr = erofs_read_metabuf(&buf, sb, map.m_pa,
+					 erofs_inode_in_metabox(inode));
 		if (IS_ERR(ptr))
 			return PTR_ERR(ptr);
 		iomap->inline_data = ptr;
diff --git a/fs/erofs/decompressor.c b/fs/erofs/decompressor.c
index 358061d7b660..354762c9723f 100644
--- a/fs/erofs/decompressor.c
+++ b/fs/erofs/decompressor.c
@@ -467,7 +467,7 @@ int z_erofs_parse_cfgs(struct super_block *sb, struct erofs_super_block *dsb)
 		return -EOPNOTSUPP;
 	}
 
-	erofs_init_metabuf(&buf, sb);
+	(void)erofs_init_metabuf(&buf, sb, false);
 	offset = EROFS_SUPER_OFFSET + sbi->sb_size;
 	alg = 0;
 	for (algs = sbi->available_compr_algs; algs; algs >>= 1, ++alg) {
diff --git a/fs/erofs/dir.c b/fs/erofs/dir.c
index 3e4b38bec0aa..d1c6cd1a45e8 100644
--- a/fs/erofs/dir.c
+++ b/fs/erofs/dir.c
@@ -34,7 +34,8 @@ static int erofs_fill_dentries(struct inode *dir, struct dir_context *ctx,
 		}
 
 		if (!dir_emit(ctx, de_name, de_namelen,
-			      le64_to_cpu(de->nid), d_type))
+			      erofs_nid_to_ino64(EROFS_SB(dir->i_sb),
+						 le64_to_cpu(de->nid)), d_type))
 			return 1;
 		++de;
 		ctx->pos += sizeof(struct erofs_dirent);
diff --git a/fs/erofs/erofs_fs.h b/fs/erofs/erofs_fs.h
index 7b150ac64742..377ee12b8b96 100644
--- a/fs/erofs/erofs_fs.h
+++ b/fs/erofs/erofs_fs.h
@@ -34,7 +34,7 @@
 #define EROFS_FEATURE_INCOMPAT_48BIT		0x00000080
 #define EROFS_FEATURE_INCOMPAT_METABOX		0x00000100
 #define EROFS_ALL_FEATURE_INCOMPAT		\
-	((EROFS_FEATURE_INCOMPAT_48BIT << 1) - 1)
+	((EROFS_FEATURE_INCOMPAT_METABOX << 1) - 1)
 
 #define EROFS_SB_EXTSLOT_SIZE	16
 
diff --git a/fs/erofs/fileio.c b/fs/erofs/fileio.c
index 3ee082476c8c..b7b3432a9882 100644
--- a/fs/erofs/fileio.c
+++ b/fs/erofs/fileio.c
@@ -115,7 +115,7 @@ static int erofs_fileio_scan_folio(struct erofs_fileio *io, struct folio *folio)
 			void *src;
 
 			src = erofs_read_metabuf(&buf, inode->i_sb,
-						 map->m_pa + ofs);
+				map->m_pa + ofs, erofs_inode_in_metabox(inode));
 			if (IS_ERR(src)) {
 				err = PTR_ERR(src);
 				break;
diff --git a/fs/erofs/fscache.c b/fs/erofs/fscache.c
index 9a8ee646e51d..362acf828279 100644
--- a/fs/erofs/fscache.c
+++ b/fs/erofs/fscache.c
@@ -274,7 +274,8 @@ static int erofs_fscache_data_read_slice(struct erofs_fscache_rq *req)
 		size_t size = map.m_llen;
 		void *src;
 
-		src = erofs_read_metabuf(&buf, sb, map.m_pa);
+		src = erofs_read_metabuf(&buf, sb, map.m_pa,
+					 erofs_inode_in_metabox(inode));
 		if (IS_ERR(src))
 			return PTR_ERR(src);
 
diff --git a/fs/erofs/inode.c b/fs/erofs/inode.c
index 47215c5e3385..9a2f59721522 100644
--- a/fs/erofs/inode.c
+++ b/fs/erofs/inode.c
@@ -29,6 +29,7 @@ static int erofs_read_inode(struct inode *inode)
 	struct super_block *sb = inode->i_sb;
 	erofs_blk_t blkaddr = erofs_blknr(sb, erofs_iloc(inode));
 	unsigned int ofs = erofs_blkoff(sb, erofs_iloc(inode));
+	bool in_mbox = erofs_inode_in_metabox(inode);
 	struct erofs_buf buf = __EROFS_BUF_INITIALIZER;
 	struct erofs_sb_info *sbi = EROFS_SB(sb);
 	erofs_blk_t addrmask = BIT_ULL(48) - 1;
@@ -39,7 +40,7 @@ static int erofs_read_inode(struct inode *inode)
 	void *ptr;
 	int err = 0;
 
-	ptr = erofs_read_metabuf(&buf, sb, erofs_pos(sb, blkaddr));
+	ptr = erofs_read_metabuf(&buf, sb, erofs_pos(sb, blkaddr), in_mbox);
 	if (IS_ERR(ptr)) {
 		err = PTR_ERR(ptr);
 		erofs_err(sb, "failed to read inode meta block (nid: %llu): %d",
@@ -78,7 +79,7 @@ static int erofs_read_inode(struct inode *inode)
 
 			memcpy(&copied, dic, gotten);
 			ptr = erofs_read_metabuf(&buf, sb,
-					erofs_pos(sb, blkaddr + 1));
+					erofs_pos(sb, blkaddr + 1), in_mbox);
 			if (IS_ERR(ptr)) {
 				err = PTR_ERR(ptr);
 				erofs_err(sb, "failed to read inode payload block (nid: %llu): %d",
@@ -264,13 +265,13 @@ static int erofs_fill_inode(struct inode *inode)
  * ino_t is 32-bits on 32-bit arch. We have to squash the 64-bit value down
  * so that it will fit.
  */
-static ino_t erofs_squash_ino(erofs_nid_t nid)
+static ino_t erofs_squash_ino(struct super_block *sb, erofs_nid_t nid)
 {
-	ino_t ino = (ino_t)nid;
+	u64 ino64 = erofs_nid_to_ino64(EROFS_SB(sb), nid);
 
 	if (sizeof(ino_t) < sizeof(erofs_nid_t))
-		ino ^= nid >> (sizeof(erofs_nid_t) - sizeof(ino_t)) * 8;
-	return ino;
+		ino64 ^= ino64 >> (sizeof(erofs_nid_t) - sizeof(ino_t)) * 8;
+	return (ino_t)ino64;
 }
 
 static int erofs_iget5_eq(struct inode *inode, void *opaque)
@@ -282,7 +283,7 @@ static int erofs_iget5_set(struct inode *inode, void *opaque)
 {
 	const erofs_nid_t nid = *(erofs_nid_t *)opaque;
 
-	inode->i_ino = erofs_squash_ino(nid);
+	inode->i_ino = erofs_squash_ino(inode->i_sb, nid);
 	EROFS_I(inode)->nid = nid;
 	return 0;
 }
@@ -291,7 +292,7 @@ struct inode *erofs_iget(struct super_block *sb, erofs_nid_t nid)
 {
 	struct inode *inode;
 
-	inode = iget5_locked(sb, erofs_squash_ino(nid), erofs_iget5_eq,
+	inode = iget5_locked(sb, erofs_squash_ino(sb, nid), erofs_iget5_eq,
 			     erofs_iget5_set, &nid);
 	if (!inode)
 		return ERR_PTR(-ENOMEM);
diff --git a/fs/erofs/internal.h b/fs/erofs/internal.h
index ad932f670bb6..26f9aa57ff3c 100644
--- a/fs/erofs/internal.h
+++ b/fs/erofs/internal.h
@@ -125,6 +125,7 @@ struct erofs_sb_info {
 	struct erofs_sb_lz4_info lz4;
 #endif	/* CONFIG_EROFS_FS_ZIP */
 	struct inode *packed_inode;
+	struct inode *metabox_inode;
 	struct erofs_dev_context *devs;
 	u64 total_blocks;
 
@@ -148,6 +149,7 @@ struct erofs_sb_info {
 	/* what we really care is nid, rather than ino.. */
 	erofs_nid_t root_nid;
 	erofs_nid_t packed_nid;
+	erofs_nid_t metabox_nid;
 	/* used for statfs, f_files - f_favail */
 	u64 inos;
 
@@ -232,6 +234,23 @@ EROFS_FEATURE_FUNCS(sb_chksum, compat, COMPAT_SB_CHKSUM)
 EROFS_FEATURE_FUNCS(xattr_filter, compat, COMPAT_XATTR_FILTER)
 EROFS_FEATURE_FUNCS(shared_ea_in_metabox, compat, COMPAT_SHARED_EA_IN_METABOX)
 
+static inline u64 erofs_nid_to_ino64(struct erofs_sb_info *sbi, erofs_nid_t nid)
+{
+	if (!erofs_sb_has_metabox(sbi))
+		return nid;
+
+	/*
+	 * When metadata compression is enabled, avoid generating excessively
+	 * large inode numbers for metadata-compressed inodes.  Shift NIDs in
+	 * the 31-62 bit range left by one and move the metabox flag to bit 31.
+	 *
+	 * Note: on-disk NIDs remain unchanged as they are primarily used for
+	 * compatibility with non-LFS 32-bit applications.
+	 */
+	return ((nid << 1) & GENMASK_ULL(63, 32)) | (nid & GENMASK(30, 0)) |
+		((nid >> EROFS_DIRENT_NID_METABOX_BIT) << 31);
+}
+
 /* atomic flag definitions */
 #define EROFS_I_EA_INITED_BIT	0
 #define EROFS_I_Z_INITED_BIT	1
@@ -281,12 +300,20 @@ struct erofs_inode {
 
 #define EROFS_I(ptr)	container_of(ptr, struct erofs_inode, vfs_inode)
 
+static inline bool erofs_inode_in_metabox(struct inode *inode)
+{
+	return EROFS_I(inode)->nid & BIT_ULL(EROFS_DIRENT_NID_METABOX_BIT);
+}
+
 static inline erofs_off_t erofs_iloc(struct inode *inode)
 {
 	struct erofs_sb_info *sbi = EROFS_I_SB(inode);
+	erofs_nid_t nid_lo = EROFS_I(inode)->nid & EROFS_DIRENT_NID_MASK;
 
+	if (erofs_inode_in_metabox(inode))
+		return nid_lo << sbi->islotbits;
 	return erofs_pos(inode->i_sb, sbi->meta_blkaddr) +
-		(EROFS_I(inode)->nid << sbi->islotbits);
+		(nid_lo << sbi->islotbits);
 }
 
 static inline unsigned int erofs_inode_version(unsigned int ifmt)
@@ -385,9 +412,10 @@ void *erofs_read_metadata(struct super_block *sb, struct erofs_buf *buf,
 void erofs_unmap_metabuf(struct erofs_buf *buf);
 void erofs_put_metabuf(struct erofs_buf *buf);
 void *erofs_bread(struct erofs_buf *buf, erofs_off_t offset, bool need_kmap);
-void erofs_init_metabuf(struct erofs_buf *buf, struct super_block *sb);
+int erofs_init_metabuf(struct erofs_buf *buf, struct super_block *sb,
+		       bool in_metabox);
 void *erofs_read_metabuf(struct erofs_buf *buf, struct super_block *sb,
-			 erofs_off_t offset);
+			 erofs_off_t offset, bool in_metabox);
 int erofs_map_dev(struct super_block *sb, struct erofs_map_dev *dev);
 int erofs_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
 		 u64 start, u64 len);
diff --git a/fs/erofs/super.c b/fs/erofs/super.c
index bc27fa3bd678..c7b9b784dc3d 100644
--- a/fs/erofs/super.c
+++ b/fs/erofs/super.c
@@ -141,7 +141,7 @@ static int erofs_init_device(struct erofs_buf *buf, struct super_block *sb,
 	struct erofs_deviceslot *dis;
 	struct file *file;
 
-	dis = erofs_read_metabuf(buf, sb, *pos);
+	dis = erofs_read_metabuf(buf, sb, *pos, false);
 	if (IS_ERR(dis))
 		return PTR_ERR(dis);
 
@@ -258,7 +258,7 @@ static int erofs_read_superblock(struct super_block *sb)
 	void *data;
 	int ret;
 
-	data = erofs_read_metabuf(&buf, sb, 0);
+	data = erofs_read_metabuf(&buf, sb, 0, false);
 	if (IS_ERR(data)) {
 		erofs_err(sb, "cannot read erofs superblock");
 		return PTR_ERR(data);
@@ -319,6 +319,14 @@ static int erofs_read_superblock(struct super_block *sb)
 		sbi->root_nid = le16_to_cpu(dsb->rb.rootnid_2b);
 	}
 	sbi->packed_nid = le64_to_cpu(dsb->packed_nid);
+	if (erofs_sb_has_metabox(sbi)) {
+		if (sbi->sb_size <= offsetof(struct erofs_super_block,
+					     metabox_nid))
+			return -EFSCORRUPTED;
+		sbi->metabox_nid = le64_to_cpu(dsb->metabox_nid);
+		if (sbi->metabox_nid & BIT_ULL(EROFS_DIRENT_NID_METABOX_BIT))
+			return -EFSCORRUPTED;	/* self-loop detection */
+	}
 	sbi->inos = le64_to_cpu(dsb->inos);
 
 	sbi->epoch = (s64)le64_to_cpu(dsb->epoch);
@@ -335,6 +343,8 @@ static int erofs_read_superblock(struct super_block *sb)
 
 	if (erofs_sb_has_48bit(sbi))
 		erofs_info(sb, "EXPERIMENTAL 48-bit layout support in use. Use at your own risk!");
+	if (erofs_sb_has_metabox(sbi))
+		erofs_info(sb, "EXPERIMENTAL metadata compression support in use. Use at your own risk!");
 	if (erofs_is_fscache_mode(sb))
 		erofs_info(sb, "[deprecated] fscache-based on-demand read feature in use. Use at your own risk!");
 out:
@@ -690,6 +700,12 @@ static int erofs_fc_fill_super(struct super_block *sb, struct fs_context *fc)
 			return PTR_ERR(inode);
 		sbi->packed_inode = inode;
 	}
+	if (erofs_sb_has_metabox(sbi)) {
+		inode = erofs_iget(sb, sbi->metabox_nid);
+		if (IS_ERR(inode))
+			return PTR_ERR(inode);
+		sbi->metabox_inode = inode;
+	}
 
 	inode = erofs_iget(sb, sbi->root_nid);
 	if (IS_ERR(inode))
@@ -845,6 +861,8 @@ static void erofs_drop_internal_inodes(struct erofs_sb_info *sbi)
 {
 	iput(sbi->packed_inode);
 	sbi->packed_inode = NULL;
+	iput(sbi->metabox_inode);
+	sbi->metabox_inode = NULL;
 #ifdef CONFIG_EROFS_FS_ZIP
 	iput(sbi->managed_cache);
 	sbi->managed_cache = NULL;
diff --git a/fs/erofs/sysfs.c b/fs/erofs/sysfs.c
index eed8797a193f..a9fe35ba0906 100644
--- a/fs/erofs/sysfs.c
+++ b/fs/erofs/sysfs.c
@@ -95,6 +95,7 @@ EROFS_ATTR_FEATURE(ztailpacking);
 EROFS_ATTR_FEATURE(fragments);
 EROFS_ATTR_FEATURE(dedupe);
 EROFS_ATTR_FEATURE(48bit);
+EROFS_ATTR_FEATURE(metabox);
 
 static struct attribute *erofs_feat_attrs[] = {
 	ATTR_LIST(zero_padding),
@@ -108,6 +109,7 @@ static struct attribute *erofs_feat_attrs[] = {
 	ATTR_LIST(fragments),
 	ATTR_LIST(dedupe),
 	ATTR_LIST(48bit),
+	ATTR_LIST(metabox),
 	NULL,
 };
 ATTRIBUTE_GROUPS(erofs_feat);
diff --git a/fs/erofs/xattr.c b/fs/erofs/xattr.c
index d61110f511e0..eaa9efd766ee 100644
--- a/fs/erofs/xattr.c
+++ b/fs/erofs/xattr.c
@@ -77,7 +77,9 @@ static int erofs_init_inode_xattrs(struct inode *inode)
 	}
 
 	it.buf = __EROFS_BUF_INITIALIZER;
-	erofs_init_metabuf(&it.buf, sb);
+	ret = erofs_init_metabuf(&it.buf, sb, erofs_inode_in_metabox(inode));
+	if (ret)
+		goto out_unlock;
 	it.pos = erofs_iloc(inode) + vi->inode_isize;
 
 	/* read in shared xattr array (non-atomic, see kmalloc below) */
@@ -326,6 +328,9 @@ static int erofs_xattr_iter_inline(struct erofs_xattr_iter *it,
 		return -ENODATA;
 	}
 
+	ret = erofs_init_metabuf(&it->buf, it->sb, erofs_inode_in_metabox(inode));
+	if (ret)
+		return ret;
 	remaining = vi->xattr_isize - xattr_header_sz;
 	it->pos = erofs_iloc(inode) + vi->inode_isize + xattr_header_sz;
 
@@ -361,12 +366,17 @@ static int erofs_xattr_iter_shared(struct erofs_xattr_iter *it,
 	struct erofs_inode *const vi = EROFS_I(inode);
 	struct super_block *const sb = it->sb;
 	struct erofs_sb_info *sbi = EROFS_SB(sb);
-	unsigned int i;
-	int ret = -ENODATA;
+	unsigned int i = 0;
+	int ret;
 
-	for (i = 0; i < vi->xattr_shared_count; ++i) {
+	ret = erofs_init_metabuf(&it->buf, sb,
+				 erofs_sb_has_shared_ea_in_metabox(sbi));
+	if (ret)
+		return ret;
+
+	while (i < vi->xattr_shared_count) {
 		it->pos = erofs_pos(sb, sbi->xattr_blkaddr) +
-				vi->xattr_shared_xattrs[i] * sizeof(__le32);
+				vi->xattr_shared_xattrs[i++] * sizeof(__le32);
 		it->kaddr = erofs_bread(&it->buf, it->pos, true);
 		if (IS_ERR(it->kaddr))
 			return PTR_ERR(it->kaddr);
@@ -378,7 +388,7 @@ static int erofs_xattr_iter_shared(struct erofs_xattr_iter *it,
 		if ((getxattr && ret != -ENODATA) || (!getxattr && ret))
 			break;
 	}
-	return ret;
+	return i ? ret : -ENODATA;
 }
 
 int erofs_getxattr(struct inode *inode, int index, const char *name,
@@ -413,7 +423,6 @@ int erofs_getxattr(struct inode *inode, int index, const char *name,
 
 	it.sb = inode->i_sb;
 	it.buf = __EROFS_BUF_INITIALIZER;
-	erofs_init_metabuf(&it.buf, it.sb);
 	it.buffer = buffer;
 	it.buffer_size = buffer_size;
 	it.buffer_ofs = 0;
@@ -439,7 +448,6 @@ ssize_t erofs_listxattr(struct dentry *dentry, char *buffer, size_t buffer_size)
 
 	it.sb = dentry->d_sb;
 	it.buf = __EROFS_BUF_INITIALIZER;
-	erofs_init_metabuf(&it.buf, it.sb);
 	it.dentry = dentry;
 	it.buffer = buffer;
 	it.buffer_size = buffer_size;
@@ -485,7 +493,7 @@ int erofs_xattr_prefixes_init(struct super_block *sb)
 	if (sbi->packed_inode)
 		buf.mapping = sbi->packed_inode->i_mapping;
 	else
-		erofs_init_metabuf(&buf, sb);
+		(void)erofs_init_metabuf(&buf, sb, false);
 
 	for (i = 0; i < sbi->xattr_prefix_count; i++) {
 		void *ptr = erofs_read_metadata(sb, &buf, &pos, &len);
diff --git a/fs/erofs/zdata.c b/fs/erofs/zdata.c
index 0d1ddd9b15de..792f20888a8f 100644
--- a/fs/erofs/zdata.c
+++ b/fs/erofs/zdata.c
@@ -855,7 +855,10 @@ static int z_erofs_pcluster_begin(struct z_erofs_frontend *fe)
 		/* bind cache first when cached decompression is preferred */
 		z_erofs_bind_cache(fe);
 	} else {
-		erofs_init_metabuf(&map->buf, sb);
+		ret = erofs_init_metabuf(&map->buf, sb,
+					 erofs_inode_in_metabox(fe->inode));
+		if (ret)
+			return ret;
 		ptr = erofs_bread(&map->buf, map->m_pa, false);
 		if (IS_ERR(ptr)) {
 			ret = PTR_ERR(ptr);
diff --git a/fs/erofs/zmap.c b/fs/erofs/zmap.c
index b72a0e3f9362..a93efd95c555 100644
--- a/fs/erofs/zmap.c
+++ b/fs/erofs/zmap.c
@@ -17,7 +17,7 @@ struct z_erofs_maprecorder {
 	u16 delta[2];
 	erofs_blk_t pblk, compressedblks;
 	erofs_off_t nextpackoff;
-	bool partialref;
+	bool partialref, in_mbox;
 };
 
 static int z_erofs_load_full_lcluster(struct z_erofs_maprecorder *m,
@@ -31,7 +31,7 @@ static int z_erofs_load_full_lcluster(struct z_erofs_maprecorder *m,
 	struct z_erofs_lcluster_index *di;
 	unsigned int advise;
 
-	di = erofs_read_metabuf(&m->map->buf, inode->i_sb, pos);
+	di = erofs_read_metabuf(&m->map->buf, inode->i_sb, pos, m->in_mbox);
 	if (IS_ERR(di))
 		return PTR_ERR(di);
 	m->lcn = lcn;
@@ -146,7 +146,7 @@ static int z_erofs_load_compact_lcluster(struct z_erofs_maprecorder *m,
 	else
 		return -EOPNOTSUPP;
 
-	in = erofs_read_metabuf(&m->map->buf, m->inode->i_sb, pos);
+	in = erofs_read_metabuf(&m->map->buf, inode->i_sb, pos, m->in_mbox);
 	if (IS_ERR(in))
 		return PTR_ERR(in);
 
@@ -392,6 +392,7 @@ static int z_erofs_map_blocks_fo(struct inode *inode,
 	struct z_erofs_maprecorder m = {
 		.inode = inode,
 		.map = map,
+		.in_mbox = erofs_inode_in_metabox(inode),
 	};
 	int err = 0;
 	unsigned int endoff, afmt;
@@ -521,6 +522,7 @@ static int z_erofs_map_blocks_ext(struct inode *inode,
 	unsigned int recsz = z_erofs_extent_recsize(vi->z_advise);
 	erofs_off_t pos = round_up(Z_EROFS_MAP_HEADER_END(erofs_iloc(inode) +
 				   vi->inode_isize + vi->xattr_isize), recsz);
+	bool in_mbox = erofs_inode_in_metabox(inode);
 	erofs_off_t lend = inode->i_size;
 	erofs_off_t l, r, mid, pa, la, lstart;
 	struct z_erofs_extent *ext;
@@ -530,7 +532,7 @@ static int z_erofs_map_blocks_ext(struct inode *inode,
 	map->m_flags = 0;
 	if (recsz <= offsetof(struct z_erofs_extent, pstart_hi)) {
 		if (recsz <= offsetof(struct z_erofs_extent, pstart_lo)) {
-			ext = erofs_read_metabuf(&map->buf, sb, pos);
+			ext = erofs_read_metabuf(&map->buf, sb, pos, in_mbox);
 			if (IS_ERR(ext))
 				return PTR_ERR(ext);
 			pa = le64_to_cpu(*(__le64 *)ext);
@@ -543,7 +545,7 @@ static int z_erofs_map_blocks_ext(struct inode *inode,
 		}
 
 		for (; lstart <= map->m_la; lstart += 1 << vi->z_lclusterbits) {
-			ext = erofs_read_metabuf(&map->buf, sb, pos);
+			ext = erofs_read_metabuf(&map->buf, sb, pos, in_mbox);
 			if (IS_ERR(ext))
 				return PTR_ERR(ext);
 			map->m_plen = le32_to_cpu(ext->plen);
@@ -563,7 +565,7 @@ static int z_erofs_map_blocks_ext(struct inode *inode,
 		for (l = 0, r = vi->z_extents; l < r; ) {
 			mid = l + (r - l) / 2;
 			ext = erofs_read_metabuf(&map->buf, sb,
-						 pos + mid * recsz);
+						 pos + mid * recsz, in_mbox);
 			if (IS_ERR(ext))
 				return PTR_ERR(ext);
 
@@ -645,7 +647,7 @@ static int z_erofs_fill_inode(struct inode *inode, struct erofs_map_blocks *map)
 		goto out_unlock;
 
 	pos = ALIGN(erofs_iloc(inode) + vi->inode_isize + vi->xattr_isize, 8);
-	h = erofs_read_metabuf(&map->buf, sb, pos);
+	h = erofs_read_metabuf(&map->buf, sb, pos, erofs_inode_in_metabox(inode));
 	if (IS_ERR(h)) {
 		err = PTR_ERR(h);
 		goto out_unlock;
-- 
2.43.5