[PATCH 2/2] erofs-utils: mkfs: add `--inode-digest-xattr` option

Hongbo Li lihongbo22 at huawei.com
Tue Dec 30 00:06:10 AEDT 2025


Hi Xiang,

On 2025/12/29 15:49, Gao Xiang wrote:
> Based on the original Hongbo's version [1], it enables storing the
> SHA-256 digest of each inode as an extended attribute, in preparation
> for the upcoming page cache sharing feature.
> 
> Example usage:
>   $ mkfs.erofs --inode-digest-xattr=trusted.erofs.fingerprint [-zlz4hc] foo.erofs foo/
> 
> Co-developed-by: Hongbo Li <lihongbo22 at huawei.com>
> [1] https://lore.kernel.org/r/20251118015849.228939-1-lihongbo22@huawei.com
> Signed-off-by: Gao Xiang <hsiangkao at linux.alibaba.com>
> ---
>   include/erofs/internal.h |  2 ++
>   include/erofs/xattr.h    |  2 ++
>   include/erofs_fs.h       |  4 +++-
>   lib/inode.c              | 44 ++++++++++++++++++++++++++++++++++++++--
>   lib/super.c              | 13 ++++++++++--
>   lib/xattr.c              | 14 ++++++++++++-
>   mkfs/main.c              | 27 ++++++++++++++++++++----
>   7 files changed, 96 insertions(+), 10 deletions(-)
> 
> diff --git a/include/erofs/internal.h b/include/erofs/internal.h
> index 62594b877151..5798f10e89c2 100644
> --- a/include/erofs/internal.h
> +++ b/include/erofs/internal.h
> @@ -130,6 +130,7 @@ struct erofs_sb_info {
>   
>   	u32 xattr_prefix_start;
>   	u8 xattr_prefix_count;
> +	u8 ishare_xattr_prefix_id;
>   	struct erofs_xattr_prefix_item *xattr_prefixes;
>   
>   	struct erofs_vfile bdev;
> @@ -189,6 +190,7 @@ 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(plain_xattr_pfx, compat, COMPAT_PLAIN_XATTR_PFX)
> +EROFS_FEATURE_FUNCS(ishare_xattrs, compat, COMPAT_ISHARE_XATTRS)
>   
>   #define EROFS_I_EA_INITED_BIT	0
>   #define EROFS_I_Z_INITED_BIT	1
> diff --git a/include/erofs/xattr.h b/include/erofs/xattr.h
> index 941bed778956..96546364f316 100644
> --- a/include/erofs/xattr.h
> +++ b/include/erofs/xattr.h
> @@ -33,6 +33,8 @@ char *erofs_export_xattr_ibody(struct erofs_inode *inode);
>   int erofs_load_shared_xattrs_from_path(struct erofs_sb_info *sbi, const char *path,
>   				       long inlinexattr_tolerance);
>   int erofs_xattr_insert_name_prefix(const char *prefix);
> +int erofs_xattr_set_ishare_prefix(struct erofs_sb_info *sbi,
> +				  const char *prefix);
>   void erofs_xattr_cleanup_name_prefixes(void);
>   int erofs_xattr_flush_name_prefixes(struct erofs_importer *im, bool plain);
>   int erofs_xattr_prefixes_init(struct erofs_sb_info *sbi);
> diff --git a/include/erofs_fs.h b/include/erofs_fs.h
> index 887f37faba59..8b0d155f8c4c 100644
> --- a/include/erofs_fs.h
> +++ b/include/erofs_fs.h
> @@ -17,6 +17,7 @@
>   #define EROFS_FEATURE_COMPAT_MTIME              0x00000002
>   #define EROFS_FEATURE_COMPAT_XATTR_FILTER	0x00000004
>   #define EROFS_FEATURE_COMPAT_PLAIN_XATTR_PFX	0x00000010
> +#define EROFS_FEATURE_COMPAT_ISHARE_XATTRS	0x00000020
>   
>   /*
>    * Any bits that aren't in EROFS_ALL_FEATURE_INCOMPAT should
> @@ -82,7 +83,8 @@ struct erofs_super_block {
>   	__le32 xattr_prefix_start;	/* start of long xattr prefixes */
>   	__le64 packed_nid;	/* nid of the special packed inode */
>   	__u8 xattr_filter_reserved; /* reserved for xattr name filter */
> -	__u8 reserved[3];
> +	__u8 ishare_xattr_prefix_id;
> +	__u8 reserved[2];
>   	__le32 build_time;	/* seconds added to epoch for mkfs time */
>   	__le64 rootnid_8b;	/* (48BIT on) nid of root directory */
>   	__le64 reserved2;
> diff --git a/lib/inode.c b/lib/inode.c
> index 1d08e39317c0..c6085a52281a 100644
> --- a/lib/inode.c
> +++ b/lib/inode.c
> @@ -31,6 +31,7 @@
>   #include "liberofs_metabox.h"
>   #include "liberofs_private.h"
>   #include "liberofs_rebuild.h"
> +#include "sha256.h"
>   
>   static inline bool erofs_is_special_identifier(const char *path)
>   {
> @@ -1879,6 +1880,37 @@ static int erofs_prepare_dir_inode(struct erofs_importer *im,
>   	return 0;
>   }
>   
> +static int erofs_set_inode_fingerprint(struct erofs_inode *inode, int fd,
> +				       erofs_off_t pos)
> +{
> +	u8 ishare_xattr_prefix_id = inode->sbi->ishare_xattr_prefix_id;
> +	erofs_off_t remaining = inode->i_size;
> +	struct erofs_vfile vf = { .fd = fd };
> +	struct sha256_state md;
> +	u8 out[32 + sizeof("sha256:") - 1];
> +	int ret;
> +
> +	if (!ishare_xattr_prefix_id)
> +		return 0;
> +	erofs_sha256_init(&md);
> +	do {
> +		u8 buf[32768];
> +
> +		ret = erofs_io_pread(&vf, buf,
> +				     min_t(u64, remaining, sizeof(buf)), pos);
> +		if (ret < 0)
> +			return ret;
> +		if (ret > 0)
> +			erofs_sha256_process(&md, buf, ret);
> +		remaining -= ret;
> +		pos += ret;
> +	} while (remaining);
> +	erofs_sha256_done(&md, out + sizeof("sha256:") - 1);
> +	memcpy(out, "sha256:", sizeof("sha256:") - 1);
> +	return erofs_setxattr(inode, ishare_xattr_prefix_id, "",
> +			      out, sizeof(out));
> +}
> +
>   static int erofs_mkfs_begin_nondirectory(struct erofs_importer *im,
>   					 struct erofs_inode *inode)
>   {
> @@ -1899,9 +1931,16 @@ static int erofs_mkfs_begin_nondirectory(struct erofs_importer *im,
>   				return -errno;
>   			__erofs_fallthrough;
>   		default:
> -			break;
> +			goto out;

We should keep use break here to set inode fingerprint in following step.

Another is that, here we use "erofs_set_inode_fingerprint" to hook the 
fingerprint xattr in inode->xattrs, but the "erofs_prepare_xattr_ibody" 
has calculated the xattr_size before. So should we move the 
"erofs_prepare_xattr_ibody" later so that to make sure the fingerprint 
xattr is set?

Thanks,
Hongbo

> +		}
> +
> +		if (S_ISREG(inode->i_mode) && inode->i_size) {
> +			ret = erofs_set_inode_fingerprint(inode, ctx.fd, ctx.fpos);
> +			if (ret < 0)
> +				return ret;
>   		}
> -		if (ctx.fd >= 0 && cfg.c_compr_opts[0].alg &&
> +
> +		if (cfg.c_compr_opts[0].alg &&
>   		    erofs_file_is_compressible(im, inode)) {
>   			ctx.ictx = erofs_prepare_compressed_file(im, inode);
>   			if (IS_ERR(ctx.ictx))
> @@ -1913,6 +1952,7 @@ static int erofs_mkfs_begin_nondirectory(struct erofs_importer *im,
>   				return ret;
>   		}
>   	}
> +out:
>   	return erofs_mkfs_go(im, EROFS_MKFS_JOB_NDIR, &ctx, sizeof(ctx));
>   }
>   
> diff --git a/lib/super.c b/lib/super.c
> index e54aff2d4ab7..a4837e5702ed 100644
> --- a/lib/super.c
> +++ b/lib/super.c
> @@ -146,7 +146,15 @@ int erofs_read_superblock(struct erofs_sb_info *sbi)
>   	sbi->build_time = le32_to_cpu(dsb->build_time);
>   
>   	memcpy(&sbi->uuid, dsb->uuid, sizeof(dsb->uuid));
> -
> +	if (erofs_sb_has_ishare_xattrs(sbi)) {
> +		if (dsb->ishare_xattr_prefix_id >= sbi->xattr_prefix_count) {
> +			erofs_err("invalid ishare xattr prefix id %d",
> +				  dsb->ishare_xattr_prefix_id);
> +			return -EFSCORRUPTED;
> +		}
> +		sbi->ishare_xattr_prefix_id =
> +			dsb->ishare_xattr_prefix_id | EROFS_XATTR_LONG_PREFIX;
> +	}
>   	ret = z_erofs_parse_cfgs(sbi, dsb);
>   	if (ret)
>   		return ret;
> @@ -160,7 +168,6 @@ int erofs_read_superblock(struct erofs_sb_info *sbi)
>   		free(sbi->devs);
>   		sbi->devs = NULL;
>   	}
> -
>   	sbi->sb_valid = !ret;
>   	return ret;
>   }
> @@ -206,6 +213,8 @@ int erofs_writesb(struct erofs_sb_info *sbi)
>   		.extra_devices = cpu_to_le16(sbi->extra_devices),
>   		.devt_slotoff = cpu_to_le16(sbi->devt_slotoff),
>   		.packed_nid = cpu_to_le64(sbi->packed_nid),
> +		.ishare_xattr_prefix_id = sbi->ishare_xattr_prefix_id &
> +			EROFS_XATTR_LONG_PREFIX_MASK,
>   	};
>   	char *buf;
>   	int ret;
> diff --git a/lib/xattr.c b/lib/xattr.c
> index b6b1a5e600fb..764aee3be3c3 100644
> --- a/lib/xattr.c
> +++ b/lib/xattr.c
> @@ -1495,9 +1495,21 @@ int erofs_xattr_insert_name_prefix(const char *prefix)
>   	}
>   
>   	tnode->index = EROFS_XATTR_LONG_PREFIX | ea_prefix_count;
> -	ea_prefix_count++;
>   	init_list_head(&tnode->list);
>   	list_add_tail(&tnode->list, &ea_name_prefixes);
> +	return ea_prefix_count++;
> +}
> +
> +int erofs_xattr_set_ishare_prefix(struct erofs_sb_info *sbi,
> +				  const char *prefix)
> +{
> +	int err;
> +
> +	err = erofs_xattr_insert_name_prefix(prefix);
> +	if (err < 0)
> +		return err;
> +	sbi->ishare_xattr_prefix_id = EROFS_XATTR_LONG_PREFIX | err;
> +	erofs_sb_set_ishare_xattrs(sbi);
>   	return 0;
>   }
>   
> diff --git a/mkfs/main.c b/mkfs/main.c
> index aaa0300bca1b..7fd6e8419d44 100644
> --- a/mkfs/main.c
> +++ b/mkfs/main.c
> @@ -102,6 +102,7 @@ static struct option long_options[] = {
>   #endif
>   	{"zD", optional_argument, NULL, 536},
>   	{"ZI", optional_argument, NULL, 537},
> +	{"inode-digest-xattr", required_argument, NULL, 538},
>   	{0, 0, 0, 0},
>   };
>   
> @@ -1262,7 +1263,7 @@ static int mkfs_parse_options_cfg(struct erofs_importer_params *params,
>   		case 19:
>   			errno = 0;
>   			opt = erofs_xattr_insert_name_prefix(optarg);
> -			if (opt) {
> +			if (opt < 0) {
>   				erofs_err("failed to parse xattr name prefix: %s",
>   					  erofs_strerror(opt));
>   				return opt;
> @@ -1424,6 +1425,14 @@ static int mkfs_parse_options_cfg(struct erofs_importer_params *params,
>   			else
>   				mkfscfg.inode_metazone = false;
>   			break;
> +		case 538:
> +			err = erofs_xattr_set_ishare_prefix(&g_sbi, optarg);
> +			if (err < 0) {
> +				erofs_err("failed to parse ishare name: %s",
> +					  erofs_strerror(err));
> +				return err;
> +			}
> +			break;
>   		case 'V':
>   			version();
>   			exit(0);
> @@ -1878,9 +1887,12 @@ int main(int argc, char **argv)
>   			goto exit;
>   		}
>   
> -		if (cfg.c_extra_ea_name_prefixes)
> -			erofs_xattr_flush_name_prefixes(&importer,
> -							mkfs_plain_xattr_pfx);
> +		err = erofs_xattr_flush_name_prefixes(&importer,
> +						      mkfs_plain_xattr_pfx);
> +		if (err) {
> +			erofs_err("failed to flush long xattr prefixes: %s",
> +				  erofs_strerror(err));
> +		}
>   
>   		root = erofs_new_inode(&g_sbi);
>   		if (IS_ERR(root)) {
> @@ -1888,6 +1900,13 @@ int main(int argc, char **argv)
>   			goto exit;
>   		}
>   	} else {
> +		err = erofs_xattr_flush_name_prefixes(&importer,
> +						      mkfs_plain_xattr_pfx);
> +		if (err) {
> +			erofs_err("failed to flush long xattr prefixes: %s",
> +				  erofs_strerror(err));
> +		}
> +
>   		root = erofs_rebuild_make_root(&g_sbi);
>   		if (IS_ERR(root)) {
>   			err = PTR_ERR(root);


More information about the Linux-erofs mailing list