Re: [PATCH v4] erofs: reuse superblock for file-backed mounts

Hi Christian,

Christian Brauner <[email protected]> writes:

> Hi Giuseppe,
>
>> When the same file-backed image is mounted multiple times (via path or
>> fd) with the superblock_share mount option, reuse the existing
>> superblock instead of creating a new one.  This allows multiple mounts
>> of the same image to share in-kernel data structures more efficiently.
>> 
>> The backing file is identified by its inode and fsoffset.  If mount
>> options conflict or extra devices are used, a separate superblock
>> is created transparently as a fallback.  Remount is not allowed on
>> shared superblocks.
>> 
>> Tested by mounting a 177M Fedora EROFS image 20 times with full
>> traversal:
>> 
>>   #!/bin/sh
>>   IMG=${1:-/root/fedora.erofs}
>>   OPTS=${2:+-o $2}
>>   N=20
>>   DIR=$(mktemp -d)
>>   trap "umount $DIR/m* 2>/dev/null; rm -rf $DIR" EXIT
>> 
>>   sync; echo 3 > /proc/sys/vm/drop_caches
>>   INODES_BEFORE=$(grep erofs_inode /proc/slabinfo | awk '{print $2}')
>>   MEM_BEFORE=$(grep ^Slab: /proc/meminfo | awk '{print $2}')
>> 
>>   for i in $(seq 1 $N); do
>>       mkdir $DIR/m$i
>>       mount -t erofs $OPTS "$IMG" $DIR/m$i
>>       find $DIR/m$i > /dev/null
>>   done
>> 
>>   echo "Superblocks: $(grep $DIR /proc/self/mountinfo | \
>>       awk '{print $3}' | sort -u | wc -l)"
>>   echo "erofs_inode delta: +$(( $(grep erofs_inode /proc/slabinfo | \
>>       awk '{print $2}') - INODES_BEFORE ))"
>>   echo "Slab delta: +$(( $(grep ^Slab: /proc/meminfo | \
>>       awk '{print $2}') - MEM_BEFORE )) kB"
>> 
>> without superblock_share:
>> 
>>   # time ./test.sh /root/fedora.erofs
>>   Superblocks: 20
>>   erofs_inode delta: +45864
>>   Slab delta: +37448 kB
>>   real	0m2.311s
>>   user	0m0.222s
>>   sys	0m1.998s
>> 
>> with superblock_share:
>> 
>>   # time ./test.sh /root/fedora.erofs superblock_share
>>   Superblocks: 1
>>   erofs_inode delta: +2044
>>   Slab delta: +284 kB
>>   real	0m0.679s
>>   user	0m0.157s
>>   sys	0m0.469s
>> 
>> The time difference shows that sharing the superblock also benefits
>> the page cache and inode cache, as subsequent mounts of the same image
>> avoid re-reading the backing file.  This is particularly useful for
>> container hosts running multiple containers from the same base image.
>> 
>> Signed-off-by: Giuseppe Scrivano <[email protected]>
>>
>> diff --git a/Documentation/filesystems/erofs.rst b/Documentation/filesystems/erofs.rst
>> index d301d9ac946a..49c4e7dbc5d6 100644
>> --- a/Documentation/filesystems/erofs.rst
>> +++ b/Documentation/filesystems/erofs.rst
>> @@ -139,6 +139,9 @@ inode_share            Enable inode page sharing for this filesystem.  Inodes wi
>>                         page cache.
>>  source=%s              (For file-backed mounts) Specify the backing image as a path
>>                         or as an already-opened file descriptor.
>> +superblock_share       (For file-backed mounts) Share the superblock when the same
>> +                       backing file is mounted more than once with compatible
>> +                       options.  Remount is not allowed on shared superblocks.
>>  ===================    =========================================================
>>  
>>  File-backed mounts
>> @@ -156,6 +159,11 @@ Only regular files are accepted as backing files; to mount an image that
>>  resides on a block device, use the traditional block device mount path
>>  instead.
>>  
>> +When superblock_share is specified and the same backing file (identified
>> +by inode and fsoffset) is mounted more than once with compatible mount
>> +options, the kernel reuses the existing superblock instead of creating a
>> +new one.  Remount is not allowed on shared superblocks.
>> +
>>  Sysfs Entries
>>  =============
>>  
>> diff --git a/fs/erofs/internal.h b/fs/erofs/internal.h
>> index 57bd21859c65..86da219f8f4e 100644
>> --- a/fs/erofs/internal.h
>> +++ b/fs/erofs/internal.h
>> @@ -156,6 +156,7 @@ struct erofs_sb_info {
>>  #define EROFS_MOUNT_DAX_NEVER		0x00000080
>>  #define EROFS_MOUNT_DIRECT_IO		0x00000100
>>  #define EROFS_MOUNT_INODE_SHARE		0x00000200
>> +#define EROFS_MOUNT_SUPERBLOCK_SHARE	0x00000400
>>  
>>  #define clear_opt(opt, option)	((opt)->mount_opt &= ~EROFS_MOUNT_##option)
>>  #define set_opt(opt, option)	((opt)->mount_opt |= EROFS_MOUNT_##option)
>> diff --git a/fs/erofs/super.c b/fs/erofs/super.c
>> index 3d92caec8d3a..4e926376e17b 100644
>> --- a/fs/erofs/super.c
>> +++ b/fs/erofs/super.c
>> @@ -386,7 +386,7 @@ static void erofs_default_options(struct erofs_sb_info *sbi)
>>  enum {
>>  	Opt_user_xattr, Opt_acl, Opt_cache_strategy, Opt_dax, Opt_dax_enum,
>>  	Opt_device, Opt_domain_id, Opt_directio, Opt_fsoffset, Opt_inode_share,
>> -	Opt_source,
>> +	Opt_source, Opt_superblock_share,
>>  };
>>  
>>  static const struct constant_table erofs_param_cache_strategy[] = {
>> @@ -415,6 +415,7 @@ static const struct fs_parameter_spec erofs_fs_parameters[] = {
>>  	fsparam_u64("fsoffset",			Opt_fsoffset),
>>  	fsparam_flag("inode_share",		Opt_inode_share),
>>  	fsparam_file_or_string("source",	Opt_source),
>> +	fsparam_flag("superblock_share",	Opt_superblock_share),
>>  	{}
>>  };
>>  
>> @@ -560,6 +561,12 @@ static int erofs_fc_parse_param(struct fs_context *fc,
>>  		else
>>  			set_opt(&sbi->opt, INODE_SHARE);
>>  		break;
>> +	case Opt_superblock_share:
>> +		if (!IS_ENABLED(CONFIG_EROFS_FS_BACKED_BY_FILE))
>> +			errorfc(fc, "%s option not supported", erofs_fs_parameters[opt].name);
>> +		else
>> +			set_opt(&sbi->opt, SUPERBLOCK_SHARE);
>> +		break;
>>  	case Opt_source:
>>  		return erofs_fc_parse_source(fc, param);
>>  	}
>> @@ -652,6 +659,17 @@ static int erofs_fc_fill_super(struct super_block *sb, struct fs_context *fc)
>>  		errorfc(fc, "FSDAX is not allowed when inode_share is on");
>>  		return -EINVAL;
>>  	}
>> +	if (test_opt(&sbi->opt, SUPERBLOCK_SHARE) &&
>> +	    test_opt(&sbi->opt, INODE_SHARE)) {
>> +		errorfc(fc, "superblock_share is not allowed when inode_share is on");
>> +		return -EINVAL;
>> +	}
>> +	/* Extra devices are not yet supported with superblock sharing.  */
>> +	if (test_opt(&sbi->opt, SUPERBLOCK_SHARE) &&
>> +	    sbi->devs->extra_devices) {
>> +		errorfc(fc, "superblock_share does not support extra devices");
>> +		return -EINVAL;
>> +	}
>>  
>>  	sbi->blkszbits = PAGE_SHIFT;
>>  	if (!sb->s_bdev) {
>> @@ -779,6 +797,51 @@ static int erofs_fc_fill_super(struct super_block *sb, struct fs_context *fc)
>>  	return 0;
>>  }
>>  
>> +static int erofs_fc_test_file_super(struct super_block *sb,
>> +				    struct fs_context *fc)
>> +{
>> +	struct erofs_sb_info *sbi = EROFS_SB(sb);
>> +	struct erofs_sb_info *new_sbi = fc->s_fs_info;
>> +
>> +	if (sb->s_iflags & SB_I_RETIRED)
>> +		return 0;
>> +	if (!sbi->dif0.file || !new_sbi->dif0.file)
>> +		return 0;
>> +	if (!test_opt(&new_sbi->opt, SUPERBLOCK_SHARE))
>> +		return 0;
>> +	return file_inode(sbi->dif0.file) == file_inode(new_sbi->dif0.file) &&
>> +	       sbi->dif0.fsoff == new_sbi->dif0.fsoff &&
>> +	       !memcmp(&sbi->opt, &new_sbi->opt, sizeof(sbi->opt));
>> +}
>> +
>> +static int erofs_fc_get_tree_file(struct fs_context *fc)
>> +{
>> +	struct erofs_sb_info *sbi = fc->s_fs_info;
>> +	struct super_block *sb;
>> +	int err;
>> +
>> +	if (!test_opt(&sbi->opt, SUPERBLOCK_SHARE))
>> +		return get_tree_nodev(fc, erofs_fc_fill_super);
>> +
>> +	sb = sget_fc(fc, erofs_fc_test_file_super, set_anon_super_fc);
>> +	if (IS_ERR(sb))
>> +		return PTR_ERR(sb);
>> +
>> +	if (sb->s_root) {
>> +		erofs_info(sb, "sharing superblock for the same backing file");
>> +	} else {
>> +		err = erofs_fc_fill_super(sb, fc);
>> +		if (err) {
>> +			deactivate_locked_super(sb);
>> +			return err;
>> +		}
>> +		sb->s_flags |= SB_ACTIVE;
>> +	}
>> +
>> +	fc->root = dget(sb->s_root);
>> +	return 0;
>> +}
>
> I think you should rename vfs_get_super() to get_tree_super() and expose
> it to modules. Then this function collapses to:
>
> if (!test_opt(&sbi->opt, SUPERBLOCK_SHARE))
> 	get_tree_super(fc, NULL, erofs_fc_fill_super);
>
> return get_tree_super(fc, erofs_fc_test_file_super, erofs_fc_fill_super);
>
> or simpler, if you move the sbi test for SUPERBLOCK_SHARE into
> erofs_fc_test_file_super:
>
> get_tree_super(fc, erofs_fc_test_file_super, erofs_fc_fill_super);

that is neat!  I'll send a new version with this fix.

Regards,
Giuseppe