Skip to contentSkip to conversation

f2fs: clean up /sys/fs/f2fs/<disk>/features

PATCH In review Jaegeuk Kim 2 participants 7 messages Jun 3, 2021 – Jun 4, 2021 F2FS file system

lore
Share & tools
lore

Jaegeuk Kim

author maintainer
#0 lore raw

PATCH 1/1 f2fs: clean up /sys/fs/f2fs/<disk>/features

Let's create /sys/fs/f2fs/<disk>/feature_list/ to meet sysfs rule.

---
 Documentation/ABI/testing/sysfs-fs-f2fs |  18 ++-
 fs/f2fs/f2fs.h                          |   3 +
 fs/f2fs/sysfs.c                         | 152 +++++++++++++++++++++++-
 3 files changed, 168 insertions(+), 5 deletions(-)
Documentation/ABI/testing/sysfs-fs-f2fs +17 -1
index 5088281e312e..43b2cde80b70 100644
@@ -203,7 +203,23 @@ Description:	Shows total written kbytes issued to disk.
 What:		/sys/fs/f2fs/<disk>/features
 Date:		July 2017
 Contact:	"Jaegeuk Kim" <[email protected]>
-Description:	Shows all enabled features in current device.
+Description:	<deprecated: should use /sys/fs/f2fs/<disk>/feature_list/
+		Shows all enabled features in current device.
+		Supported features:
+		encryption, blkzoned, extra_attr, projquota, inode_checksum,
+		flexible_inline_xattr, quota_ino, inode_crtime, lost_found,
+		verity, sb_checksum, casefold, readonly, compression,
+		encrypted_casefold, pin_file.
+
+What:		/sys/fs/f2fs/<disk>/feature_list/
+Date:		June 2021
+Contact:	"Jaegeuk Kim" <[email protected]>
+Description:	Expand /sys/fs/f2fs/<disk>/features to meet sysfs rule.
+		Supported features:
+		encryption, block_zoned, extra_attr, projquota, inode_checksum,
+		flexible_inline_xattr, quota_ino, inode_crtime, lost_found,
+		verity, sb_checksum, casefold, readonly, compression,
+		encrypted_casefold, pin_file.
 
 What:		/sys/fs/f2fs/<disk>/inject_rate
 Date:		May 2016
fs/f2fs/f2fs.h +3 -0
index 8903c43091f8..17ade71547f1 100644
@@ -1665,6 +1665,9 @@ struct f2fs_sb_info {
 	struct kobject s_stat_kobj;		/* /sys/fs/f2fs/<devname>/stat */
 	struct completion s_stat_kobj_unregister;
 
+	struct kobject s_disk_feat_kobj;		/* /sys/fs/f2fs/<devname>/feature_list */
+	struct completion s_disk_feat_kobj_unregister;
+
 	/* For shrinker support */
 	struct list_head s_list;
 	int s_ndevs;				/* number of devices */
fs/f2fs/sysfs.c +148 -4

228 lines not on this page · raw message

Signature · 1 line
2.32.0.rc1.229.g3e70b5a671-goog
#1 lore raw
Quoted 15 lines · Jaegeuk Kim

Let's create /sys/fs/f2fs/<disk>/feature_list/ to meet sysfs rule.

---
 Documentation/ABI/testing/sysfs-fs-f2fs |  18 ++-
 fs/f2fs/f2fs.h                          |   3 +
 fs/f2fs/sysfs.c                         | 152 +++++++++++++++++++++++-
 3 files changed, 168 insertions(+), 5 deletions(-)
Documentation/ABI/testing/sysfs-fs-f2fs +0 -0
index 5088281e312e..43b2cde80b70 100644
@@ -203,7 +203,23 @@ Description:	Shows total written kbytes issued to disk.
 What:		/sys/fs/f2fs/<disk>/features

Documentation/ABI/testing/sysfs-fs-f2fs Description: Shows total written kbytes issued to disk. · line 212

Quoted 10 lines
Documentation/ABI/testing/sysfs-fs-f2fs +7 -1
 Date:		July 2017
 Contact:	"Jaegeuk Kim" <[email protected]>
-Description:	Shows all enabled features in current device.
+Description:	<deprecated: should use /sys/fs/f2fs/<disk>/feature_list/
+		Shows all enabled features in current device.
+		Supported features:
+		encryption, blkzoned, extra_attr, projquota, inode_checksum,
+		flexible_inline_xattr, quota_ino, inode_crtime, lost_found,
+		verity, sb_checksum, casefold, readonly, compression,
+		encrypted_casefold, pin_file.

Isn't pin_file a feature of the implementation, not of a particular filesystem instance? I.e. something that should go in /sys/fs/f2fs/features/, not here.

Likewise for encrypted_casefold, as it is implied by encryption && casefold.

Documentation/ABI/testing/sysfs-fs-f2fs Description: Shows total written kbytes issued to disk. · line 222

Quoted 10 lines
Documentation/ABI/testing/sysfs-fs-f2fs +10 -0
+
+What:		/sys/fs/f2fs/<disk>/feature_list/
+Date:		June 2021
+Contact:	"Jaegeuk Kim" <[email protected]>
+Description:	Expand /sys/fs/f2fs/<disk>/features to meet sysfs rule.
+		Supported features:
+		encryption, block_zoned, extra_attr, projquota, inode_checksum,
+		flexible_inline_xattr, quota_ino, inode_crtime, lost_found,
+		verity, sb_checksum, casefold, readonly, compression,
+		encrypted_casefold, pin_file.

Is it intentional that the file has "blkzoned" but the directory has "blk_zoned"?

Also, your code has another difference -- "project_quota" is used instead of "projquota". But that's not mentioned above.

And encrypted_casefold and pin_file don't seem appropriate to include here, as mentioned above.

Documentation/ABI/testing/sysfs-fs-f2fs struct f2fs_sb_info { · line 1669

Quoted 6 lines
Documentation/ABI/testing/sysfs-fs-f2fs +2 -0
@@ -1665,6 +1665,9 @@ struct f2fs_sb_info {
 	struct kobject s_stat_kobj;		/* /sys/fs/f2fs/<devname>/stat */
 	struct completion s_stat_kobj_unregister;
 
+	struct kobject s_disk_feat_kobj;		/* /sys/fs/f2fs/<devname>/feature_list */
+	struct completion s_disk_feat_kobj_unregister;

This is for a particular filesystem instance, not a disk per se. (A f2fs filesystem can use multiple disks.) So having "disk" in the name doesn't make sense.

Please use more logical names like s_feature_list_kobj, f2fs_sb_feature_list_ktype, f2fs_sb_feat_attrs, etc.

Quoted 71 lines
Documentation/ABI/testing/sysfs-fs-f2fs +49 -3
 static ssize_t f2fs_feature_show(struct f2fs_attr *a,
 		struct f2fs_sb_info *sbi, char *buf)
 {
+	unsigned long feat_supp = 0;
+
 	switch (a->id) {
 	case FEAT_CRYPTO:
+		feat_supp |= f2fs_sb_has_encrypt(sbi) ?
+					(1 << FEAT_CRYPTO) : 0;
+		fallthrough;
 	case FEAT_BLKZONED:
-	case FEAT_ATOMIC_WRITE:
+		feat_supp |= f2fs_sb_has_blkzoned(sbi) ?
+					(1 << FEAT_BLKZONED) : 0;
+		fallthrough;
 	case FEAT_EXTRA_ATTR:
+		feat_supp |= f2fs_sb_has_extra_attr(sbi) ?
+					(1 << FEAT_EXTRA_ATTR) : 0;
+		fallthrough;
 	case FEAT_PROJECT_QUOTA:
+		feat_supp |= f2fs_sb_has_project_quota(sbi) ?
+					(1 << FEAT_PROJECT_QUOTA) : 0;
+		fallthrough;
 	case FEAT_INODE_CHECKSUM:
+		feat_supp |= f2fs_sb_has_inode_chksum(sbi) ?
+					(1 << FEAT_INODE_CHECKSUM) : 0;
+		fallthrough;
 	case FEAT_FLEXIBLE_INLINE_XATTR:
+		feat_supp |= f2fs_sb_has_flexible_inline_xattr(sbi) ?
+					(1 << FEAT_FLEXIBLE_INLINE_XATTR) : 0;
+		fallthrough;
 	case FEAT_QUOTA_INO:
+		feat_supp |= f2fs_sb_has_quota_ino(sbi) ?
+					(1 << FEAT_QUOTA_INO) : 0;
+		fallthrough;
 	case FEAT_INODE_CRTIME:
+		feat_supp |= f2fs_sb_has_inode_crtime(sbi) ?
+					(1 << FEAT_INODE_CRTIME) : 0;
+		fallthrough;
 	case FEAT_LOST_FOUND:
+		feat_supp |= f2fs_sb_has_lost_found(sbi) ?
+					(1 << FEAT_LOST_FOUND) : 0;
+		fallthrough;
 	case FEAT_VERITY:
+		feat_supp |= f2fs_sb_has_verity(sbi) ?
+					(1 << FEAT_VERITY) : 0;
+		fallthrough;
 	case FEAT_SB_CHECKSUM:
+		feat_supp |= f2fs_sb_has_sb_chksum(sbi) ?
+					(1 << FEAT_SB_CHECKSUM) : 0;
+		fallthrough;
 	case FEAT_CASEFOLD:
+		feat_supp |= f2fs_sb_has_casefold(sbi) ?
+					(1 << FEAT_CASEFOLD) : 0;
+		fallthrough;
 	case FEAT_COMPRESSION:
+		feat_supp |= f2fs_sb_has_compression(sbi) ?
+					(1 << FEAT_COMPRESSION) : 0;
+		fallthrough;
 	case FEAT_RO:
-	case FEAT_TEST_DUMMY_ENCRYPTION_V2:
+		feat_supp |= f2fs_sb_has_readonly(sbi) ?
+					(1 << FEAT_RO) : 0;
+		fallthrough;
 	case FEAT_ENCRYPTED_CASEFOLD:
-		return sprintf(buf, "supported\n");
+		feat_supp |= (f2fs_sb_has_casefold(sbi) &&
+				f2fs_sb_has_encrypt(sbi)) ?
+					(1 << FEAT_ENCRYPTED_CASEFOLD) : 0;
+		fallthrough;
+	case FEAT_PIN_FILE:

Documentation/ABI/testing/sysfs-fs-f2fs struct f2fs_sb_info {

Quoted 10 lines
Documentation/ABI/testing/sysfs-fs-f2fs +7 -1
+		feat_supp |= (1 << FEAT_PIN_FILE);
+		fallthrough;
+	case FEAT_TEST_DUMMY_ENCRYPTION_V2:
+	case FEAT_ATOMIC_WRITE:
+		if (!a->offset || feat_supp & (1 << a->id))
+			return sprintf(buf, "supported\n");
 	}
-	return 0;
+	return sprintf(buf, "not supported\n");
 }

This function doesn't make much sense.

Part of the problem is that the same function is handling both /sys/fs/f2fs/features/ and /sys/fs/f2fs/$s_id/feature_list/.

All the former needs is to print "supported", since unsupported is indicated by the file not being there at all. So it should simply have its own ->show function separate from the one for feature_list/.

And the feature_list/ ones could just store the F2FS_FEATURE_* bit in f2fs_attr::id and check for it using F2FS_HAS_FEATURE(). That would be much simpler -- no need for the feat_id enum or the long switch statement.

Also for feature_list/ it might be better to use "unsupported" than "not supported", so that \<supported\> doesn't match...

Documentation/ABI/testing/sysfs-fs-f2fs

Quoted 6 lines

static struct f2fs_attr f2fs_attr_##_name = { \

Documentation/ABI/testing/sysfs-fs-f2fs +1 -0
 	.attr = {.name = __stringify(_name), .mode = 0444 },	\
 	.show	= f2fs_feature_show,				\
+	.offset	= 0,						\
 	.id	= _id,						\
 }

There's no need to use the .offset argument if features/ and $s_id/feature_list/ just used different ->show functions.

Quoted 1 line

+#define F2FS_DISK_FEATURE_RO_ATTR(_name, _id) \

F2FS_SB_FEATURE_ATTR would be a much better name, since these pertain to a filesystem instance (not necessarily a disk), and all the feature attributes are read-only.

Quoted 20 lines

static int __maybe_unused segment_info_seq_show(struct seq_file *seq,

Documentation/ABI/testing/sysfs-fs-f2fs +9 -0
 						void *offset)
 {
@@ -1149,6 +1279,15 @@ int f2fs_register_sysfs(struct f2fs_sb_info *sbi)
 	if (err)
 		goto put_stat_kobj;
 
+	sbi->s_disk_feat_kobj.kset = &f2fs_kset;
+	init_completion(&sbi->s_disk_feat_kobj_unregister);
+	err = kobject_init_and_add(&sbi->s_disk_feat_kobj,
+						&f2fs_disk_feat_ktype,
+						&sbi->s_kobj, "feature_list");
+	if (err)
+		goto put_stat_kobj;
+
+
 	if (f2fs_proc_root)
 		sbi->s_proc = proc_mkdir(sb->s_id, f2fs_proc_root);
 

Documentation/ABI/testing/sysfs-fs-f2fs int f2fs_register_sysfs(struct f2fs_sb_info *sbi) · line 1309

Quoted 5 lines
Documentation/ABI/testing/sysfs-fs-f2fs +2 -0
@@ -1166,6 +1305,8 @@ int f2fs_register_sysfs(struct f2fs_sb_info *sbi)
 put_stat_kobj:
 	kobject_put(&sbi->s_stat_kobj);
 	wait_for_completion(&sbi->s_stat_kobj_unregister);
+	kobject_put(&sbi->s_disk_feat_kobj);
+	wait_for_completion(&sbi->s_disk_feat_kobj_unregister);

It seems this should go to its own label.

Also, please note that it's very easy to get confused between /sys/fs/f2fs/features/, /sys/fs/f2fs/$s_id/features, and /sys/fs/f2fs/$s_id/feature_list/. Adding some comments could clarify things a lot.

- Eric

author maintainer
#2 lore raw
Quoted 21 lines · Eric Biggers
Quoted 20 lines · Jaegeuk Kim

Let's create /sys/fs/f2fs/<disk>/feature_list/ to meet sysfs rule.

---
 Documentation/ABI/testing/sysfs-fs-f2fs |  18 ++-
 fs/f2fs/f2fs.h                          |   3 +
 fs/f2fs/sysfs.c                         | 152 +++++++++++++++++++++++-
 3 files changed, 168 insertions(+), 5 deletions(-)
Documentation/ABI/testing/sysfs-fs-f2fs +2 -1
index 5088281e312e..43b2cde80b70 100644
@@ -203,7 +203,23 @@ Description:	Shows total written kbytes issued to disk.
 What:		/sys/fs/f2fs/<disk>/features
 Date:		July 2017
 Contact:	"Jaegeuk Kim" <[email protected]>
-Description:	Shows all enabled features in current device.
+Description:	<deprecated: should use /sys/fs/f2fs/<disk>/feature_list/
+		Shows all enabled features in current device.
Quoted 8 lines

Documentation/ABI/testing/sysfs-fs-f2fs Description: Shows total written kbytes issued to disk. · line 212

Quoted 5 lines
Documentation/ABI/testing/sysfs-fs-f2fs +5 -0
+		Supported features:
+		encryption, blkzoned, extra_attr, projquota, inode_checksum,
+		flexible_inline_xattr, quota_ino, inode_crtime, lost_found,
+		verity, sb_checksum, casefold, readonly, compression,
+		encrypted_casefold, pin_file.

Isn't pin_file a feature of the implementation, not of a particular filesystem instance? I.e. something that should go in /sys/fs/f2fs/features/, not here.

Done.

Quoted 2 lines

Likewise for encrypted_casefold, as it is implied by encryption && casefold.

I think encrypted_casefold needs to be in both places:
1) kernel support by kconfig, 2) on-disk support by feature support.

Quoted 14 lines

Documentation/ABI/testing/sysfs-fs-f2fs Description: Shows total written kbytes issued to disk. · line 222

Quoted 11 lines
Documentation/ABI/testing/sysfs-fs-f2fs +10 -0
+
+What:		/sys/fs/f2fs/<disk>/feature_list/
+Date:		June 2021
+Contact:	"Jaegeuk Kim" <[email protected]>
+Description:	Expand /sys/fs/f2fs/<disk>/features to meet sysfs rule.
+		Supported features:
+		encryption, block_zoned, extra_attr, projquota, inode_checksum,
+		flexible_inline_xattr, quota_ino, inode_crtime, lost_found,
+		verity, sb_checksum, casefold, readonly, compression,
+		encrypted_casefold, pin_file.

Is it intentional that the file has "blkzoned" but the directory has "blk_zoned"?

I tried to sync this with /sys/fs/f2fs/features/block_zoned. Let me add a comment in the doc.

Quoted 3 lines

Also, your code has another difference -- "project_quota" is used instead of "projquota". But that's not mentioned above.

Ditto.

Quoted 3 lines

And encrypted_casefold and pin_file don't seem appropriate to include here, as mentioned above.

Done.

Quoted 12 lines

Documentation/ABI/testing/sysfs-fs-f2fs struct f2fs_sb_info { · line 1669

Quoted 7 lines
Documentation/ABI/testing/sysfs-fs-f2fs +2 -0
@@ -1665,6 +1665,9 @@ struct f2fs_sb_info {
 	struct kobject s_stat_kobj;		/* /sys/fs/f2fs/<devname>/stat */
 	struct completion s_stat_kobj_unregister;
 
+	struct kobject s_disk_feat_kobj;		/* /sys/fs/f2fs/<devname>/feature_list */
+	struct completion s_disk_feat_kobj_unregister;

This is for a particular filesystem instance, not a disk per se. (A f2fs filesystem can use multiple disks.) So having "disk" in the name doesn't make sense.

Quoted 2 lines

Please use more logical names like s_feature_list_kobj, f2fs_sb_feature_list_ktype, f2fs_sb_feat_attrs, etc.

Done.

Quoted 92 lines
Quoted 72 lines
Documentation/ABI/testing/sysfs-fs-f2fs +49 -3
 static ssize_t f2fs_feature_show(struct f2fs_attr *a,
 		struct f2fs_sb_info *sbi, char *buf)
 {
+	unsigned long feat_supp = 0;
+
 	switch (a->id) {
 	case FEAT_CRYPTO:
+		feat_supp |= f2fs_sb_has_encrypt(sbi) ?
+					(1 << FEAT_CRYPTO) : 0;
+		fallthrough;
 	case FEAT_BLKZONED:
-	case FEAT_ATOMIC_WRITE:
+		feat_supp |= f2fs_sb_has_blkzoned(sbi) ?
+					(1 << FEAT_BLKZONED) : 0;
+		fallthrough;
 	case FEAT_EXTRA_ATTR:
+		feat_supp |= f2fs_sb_has_extra_attr(sbi) ?
+					(1 << FEAT_EXTRA_ATTR) : 0;
+		fallthrough;
 	case FEAT_PROJECT_QUOTA:
+		feat_supp |= f2fs_sb_has_project_quota(sbi) ?
+					(1 << FEAT_PROJECT_QUOTA) : 0;
+		fallthrough;
 	case FEAT_INODE_CHECKSUM:
+		feat_supp |= f2fs_sb_has_inode_chksum(sbi) ?
+					(1 << FEAT_INODE_CHECKSUM) : 0;
+		fallthrough;
 	case FEAT_FLEXIBLE_INLINE_XATTR:
+		feat_supp |= f2fs_sb_has_flexible_inline_xattr(sbi) ?
+					(1 << FEAT_FLEXIBLE_INLINE_XATTR) : 0;
+		fallthrough;
 	case FEAT_QUOTA_INO:
+		feat_supp |= f2fs_sb_has_quota_ino(sbi) ?
+					(1 << FEAT_QUOTA_INO) : 0;
+		fallthrough;
 	case FEAT_INODE_CRTIME:
+		feat_supp |= f2fs_sb_has_inode_crtime(sbi) ?
+					(1 << FEAT_INODE_CRTIME) : 0;
+		fallthrough;
 	case FEAT_LOST_FOUND:
+		feat_supp |= f2fs_sb_has_lost_found(sbi) ?
+					(1 << FEAT_LOST_FOUND) : 0;
+		fallthrough;
 	case FEAT_VERITY:
+		feat_supp |= f2fs_sb_has_verity(sbi) ?
+					(1 << FEAT_VERITY) : 0;
+		fallthrough;
 	case FEAT_SB_CHECKSUM:
+		feat_supp |= f2fs_sb_has_sb_chksum(sbi) ?
+					(1 << FEAT_SB_CHECKSUM) : 0;
+		fallthrough;
 	case FEAT_CASEFOLD:
+		feat_supp |= f2fs_sb_has_casefold(sbi) ?
+					(1 << FEAT_CASEFOLD) : 0;
+		fallthrough;
 	case FEAT_COMPRESSION:
+		feat_supp |= f2fs_sb_has_compression(sbi) ?
+					(1 << FEAT_COMPRESSION) : 0;
+		fallthrough;
 	case FEAT_RO:
-	case FEAT_TEST_DUMMY_ENCRYPTION_V2:
+		feat_supp |= f2fs_sb_has_readonly(sbi) ?
+					(1 << FEAT_RO) : 0;
+		fallthrough;
 	case FEAT_ENCRYPTED_CASEFOLD:
-		return sprintf(buf, "supported\n");
+		feat_supp |= (f2fs_sb_has_casefold(sbi) &&
+				f2fs_sb_has_encrypt(sbi)) ?
+					(1 << FEAT_ENCRYPTED_CASEFOLD) : 0;
+		fallthrough;
+	case FEAT_PIN_FILE:

Documentation/ABI/testing/sysfs-fs-f2fs struct f2fs_sb_info {

Quoted 10 lines
Documentation/ABI/testing/sysfs-fs-f2fs +7 -1
+		feat_supp |= (1 << FEAT_PIN_FILE);
+		fallthrough;
+	case FEAT_TEST_DUMMY_ENCRYPTION_V2:
+	case FEAT_ATOMIC_WRITE:
+		if (!a->offset || feat_supp & (1 << a->id))
+			return sprintf(buf, "supported\n");
 	}
-	return 0;
+	return sprintf(buf, "not supported\n");
 }

This function doesn't make much sense.

Part of the problem is that the same function is handling both /sys/fs/f2fs/features/ and /sys/fs/f2fs/$s_id/feature_list/.

All the former needs is to print "supported", since unsupported is indicated by the file not being there at all. So it should simply have its own ->show function separate from the one for feature_list/.

Quoted 6 lines

And the feature_list/ ones could just store the F2FS_FEATURE_* bit in f2fs_attr::id and check for it using F2FS_HAS_FEATURE(). That would be much simpler -- no need for the feat_id enum or the long switch statement.

Also for feature_list/ it might be better to use "unsupported" than "not supported", so that \<supported\> doesn't match...

Done.

Quoted 11 lines

Documentation/ABI/testing/sysfs-fs-f2fs

Quoted 7 lines

static struct f2fs_attr f2fs_attr_##_name = { \

Documentation/ABI/testing/sysfs-fs-f2fs +1 -0
 	.attr = {.name = __stringify(_name), .mode = 0444 },	\
 	.show	= f2fs_feature_show,				\
+	.offset	= 0,						\
 	.id	= _id,						\
 }

There's no need to use the .offset argument if features/ and $s_id/feature_list/ just used different ->show functions.

Quoted 5 lines
Quoted 1 line

+#define F2FS_DISK_FEATURE_RO_ATTR(_name, _id) \

F2FS_SB_FEATURE_ATTR would be a much better name, since these pertain to a filesystem instance (not necessarily a disk), and all the feature attributes are read-only.

Used F2FS_SB_FEATURE_RO_ATTR.

Quoted 29 lines
Quoted 21 lines

static int __maybe_unused segment_info_seq_show(struct seq_file *seq,

Documentation/ABI/testing/sysfs-fs-f2fs +9 -0
 						void *offset)
 {
@@ -1149,6 +1279,15 @@ int f2fs_register_sysfs(struct f2fs_sb_info *sbi)
 	if (err)
 		goto put_stat_kobj;
 
+	sbi->s_disk_feat_kobj.kset = &f2fs_kset;
+	init_completion(&sbi->s_disk_feat_kobj_unregister);
+	err = kobject_init_and_add(&sbi->s_disk_feat_kobj,
+						&f2fs_disk_feat_ktype,
+						&sbi->s_kobj, "feature_list");
+	if (err)
+		goto put_stat_kobj;
+
+
 	if (f2fs_proc_root)
 		sbi->s_proc = proc_mkdir(sb->s_id, f2fs_proc_root);
 

Documentation/ABI/testing/sysfs-fs-f2fs int f2fs_register_sysfs(struct f2fs_sb_info *sbi) · line 1309

Quoted 5 lines
Documentation/ABI/testing/sysfs-fs-f2fs +2 -0
@@ -1166,6 +1305,8 @@ int f2fs_register_sysfs(struct f2fs_sb_info *sbi)
 put_stat_kobj:
 	kobject_put(&sbi->s_stat_kobj);
 	wait_for_completion(&sbi->s_stat_kobj_unregister);
+	kobject_put(&sbi->s_disk_feat_kobj);
+	wait_for_completion(&sbi->s_disk_feat_kobj_unregister);

It seems this should go to its own label.

Quoted 4 lines

Also, please note that it's very easy to get confused between /sys/fs/f2fs/features/, /sys/fs/f2fs/$s_id/features, and /sys/fs/f2fs/$s_id/feature_list/. Adding some comments could clarify things a lot.

Done.

Quoted 2 lines

- Eric

author maintainer
#3 lore raw
In reply to Jaegeuk Kim

Let's create /sys/fs/f2fs/<disk>/feature_list/ to meet sysfs rule.

Note that there are three feature list entries:
1) /sys/fs/f2fs/features
: shows runtime features supported by in-kernel f2fs along with Kconfig
- ref. F2FS_FEATURE_RO_ATTR()

2) /sys/fs/f2fs/$s_id/features <deprecated>
: shows on-disk features enabled by mkfs.f2fs, used for old kernels

3) /sys/fs/f2fs/$s_id/feature_list
: shows on-disk features enabled by mkfs.f2fs per instance.
this list covers old feature list provided by 2) and beyond
- ref. F2FS_SB_FEATURE_RO_ATTR()

---
 Documentation/ABI/testing/sysfs-fs-f2fs |  29 +++++-
 fs/f2fs/f2fs.h                          |   5 +
 fs/f2fs/sysfs.c                         | 118 ++++++++++++++++++++++++
 3 files changed, 151 insertions(+), 1 deletion(-)
Documentation/ABI/testing/sysfs-fs-f2fs +28 -1
index 5088281e312e..86d5e0e12a07 100644
@@ -203,7 +203,34 @@ Description:	Shows total written kbytes issued to disk.
 What:		/sys/fs/f2fs/<disk>/features
 Date:		July 2017
 Contact:	"Jaegeuk Kim" <[email protected]>
-Description:	Shows all enabled features in current device.
+Description:	<deprecated: should use /sys/fs/f2fs/<disk>/feature_list/
+		Shows all enabled features in current device.
+		Supported features:
+		encryption, blkzoned, extra_attr, projquota, inode_checksum,
+		flexible_inline_xattr, quota_ino, inode_crtime, lost_found,
+		verity, sb_checksum, casefold, readonly, compression,
+		encrypted_casefold, pin_file.
+
+What:		/sys/fs/f2fs/<disk>/feature_list/
+Date:		June 2021
+Contact:	"Jaegeuk Kim" <[email protected]>
+Description:	Expand /sys/fs/f2fs/<disk>/features to meet sysfs rule.
+		Supported on-disk features:
+		encryption, block_zoned (aka blkzoned), extra_attr,
+		project_quota (aka projquota), inode_checksum,
+		flexible_inline_xattr, quota_ino, inode_crtime, lost_found,
+		verity, sb_checksum, casefold, readonly, compression.
+		Note that, encrypted_casefold and pin_file were moved into
+		/sys/fs/f2fs/features/.
+
+What:		/sys/fs/f2fs/feature/
+Date:		July 2017
+Contact:	"Jaegeuk Kim" <[email protected]>
+Description:	Shows all enabled kernel features.
+		Supported features:
+		in addition to all in /sys/fs/f2fs/<disk>/feature_list/,
+		test_dummy_encryption_v2, encrypted_casefold, pin_file,
+		atomic_write
 
 What:		/sys/fs/f2fs/<disk>/inject_rate
 Date:		May 2016
fs/f2fs/f2fs.h +5 -0
index 8903c43091f8..c8c29effd844 100644
@@ -172,6 +172,8 @@ struct f2fs_mount_info {
 
 #define __F2FS_HAS_FEATURE(raw_super, mask)				\
 	((raw_super->feature & cpu_to_le32(mask)) != 0)
+#define F2FS_MATCH_FEATURE(sbi, mask)					\
+	((sbi->raw_super->feature & cpu_to_le32(mask)) == cpu_to_le32(mask))
 #define F2FS_HAS_FEATURE(sbi, mask)	__F2FS_HAS_FEATURE(sbi->raw_super, mask)
 #define F2FS_SET_FEATURE(sbi, mask)					\
 	(sbi->raw_super->feature |= cpu_to_le32(mask))
@@ -1665,6 +1667,9 @@ struct f2fs_sb_info {
 	struct kobject s_stat_kobj;		/* /sys/fs/f2fs/<devname>/stat */
 	struct completion s_stat_kobj_unregister;
 
+	struct kobject s_feature_list_kobj;		/* /sys/fs/f2fs/<devname>/feature_list */
+	struct completion s_feature_list_kobj_unregister;
+
 	/* For shrinker support */
 	struct list_head s_list;
 	int s_ndevs;				/* number of devices */
fs/f2fs/sysfs.c +118 -0

178 lines not on this page · raw message

Signature · 1 line
2.32.0.rc1.229.g3e70b5a671-goog
#4 lore raw
Quoted 12 lines · Jaegeuk Kim

enum feat_id {

fs/f2fs/sysfs.c +1 -0
 	FEAT_CRYPTO = 0,
 	FEAT_BLKZONED,
@@ -587,6 +601,7 @@ enum feat_id {
 	FEAT_RO,
 	FEAT_TEST_DUMMY_ENCRYPTION_V2,
 	FEAT_ENCRYPTED_CASEFOLD,
+	FEAT_PIN_FILE,
 };
 
 static ssize_t f2fs_feature_show(struct f2fs_attr *a,

fs/f2fs/sysfs.c static ssize_t f2fs_feature_show(struct f2fs_attr *a, · line 631

Quoted 7 lines
fs/f2fs/sysfs.c +1 -0
@@ -610,6 +625,7 @@ static ssize_t f2fs_feature_show(struct f2fs_attr *a,
 	case FEAT_RO:
 	case FEAT_TEST_DUMMY_ENCRYPTION_V2:
 	case FEAT_ENCRYPTED_CASEFOLD:
+	case FEAT_PIN_FILE:
 		return sprintf(buf, "supported\n");
 	}
 	return 0;

There's no need for the feat_id enum to exist. If f2fs_feature_show() just always printed "supported\n", it will do the right thing.

Also, adding pin_file probably should be a separate patch. That seems to be a bug fix, as pin_file was mistakenly added to the per-sb feature list instead of to the kernel feature list?

fs/f2fs/sysfs.c static struct attribute *f2fs_stat_attrs[] = { · line 894

Quoted 7 lines
fs/f2fs/sysfs.c +7 -0
+static ssize_t f2fs_sb_feature_show(struct f2fs_attr *a,
+		struct f2fs_sb_info *sbi, char *buf)
+{
+	if (F2FS_MATCH_FEATURE(sbi, a->id))
+		return sprintf(buf, "supported\n");
+	return sprintf(buf, "unsupported\n");
+}

This can just use F2FS_HAS_FEATURE(), provided that encrypted_casefold isn't included here, which it shouldn't be (as discussed elsewhere).

- Eric

author maintainer
#5 lore raw
Quoted 11 lines · Eric Biggers
Quoted 10 lines · Jaegeuk Kim

enum feat_id {

fs/f2fs/sysfs.c +1 -0
 	FEAT_CRYPTO = 0,
 	FEAT_BLKZONED,
@@ -587,6 +601,7 @@ enum feat_id {
 	FEAT_RO,
 	FEAT_TEST_DUMMY_ENCRYPTION_V2,
 	FEAT_ENCRYPTED_CASEFOLD,
+	FEAT_PIN_FILE,
 };
 
Quoted 12 lines

fs/f2fs/sysfs.c static ssize_t f2fs_feature_show(struct f2fs_attr *a, · line 631

Quoted 9 lines
fs/f2fs/sysfs.c +1 -0
 static ssize_t f2fs_feature_show(struct f2fs_attr *a,
@@ -610,6 +625,7 @@ static ssize_t f2fs_feature_show(struct f2fs_attr *a,
 	case FEAT_RO:
 	case FEAT_TEST_DUMMY_ENCRYPTION_V2:
 	case FEAT_ENCRYPTED_CASEFOLD:
+	case FEAT_PIN_FILE:
 		return sprintf(buf, "supported\n");
 	}
 	return 0;

There's no need for the feat_id enum to exist. If f2fs_feature_show() just always printed "supported\n", it will do the right thing.

Done.

Quoted 4 lines

Also, adding pin_file probably should be a separate patch. That seems to be a bug fix, as pin_file was mistakenly added to the per-sb feature list instead of to the kernel feature list?

Done.

Quoted 11 lines

fs/f2fs/sysfs.c static struct attribute *f2fs_stat_attrs[] = { · line 894

Quoted 8 lines
fs/f2fs/sysfs.c +7 -0
+static ssize_t f2fs_sb_feature_show(struct f2fs_attr *a,
+		struct f2fs_sb_info *sbi, char *buf)
+{
+	if (F2FS_MATCH_FEATURE(sbi, a->id))
+		return sprintf(buf, "supported\n");
+	return sprintf(buf, "unsupported\n");
+}

This can just use F2FS_HAS_FEATURE(), provided that encrypted_casefold isn't included here, which it shouldn't be (as discussed elsewhere).

I know, but I think it'd be good to sync with kernel feature.

Quoted 2 lines

- Eric

author maintainer
#6 lore raw
In reply to Jaegeuk Kim

Let's create /sys/fs/f2fs/<disk>/feature_list/ to meet sysfs rule.

Note that there are three feature list entries:
1) /sys/fs/f2fs/features
: shows runtime features supported by in-kernel f2fs along with Kconfig
- ref. F2FS_FEATURE_RO_ATTR()

2) /sys/fs/f2fs/$s_id/features <deprecated>
: shows on-disk features enabled by mkfs.f2fs, used for old kernels

3) /sys/fs/f2fs/$s_id/feature_list
: shows on-disk features enabled by mkfs.f2fs per instance.
this list covers old feature list provided by 2) and beyond
- ref. F2FS_SB_FEATURE_RO_ATTR()

---
 Documentation/ABI/testing/sysfs-fs-f2fs |  28 +++-
 fs/f2fs/f2fs.h                          |   5 +
 fs/f2fs/sysfs.c                         | 200 ++++++++++++++++--------
 3 files changed, 169 insertions(+), 64 deletions(-)
Documentation/ABI/testing/sysfs-fs-f2fs +27 -1
index 5088281e312e..ca697cc1e765 100644
@@ -203,7 +203,33 @@ Description:	Shows total written kbytes issued to disk.
 What:		/sys/fs/f2fs/<disk>/features
 Date:		July 2017
 Contact:	"Jaegeuk Kim" <[email protected]>
-Description:	Shows all enabled features in current device.
+Description:	<deprecated: should use /sys/fs/f2fs/<disk>/feature_list/
+		Shows all enabled features in current device.
+		Supported features:
+		encryption, blkzoned, extra_attr, projquota, inode_checksum,
+		flexible_inline_xattr, quota_ino, inode_crtime, lost_found,
+		verity, sb_checksum, casefold, readonly, compression,
+		encrypted_casefold, pin_file.
+
+What:		/sys/fs/f2fs/<disk>/feature_list/
+Date:		June 2021
+Contact:	"Jaegeuk Kim" <[email protected]>
+Description:	Expand /sys/fs/f2fs/<disk>/features to meet sysfs rule.
+		Supported on-disk features:
+		encryption, block_zoned (aka blkzoned), extra_attr,
+		project_quota (aka projquota), inode_checksum,
+		flexible_inline_xattr, quota_ino, inode_crtime, lost_found,
+		verity, sb_checksum, casefold, readonly, compression,
+		encrypted_casefold.
+		Note that, pin_file is moved into /sys/fs/f2fs/features/.
+
+What:		/sys/fs/f2fs/feature/
+Date:		July 2017
+Contact:	"Jaegeuk Kim" <[email protected]>
+Description:	Shows all enabled kernel features.
+		Supported features:
+		in addition to all in /sys/fs/f2fs/<disk>/feature_list/,
+		test_dummy_encryption_v2, atomic_write, pin_file
 
 What:		/sys/fs/f2fs/<disk>/inject_rate
 Date:		May 2016
fs/f2fs/f2fs.h +5 -0
index 8903c43091f8..c8c29effd844 100644
@@ -172,6 +172,8 @@ struct f2fs_mount_info {
 
 #define __F2FS_HAS_FEATURE(raw_super, mask)				\
 	((raw_super->feature & cpu_to_le32(mask)) != 0)
+#define F2FS_MATCH_FEATURE(sbi, mask)					\
+	((sbi->raw_super->feature & cpu_to_le32(mask)) == cpu_to_le32(mask))
 #define F2FS_HAS_FEATURE(sbi, mask)	__F2FS_HAS_FEATURE(sbi->raw_super, mask)
 #define F2FS_SET_FEATURE(sbi, mask)					\
 	(sbi->raw_super->feature |= cpu_to_le32(mask))
@@ -1665,6 +1667,9 @@ struct f2fs_sb_info {
 	struct kobject s_stat_kobj;		/* /sys/fs/f2fs/<devname>/stat */
 	struct completion s_stat_kobj_unregister;
 
+	struct kobject s_feature_list_kobj;		/* /sys/fs/f2fs/<devname>/feature_list */
+	struct completion s_feature_list_kobj_unregister;
+
 	/* For shrinker support */
 	struct list_head s_list;
 	int s_ndevs;				/* number of devices */
fs/f2fs/sysfs.c +137 -63

263 lines not on this page · raw message

Signature · 1 line
2.32.0.rc1.229.g3e70b5a671-goog