Skip to contentSkip to conversation

erofs: support idmapped mounts

PATCH Superseded Chao Yu 3 participants 6 messages May 17, 2022 EROFS file system

v1 of 2 Superseded Latest is v2 →

lore
Share & tools
lore

Chao Yu

author maintainer
#0 lore raw

PATCH 1/1 erofs: support idmapped mounts

This patch enables idmapped mounts for erofs, since all dedicated helpers for this functionality existsm, so, in this patch we just pass down the user_namespace argument from the VFS methods to the relevant helpers.

Simple idmap example on erofs image:

1. mkdir dir
2. touch dir/file
3. mkfs.erofs erofs.img dir
4. mount -t erofs -o loop erofs.img /mnt/erofs/

5. ls -ln /mnt/erofs/
total 0
-rw-rw-r-- 1 1000 1000 0 May 17 15:26 file

6. mount-idmapped --map-mount b:0:1001:1 /mnt/erofs/ /mnt/scratch_erofs/

7. ls -ln /mnt/scratch_erofs/
total 0
-rw-rw-r-- 1 65534 65534 0 May 17 15:26 file

---
 fs/erofs/inode.c | 2 +-
 fs/erofs/super.c | 2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)
fs/erofs/inode.c +1 -1
index e8b37ba5e9ad..5320bf52c1ce 100644
@@ -370,7 +370,7 @@ int erofs_getattr(struct user_namespace *mnt_userns, const struct path *path,
 	stat->attributes_mask |= (STATX_ATTR_COMPRESSED |
 				  STATX_ATTR_IMMUTABLE);
 
-	generic_fillattr(&init_user_ns, inode, stat);
+	generic_fillattr(mnt_userns, inode, stat);
 	return 0;
 }
 
fs/erofs/super.c +1 -1
index 0c4b41130c2f..7dc5f2e8ddee 100644
@@ -781,7 +781,7 @@ static struct file_system_type erofs_fs_type = {
 	.name           = "erofs",
 	.init_fs_context = erofs_init_fs_context,
 	.kill_sb        = erofs_kill_sb,
-	.fs_flags       = FS_REQUIRES_DEV,
+	.fs_flags       = FS_REQUIRES_DEV | FS_ALLOW_IDMAP,
 };
 MODULE_ALIAS_FS("erofs");
 
Signature · 1 line
2.25.1
Reviewed-by on v1 1/1
#1 lore raw

Hi Chao,

Quoted 15 lines · Chao Yu

This patch enables idmapped mounts for erofs, since all dedicated helpers for this functionality existsm, so, in this patch we just pass down the user_namespace argument from the VFS methods to the relevant helpers.

Simple idmap example on erofs image:

1. mkdir dir
2. touch dir/file
3. mkfs.erofs erofs.img dir
4. mount -t erofs -o loop erofs.img /mnt/erofs/

5. ls -ln /mnt/erofs/
total 0
-rw-rw-r-- 1 1000 1000 0 May 17 15:26 file

Quoted 7 lines

6. mount-idmapped --map-mount b:0:1001:1 /mnt/erofs/ /mnt/scratch_erofs/

7. ls -ln /mnt/scratch_erofs/
total 0
-rw-rw-r-- 1 65534 65534 0 May 17 15:26 file

Thanks, yeah, I think it's enough to enable idmapped mount for EROFS:

Thanks,
Gao Xiang

Quoted 33 lines
---
 fs/erofs/inode.c | 2 +-
 fs/erofs/super.c | 2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)
fs/erofs/inode.c +1 -1
index e8b37ba5e9ad..5320bf52c1ce 100644
@@ -370,7 +370,7 @@ int erofs_getattr(struct user_namespace *mnt_userns, const struct path *path,
 	stat->attributes_mask |= (STATX_ATTR_COMPRESSED |
 				  STATX_ATTR_IMMUTABLE);
 
-	generic_fillattr(&init_user_ns, inode, stat);
+	generic_fillattr(mnt_userns, inode, stat);
 	return 0;
 }
 
fs/erofs/super.c +1 -1
index 0c4b41130c2f..7dc5f2e8ddee 100644
@@ -781,7 +781,7 @@ static struct file_system_type erofs_fs_type = {
 	.name           = "erofs",
 	.init_fs_context = erofs_init_fs_context,
 	.kill_sb        = erofs_kill_sb,
-	.fs_flags       = FS_REQUIRES_DEV,
+	.fs_flags       = FS_REQUIRES_DEV | FS_ALLOW_IDMAP,
 };
 MODULE_ALIAS_FS("erofs");
 
Signature · 1 line
2.25.1
Reviewed-by on v1 1/1
#2 lore raw
In reply to Chao Yu
Quoted 15 lines · Chao Yu

This patch enables idmapped mounts for erofs, since all dedicated helpers for this functionality existsm, so, in this patch we just pass down the user_namespace argument from the VFS methods to the relevant helpers.

Simple idmap example on erofs image:

1. mkdir dir
2. touch dir/file
3. mkfs.erofs erofs.img dir
4. mount -t erofs -o loop erofs.img /mnt/erofs/

5. ls -ln /mnt/erofs/
total 0
-rw-rw-r-- 1 1000 1000 0 May 17 15:26 file

Quoted 5 lines

6. mount-idmapped --map-mount b:0:1001:1 /mnt/erofs/ /mnt/scratch_erofs/

7. ls -ln /mnt/scratch_erofs/
total 0
-rw-rw-r-- 1 65534 65534 0 May 17 15:26 file

Your current example maps id 0 in the filesystem to id 1001 in the mount. But since no files with id 0 exist in the filesystem you're illustrating that unmapped ids are correctly reported as overflow{g,u}id.

I think what you'd rather want to show is something like this:

5. ls -ln /mnt/erofs/
total 0
-rw-rw-r-- 1 1000 1000 0 May 17 15:26 file

6. mount-idmapped --map-mount b:1000:1001:1 /mnt/erofs/ /mnt/scratch_erofs/

7. ls -ln /mnt/scratch_erofs/
total 0
-rw-rw-r-- 1 1001 1001 0 May 17 15:26 file

where id 1000 in the filesystem maps to id 1001 in the mount.

Quoted 3 lines

---

Overall this is currently the smallest patch to support idmapped mounts.

Is erofs integrated with xfstests in any way?
For read-only filesystems we probably only need to verify that {g,u}id are correctly reported. All the writable aspects are irrelevant.

Looks good,

Quoted 33 lines
 fs/erofs/inode.c | 2 +-
 fs/erofs/super.c | 2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)
fs/erofs/inode.c +1 -1
index e8b37ba5e9ad..5320bf52c1ce 100644
@@ -370,7 +370,7 @@ int erofs_getattr(struct user_namespace *mnt_userns, const struct path *path,
 	stat->attributes_mask |= (STATX_ATTR_COMPRESSED |
 				  STATX_ATTR_IMMUTABLE);
 
-	generic_fillattr(&init_user_ns, inode, stat);
+	generic_fillattr(mnt_userns, inode, stat);
 	return 0;
 }
 
fs/erofs/super.c +1 -1
index 0c4b41130c2f..7dc5f2e8ddee 100644
@@ -781,7 +781,7 @@ static struct file_system_type erofs_fs_type = {
 	.name           = "erofs",
 	.init_fs_context = erofs_init_fs_context,
 	.kill_sb        = erofs_kill_sb,
-	.fs_flags       = FS_REQUIRES_DEV,
+	.fs_flags       = FS_REQUIRES_DEV | FS_ALLOW_IDMAP,
 };
 MODULE_ALIAS_FS("erofs");
 
Signature · 1 line
2.25.1
#3 lore raw

Hi Christian,

Quoted 41 lines · Christian Brauner
Quoted 15 lines · Chao Yu

This patch enables idmapped mounts for erofs, since all dedicated helpers for this functionality existsm, so, in this patch we just pass down the user_namespace argument from the VFS methods to the relevant helpers.

Simple idmap example on erofs image:

1. mkdir dir
2. touch dir/file
3. mkfs.erofs erofs.img dir
4. mount -t erofs -o loop erofs.img /mnt/erofs/

5. ls -ln /mnt/erofs/
total 0
-rw-rw-r-- 1 1000 1000 0 May 17 15:26 file

Quoted 5 lines

6. mount-idmapped --map-mount b:0:1001:1 /mnt/erofs/ /mnt/scratch_erofs/

7. ls -ln /mnt/scratch_erofs/
total 0
-rw-rw-r-- 1 65534 65534 0 May 17 15:26 file

Your current example maps id 0 in the filesystem to id 1001 in the mount. But since no files with id 0 exist in the filesystem you're illustrating that unmapped ids are correctly reported as overflow{g,u}id.

I think what you'd rather want to show is something like this:

5. ls -ln /mnt/erofs/
total 0
-rw-rw-r-- 1 1000 1000 0 May 17 15:26 file

6. mount-idmapped --map-mount b:1000:1001:1 /mnt/erofs/ /mnt/scratch_erofs/

7. ls -ln /mnt/scratch_erofs/
total 0
-rw-rw-r-- 1 1001 1001 0 May 17 15:26 file

where id 1000 in the filesystem maps to id 1001 in the mount.

Quoted 1 line
Quoted 8 lines
Quoted 2 lines

---

Overall this is currently the smallest patch to support idmapped mounts.

Is erofs integrated with xfstests in any way?
For read-only filesystems we probably only need to verify that {g,u}id are correctly reported. All the writable aspects are irrelevant.

Currently most generic xfstests test cases are unsuitable for erofs.

Instead we have regression testcases for EROFS specific since it needs to generate images with care,
https://git.kernel.org/pub/scm/linux/kernel/git/xiang/erofs-utils.git/log/?h=experimental-tests

Also we have an erofsstress to do long time random stress workloads,
https://github.com/erofs/erofsstress

But yeah, it's some awkward that fstests idmapped mount testcases may be unsuitable for EROFS for now. I will add some new testcases to build images and test for this behavior.

Quoted 3 lines

Looks good,

Thanks for your review!

Thanks,
Gao Xiang

Quoted 34 lines
Quoted 34 lines
 fs/erofs/inode.c | 2 +-
 fs/erofs/super.c | 2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)
fs/erofs/inode.c +1 -1
index e8b37ba5e9ad..5320bf52c1ce 100644
@@ -370,7 +370,7 @@ int erofs_getattr(struct user_namespace *mnt_userns, const struct path *path,
 	stat->attributes_mask |= (STATX_ATTR_COMPRESSED |
 				  STATX_ATTR_IMMUTABLE);
 
-	generic_fillattr(&init_user_ns, inode, stat);
+	generic_fillattr(mnt_userns, inode, stat);
 	return 0;
 }
 
fs/erofs/super.c +1 -1
index 0c4b41130c2f..7dc5f2e8ddee 100644
@@ -781,7 +781,7 @@ static struct file_system_type erofs_fs_type = {
 	.name           = "erofs",
 	.init_fs_context = erofs_init_fs_context,
 	.kill_sb        = erofs_kill_sb,
-	.fs_flags       = FS_REQUIRES_DEV,
+	.fs_flags       = FS_REQUIRES_DEV | FS_ALLOW_IDMAP,
 };
 MODULE_ALIAS_FS("erofs");
 
Signature · 1 line
2.25.1
#4 lore raw
Quoted 67 lines · Gao Xiang

Hi Christian,

Quoted 41 lines · Christian Brauner
Quoted 15 lines · Chao Yu

This patch enables idmapped mounts for erofs, since all dedicated helpers for this functionality existsm, so, in this patch we just pass down the user_namespace argument from the VFS methods to the relevant helpers.

Simple idmap example on erofs image:

1. mkdir dir
2. touch dir/file
3. mkfs.erofs erofs.img dir
4. mount -t erofs -o loop erofs.img /mnt/erofs/

5. ls -ln /mnt/erofs/
total 0
-rw-rw-r-- 1 1000 1000 0 May 17 15:26 file

Quoted 5 lines

6. mount-idmapped --map-mount b:0:1001:1 /mnt/erofs/ /mnt/scratch_erofs/

7. ls -ln /mnt/scratch_erofs/
total 0
-rw-rw-r-- 1 65534 65534 0 May 17 15:26 file

Your current example maps id 0 in the filesystem to id 1001 in the mount. But since no files with id 0 exist in the filesystem you're illustrating that unmapped ids are correctly reported as overflow{g,u}id.

I think what you'd rather want to show is something like this:

5. ls -ln /mnt/erofs/
total 0
-rw-rw-r-- 1 1000 1000 0 May 17 15:26 file

6. mount-idmapped --map-mount b:1000:1001:1 /mnt/erofs/ /mnt/scratch_erofs/

7. ls -ln /mnt/scratch_erofs/
total 0
-rw-rw-r-- 1 1001 1001 0 May 17 15:26 file

where id 1000 in the filesystem maps to id 1001 in the mount.

Quoted 1 line
Quoted 8 lines
Quoted 2 lines

---

Overall this is currently the smallest patch to support idmapped mounts.

Is erofs integrated with xfstests in any way?
For read-only filesystems we probably only need to verify that {g,u}id are correctly reported. All the writable aspects are irrelevant.

Currently most generic xfstests test cases are unsuitable for erofs.

Instead we have regression testcases for EROFS specific since it needs to generate images with care,
https://git.kernel.org/pub/scm/linux/kernel/git/xiang/erofs-utils.git/log/?h=experimental-tests

Also we have an erofsstress to do long time random stress workloads,
https://github.com/erofs/erofsstress

But yeah, it's some awkward that fstests idmapped mount testcases may be unsuitable for EROFS for now. I will add some new testcases to build images and test for this behavior.

Quoted 1 line
Quoted 4 lines
Quoted 2 lines

Looks good,

Thanks for your review!

Thanks for supporting this in erofs!
Christian

#5 lore raw
Quoted 36 lines · Christian Brauner
Quoted 35 lines · Gao Xiang

Hi Christian,

Quoted 32 lines · Christian Brauner
Quoted 15 lines · Chao Yu

This patch enables idmapped mounts for erofs, since all dedicated helpers for this functionality existsm, so, in this patch we just pass down the user_namespace argument from the VFS methods to the relevant helpers.

Simple idmap example on erofs image:

1. mkdir dir
2. touch dir/file
3. mkfs.erofs erofs.img dir
4. mount -t erofs -o loop erofs.img /mnt/erofs/

5. ls -ln /mnt/erofs/
total 0
-rw-rw-r-- 1 1000 1000 0 May 17 15:26 file

Quoted 5 lines

6. mount-idmapped --map-mount b:0:1001:1 /mnt/erofs/ /mnt/scratch_erofs/

7. ls -ln /mnt/scratch_erofs/
total 0
-rw-rw-r-- 1 65534 65534 0 May 17 15:26 file

Your current example maps id 0 in the filesystem to id 1001 in the mount. But since no files with id 0 exist in the filesystem you're illustrating that unmapped ids are correctly reported as overflow{g,u}id.

I think what you'd rather want to show is something like this:

5. ls -ln /mnt/erofs/
total 0
-rw-rw-r-- 1 1000 1000 0 May 17 15:26 file

Quoted 7 lines
Quoted 7 lines
Quoted 7 lines

6. mount-idmapped --map-mount b:1000:1001:1 /mnt/erofs/ /mnt/scratch_erofs/

7. ls -ln /mnt/scratch_erofs/
total 0
-rw-rw-r-- 1 1001 1001 0 May 17 15:26 file

where id 1000 in the filesystem maps to id 1001 in the mount.

Yeah, I just manually tested, although some steps assume user 1000 and some steps assume the root user. But it works.

I will rephrase such commit messages when applying later...

Thanks,
Gao Xiang