[PATCH 1/2] fs/super.c: introduce get_tree_bdev_by_dev()

[PATCH 1/2] fs/super.c: introduce get_tree_bdev_by_dev()

[Gao Xiang]

As Allison reported [1], currently get_tree_bdev() will store
"Can't lookup blockdev" error message.  Although it makes sense for
pure bdev-based fses, this message may mislead users who try to use
EROFS file-backed mounts since get_tree_nodev() is used as a fallback
then.

Add get_tree_bdev_by_dev() to specify a device number explicitly
instead of the hardcoded fc->source as mentioned in [2], there are
other benefits like:
  - Filesystems can have other ways to get a bdev-based sb
    in addition to the current hard-coded source path;

  - Pseudo-filesystems can utilize this method to generate a
    filesystem from given device numbers too.

  - Like get_tree_nodev(), it doesn't strictly tie to fc->source
    either.

[1] https://lore.kernel.org/r/CAOYeF9VQ8jKVmpy5Zy9DNhO6xmWSKMB-DO8yvBB0XvBE7=3Ugg@mail.gmail.com
[2] https://lore.kernel.org/r/b9565874-7018-46ef-b123-b524a1dffb21@linux.alibaba.com
Signed-off-by: Gao Xiang <hsiangkao@linux.alibaba.com>
---
 fs/super.c                 | 41 ++++++++++++++++++++++++++------------
 include/linux/fs_context.h |  3 +++
 2 files changed, 31 insertions(+), 13 deletions(-)

diff --git a/fs/super.c b/fs/super.c
index 1db230432960..8cc8350b9ba6 100644
--- a/fs/super.c
+++ b/fs/super.c
@@ -1596,26 +1596,17 @@ int setup_bdev_super(struct super_block *sb, int sb_flags,
 EXPORT_SYMBOL_GPL(setup_bdev_super);
 
 /**
- * get_tree_bdev - Get a superblock based on a single block device
+ * get_tree_bdev_by_dev - Get a bdev-based superblock with a given device number
  * @fc: The filesystem context holding the parameters
  * @fill_super: Helper to initialise a new superblock
+ * @dev: The device number indicating the target block device
  */
-int get_tree_bdev(struct fs_context *fc,
+int get_tree_bdev_by_dev(struct fs_context *fc,
 		int (*fill_super)(struct super_block *,
-				  struct fs_context *))
+				  struct fs_context *), dev_t dev)
 {
 	struct super_block *s;
 	int error = 0;
-	dev_t dev;
-
-	if (!fc->source)
-		return invalf(fc, "No source specified");
-
-	error = lookup_bdev(fc->source, &dev);
-	if (error) {
-		errorf(fc, "%s: Can't lookup blockdev", fc->source);
-		return error;
-	}
 
 	fc->sb_flags |= SB_NOSEC;
 	s = sget_dev(fc, dev);
@@ -1644,6 +1635,30 @@ int get_tree_bdev(struct fs_context *fc,
 	fc->root = dget(s->s_root);
 	return 0;
 }
+EXPORT_SYMBOL_GPL(get_tree_bdev_by_dev);
+
+/**
+ * get_tree_bdev - Get a superblock based on a single block device
+ * @fc: The filesystem context holding the parameters
+ * @fill_super: Helper to initialise a new superblock
+ */
+int get_tree_bdev(struct fs_context *fc,
+		int (*fill_super)(struct super_block *,
+				  struct fs_context *))
+{
+	int error;
+	dev_t dev;
+
+	if (!fc->source)
+		return invalf(fc, "No source specified");
+
+	error = lookup_bdev(fc->source, &dev);
+	if (error) {
+		errorf(fc, "%s: Can't lookup blockdev", fc->source);
+		return error;
+	}
+	return get_tree_bdev_by_dev(fc, fill_super, dev);
+}
 EXPORT_SYMBOL(get_tree_bdev);
 
 static int test_bdev_super(struct super_block *s, void *data)
diff --git a/include/linux/fs_context.h b/include/linux/fs_context.h
index c13e99cbbf81..54f23589ad5b 100644
--- a/include/linux/fs_context.h
+++ b/include/linux/fs_context.h
@@ -160,6 +160,9 @@ extern int get_tree_keyed(struct fs_context *fc,
 
 int setup_bdev_super(struct super_block *sb, int sb_flags,
 		struct fs_context *fc);
+int get_tree_bdev_by_dev(struct fs_context *fc,
+			 int (*fill_super)(struct super_block *sb,
+					   struct fs_context *fc), dev_t dev);
 extern int get_tree_bdev(struct fs_context *fc,
 			       int (*fill_super)(struct super_block *sb,
 						 struct fs_context *fc));
-- 
2.43.5

[PATCH 2/2] erofs: use get_tree_bdev_by_dev() to avoid misleading messages

[Gao Xiang]

Users can pass in an arbitrary source path for the proper type of
a mount then without "Can't lookup blockdev" error message.

Reported-by: Allison Karlitskaya <allison.karlitskaya@redhat.com>
Closes: https://lore.kernel.org/r/CAOYeF9VQ8jKVmpy5Zy9DNhO6xmWSKMB-DO8yvBB0XvBE7=3Ugg@mail.gmail.com
Signed-off-by: Gao Xiang <hsiangkao@linux.alibaba.com>
---
 fs/erofs/super.c | 9 ++++++---
 1 file changed, 6 insertions(+), 3 deletions(-)

diff --git a/fs/erofs/super.c b/fs/erofs/super.c
index 666873f745da..04a5873c1594 100644
--- a/fs/erofs/super.c
+++ b/fs/erofs/super.c
@@ -700,16 +700,19 @@ static int erofs_fc_fill_super(struct super_block *sb, struct fs_context *fc)
 static int erofs_fc_get_tree(struct fs_context *fc)
 {
 	struct erofs_sb_info *sbi = fc->s_fs_info;
+	dev_t dev;
 	int ret;
 
 	if (IS_ENABLED(CONFIG_EROFS_FS_ONDEMAND) && sbi->fsid)
 		return get_tree_nodev(fc, erofs_fc_fill_super);
 
-	ret = get_tree_bdev(fc, erofs_fc_fill_super);
+	if (!fc->source)
+		return invalf(fc, "No source specified");
+	ret = lookup_bdev(fc->source, &dev);
+	if (!ret)
+		return get_tree_bdev_by_dev(fc, erofs_fc_fill_super, dev);
 #ifdef CONFIG_EROFS_FS_BACKED_BY_FILE
 	if (ret == -ENOTBLK) {
-		if (!fc->source)
-			return invalf(fc, "No source specified");
 		sbi->fdev = filp_open(fc->source, O_RDONLY | O_LARGEFILE, 0);
 		if (IS_ERR(sbi->fdev))
 			return PTR_ERR(sbi->fdev);
-- 
2.43.5

Re: [PATCH 1/2] fs/super.c: introduce get_tree_bdev_by_dev()

[Christoph Hellwig]

On Tue, Oct 08, 2024 at 05:56:05PM +0800, Gao Xiang wrote:
> As Allison reported [1], currently get_tree_bdev() will store
> "Can't lookup blockdev" error message.  Although it makes sense for
> pure bdev-based fses, this message may mislead users who try to use
> EROFS file-backed mounts since get_tree_nodev() is used as a fallback
> then.
> 
> Add get_tree_bdev_by_dev() to specify a device number explicitly
> instead of the hardcoded fc->source as mentioned in [2], there are
> other benefits like:
>   - Filesystems can have other ways to get a bdev-based sb
>     in addition to the current hard-coded source path;
> 
>   - Pseudo-filesystems can utilize this method to generate a
>     filesystem from given device numbers too.
> 
>   - Like get_tree_nodev(), it doesn't strictly tie to fc->source
>     either.

Do you have concrete plans for any of those?  If so send pointers.
Otherwise just passing a quiet flag of some form feels like a much
saner interface.

Re: [PATCH 1/2] fs/super.c: introduce get_tree_bdev_by_dev()

[Gao Xiang]

Hi Christoph,

On 2024/10/8 19:49, Christoph Hellwig wrote:
> On Tue, Oct 08, 2024 at 05:56:05PM +0800, Gao Xiang wrote:
>> As Allison reported [1], currently get_tree_bdev() will store
>> "Can't lookup blockdev" error message.  Although it makes sense for
>> pure bdev-based fses, this message may mislead users who try to use
>> EROFS file-backed mounts since get_tree_nodev() is used as a fallback
>> then.
>>
>> Add get_tree_bdev_by_dev() to specify a device number explicitly
>> instead of the hardcoded fc->source as mentioned in [2], there are
>> other benefits like:
>>    - Filesystems can have other ways to get a bdev-based sb
>>      in addition to the current hard-coded source path;
>>
>>    - Pseudo-filesystems can utilize this method to generate a
>>      filesystem from given device numbers too.
>>
>>    - Like get_tree_nodev(), it doesn't strictly tie to fc->source
>>      either.
> 
> Do you have concrete plans for any of those?  If so send pointers.

Thanks for your comment.

I don't have any pointer for now, just listing the potential use
cases for reference.

But the error message out of get_tree_bdev() is inflexible and
IMHO it's too coupled to `fc->source`.

> Otherwise just passing a quiet flag of some form feels like a much
> saner interface.

I'm fine with this way, but that will be a treewide change, I
will send out a version with a flag later.

Thanks,
Gao Xiang

Re: [PATCH 1/2] fs/super.c: introduce get_tree_bdev_by_dev()

[Christoph Hellwig]

On Tue, Oct 08, 2024 at 08:10:45PM +0800, Gao Xiang wrote:
> But the error message out of get_tree_bdev() is inflexible and
> IMHO it's too coupled to `fc->source`.
> 
> > Otherwise just passing a quiet flag of some form feels like a much
> > saner interface.
> 
> I'm fine with this way, but that will be a treewide change, I
> will send out a version with a flag later.

I'd probably just add a get_tree_bdev_flags and pass 0 flags from
get_tree_bdev.

Re: [PATCH 1/2] fs/super.c: introduce get_tree_bdev_by_dev()

[Gao Xiang]

On 2024/10/8 20:23, Christoph Hellwig wrote:
> On Tue, Oct 08, 2024 at 08:10:45PM +0800, Gao Xiang wrote:
>> But the error message out of get_tree_bdev() is inflexible and
>> IMHO it's too coupled to `fc->source`.
>>
>>> Otherwise just passing a quiet flag of some form feels like a much
>>> saner interface.
>>
>> I'm fine with this way, but that will be a treewide change, I
>> will send out a version with a flag later.
> 
> I'd probably just add a get_tree_bdev_flags and pass 0 flags from
> get_tree_bdev.

how about
int get_tree_bdev_flags(struct fs_context *fc,
		int (*fill_super)(struct super_block *,
				  struct fs_context *), bool quiet)

for now? it can be turned into `int flags` if other needs
are shown later (and I don't need to define an enum.)

Does it look good to you? if okay, I will follow this way.

Thanks,
Gao Xiang

Re: [PATCH 1/2] fs/super.c: introduce get_tree_bdev_by_dev()

[Christoph Hellwig]

On Tue, Oct 08, 2024 at 08:33:27PM +0800, Gao Xiang wrote:
> how about
> int get_tree_bdev_flags(struct fs_context *fc,
> 		int (*fill_super)(struct super_block *,
> 				  struct fs_context *), bool quiet)
> 
> for now? it can be turned into `int flags` if other needs
> are shown later (and I don't need to define an enum.)

I'd pass an unsigned int flags with a clearly spellt out (and
extensible) flags namespae.

Re: [PATCH 1/2] fs/super.c: introduce get_tree_bdev_by_dev()

[Gao Xiang]

On 2024/10/8 20:55, Christoph Hellwig wrote:
> On Tue, Oct 08, 2024 at 08:33:27PM +0800, Gao Xiang wrote:
>> how about
>> int get_tree_bdev_flags(struct fs_context *fc,
>> 		int (*fill_super)(struct super_block *,
>> 				  struct fs_context *), bool quiet)
>>
>> for now? it can be turned into `int flags` if other needs
>> are shown later (and I don't need to define an enum.)
> 
> I'd pass an unsigned int flags with a clearly spellt out (and
> extensible) flags namespae.

okay, got it.

Thanks,
Gao Xiang