Re: Upcoming: Notifications, FS notifications and fsinfo()

From: Miklos Szeredi
Date: Mon Apr 06 2020 - 04:36:13 EST


On Fri, Apr 3, 2020 at 10:30 PM J. Bruce Fields <bfields@xxxxxxxxxxxx> wrote:
>
> On Fri, Apr 03, 2020 at 05:12:23PM +0200, Lennart Poettering wrote:
> > BTW, while we are at it: one more thing I'd love to see exposed by
> > statx() is a simple flag whether the inode is a mount point. There's
> > plenty code that implements a test like this all over the place, and
> > it usually isn't very safe. There's one implementation in util-linux
> > for example (in the /usr/bin/mountpoint binary), and another one in
> > systemd. Would be awesome to just have a statx() return flag for that,
> > that would make things *so* much easier and more robust. because in
> > fact most code isn't very good that implements this, as much of it
> > just compares st_dev of the specified file and its parent. Better code
> > compares the mount ID, but as mentioned that's not as pretty as it
> > could be so far...
>
> nfs-utils/support/misc/mountpoint.c:check_is_mountpoint() stats the file
> and ".." and returns true if they have different st_dev or the same
> st_ino. Comparing mount ids sounds better.
>
> So anyway, yes, everybody reinvents the wheel here, and this would be
> useful. (And, yes, we want to know for the vfsmount, we don't care
> whether the same inode is used as a mountpoint someplace else.)

Attaching a patch.

There's some ambiguity about what is a "mountpoint" and what these
tools are interested in. My guess is that they are not interested in
an object being a mount point (something where another object is
mounted) but being a mount root (this is the object mounted at the
mount point). I.e

fd = open("/mnt", O_PATH);
mount("/bin", "/mnt", NULL, MS_BIND, NULL);
statx(AT_FDCWD, "/mnt", 0, 0, &stx1);
statx(fd, "", AT_EMPTY_PATH, 0, &stx2);
printf("mount_root(/mnt) = %c, mount_root(fd) = %c\n",
stx1.stx_attributes & STATX_ATTR_MOUNT_ROOT ? 'y' : 'n',
stx2.stx_attributes & STATX_ATTR_MOUNT_ROOT ? 'y' : 'n');

Would print:
mount_root(/mnt) = y, mount_root(fd) = n

Thanks,
Miklos
From: Miklos Szeredi <mszeredi@xxxxxxxxxx>
Subject: statx: add mount_root

Determining whether a path or file descriptor refers to a mountpoint (or
more precisely a mount root) is not trivial using current tools.

Add a flag to statx that indicates whether the path or fd refers to the
root of a mount or not.

Reported-by: Lennart Poettering <mzxreary@xxxxxxxxxxx>
Reported-by: J. Bruce Fields <bfields@xxxxxxxxxxxx>
Signed-off-by: Miklos Szeredi <mszeredi@xxxxxxxxxx>
---
fs/stat.c | 3 +++
include/uapi/linux/stat.h | 1 +
2 files changed, 4 insertions(+)

--- a/include/uapi/linux/stat.h
+++ b/include/uapi/linux/stat.h
@@ -172,6 +172,7 @@ struct statx {
#define STATX_ATTR_NODUMP 0x00000040 /* [I] File is not to be dumped */
#define STATX_ATTR_ENCRYPTED 0x00000800 /* [I] File requires key to decrypt in fs */
#define STATX_ATTR_AUTOMOUNT 0x00001000 /* Dir: Automount trigger */
+#define STATX_ATTR_MOUNT_ROOT 0x00002000 /* Root of a mount */
#define STATX_ATTR_VERITY 0x00100000 /* [I] Verity protected file */


--- a/fs/stat.c
+++ b/fs/stat.c
@@ -202,6 +202,9 @@ int vfs_statx(int dfd, const char __user
error = vfs_getattr(&path, stat, request_mask, flags);
stat->mnt_id = real_mount(path.mnt)->mnt_id;
stat->result_mask |= STATX_MNT_ID;
+ if (path.mnt->mnt_root == path.dentry)
+ stat->attributes |= STATX_ATTR_MOUNT_ROOT;
+ stat->attributes_mask |= STATX_ATTR_MOUNT_ROOT;
path_put(&path);
if (retry_estale(error, lookup_flags)) {
lookup_flags |= LOOKUP_REVAL;