diff mbox

[v2] vfs: use "none" if mount source is empty string

Message ID 1524109641-45617-1-git-send-email-cofyc.jackson@gmail.com (mailing list archive)
State New, archived
Headers show

Commit Message

Yecheng Fu April 19, 2018, 3:47 a.m. UTC
`libmount` from util-linux and many softwares in userspace (e.g.
kubelet) did not expect empty string as mount source:

```
$ mount -t tmpfs "" /mnt/tmpfs
$ findmnt /mnt/tmpfs
findmnt: /proc/self/mountinfo: parse error at line 51
$ cat /proc/self/mountinfo | grep -P '\/mnt\/tmpfs'
74 25 0:59 / /mnt/tmpfs rw,relatime shared:38 - tmpfs  rw
$ cat /proc/self/mounts | grep -P '\/mnt\/tmpfs'
 /mnt/tmpfs tmpfs rw,relatime 0 0
```

`source` field in mounts/mountinfo is empty, which breaks a lot of
mounts/mountinfo parsers.

This fixes issues in parsing when user uses empty string as mount
source.

Cc: Karel Zak <kzak@redhat.com>
Signed-off-by: Yecheng Fu <cofyc.jackson@gmail.com>
---

PATCH v2 updates `show_vfsstat()` too.

 fs/proc_namespace.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

Comments

Matthew Wilcox April 19, 2018, 11:32 a.m. UTC | #1
On Thu, Apr 19, 2018 at 11:47:21AM +0800, Yecheng Fu wrote:
> `libmount` from util-linux and many softwares in userspace (e.g.
> kubelet) did not expect empty string as mount source:
> 
> ```
> $ mount -t tmpfs "" /mnt/tmpfs
> $ findmnt /mnt/tmpfs
> findmnt: /proc/self/mountinfo: parse error at line 51
> $ cat /proc/self/mountinfo | grep -P '\/mnt\/tmpfs'
> 74 25 0:59 / /mnt/tmpfs rw,relatime shared:38 - tmpfs  rw
> $ cat /proc/self/mounts | grep -P '\/mnt\/tmpfs'
>  /mnt/tmpfs tmpfs rw,relatime 0 0
> ```
> 
> `source` field in mounts/mountinfo is empty, which breaks a lot of
> mounts/mountinfo parsers.
> 
> This fixes issues in parsing when user uses empty string as mount
> source.
> 
> Cc: Karel Zak <kzak@redhat.com>
> Signed-off-by: Yecheng Fu <cofyc.jackson@gmail.com>

Reviewed-by: Matthew Wilcox <mawilcox@microsoft.com>
Yecheng Fu May 15, 2018, 5:18 a.m. UTC | #2
On Thu, Apr 19, 2018 at 04:32:56AM -0700, Matthew Wilcox wrote:
> On Thu, Apr 19, 2018 at 11:47:21AM +0800, Yecheng Fu wrote:
> > `libmount` from util-linux and many softwares in userspace (e.g.
> > kubelet) did not expect empty string as mount source:
> > 
> > ```
> > $ mount -t tmpfs "" /mnt/tmpfs
> > $ findmnt /mnt/tmpfs
> > findmnt: /proc/self/mountinfo: parse error at line 51
> > $ cat /proc/self/mountinfo | grep -P '\/mnt\/tmpfs'
> > 74 25 0:59 / /mnt/tmpfs rw,relatime shared:38 - tmpfs  rw
> > $ cat /proc/self/mounts | grep -P '\/mnt\/tmpfs'
> >  /mnt/tmpfs tmpfs rw,relatime 0 0
> > ```
> > 
> > `source` field in mounts/mountinfo is empty, which breaks a lot of
> > mounts/mountinfo parsers.
> > 
> > This fixes issues in parsing when user uses empty string as mount
> > source.
> > 
> > Cc: Karel Zak <kzak@redhat.com>
> > Signed-off-by: Yecheng Fu <cofyc.jackson@gmail.com>
> 
> Reviewed-by: Matthew Wilcox <mawilcox@microsoft.com>

hi, I'm a newbie here. Will this be merged or is there still something I
need to do?

I had submitted a patch to [util-linux](https://github.com/karelzak/util-linux/pull/619). Karel prefer kernel side bugfix than rewrite all the sscanf() stuff and I agree. Kernel side bugfix also fixes all parsers too.
Yecheng Fu July 12, 2018, 2:29 a.m. UTC | #3
On Tue, May 15, 2018 at 01:18:00PM +0800, Yecheng Fu wrote:
> On Thu, Apr 19, 2018 at 04:32:56AM -0700, Matthew Wilcox wrote:
> > On Thu, Apr 19, 2018 at 11:47:21AM +0800, Yecheng Fu wrote:
> > > `libmount` from util-linux and many softwares in userspace (e.g.
> > > kubelet) did not expect empty string as mount source:
> > > 
> > > ```
> > > $ mount -t tmpfs "" /mnt/tmpfs
> > > $ findmnt /mnt/tmpfs
> > > findmnt: /proc/self/mountinfo: parse error at line 51
> > > $ cat /proc/self/mountinfo | grep -P '\/mnt\/tmpfs'
> > > 74 25 0:59 / /mnt/tmpfs rw,relatime shared:38 - tmpfs  rw
> > > $ cat /proc/self/mounts | grep -P '\/mnt\/tmpfs'
> > >  /mnt/tmpfs tmpfs rw,relatime 0 0
> > > ```
> > > 
> > > `source` field in mounts/mountinfo is empty, which breaks a lot of
> > > mounts/mountinfo parsers.
> > > 
> > > This fixes issues in parsing when user uses empty string as mount
> > > source.
> > > 
> > > Cc: Karel Zak <kzak@redhat.com>
> > > Signed-off-by: Yecheng Fu <cofyc.jackson@gmail.com>
> > 
> > Reviewed-by: Matthew Wilcox <mawilcox@microsoft.com>
> 
> hi, I'm a newbie here. Will this be merged or is there still something I
> need to do?
> 
> I had submitted a patch to [util-linux](https://github.com/karelzak/util-linux/pull/619). Karel prefer kernel side bugfix than rewrite all the sscanf() stuff and I agree. Kernel side bugfix also fixes all parsers too.
> 
> -- 
> Yecheng Fu

hi, didn't receive any updates for months, sorry to ping again in case
someone missed it. 

Karel patched util-linux to work around this issue now [^1], but I hope this
can be fixed on kernel side too to improve compatibility.

[^1]: https://github.com/karelzak/util-linux/commit/18a52a5094f820b5da013daf5972eb8e65be9680
Karel Zak July 12, 2018, 8:43 a.m. UTC | #4
On Thu, Jul 12, 2018 at 10:29:50AM +0800, Yecheng Fu wrote:
> On Tue, May 15, 2018 at 01:18:00PM +0800, Yecheng Fu wrote:
> > On Thu, Apr 19, 2018 at 04:32:56AM -0700, Matthew Wilcox wrote:
> > > On Thu, Apr 19, 2018 at 11:47:21AM +0800, Yecheng Fu wrote:
> > > > `libmount` from util-linux and many softwares in userspace (e.g.
> > > > kubelet) did not expect empty string as mount source:
> > > > 
> > > > ```
> > > > $ mount -t tmpfs "" /mnt/tmpfs
> > > > $ findmnt /mnt/tmpfs
> > > > findmnt: /proc/self/mountinfo: parse error at line 51
> > > > $ cat /proc/self/mountinfo | grep -P '\/mnt\/tmpfs'
> > > > 74 25 0:59 / /mnt/tmpfs rw,relatime shared:38 - tmpfs  rw
> > > > $ cat /proc/self/mounts | grep -P '\/mnt\/tmpfs'
> > > >  /mnt/tmpfs tmpfs rw,relatime 0 0
> > > > ```
> > > > 
> > > > `source` field in mounts/mountinfo is empty, which breaks a lot of
> > > > mounts/mountinfo parsers.
> > > > 
> > > > This fixes issues in parsing when user uses empty string as mount
> > > > source.
> > > > 
> > > > Cc: Karel Zak <kzak@redhat.com>
> > > > Signed-off-by: Yecheng Fu <cofyc.jackson@gmail.com>
> > > 
> > > Reviewed-by: Matthew Wilcox <mawilcox@microsoft.com>
> > 
> > hi, I'm a newbie here. Will this be merged or is there still something I
> > need to do?
> > 
> > I had submitted a patch to [util-linux](https://github.com/karelzak/util-linux/pull/619). Karel prefer kernel side bugfix than rewrite all the sscanf() stuff and I agree. Kernel side bugfix also fixes all parsers too.
> > 
> > -- 
> > Yecheng Fu
> 
> hi, didn't receive any updates for months, sorry to ping again in case
> someone missed it. 
> 
> Karel patched util-linux to work around this issue now [^1], but I hope this
> can be fixed on kernel side too to improve compatibility.

I agree. I'm absolutely sure that in userspace are many many another
places where nobody expects empty mount source in /proc/self/mountinfo. 
It would be really nice to use there "none" as placeholder.

    Karel
diff mbox

Patch

diff --git a/fs/proc_namespace.c b/fs/proc_namespace.c
index 3f1190d..24591b8 100644
--- a/fs/proc_namespace.c
+++ b/fs/proc_namespace.c
@@ -104,7 +104,7 @@  static int show_vfsmnt(struct seq_file *m, struct vfsmount *mnt)
 		if (err)
 			goto out;
 	} else {
-		mangle(m, r->mnt_devname ? r->mnt_devname : "none");
+		mangle(m, (r->mnt_devname && r->mnt_devname[0] != '\0') ? r->mnt_devname : "none");
 	}
 	seq_putc(m, ' ');
 	/* mountpoints outside of chroot jail will give SEQ_SKIP on this */
@@ -174,7 +174,7 @@  static int show_mountinfo(struct seq_file *m, struct vfsmount *mnt)
 		if (err)
 			goto out;
 	} else {
-		mangle(m, r->mnt_devname ? r->mnt_devname : "none");
+		mangle(m, (r->mnt_devname && r->mnt_devname[0] != '\0') ? r->mnt_devname : "none");
 	}
 	seq_puts(m, sb->s_flags & MS_RDONLY ? " ro" : " rw");
 	err = show_sb_opts(m, sb);
@@ -202,7 +202,7 @@  static int show_vfsstat(struct seq_file *m, struct vfsmount *mnt)
 		if (err)
 			goto out;
 	} else {
-		if (r->mnt_devname) {
+		if (r->mnt_devname && r->mnt_devname[0] != '\0') {
 			seq_puts(m, "device ");
 			mangle(m, r->mnt_devname);
 		} else