Re: [sshfs] sshfs-fuse mounting bug

[Miklos Szeredi <miklos@sz...>]

> Good work with sshfs, it certainly makes life easier.
> 
> I have found what appears to be a bug by accidentally mounting an 
> invalid path.
> 
> Instead of receiving an invalid path error, the mount operation 
> succeeded.  Subsequent attempts to access the mounted file system were, 
> of course, unsuccessful.  There wasn't a problem unmounting.
> 
> I don't know if this is the intended sshfs behaviour, but I doubt it.  I 
> think that the mount should fail and give an invalid path message (e.g. 
> sshfs: /invalid/path: No such file or directory).
> 
> Below you will find an example of the bug and also some relevant 
> configuration information.

This is more or less a known situation.  Fixing it wouldn't be too
hard, but I'm reluctant to add an extra client-server roundtrip delay
to the mounting just for the error case.

An option could be added which would enable checking the remote path.
But it would be turned off by default, which would make it sort of
useless.

Miklos

[sshfs] sshfs-fuse mounting bug

[Joseph Spadavecchia <j.spadavecchia@ed...>]

Good work with sshfs, it certainly makes life easier.

I have found what appears to be a bug by accidentally mounting an 
invalid path.

Instead of receiving an invalid path error, the mount operation 
succeeded.  Subsequent attempts to access the mounted file system were, 
of course, unsuccessful.  There wasn't a problem unmounting.

I don't know if this is the intended sshfs behaviour, but I doubt it.  I 
think that the mount should fail and give an invalid path message (e.g. 
sshfs: /invalid/path: No such file or directory).

Below you will find an example of the bug and also some relevant 
configuration information.

Kind regards,
Joseph

P.S. Although different, in practical terms how do you compare your work 
with that of SFS (http://www.fs.net)?

Here's an example, where foo is the (empty) mount point:

joe@...:~/test > ls -lh
total 4.0K
drwxr-xr-x  2 joe joe 4.0K Dec  1 15:47 foo
joe@...:~/test >

joe@...:~/test > sshfs joe@...:/invalid/path ./foo
Password:
joe@...:~/test >

joe@...:~/test > ls
foo
joe@...:~/test >

joe@...:~/test > ls -lh
ls: foo: No such file or directory
total 0
joe@...:~/test >

joe@...:~/test > cd foo
joe@...:~/test/foo >

joe@...:~/test/foo > ls -lh
ls: .: No such file or directory
joe@...:~/test/foo >

joe@...:~/test/foo > cd ../
joe@...:~/test > fusermount -u ./foo
joe@...:~/test > ls -lh foo
total 0
joe@...:~/test >

dmesg and syslogs report no errors:

fuse init (API version 7.2)
fuse distribution version: 2.4.1

My configuration is as follows:

client:
~~~~

1) vanilla Linux  2.6.11.11 kernel
2) fuse 2.4.1
3) sshfs 1.2
4) OpenSSH_3.9p1, OpenSSL 0.9.7e

server:
~~~~~

1) vanilla Linux 2.6.12.5 kernel
2) OpenSSH_3.9p1, OpenSSL 0.9.7e

Re: [sshfs] sshfs-fuse mounting bug

[Miklos Szeredi <miklos@sz...>]

> Good work with sshfs, it certainly makes life easier.
> 
> I have found what appears to be a bug by accidentally mounting an 
> invalid path.
> 
> Instead of receiving an invalid path error, the mount operation 
> succeeded.  Subsequent attempts to access the mounted file system were, 
> of course, unsuccessful.  There wasn't a problem unmounting.
> 
> I don't know if this is the intended sshfs behaviour, but I doubt it.  I 
> think that the mount should fail and give an invalid path message (e.g. 
> sshfs: /invalid/path: No such file or directory).
> 
> Below you will find an example of the bug and also some relevant 
> configuration information.

This is more or less a known situation.  Fixing it wouldn't be too
hard, but I'm reluctant to add an extra client-server roundtrip delay
to the mounting just for the error case.

An option could be added which would enable checking the remote path.
But it would be turned off by default, which would make it sort of
useless.

Miklos