#49 sshfs Failure

closed-fixed
Florin Malita
None
6
2003-06-10
2003-03-16
John L. Clark
No

This bug report corresponds to the e-mail I sent to the
mailing list on 15 March.

Using current cvs sources, sshfs doesn't work. Compiled
with --enable-debug and --enable-kdebug, you can find
the cli debug log in the attached file debug.gz, and the
kernel debug log in the attached file kdebug.gz. I am
running stock kernel 2.4.20, gcc 3.2, and new or recent
versions of ssh on the relevant machines. The module is
successfully loaded, and I proceed as follows:

machine:~/mnt # lufsmount
sshfs://user@othermachine/home/user dir/
[... seemingly successful output can be found in debug.gz
...]

machine:~/mnt # ls dir/
[... hangs seemingly indefinitely ...]

At this point I interrupt out of the `ls`, and I get the two
additional messages:

Mar 14 17:41:26 machine kernel: (lu_execute) -
do_execute failed!
Mar 14 17:41:26 machine kernel: (lu_execute) -
interrupted by a signal. disconnecting this slot...

Discussion

  • John L. Clark
    John L. Clark
    2003-03-16

    • priority: 5 --> 6
     
  • John L. Clark
    John L. Clark
    2003-03-16

    Console debugging output

     
    Attachments
  • John L. Clark
    John L. Clark
    2003-03-16

    Kernel debugging output

     
    Attachments
  • Logged In: YES
    user_id=431383

    don't know if this will be useful or not, but i traced the
    error (for me at least) to kernel/Linux/2.4/proc.c. inside
    the do_execute() function the call to sock_recv() is
    returning -ERESTARTSYS which causes do_execute() to return
    an error. something also a little odd is that SIGIO is set
    in the signal mask when lu_execute() checks it after the
    failed do_execute().

    i also noticed that at the time the problem occurs another
    client attempts to connect to the daemon. this seems odd,
    but perhaps i just don't understand the code...

    for me, accesses to the file system from a bash shell work
    fine, but accesses from emacs cause the lockup.

     
  • Logged In: NO

    I just wanted to note that I am having the same problem with
    sshfs. I tried several older versions of lufs, and I found that this problem
    was introduced in lufs-0.9.0pre1. In version 0.8.3 it works fine. Hopefully
    that will help the developers track the problem down.

    Brandon

     
  • Logged In: NO

    I just wanted to note that I am having the same problem with
    sshfs. I tried several older versions of lufs, and I found that this problem
    was introduced in lufs-0.9.0pre1. In version 0.8.3 it works fine. Hopefully
    that will help the developers track the problem down.

    Brandon

     
  • Florin Malita
    Florin Malita
    2003-05-11

    • assigned_to: nobody --> mali
     
  • Florin Malita
    Florin Malita
    2003-05-11

    Logged In: YES
    user_id=154306

    Try adding "-c 1" (with lufsmount) or "channels=1" (with
    mount).

    Let me know whether that helps.

     
  • Florin Malita
    Florin Malita
    2003-05-27

    Logged In: YES
    user_id=154306

    could somebody try the CVS version and verify whether it fixes this
    issue?

     
  • John L. Clark
    John L. Clark
    2003-05-28

    Logged In: YES
    user_id=143160

    I initially confirm that this does in fact work. I haven't tested it
    rigorously, but all initial tests seem to indicate proper
    performance. And... woohoo! Thanks, Florin!

    Now, off to build a version without command line debug
    output, so that I can use it!

     
  • Logged In: YES
    user_id=736298

    I've got the same problem but with more message :

    [root@cassiope gnumdk]# cd /mnt/cerela/
    [root@cassiope cerela]# ls
    ls: .: Stale NFS file handle
    [root@cassiope cerela]# ls
    bin/ download/ home/ mnt/ proc/ sbin/ var/
    boot/ etc/ initrd/ opt/ root/ tmp/ Videos/
    dev/ halt lib/ poweroff save/ usr/ Zik/
    [root@cassiope cerela]# ls
    ls: lecture du rpertoire .: Connection reset by peer
    [root@cassiope cerela]#

    It works one time and failed two time :(

     
  • Florin Malita
    Florin Malita
    2003-06-10

    Logged In: YES
    user_id=154306

    your problem is different:

    you either didn't configure public key authentication on the
    remote server or the server only allows one connection/ip.

    channels=1 should be a quick workaround.

     
  • Florin Malita
    Florin Malita
    2003-06-10

    • status: open --> closed-fixed