[Fwd: [Fwd: SMBMOUNT]]

Urban Widmark urban at svenskatest.se
Fri Sep 1 21:46:39 GMT 2000

[please, do not forward the message as an attachment if you expect a reply
 to that text]

On Fri, 1 Sep 2000, Brendon Standing wrote:

> change into that directory it is mounted to. The problem comes, however,
> when I try list that share or directory. It just hangs here and I have
> to log onto another screen and kill the processes running and even
> umount that share. If , hovever, don't list this share but change into
> another subdirectory ( which I already know is there ) off of this
> share, i can list the file and do what I need to do without a problem. I
> believe that the problem might be in the number of folders it is trying
> to list within that share. I think that the answer might be in the

Are you sure you wait long enough? ls can take a lot of time for large
directories, especially if the ls needs to stat() each file to check it's
type (color ls, ls -F or ls -l).

Try '/bin/ls --color=never /some_path' and go for a coffe break ... :)

I think smbfs needs to be more clever when reading directories. "aa" is a
test directory with 568 files on a NT4 box, a short 100Mbps trip away.

% time /bin/ls /mnt/auto/ntfs/nls/aa/ > /dev/null
0.000u 0.010s 0:00.14 7.1%      0+0k 0+0io 117pf+0w
% time /bin/ls -alF /mnt/auto/ntfs/nls/aa/ > /dev/null 
0.010u 0.030s 0:01.25 3.2%      0+0k 0+0io 150pf+0w

That is a rather large difference ... (time measured on the 2nd run).

> output once I have run the kill command once that process has hung. The
> errors are such:
> smb_get_length : recv-error=512
> smb_trans2_request : result -512, setting invalid
> smb_retry : caught signal
> smb_proc_eraddir_long : error=512 breaking
> smb_refill_dircashe:readdir failed,result=-512

That is caused by the current process (the ls) getting a signal while
receiving data in the smbfs code. A bug causes the mount point to become
unusable after this (smbfs tells smbmount to make a new connection but
doesn't wait for it to complete, causing smbmount to terminate).

But since this happens after you kill the process it is not related to the
ls problem.

(A "better than nothing" fix for this is in 2.4.0-test7 and hopefully soon
 in 2.2.18-pre3, anyone wanting this bugfix, nls support or OS/2 support
 can get the not-yet included patch from here:


More information about the samba mailing list