I noticed that smbfs no longer respects the "uid" and "gid" mount
options passed to it by mount.(I think it stopped when the server was
upgraded to Samba 3.0. Not sure though, since my client was upgraded to
Linux 2.6.3 at around the same time). I've made this small patch that
fixes it (bear with me, this is my first patch to the kernel :-) ):
======== Start patch ========
--- kernel-source-2.6.3.orig/fs/smbfs/proc.c 2004-02-19
08:55:44.000000000 +0 000
+++ kernel-source-2.6.3/fs/smbfs/proc.c 2004-03-04
13:56:04.000000000 +0 000
@@ -1834,7 +1834,13 @@
static void
smb_finish_dirent(struct smb_sb_info *server, struct smb_fattr *fattr)
{
- if (fattr->f_unix)
+
+ if (server->mnt->uid)
+ fattr->f_uid = server->mnt->uid;
+ if (server->mnt->gid)
+ fattr->f_gid = server->mnt->gid;
+
+ if (fattr->f_unix)
return;
fattr->f_mode = server->mnt->file_mode;
======= End patch ========
--
Søren Hansen
tor, 04.03.2004 kl. 23.20 skrev S?ren Hansen:
> I noticed that smbfs no longer respects the "uid" and "gid" mount
> options passed to it by mount.(I think it stopped when the server was
> upgraded to Samba 3.0. Not sure though, since my client was upgraded to
> Linux 2.6.3 at around the same time). I've made this small patch that
> fixes it (bear with me, this is my first patch to the kernel :-) ):
I have the same problem as you have. I have not gotten around to test
your patch yet (but will later tonight), but I read in another mail from
you that you had discussed the patch with Urban Widmark. Is the any
opinions against this patch? If not, will you try to submit it to
Andrew/Linus?
Anyway, thanks for the patch :)
Best regards,
Stian
On Mon, 8 Mar 2004, Stian Jordet wrote:
> tor, 04.03.2004 kl. 23.20 skrev S?ren Hansen:
> > I noticed that smbfs no longer respects the "uid" and "gid" mount
> > options passed to it by mount.(I think it stopped when the server was
> > upgraded to Samba 3.0. Not sure though, since my client was upgraded to
> > Linux 2.6.3 at around the same time). I've made this small patch that
> > fixes it (bear with me, this is my first patch to the kernel :-) ):
>
> I have the same problem as you have. I have not gotten around to test
> your patch yet (but will later tonight), but I read in another mail from
> you that you had discussed the patch with Urban Widmark. Is the any
> opinions against this patch? If not, will you try to submit it to
> Andrew/Linus?
There is nothing directly wrong with the patch, but maybe there are better
solutions.
Here is my summary of our off-list discussion.
1. If you configure samba and disable the unix extensions ("UE") in
smb.conf then smbfs will behave like it does with earlier samba
servers.
2. The idea with the UE is to communicate better over smb between unix
systems and uid/gid mapping isn't done in things like nfs.
Perhaps what you really want is to disable the unix extensions on
the smbfs side?
3. If this is needed for smbfs+UE perhaps it is also useful for other
filesystems in some cases. Some non-unixlike filesystems have these
mappings already and perhaps they would be generally useable.
(see also the "UID/GID mapping system" post to this list)
Sorry for taking that thread off-list, btw. I didn't read the message too
carefully when Andrew poked me for not responding.
/Urban
man, 08.03.2004 kl. 22.50 skrev Urban Widmark:
> Here is my summary of our off-list discussion.
>
> 1. If you configure samba and disable the unix extensions ("UE") in
> smb.conf then smbfs will behave like it does with earlier samba
> servers.
Uhh, I didn't know this. Sorry for my stupidity. Everything works well
without any patch now. For my use, at least :)
Thanks :)
Best regards,
Stian