Steven French | 3 May 20:20 2005
Picon

Re: Some problems with linux-cifs


> - on linux-cifs.samba.org, "1.34 for 2.6 kernels" links to 1.33
> - the tarballs for 1.33 or 1.34 both result in this message when trying
> to build a new kernel, based off of the stock config for ubuntu:
...
>    CC [M]  fs/cifs/connect.o
> fs/cifs/connect.c: In function `cifs_mount':
> fs/cifs/connect.c:1702: error: structure has no member named `s_time_gran'
> make[3]: *** [fs/cifs/connect.o] Error 1
> make[2]: *** [fs/cifs] Error 2
> make[1]: *** [fs] Error 2
> make[1]: Leaving directory `/usr/src/linux-source-2.6.10'
> make: *** [stamp-build] Error 2

This field was globally added to the kernel for proper nanosecond time rounding (presumably it was added just after 2.6.10).  The RHEL4 version of the cifs code (which handles 2.6.8/2.6.9 kernels) should be fine and does not depend on this field, although I have not tried it personally on 2.6.10

> but given the huge number of suspiciously broken-looking pages and broken links on the linux cifs client site
I don't spend much time on the webpages, but must have broken a few links when I added the two new child html pages, but I will scan through and try to find and fix any broken links.  Check back tomorrow and let me know if you see any broken links that I have not addresed.

_______________________________________________
linux-cifs-client mailing list
linux-cifs-client <at> lists.samba.org
https://lists.samba.org/mailman/listinfo/linux-cifs-client
Glyph Lefkowitz | 3 May 21:16 2005

Re: Some problems with linux-cifs

Steven French wrote:
> 
>  > - on linux-cifs.samba.org, "1.34 for 2.6 kernels" links to 1.33
>  > - the tarballs for 1.33 or 1.34 both result in this message when trying
>  > to build a new kernel, based off of the stock config for ubuntu:
> ...
>  >    CC [M]  fs/cifs/connect.o
>  > fs/cifs/connect.c: In function `cifs_mount':
>  > fs/cifs/connect.c:1702: error: structure has no member named 
> `s_time_gran'
>  > make[3]: *** [fs/cifs/connect.o] Error 1
>  > make[2]: *** [fs/cifs] Error 2
>  > make[1]: *** [fs] Error 2
>  > make[1]: Leaving directory `/usr/src/linux-source-2.6.10'
>  > make: *** [stamp-build] Error 2
> 
> This field was globally added to the kernel for proper nanosecond time 
> rounding (presumably it was added just after 2.6.10).  The RHEL4 version 
> of the cifs code (which handles 2.6.8/2.6.9 kernels) should be fine and 
> does not depend on this field, although I have not tried it personally 
> on 2.6.10

Thanks for the prompt response.  I should have sent my initial inquiry 
to the list, sorry :-).

With the RHEL version of 1.34 the build seems to be progressing 
smoothly.  It hasn't linked yet, but it's past the point where it was 
dying before.

I'll publish .ko's for a few different stock ubuntu kernels if they work 
out OK.

>  > but given the huge number of suspiciously broken-looking pages and 
>  > broken links on the linux cifs client site

> I don't spend much time on the webpages, but must have broken a few 
> links when I added the two new child html pages, but I will scan through 
> and try to find and fix any broken links.  Check back tomorrow and let 
> me know if you see any broken links that I have not addresed.

OK.  When I check tomorrow I'll make a list.  Or perhaps I could submit 
patches to the website?  Where is the repository?
Glyph Lefkowitz | 4 May 04:37 2005

Re: Some problems with linux-cifs

Glyph Lefkowitz wrote:
> Steven French wrote:

> With the RHEL version of 1.34 the build seems to be progressing 
> smoothly.  It hasn't linked yet, but it's past the point where it was 
> dying before.
> 
> I'll publish .ko's for a few different stock ubuntu kernels if they work 
> out OK.

After learning more about building kernel modules than I really cared to 
know, I managed to get a working cifs module, but it seemed to have the 
problem that I was facing with the previous version of the cifs module 
even worse.

Since I've now tested with the latest version and it's still a problem I 
guess now I'll report the issue :).  I was under the impression that 
cifs(fs) was unequivocally better than smbfs in terms of linux clients 
for samba and win32 file sharing servers.  I was having a minor issue 
with smbfs where occasionally it would seize up and the process 
accessing the drive would become defunct.  I thought cifs might fix this.

Far from fixing it, cifs instead exacerbated the problem, locking up 
nautilus within minutes of trying (rather than days) to access any of my 
shares.  This causes nautilus, gam_server, and any other processes that 
happen to be accessing the volume at the time become both <defunct> and 
completely unkillable, despite telling 'init' to reap its children 
repeatedly.

I've replicated this behavior on 3 linux machines, all running ubuntu, 
with both the 1.28 that shipped with 2.6.10 and a 1.34[RHEL] I built 
myself against that kernel.  There are two desktops on a wired network 
with different network cards (an AMD64 box and an Athlon) and a laptop 
on the wireless using an ipw2200 (a Pentium M).  I have also booted 
Windows on 2 of these machines and verified that there are no lock-ups, 
hangs or disconnects from Windows clients, nor using the "connect to 
server" feature on Mac OS X.

I'm using an "embedded" device, the Buffalo Tech TeraStation:

http://buffalotech.com/products/product-detail.php?productid=97&categoryid=19

Which according to smbclient is:

Domain=[WORKGROUP] OS=[Unix] Server=[Samba 2.2.8a-ja-1.1]

I know that samba 2 is not too popular with the devs any more... could 
an interaction with that be causing the problem?

I've got a trouble ticket open with buffalo tech now since the device is 
supposedly going to support NFS through a firmware update "any day now" 
and I imagine that NFS would be a better choice for UNIX filesharing if 
I can get it going.  Still, "any day now" could be 6 months from now, 
and I'd like to have something semi-stable for my linux clients in the 
meanwhile.
Steven French | 4 May 05:16 2005
Picon

Re: Some problems with linux-cifs

The comment about locking up Nautilus reminds me of the D_NOTIFY support ...
Make sure you turn CIFS_EXPERIMENTAL off (it is a build option in make menuconfig), the D_NOTIFY support
which is not quite complete is on when CIFS_EXPERIMENTAL is enabled


Steve French
Senior Software Engineer
Linux Technology Center - IBM Austin
phone: 512-838-2294
email: sfrench at-sign us dot ibm dot com

Glyph Lefkowitz <glyph <at> divmod.com> wrote on 05/03/2005 09:37:15 PM:

> Glyph Lefkowitz wrote:
> > Steven French wrote:
>
> > With the RHEL version of 1.34 the build seems to be progressing
> > smoothly.  It hasn't linked yet, but it's past the point where it was
> > dying before.
> >
> > I'll publish .ko's for a few different stock ubuntu kernels if they work
> > out OK.
>
> After learning more about building kernel modules than I really cared to
> know, I managed to get a working cifs module, but it seemed to have the
> problem that I was facing with the previous version of the cifs module
> even worse.
>
> Since I've now tested with the latest version and it's still a problem I
> guess now I'll report the issue :).  I was under the impression that
> cifs(fs) was unequivocally better than smbfs in terms of linux clients
> for samba and win32 file sharing servers.  I was having a minor issue
> with smbfs where occasionally it would seize up and the process
> accessing the drive would become defunct.  I thought cifs might fix this.
>
> Far from fixing it, cifs instead exacerbated the problem, locking up
> nautilus within minutes of trying (rather than days) to access any of my
> shares.  This causes nautilus, gam_server, and any other processes that
> happen to be accessing the volume at the time become both <defunct> and
> completely unkillable, despite telling 'init' to reap its children
> repeatedly.
>
> I've replicated this behavior on 3 linux machines, all running ubuntu,
> with both the 1.28 that shipped with 2.6.10 and a 1.34[RHEL] I built
> myself against that kernel.  There are two desktops on a wired network
> with different network cards (an AMD64 box and an Athlon) and a laptop
> on the wireless using an ipw2200 (a Pentium M).  I have also booted
> Windows on 2 of these machines and verified that there are no lock-ups,
> hangs or disconnects from Windows clients, nor using the "connect to
> server" feature on Mac OS X.
>
> I'm using an "embedded" device, the Buffalo Tech TeraStation:
>
> http://buffalotech.com/products/product-detail.php?productid=97&categoryid=19
>
> Which according to smbclient is:
>
> Domain=[WORKGROUP] OS=[Unix] Server=[Samba 2.2.8a-ja-1.1]
>
> I know that samba 2 is not too popular with the devs any more... could
> an interaction with that be causing the problem?
>
> I've got a trouble ticket open with buffalo tech now since the device is
> supposedly going to support NFS through a firmware update "any day now"
> and I imagine that NFS would be a better choice for UNIX filesharing if
> I can get it going.  Still, "any day now" could be 6 months from now,
> and I'd like to have something semi-stable for my linux clients in the
> meanwhile.

_______________________________________________
linux-cifs-client mailing list
linux-cifs-client <at> lists.samba.org
https://lists.samba.org/mailman/listinfo/linux-cifs-client
Steven French | 4 May 19:46 2005
Picon

Re: user get lost by writing to a share which root has mounted

Are Unix extensions enabled in the smb.conf for your Samba 3 server?

You may also want to experiment with the mount options "nosetuids" (default IIRC) and "setuids" mount option which controls whether the server or client uid is used in setting the owner of a newly created flag)



Steve French
Senior Software Engineer
Linux Technology Center - IBM Austin
phone: 512-838-2294
email: sfrench at-sign us dot ibm dot com
Christian Keller <oerray <at> yahoo.de>


          Christian Keller <oerray <at> yahoo.de>
          Sent by: linux-cifs-client-bounces+sfrench=us.ibm.com <at> lists.samba.org

          04/13/2005 04:59 PM


To

linux-cifs-client <at> lists.samba.org

cc


Subject

[linux-cifs-client] user get lost by writing to a share which root has mounted

Hello,

I want to mount the home directorys from a samba 3
server for Apache to use the public_html feature.
The Idea was to mount the /home on the Apache server
and not ever user home (we have ~1000 users).
The share was mounted from root. The usernames and groups
on the Apache box for read access are right. But if a user
makes a new file (through a php script or something like
that) the file owner is root.

Is this a configuration failure from me or
is something wronge with cifs/samba?

Thanks for your help.

sincerly
Chris

_______________________________________________
linux-cifs-client mailing list
linux-cifs-client <at> lists.samba.org
https://lists.samba.org/mailman/listinfo/linux-cifs-client

_______________________________________________
linux-cifs-client mailing list
linux-cifs-client <at> lists.samba.org
https://lists.samba.org/mailman/listinfo/linux-cifs-client
Glyph Lefkowitz | 4 May 20:21 2005

Re: Some problems with linux-cifs

Steven French wrote:
> The comment about locking up Nautilus reminds me of the D_NOTIFY support ...
> Make sure you turn CIFS_EXPERIMENTAL off (it is a build option in make 
> menuconfig), the D_NOTIFY support
> which is not quite complete is on when CIFS_EXPERIMENTAL is enabled

Unfortunately it already was off:

CONFIG_CIFS=m
# CONFIG_CIFS_STATS is not set
# CONFIG_CIFS_XATTR is not set
# CONFIG_CIFS_EXPERIMENTAL is not set
Glyph Lefkowitz | 4 May 20:35 2005

Re: Some problems with linux-cifs

Steven French wrote:

> I don't spend much time on the webpages, but must have broken a few 
> links when I added the two new child html pages, but I will scan through 
> and try to find and fix any broken links.  Check back tomorrow and let 
> me know if you see any broken links that I have not addresed.

Well, it's tomorrow, and it's still pretty bad :).

Incorrect Links:

On the body of http://linux-cifs.samba.org/

"1.34 for 2.6 kernels" links to "cifs-1.33.tar.gz"

Broken Links:

On the left sidebar of http://linux-cifs.samba.org/

"Installation" results in a python traceback.

"Todo List" results in a python traceback with a Samba logo on it.

Similarly for "Authors/Thanks"

"CIFS VFS Paper" and "Linux Network Filesystems Paper" both result in a 
page on IBM's site that says:

"We have recently made changes to the IBM developerWorks open source 
site.  The requested project cannot be found and may no longer be active."

Invalid HTML:

Many of the pages are just unbelievably ugly, to the point of 
uselessness.  for example, here's a screenshot of the CIFS Testing page:

http://twistedmatrix.com/users/glyph/images/content/screenshots/cifs-testing.png
Steven French | 4 May 21:05 2005
Picon

Re: Some problems with linux-cifs

Looks like the svn web service on samba.org is down, so I moved those links over to gitweb on kernel.org

I have not been able to track down two of the broken links (to conference presentations), but since they were a couple years old I removed them and will replace with newer information later.

Otherwise the others have been fixed on the main page.


Steve French
Senior Software Engineer
Linux Technology Center - IBM Austin
phone: 512-838-2294
email: sfrench at-sign us dot ibm dot com

Glyph Lefkowitz <glyph <at> divmod.com> wrote on 05/04/2005 01:35:23 PM:

> Steven French wrote:
>
> > I don't spend much time on the webpages, but must have broken a few
> > links when I added the two new child html pages, but I will scan through
> > and try to find and fix any broken links.  Check back tomorrow and let
> > me know if you see any broken links that I have not addresed.
>
> Well, it's tomorrow, and it's still pretty bad :).
>
> Incorrect Links:
>
> On the body of http://linux-cifs.samba.org/
>
> "1.34 for 2.6 kernels" links to "cifs-1.33.tar.gz"
>
> Broken Links:
>
> On the left sidebar of http://linux-cifs.samba.org/
>
> "Installation" results in a python traceback.
>
> "Todo List" results in a python traceback with a Samba logo on it.
>
> Similarly for "Authors/Thanks"
>
> "CIFS VFS Paper" and "Linux Network Filesystems Paper" both result in a
> page on IBM's site that says:
>
> "We have recently made changes to the IBM developerWorks open source
> site.  The requested project cannot be found and may no longer be active."
>
> Invalid HTML:
>
> Many of the pages are just unbelievably ugly, to the point of
> uselessness.  for example, here's a screenshot of the CIFS Testing page:
>
> http://twistedmatrix.
> com/users/glyph/images/content/screenshots/cifs-testing.png

_______________________________________________
linux-cifs-client mailing list
linux-cifs-client <at> lists.samba.org
https://lists.samba.org/mailman/listinfo/linux-cifs-client
Naidu Bollineni | 5 May 02:14 2005

RE: Re: Some problems with linux-cifs

How can I see long names of the files from server? Even though Flags2 says client is not bound by the 8.3 notation, the server that has about 5K files is returning them in 8.3 format. I would like to see the real names, otherwise, it is difficult to correlate the real names to the mapped/mangled names.

 

Naidu Bollineni

Senior Member Technical Staff

Kazeon Systems

naidu <at> kazeon.com

www.kazeon.com

From: linux-cifs-client-bounces+naidu=kazeon.com <at> lists.samba.org [mailto:linux-cifs-client-bounces+naidu=kazeon.com <at> lists.samba.org] On Behalf Of Steven French
Sent: Wednesday, May 04, 2005 12:05 PM
To: Glyph Lefkowitz
Cc: linux-cifs-client <at> lists.samba.org
Subject: [linux-cifs-client] Re: Some problems with linux-cifs

 

Looks like the svn web service on samba.org is down, so I moved those links over to gitweb on kernel.org

I have not been able to track down two of the broken links (to conference presentations), but since they were a couple years old I removed them and will replace with newer information later.

Otherwise the others have been fixed on the main page.


Steve French
Senior Software Engineer
LinuxTechnologyCenter - IBM Austin
phone: 512-838-2294
email: sfrench at-sign us dot ibm dot com

Glyph Lefkowitz <glyph <at> divmod.com> wrote on 05/04/2005 01:35:23 PM:

> Steven French wrote:
>
> > I don't spend much time on the webpages, but must have broken a few
> > links when I added the two new child html pages, but I will scan through
> > and try to find and fix any broken links.  Check back tomorrow and let
> > me know if you see any broken links that I have not addresed.
>
> Well, it's tomorrow, and it's still pretty bad :).
>
> Incorrect Links:
>
> On the body of http://linux-cifs.samba.org/
>
> "1.34 for 2.6 kernels" links to "cifs-1.33.tar.gz"
>
> Broken Links:
>
> On the left sidebar of http://linux-cifs.samba.org/
>
> "Installation" results in a python traceback.
>
> "Todo List" results in a python traceback with a Samba logo on it.
>
> Similarly for "Authors/Thanks"
>
> "CIFS VFS Paper" and "Linux Network Filesystems Paper" both result in a
> page on IBM's site that says:
>
> "We have recently made changes to the IBM developerWorks open source
> site.  The requested project cannot be found and may no longer be active."
>
> Invalid HTML:
>
> Many of the pages are just unbelievably ugly, to the point of
> uselessness.  for example, here's a screenshot of the CIFS Testing page:
>
> http://twistedmatrix.
> com/users/glyph/images/content/screenshots/cifs-testing.png

_______________________________________________
linux-cifs-client mailing list
linux-cifs-client <at> lists.samba.org
https://lists.samba.org/mailman/listinfo/linux-cifs-client
Steve French | 6 May 00:57 2005
Picon

Longnames (was problems with cifs)

> How can I see long names of the files from server? Even though Flags2
> says client is not bound by the 8.3 notation, the server that has about
> 5K files is returning them in 8.3 format. I would like to see the real
> names, otherwise, it is difficult to correlate the real names to the
> mapped/mangled names.
AFAIK, the only cases in which servers will return short names 
are the following:

1) the underlying filesystem (some versions of FAT e.g.) does 
not support longnames for which mangling the name is the only choice
other than rejecting the create or mkdir

2) as the alt name in some of the higher levels of FindFirst (the non-mangled
name is also typically returned)

3) (for Samba at least) when the filename contains what is considered
an illegal character (such as <>\/:)

Presumably it could be a bug in the server handling of the longnames 
flag in the header but I have not seen that.

Which server?

Gmane