Linux in-kernel CIFS client for attaching to Samba or Windows servers

headers Steven French | 3 May 2005 20:20

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

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 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

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

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

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

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

> 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?

Mon	Tue	Wed	Thu	Fri	Sat	Sun

						1
2	3	4	5	6	7	8
9	10	11	12	13	14	15
16	17	18	19	20	21	22
23	24	25	26	27	28	29
30	31

2	June 2011
1	February 2011
2	January 2011
6	December 2010
30	June 2010
155	May 2010
198	April 2010
144	March 2010
34	February 2010
69	January 2010
34	December 2009
27	November 2009
84	October 2009
98	September 2009
129	August 2009
107	July 2009
144	June 2009
222	May 2009
333	April 2009
113	March 2009
139	February 2009
99	January 2009
121	December 2008
146	November 2008
155	October 2008
123	September 2008
188	August 2008
118	July 2008
45	June 2008
145	May 2008
86	April 2008
119	March 2008
59	February 2008
89	January 2008
43	December 2007
146	November 2007
84	October 2007
81	September 2007
87	August 2007
54	July 2007
55	June 2007
63	May 2007
40	April 2007
130	March 2007
60	February 2007
42	January 2007
54	December 2006
36	November 2006
53	October 2006
48	September 2006
32	August 2006
45	July 2006
33	June 2006
61	May 2006
38	April 2006
46	March 2006
29	February 2006
36	January 2006
34	December 2005
55	November 2005
56	October 2005
21	September 2005
49	August 2005
26	July 2005
33	June 2005
59	May 2005
65	April 2005
42	March 2005
28	February 2005
30	January 2005
114	December 2004
47	November 2004
73	October 2004
23	September 2004
78	August 2004
35	July 2004
35	June 2004
60	May 2004
38	April 2004
44	March 2004
29	February 2004
26	January 2004
24	December 2003
1	January 1998

Re: Some problems with linux-cifs

Re: Some problems with linux-cifs

Re: Some problems with linux-cifs

Re: Some problems with linux-cifs

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

Re: Some problems with linux-cifs

Re: Some problems with linux-cifs

Re: Some problems with linux-cifs

RE: Re: Some problems with linux-cifs

Longnames (was problems with cifs)