SAMBA digest 1920
Multiple recipients of list
samba at samba.org
Fri Dec 25 06:47:11 GMT 1998
Headway Research is closed for the Christmas and New Year holidays. Normal
work will be resumed Monday January 4, 1999. Your email will receive the
appropriate attention upon our return. We wish you a safe and happy
holiday season!!
***************************************************************************
Headway Research, Inc. |Telephone: 972-272-5431
3713 Forest Lane |Facsimile: 972-272-7817
Garland, Texas 75042 |e-mail: verns at headres.com
U.S.A. |http://www.HeadwayResearch.com
***************************************************************************
Your message follows:
===========================================================================
> From samba at samba.org Fri Dec 25 06:47:10 1998
> Return-Path: samba at samba.org
> Received: from samba.org (uucp at localhost) by headway.hri.com (8.8.5/8.6.9) with UUCP id GAA08425 for jimh at headres.com; Fri, 25 Dec 1998 06:40:11 GMT
> From: samba at samba.org
> Received: by metronet.com id AA02776
> (5.67a/IDA1.5hp for <jimh at headres.com>); Fri, 25 Dec 1998 00:29:48 -0600
> Received: from kong.bstc.net(207.239.221.11)
> via SMTP by fiinix.metronet.com, id smtpdAAAa02928; Fri Dec 25 00:29:37 1998
> Received: (qmail 3865 invoked from network); 25 Dec 1998 06:17:34 -0000
> Received: from samba.anu.edu.au (150.203.164.44)
> by kong.bstc.net with SMTP; 25 Dec 1998 06:17:34 -0000
> Received: from localhost ([127.0.0.1]:15475 "HELO" ident: "NO-IDENT-SERVICE[2]") by samba.anu.edu.au with SMTP id <12615455-4380>; Fri, 25 Dec 1998 17:17:43 +1100
> Errors-To: listproc-errors at samba.anu.edu.au
> Reply-To: samba at samba.org
> Originator: samba at samba.org
> Sender: samba at samba.org
> Precedence: bulk
> To: Multiple recipients of list <samba at samba.org>
> Subject: SAMBA digest 1920
> X-Listprocessor-Version: 6.0c -- ListProcessor by Anastasios Kotsikonas
> X-Url: http://samba.anu.edu.au/listproc
> X-Comment: Discussion of Samba software package
> Message-Id: <19981225061749Z12615455-4380+973 at samba.anu.edu.au>
> Date: Fri, 25 Dec 1998 17:17:43 +1100
>
> SAMBA Digest 1920
>
> For information on unsubscribing see http://samba.org/listproc/
> Topics covered in this issue include:
>
> 1) Odd behavior w/ read raw and write raw under NT1
> by Peter Chen <petechen at remus.rutgers.edu>
> 2) %L, subnets and dns resolution Xmas problem
> by laurent.menu at temic.fr
> 3) Re: File locking
> by Benjamin Scott <bscott at hamptonsys.com>
> 4) Re: Problems with SMB.CONF (2nd)
> by "Robert Dahlem" <Robert.Dahlem at frankfurt.netsurf.de>
> 5) Re: %L, subnets and dns resolution Xmas problem
> by "Florian G. Pflug" <fgp at fgp.priv.at>
>
> ----------------------------------------------------------------------
>
> Date: Thu, 24 Dec 1998 08:07:59 -0500
> From: Peter Chen <petechen at remus.rutgers.edu>
> To: samba at samba.org
> Subject: Odd behavior w/ read raw and write raw under NT1
> Message-ID: <v03007818b2a7e9681358@[10.10.53.4]>
>
> A coworker and I have been doing some performance tuning with Samba 20b4
> under Linux 2.0.36 (RedHat 5.2 distribution) on a 100-T network. The
> protocol is NT1. Two settings are tweaked in particular, "read raw" and
> "write raw". When both are yes, reading a large file results in 6M/s, but
> writing is only 2M/s, which is far below writing to an NT server (6M/s).
> If any of "read raw" or "write raw" is off, the result is the other way
> around, with writing a file to be around 6M/s, and reading around 2M/s.
>
> When I changed the protocol to LANMAN2, "read raw = yes" and "write raw =
> no" result in good throughput for both reading and writing. The only
> problem is that LANMAN2 doesn't support long filenames. So we looked at
> the source searching for places where "read raw" and "write raw" are used.
> To take it a step further, we downloaded the SMB protocol specs from
> Microsoft. It turned out that there are two separate bits in LANMAN2's
> message for read raw and write raw, whereas under NT1, in the
> "Capabilities" structure, there is only a 1 bit flag for whether the server
> supports both read raw and write raw. This seems to be exactly where the
> problem is.
>
> So how does one configure the server using protocol NT1 to have read raw =
> yes, and write raw = no? Right now the flags are simply "and" together in
> the code.
> TIA.
>
> Pete <petechen at remus.rutgers.edu>
>
> Peter Chen
> email: <mailto:petechen at remus.rutgers.edu>
>
>
>
> ------------------------------
>
> Date: Thu, 24 Dec 1998 14:05:00 +0100
> From: laurent.menu at temic.fr
> Subject: %L, subnets and dns resolution Xmas problem
> Message-ID: <"d09X'SQ000000000*"@MHS>
>
> Hi,
>
> Using the %L macro in smb.conf is for me *very* usefull, I can setup as
> many logical servers as I need and that's great : no need for new machines,
> I can switch logical servers back and forth, it's a kind of System
> Administrator dream :-)
>
> *But* I've discovered a problem with the %L macro in a subnetted
> environment :
>
> - "hermes" is a Samba 1.9.18p10 server on a hpux 10.20 and is configured
> like this:
> netbios name = hermes
> netbios aliases = tet59 tet30 mc752b tet58 am740b td446
>
> Please note : tet59, tet30, mc752b, tet58, am740b and td446 are also
> hermes aliases in the DNS. I call thos aliases "fake servers".
>
> - I have a service like this :
> [public]
> ;fakeserver = tet59 tet30 mc752b tet58 am740b td446
> comment = %L Public Zone
> path = /home/dos/%L/public
> public = yes
> writeable =yes
> create mask = 664
> directory mask = 775
> that means I have a different "common" service for each fake server. I list
> the corresponding shared disks below :
> \\tet59\public
> \\tet58\public
> \\tet30\public
> \\td446\public
> \\am740b\public
> \\mc752b\public
> \\hermes\public
>
> - it works from win95 and NT4-SP3 as far as the names are resolved by
> broadcast, the pc and the server are on the same subnet
>
> - but if the client is in another subnet and the name (ie : tet59) is then
> resolved by DNS, wether you chose \\tet59\public, \\tet58\public,
> \\tet30\public, \\td446\public, \\am740b\public, \\mc752b\public or
> \\hermes\public, it connects to \\hermes\public.
>
>
> =====> Actually it seems the %L macro is mapped to the real netbios
> =====> name of the machine defined by smb.conf "netbios name ="
> =====> configuration directive when the server is called from a pc
> =====> resolving the server name with DNS. :-(
>
>
> - using for example "smbclient -L tet59" from a samba box on a different
> subnet works correctly (shows "tet59 Common Zone" as the service comment)
>
> Please note I can't (and don't want to ;-) use WINS nor configure a lmhosts
> table on the pc's
>
> I did not join any log as there are no error status in them @ debug level 3.
> Please tell me if you need some more information.
>
>
> Another (less important :-) question : is there a maximum number of netbios
> aliases for a samba server ?
>
>
> Thank you for your help and merry christmas :-)
>
>
> Laurent Menu
>
>
> ------------------------------
>
> Date: Thu, 24 Dec 1998 10:42:57 -0500 (EST)
> From: Benjamin Scott <bscott at hamptonsys.com>
> To: Joel Knight <jwknight at cyberlink.bc.ca>
> Subject: Re: File locking
> Message-ID: <Pine.LNX.4.05.9812241024200.9558-100000 at benden.hamptonsys.com>
>
> On Wed, 23 Dec 1998, Joel Knight wrote:
> > I have samba 1.9.18p10 running on a Linux 2.0.34 box serving win95 and 98
> > clients. My problem arises with file locking. When one of the clients
> > opens a file, smbstatus shows the file as being open and locked which is
> > fine. When one of the clients saves the open file, smbstatus no longer
> > shows the file as being open. Is that normal behavior? Also, if a client
> > opens a file, and then closes it without saving, smbstatus shows that the
> > file is still opened. Is this also normal?
>
> I do not know if that is intended behavior, but I seen similar behavior
> here. I first noticed it when I opened a Perl CGI script for editing in my
> favorite (DOS based) editor. Samba listed that file as locked, which makes
> some sense [1]. I saved it, exited, and tried to run the script, and it
> failed with "Text file busy". Sure enough, Samba was still listing that file
> as open and locked, even though I had exited that editor, and even the DOS
> window that spawned it!
>
> [1] But not a lot, since my editor is configured to only open the files when
> actually reading/saving from/to the file.
>
> At first, I could not get Samba to close the file without logging out or
> rebooting the 'doze client.
>
> Then I noticed something weirder -- and more worrisome. If I run a program
> (Microsoft Source Safe, as it happens) that opens a lot of files, Samba seems
> to "forget about" the locks opened on previous files. It is as if the locks
> are maintained in a fixed-size table, and LRU locks fall off the end of the
> table.
>
> This behavior occurred both on a Win98 (new install) and a Win NT 4.0 SP4
> Workstation client.
>
> So, I have multiple concerns. First, why does Samba maintain locks even
> after the client has closed the file? Second, do mandatory exclusive locks
> requested by the client also fall out of the lock table? I sure hope not!
>
> I want to look into this further, but I will not have time to do that until
> mid-January at the earliest. In the mean time, maybe this information will
> help somebody. :)
>
> Configuration highlights:
> - Linux kernel 2.0.35
> - Base distributionL: Red Hat Linux 5.1
> - Samba is 1.9.18p10, compiled from pristine sources
> - PAM is in use
> - dns proxy = true
> - wins support - true
> - {domain|local|preferred} master = true
> - oplocks = true
> - read prediction = true
> - strict locking = true
> - encrypt passwords = true
>
> I can provide full configuration files, log files at the debug level of your
> choice, smbstatus dumps, or whatever, on request.
>
> Have a nondenominational event of a nature you prefer. :-)
>
> ----------------------------------------------------------------------
> Benjamin Scott <bscott at hamptonsys.com> Phone: (603)431-7315
> Hampton Systems Group, Inc. Fax : (603)431-0822
> http://www.hamptonsys.com DSN : 852-3581
> ======================================================================
>
>
> ------------------------------
>
> Date: Thu, 24 Dec 1998 17:39:22 +0100
> From: "Robert Dahlem" <Robert.Dahlem at frankfurt.netsurf.de>
> To: "Majid Tajamolian" <majid at bol.sharif.ac.ir>
> Cc: "Multiple recipients of list" <samba at samba.org>
> Subject: Re: Problems with SMB.CONF (2nd)
> Message-ID: <199812241639.RAA18030 at beach.frankfurt.netsurf.de>
>
> Majid,
>
> >> >> >3. The "dont descend" parameter for the "root" share name has not any
> >> >> > effect!
>
> >> Can you provide a log made with debug level 8 of some test case?
>
> >OK, enclosed are log files with debug level 8 which I connect to the
> >"root" service and descend (!) to the /dev directory.
>
> Your logs show clearly that samba is using "./dev" instead of "/dev". As far as I
> remember there is a hint in "man smb.conf" that one has to experiment a little
> with i.e. "./dev" and so on.
>
> Would you please be so kind: try this and report about your (non-)success?
>
> Hasta la vista,
> Robert
>
> --
> ---------------------------------------------------------------
> Robert.Dahlem at frankfurt.netsurf.de
> Radio Bornheim - 2:2461/332 at fidonet +49-69-4930830 (ZyX, V34)
> 2:2461/326 at fidonet +49-69-94414444 (ISDN X.75)
> ---------------------------------------------------------------
>
>
> ------------------------------
>
> Date: Thu, 24 Dec 1998 17:20:28 +0100
> From: "Florian G. Pflug" <fgp at fgp.priv.at>
> To: laurent.menu at temic.fr, Multiple recipients of list <samba at samba.org>
> Subject: Re: %L, subnets and dns resolution Xmas problem
> Message-ID: <19981224172028.A1017 at fgp.priv.at>
>
> On Fri, Dec 25, 1998 at 01:55:59AM +1100, laurent.menu at temic.fr wrote:
> > Hi,
> >
> > Using the %L macro in smb.conf is for me *very* usefull, I can setup as
> > many logical servers as I need and that's great : no need for new machines,
> > I can switch logical servers back and forth, it's a kind of System
> > Administrator dream :-)
> >
> > *But* I've discovered a problem with the %L macro in a subnetted
> > environment :
> >
> > - "hermes" is a Samba 1.9.18p10 server on a hpux 10.20 and is configured
> > like this:
> > netbios name = hermes
> > netbios aliases = tet59 tet30 mc752b tet58 am740b td446
> >
> > Please note : tet59, tet30, mc752b, tet58, am740b and td446 are also
> > hermes aliases in the DNS. I call thos aliases "fake servers".
> >
> > - I have a service like this :
> > [public]
> > ;fakeserver = tet59 tet30 mc752b tet58 am740b td446
> > comment = %L Public Zone
> > path = /home/dos/%L/public
> > public = yes
> > writeable =yes
> > create mask = 664
> > directory mask = 775
> > that means I have a different "common" service for each fake server. I list
> > the corresponding shared disks below :
> > \\tet59\public
> > \\tet58\public
> > \\tet30\public
> > \\td446\public
> > \\am740b\public
> > \\mc752b\public
> > \\hermes\public
> >
> > - it works from win95 and NT4-SP3 as far as the names are resolved by
> > broadcast, the pc and the server are on the same subnet
> >
> > - but if the client is in another subnet and the name (ie : tet59) is then
> > resolved by DNS, wether you chose \\tet59\public, \\tet58\public,
> > \\tet30\public, \\td446\public, \\am740b\public, \\mc752b\public or
> > \\hermes\public, it connects to \\hermes\public.
> >
> >
> > =====> Actually it seems the %L macro is mapped to the real netbios
> > =====> name of the machine defined by smb.conf "netbios name ="
> > =====> configuration directive when the server is called from a pc
> > =====> resolving the server name with DNS. :-(
> >
> >
> > - using for example "smbclient -L tet59" from a samba box on a different
> > subnet works correctly (shows "tet59 Common Zone" as the service comment)
> >
> > Please note I can't (and don't want to ;-) use WINS nor configure a lmhosts
> > table on the pc's
> >
> > I did not join any log as there are no error status in them @ debug level 3.
> > Please tell me if you need some more information.
> >
> >
> > Another (less important :-) question : is there a maximum number of netbios
> > aliases for a samba server ?
> >
> >
> > Thank you for your help and merry christmas :-)
> >
> >
> > Laurent Menu
> >
> Hi
>
> The only solution I see, is to use network aliasing. You'd have to configure
> as many virtual interfaces as you want virtual servers. Then you start a
> nmbd & smbd for each interface with "interfaces =
> <one-of-the-logical-interfaces>".
>
> Greetings, Florian Pflug
>
>
> ------------------------------
>
> End of SAMBA Digest 1920
> ************************
More information about the samba
mailing list