[PATCH] CIFS: Add nodfs mount option

[Martin Koegler]

Older samba server (eg. 3.0.24 from Debian etch) don't work correctly,
if DFS paths are used. Such server claim that they support DFS, but fail
to process some requests with DFS paths. Starting with Linux 2.6.26,
the cifs clients starts sending DFS paths in such situations, rendering
it unuseable with older samba servers.

The nodfs mount options forces a share to be used with non DFS paths,
even if the server claims, that it supports it.

Signed-off-by: Martin Koegler <[hidden email]>
---

This patch makes my home directory on a samba server useable for Linux 2.6.26
again. It still needs more testing. Should apply to linux 2.6.27 too.

Would it be possible to support something like this in the offical CIFS client?

diff a/fs/cifs/connect.c b/fs/cifs/connect.c
--- a/fs/cifs/connect.c
+++ b/fs/cifs/connect.c
@@ -90,6 +90,7 @@ struct smb_vol {
  bool nocase:1;     /* request case insensitive filenames */
  bool nobrl:1;      /* disable sending byte range locks to srv */
  bool seal:1;       /* request transport encryption on share */
+ bool nodfs:1;
  unsigned int rsize;
  unsigned int wsize;
  unsigned int sockopt;
@@ -1223,6 +1224,8 @@ cifs_parse_mount_options(char *options,
  vol->sfu_emul = 1;
  } else if (strnicmp(data, "nosfu", 5) == 0) {
  vol->sfu_emul = 0;
+ } else if (strnicmp(data, "nodfs", 5) == 0) {
+ vol->nodfs = 1;
  } else if (strnicmp(data, "posixpaths", 10) == 0) {
  vol->posix_paths = 1;
  } else if (strnicmp(data, "noposixpaths", 12) == 0) {
@@ -2167,6 +2170,10 @@ cifs_mount(struct super_block *sb, struc
  volume_info.UNC,
  tcon, cifs_sb->local_nls);
  cFYI(1, ("CIFS Tcon rc = %d", rc));
+ if (volume_info.nodfs) {
+ tcon->Flags &= ~SMB_SHARE_IS_IN_DFS;
+ cFYI(1, ("CIFS DFS disabeld (%d)", tcon->Flags));
+ }
  }
  if (!rc) {
  atomic_inc(&pSesInfo->inUse);
--
_______________________________________________
linux-cifs-client mailing list
[hidden email]
https://lists.samba.org/mailman/listinfo/linux-cifs-client

[Jeff Layton-2]

On Thu, 16 Oct 2008 08:17:43 +0200
Martin Koegler <[hidden email]> wrote:

> Older samba server (eg. 3.0.24 from Debian etch) don't work correctly,
> if DFS paths are used. Such server claim that they support DFS, but fail
> to process some requests with DFS paths. Starting with Linux 2.6.26,
> the cifs clients starts sending DFS paths in such situations, rendering
> it unuseable with older samba servers.
>
> The nodfs mount options forces a share to be used with non DFS paths,
> even if the server claims, that it supports it.
>
> Signed-off-by: Martin Koegler <[hidden email]>
> ---
>
> This patch makes my home directory on a samba server useable for Linux 2.6.26
> again. It still needs more testing. Should apply to linux 2.6.27 too.
>
> Would it be possible to support something like this in the offical CIFS client?
>
> diff a/fs/cifs/connect.c b/fs/cifs/connect.c
> --- a/fs/cifs/connect.c
> +++ b/fs/cifs/connect.c
> @@ -90,6 +90,7 @@ struct smb_vol {
>   bool nocase:1;     /* request case insensitive filenames */
>   bool nobrl:1;      /* disable sending byte range locks to srv */
>   bool seal:1;       /* request transport encryption on share */
> + bool nodfs:1;
>   unsigned int rsize;
>   unsigned int wsize;
>   unsigned int sockopt;
> @@ -1223,6 +1224,8 @@ cifs_parse_mount_options(char *options,
>   vol->sfu_emul = 1;
>   } else if (strnicmp(data, "nosfu", 5) == 0) {
>   vol->sfu_emul = 0;
> + } else if (strnicmp(data, "nodfs", 5) == 0) {
> + vol->nodfs = 1;
>   } else if (strnicmp(data, "posixpaths", 10) == 0) {
>   vol->posix_paths = 1;
>   } else if (strnicmp(data, "noposixpaths", 12) == 0) {
> @@ -2167,6 +2170,10 @@ cifs_mount(struct super_block *sb, struc
>   volume_info.UNC,
>   tcon, cifs_sb->local_nls);
>   cFYI(1, ("CIFS Tcon rc = %d", rc));
> + if (volume_info.nodfs) {
> + tcon->Flags &= ~SMB_SHARE_IS_IN_DFS;
> + cFYI(1, ("CIFS DFS disabeld (%d)", tcon->Flags));
> + }
>   }
>   if (!rc) {
>   atomic_inc(&pSesInfo->inUse);
> --

Looks reasonable to me. Igor did most of the DFS work, so I cc'ed him
here. Igor, any thoughts?

--
Jeff Layton <[hidden email]>
_______________________________________________
linux-cifs-client mailing list
[hidden email]
https://lists.samba.org/mailman/listinfo/linux-cifs-client

[Steve French-2]

Sounds reasonable to me - I will queue it up or something similar
unless there are any objections.

On Thu, Oct 16, 2008 at 6:43 AM, Jeff Layton <[hidden email]> wrote:

> On Thu, 16 Oct 2008 08:17:43 +0200
> Martin Koegler <[hidden email]> wrote:
>
>> Older samba server (eg. 3.0.24 from Debian etch) don't work correctly,
>> if DFS paths are used. Such server claim that they support DFS, but fail
>> to process some requests with DFS paths. Starting with Linux 2.6.26,
>> the cifs clients starts sending DFS paths in such situations, rendering
>> it unuseable with older samba servers.
>>
>> The nodfs mount options forces a share to be used with non DFS paths,
>> even if the server claims, that it supports it.
>>
>> Signed-off-by: Martin Koegler <[hidden email]>
>> ---
>>
>> This patch makes my home directory on a samba server useable for Linux 2.6.26
>> again. It still needs more testing. Should apply to linux 2.6.27 too.
>>
>> Would it be possible to support something like this in the offical CIFS client?
>>
>> diff a/fs/cifs/connect.c b/fs/cifs/connect.c
>> --- a/fs/cifs/connect.c
>> +++ b/fs/cifs/connect.c
>> @@ -90,6 +90,7 @@ struct smb_vol {
>>       bool nocase:1;     /* request case insensitive filenames */
>>       bool nobrl:1;      /* disable sending byte range locks to srv */
>>       bool seal:1;       /* request transport encryption on share */
>> +     bool nodfs:1;
>>       unsigned int rsize;
>>       unsigned int wsize;
>>       unsigned int sockopt;
>> @@ -1223,6 +1224,8 @@ cifs_parse_mount_options(char *options,
>>                       vol->sfu_emul = 1;
>>               } else if (strnicmp(data, "nosfu", 5) == 0) {
>>                       vol->sfu_emul = 0;
>> +             } else if (strnicmp(data, "nodfs", 5) == 0) {
>> +                     vol->nodfs = 1;
>>               } else if (strnicmp(data, "posixpaths", 10) == 0) {
>>                       vol->posix_paths = 1;
>>               } else if (strnicmp(data, "noposixpaths", 12) == 0) {
>> @@ -2167,6 +2170,10 @@ cifs_mount(struct super_block *sb, struc
>>                                               volume_info.UNC,
>>                                               tcon, cifs_sb->local_nls);
>>                                       cFYI(1, ("CIFS Tcon rc = %d", rc));
>> +                                     if (volume_info.nodfs) {
>> +                                             tcon->Flags &= ~SMB_SHARE_IS_IN_DFS;
>> +                                             cFYI(1, ("CIFS DFS disabeld (%d)", tcon->Flags));
>> +                                     }
>>                               }
>>                               if (!rc) {
>>                                       atomic_inc(&pSesInfo->inUse);
>> --
>
>
> Looks reasonable to me. Igor did most of the DFS work, so I cc'ed him
> here. Igor, any thoughts?
>
> --
> Jeff Layton <[hidden email]>
>

--
Thanks,

Steve
_______________________________________________
linux-cifs-client mailing list
[hidden email]
https://lists.samba.org/mailman/listinfo/linux-cifs-client

[Igor Mammedov]

On Thu, Oct 16, 2008 at 3:43 PM, Jeff Layton <[hidden email]> wrote:

> On Thu, 16 Oct 2008 08:17:43 +0200
> Martin Koegler <[hidden email]> wrote:
>
>> Older samba server (eg. 3.0.24 from Debian etch) don't work correctly,
>> if DFS paths are used. Such server claim that they support DFS, but fail
>> to process some requests with DFS paths. Starting with Linux 2.6.26,
>> the cifs clients starts sending DFS paths in such situations, rendering
>> it unuseable with older samba servers.
>>
>> The nodfs mount options forces a share to be used with non DFS paths,
>> even if the server claims, that it supports it.
>>
>> Signed-off-by: Martin Koegler <[hidden email]>
>> ---
>>
>> This patch makes my home directory on a samba server useable for Linux 2.6.26
>> again. It still needs more testing. Should apply to linux 2.6.27 too.
>>
>> Would it be possible to support something like this in the offical CIFS client?
>>
>> diff a/fs/cifs/connect.c b/fs/cifs/connect.c
>> --- a/fs/cifs/connect.c
>> +++ b/fs/cifs/connect.c
>> @@ -90,6 +90,7 @@ struct smb_vol {
>>       bool nocase:1;     /* request case insensitive filenames */
>>       bool nobrl:1;      /* disable sending byte range locks to srv */
>>       bool seal:1;       /* request transport encryption on share */
>> +     bool nodfs:1;
>>       unsigned int rsize;
>>       unsigned int wsize;
>>       unsigned int sockopt;
>> @@ -1223,6 +1224,8 @@ cifs_parse_mount_options(char *options,
>>                       vol->sfu_emul = 1;
>>               } else if (strnicmp(data, "nosfu", 5) == 0) {
>>                       vol->sfu_emul = 0;
>> +             } else if (strnicmp(data, "nodfs", 5) == 0) {
>> +                     vol->nodfs = 1;
>>               } else if (strnicmp(data, "posixpaths", 10) == 0) {
>>                       vol->posix_paths = 1;
>>               } else if (strnicmp(data, "noposixpaths", 12) == 0) {
>> @@ -2167,6 +2170,10 @@ cifs_mount(struct super_block *sb, struc
>>                                               volume_info.UNC,
>>                                               tcon, cifs_sb->local_nls);
>>                                       cFYI(1, ("CIFS Tcon rc = %d", rc));
>> +                                     if (volume_info.nodfs) {
>> +                                             tcon->Flags &= ~SMB_SHARE_IS_IN_DFS;
>> +                                             cFYI(1, ("CIFS DFS disabeld (%d)", tcon->Flags));
>> +                                     }
>>                               }
>>                               if (!rc) {
>>                                       atomic_inc(&pSesInfo->inUse);
>> --
>
>
> Looks reasonable to me. Igor did most of the DFS work, so I cc'ed him
> here. Igor, any thoughts?

Acked.

> --
> Jeff Layton <[hidden email]>
>
_______________________________________________
linux-cifs-client mailing list
[hidden email]
https://lists.samba.org/mailman/listinfo/linux-cifs-client

[Steve French-2]

Merged - after cleaning up some long (>80) lines and adding an
fs/cifs/README entry

On Thu, Oct 16, 2008 at 11:23 AM, Q (Igor Mammedov) <[hidden email]> wrote:

> On Thu, Oct 16, 2008 at 3:43 PM, Jeff Layton <[hidden email]> wrote:
>> On Thu, 16 Oct 2008 08:17:43 +0200
>> Martin Koegler <[hidden email]> wrote:
>>
>>> Older samba server (eg. 3.0.24 from Debian etch) don't work correctly,
>>> if DFS paths are used. Such server claim that they support DFS, but fail
>>> to process some requests with DFS paths. Starting with Linux 2.6.26,
>>> the cifs clients starts sending DFS paths in such situations, rendering
>>> it unuseable with older samba servers.
>>>
>>> The nodfs mount options forces a share to be used with non DFS paths,
>>> even if the server claims, that it supports it.
>>>
>>> Signed-off-by: Martin Koegler <[hidden email]>
>>> ---
>>>
>>> This patch makes my home directory on a samba server useable for Linux 2.6.26
>>> again. It still needs more testing. Should apply to linux 2.6.27 too.
>>>
>>> Would it be possible to support something like this in the offical CIFS client?
>>>
>>> diff a/fs/cifs/connect.c b/fs/cifs/connect.c
>>> --- a/fs/cifs/connect.c
>>> +++ b/fs/cifs/connect.c
>>> @@ -90,6 +90,7 @@ struct smb_vol {
>>>       bool nocase:1;     /* request case insensitive filenames */
>>>       bool nobrl:1;      /* disable sending byte range locks to srv */
>>>       bool seal:1;       /* request transport encryption on share */
>>> +     bool nodfs:1;
>>>       unsigned int rsize;
>>>       unsigned int wsize;
>>>       unsigned int sockopt;
>>> @@ -1223,6 +1224,8 @@ cifs_parse_mount_options(char *options,
>>>                       vol->sfu_emul = 1;
>>>               } else if (strnicmp(data, "nosfu", 5) == 0) {
>>>                       vol->sfu_emul = 0;
>>> +             } else if (strnicmp(data, "nodfs", 5) == 0) {
>>> +                     vol->nodfs = 1;
>>>               } else if (strnicmp(data, "posixpaths", 10) == 0) {
>>>                       vol->posix_paths = 1;
>>>               } else if (strnicmp(data, "noposixpaths", 12) == 0) {
>>> @@ -2167,6 +2170,10 @@ cifs_mount(struct super_block *sb, struc
>>>                                               volume_info.UNC,
>>>                                               tcon, cifs_sb->local_nls);
>>>                                       cFYI(1, ("CIFS Tcon rc = %d", rc));
>>> +                                     if (volume_info.nodfs) {
>>> +                                             tcon->Flags &= ~SMB_SHARE_IS_IN_DFS;
>>> +                                             cFYI(1, ("CIFS DFS disabeld (%d)", tcon->Flags));
>>> +                                     }
>>>                               }
>>>                               if (!rc) {
>>>                                       atomic_inc(&pSesInfo->inUse);
>>> --
>>
>>
>> Looks reasonable to me. Igor did most of the DFS work, so I cc'ed him
>> here. Igor, any thoughts?
>
> Acked.
>
>> --
>> Jeff Layton <[hidden email]>
>>
>

--
Thanks,

Steve
_______________________________________________
linux-cifs-client mailing list
[hidden email]
https://lists.samba.org/mailman/listinfo/linux-cifs-client

[Michael-165]

This patch is a good one. It helped me with the same problem on my Samba
3.0.24 server and 2.6.27 client.

I am just wondering, why it doesn't appear in the newly released kernels. It
is neither in 2.6.27.2 nor in 2.6.27.3. It is not even scheduled for
2.6.27.4 ???


Steve French wrote:

> Merged - after cleaning up some long (>80) lines and adding an
> fs/cifs/README entry
>


_______________________________________________
linux-cifs-client mailing list
[hidden email]
https://lists.samba.org/mailman/listinfo/linux-cifs-client

[Timothee Besset-2]

If anything, I'm pretty sure you won't see this before 2.6.28. This does
not classify as a bug/regression fix, it's new functionality (2.6.28
merge windows is open however)

TTimo

Michael wrote:

> This patch is a good one. It helped me with the same problem on my Samba
> 3.0.24 server and 2.6.27 client.
>
> I am just wondering, why it doesn't appear in the newly released kernels. It
> is neither in 2.6.27.2 nor in 2.6.27.3. It is not even scheduled for
> 2.6.27.4 ???
>
>
> Steve French wrote:
>
>  
>> Merged - after cleaning up some long (>80) lines and adding an
>> fs/cifs/README entry
>>
>>    
>
>
> _______________________________________________
> linux-cifs-client mailing list
> [hidden email]
> https://lists.samba.org/mailman/listinfo/linux-cifs-client
>  

_______________________________________________
linux-cifs-client mailing list
[hidden email]
https://lists.samba.org/mailman/listinfo/linux-cifs-client

[Michael-165]

Well, you may be right from a technical point of view, but from a users
point of view it is very much a regression. Before kernel 2.6.25, I could
use my Samba server perfectly. Now it does no longer work (and there is no
workaround).

So, either the Samba guys correct the DFS coding in their 3.0.24 or we need
the nodfs patch in the kernel.

At the moment I resort to compiling the kernels myself (although I am no
kernel hacker :-( ), but that shouldn't be a long term perspective, I hope.

Timothee Besset wrote:

> If anything, I'm pretty sure you won't see this before 2.6.28. This does
> not classify as a bug/regression fix, it's new functionality (2.6.28
> merge windows is open however)
>
> TTimo
>
> Michael wrote:
>> This patch is a good one. It helped me with the same problem on my Samba
>> 3.0.24 server and 2.6.27 client.
>>
>> I am just wondering, why it doesn't appear in the newly released kernels.
>> It is neither in 2.6.27.2 nor in 2.6.27.3. It is not even scheduled for
>> 2.6.27.4 ???
>>
>>
>> Steve French wrote:
>>
>>  
>>> Merged - after cleaning up some long (>80) lines and adding an
>>> fs/cifs/README entry
>>>
>>>    
>>
>>
>> _______________________________________________
>> linux-cifs-client mailing list
>> [hidden email]
>> https://lists.samba.org/mailman/listinfo/linux-cifs-client
>>

_______________________________________________
linux-cifs-client mailing list
[hidden email]
https://lists.samba.org/mailman/listinfo/linux-cifs-client

[Timothee Besset-2]

I am running on a patched 2.6.27 kernel myself for the same reasons. I
don't think any distro (Fedora/Debian/Ubuntu) has a stable release that
is running anything newer than 2.6.24 atm, and they might pickup this
patch and backport it to their own distro-flavored kernels if they want to.

We'd all like things to go faster and run better, but from what I know
about patches going into kernel releases, I doubt that one will make it
into a 2.6.27.x - after all we're looking at several months of not
working (since 2.6.25 I believe), and people would still need to add the
nodfs option to their mounts to fix things .. for a bug that is on the
samba server side, so this is new functionality, which incidentally
allows a workaround for some buggy servers .. just saying :)

TTimo

Michael wrote:

> Well, you may be right from a technical point of view, but from a users
> point of view it is very much a regression. Before kernel 2.6.25, I could
> use my Samba server perfectly. Now it does no longer work (and there is no
> workaround).
>
> So, either the Samba guys correct the DFS coding in their 3.0.24 or we need
> the nodfs patch in the kernel.
>
> At the moment I resort to compiling the kernels myself (although I am no
> kernel hacker :-( ), but that shouldn't be a long term perspective, I hope.
>
> Timothee Besset wrote:
>
>  
>> If anything, I'm pretty sure you won't see this before 2.6.28. This does
>> not classify as a bug/regression fix, it's new functionality (2.6.28
>> merge windows is open however)
>>
>> TTimo
>>
>> Michael wrote:
>>    
>>> This patch is a good one. It helped me with the same problem on my Samba
>>> 3.0.24 server and 2.6.27 client.
>>>
>>> I am just wondering, why it doesn't appear in the newly released kernels.
>>> It is neither in 2.6.27.2 nor in 2.6.27.3. It is not even scheduled for
>>> 2.6.27.4 ???
>>>
>>>
>>> Steve French wrote:
>>>
>>>  
>>>      
>>>> Merged - after cleaning up some long (>80) lines and adding an
>>>> fs/cifs/README entry
>>>>
>>>>    
>>>>        
>>> _______________________________________________
>>> linux-cifs-client mailing list
>>> [hidden email]
>>> https://lists.samba.org/mailman/listinfo/linux-cifs-client
>>>
>>>      
>
>
> _______________________________________________
> linux-cifs-client mailing list
> [hidden email]
> https://lists.samba.org/mailman/listinfo/linux-cifs-client
>  

_______________________________________________
linux-cifs-client mailing list
[hidden email]
https://lists.samba.org/mailman/listinfo/linux-cifs-client

[Jeff Layton-2]

On Fri, 24 Oct 2008 17:30:02 +0200
Michael <[hidden email]> wrote:

> Well, you may be right from a technical point of view, but from a users
> point of view it is very much a regression. Before kernel 2.6.25, I could
> use my Samba server perfectly. Now it does no longer work (and there is no
> workaround).
>
> So, either the Samba guys correct the DFS coding in their 3.0.24 or we need
> the nodfs patch in the kernel.
>

...or you need to update to a newer samba version. AFAIK, more recent 3.0.z
samba versions have shaken out most of these bugs.

--
Jeff Layton <[hidden email]>
_______________________________________________
linux-cifs-client mailing list
[hidden email]
https://lists.samba.org/mailman/listinfo/linux-cifs-client

[Timothee Besset-2]

Jeff Layton wrote:

> On Fri, 24 Oct 2008 17:30:02 +0200
> Michael <[hidden email]> wrote:
>
>  
>> Well, you may be right from a technical point of view, but from a users
>> point of view it is very much a regression. Before kernel 2.6.25, I could
>> use my Samba server perfectly. Now it does no longer work (and there is no
>> workaround).
>>
>> So, either the Samba guys correct the DFS coding in their 3.0.24 or we need
>> the nodfs patch in the kernel.
>>
>>    
>
> ...or you need to update to a newer samba version. AFAIK, more recent 3.0.z
> samba versions have shaken out most of these bugs.
>
>  

..or more simply turn off DFS on the server (I couldn't do that though,
it breaks things for my Windows users)

TTimo

_______________________________________________
linux-cifs-client mailing list
[hidden email]
https://lists.samba.org/mailman/listinfo/linux-cifs-client

[Michael-165]

Thanks for the suggestions :-)

The server is debian etch based. Unfortunately, 3.0.24 is the latest stable
debian package. The package from the testing repository doesn't run on
debian etch.

Unfortunately, I also have a windows computer on the net (otherwise I would
have used nfs).

Let's hope that the patch makes it into 2.6.28.

Michael

Timothee Besset wrote:

> Jeff Layton wrote:
>
>> ...or you need to update to a newer samba version. AFAIK, more recent
>> 3.0.z samba versions have shaken out most of these bugs.
>>
>>  
> ..or more simply turn off DFS on the server (I couldn't do that though,
> it breaks things for my Windows users)
>
> TTimo

_______________________________________________
linux-cifs-client mailing list
[hidden email]
https://lists.samba.org/mailman/listinfo/linux-cifs-client