Re: [PATCH v2 0/6] Add support for DFS in SMB2+

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



2017-02-23 6:43 GMT-08:00 Aurelien Aptel <aaptel@xxxxxxxx>:
> This series of patch tries to implement the get_dfs_refer operation
> for SMB2+.
>
> In SMB2+, DFS resolving is now done through an FSCTL (patch #3). The
> relevant Microsoft specifications for this are:
>
> MS-SMB2: 3.2.4.20   Application Requests an IO Control Code Operation
> MS-SMB2: 3.2.4.20.3 Application Requests DFS Referral Information
> MS-SMB2: 3.2.5.14   Receiving an SMB2 IOCTL Response
> MS-SMB2: 3.2.5.14.4 Handling a DFS Referral Information Response
>
> MS-DFSC: 2.2  Message Syntax (but really the whole document is useful)
>
> The DFS response payload however is identical. Patch #1 moves the
> DFS response parsing out of SMB1 code and makes it work on any version
> of SMB.
>
> DFS code has 2 "main" entry points: initial mounting and automount
> (when a DFS link is accessed/traversed).
>
> When automounting, cifs.ko calls build_path_from_dentry() which only
> makes treename-prefixed paths when the tcon has the
> SMB_IN_DFS_SHARE. This flag is checked in tcon->Flags which is never
> touched by SMB2 code as it sets tcon->share_flags on connexion.
>
> * CIFS requires to prefix all pathnames with the treename prefix when
>   connected to a DFS server, so the current build_path_from_dentry()
>   makes sense for CIFS.
> * For SMB2+ it seems only the Create request requires the treename
>   prefix. Simply making the function check for both CIFS SMB2+ flag
>   for DFS to add a prefix does not work as the server has different
>   expectations about which packet can have/require a DFS pathname.
>
> Patch #2 adds build_path_from_dentry_optional_prefix() with an extra
> bool arg to decide to prefix or not. The automount code path always
> ask for it. Patch #5 modifies SMB2_open() to add the treename prefix
> to the given path if the server requires it.
>
> As part of the mouting process a IPC tcon is made (I suspect we don't
> need it anymore in SMB3). This tcon doesn't respect the signing
> requirement. This is fixed by patch #4.
>
> Finally the SMB2+ implementation of the get_dfs_referral operation is
> added in all supported SMB versions in patch #6.
>
> I've sucessfuly tested this against a Windows Server 2016 DFS
> setup. Samba requires a patch as currently it only accepts DFS
> referrals requests on an IPC connexion, which is in violation of the
> spec. A patch was sent on samba-technical:
>
> https://lists.samba.org/archive/samba-technical/2017-February/118859.html
>
> Changes since v1:
>
> * The fsctl is made on the first available tcon to the server, which
>   is allowed by the spec. This way we avoid hardcoding additional
>   rules in SMB2_ioctl().
> * add signing flag in smb2 header if required
> * no copy/pasting, the DFS parsing code is now generic and works for
>   both SMB versions
> * add treename prefix in SMB2_open() when required
>
> Aurelien Aptel (6):
>   CIFS: move DFS response parsing out of SMB1 code
>   CIFS: add build_path_from_dentry_optional_prefix()
>   CIFS: implement get_dfs_refer for SMB2+
>   CIFS: set signing flag in SMB2+ TreeConnect if needed
>   CIFS: use DFS pathnames in SMB2+ Create requests
>   CIFS: enable get_dfs_refer for SMB2+
>
>  fs/cifs/cifs_dfs_ref.c |   4 +-
>  fs/cifs/cifspdu.h      |  16 ++++---
>  fs/cifs/cifsproto.h    |   7 +++
>  fs/cifs/cifssmb.c      | 119 +++----------------------------------------------
>  fs/cifs/dir.c          |  13 +++++-
>  fs/cifs/misc.c         | 105 +++++++++++++++++++++++++++++++++++++++++++
>  fs/cifs/smb2ops.c      |  76 +++++++++++++++++++++++++++++++
>  fs/cifs/smb2pdu.c      |  97 ++++++++++++++++++++++++++++++++--------
>  fs/cifs/smb2pdu.h      |   8 ++++
>  9 files changed, 305 insertions(+), 140 deletions(-)
>
> --
> 2.10.2
>
> --
> To unsubscribe from this list: send the line "unsubscribe linux-cifs" in
> the body of a message to majordomo@xxxxxxxxxxxxxxx
> More majordomo info at  http://vger.kernel.org/majordomo-info.html

I would merge patches #3 and #6 because a static function which is
introduced in #3 should raise compiler warnings. Other than that the
patchset looks good.

Reviewed-by: Pavel Shilovsky <pshilov@xxxxxxxxxxxxx>

-- 
Best regards,
Pavel Shilovsky
--
To unsubscribe from this list: send the line "unsubscribe linux-cifs" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html



[Linux USB Devel]     [Video for Linux]     [Linux Audio Users]     [Yosemite News]     [Linux Kernel]     [Linux SCSI]

  Powered by Linux