[cifs-protocol] Help regarding the security descriptor creation algorithms
Nadezhda Ivanova
nadezhda.ivanova at postpath.com
Fri Aug 21 02:09:03 MDT 2009
Hi Obaid,
We will need some time to make some tests and see if there is still behavior we do not understand. If you prefer, we can consider this issue closed, and if there is something more I will raise a new issue. Thank you for your assistance!
Best Regards,
Nadezhda Ivanova
From: Obaid Farooqi [mailto:obaidf at microsoft.com]
Sent: Thursday, August 20, 2009 6:27 PM
To: Nadezhda Ivanova
Cc: 'pfif at tridgell.net'; 'cifs-protocol at samba.org'
Subject: RE: Help regarding the security descriptor creation algorithms
Hi Nadezhda:
Please let me know if I answered your question. As mentioned before, if you have any further questions on ACE inheritance, please feel free to contact us.
Regards,
Obaid Farooqi
Sr. Support Escalation Engineer | Microsoft
From: Obaid Farooqi
Sent: Thursday, August 13, 2009 5:36 PM
To: 'Nadezhda Ivanova'
Cc: 'pfif at tridgell.net'; 'cifs-protocol at samba.org'
Subject: RE: Help regarding the security descriptor creation algorithms
Hi Nadezhda:
The same information that I sent you will appear in a future version of the document MS-DTYP. The format may be different but content will be more or less same.
If you still have questions about ACE inheritance, please let me know. The intention of the reply is to allow you to move ahead with your implementation and not wait for the documentation. We don’t have a publically available date that I can provide you for the next documentation refresh at this time. However, we refresh the MSDN documentation routinely.
Regards,
Obaid Farooqi
Sr. Support Escalation Engineer | Microsoft
From: Nadezhda Ivanova [mailto:nadezhda.ivanova at postpath.com]
Sent: Wednesday, August 12, 2009 3:10 AM
To: Obaid Farooqi
Subject: RE: Help regarding the security descriptor creation algorithms
Hi Obaid,
Thank you for the information. It confirms our current knowledge about how ACE’s are being inherited and we will adjust our implementation accordingly while waiting for the new version of MS-DTYP. Do you know whe we can expect that?
Regards,
Nadezhda Ivanova
From: Obaid Farooqi [mailto:obaidf at microsoft.com]
Sent: Tuesday, August 11, 2009 6:32 PM
To: Nadezhda Ivanova
Cc: 'pfif at tridgell.net'; 'cifs-protocol at samba.org'
Subject: RE: Help regarding the security descriptor creation algorithms
Hi Nadezhda:
I just realized that the table in my previous email may not appear correctly. So, I am attaching a PDF version of my reply. Please let me know if you have any difficulty reading it.
Regards,
Obaid Farooqi
Sr. Support Escalation Engineer | Microsoft
From: Obaid Farooqi
Sent: Tuesday, August 11, 2009 10:02 AM
To: Nadezhda Ivanova
Cc: pfif at tridgell.net; cifs-protocol at samba.org
Subject: RE: Help regarding the security descriptor creation algorithms
Hi Nadezhda:
We have finished our investigation on your question about ComputeInheritedACLfromParent algorithm. The documentation is in progress for a newer version of the algorithm that will appear in a future version of MS-DTYP.
For your implementation, you can use the following rules to derive inherited ACE’s from the parent object.
There are seven flags that can appear in an ACE. Of the seven flags, the following pertains to inheritance:
CI CONTAINER_INHERIT_ACE
OI OBJECT_INHERIT_ACE
NP NO_PROPAGATE_INHERIT_ACE
IO INHERIT_ONLY_ACE
ID INHERITED_ACE
IO and ID do not play a part when it comes to making decisions about inheritance. ID flag is added to any ACE that is inherited to mention the fact that it was inherited. IO flag is used to render an ACE ineffective for the child that inherits the ACE. An ACE that has IO flag can be inherited but the decision is based on other flags, if present. As such, the following table mostly ignores ID and IO flags. The following table outlines what would be the flags of the inherited ACE based on the flags that parent has.
Parent ACE Flags Child Container Object Child Leaf Object
No Flags, IO No Inheritance No Inheritance
OI IO,OI Inherited, No flags
OI,NP No Inheritance Inherited, No flags
CI CI No Inheritance
CI,NP Inherited, No flags No Inheritance
CI,OI IO,CI,OI Inherited, No flags
CI,OI,NP Inherited, No flags Inherited, No flags
For the cases in which a container inherits an ACE that is both effective on the container and inheritable by its descendents, the container may inherit two ACEs. This occurs when inheritable ACE contains generic information. The container inherits an ACE with an additional IO flag with generic information and an effective-only ACE in which the generic information has been mapped.
The following page at MSDN also has similar information that you may find helpful:
http://msdn.microsoft.com/en-us/library/aa374924(VS.85).aspx
Please let me know if this answers your question. If yes, then I’ll consider this issue resolved.
Regards,
Obaid Farooqi
Sr. Support Escalation Engineer | Microsoft
From: Nadezhda Ivanova [mailto:nadezhda.ivanova at postpath.com]
Sent: Friday, July 10, 2009 6:09 AM
To: Interoperability Documentation Help
Cc: pfif at tridgell.net; cifs-protocol at samba.org
Subject: Help regarding the security descriptor creation algorithms
Hi,
I have been working on implementing correct nTSecurityDeascriptor creation in the directory service of Samba 4, and have come upon a problem in the ComputeInheritedACLfromParent subroutine described in MS-DTYP 2.5.2.6. The way the algorithm is described, the purpose of this algorithm is to determine which ACE’s from an object’s parent are to be inherited by the new object actively, and which are to be inherited only. The ComputeInheritedACLfromParent as described, walks the parent ACL twice. The first time it determines the active inherited ACE’s, the second time the ones that are inherited but inactive.
I have been testing our implementation with the CN=Schema partition, as the attributes and objects by default are not given a security descriptor during creation, and the defaultSecurityDescriptor of attribute-Schema is empty DACL and SACL.
So, they inherit all their DACL ACE’s from their parent, CN=Schema.
In a Win2008R2, CN=Schema has three inheritable DACL ACE’s:
(A;CI;RPLCLORC;;;AU)
(A;CI;RPWPCRCCLCLORCWOWDSW;;;SA)
(A;CI;RPWPCRCCDCLCLORCWOWDSDDTSW;;;SY)
ComputeInheritedACLfromParent has the following arguments:
ACL: ACL that contains the parent's ACEs from which to compute the inherited ACL.
IsContainerObject: TRUE if the object is a container, FALSE otherwise.
ObjectTypes: Array of GUIDs for the object type being created.
So if we invoke the ComputeInheritedACLfromParent with the above DACL,and isConatinerObject = true (According to MS-ADTS 7.1.3, true is always the value), the first walk of the input
Initialize ExplicitACL to Empty ACL
FOR each ACE in ACL DO
IF ACE.Flags contains INHERIT_ONLY
THEN
CONTINUE
ENDIF
IF(((ACE.Flags contains CONTAINER_INHERIT) AND
(IsContainerObject = TRUE))OR
((ACE.Flags contains OBJECT_INHERIT) AND (IsContainerObject = FALSE)))
THEN
CASE ACE.Type OF
ALLOW:
DENY:
Set NewACE to ACE
Set NewACE.Flags to INHERITED
Append NewACE to ExplicitACL
OBJECT_ALLOW:
OBJECT_DENY:
IF (ObjectTypes contains ACE.ObjectGUID) THEN
Set NewACE to ACE
Set NewACE.Flags to INHERITED
Append NewACE to ExplicitACL
ENDIF
ENDCASE
ENDIF
END FOR
Will give:
D:AI(A;CIID;RPLCLORC;;;AU)(A;CIID;RPWPCRCCLCLORCWOWDSW;;;SA)(A;CIID;RPWPCRCCDCLCLORCWOWDSDDTSW;;;SY)
Which is as expected, as this is the DACL of all attributes and classes in Win 2008.
However, the algorithm then walks the input a second time:
Initialize InheritableACL to Empty ACL
IF (IsContainerObject = TRUE) THEN //In our case this is always true
FOR each ACE in ACL DO
IF ACE.Flags contains NO_PROPAGATE THEN //This flag is not set
CONTINUE
ENDIF
IF((ACE.Flags contains CONTAINER_INHERIT) OR
(ACE.Flags contains OBJECT_INHERIT))
THEN
Set NewACE to ACE
Add INHERITED to NewACE.Flags
Add INHERIT_ONLY to NewACE.Flags
Append NewACE to InheritableACL
ENDIF
END FOR
ENDIF
This second loop yields:
(A;CIIOID;RPLCLORC;;;AU)(A;CIIOID;RPWPCRCCLCLORCWOWDSW;;;SA)(A;CIIOID;RPWPCRCCDCLCLORCWOWDSDDTSW;;;SY)
Which after:
RETURN concatenation of ExplicitACL and InheritableACL
Makes the final DACL look like:
D:AI(A;CIID;RPLCLORC;;;AU)(A;CIID;RPWPCRCCLCLORCWOWDSW;;;SA)(A;CIID;RPWPCRCCDCLCLORCWOWDSDDTSW;;;SY)(A;CIIOID;RPLCLORC;;;AU)(A;CIIOID;RPWPCRCCLCLORCWOWDSW;;;SA)(A;CIIOID;RPWPCRCCDCLCLORCWOWDSDDTSW;;;SY)
So ACE’s are duplicated.
However, an attribute’s DACL in Win2008 does not have these last three ACE’s, so I am obviously missing something. How should the flow actually go with this same example in order to avoid this duplication? Or am I providing the wrong argument?
Best Regards,
Nadezhda Ivanova
Nadezhda Ivanova
Software EngineerSoftware Development
nadezhda.ivanova at postpath.com CISCO SYSTEMS BULGARIA EOOD
18 Macedonia Blvd. Sofia 1606
Bulgaria
Think before you print.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.samba.org/pipermail/cifs-protocol/attachments/20090821/7f640b45/attachment-0001.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: image/gif
Size: 87 bytes
Desc: not available
URL: <http://lists.samba.org/pipermail/cifs-protocol/attachments/20090821/7f640b45/attachment-0002.gif>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: image/gif
Size: 837 bytes
Desc: not available
URL: <http://lists.samba.org/pipermail/cifs-protocol/attachments/20090821/7f640b45/attachment-0003.gif>
More information about the cifs-protocol
mailing list