Re: [PATCH 4/4] server: Add primitive support for setting and getting the security descriptor of files based on their Unix permissions.
Robert Shearman <rob(a)codeweavers.com> writes:
+ /* assume we are the owner */ + if (st.st_mode & S_IRWXU) + {
Why are you assuming that? There doesn't seem to be much sense in checking permissions without checking the owner. -- Alexandre Julliard julliard(a)winehq.org
Alexandre Julliard wrote:
Robert Shearman <rob(a)codeweavers.com> writes:
+ /* assume we are the owner */ + if (st.st_mode & S_IRWXU) + {
Why are you assuming that? There doesn't seem to be much sense in checking permissions without checking the owner.
We don't have a mapping between NT SIDs and Unix user IDs in Wine, so if the owner is different then there isn't much we can do with it. -- Rob Shearman
On Wednesday 03 October 2007 11:18:48 Robert Shearman wrote:
Alexandre Julliard wrote:
Robert Shearman <rob(a)codeweavers.com> writes:
+ /* assume we are the owner */ + if (st.st_mode & S_IRWXU) + {
Why are you assuming that? There doesn't seem to be much sense in checking permissions without checking the owner.
We don't have a mapping between NT SIDs and Unix user IDs in Wine, so if the owner is different then there isn't much we can do with it.
Would it make sense to try and get winbind to do that for us, if possible? Or would we be unable to use the results of that in a sensible way? Jerry Carter is currently working on an LGPL winbind client library, we could use that. In the long run, winbind will convert to an idl-based protocol and we could just talk to winbindd directly, thus dropping a dependency. Cheers, Kai -- Kai Blin WorldForge developer http://www.worldforge.org/ Wine developer http://wiki.winehq.org/KaiBlin Samba team member http://www.samba.org/samba/team/ -- Will code for cotton.
Kai Blin wrote:
On Wednesday 03 October 2007 11:18:48 Robert Shearman wrote:
Alexandre Julliard wrote:
Robert Shearman <rob(a)codeweavers.com> writes:
+ /* assume we are the owner */ + if (st.st_mode & S_IRWXU) + {
Why are you assuming that? There doesn't seem to be much sense in checking permissions without checking the owner.
We don't have a mapping between NT SIDs and Unix user IDs in Wine, so if the owner is different then there isn't much we can do with it.
Would it make sense to try and get winbind to do that for us, if possible? Or would we be unable to use the results of that in a sensible way?
Jerry Carter is currently working on an LGPL winbind client library, we could use that. In the long run, winbind will convert to an idl-based protocol and we could just talk to winbindd directly, thus dropping a dependency.
Good suggestion for a discussion at WineConf :-) However, we can't do any blocking calls while executing a server call on the behalf of a client, otherwise it will cause performance problems for all the applications that the user is running. Also, we don't have an IDL compiler / runtime that is compatible with wineserver. Further more, our mapping of users / groups will be slightly different from Samba's due to most Win32 installers requiring the user to be an administrator. I think all of these issues can be worked around, but it's not something we can rush into (and in case the Samba developers are doing this just for us, an winbind client library is not going to useful.) -- Rob Shearman
On Wednesday 03 October 2007 12:24:52 Robert Shearman wrote:
Kai Blin wrote:
Would it make sense to try and get winbind to do that for us, if possible? Or would we be unable to use the results of that in a sensible way?
Jerry Carter is currently working on an LGPL winbind client library, we could use that. In the long run, winbind will convert to an idl-based protocol and we could just talk to winbindd directly, thus dropping a dependency.
Good suggestion for a discussion at WineConf :-)
Collecting topics for my presentation all the time.
However, we can't do any blocking calls while executing a server call on the behalf of a client, otherwise it will cause performance problems for all the applications that the user is running.
In Samba4's winbind, all the calls are async, in Samba3's winbindd, all calls that can block over an extended period of time have an async version.
Also, we don't have an IDL compiler / runtime that is compatible with wineserver. Further more, our mapping of users / groups will be slightly different from Samba's due to most Win32 installers requiring the user to be an administrator.
Possibly this could be fixed by making sure the user always is a local administrator, I take. Or we can add a "switch to administrator" button.
I think all of these issues can be worked around, but it's not something we can rush into (and in case the Samba developers are doing this just for us, an winbind client library is not going to useful.)
No, I think Jerry has some other use cases in mind with that. No fuss. Cheers, Kai -- Kai Blin WorldForge developer http://www.worldforge.org/ Wine developer http://wiki.winehq.org/KaiBlin Samba team member http://www.samba.org/samba/team/ -- Will code for cotton.
Robert Shearman <rob(a)codeweavers.com> writes:
We don't have a mapping between NT SIDs and Unix user IDs in Wine, so if the owner is different then there isn't much we can do with it.
You can't report the correct user SID, but you can still report more correct permissions for the current user by checking IRWXO instead of IRWXU if the user is not the owner. Another possibility would be to have some sort of "other user" SID that all uids map to except the current one. -- Alexandre Julliard julliard(a)winehq.org
Alexandre Julliard wrote:
Robert Shearman <rob(a)codeweavers.com> writes:
We don't have a mapping between NT SIDs and Unix user IDs in Wine, so if the owner is different then there isn't much we can do with it.
You can't report the correct user SID, but you can still report more correct permissions for the current user by checking IRWXO instead of IRWXU if the user is not the owner. Another possibility would be to have some sort of "other user" SID that all uids map to except the current one.
The SID for "anonymous logon" seems to be the most appropriate choice for the latter suggestion. Also, tests on Windows show that "Builtin Users" may be a better choice to map the Unix others permissions to than "Everyone". -- Rob Shearman
participants (3)
-
Alexandre Julliard -
Kai Blin -
Robert Shearman