Hi all,

I created the Jira task https://bugs.tizen.org/jira/browse/TC-2154 to
track that issue.

Also, I threw some notes on the subject in the wiki:
https://wiki.tizen.org/wiki/User:Jobol/security_manager_improvements_proposal

Best regards
José Bollo

Le jeudi 20 novembre 2014 à 17:26 +0000, Schaufler, Casey a écrit :
> > -----Original Message-----
> > From: Dev [mailto:[email protected]] On Behalf Of Rafal Krypa
> > Sent: Thursday, November 20, 2014 8:30 AM
> > To: [email protected]
> > Subject: Re: [Dev] transferring files from and to a service
> > 
> > On 2014-11-20 10:12, Tomasz Swierczek wrote:
> > > Okay, so if this is the usecase, then we need to have, like I mentioned, 
> > > the
> > > mv_chmod_chown() API in security-server PLUS some way of authorizing
> > this
> > > action (ie. guard it with http://tizen.org/privilege/system like it was
> > > proposed or some other, more precise privilege, allowed only for
> > > system-level things).
> > >
> > >
> > > @Rafal, Casey, others - what is your opinion?
> > 
> > 
> > We could get a simpler interface if the file would be created already in the
> > final destination, at user's home directory.
> > I'mthinking about a security-manager API that would implement creation of
> > an empty file in the user's home directory and pass the file descriptor to 
> > the
> > caller. The caller (e.g. obexd) would then write to that file descriptor to 
> > fill it
> > with data. Things that security-manager should do in such case:
> > - take target user name, file name and type as argument
> > - verify privilege, can also trigger popup to get user consent
> > - check if the file already exists, guard against overwriting it (either by 
> > failing
> > orappending some string to the file name)
> > - create the empty file, set ownership and Smack label
> > - send the file descriptor back
> 
> In another thread someone explicitly wanted to create the file
> in the service's private place and only pass it to the user when
> the file was complete. Something to do with an App getting unhappy
> if it tried to read a partial database. The interface you suggest here
> might be good, but it doesn't provide the desired behavior.
> 
> Don't even think of trying to provide one interface that sets all
> of the attributes. It will be broken the instant some service decides
> it needs to use an interface we haven't thought of. It will be safe so
> long as the "secman_rename()" gets done last.
> 
> secman_create()
>       This is special. Create a file that the other
>       security manager function can act upon.
> secman_own()
>       If the service needs to use an existing file,
>       call this. If someone else called secman_create()
>       or secman_own(), this should fail.
> 
> The security manager will remember who "owns" a
> file (based on the UID and Smack label of the service)
> and only allow that service to make changes.
> 
> secman_chmod()
> secman_chacl()
> secman_chsmack()
> secman_chown()
> 
>       These should all be self explanatory. Add more
>       as the mood suits.
> 
> secman_rename()
> 
>       This is move, but use the system call name to
>       emphasize that you can't cross filesystem boundries.
> 
> secman_close()
> 
>       Declare that you're done with this file.
> 
> Oh, and just to be a pain, all paths must be absolute.
> 
>  
> > Would that satisfy your needs? Or ismore flexibility required for managing
> > the downloaded files?
> > _______________________________________________
> > Dev mailing list
> > [email protected]
> > https://lists.tizen.org/listinfo/dev
> _______________________________________________
> Dev mailing list
> [email protected]
> https://lists.tizen.org/listinfo/dev


_______________________________________________
Dev mailing list
[email protected]
https://lists.tizen.org/listinfo/dev

Reply via email to