On Wed, Mar 15, 2017 at 8:49 PM, Andrew Morgan <[email protected]> wrote:
> On 03/15/2017 05:20 PM, Jean-Philippe Ouellet wrote:
> Thanks for the insight on how existing programs already utilize the EFA.
> I did not know some set them as untrusted.
>
>> Linux uses "security", "system", "trusted", and "user".
>
> I assume you're referring to this line on the freedesktop spec? On my
> own system however, I don't seem to find the same mark of trust when
> downloading a file:
>
> * Download repo file .zip from GitHub *
>
> [user@untrusted ~]$ getfattr -d ~/Downloads/a53420.zip
> [user@untrusted ~]$
>
> Firefox hasn't added anything in terms of extended attributes to the
> file, and I haven't changed any settings related to this as far as I
> know. Perhaps I'm doing something wrong here?

Ah, indeed. I can reproduce your observed behavior.

I use chrome nearly exclusively and hadn't noticed.

>> you likely do not need to worry about a way to set them manually, but
>> rather a way to determine how to handle them when they are found.
>
> This would be nice but rather tricky to implement I feel, as we would
> have to keep a record of which program namespaces and which values would
> represent a low level of trust and act upon that. This is of course
> assuming I am only able to read these attributes based on their textual
> values, but so far I do not know any other way.
>
> When a program fails to mark that trust correctly or at all, it is
> potentially putting the user in danger or at the least
> inconveniencing/confusing them.
>
> I still believe that only setting a file to open in a certain domain
> based on the user's own actions, such as setting an option for that
> specific file, is clearest for the user in what is going to happen when
> they open a particular file.

Upon reviewing the arguments in [1] again, and checking more programs
to see if they set the xdg flags, I now agree with you.

[1]: https://github.com/QubesOS/qubes-issues/issues/441#issuecomment-253731556

>> The task then would be to extend qfilecopy [1] to transfer these attributes
>
> In the issue on the GSoC page it states:
>
>> Allow setting persistent flag for a file that should be opened in specific 
>> way (“locally”); this flag should local to the VM - it shouldn’t be possible 
>> to preserve (or even fabricate) the flag while transferring the file from/to 
>> VM.
>
> I assuming by this bullet point we don't want the attribute transferred?
> From my own testing the flags are already stripped when using qfilecopy,
> thus we shouldn't have to do any work here, unless I'm misunderstanding
> you? :)

I have not thought this through, and defer to those who have.

-- 
You received this message because you are subscribed to the Google Groups 
"qubes-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
To post to this group, send email to [email protected].
To view this discussion on the web visit 
https://groups.google.com/d/msgid/qubes-devel/CABQWM_Dk%2BLRpgbdO2Mg1CvOdSq%3DYr75VwMzp8Fc-AM%2BWM7VXHQ%40mail.gmail.com.
For more options, visit https://groups.google.com/d/optout.

Reply via email to