touch-packages team mailing list archive

[Jamie Strandboge <jamie@xxxxxxxxxx>]

This is likely too specific (ActiveConnection/0):
+  dbus (send)
+       bus=system
+       interface="org.freedesktop.DBus.Properties"
+       path=/org/freedesktop/NetworkManager/ActiveConnection/0
+       member=GetAll
+       peer=(name=org.freedesktop.NetworkManager),

This is too lenient:
+  dbus (send)
+       bus=system
+       interface="org.freedesktop.DBus.Properties"
+       path=/org/freedesktop/NetworkManager
+       member=GetAll
+       peer=(name=org.freedesktop.NetworkManager),

The Get*() and Set() methods are precisely what I was talking about
before because these methods specify what to get and set based on member
arguments and thus cannot be further restriceted by apparmor. In other
words, this method allows an app to get any property from network
manager, which is far too lenient. Network manager is designed with a
different trust model that does not fit within the context of trusted
helpers. Namely, network manager relies on policy kit and assumes the
user that is able to talk to it (via policy kit ACLs) is trusted. On
touch, the app is not trusted but the user is allowed to talk to network
manager. Therefore, network manager's current design does not afford
itself to opening up its api in the manner requested by this patch.

Which gets us back to the several choices:
* use connectivity api instead. It was specifically designed for this. apps talk to connectivity api via its simple and safe api, and it talks to network manager.
* write a proxy that is able to inspect the member data and filter it accordingly.
* patch network manager to provide a safe api for the things apps need

connectivity api is implemented precisely to address the concerns in
this bug and it seems clear (to me) that we should be using it. The
proxy idea would work and other libraries wouldn't have to be patched
(much) to use it, but is yet another service, is error prone and the
hardest to implement in the short term. Patching network manager
requires effort, increases maintenance costs and ultimately the same
work for apps to use it as adjusting them to use connectivity api.

-- 
You received this bug notification because you are a member of Ubuntu
Touch seeded packages, which is subscribed to apparmor-easyprof-ubuntu
in Ubuntu.
https://bugs.launchpad.net/bugs/1404188

Title:
  QNetworkSession::isOpen() always returns false

Status in Canonical System Image:
  Confirmed
Status in apparmor-easyprof-ubuntu package in Ubuntu:
  Confirmed
Status in qtbase-opensource-src package in Ubuntu:
  Confirmed

Bug description:
  Using QNetworkSession::isOpen() in confined apps on a phone running
  vivid always returns false. This might be an apparmor thing, however,
  I couldn't find any REJECTED entries in log files.

  The test app in lp:~mzanetti/+junk/nmsessiontest can reproduce the
  issue. Open this project in ubuntu-sdk's qtcreator and run it on a
  vivid device. Press the button and watch the debug prints.

  On a vivid-desktop or a RTM based phone it will print "all is well".
  On a vivid phone however, it'll print "network session not open..."

  This used to work fine at least back in utopic images.

To manage notifications about this bug go to:
https://bugs.launchpad.net/canonical-devices-system-image/+bug/1404188/+subscriptions