Conversation
|
Started test build 180823 |
|
Build 180823 successful |
|
During testing, I encountered the following error: I'm not sure how to resolve this. Any guidance or suggestions would be appreciated. |
|
There's probably a LSM (Linux Security Module) preventing passt from detaching its user namespace, which is done for (security) sandboxing reasons, see https://passt.top/passt/tree/isolation.c. Is this on a distribution using AppArmor? Or SELinux? For Debian, openSUSE, and Ubuntu, see https://passt.top/passt/tree/contrib/apparmor/ and, maybe also helpful: https://salsa.debian.org/sbrivio/passt/-/blob/master/debian/rules?ref_type=heads. SELinux? Then https://passt.top/passt/tree/contrib/selinux. |
|
@sbrivio-rh Thank you for the information! I tried patching After further research, I found that the issue arises because Flatpak restricts applications from creating user namespaces: flatpak/flatpak#5921. Based on the discussion, it seems that the best solution for now would be to adapt |
|
Yeeah its the sandboxing that breaks it, we might just need to disable the sandboxing for this, similar to what they do with chromium-based flatpaks |
Actually, zypak still provides some level of sandboxing, so it's not the same as having no sandboxing at all. |
Please, let's not go that way. If it breaks, the pieces are all yours: I'm not going to support this. By the way, is this obvious security weakness sufficiently reported to Flathub/Flatpak maintainers?
I didn't understand exactly how this (or zypak) would work:
|
I agree.
Disabling user namespaces may decrease the attack surface to the kernel, so it may not necessarily be a "security weakness". Although issue flatpak/flatpak#5921 was opened a few months ago, the discussion has only become heated this week. It might be helpful if you could ask some of your colleagues who are working on Flatpak to take a look at this.
Yes, we should use the sandbox created by |
Resolves #16 .