Skip to content

Conversation

@a-nogikh
Copy link
Collaborator

At the moment of its introduction, the attribute was overridable by enable_syscalls. However, since some of the malicious calls are actually variants of generic syscalls (e.g. ioctl), it has de-facto made it very easy to enable them also on non-snapshot instances.

Do not let enable_syscalls override this. Instead, only activate snapshot calls on the snapshot instances.

At the moment of its introduction, the attribute was overridable by
enable_syscalls. However, since some of the malicious calls are actually
variants of generic syscalls (e.g. ioctl), it has de-facto made it very
easy to enable them also on non-snapshot instances.

Do not let enable_syscalls override this. Instead, only activate
snapshot calls on the snapshot instances.
@a-nogikh a-nogikh requested a review from dvyukov January 21, 2026 18:08
@dvyukov
Copy link
Collaborator

dvyukov commented Jan 21, 2026

We need to at least produce an error, if a syscall is enabled by a full name, but we won't enable it.
If we I explicitly write full name, got no errors, but it's not enabled, looks like very bad failure mode.

But perhaps enabling by full name will be comparable amount of code to the check. If yes, then it looks even better.

@a-nogikh a-nogikh closed this Jan 21, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants