8299cef200
A guest can enter an IPC namespace through setns() with either an explicit CLONE_NEWIPC type or a zero type inferred from the namespace fd. Record both successful paths so the AOT exit worker remains synchronous and cannot keep the namespace alive after guest exit. Inspect NS_GET_NSTYPE only for the zero-type namespace-fd form and preserve the real setns() result when that ioctl is not applicable. This keeps pidfd namespace masks working while handling invalid descriptors explicitly. Add integration coverage for invalid descriptors, pidfd namespace joins, and both namespace-fd forms with AOT and mqueue lifecycle validation. Signed-off-by: yuerengan <y347812075@163.com> |
||
|---|---|---|
| .. | ||
| cef-userns-exec.S | ||
| cef-userns-nested.S | ||
| clone-namespaces.S | ||
| exec-with-pidfd.py | ||
| meson.build | ||
| proc-readdir.S | ||
| rcu-thread-visibility.S | ||
| setns-ipc-mqueue.S | ||
| test-cef-userns-exec.sh | ||
| test-cef-userns-nested.sh | ||
| test-clone-namespaces.sh | ||
| test-proc-readdir.sh | ||
| test-rcu-thread-visibility.sh | ||
| test-setns-ipc-mqueue.sh | ||