Container-init must behave like global-init to processes within the
container and hence it must be immune to unhandled fatal signals from
within the container (i.e SIG_DFL signals that terminate the process).
But the same container-init must behave like a normal process to
processes in ancestor namespaces and so if it receives the same fatal
signal from a process in ancestor namespace, the signal must be
processed.
Implementing these semantics requires that send_signal() determine pid
namespace of the sender but since signals can originate from workqueues/
interrupt-handlers, determining pid namespace of sender may not always
be possible or safe.
Changelog[v3]:
Changes based on discussions of previous version:
http://lkml.org/lkml/2008/11/25/458
Major changes:
- Define SIGNAL_UNKILLABLE_FROM_NS and use in container-inits to
skip fatal signals from same namespace but process SIGKILL/SIGSTOP
from ancestor namespace.
- Use SI_FROMUSER() and si_code != SI_ASYNCIO to determine if
it is safe to dereference pid-namespace of caller. Highly
experimental :-)
- Masquerading si_pid when crossing namespace boundary: relevant
patches merged in -mm and dropped from this set.
Minor changes:
- Remove 'handler' parameter to tracehook functions
- Update sig_ignored() to drop SIG_DFL signals to global init early
(tried to address Roland's and Oleg's comments)
- Use 'same_ns' flag to drop SIGKILL/SIGSTOP to cinit from same
namespace
This patchset implements the design/simplified semantics suggested by
Oleg Nesterov. The simplified semantics for container-init are:
- container-init must never be terminated by a signal from a
descendant process.
- container-init must never be immune to SIGKILL from an ancestor
namespace (so a process in parent namespace must always be able
to terminate a descendant container).
- container-init may be immune to unhandled fatal signals (like
SIGUSR1) even if they are from ancestor namespace (SIGKILL is
the only reliable signal from ancestor namespace).
Patches in this set:
[PATCH 1/6] Remove 'handler' parameter to tracehook functions
[PATCH 2/6] Protect init from unwanted signals more
[PATCH 3/6] Define/set SIGNAL_UNKILLABLE_FROM_NS
[PATCH 4/6] Define siginfo_from_ancestor_ns()
[PATCH 5/6] Protect cinit from unblocked SIG_DFL signals
[PATCH 6/6] Protect cinit from blocked fatal signals
TODO:
- Use sig_task_unkillable() in fs/proc/array.c:task_sig() to
correctly report ignored signals for container/global init.
- Make SI_ASYNCIO a kernel signal ?
- Compile/touch tested. Need so real testing ;-)
Limitations/side-effects of current design
- Container-init is immune to suicide - kill(getpid(), SIGKILL) is
ignored. Use exit() :-)
--
