| Message ID | 20221207164338.1535591-1-mclapinski@google.com (mailing list archive) |
|---|---|
| Headers | show |
| Series | sched/membarrier, selftests: Introduce MEMBARRIER_CMD_GET_REGISTRATIONS | expand |
On 2022-12-07 11:43, Michal Clapinski wrote: > This change provides a method to query previously issued registrations. > It's needed for CRIU (checkpoint/restore in userspace). Before this > change we had to issue private membarrier commands during checkpoint - > if they succeeded, they must have been registered. Unfortunately global > membarrier succeeds even on unregistered processes, so there was no way to > tell if MEMBARRIER_CMD_REGISTER_GLOBAL_EXPEDITED had been issued or not. > > CRIU is run after the process has been frozen with ptrace, so we don't > have to worry too much about the result of running this command in parallel > with registration commands. Peter, Paul, I'm OK with the proposed changes. Should we route this through sched/core from the tip tree ? For both patches: Acked-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com> Thanks, Mathieu > > Michal Clapinski (2): > sched/membarrier: Introduce MEMBARRIER_CMD_GET_REGISTRATIONS > selftests/membarrier: Test MEMBARRIER_CMD_GET_REGISTRATIONS > > include/uapi/linux/membarrier.h | 4 ++ > kernel/sched/membarrier.c | 39 ++++++++++++++++++- > .../membarrier/membarrier_test_impl.h | 33 ++++++++++++++++ > .../membarrier/membarrier_test_multi_thread.c | 2 +- > .../membarrier_test_single_thread.c | 6 ++- > 5 files changed, 81 insertions(+), 3 deletions(-) >
On Thu, Dec 22, 2022 at 10:28:28AM -0500, Mathieu Desnoyers wrote: > On 2022-12-07 11:43, Michal Clapinski wrote: > > This change provides a method to query previously issued registrations. > > It's needed for CRIU (checkpoint/restore in userspace). Before this > > change we had to issue private membarrier commands during checkpoint - > > if they succeeded, they must have been registered. Unfortunately global > > membarrier succeeds even on unregistered processes, so there was no way to > > tell if MEMBARRIER_CMD_REGISTER_GLOBAL_EXPEDITED had been issued or not. > > > > CRIU is run after the process has been frozen with ptrace, so we don't > > have to worry too much about the result of running this command in parallel > > with registration commands. > > Peter, Paul, I'm OK with the proposed changes. Should we route this through > sched/core from the tip tree ? > > For both patches: > > Acked-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com> Also for both patches: Acked-by: Paul E. McKenney <paulmck@kernel.org> > Thanks, > > Mathieu > > > > > Michal Clapinski (2): > > sched/membarrier: Introduce MEMBARRIER_CMD_GET_REGISTRATIONS > > selftests/membarrier: Test MEMBARRIER_CMD_GET_REGISTRATIONS > > > > include/uapi/linux/membarrier.h | 4 ++ > > kernel/sched/membarrier.c | 39 ++++++++++++++++++- > > .../membarrier/membarrier_test_impl.h | 33 ++++++++++++++++ > > .../membarrier/membarrier_test_multi_thread.c | 2 +- > > .../membarrier_test_single_thread.c | 6 ++- > > 5 files changed, 81 insertions(+), 3 deletions(-) > > > > -- > Mathieu Desnoyers > EfficiOS Inc. > https://www.efficios.com >