membarrier commits for 4.14
Hello, Linus,
This urgent pull request contains five commits that fix a problem that arose when designing requested extensions to sys_membarrier() to allow JITs to efficiently flush old code from instruction caches. Several potential algorithms are much less painful if the user register intent to use this functionality early on, for example, before the process spawns the second thread. Registering at this time removes the need to interrupt each and every thread in that process at the first expedited sys_membarrier() system call.
However, if we let the current function out, then user programs might be written without registration, which would make it more difficult to add required registration after the fact. Hence the late-in-merge-window pull request.
The last two commits remove unused code from architectures not supporting sys_membarrier() and fix a problem that arises when tasks are created with CLONE_VM but !CLONE_THREAD on powerpc.
These have been subject to 0day test robot and -next testing and are available in the git repository at:
git://git.kernel.org/pub/scm/linux/kernel/git/paulmck/linux-rcu.git membarrier.2017.10.09a
for you to fetch changes up to f7399e01ccd0850786bdedbee2fd351651ce99f8:
Fix: membarrier: Handle CLONE_VM + !CLONE_THREAD correctly on powerpc (2017-10-09 14:24:49 -0700)
----------------------------------------------------------------
Mathieu Desnoyers (5):
membarrier: Provide register expedited private command
membarrier: selftest: Test private expedited cmd
membarrier: Document scheduler barrier requirements
membarrier: Remove unused code for architectures without membarrier hooks
Fix: membarrier: Handle CLONE_VM + !CLONE_THREAD correctly on powerpc
MAINTAINERS | 2 + arch/powerpc/Kconfig | 1 + arch/powerpc/include/asm/membarrier.h | 26 +++++ arch/powerpc/kernel/Makefile | 2 + arch/powerpc/kernel/membarrier.c | 36 +++++++ arch/powerpc/mm/mmu_context.c | 7 ++ arch/x86/mm/tlb.c | 5 + fs/exec.c | 1 + include/linux/mm_types.h | 3 + include/linux/sched/mm.h | 37 +++++++ include/uapi/linux/membarrier.h | 23 +++-- init/Kconfig | 3 + kernel/sched/core.c | 47 +++++---- kernel/sched/membarrier.c | 35 ++++++- .../testing/selftests/membarrier/membarrier_test.c | 109 ++++++++++++++++++--- 15 files changed, 290 insertions(+), 47 deletions(-) create mode 100644 arch/powerpc/include/asm/membarrier.h create mode 100644 arch/powerpc/kernel/membarrier.c
Quoted 1 line · ----- On Oct 18, 2017, at 3:59 PM, Paul E. McKenney [email protected]…
Hello, Linus,
Hi Linus,
Time did fly since I submitted this series near the beginning of the 4.14 rc cycle. Now that we are closer to the end of the cycle, a more surgical approach would be more appropriate.
I have prepared the minimal patch that needs to be applied to modify the user-space visible ABI before the end of the 4.14 cycle.
I will send that patch to you directly, hoping you can apply it before the end of the 4.14 cycle.
Thanks,
Mathieu
Quoted 56 lines
This urgent pull request contains five commits that fix a problem that arose when designing requested extensions to
sys_membarrier()to allow JITs to efficiently flush old code from instruction caches. Several potential algorithms are much less painful if the user register intent to use this functionality early on, for example, before the process spawns the second thread. Registering at this time removes the need to interrupt each and every thread in that process at the first expeditedsys_membarrier()system call.However, if we let the current function out, then user programs might be written without registration, which would make it more difficult to add required registration after the fact. Hence the late-in-merge-window pull request.
The last two commits remove unused code from architectures not supporting
sys_membarrier()and fix a problem that arises when tasks are created withCLONE_VMbut !CLONE_THREAD on powerpc.These have been subject to 0day test robot and -next testing and are available in the git repository at:
git://git.kernel.org/pub/scm/linux/kernel/git/paulmck/linux-rcu.git
membarrier.2017.10.09afor you to fetch changes up to
f7399e01ccd0850786bdedbee2fd351651ce99f8:Fix: membarrier: Handle
CLONE_VM+ !CLONE_THREAD correctly on powerpc (2017-10-09 14:24:49 -0700)----------------------------------------------------------------
Mathieu Desnoyers (5):
membarrier: Provide register expedited private command
membarrier: selftest: Test private expedited cmd
membarrier: Document scheduler barrier requirements
membarrier: Remove unused code for architectures without membarrier hooks
Fix: membarrier: HandleCLONE_VM+ !CLONE_THREAD correctly on powerpcMAINTAINERS | 2 + arch/powerpc/Kconfig | 1 + arch/powerpc/include/asm/membarrier.h | 26 +++++ arch/powerpc/kernel/Makefile | 2 + arch/powerpc/kernel/membarrier.c | 36 +++++++ arch/powerpc/mm/mmu_context.c | 7 ++ arch/x86/mm/tlb.c | 5 + fs/exec.c | 1 + include/linux/mm_types.h | 3 + include/linux/sched/mm.h | 37 +++++++ include/uapi/linux/membarrier.h | 23 +++-- init/Kconfig | 3 + kernel/sched/core.c | 47 +++++---- kernel/sched/membarrier.c | 35 ++++++- .../testing/selftests/membarrier/membarrier_test.c | 109 ++++++++++++++++++--- 15 files changed, 290 insertions(+), 47 deletions(-) create mode 100644 arch/powerpc/include/asm/membarrier.h create mode 100644 arch/powerpc/kernel/membarrier.c
Signature · 3 lines
Mathieu Desnoyers EfficiOS Inc. http://www.efficios.com
Quoted 7 lines · Mathieu Desnoyers
I have prepared the minimal patch that needs to be applied to modify the user-space visible ABI before the end of the 4.14 cycle.
I will send that patch to you directly, hoping you can apply it before the end of the 4.14 cycle.
Thanks, that sounds much better than the big pull request at this stage.
Linus