[PATCH 0/5 v0.6] sched/umcg: RFC UMCG patchset

From: Peter Oskolkov
Date: Fri Sep 17 2021 - 14:03:32 EST


Many thanks to Jann Horn for a detailed review of v0.5:
https://lore.kernel.org/lkml/20210908184905.163787-1-posk@xxxxxxxxxx/

The main change in this patchet vs v0.5 is pinning of
pages containing struct umcg_task of RUNNING workers
and their servers. This allows uaccess operations from
sched_submit_work (preempt-disabled). The pages are
pinned only for RUNNING workers, i.e. at most NR_CPUS
pages are pinned by UMCG at any one time, across the
whole machine.

A smaller change is the move of worker wakeup processing
from sched_update_worker() to exit_to_user_mode_loop(),
to ensure that no kernel locks are held if the worker
is descheduled there.

Commit messages in patches 2 and 3 list more changes.

Peter Oskolkov (5):
sched/umcg: add WF_CURRENT_CPU and externise ttwu
sched/umcg: RFC: add userspace atomic helpers
sched/umcg: RFC: implement UMCG syscalls
sched/umcg: add Documentation/userspace-api/umcg.rst
sched/umcg: add Documentation/userspace-api/umcg.txt

Documentation/userspace-api/umcg.rst | 550 ++++++++++++++++++
Documentation/userspace-api/umcg.txt | 537 ++++++++++++++++++
arch/x86/entry/syscalls/syscall_64.tbl | 2 +
fs/exec.c | 1 +
include/linux/sched.h | 56 ++
include/linux/syscalls.h | 4 +
include/uapi/asm-generic/unistd.h | 8 +-
include/uapi/linux/umcg.h | 117 ++++
init/Kconfig | 10 +
kernel/entry/common.c | 1 +
kernel/exit.c | 2 +
kernel/sched/Makefile | 1 +
kernel/sched/core.c | 18 +-
kernel/sched/fair.c | 4 +
kernel/sched/sched.h | 15 +-
kernel/sched/umcg.c | 745 +++++++++++++++++++++++++
kernel/sched/umcg_uaccess.h | 344 ++++++++++++
kernel/sys_ni.c | 4 +
18 files changed, 2408 insertions(+), 11 deletions(-)
create mode 100644 Documentation/userspace-api/umcg.rst
create mode 100644 Documentation/userspace-api/umcg.txt
create mode 100644 include/uapi/linux/umcg.h
create mode 100644 kernel/sched/umcg.c
create mode 100644 kernel/sched/umcg_uaccess.h

--
2.25.1