From 44ee678655a096a418e1fdabd0c972749017b5ec Mon Sep 17 00:00:00 2001 From: Todd Kjos Date: Fri, 27 Mar 2020 10:04:16 -0700 Subject: [PATCH] Revert "sched/core: Prevent race condition between cpuset and __sched_setscheduler()" This reverts commit 710da3c8ea7dfbd327920afd3831d8c82c42789d. When changing a thread's scheduling priority, binder calls sched_setscheduler_nocheck() while holding the node lock and proc inner lock. This was safe until v5.3 when this change was introduced where cpuset_read_lock() is called in this path which can sleep. This patch was introduced to fix a possible accounting error in sched deadline (potential oversell of CPU bandwidth) due to a race condition between cpusets and __sched_setscheduler(). This is not an issue for Android. This should be fixed in the binder driver, but that may take some time. Bug: 151861772 Change-Id: Ica1ef71b3cdcdc509b341ea1b57c41f8ee73794a Signed-off-by: Todd Kjos --- include/linux/cpuset.h | 5 ----- kernel/cgroup/cpuset.c | 11 ----------- kernel/sched/core.c | 20 +++----------------- 3 files changed, 3 insertions(+), 33 deletions(-) diff --git a/include/linux/cpuset.h b/include/linux/cpuset.h index 04c20de66afc..7f1478c26a33 100644 --- a/include/linux/cpuset.h +++ b/include/linux/cpuset.h @@ -55,8 +55,6 @@ extern void cpuset_init_smp(void); extern void cpuset_force_rebuild(void); extern void cpuset_update_active_cpus(void); extern void cpuset_wait_for_hotplug(void); -extern void cpuset_read_lock(void); -extern void cpuset_read_unlock(void); extern void cpuset_cpus_allowed(struct task_struct *p, struct cpumask *mask); extern void cpuset_cpus_allowed_fallback(struct task_struct *p); extern nodemask_t cpuset_mems_allowed(struct task_struct *p); @@ -178,9 +176,6 @@ static inline void cpuset_update_active_cpus(void) static inline void cpuset_wait_for_hotplug(void) { } -static inline void cpuset_read_lock(void) { } -static inline void cpuset_read_unlock(void) { } - static inline void cpuset_cpus_allowed(struct task_struct *p, struct cpumask *mask) { diff --git a/kernel/cgroup/cpuset.c b/kernel/cgroup/cpuset.c index 58f5073acff7..5c2c872c04a2 100644 --- a/kernel/cgroup/cpuset.c +++ b/kernel/cgroup/cpuset.c @@ -334,17 +334,6 @@ static struct cpuset top_cpuset = { */ DEFINE_STATIC_PERCPU_RWSEM(cpuset_rwsem); - -void cpuset_read_lock(void) -{ - percpu_down_read(&cpuset_rwsem); -} - -void cpuset_read_unlock(void) -{ - percpu_up_read(&cpuset_rwsem); -} - static DEFINE_SPINLOCK(callback_lock); static struct workqueue_struct *cpuset_migrate_mm_wq; diff --git a/kernel/sched/core.c b/kernel/sched/core.c index 9065fab507bd..1ae88d287e54 100644 --- a/kernel/sched/core.c +++ b/kernel/sched/core.c @@ -4866,9 +4866,6 @@ recheck: return retval; } - if (pi) - cpuset_read_lock(); - /* * Make sure no PI-waiters arrive (or leave) while we are * changing the priority of the task: @@ -4943,8 +4940,6 @@ change: if (unlikely(oldpolicy != -1 && oldpolicy != p->policy)) { policy = oldpolicy = -1; task_rq_unlock(rq, p, &rf); - if (pi) - cpuset_read_unlock(); goto recheck; } @@ -5005,10 +5000,8 @@ change: preempt_disable(); task_rq_unlock(rq, p, &rf); - if (pi) { - cpuset_read_unlock(); + if (pi) rt_mutex_adjust_pi(p); - } /* Run balance callbacks after we've adjusted the PI chain: */ balance_callback(rq); @@ -5018,8 +5011,6 @@ change: unlock: task_rq_unlock(rq, p, &rf); - if (pi) - cpuset_read_unlock(); return retval; } @@ -5104,14 +5095,9 @@ do_sched_setscheduler(pid_t pid, int policy, struct sched_param __user *param) rcu_read_lock(); retval = -ESRCH; p = find_process_by_pid(pid); - if (likely(p)) - get_task_struct(p); - rcu_read_unlock(); - - if (likely(p)) { + if (p != NULL) retval = sched_setscheduler(p, policy, &lparam); - put_task_struct(p); - } + rcu_read_unlock(); return retval; }