diff options
author | Tejun Heo <tj@kernel.org> | 2024-09-27 10:02:40 -1000 |
---|---|---|
committer | Tejun Heo <tj@kernel.org> | 2024-09-27 10:02:40 -1000 |
commit | efe231d9debf6db812bebb262407c95b21cdb8a2 (patch) | |
tree | 0746479ba3f2e723c3e70846397dd2ea1829b915 /kernel/sched | |
parent | 160216568cddc9d6e7f36133ba41d25459d90de4 (diff) |
sched_ext: Decouple locks in scx_ops_enable()
The enable path uses three big locks - scx_fork_rwsem, scx_cgroup_rwsem and
cpus_read_lock. Currently, the locks are grabbed together which is prone to
locking order problems.
For example, currently, there is a possible deadlock involving
scx_fork_rwsem and cpus_read_lock. cpus_read_lock has to nest inside
scx_fork_rwsem due to locking order existing in other subsystems. However,
there exists a dependency in the other direction during hotplug if hotplug
needs to fork a new task, which happens in some cases. This leads to the
following deadlock:
scx_ops_enable() hotplug
percpu_down_write(&cpu_hotplug_lock)
percpu_down_write(&scx_fork_rwsem)
block on cpu_hotplug_lock
kthread_create() waits for kthreadd
kthreadd blocks on scx_fork_rwsem
Note that this doesn't trigger lockdep because the hotplug side dependency
bounces through kthreadd.
With the preceding scx_cgroup_enabled change, this can be solved by
decoupling cpus_read_lock, which is needed for static_key manipulations,
from the other two locks.
- Move the first block of static_key manipulations outside of scx_fork_rwsem
and scx_cgroup_rwsem. This is now safe with the preceding
scx_cgroup_enabled change.
- Drop scx_cgroup_rwsem and scx_fork_rwsem between the two task iteration
blocks so that __scx_ops_enabled static_key enabling is outside the two
rwsems.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reported-and-tested-by: Aboorva Devarajan <aboorvad@linux.ibm.com>
Link: http://lkml.kernel.org/r/8cd0ec0c4c7c1bc0119e61fbef0bee9d5e24022d.camel@linux.ibm.com
Diffstat (limited to 'kernel/sched')
-rw-r--r-- | kernel/sched/ext.c | 67 |
1 files changed, 27 insertions, 40 deletions
diff --git a/kernel/sched/ext.c b/kernel/sched/ext.c index 7d0ce7e6ea1f..0c398ecdb938 100644 --- a/kernel/sched/ext.c +++ b/kernel/sched/ext.c @@ -5049,7 +5049,8 @@ static int scx_ops_enable(struct sched_ext_ops *ops, struct bpf_link *link) ret = SCX_CALL_OP_RET(SCX_KF_UNLOCKED, init); if (ret) { ret = ops_sanitize_err("init", ret); - goto err_disable_unlock_cpus; + cpus_read_unlock(); + goto err_disable; } } @@ -5092,54 +5093,30 @@ static int scx_ops_enable(struct sched_ext_ops *ops, struct bpf_link *link) */ scx_ops_bypass(true); - /* - * Lock out forks, cgroup on/offlining and moves before opening the - * floodgate so that they don't wander into the operations prematurely. - * - * We don't need to keep the CPUs stable but static_branch_*() requires - * cpus_read_lock() and scx_cgroup_rwsem must nest inside - * cpu_hotplug_lock because of the following dependency chain: - * - * cpu_hotplug_lock --> cgroup_threadgroup_rwsem --> scx_cgroup_rwsem - * - * So, we need to do cpus_read_lock() before scx_cgroup_lock() and use - * static_branch_*_cpuslocked(). - * - * Note that cpu_hotplug_lock must nest inside scx_fork_rwsem due to the - * following dependency chain: - * - * scx_fork_rwsem --> pernet_ops_rwsem --> cpu_hotplug_lock - */ - percpu_down_write(&scx_fork_rwsem); - cpus_read_lock(); - scx_cgroup_lock(); - for (i = SCX_OPI_NORMAL_BEGIN; i < SCX_OPI_NORMAL_END; i++) if (((void (**)(void))ops)[i]) - static_branch_enable_cpuslocked(&scx_has_op[i]); + static_branch_enable(&scx_has_op[i]); if (ops->flags & SCX_OPS_ENQ_LAST) - static_branch_enable_cpuslocked(&scx_ops_enq_last); + static_branch_enable(&scx_ops_enq_last); if (ops->flags & SCX_OPS_ENQ_EXITING) - static_branch_enable_cpuslocked(&scx_ops_enq_exiting); + static_branch_enable(&scx_ops_enq_exiting); if (scx_ops.cpu_acquire || scx_ops.cpu_release) - static_branch_enable_cpuslocked(&scx_ops_cpu_preempt); + static_branch_enable(&scx_ops_cpu_preempt); if (!ops->update_idle || (ops->flags & SCX_OPS_KEEP_BUILTIN_IDLE)) { reset_idle_masks(); - static_branch_enable_cpuslocked(&scx_builtin_idle_enabled); + static_branch_enable(&scx_builtin_idle_enabled); } else { - static_branch_disable_cpuslocked(&scx_builtin_idle_enabled); + static_branch_disable(&scx_builtin_idle_enabled); } /* - * All cgroups should be initialized before letting in tasks. cgroup - * on/offlining and task migrations are already locked out. + * Lock out forks, cgroup on/offlining and moves before opening the + * floodgate so that they don't wander into the operations prematurely. */ - ret = scx_cgroup_init(); - if (ret) - goto err_disable_unlock_all; + percpu_down_write(&scx_fork_rwsem); WARN_ON_ONCE(scx_ops_init_task_enabled); scx_ops_init_task_enabled = true; @@ -5150,7 +5127,18 @@ static int scx_ops_enable(struct sched_ext_ops *ops, struct bpf_link *link) * leaving as sched_ext_free() can handle both prepped and enabled * tasks. Prep all tasks first and then enable them with preemption * disabled. + * + * All cgroups should be initialized before scx_ops_init_task() so that + * the BPF scheduler can reliably track each task's cgroup membership + * from scx_ops_init_task(). Lock out cgroup on/offlining and task + * migrations while tasks are being initialized so that + * scx_cgroup_can_attach() never sees uninitialized tasks. */ + scx_cgroup_lock(); + ret = scx_cgroup_init(); + if (ret) + goto err_disable_unlock_all; + spin_lock_irq(&scx_tasks_lock); scx_task_iter_init(&sti); while ((p = scx_task_iter_next_locked(&sti))) { @@ -5183,19 +5171,22 @@ static int scx_ops_enable(struct sched_ext_ops *ops, struct bpf_link *link) } scx_task_iter_exit(&sti); spin_unlock_irq(&scx_tasks_lock); + scx_cgroup_unlock(); + percpu_up_write(&scx_fork_rwsem); /* * All tasks are READY. It's safe to turn on scx_enabled() and switch * all eligible tasks. */ WRITE_ONCE(scx_switching_all, !(ops->flags & SCX_OPS_SWITCH_PARTIAL)); - static_branch_enable_cpuslocked(&__scx_ops_enabled); + static_branch_enable(&__scx_ops_enabled); /* * We're fully committed and can't fail. The task READY -> ENABLED * transitions here are synchronized against sched_ext_free() through * scx_tasks_lock. */ + percpu_down_write(&scx_fork_rwsem); spin_lock_irq(&scx_tasks_lock); scx_task_iter_init(&sti); while ((p = scx_task_iter_next_locked(&sti))) { @@ -5213,10 +5204,8 @@ static int scx_ops_enable(struct sched_ext_ops *ops, struct bpf_link *link) } scx_task_iter_exit(&sti); spin_unlock_irq(&scx_tasks_lock); - - scx_cgroup_unlock(); - cpus_read_unlock(); percpu_up_write(&scx_fork_rwsem); + scx_ops_bypass(false); /* @@ -5259,8 +5248,6 @@ err_disable_unlock_all: scx_cgroup_unlock(); percpu_up_write(&scx_fork_rwsem); scx_ops_bypass(false); -err_disable_unlock_cpus: - cpus_read_unlock(); err_disable: mutex_unlock(&scx_ops_enable_mutex); /* must be fully disabled before returning */ |