From: Tejun Heo <tj@kernel.org>
Date: Sun, 10 May 2026 22:51:09 +0000 (-1000)
Subject: Merge branch 'for-7.1-fixes' into for-7.2
X-Git-Url: http://git.ipfire.org/cgi-bin/gitweb.cgi?a=commitdiff_plain;h=b5646c627652241f248f0a4ff31b1d32316b0068;p=thirdparty%2Fkernel%2Flinux.git

Merge branch 'for-7.1-fixes' into for-7.2

Conflict between:

 [1] 41e3312861ea ("sched_ext: add p->scx.tid and SCX_OPS_TID_TO_TASK lookup")
 [2] c941d7391f25 ("sched_ext: Close root-enable vs sched_ext_dead() race with SCX_TASK_INIT_BEGIN")

in scx_root_enable_workfn()'s post-init block. [1] added a tid hash
insertion under a scoped_guard() for scx_tasks_lock; [2] wraps the same
region in task_rq_lock() for a DEAD recheck. A naive merge would invert the
iter's outer/inner order.

 [3] f25ad1e3cbaa ("sched_ext: Add scx_task_iter_relock() and use it in scx_root_enable_workfn()")

was added to for-7.2 for a clean resolution: scx_task_iter_relock(iter, p)
takes both scx_tasks_lock and @p's rq lock in iter order.

Resolved by routing both sides through [3]'s dual-lock helper: the post-init
region runs under a single scx_task_iter_relock() acquisition, with [2]'s
state machine and [1]'s hash insert in sequence inside it.

Signed-off-by: Tejun Heo <tj@kernel.org>
---

b5646c627652241f248f0a4ff31b1d32316b0068
diff --cc kernel/sched/ext.c
index fbdc1819d4cb0,4efe0099f79af..b8dd3358959d7
--- a/kernel/sched/ext.c
+++ b/kernel/sched/ext.c
@@@ -7079,13 -6936,26 +7188,25 @@@ static void scx_root_enable_workfn(stru
  		if (!tryget_task_struct(p))
  			continue;
  
+ 		/*
+ 		 * Set %INIT_BEGIN under the iter's rq lock so that a concurrent
+ 		 * sched_ext_dead() does not call ops.exit_task() on @p while
+ 		 * ops.init_task() is running. If sched_ext_dead() runs before
+ 		 * this store, it has already removed @p from scx_tasks and the
+ 		 * iter won't visit @p; if it runs after, it observes
+ 		 * %INIT_BEGIN and transitions to %DEAD without calling ops,
+ 		 * leaving the post-init recheck below to unwind.
+ 		 */
+ 		scx_set_task_state(p, SCX_TASK_INIT_BEGIN);
  		scx_task_iter_unlock(&sti);
  
- 		ret = scx_init_task(sch, p, false);
+ 		ret = __scx_init_task(sch, p, false);
  
 -		rq = task_rq_lock(p, &rf);
 +		scx_task_iter_relock(&sti, p);
  
- 		if (ret) {
+ 		if (unlikely(ret)) {
+ 			if (scx_get_task_state(p) != SCX_TASK_DEAD)
+ 				scx_set_task_state(p, SCX_TASK_NONE);
 -			task_rq_unlock(rq, p, &rf);
  			put_task_struct(p);
  			scx_task_iter_stop(&sti);
  			scx_error(sch, "ops.init_task() failed (%d) for %s[%d]",
@@@ -7093,17 -6963,20 +7214,27 @@@
  			goto err_disable_unlock_all;
  		}
  
- 		scx_set_task_sched(p, sch);
- 		scx_set_task_state(p, SCX_TASK_READY);
+ 		if (scx_get_task_state(p) == SCX_TASK_DEAD) {
+ 			/*
+ 			 * sched_ext_dead() observed %INIT_BEGIN and set %DEAD.
+ 			 * ops.exit_task() is owed to the sched __scx_init_task()
+ 			 * ran against; call it now.
+ 			 */
+ 			scx_sub_init_cancel_task(sch, p);
+ 		} else {
+ 			scx_set_task_state(p, SCX_TASK_INIT);
+ 			scx_set_task_sched(p, sch);
+ 			scx_set_task_state(p, SCX_TASK_READY);
+ 		}
  
 -		task_rq_unlock(rq, p, &rf);
 +		/*
 +		 * Insert into the tid hash. scx_tasks_lock is held by the iter;
 +		 * list_empty() guards against sched_ext_dead() having taken @p
 +		 * off the list while init ran unlocked.
 +		 */
 +		if (scx_tid_to_task_enabled() && !list_empty(&p->scx.tasks_node))
 +			scx_tid_hash_insert(p);
 +
  		put_task_struct(p);
  	}
  	scx_task_iter_stop(&sti);