Merge tag 'core-rseq-2025-11-30' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip

Pull rseq updates from Thomas Gleixner:
 "A large overhaul of the restartable sequences and CID management:

  The recent enablement of RSEQ in glibc resulted in regressions which
  are caused by the related overhead. It turned out that the decision to
  invoke the exit to user work was not really a decision. More or less
  each context switch caused that. There is a long list of small issues
  which sums up nicely and results in a 3-4% regression in I/O
  benchmarks.

  The other detail which caused issues due to extra work in context
  switch and task migration is the CID (memory context ID) management.
  It also requires to use a task work to consolidate the CID space,
  which is executed in the context of an arbitrary task and results in
  sporadic uncontrolled exit latencies.

  The rewrite addresses this by:

   - Removing deprecated and long unsupported functionality

   - Moving the related data into dedicated data structures which are
     optimized for fast path processing.

   - Caching values so actual decisions can be made

   - Replacing the current implementation with a optimized inlined
     variant.

   - Separating fast and slow path for architectures which use the
     generic entry code, so that only fault and error handling goes into
     the TIF_NOTIFY_RESUME handler.

   - Rewriting the CID management so that it becomes mostly invisible in
     the context switch path. That moves the work of switching modes
     into the fork/exit path, which is a reasonable tradeoff. That work
     is only required when a process creates more threads than the
     cpuset it is allowed to run on or when enough threads exit after
     that. An artificial thread pool benchmarks which triggers this did
     not degrade, it actually improved significantly.

     The main effect in migration heavy scenarios is that runqueue lock
     held time and therefore contention goes down significantly"

* tag 'core-rseq-2025-11-30' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: (54 commits)
  sched/mmcid: Switch over to the new mechanism
  sched/mmcid: Implement deferred mode change
  irqwork: Move data struct to a types header
  sched/mmcid: Provide CID ownership mode fixup functions
  sched/mmcid: Provide new scheduler CID mechanism
  sched/mmcid: Introduce per task/CPU ownership infrastructure
  sched/mmcid: Serialize sched_mm_cid_fork()/exit() with a mutex
  sched/mmcid: Provide precomputed maximal value
  sched/mmcid: Move initialization out of line
  signal: Move MMCID exit out of sighand lock
  sched/mmcid: Convert mm CID mask to a bitmap
  cpumask: Cache num_possible_cpus()
  sched/mmcid: Use cpumask_weighted_or()
  cpumask: Introduce cpumask_weighted_or()
  sched/mmcid: Prevent pointless work in mm_update_cpus_allowed()
  sched/mmcid: Move scheduler code out of global header
  sched: Fixup whitespace damage
  sched/mmcid: Cacheline align MM CID storage
  sched/mmcid: Use proper data structures
  sched/mmcid: Revert the complex CID management
  ...

1 files changed, 161 insertions, 231 deletions

diff --git a/kernel/sched/sched.h b/kernel/sched/sched.h
index b419a4d98461..8590113e4a60 100644
--- a/kernel/sched/sched.h
+++ b/kernel/sched/sched.h
@@ -2223,6 +2223,7 @@ static inline void __set_task_cpu(struct task_struct *p, unsigned int cpu)
 	smp_wmb();
 	WRITE_ONCE(task_thread_info(p)->cpu, cpu);
 	p->wake_cpu = cpu;
+	rseq_sched_set_ids_changed(p);
 #endif /* CONFIG_SMP */
 }
 
@@ -3679,283 +3680,212 @@ extern const char *preempt_modes[];
 
 #ifdef CONFIG_SCHED_MM_CID
 
-#define SCHED_MM_CID_PERIOD_NS	(100ULL * 1000000)	/* 100ms */
-#define MM_CID_SCAN_DELAY	100			/* 100ms */
+static __always_inline bool cid_on_cpu(unsigned int cid)
+{
+	return cid & MM_CID_ONCPU;
+}
 
-extern raw_spinlock_t cid_lock;
-extern int use_cid_lock;
+static __always_inline bool cid_in_transit(unsigned int cid)
+{
+	return cid & MM_CID_TRANSIT;
+}
 
-extern void sched_mm_cid_migrate_from(struct task_struct *t);
-extern void sched_mm_cid_migrate_to(struct rq *dst_rq, struct task_struct *t);
-extern void task_tick_mm_cid(struct rq *rq, struct task_struct *curr);
-extern void init_sched_mm_cid(struct task_struct *t);
+static __always_inline unsigned int cpu_cid_to_cid(unsigned int cid)
+{
+	return cid & ~MM_CID_ONCPU;
+}
 
-static inline void __mm_cid_put(struct mm_struct *mm, int cid)
+static __always_inline unsigned int cid_to_cpu_cid(unsigned int cid)
 {
-	if (cid < 0)
-		return;
-	cpumask_clear_cpu(cid, mm_cidmask(mm));
+	return cid | MM_CID_ONCPU;
 }
 
-/*
- * The per-mm/cpu cid can have the MM_CID_LAZY_PUT flag set or transition to
- * the MM_CID_UNSET state without holding the rq lock, but the rq lock needs to
- * be held to transition to other states.
- *
- * State transitions synchronized with cmpxchg or try_cmpxchg need to be
- * consistent across CPUs, which prevents use of this_cpu_cmpxchg.
- */
-static inline void mm_cid_put_lazy(struct task_struct *t)
+static __always_inline unsigned int cid_to_transit_cid(unsigned int cid)
 {
-	struct mm_struct *mm = t->mm;
-	struct mm_cid __percpu *pcpu_cid = mm->pcpu_cid;
-	int cid;
+	return cid | MM_CID_TRANSIT;
+}
 
-	lockdep_assert_irqs_disabled();
-	cid = __this_cpu_read(pcpu_cid->cid);
-	if (!mm_cid_is_lazy_put(cid) ||
-	    !try_cmpxchg(&this_cpu_ptr(pcpu_cid)->cid, &cid, MM_CID_UNSET))
-		return;
-	__mm_cid_put(mm, mm_cid_clear_lazy_put(cid));
+static __always_inline unsigned int cid_from_transit_cid(unsigned int cid)
+{
+	return cid & ~MM_CID_TRANSIT;
 }
 
-static inline int mm_cid_pcpu_unset(struct mm_struct *mm)
+static __always_inline bool cid_on_task(unsigned int cid)
 {
-	struct mm_cid __percpu *pcpu_cid = mm->pcpu_cid;
-	int cid, res;
+	/* True if none of the MM_CID_ONCPU, MM_CID_TRANSIT, MM_CID_UNSET bits is set */
+	return cid < MM_CID_TRANSIT;
+}
 
-	lockdep_assert_irqs_disabled();
-	cid = __this_cpu_read(pcpu_cid->cid);
-	for (;;) {
-		if (mm_cid_is_unset(cid))
-			return MM_CID_UNSET;
-		/*
-		 * Attempt transition from valid or lazy-put to unset.
-		 */
-		res = cmpxchg(&this_cpu_ptr(pcpu_cid)->cid, cid, MM_CID_UNSET);
-		if (res == cid)
-			break;
-		cid = res;
-	}
-	return cid;
+static __always_inline void mm_drop_cid(struct mm_struct *mm, unsigned int cid)
+{
+	clear_bit(cid, mm_cidmask(mm));
 }
 
-static inline void mm_cid_put(struct mm_struct *mm)
+static __always_inline void mm_unset_cid_on_task(struct task_struct *t)
 {
-	int cid;
+	unsigned int cid = t->mm_cid.cid;
 
-	lockdep_assert_irqs_disabled();
-	cid = mm_cid_pcpu_unset(mm);
-	if (cid == MM_CID_UNSET)
-		return;
-	__mm_cid_put(mm, mm_cid_clear_lazy_put(cid));
+	t->mm_cid.cid = MM_CID_UNSET;
+	if (cid_on_task(cid))
+		mm_drop_cid(t->mm, cid);
 }
 
-static inline int __mm_cid_try_get(struct task_struct *t, struct mm_struct *mm)
+static __always_inline void mm_drop_cid_on_cpu(struct mm_struct *mm, struct mm_cid_pcpu *pcp)
 {
-	struct cpumask *cidmask = mm_cidmask(mm);
-	struct mm_cid __percpu *pcpu_cid = mm->pcpu_cid;
-	int cid, max_nr_cid, allowed_max_nr_cid;
+	/* Clear the ONCPU bit, but do not set UNSET in the per CPU storage */
+	pcp->cid = cpu_cid_to_cid(pcp->cid);
+	mm_drop_cid(mm, pcp->cid);
+}
 
-	/*
-	 * After shrinking the number of threads or reducing the number
-	 * of allowed cpus, reduce the value of max_nr_cid so expansion
-	 * of cid allocation will preserve cache locality if the number
-	 * of threads or allowed cpus increase again.
-	 */
-	max_nr_cid = atomic_read(&mm->max_nr_cid);
-	while ((allowed_max_nr_cid = min_t(int, READ_ONCE(mm->nr_cpus_allowed),
-					   atomic_read(&mm->mm_users))),
-	       max_nr_cid > allowed_max_nr_cid) {
-		/* atomic_try_cmpxchg loads previous mm->max_nr_cid into max_nr_cid. */
-		if (atomic_try_cmpxchg(&mm->max_nr_cid, &max_nr_cid, allowed_max_nr_cid)) {
-			max_nr_cid = allowed_max_nr_cid;
-			break;
-		}
-	}
-	/* Try to re-use recent cid. This improves cache locality. */
-	cid = __this_cpu_read(pcpu_cid->recent_cid);
-	if (!mm_cid_is_unset(cid) && cid < max_nr_cid &&
-	    !cpumask_test_and_set_cpu(cid, cidmask))
-		return cid;
-	/*
-	 * Expand cid allocation if the maximum number of concurrency
-	 * IDs allocated (max_nr_cid) is below the number cpus allowed
-	 * and number of threads. Expanding cid allocation as much as
-	 * possible improves cache locality.
-	 */
-	cid = max_nr_cid;
-	while (cid < READ_ONCE(mm->nr_cpus_allowed) && cid < atomic_read(&mm->mm_users)) {
-		/* atomic_try_cmpxchg loads previous mm->max_nr_cid into cid. */
-		if (!atomic_try_cmpxchg(&mm->max_nr_cid, &cid, cid + 1))
-			continue;
-		if (!cpumask_test_and_set_cpu(cid, cidmask))
-			return cid;
-	}
-	/*
-	 * Find the first available concurrency id.
-	 * Retry finding first zero bit if the mask is temporarily
-	 * filled. This only happens during concurrent remote-clear
-	 * which owns a cid without holding a rq lock.
-	 */
-	for (;;) {
-		cid = cpumask_first_zero(cidmask);
-		if (cid < READ_ONCE(mm->nr_cpus_allowed))
-			break;
-		cpu_relax();
-	}
-	if (cpumask_test_and_set_cpu(cid, cidmask))
-		return -1;
+static inline unsigned int __mm_get_cid(struct mm_struct *mm, unsigned int max_cids)
+{
+	unsigned int cid = find_first_zero_bit(mm_cidmask(mm), max_cids);
 
+	if (cid >= max_cids)
+		return MM_CID_UNSET;
+	if (test_and_set_bit(cid, mm_cidmask(mm)))
+		return MM_CID_UNSET;
 	return cid;
 }
 
-/*
- * Save a snapshot of the current runqueue time of this cpu
- * with the per-cpu cid value, allowing to estimate how recently it was used.
- */
-static inline void mm_cid_snapshot_time(struct rq *rq, struct mm_struct *mm)
+static inline unsigned int mm_get_cid(struct mm_struct *mm)
 {
-	struct mm_cid *pcpu_cid = per_cpu_ptr(mm->pcpu_cid, cpu_of(rq));
+	unsigned int cid = __mm_get_cid(mm, READ_ONCE(mm->mm_cid.max_cids));
 
-	lockdep_assert_rq_held(rq);
-	WRITE_ONCE(pcpu_cid->time, rq->clock);
+	while (cid == MM_CID_UNSET) {
+		cpu_relax();
+		cid = __mm_get_cid(mm, num_possible_cpus());
+	}
+	return cid;
 }
 
-static inline int __mm_cid_get(struct rq *rq, struct task_struct *t,
-			       struct mm_struct *mm)
+static inline unsigned int mm_cid_converge(struct mm_struct *mm, unsigned int orig_cid,
+					   unsigned int max_cids)
 {
-	int cid;
+	unsigned int new_cid, cid = cpu_cid_to_cid(orig_cid);
 
-	/*
-	 * All allocations (even those using the cid_lock) are lock-free. If
-	 * use_cid_lock is set, hold the cid_lock to perform cid allocation to
-	 * guarantee forward progress.
-	 */
-	if (!READ_ONCE(use_cid_lock)) {
-		cid = __mm_cid_try_get(t, mm);
-		if (cid >= 0)
-			goto end;
-		raw_spin_lock(&cid_lock);
-	} else {
-		raw_spin_lock(&cid_lock);
-		cid = __mm_cid_try_get(t, mm);
-		if (cid >= 0)
-			goto unlock;
+	/* Is it in the optimal CID space? */
+	if (likely(cid < max_cids))
+		return orig_cid;
+
+	/* Try to find one in the optimal space. Otherwise keep the provided. */
+	new_cid = __mm_get_cid(mm, max_cids);
+	if (new_cid != MM_CID_UNSET) {
+		mm_drop_cid(mm, cid);
+		/* Preserve the ONCPU mode of the original CID */
+		return new_cid | (orig_cid & MM_CID_ONCPU);
 	}
+	return orig_cid;
+}
 
-	/*
-	 * cid concurrently allocated. Retry while forcing following
-	 * allocations to use the cid_lock to ensure forward progress.
-	 */
-	WRITE_ONCE(use_cid_lock, 1);
-	/*
-	 * Set use_cid_lock before allocation. Only care about program order
-	 * because this is only required for forward progress.
-	 */
-	barrier();
-	/*
-	 * Retry until it succeeds. It is guaranteed to eventually succeed once
-	 * all newcoming allocations observe the use_cid_lock flag set.
-	 */
-	do {
-		cid = __mm_cid_try_get(t, mm);
-		cpu_relax();
-	} while (cid < 0);
-	/*
-	 * Allocate before clearing use_cid_lock. Only care about
-	 * program order because this is for forward progress.
-	 */
-	barrier();
-	WRITE_ONCE(use_cid_lock, 0);
-unlock:
-	raw_spin_unlock(&cid_lock);
-end:
-	mm_cid_snapshot_time(rq, mm);
+static __always_inline void mm_cid_update_task_cid(struct task_struct *t, unsigned int cid)
+{
+	if (t->mm_cid.cid != cid) {
+		t->mm_cid.cid = cid;
+		rseq_sched_set_ids_changed(t);
+	}
+}
 
-	return cid;
+static __always_inline void mm_cid_update_pcpu_cid(struct mm_struct *mm, unsigned int cid)
+{
+	__this_cpu_write(mm->mm_cid.pcpu->cid, cid);
 }
 
-static inline int mm_cid_get(struct rq *rq, struct task_struct *t,
-			     struct mm_struct *mm)
+static __always_inline void mm_cid_from_cpu(struct task_struct *t, unsigned int cpu_cid)
 {
-	struct mm_cid __percpu *pcpu_cid = mm->pcpu_cid;
-	int cid;
+	unsigned int max_cids, tcid = t->mm_cid.cid;
+	struct mm_struct *mm = t->mm;
 
-	lockdep_assert_rq_held(rq);
-	cid = __this_cpu_read(pcpu_cid->cid);
-	if (mm_cid_is_valid(cid)) {
-		mm_cid_snapshot_time(rq, mm);
-		return cid;
-	}
-	if (mm_cid_is_lazy_put(cid)) {
-		if (try_cmpxchg(&this_cpu_ptr(pcpu_cid)->cid, &cid, MM_CID_UNSET))
-			__mm_cid_put(mm, mm_cid_clear_lazy_put(cid));
+	max_cids = READ_ONCE(mm->mm_cid.max_cids);
+	/* Optimize for the common case where both have the ONCPU bit set */
+	if (likely(cid_on_cpu(cpu_cid & tcid))) {
+		if (likely(cpu_cid_to_cid(cpu_cid) < max_cids)) {
+			mm_cid_update_task_cid(t, cpu_cid);
+			return;
+		}
+		/* Try to converge into the optimal CID space */
+		cpu_cid = mm_cid_converge(mm, cpu_cid, max_cids);
+	} else {
+		/* Hand over or drop the task owned CID */
+		if (cid_on_task(tcid)) {
+			if (cid_on_cpu(cpu_cid))
+				mm_unset_cid_on_task(t);
+			else
+				cpu_cid = cid_to_cpu_cid(tcid);
+		}
+		/* Still nothing, allocate a new one */
+		if (!cid_on_cpu(cpu_cid))
+			cpu_cid = cid_to_cpu_cid(mm_get_cid(mm));
 	}
-	cid = __mm_cid_get(rq, t, mm);
-	__this_cpu_write(pcpu_cid->cid, cid);
-	__this_cpu_write(pcpu_cid->recent_cid, cid);
-
-	return cid;
+	mm_cid_update_pcpu_cid(mm, cpu_cid);
+	mm_cid_update_task_cid(t, cpu_cid);
 }
 
-static inline void switch_mm_cid(struct rq *rq,
-				 struct task_struct *prev,
-				 struct task_struct *next)
+static __always_inline void mm_cid_from_task(struct task_struct *t, unsigned int cpu_cid)
 {
-	/*
-	 * Provide a memory barrier between rq->curr store and load of
-	 * {prev,next}->mm->pcpu_cid[cpu] on rq->curr->mm transition.
-	 *
-	 * Should be adapted if context_switch() is modified.
-	 */
-	if (!next->mm) {                                // to kernel
-		/*
-		 * user -> kernel transition does not guarantee a barrier, but
-		 * we can use the fact that it performs an atomic operation in
-		 * mmgrab().
-		 */
-		if (prev->mm)                           // from user
-			smp_mb__after_mmgrab();
-		/*
-		 * kernel -> kernel transition does not change rq->curr->mm
-		 * state. It stays NULL.
-		 */
-	} else {                                        // to user
-		/*
-		 * kernel -> user transition does not provide a barrier
-		 * between rq->curr store and load of {prev,next}->mm->pcpu_cid[cpu].
-		 * Provide it here.
-		 */
-		if (!prev->mm) {                        // from kernel
-			smp_mb();
-		} else {				// from user
-			/*
-			 * user->user transition relies on an implicit
-			 * memory barrier in switch_mm() when
-			 * current->mm changes. If the architecture
-			 * switch_mm() does not have an implicit memory
-			 * barrier, it is emitted here.  If current->mm
-			 * is unchanged, no barrier is needed.
-			 */
-			smp_mb__after_switch_mm();
+	unsigned int max_cids, tcid = t->mm_cid.cid;
+	struct mm_struct *mm = t->mm;
+
+	max_cids = READ_ONCE(mm->mm_cid.max_cids);
+	/* Optimize for the common case, where both have the ONCPU bit clear */
+	if (likely(cid_on_task(tcid | cpu_cid))) {
+		if (likely(tcid < max_cids)) {
+			mm_cid_update_pcpu_cid(mm, tcid);
+			return;
 		}
+		/* Try to converge into the optimal CID space */
+		tcid = mm_cid_converge(mm, tcid, max_cids);
+	} else {
+		/* Hand over or drop the CPU owned CID */
+		if (cid_on_cpu(cpu_cid)) {
+			if (cid_on_task(tcid))
+				mm_drop_cid_on_cpu(mm, this_cpu_ptr(mm->mm_cid.pcpu));
+			else
+				tcid = cpu_cid_to_cid(cpu_cid);
+		}
+		/* Still nothing, allocate a new one */
+		if (!cid_on_task(tcid))
+			tcid = mm_get_cid(mm);
+		/* Set the transition mode flag if required */
+		tcid |= READ_ONCE(mm->mm_cid.transit);
 	}
-	if (prev->mm_cid_active) {
-		mm_cid_snapshot_time(rq, prev->mm);
-		mm_cid_put_lazy(prev);
-		prev->mm_cid = -1;
-	}
-	if (next->mm_cid_active)
-		next->last_mm_cid = next->mm_cid = mm_cid_get(rq, next, next->mm);
+	mm_cid_update_pcpu_cid(mm, tcid);
+	mm_cid_update_task_cid(t, tcid);
+}
+
+static __always_inline void mm_cid_schedin(struct task_struct *next)
+{
+	struct mm_struct *mm = next->mm;
+	unsigned int cpu_cid;
+
+	if (!next->mm_cid.active)
+		return;
+
+	cpu_cid = __this_cpu_read(mm->mm_cid.pcpu->cid);
+	if (likely(!READ_ONCE(mm->mm_cid.percpu)))
+		mm_cid_from_task(next, cpu_cid);
+	else
+		mm_cid_from_cpu(next, cpu_cid);
+}
+
+static __always_inline void mm_cid_schedout(struct task_struct *prev)
+{
+	/* During mode transitions CIDs are temporary and need to be dropped */
+	if (likely(!cid_in_transit(prev->mm_cid.cid)))
+		return;
+
+	mm_drop_cid(prev->mm, cid_from_transit_cid(prev->mm_cid.cid));
+	prev->mm_cid.cid = MM_CID_UNSET;
+}
+
+static inline void mm_cid_switch_to(struct task_struct *prev, struct task_struct *next)
+{
+	mm_cid_schedout(prev);
+	mm_cid_schedin(next);
 }
 
 #else /* !CONFIG_SCHED_MM_CID: */
-static inline void switch_mm_cid(struct rq *rq, struct task_struct *prev, struct task_struct *next) { }
-static inline void sched_mm_cid_migrate_from(struct task_struct *t) { }
-static inline void sched_mm_cid_migrate_to(struct rq *dst_rq, struct task_struct *t) { }
-static inline void task_tick_mm_cid(struct rq *rq, struct task_struct *curr) { }
-static inline void init_sched_mm_cid(struct task_struct *t) { }
+static inline void mm_cid_switch_to(struct task_struct *prev, struct task_struct *next) { }
 #endif /* !CONFIG_SCHED_MM_CID */
 
 extern u64 avg_vruntime(struct cfs_rq *cfs_rq);