From 73bd1227787bfe73eea3d04c63a89cb55db9c23e Mon Sep 17 00:00:00 2001 From: Herbert Xu Date: Sat, 18 Apr 2026 09:41:21 +0800 Subject: rhashtable: Restore insecure_elasticity toggle Some users of rhashtable cannot handle insertion failures, and are happy to accept the consequences of a hash table that having very long chains. Restore the insecure_elasticity toggle for these users. In addition to disabling the chain length checks, this also removes the emergency resize that would otherwise occur when the hash table occupancy hits 100% (an async resize is still scheduled at 75%). Signed-off-by: Herbert Xu Signed-off-by: Tejun Heo --- include/linux/rhashtable-types.h | 2 ++ include/linux/rhashtable.h | 5 +++-- 2 files changed, 5 insertions(+), 2 deletions(-) (limited to 'include') diff --git a/include/linux/rhashtable-types.h b/include/linux/rhashtable-types.h index 015c8298bebc..72082428d6c6 100644 --- a/include/linux/rhashtable-types.h +++ b/include/linux/rhashtable-types.h @@ -49,6 +49,7 @@ typedef int (*rht_obj_cmpfn_t)(struct rhashtable_compare_arg *arg, * @head_offset: Offset of rhash_head in struct to be hashed * @max_size: Maximum size while expanding * @min_size: Minimum size while shrinking + * @insecure_elasticity: Set to true to disable chain length checks * @automatic_shrinking: Enable automatic shrinking of tables * @hashfn: Hash function (default: jhash2 if !(key_len % 4), or jhash) * @obj_hashfn: Function to hash object @@ -61,6 +62,7 @@ struct rhashtable_params { u16 head_offset; unsigned int max_size; u16 min_size; + bool insecure_elasticity; bool automatic_shrinking; rht_hashfn_t hashfn; rht_obj_hashfn_t obj_hashfn; diff --git a/include/linux/rhashtable.h b/include/linux/rhashtable.h index 0480509a6339..7def3f0f556b 100644 --- a/include/linux/rhashtable.h +++ b/include/linux/rhashtable.h @@ -821,14 +821,15 @@ slow_path: goto out; } - if (elasticity <= 0) + if (elasticity <= 0 && !params.insecure_elasticity) goto slow_path; data = ERR_PTR(-E2BIG); if (unlikely(rht_grow_above_max(ht, tbl))) goto out_unlock; - if (unlikely(rht_grow_above_100(ht, tbl))) + if (unlikely(rht_grow_above_100(ht, tbl)) && + !params.insecure_elasticity) goto slow_path; /* Inserting at head of list makes unlocking free. */ -- cgit v1.2.3 From 4fe985292709eeb6a4653c71660f893e26c2f2dd Mon Sep 17 00:00:00 2001 From: Tejun Heo Date: Mon, 20 Apr 2026 20:03:26 -1000 Subject: rhashtable: Bounce deferred worker kick through irq_work Inserts past 75% load call schedule_work(&ht->run_work) to kick an async resize. If a caller holds a raw spinlock (e.g. an insecure_elasticity user), schedule_work() under that lock records caller_lock -> pool->lock -> pi_lock -> rq->__lock A cycle forms if any of these locks is acquired in the reverse direction elsewhere. sched_ext, the only current insecure_elasticity user, hits this: it holds scx_sched_lock across rhashtable inserts of sub-schedulers, while scx_bypass() takes rq->__lock -> scx_sched_lock. Exercising the resize path produces: Chain exists of: &pool->lock --> &rq->__lock --> scx_sched_lock Bounce the kick from the insert paths through irq_work so schedule_work() runs from hard IRQ context with the caller's lock no longer held. rht_deferred_worker()'s self-rearm on error stays on schedule_work(&ht->run_work) - the worker runs in process context with no caller lock held, and keeping the self-requeue on @run_work lets cancel_work_sync() in rhashtable_free_and_destroy() drain it. v3: Keep rht_deferred_worker()'s self-rearm on schedule_work(&run_work). Routing it through irq_work in v2 broke cancel_work_sync()'s self-requeue handling - an irq_work queued after irq_work_sync() returned but while cancel_work_sync() was still waiting could fire post-teardown. v2: Bounce unconditionally instead of gating on insecure_elasticity, as suggested by Herbert. Signed-off-by: Tejun Heo Acked-by: Herbert Xu --- include/linux/rhashtable-types.h | 3 +++ include/linux/rhashtable.h | 3 ++- lib/rhashtable.c | 31 ++++++++++++++++++++++++++++--- 3 files changed, 33 insertions(+), 4 deletions(-) (limited to 'include') diff --git a/include/linux/rhashtable-types.h b/include/linux/rhashtable-types.h index 72082428d6c6..fc2f596a6df1 100644 --- a/include/linux/rhashtable-types.h +++ b/include/linux/rhashtable-types.h @@ -12,6 +12,7 @@ #include #include #include +#include #include #include @@ -77,6 +78,7 @@ struct rhashtable_params { * @p: Configuration parameters * @rhlist: True if this is an rhltable * @run_work: Deferred worker to expand/shrink asynchronously + * @run_irq_work: Bounces the @run_work kick through hard IRQ context. * @mutex: Mutex to protect current/future table swapping * @lock: Spin lock to protect walker list * @nelems: Number of elements in table @@ -88,6 +90,7 @@ struct rhashtable { struct rhashtable_params p; bool rhlist; struct work_struct run_work; + struct irq_work run_irq_work; struct mutex mutex; spinlock_t lock; atomic_t nelems; diff --git a/include/linux/rhashtable.h b/include/linux/rhashtable.h index 7def3f0f556b..ef5230cece36 100644 --- a/include/linux/rhashtable.h +++ b/include/linux/rhashtable.h @@ -20,6 +20,7 @@ #include #include +#include #include #include #include @@ -847,7 +848,7 @@ slow_path: rht_assign_unlock(tbl, bkt, obj, flags); if (rht_grow_above_75(ht, tbl)) - schedule_work(&ht->run_work); + irq_work_queue(&ht->run_irq_work); data = NULL; out: diff --git a/lib/rhashtable.c b/lib/rhashtable.c index fb2b7bc137ba..7a67ef5b67b6 100644 --- a/lib/rhashtable.c +++ b/lib/rhashtable.c @@ -441,10 +441,33 @@ static void rht_deferred_worker(struct work_struct *work) mutex_unlock(&ht->mutex); + /* + * Re-arm via @run_work, not @run_irq_work. + * rhashtable_free_and_destroy() drains async work as irq_work_sync() + * followed by cancel_work_sync(). If this site queued irq_work while + * cancel_work_sync() was waiting for us, irq_work_sync() would already + * have returned and the stale irq_work could fire post-teardown. + * cancel_work_sync() natively handles self-requeue on @run_work. + */ if (err) schedule_work(&ht->run_work); } +/* + * Insert-path callers can run under a raw spinlock (e.g. an insecure_elasticity + * user). Calling schedule_work() under that lock records caller_lock -> + * pool->lock -> pi_lock -> rq->__lock, closing a locking cycle if any of + * these is acquired in the reverse direction elsewhere. Bounce through + * irq_work so the schedule_work() runs with the caller's lock no longer held. + */ +static void rht_deferred_irq_work(struct irq_work *irq_work) +{ + struct rhashtable *ht = container_of(irq_work, struct rhashtable, + run_irq_work); + + schedule_work(&ht->run_work); +} + static int rhashtable_insert_rehash(struct rhashtable *ht, struct bucket_table *tbl) { @@ -477,7 +500,7 @@ static int rhashtable_insert_rehash(struct rhashtable *ht, if (err == -EEXIST) err = 0; } else - schedule_work(&ht->run_work); + irq_work_queue(&ht->run_irq_work); return err; @@ -488,7 +511,7 @@ fail: /* Schedule async rehash to retry allocation in process context. */ if (err == -ENOMEM) - schedule_work(&ht->run_work); + irq_work_queue(&ht->run_irq_work); return err; } @@ -630,7 +653,7 @@ static void *rhashtable_try_insert(struct rhashtable *ht, const void *key, rht_unlock(tbl, bkt, flags); if (inserted && rht_grow_above_75(ht, tbl)) - schedule_work(&ht->run_work); + irq_work_queue(&ht->run_irq_work); } } while (!IS_ERR_OR_NULL(new_tbl)); @@ -1085,6 +1108,7 @@ int rhashtable_init_noprof(struct rhashtable *ht, RCU_INIT_POINTER(ht->tbl, tbl); INIT_WORK(&ht->run_work, rht_deferred_worker); + init_irq_work(&ht->run_irq_work, rht_deferred_irq_work); return 0; } @@ -1150,6 +1174,7 @@ void rhashtable_free_and_destroy(struct rhashtable *ht, struct bucket_table *tbl, *next_tbl; unsigned int i; + irq_work_sync(&ht->run_irq_work); cancel_work_sync(&ht->run_work); mutex_lock(&ht->mutex); -- cgit v1.2.3