Merge tag 'perf-kprobes-2020-12-14' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip

Pull perf/kprobes updates from Thomas Gleixner:
 "Make kretprobes lockless to avoid the rp->lock performance and
  potential lock ordering issues"

* tag 'perf-kprobes-2020-12-14' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  locking/atomics: Regenerate the atomics-check SHA1's
  kprobes: Replace rp->free_instance with freelist
  freelist: Implement lockless freelist
  asm-generic/atomic: Add try_cmpxchg() fallbacks
  kprobes: Remove kretprobe hash
  llist: Add nonatomic __llist_add() and __llist_dell_all()

3 files changed, 98 insertions, 223 deletions

diff --git a/kernel/fork.c b/kernel/fork.c
index 09be1be28cde..1513832de220 100644
--- a/kernel/fork.c
+++ b/kernel/fork.c
@@ -2183,6 +2183,10 @@ static __latent_entropy struct task_struct *copy_process(
 	INIT_LIST_HEAD(&p->thread_group);
 	p->task_works = NULL;
 
+#ifdef CONFIG_KRETPROBES
+	p->kretprobe_instances.first = NULL;
+#endif
+
 	/*
 	 * Ensure that the cgroup subsystem policies allow the new process to be
 	 * forked. It should be noted that the new process's css_set can be changed
diff --git a/kernel/kprobes.c b/kernel/kprobes.c
index 41fdbb7953c6..f7fb5d135930 100644
--- a/kernel/kprobes.c
+++ b/kernel/kprobes.c
@@ -36,7 +36,6 @@
 #include <linux/cpu.h>
 #include <linux/jump_label.h>
 #include <linux/perf_event.h>
-#include <linux/static_call.h>
 
 #include <asm/sections.h>
 #include <asm/cacheflush.h>
@@ -54,7 +53,6 @@ static int kprobes_initialized;
  * - RCU hlist traversal under disabling preempt (breakpoint handlers)
  */
 static struct hlist_head kprobe_table[KPROBE_TABLE_SIZE];
-static struct hlist_head kretprobe_inst_table[KPROBE_TABLE_SIZE];
 
 /* NOTE: change this value only with kprobe_mutex held */
 static bool kprobes_all_disarmed;
@@ -62,9 +60,6 @@ static bool kprobes_all_disarmed;
 /* This protects kprobe_table and optimizing_list */
 static DEFINE_MUTEX(kprobe_mutex);
 static DEFINE_PER_CPU(struct kprobe *, kprobe_instance) = NULL;
-static struct {
-	raw_spinlock_t lock ____cacheline_aligned_in_smp;
-} kretprobe_table_locks[KPROBE_TABLE_SIZE];
 
 kprobe_opcode_t * __weak kprobe_lookup_name(const char *name,
 					unsigned int __unused)
@@ -72,11 +67,6 @@ kprobe_opcode_t * __weak kprobe_lookup_name(const char *name,
 	return ((kprobe_opcode_t *)(kallsyms_lookup_name(name)));
 }
 
-static raw_spinlock_t *kretprobe_table_lock_ptr(unsigned long hash)
-{
-	return &(kretprobe_table_locks[hash].lock);
-}
-
 /* Blacklist -- list of struct kprobe_blacklist_entry */
 static LIST_HEAD(kprobe_blacklist);
 
@@ -1224,76 +1214,26 @@ void kprobes_inc_nmissed_count(struct kprobe *p)
 }
 NOKPROBE_SYMBOL(kprobes_inc_nmissed_count);
 
-static void recycle_rp_inst(struct kretprobe_instance *ri)
+static void free_rp_inst_rcu(struct rcu_head *head)
 {
-	struct kretprobe *rp = ri->rp;
+	struct kretprobe_instance *ri = container_of(head, struct kretprobe_instance, rcu);
 
-	/* remove rp inst off the rprobe_inst_table */
-	hlist_del(&ri->hlist);
-	INIT_HLIST_NODE(&ri->hlist);
-	if (likely(rp)) {
-		raw_spin_lock(&rp->lock);
-		hlist_add_head(&ri->hlist, &rp->free_instances);
-		raw_spin_unlock(&rp->lock);
-	} else
-		kfree_rcu(ri, rcu);
+	if (refcount_dec_and_test(&ri->rph->ref))
+		kfree(ri->rph);
+	kfree(ri);
 }
-NOKPROBE_SYMBOL(recycle_rp_inst);
-
-static void kretprobe_hash_lock(struct task_struct *tsk,
-			 struct hlist_head **head, unsigned long *flags)
-__acquires(hlist_lock)
-{
-	unsigned long hash = hash_ptr(tsk, KPROBE_HASH_BITS);
-	raw_spinlock_t *hlist_lock;
+NOKPROBE_SYMBOL(free_rp_inst_rcu);
 
-	*head = &kretprobe_inst_table[hash];
-	hlist_lock = kretprobe_table_lock_ptr(hash);
-	/*
-	 * Nested is a workaround that will soon not be needed.
-	 * There's other protections that make sure the same lock
-	 * is not taken on the same CPU that lockdep is unaware of.
-	 * Differentiate when it is taken in NMI context.
-	 */
-	raw_spin_lock_irqsave_nested(hlist_lock, *flags, !!in_nmi());
-}
-NOKPROBE_SYMBOL(kretprobe_hash_lock);
-
-static void kretprobe_table_lock(unsigned long hash,
-				 unsigned long *flags)
-__acquires(hlist_lock)
-{
-	raw_spinlock_t *hlist_lock = kretprobe_table_lock_ptr(hash);
-	/*
-	 * Nested is a workaround that will soon not be needed.
-	 * There's other protections that make sure the same lock
-	 * is not taken on the same CPU that lockdep is unaware of.
-	 * Differentiate when it is taken in NMI context.
-	 */
-	raw_spin_lock_irqsave_nested(hlist_lock, *flags, !!in_nmi());
-}
-NOKPROBE_SYMBOL(kretprobe_table_lock);
-
-static void kretprobe_hash_unlock(struct task_struct *tsk,
-			   unsigned long *flags)
-__releases(hlist_lock)
+static void recycle_rp_inst(struct kretprobe_instance *ri)
 {
-	unsigned long hash = hash_ptr(tsk, KPROBE_HASH_BITS);
-	raw_spinlock_t *hlist_lock;
-
-	hlist_lock = kretprobe_table_lock_ptr(hash);
-	raw_spin_unlock_irqrestore(hlist_lock, *flags);
-}
-NOKPROBE_SYMBOL(kretprobe_hash_unlock);
+	struct kretprobe *rp = get_kretprobe(ri);
 
-static void kretprobe_table_unlock(unsigned long hash,
-				   unsigned long *flags)
-__releases(hlist_lock)
-{
-	raw_spinlock_t *hlist_lock = kretprobe_table_lock_ptr(hash);
-	raw_spin_unlock_irqrestore(hlist_lock, *flags);
+	if (likely(rp)) {
+		freelist_add(&ri->freelist, &rp->freelist);
+	} else
+		call_rcu(&ri->rcu, free_rp_inst_rcu);
 }
-NOKPROBE_SYMBOL(kretprobe_table_unlock);
+NOKPROBE_SYMBOL(recycle_rp_inst);
 
 static struct kprobe kprobe_busy = {
 	.addr = (void *) get_kprobe,
@@ -1324,24 +1264,21 @@ void kprobe_busy_end(void)
 void kprobe_flush_task(struct task_struct *tk)
 {
 	struct kretprobe_instance *ri;
-	struct hlist_head *head;
-	struct hlist_node *tmp;
-	unsigned long hash, flags = 0;
+	struct llist_node *node;
 
+	/* Early boot, not yet initialized. */
 	if (unlikely(!kprobes_initialized))
-		/* Early boot.  kretprobe_table_locks not yet initialized. */
 		return;
 
 	kprobe_busy_begin();
 
-	hash = hash_ptr(tk, KPROBE_HASH_BITS);
-	head = &kretprobe_inst_table[hash];
-	kretprobe_table_lock(hash, &flags);
-	hlist_for_each_entry_safe(ri, tmp, head, hlist) {
-		if (ri->task == tk)
-			recycle_rp_inst(ri);
+	node = __llist_del_all(&tk->kretprobe_instances);
+	while (node) {
+		ri = container_of(node, struct kretprobe_instance, llist);
+		node = node->next;
+
+		recycle_rp_inst(ri);
 	}
-	kretprobe_table_unlock(hash, &flags);
 
 	kprobe_busy_end();
 }
@@ -1350,37 +1287,23 @@ NOKPROBE_SYMBOL(kprobe_flush_task);
 static inline void free_rp_inst(struct kretprobe *rp)
 {
 	struct kretprobe_instance *ri;
-	struct hlist_node *next;
+	struct freelist_node *node;
+	int count = 0;
+
+	node = rp->freelist.head;
+	while (node) {
+		ri = container_of(node, struct kretprobe_instance, freelist);
+		node = node->next;
 
-	hlist_for_each_entry_safe(ri, next, &rp->free_instances, hlist) {
-		hlist_del(&ri->hlist);
 		kfree(ri);
+		count++;
 	}
-}
-
-static void cleanup_rp_inst(struct kretprobe *rp)
-{
-	unsigned long flags, hash;
-	struct kretprobe_instance *ri;
-	struct hlist_node *next;
-	struct hlist_head *head;
 
-	/* To avoid recursive kretprobe by NMI, set kprobe busy here */
-	kprobe_busy_begin();
-	for (hash = 0; hash < KPROBE_TABLE_SIZE; hash++) {
-		kretprobe_table_lock(hash, &flags);
-		head = &kretprobe_inst_table[hash];
-		hlist_for_each_entry_safe(ri, next, head, hlist) {
-			if (ri->rp == rp)
-				ri->rp = NULL;
-		}
-		kretprobe_table_unlock(hash, &flags);
+	if (refcount_sub_and_test(count, &rp->rph->ref)) {
+		kfree(rp->rph);
+		rp->rph = NULL;
 	}
-	kprobe_busy_end();
-
-	free_rp_inst(rp);
 }
-NOKPROBE_SYMBOL(cleanup_rp_inst);
 
 /* Add the new probe to ap->list */
 static int add_new_kprobe(struct kprobe *ap, struct kprobe *p)
@@ -1643,7 +1566,6 @@ static int check_kprobe_address_safe(struct kprobe *p,
 	if (!kernel_text_address((unsigned long) p->addr) ||
 	    within_kprobe_blacklist((unsigned long) p->addr) ||
 	    jump_label_text_reserved(p->addr, p->addr) ||
-	    static_call_text_reserved(p->addr, p->addr) ||
 	    find_bug((unsigned long)p->addr)) {
 		ret = -EINVAL;
 		goto out;
@@ -1942,88 +1864,56 @@ unsigned long __kretprobe_trampoline_handler(struct pt_regs *regs,
 					     void *trampoline_address,
 					     void *frame_pointer)
 {
-	struct kretprobe_instance *ri = NULL, *last = NULL;
-	struct hlist_head *head;
-	struct hlist_node *tmp;
-	unsigned long flags;
 	kprobe_opcode_t *correct_ret_addr = NULL;
-	bool skipped = false;
+	struct kretprobe_instance *ri = NULL;
+	struct llist_node *first, *node;
+	struct kretprobe *rp;
 
-	kretprobe_hash_lock(current, &head, &flags);
+	/* Find all nodes for this frame. */
+	first = node = current->kretprobe_instances.first;
+	while (node) {
+		ri = container_of(node, struct kretprobe_instance, llist);
 
-	/*
-	 * It is possible to have multiple instances associated with a given
-	 * task either because multiple functions in the call path have
-	 * return probes installed on them, and/or more than one
-	 * return probe was registered for a target function.
-	 *
-	 * We can handle this because:
-	 *     - instances are always pushed into the head of the list
-	 *     - when multiple return probes are registered for the same
-	 *	 function, the (chronologically) first instance's ret_addr
-	 *	 will be the real return address, and all the rest will
-	 *	 point to kretprobe_trampoline.
-	 */
-	hlist_for_each_entry(ri, head, hlist) {
-		if (ri->task != current)
-			/* another task is sharing our hash bucket */
-			continue;
-		/*
-		 * Return probes must be pushed on this hash list correct
-		 * order (same as return order) so that it can be popped
-		 * correctly. However, if we find it is pushed it incorrect
-		 * order, this means we find a function which should not be
-		 * probed, because the wrong order entry is pushed on the
-		 * path of processing other kretprobe itself.
-		 */
-		if (ri->fp != frame_pointer) {
-			if (!skipped)
-				pr_warn("kretprobe is stacked incorrectly. Trying to fixup.\n");
-			skipped = true;
-			continue;
-		}
+		BUG_ON(ri->fp != frame_pointer);
 
-		correct_ret_addr = ri->ret_addr;
-		if (skipped)
-			pr_warn("%ps must be blacklisted because of incorrect kretprobe order\n",
-				ri->rp->kp.addr);
-
-		if (correct_ret_addr != trampoline_address)
+		if (ri->ret_addr != trampoline_address) {
+			correct_ret_addr = ri->ret_addr;
 			/*
 			 * This is the real return address. Any other
 			 * instances associated with this task are for
 			 * other calls deeper on the call stack
 			 */
-			break;
+			goto found;
+		}
+
+		node = node->next;
 	}
+	pr_err("Oops! Kretprobe fails to find correct return address.\n");
+	BUG_ON(1);
 
-	BUG_ON(!correct_ret_addr || (correct_ret_addr == trampoline_address));
-	last = ri;
+found:
+	/* Unlink all nodes for this frame. */
+	current->kretprobe_instances.first = node->next;
+	node->next = NULL;
 
-	hlist_for_each_entry_safe(ri, tmp, head, hlist) {
-		if (ri->task != current)
-			/* another task is sharing our hash bucket */
-			continue;
-		if (ri->fp != frame_pointer)
-			continue;
+	/* Run them..  */
+	while (first) {
+		ri = container_of(first, struct kretprobe_instance, llist);
+		first = first->next;
 
-		if (ri->rp && ri->rp->handler) {
+		rp = get_kretprobe(ri);
+		if (rp && rp->handler) {
 			struct kprobe *prev = kprobe_running();
 
-			__this_cpu_write(current_kprobe, &ri->rp->kp);
+			__this_cpu_write(current_kprobe, &rp->kp);
 			ri->ret_addr = correct_ret_addr;
-			ri->rp->handler(ri, regs);
+			rp->handler(ri, regs);
 			__this_cpu_write(current_kprobe, prev);
 		}
 
 		recycle_rp_inst(ri);
-
-		if (ri == last)
-			break;
 	}
 
-	kretprobe_hash_unlock(current, &flags);
-
 	return (unsigned long)correct_ret_addr;
 }
 NOKPROBE_SYMBOL(__kretprobe_trampoline_handler)
@@ -2035,44 +1925,26 @@ NOKPROBE_SYMBOL(__kretprobe_trampoline_handler)
 static int pre_handler_kretprobe(struct kprobe *p, struct pt_regs *regs)
 {
 	struct kretprobe *rp = container_of(p, struct kretprobe, kp);
-	unsigned long hash, flags = 0;
 	struct kretprobe_instance *ri;
+	struct freelist_node *fn;
 
-	/* TODO: consider to only swap the RA after the last pre_handler fired */
-	hash = hash_ptr(current, KPROBE_HASH_BITS);
-	/*
-	 * Nested is a workaround that will soon not be needed.
-	 * There's other protections that make sure the same lock
-	 * is not taken on the same CPU that lockdep is unaware of.
-	 */
-	raw_spin_lock_irqsave_nested(&rp->lock, flags, 1);
-	if (!hlist_empty(&rp->free_instances)) {
-		ri = hlist_entry(rp->free_instances.first,
-				struct kretprobe_instance, hlist);
-		hlist_del(&ri->hlist);
-		raw_spin_unlock_irqrestore(&rp->lock, flags);
-
-		ri->rp = rp;
-		ri->task = current;
-
-		if (rp->entry_handler && rp->entry_handler(ri, regs)) {
-			raw_spin_lock_irqsave_nested(&rp->lock, flags, 1);
-			hlist_add_head(&ri->hlist, &rp->free_instances);
-			raw_spin_unlock_irqrestore(&rp->lock, flags);
-			return 0;
-		}
+	fn = freelist_try_get(&rp->freelist);
+	if (!fn) {
+		rp->nmissed++;
+		return 0;
+	}
 
-		arch_prepare_kretprobe(ri, regs);
+	ri = container_of(fn, struct kretprobe_instance, freelist);
 
-		/* XXX(hch): why is there no hlist_move_head? */
-		INIT_HLIST_NODE(&ri->hlist);
-		kretprobe_table_lock(hash, &flags);
-		hlist_add_head(&ri->hlist, &kretprobe_inst_table[hash]);
-		kretprobe_table_unlock(hash, &flags);
-	} else {
-		rp->nmissed++;
-		raw_spin_unlock_irqrestore(&rp->lock, flags);
+	if (rp->entry_handler && rp->entry_handler(ri, regs)) {
+		freelist_add(&ri->freelist, &rp->freelist);
+		return 0;
 	}
+
+	arch_prepare_kretprobe(ri, regs);
+
+	__llist_add(&ri->llist, &current->kretprobe_instances);
+
 	return 0;
 }
 NOKPROBE_SYMBOL(pre_handler_kretprobe);
@@ -2129,18 +2001,24 @@ int register_kretprobe(struct kretprobe *rp)
 		rp->maxactive = num_possible_cpus();
 #endif
 	}
-	raw_spin_lock_init(&rp->lock);
-	INIT_HLIST_HEAD(&rp->free_instances);
+	rp->freelist.head = NULL;
+	rp->rph = kzalloc(sizeof(struct kretprobe_holder), GFP_KERNEL);
+	if (!rp->rph)
+		return -ENOMEM;
+
+	rp->rph->rp = rp;
 	for (i = 0; i < rp->maxactive; i++) {
-		inst = kmalloc(sizeof(struct kretprobe_instance) +
+		inst = kzalloc(sizeof(struct kretprobe_instance) +
 			       rp->data_size, GFP_KERNEL);
 		if (inst == NULL) {
+			refcount_set(&rp->rph->ref, i);
 			free_rp_inst(rp);
 			return -ENOMEM;
 		}
-		INIT_HLIST_NODE(&inst->hlist);
-		hlist_add_head(&inst->hlist, &rp->free_instances);
+		inst->rph = rp->rph;
+		freelist_add(&inst->freelist, &rp->freelist);
 	}
+	refcount_set(&rp->rph->ref, i);
 
 	rp->nmissed = 0;
 	/* Establish function entry probe point */
@@ -2182,16 +2060,18 @@ void unregister_kretprobes(struct kretprobe **rps, int num)
 	if (num <= 0)
 		return;
 	mutex_lock(&kprobe_mutex);
-	for (i = 0; i < num; i++)
+	for (i = 0; i < num; i++) {
 		if (__unregister_kprobe_top(&rps[i]->kp) < 0)
 			rps[i]->kp.addr = NULL;
+		rps[i]->rph->rp = NULL;
+	}
 	mutex_unlock(&kprobe_mutex);
 
 	synchronize_rcu();
 	for (i = 0; i < num; i++) {
 		if (rps[i]->kp.addr) {
 			__unregister_kprobe_bottom(&rps[i]->kp);
-			cleanup_rp_inst(rps[i]);
+			free_rp_inst(rps[i]);
 		}
 	}
 }
@@ -2235,9 +2115,6 @@ static void kill_kprobe(struct kprobe *p)
 
 	lockdep_assert_held(&kprobe_mutex);
 
-	if (WARN_ON_ONCE(kprobe_gone(p)))
-		return;
-
 	p->flags |= KPROBE_FLAG_GONE;
 	if (kprobe_aggrprobe(p)) {
 		/*
@@ -2518,10 +2395,7 @@ static int kprobes_module_callback(struct notifier_block *nb,
 	mutex_lock(&kprobe_mutex);
 	for (i = 0; i < KPROBE_TABLE_SIZE; i++) {
 		head = &kprobe_table[i];
-		hlist_for_each_entry(p, head, hlist) {
-			if (kprobe_gone(p))
-				continue;
-
+		hlist_for_each_entry(p, head, hlist)
 			if (within_module_init((unsigned long)p->addr, mod) ||
 			    (checkcore &&
 			     within_module_core((unsigned long)p->addr, mod))) {
@@ -2538,7 +2412,6 @@ static int kprobes_module_callback(struct notifier_block *nb,
 				 */
 				kill_kprobe(p);
 			}
-		}
 	}
 	if (val == MODULE_STATE_GOING)
 		remove_module_kprobe_blacklist(mod);
@@ -2583,11 +2456,8 @@ static int __init init_kprobes(void)
 
 	/* FIXME allocate the probe table, currently defined statically */
 	/* initialize all list heads */
-	for (i = 0; i < KPROBE_TABLE_SIZE; i++) {
+	for (i = 0; i < KPROBE_TABLE_SIZE; i++)
 		INIT_HLIST_HEAD(&kprobe_table[i]);
-		INIT_HLIST_HEAD(&kretprobe_inst_table[i]);
-		raw_spin_lock_init(&(kretprobe_table_locks[i].lock));
-	}
 
 	err = populate_kprobe_blacklist(__start_kprobe_blacklist,
 					__stop_kprobe_blacklist);
diff --git a/kernel/trace/trace_kprobe.c b/kernel/trace/trace_kprobe.c
index b911e9f6d9f5..97c7a7782db7 100644
--- a/kernel/trace/trace_kprobe.c
+++ b/kernel/trace/trace_kprobe.c
@@ -1731,7 +1731,8 @@ NOKPROBE_SYMBOL(kprobe_dispatcher);
 static int
 kretprobe_dispatcher(struct kretprobe_instance *ri, struct pt_regs *regs)
 {
-	struct trace_kprobe *tk = container_of(ri->rp, struct trace_kprobe, rp);
+	struct kretprobe *rp = get_kretprobe(ri);
+	struct trace_kprobe *tk = container_of(rp, struct trace_kprobe, rp);
 
 	raw_cpu_inc(*tk->nhit);