[mirror_ubuntu-bionic-kernel.git] / mm / kasan / quarantine.c

/*
 * KASAN quarantine.
 *
 * Author: Alexander Potapenko <glider@google.com>
 * Copyright (C) 2016 Google, Inc.
 *
 * Based on code by Dmitry Chernenkov.
 *
 * This program is free software; you can redistribute it and/or
 * modify it under the terms of the GNU General Public License
 * version 2 as published by the Free Software Foundation.
 *
 * This program is distributed in the hope that it will be useful, but
 * WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * General Public License for more details.
 *
 */

#include <linux/gfp.h>
#include <linux/hash.h>
#include <linux/kernel.h>
#include <linux/mm.h>
#include <linux/percpu.h>
#include <linux/printk.h>
#include <linux/shrinker.h>
#include <linux/slab.h>
#include <linux/srcu.h>
#include <linux/string.h>
#include <linux/types.h>

#include "../slab.h"
#include "kasan.h"

/* Data structure and operations for quarantine queues. */

/*
 * Each queue is a signle-linked list, which also stores the total size of
 * objects inside of it.
 */
struct qlist_head {
	struct qlist_node *head;
	struct qlist_node *tail;
	size_t bytes;
};

#define QLIST_INIT { NULL, NULL, 0 }

static bool qlist_empty(struct qlist_head *q)
{
	return !q->head;
}

static void qlist_init(struct qlist_head *q)
{
	q->head = q->tail = NULL;
	q->bytes = 0;
}

static void qlist_put(struct qlist_head *q, struct qlist_node *qlink,
		size_t size)
{
	if (unlikely(qlist_empty(q)))
		q->head = qlink;
	else
		q->tail->next = qlink;
	q->tail = qlink;
	qlink->next = NULL;
	q->bytes += size;
}

static void qlist_move_all(struct qlist_head *from, struct qlist_head *to)
{
	if (unlikely(qlist_empty(from)))
		return;

	if (qlist_empty(to)) {
		*to = *from;
		qlist_init(from);
		return;
	}

	to->tail->next = from->head;
	to->tail = from->tail;
	to->bytes += from->bytes;

	qlist_init(from);
}

#define QUARANTINE_PERCPU_SIZE (1 << 20)
#define QUARANTINE_BATCHES \
	(1024 > 4 * CONFIG_NR_CPUS ? 1024 : 4 * CONFIG_NR_CPUS)

/*
 * The object quarantine consists of per-cpu queues and a global queue,
 * guarded by quarantine_lock.
 */
static DEFINE_PER_CPU(struct qlist_head, cpu_quarantine);

/* Round-robin FIFO array of batches. */
static struct qlist_head global_quarantine[QUARANTINE_BATCHES];
static int quarantine_head;
static int quarantine_tail;
/* Total size of all objects in global_quarantine across all batches. */
static unsigned long quarantine_size;
static DEFINE_SPINLOCK(quarantine_lock);
DEFINE_STATIC_SRCU(remove_cache_srcu);

/* Maximum size of the global queue. */
static unsigned long quarantine_max_size;

/*
 * Target size of a batch in global_quarantine.
 * Usually equal to QUARANTINE_PERCPU_SIZE unless we have too much RAM.
 */
static unsigned long quarantine_batch_size;

/*
 * The fraction of physical memory the quarantine is allowed to occupy.
 * Quarantine doesn't support memory shrinker with SLAB allocator, so we keep
 * the ratio low to avoid OOM.
 */
#define QUARANTINE_FRACTION 32

static struct kmem_cache *qlink_to_cache(struct qlist_node *qlink)
{
	return virt_to_head_page(qlink)->slab_cache;
}

static void *qlink_to_object(struct qlist_node *qlink, struct kmem_cache *cache)
{
	struct kasan_free_meta *free_info =
		container_of(qlink, struct kasan_free_meta,
			     quarantine_link);

	return ((void *)free_info) - cache->kasan_info.free_meta_offset;
}

static void qlink_free(struct qlist_node *qlink, struct kmem_cache *cache)
{
	void *object = qlink_to_object(qlink, cache);
	unsigned long flags;

	if (IS_ENABLED(CONFIG_SLAB))
		local_irq_save(flags);

	___cache_free(cache, object, _THIS_IP_);

	if (IS_ENABLED(CONFIG_SLAB))
		local_irq_restore(flags);
}

static void qlist_free_all(struct qlist_head *q, struct kmem_cache *cache)
{
	struct qlist_node *qlink;

	if (unlikely(qlist_empty(q)))
		return;

	qlink = q->head;
	while (qlink) {
		struct kmem_cache *obj_cache =
			cache ? cache :	qlink_to_cache(qlink);
		struct qlist_node *next = qlink->next;

		qlink_free(qlink, obj_cache);
		qlink = next;
	}
	qlist_init(q);
}

void quarantine_put(struct kasan_free_meta *info, struct kmem_cache *cache)
{
	unsigned long flags;
	struct qlist_head *q;
	struct qlist_head temp = QLIST_INIT;

	/*
	 * Note: irq must be disabled until after we move the batch to the
	 * global quarantine. Otherwise quarantine_remove_cache() can miss
	 * some objects belonging to the cache if they are in our local temp
	 * list. quarantine_remove_cache() executes on_each_cpu() at the
	 * beginning which ensures that it either sees the objects in per-cpu
	 * lists or in the global quarantine.
	 */
	local_irq_save(flags);

	q = this_cpu_ptr(&cpu_quarantine);
	qlist_put(q, &info->quarantine_link, cache->size);
	if (unlikely(q->bytes > QUARANTINE_PERCPU_SIZE)) {
		qlist_move_all(q, &temp);

		spin_lock(&quarantine_lock);
		WRITE_ONCE(quarantine_size, quarantine_size + temp.bytes);
		qlist_move_all(&temp, &global_quarantine[quarantine_tail]);
		if (global_quarantine[quarantine_tail].bytes >=
				READ_ONCE(quarantine_batch_size)) {
			int new_tail;

			new_tail = quarantine_tail + 1;
			if (new_tail == QUARANTINE_BATCHES)
				new_tail = 0;
			if (new_tail != quarantine_head)
				quarantine_tail = new_tail;
		}
		spin_unlock(&quarantine_lock);
	}

	local_irq_restore(flags);
}

void quarantine_reduce(void)
{
	size_t total_size, new_quarantine_size, percpu_quarantines;
	unsigned long flags;
	int srcu_idx;
	struct qlist_head to_free = QLIST_INIT;

	if (likely(READ_ONCE(quarantine_size) <=
		   READ_ONCE(quarantine_max_size)))
		return;

	/*
	 * srcu critical section ensures that quarantine_remove_cache()
	 * will not miss objects belonging to the cache while they are in our
	 * local to_free list. srcu is chosen because (1) it gives us private
	 * grace period domain that does not interfere with anything else,
	 * and (2) it allows synchronize_srcu() to return without waiting
	 * if there are no pending read critical sections (which is the
	 * expected case).
	 */
	srcu_idx = srcu_read_lock(&remove_cache_srcu);
	spin_lock_irqsave(&quarantine_lock, flags);

	/*
	 * Update quarantine size in case of hotplug. Allocate a fraction of
	 * the installed memory to quarantine minus per-cpu queue limits.
	 */
	total_size = (READ_ONCE(totalram_pages) << PAGE_SHIFT) /
		QUARANTINE_FRACTION;
	percpu_quarantines = QUARANTINE_PERCPU_SIZE * num_online_cpus();
	new_quarantine_size = (total_size < percpu_quarantines) ?
		0 : total_size - percpu_quarantines;
	WRITE_ONCE(quarantine_max_size, new_quarantine_size);
	/* Aim at consuming at most 1/2 of slots in quarantine. */
	WRITE_ONCE(quarantine_batch_size, max((size_t)QUARANTINE_PERCPU_SIZE,
		2 * total_size / QUARANTINE_BATCHES));

	if (likely(quarantine_size > quarantine_max_size)) {
		qlist_move_all(&global_quarantine[quarantine_head], &to_free);
		WRITE_ONCE(quarantine_size, quarantine_size - to_free.bytes);
		quarantine_head++;
		if (quarantine_head == QUARANTINE_BATCHES)
			quarantine_head = 0;
	}

	spin_unlock_irqrestore(&quarantine_lock, flags);

	qlist_free_all(&to_free, NULL);
	srcu_read_unlock(&remove_cache_srcu, srcu_idx);
}

static void qlist_move_cache(struct qlist_head *from,
				   struct qlist_head *to,
				   struct kmem_cache *cache)
{
	struct qlist_node *curr;

	if (unlikely(qlist_empty(from)))
		return;

	curr = from->head;
	qlist_init(from);
	while (curr) {
		struct qlist_node *next = curr->next;
		struct kmem_cache *obj_cache = qlink_to_cache(curr);

		if (obj_cache == cache)
			qlist_put(to, curr, obj_cache->size);
		else
			qlist_put(from, curr, obj_cache->size);

		curr = next;
	}
}

static void per_cpu_remove_cache(void *arg)
{
	struct kmem_cache *cache = arg;
	struct qlist_head to_free = QLIST_INIT;
	struct qlist_head *q;

	q = this_cpu_ptr(&cpu_quarantine);
	qlist_move_cache(q, &to_free, cache);
	qlist_free_all(&to_free, cache);
}

/* Free all quarantined objects belonging to cache. */
void quarantine_remove_cache(struct kmem_cache *cache)
{
	unsigned long flags, i;
	struct qlist_head to_free = QLIST_INIT;

	/*
	 * Must be careful to not miss any objects that are being moved from
	 * per-cpu list to the global quarantine in quarantine_put(),
	 * nor objects being freed in quarantine_reduce(). on_each_cpu()
	 * achieves the first goal, while synchronize_srcu() achieves the
	 * second.
	 */
	on_each_cpu(per_cpu_remove_cache, cache, 1);

	spin_lock_irqsave(&quarantine_lock, flags);
	for (i = 0; i < QUARANTINE_BATCHES; i++) {
		if (qlist_empty(&global_quarantine[i]))
			continue;
		qlist_move_cache(&global_quarantine[i], &to_free, cache);
		/* Scanning whole quarantine can take a while. */
		spin_unlock_irqrestore(&quarantine_lock, flags);
		cond_resched();
		spin_lock_irqsave(&quarantine_lock, flags);
	}
	spin_unlock_irqrestore(&quarantine_lock, flags);

	qlist_free_all(&to_free, cache);

	synchronize_srcu(&remove_cache_srcu);
}
Commit	Line	Data
55834c59 AP	1	/*
	2	* KASAN quarantine.
	3	*
	4	* Author: Alexander Potapenko <glider@google.com>
	5	* Copyright (C) 2016 Google, Inc.
	6	*
	7	* Based on code by Dmitry Chernenkov.
	8	*
	9	* This program is free software; you can redistribute it and/or
	10	* modify it under the terms of the GNU General Public License
	11	* version 2 as published by the Free Software Foundation.
	12	*
	13	* This program is distributed in the hope that it will be useful, but
	14	* WITHOUT ANY WARRANTY; without even the implied warranty of
	15	* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
	16	* General Public License for more details.
	17	*
	18	*/
	19
	20	#include <linux/gfp.h>
	21	#include <linux/hash.h>
	22	#include <linux/kernel.h>
	23	#include <linux/mm.h>
	24	#include <linux/percpu.h>
	25	#include <linux/printk.h>
	26	#include <linux/shrinker.h>
	27	#include <linux/slab.h>
ce5bec54	28	#include <linux/srcu.h>
55834c59 AP	29	#include <linux/string.h>
	30	#include <linux/types.h>
	31
	32	#include "../slab.h"
	33	#include "kasan.h"
	34
	35	/* Data structure and operations for quarantine queues. */
	36
	37	/*
	38	* Each queue is a signle-linked list, which also stores the total size of
	39	* objects inside of it.
	40	*/
	41	struct qlist_head {
	42	struct qlist_node *head;
	43	struct qlist_node *tail;
	44	size_t bytes;
	45	};
	46
	47	#define QLIST_INIT { NULL, NULL, 0 }
	48
	49	static bool qlist_empty(struct qlist_head *q)
	50	{
	51	return !q->head;
	52	}
	53
	54	static void qlist_init(struct qlist_head *q)
	55	{
	56	q->head = q->tail = NULL;
	57	q->bytes = 0;
	58	}
	59
	60	static void qlist_put(struct qlist_head q, struct qlist_node qlink,
	61	size_t size)
	62	{
	63	if (unlikely(qlist_empty(q)))
	64	q->head = qlink;
	65	else
	66	q->tail->next = qlink;
	67	q->tail = qlink;
	68	qlink->next = NULL;
	69	q->bytes += size;
	70	}
	71
	72	static void qlist_move_all(struct qlist_head from, struct qlist_head to)
	73	{
	74	if (unlikely(qlist_empty(from)))
	75	return;
	76
	77	if (qlist_empty(to)) {
	78	to = from;
	79	qlist_init(from);
	80	return;
	81	}
	82
	83	to->tail->next = from->head;
	84	to->tail = from->tail;
	85	to->bytes += from->bytes;
	86
	87	qlist_init(from);
	88	}
	89
64abdcb2 DV	90	#define QUARANTINE_PERCPU_SIZE (1 << 20)
	91	#define QUARANTINE_BATCHES \
	92	(1024 > 4 * CONFIG_NR_CPUS ? 1024 : 4 * CONFIG_NR_CPUS)
55834c59 AP	93
	94	/*
	95	* The object quarantine consists of per-cpu queues and a global queue,
	96	* guarded by quarantine_lock.
	97	*/
	98	static DEFINE_PER_CPU(struct qlist_head, cpu_quarantine);
	99
64abdcb2 DV	100	/* Round-robin FIFO array of batches. */
	101	static struct qlist_head global_quarantine[QUARANTINE_BATCHES];
	102	static int quarantine_head;
	103	static int quarantine_tail;
	104	/* Total size of all objects in global_quarantine across all batches. */
	105	static unsigned long quarantine_size;
55834c59	106	static DEFINE_SPINLOCK(quarantine_lock);
ce5bec54	107	DEFINE_STATIC_SRCU(remove_cache_srcu);
55834c59 AP	108
55834c59 AP	109	/* Maximum size of the global queue. */
64abdcb2 DV	110	static unsigned long quarantine_max_size;
	111
	112	/*
	113	* Target size of a batch in global_quarantine.
	114	* Usually equal to QUARANTINE_PERCPU_SIZE unless we have too much RAM.
	115	*/
	116	static unsigned long quarantine_batch_size;
55834c59 AP	117
	118	/*
	119	* The fraction of physical memory the quarantine is allowed to occupy.
	120	* Quarantine doesn't support memory shrinker with SLAB allocator, so we keep
	121	* the ratio low to avoid OOM.
	122	*/
	123	#define QUARANTINE_FRACTION 32
	124
55834c59 AP	125	static struct kmem_cache qlink_to_cache(struct qlist_node qlink)
	126	{
	127	return virt_to_head_page(qlink)->slab_cache;
	128	}
	129
	130	static void qlink_to_object(struct qlist_node qlink, struct kmem_cache *cache)
	131	{
	132	struct kasan_free_meta *free_info =
	133	container_of(qlink, struct kasan_free_meta,
	134	quarantine_link);
	135
	136	return ((void *)free_info) - cache->kasan_info.free_meta_offset;
	137	}
	138
	139	static void qlink_free(struct qlist_node qlink, struct kmem_cache cache)
	140	{
	141	void *object = qlink_to_object(qlink, cache);
55834c59 AP	142	unsigned long flags;
55834c59 AP	143
f7376aed AR	144	if (IS_ENABLED(CONFIG_SLAB))
	145	local_irq_save(flags);
	146
55834c59	147	___cache_free(cache, object, _THIS_IP_);
f7376aed AR	148
	149	if (IS_ENABLED(CONFIG_SLAB))
	150	local_irq_restore(flags);
55834c59 AP	151	}
	152
	153	static void qlist_free_all(struct qlist_head q, struct kmem_cache cache)
	154	{
	155	struct qlist_node *qlink;
	156
	157	if (unlikely(qlist_empty(q)))
	158	return;
	159
	160	qlink = q->head;
	161	while (qlink) {
	162	struct kmem_cache *obj_cache =
	163	cache ? cache : qlink_to_cache(qlink);
	164	struct qlist_node *next = qlink->next;
	165
	166	qlink_free(qlink, obj_cache);
	167	qlink = next;
	168	}
	169	qlist_init(q);
	170	}
	171
	172	void quarantine_put(struct kasan_free_meta info, struct kmem_cache cache)
	173	{
	174	unsigned long flags;
	175	struct qlist_head *q;
	176	struct qlist_head temp = QLIST_INIT;
	177
ce5bec54 DV	178	/*
	179	* Note: irq must be disabled until after we move the batch to the
	180	* global quarantine. Otherwise quarantine_remove_cache() can miss
	181	* some objects belonging to the cache if they are in our local temp
	182	* list. quarantine_remove_cache() executes on_each_cpu() at the
	183	* beginning which ensures that it either sees the objects in per-cpu
	184	* lists or in the global quarantine.
	185	*/
55834c59 AP	186	local_irq_save(flags);
	187
	188	q = this_cpu_ptr(&cpu_quarantine);
	189	qlist_put(q, &info->quarantine_link, cache->size);
ce5bec54	190	if (unlikely(q->bytes > QUARANTINE_PERCPU_SIZE)) {
55834c59 AP	191	qlist_move_all(q, &temp);
55834c59 AP	192
ce5bec54	193	spin_lock(&quarantine_lock);
64abdcb2 DV	194	WRITE_ONCE(quarantine_size, quarantine_size + temp.bytes);
	195	qlist_move_all(&temp, &global_quarantine[quarantine_tail]);
	196	if (global_quarantine[quarantine_tail].bytes >=
	197	READ_ONCE(quarantine_batch_size)) {
	198	int new_tail;
	199
	200	new_tail = quarantine_tail + 1;
	201	if (new_tail == QUARANTINE_BATCHES)
	202	new_tail = 0;
	203	if (new_tail != quarantine_head)
	204	quarantine_tail = new_tail;
	205	}
ce5bec54	206	spin_unlock(&quarantine_lock);
55834c59	207	}
ce5bec54 DV	208
ce5bec54 DV	209	local_irq_restore(flags);
55834c59 AP	210	}
	211
	212	void quarantine_reduce(void)
	213	{
64abdcb2	214	size_t total_size, new_quarantine_size, percpu_quarantines;
55834c59	215	unsigned long flags;
ce5bec54	216	int srcu_idx;
55834c59	217	struct qlist_head to_free = QLIST_INIT;
55834c59	218
64abdcb2 DV	219	if (likely(READ_ONCE(quarantine_size) <=
64abdcb2 DV	220	READ_ONCE(quarantine_max_size)))
55834c59 AP	221	return;
55834c59 AP	222
ce5bec54 DV	223	/*
	224	* srcu critical section ensures that quarantine_remove_cache()
	225	* will not miss objects belonging to the cache while they are in our
	226	* local to_free list. srcu is chosen because (1) it gives us private
	227	* grace period domain that does not interfere with anything else,
	228	* and (2) it allows synchronize_srcu() to return without waiting
	229	* if there are no pending read critical sections (which is the
	230	* expected case).
	231	*/
	232	srcu_idx = srcu_read_lock(&remove_cache_srcu);
55834c59 AP	233	spin_lock_irqsave(&quarantine_lock, flags);
	234
	235	/*
	236	* Update quarantine size in case of hotplug. Allocate a fraction of
	237	* the installed memory to quarantine minus per-cpu queue limits.
	238	*/
64abdcb2	239	total_size = (READ_ONCE(totalram_pages) << PAGE_SHIFT) /
55834c59	240	QUARANTINE_FRACTION;
c3cee372	241	percpu_quarantines = QUARANTINE_PERCPU_SIZE * num_online_cpus();
64abdcb2 DV	242	new_quarantine_size = (total_size < percpu_quarantines) ?
	243	0 : total_size - percpu_quarantines;
	244	WRITE_ONCE(quarantine_max_size, new_quarantine_size);
	245	/* Aim at consuming at most 1/2 of slots in quarantine. */
	246	WRITE_ONCE(quarantine_batch_size, max((size_t)QUARANTINE_PERCPU_SIZE,
	247	2 * total_size / QUARANTINE_BATCHES));
	248
	249	if (likely(quarantine_size > quarantine_max_size)) {
	250	qlist_move_all(&global_quarantine[quarantine_head], &to_free);
	251	WRITE_ONCE(quarantine_size, quarantine_size - to_free.bytes);
	252	quarantine_head++;
	253	if (quarantine_head == QUARANTINE_BATCHES)
	254	quarantine_head = 0;
55834c59	255	}
55834c59 AP	256
	257	spin_unlock_irqrestore(&quarantine_lock, flags);
	258
	259	qlist_free_all(&to_free, NULL);
ce5bec54	260	srcu_read_unlock(&remove_cache_srcu, srcu_idx);
55834c59 AP	261	}
	262
	263	static void qlist_move_cache(struct qlist_head *from,
	264	struct qlist_head *to,
	265	struct kmem_cache *cache)
	266	{
0ab686d8	267	struct qlist_node *curr;
55834c59 AP	268
	269	if (unlikely(qlist_empty(from)))
	270	return;
	271
	272	curr = from->head;
0ab686d8	273	qlist_init(from);
55834c59	274	while (curr) {
0ab686d8 JK	275	struct qlist_node *next = curr->next;
	276	struct kmem_cache *obj_cache = qlink_to_cache(curr);
	277
	278	if (obj_cache == cache)
	279	qlist_put(to, curr, obj_cache->size);
	280	else
	281	qlist_put(from, curr, obj_cache->size);
	282
	283	curr = next;
55834c59 AP	284	}
	285	}
	286
	287	static void per_cpu_remove_cache(void *arg)
	288	{
	289	struct kmem_cache *cache = arg;
	290	struct qlist_head to_free = QLIST_INIT;
	291	struct qlist_head *q;
	292
	293	q = this_cpu_ptr(&cpu_quarantine);
	294	qlist_move_cache(q, &to_free, cache);
	295	qlist_free_all(&to_free, cache);
	296	}
	297
f9fa1d91	298	/* Free all quarantined objects belonging to cache. */
55834c59 AP	299	void quarantine_remove_cache(struct kmem_cache *cache)
55834c59 AP	300	{
64abdcb2	301	unsigned long flags, i;
55834c59 AP	302	struct qlist_head to_free = QLIST_INIT;
55834c59 AP	303
ce5bec54 DV	304	/*
	305	* Must be careful to not miss any objects that are being moved from
	306	* per-cpu list to the global quarantine in quarantine_put(),
	307	* nor objects being freed in quarantine_reduce(). on_each_cpu()
	308	* achieves the first goal, while synchronize_srcu() achieves the
	309	* second.
	310	*/
55834c59 AP	311	on_each_cpu(per_cpu_remove_cache, cache, 1);
	312
	313	spin_lock_irqsave(&quarantine_lock, flags);
68fd814a DV	314	for (i = 0; i < QUARANTINE_BATCHES; i++) {
	315	if (qlist_empty(&global_quarantine[i]))
	316	continue;
64abdcb2	317	qlist_move_cache(&global_quarantine[i], &to_free, cache);
68fd814a DV	318	/* Scanning whole quarantine can take a while. */
	319	spin_unlock_irqrestore(&quarantine_lock, flags);
	320	cond_resched();
	321	spin_lock_irqsave(&quarantine_lock, flags);
	322	}
55834c59 AP	323	spin_unlock_irqrestore(&quarantine_lock, flags);
	324
	325	qlist_free_all(&to_free, cache);
ce5bec54 DV	326
ce5bec54 DV	327	synchronize_srcu(&remove_cache_srcu);
55834c59	328	}