[deliverable/linux.git] / lib / stackdepot.c

/*
 * Generic stack depot for storing stack traces.
 *
 * Some debugging tools need to save stack traces of certain events which can
 * be later presented to the user. For example, KASAN needs to safe alloc and
 * free stacks for each object, but storing two stack traces per object
 * requires too much memory (e.g. SLUB_DEBUG needs 256 bytes per object for
 * that).
 *
 * Instead, stack depot maintains a hashtable of unique stacktraces. Since alloc
 * and free stacks repeat a lot, we save about 100x space.
 * Stacks are never removed from depot, so we store them contiguously one after
 * another in a contiguos memory allocation.
 *
 * 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/jhash.h>
#include <linux/kernel.h>
#include <linux/mm.h>
#include <linux/percpu.h>
#include <linux/printk.h>
#include <linux/slab.h>
#include <linux/stacktrace.h>
#include <linux/stackdepot.h>
#include <linux/string.h>
#include <linux/types.h>

#define DEPOT_STACK_BITS (sizeof(depot_stack_handle_t) * 8)

#define STACK_ALLOC_NULL_PROTECTION_BITS 1
#define STACK_ALLOC_ORDER 2 /* 'Slab' size order for stack depot, 4 pages */
#define STACK_ALLOC_SIZE (1LL << (PAGE_SHIFT + STACK_ALLOC_ORDER))
#define STACK_ALLOC_ALIGN 4
#define STACK_ALLOC_OFFSET_BITS (STACK_ALLOC_ORDER + PAGE_SHIFT - \
					STACK_ALLOC_ALIGN)
#define STACK_ALLOC_INDEX_BITS (DEPOT_STACK_BITS - \
		STACK_ALLOC_NULL_PROTECTION_BITS - STACK_ALLOC_OFFSET_BITS)
#define STACK_ALLOC_SLABS_CAP 1024
#define STACK_ALLOC_MAX_SLABS \
	(((1LL << (STACK_ALLOC_INDEX_BITS)) < STACK_ALLOC_SLABS_CAP) ? \
	 (1LL << (STACK_ALLOC_INDEX_BITS)) : STACK_ALLOC_SLABS_CAP)

/* The compact structure to store the reference to stacks. */
union handle_parts {
	depot_stack_handle_t handle;
	struct {
		u32 slabindex : STACK_ALLOC_INDEX_BITS;
		u32 offset : STACK_ALLOC_OFFSET_BITS;
		u32 valid : STACK_ALLOC_NULL_PROTECTION_BITS;
	};
};

struct stack_record {
	struct stack_record *next;	/* Link in the hashtable */
	u32 hash;			/* Hash in the hastable */
	u32 size;			/* Number of frames in the stack */
	union handle_parts handle;
	unsigned long entries[1];	/* Variable-sized array of entries. */
};

static void *stack_slabs[STACK_ALLOC_MAX_SLABS];

static int depot_index;
static int next_slab_inited;
static size_t depot_offset;
static DEFINE_SPINLOCK(depot_lock);

static bool init_stack_slab(void **prealloc)
{
	if (!*prealloc)
		return false;
	/*
	 * This smp_load_acquire() pairs with smp_store_release() to
	 * |next_slab_inited| below and in depot_alloc_stack().
	 */
	if (smp_load_acquire(&next_slab_inited))
		return true;
	if (stack_slabs[depot_index] == NULL) {
		stack_slabs[depot_index] = *prealloc;
	} else {
		stack_slabs[depot_index + 1] = *prealloc;
		/*
		 * This smp_store_release pairs with smp_load_acquire() from
		 * |next_slab_inited| above and in depot_save_stack().
		 */
		smp_store_release(&next_slab_inited, 1);
	}
	*prealloc = NULL;
	return true;
}

/* Allocation of a new stack in raw storage */
static struct stack_record *depot_alloc_stack(unsigned long *entries, int size,
		u32 hash, void **prealloc, gfp_t alloc_flags)
{
	int required_size = offsetof(struct stack_record, entries) +
		sizeof(unsigned long) * size;
	struct stack_record *stack;

	required_size = ALIGN(required_size, 1 << STACK_ALLOC_ALIGN);

	if (unlikely(depot_offset + required_size > STACK_ALLOC_SIZE)) {
		if (unlikely(depot_index + 1 >= STACK_ALLOC_MAX_SLABS)) {
			WARN_ONCE(1, "Stack depot reached limit capacity");
			return NULL;
		}
		depot_index++;
		depot_offset = 0;
		/*
		 * smp_store_release() here pairs with smp_load_acquire() from
		 * |next_slab_inited| in depot_save_stack() and
		 * init_stack_slab().
		 */
		if (depot_index + 1 < STACK_ALLOC_MAX_SLABS)
			smp_store_release(&next_slab_inited, 0);
	}
	init_stack_slab(prealloc);
	if (stack_slabs[depot_index] == NULL)
		return NULL;

	stack = stack_slabs[depot_index] + depot_offset;

	stack->hash = hash;
	stack->size = size;
	stack->handle.slabindex = depot_index;
	stack->handle.offset = depot_offset >> STACK_ALLOC_ALIGN;
	stack->handle.valid = 1;
	memcpy(stack->entries, entries, size * sizeof(unsigned long));
	depot_offset += required_size;

	return stack;
}

#define STACK_HASH_ORDER 20
#define STACK_HASH_SIZE (1L << STACK_HASH_ORDER)
#define STACK_HASH_MASK (STACK_HASH_SIZE - 1)
#define STACK_HASH_SEED 0x9747b28c

static struct stack_record *stack_table[STACK_HASH_SIZE] = {
	[0 ...	STACK_HASH_SIZE - 1] = NULL
};

/* Calculate hash for a stack */
static inline u32 hash_stack(unsigned long *entries, unsigned int size)
{
	return jhash2((u32 *)entries,
			       size * sizeof(unsigned long) / sizeof(u32),
			       STACK_HASH_SEED);
}

/* Find a stack that is equal to the one stored in entries in the hash */
static inline struct stack_record *find_stack(struct stack_record *bucket,
					     unsigned long *entries, int size,
					     u32 hash)
{
	struct stack_record *found;

	for (found = bucket; found; found = found->next) {
		if (found->hash == hash &&
		    found->size == size &&
		    !memcmp(entries, found->entries,
			    size * sizeof(unsigned long))) {
			return found;
		}
	}
	return NULL;
}

void depot_fetch_stack(depot_stack_handle_t handle, struct stack_trace *trace)
{
	union handle_parts parts = { .handle = handle };
	void *slab = stack_slabs[parts.slabindex];
	size_t offset = parts.offset << STACK_ALLOC_ALIGN;
	struct stack_record *stack = slab + offset;

	trace->nr_entries = trace->max_entries = stack->size;
	trace->entries = stack->entries;
	trace->skip = 0;
}

/**
 * depot_save_stack - save stack in a stack depot.
 * @trace - the stacktrace to save.
 * @alloc_flags - flags for allocating additional memory if required.
 *
 * Returns the handle of the stack struct stored in depot.
 */
depot_stack_handle_t depot_save_stack(struct stack_trace *trace,
				    gfp_t alloc_flags)
{
	u32 hash;
	depot_stack_handle_t retval = 0;
	struct stack_record *found = NULL, **bucket;
	unsigned long flags;
	struct page *page = NULL;
	void *prealloc = NULL;

	if (unlikely(trace->nr_entries == 0))
		goto fast_exit;

	hash = hash_stack(trace->entries, trace->nr_entries);
	bucket = &stack_table[hash & STACK_HASH_MASK];

	/*
	 * Fast path: look the stack trace up without locking.
	 * The smp_load_acquire() here pairs with smp_store_release() to
	 * |bucket| below.
	 */
	found = find_stack(smp_load_acquire(bucket), trace->entries,
			   trace->nr_entries, hash);
	if (found)
		goto exit;

	/*
	 * Check if the current or the next stack slab need to be initialized.
	 * If so, allocate the memory - we won't be able to do that under the
	 * lock.
	 *
	 * The smp_load_acquire() here pairs with smp_store_release() to
	 * |next_slab_inited| in depot_alloc_stack() and init_stack_slab().
	 */
	if (unlikely(!smp_load_acquire(&next_slab_inited))) {
		/*
		 * Zero out zone modifiers, as we don't have specific zone
		 * requirements. Keep the flags related to allocation in atomic
		 * contexts and I/O.
		 */
		alloc_flags &= ~GFP_ZONEMASK;
		alloc_flags &= (GFP_ATOMIC | GFP_KERNEL);
		alloc_flags |= __GFP_NOWARN;
		page = alloc_pages(alloc_flags, STACK_ALLOC_ORDER);
		if (page)
			prealloc = page_address(page);
	}

	spin_lock_irqsave(&depot_lock, flags);

	found = find_stack(*bucket, trace->entries, trace->nr_entries, hash);
	if (!found) {
		struct stack_record *new =
			depot_alloc_stack(trace->entries, trace->nr_entries,
					  hash, &prealloc, alloc_flags);
		if (new) {
			new->next = *bucket;
			/*
			 * This smp_store_release() pairs with
			 * smp_load_acquire() from |bucket| above.
			 */
			smp_store_release(bucket, new);
			found = new;
		}
	} else if (prealloc) {
		/*
		 * We didn't need to store this stack trace, but let's keep
		 * the preallocated memory for the future.
		 */
		WARN_ON(!init_stack_slab(&prealloc));
	}

	spin_unlock_irqrestore(&depot_lock, flags);
exit:
	if (prealloc) {
		/* Nobody used this memory, ok to free it. */
		free_pages((unsigned long)prealloc, STACK_ALLOC_ORDER);
	}
	if (found)
		retval = found->handle.handle;
fast_exit:
	return retval;
}
Commit	Line	Data
cd11016e AP	1	/*
	2	* Generic stack depot for storing stack traces.
	3	*
	4	* Some debugging tools need to save stack traces of certain events which can
	5	* be later presented to the user. For example, KASAN needs to safe alloc and
	6	* free stacks for each object, but storing two stack traces per object
	7	* requires too much memory (e.g. SLUB_DEBUG needs 256 bytes per object for
	8	* that).
	9	*
	10	* Instead, stack depot maintains a hashtable of unique stacktraces. Since alloc
	11	* and free stacks repeat a lot, we save about 100x space.
	12	* Stacks are never removed from depot, so we store them contiguously one after
	13	* another in a contiguos memory allocation.
	14	*
	15	* Author: Alexander Potapenko <glider@google.com>
	16	* Copyright (C) 2016 Google, Inc.
	17	*
	18	* Based on code by Dmitry Chernenkov.
	19	*
	20	* This program is free software; you can redistribute it and/or
	21	* modify it under the terms of the GNU General Public License
	22	* version 2 as published by the Free Software Foundation.
	23	*
	24	* This program is distributed in the hope that it will be useful, but
	25	* WITHOUT ANY WARRANTY; without even the implied warranty of
	26	* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
	27	* General Public License for more details.
	28	*
	29	*/
	30
	31	#include <linux/gfp.h>
	32	#include <linux/jhash.h>
	33	#include <linux/kernel.h>
	34	#include <linux/mm.h>
	35	#include <linux/percpu.h>
	36	#include <linux/printk.h>
	37	#include <linux/slab.h>
	38	#include <linux/stacktrace.h>
	39	#include <linux/stackdepot.h>
	40	#include <linux/string.h>
	41	#include <linux/types.h>
	42
	43	#define DEPOT_STACK_BITS (sizeof(depot_stack_handle_t) * 8)
	44
7c31190b	45	#define STACK_ALLOC_NULL_PROTECTION_BITS 1
cd11016e AP	46	#define STACK_ALLOC_ORDER 2 /* 'Slab' size order for stack depot, 4 pages */
	47	#define STACK_ALLOC_SIZE (1LL << (PAGE_SHIFT + STACK_ALLOC_ORDER))
	48	#define STACK_ALLOC_ALIGN 4
	49	#define STACK_ALLOC_OFFSET_BITS (STACK_ALLOC_ORDER + PAGE_SHIFT - \
	50	STACK_ALLOC_ALIGN)
7c31190b JK	51	#define STACK_ALLOC_INDEX_BITS (DEPOT_STACK_BITS - \
7c31190b JK	52	STACK_ALLOC_NULL_PROTECTION_BITS - STACK_ALLOC_OFFSET_BITS)
cd11016e AP	53	#define STACK_ALLOC_SLABS_CAP 1024
	54	#define STACK_ALLOC_MAX_SLABS \
	55	(((1LL << (STACK_ALLOC_INDEX_BITS)) < STACK_ALLOC_SLABS_CAP) ? \
	56	(1LL << (STACK_ALLOC_INDEX_BITS)) : STACK_ALLOC_SLABS_CAP)
	57
	58	/* The compact structure to store the reference to stacks. */
	59	union handle_parts {
	60	depot_stack_handle_t handle;
	61	struct {
	62	u32 slabindex : STACK_ALLOC_INDEX_BITS;
	63	u32 offset : STACK_ALLOC_OFFSET_BITS;
7c31190b	64	u32 valid : STACK_ALLOC_NULL_PROTECTION_BITS;
cd11016e AP	65	};
	66	};
	67
	68	struct stack_record {
	69	struct stack_record next; / Link in the hashtable */
	70	u32 hash; /* Hash in the hastable */
	71	u32 size; /* Number of frames in the stack */
	72	union handle_parts handle;
	73	unsigned long entries[1]; /* Variable-sized array of entries. */
	74	};
	75
	76	static void *stack_slabs[STACK_ALLOC_MAX_SLABS];
	77
	78	static int depot_index;
	79	static int next_slab_inited;
	80	static size_t depot_offset;
	81	static DEFINE_SPINLOCK(depot_lock);
	82
	83	static bool init_stack_slab(void **prealloc)
	84	{
	85	if (!*prealloc)
	86	return false;
	87	/*
	88	* This smp_load_acquire() pairs with smp_store_release() to
	89	* \|next_slab_inited\| below and in depot_alloc_stack().
	90	*/
	91	if (smp_load_acquire(&next_slab_inited))
	92	return true;
	93	if (stack_slabs[depot_index] == NULL) {
	94	stack_slabs[depot_index] = *prealloc;
	95	} else {
	96	stack_slabs[depot_index + 1] = *prealloc;
	97	/*
	98	* This smp_store_release pairs with smp_load_acquire() from
	99	* \|next_slab_inited\| above and in depot_save_stack().
	100	*/
	101	smp_store_release(&next_slab_inited, 1);
	102	}
	103	*prealloc = NULL;
	104	return true;
	105	}
	106
	107	/* Allocation of a new stack in raw storage */
	108	static struct stack_record depot_alloc_stack(unsigned long entries, int size,
	109	u32 hash, void **prealloc, gfp_t alloc_flags)
	110	{
	111	int required_size = offsetof(struct stack_record, entries) +
	112	sizeof(unsigned long) * size;
	113	struct stack_record *stack;
	114
	115	required_size = ALIGN(required_size, 1 << STACK_ALLOC_ALIGN);
	116
	117	if (unlikely(depot_offset + required_size > STACK_ALLOC_SIZE)) {
	118	if (unlikely(depot_index + 1 >= STACK_ALLOC_MAX_SLABS)) {
	119	WARN_ONCE(1, "Stack depot reached limit capacity");
	120	return NULL;
	121	}
	122	depot_index++;
	123	depot_offset = 0;
	124	/*
	125	* smp_store_release() here pairs with smp_load_acquire() from
	126	* \|next_slab_inited\| in depot_save_stack() and
	127	* init_stack_slab().
	128	*/
129	if (depot_index + 1 < STACK_ALLOC_MAX_SLABS)
130	smp_store_release(&next_slab_inited, 0);
131	}
132	init_stack_slab(prealloc);
133	if (stack_slabs[depot_index] == NULL)
134	return NULL;
135
136	stack = stack_slabs[depot_index] + depot_offset;
137
138	stack->hash = hash;
139	stack->size = size;
140	stack->handle.slabindex = depot_index;
141	stack->handle.offset = depot_offset >> STACK_ALLOC_ALIGN;
7c31190b	142	stack->handle.valid = 1;
cd11016e AP	143	memcpy(stack->entries, entries, size * sizeof(unsigned long));
	144	depot_offset += required_size;
	145
	146	return stack;
	147	}
	148
	149	#define STACK_HASH_ORDER 20
	150	#define STACK_HASH_SIZE (1L << STACK_HASH_ORDER)
	151	#define STACK_HASH_MASK (STACK_HASH_SIZE - 1)
	152	#define STACK_HASH_SEED 0x9747b28c
	153
	154	static struct stack_record *stack_table[STACK_HASH_SIZE] = {
	155	[0 ... STACK_HASH_SIZE - 1] = NULL
	156	};
	157
	158	/* Calculate hash for a stack */
	159	static inline u32 hash_stack(unsigned long *entries, unsigned int size)
	160	{
	161	return jhash2((u32 *)entries,
	162	size * sizeof(unsigned long) / sizeof(u32),
	163	STACK_HASH_SEED);
	164	}
	165
	166	/* Find a stack that is equal to the one stored in entries in the hash */
	167	static inline struct stack_record find_stack(struct stack_record bucket,
	168	unsigned long *entries, int size,
	169	u32 hash)
	170	{
	171	struct stack_record *found;
	172
	173	for (found = bucket; found; found = found->next) {
	174	if (found->hash == hash &&
	175	found->size == size &&
	176	!memcmp(entries, found->entries,
	177	size * sizeof(unsigned long))) {
	178	return found;
	179	}
	180	}
	181	return NULL;
	182	}
	183
	184	void depot_fetch_stack(depot_stack_handle_t handle, struct stack_trace *trace)
	185	{
	186	union handle_parts parts = { .handle = handle };
	187	void *slab = stack_slabs[parts.slabindex];
	188	size_t offset = parts.offset << STACK_ALLOC_ALIGN;
	189	struct stack_record *stack = slab + offset;
	190
	191	trace->nr_entries = trace->max_entries = stack->size;
	192	trace->entries = stack->entries;
	193	trace->skip = 0;
	194	}
	195
	196	/**
	197	* depot_save_stack - save stack in a stack depot.
	198	* @trace - the stacktrace to save.
	199	* @alloc_flags - flags for allocating additional memory if required.
	200	*
	201	* Returns the handle of the stack struct stored in depot.
	202	*/
	203	depot_stack_handle_t depot_save_stack(struct stack_trace *trace,
	204	gfp_t alloc_flags)
	205	{
	206	u32 hash;
207	depot_stack_handle_t retval = 0;
208	struct stack_record found = NULL, *bucket;
209	unsigned long flags;
210	struct page *page = NULL;
211	void *prealloc = NULL;
212
213	if (unlikely(trace->nr_entries == 0))
214	goto fast_exit;
215
216	hash = hash_stack(trace->entries, trace->nr_entries);
cd11016e AP	217	bucket = &stack_table[hash & STACK_HASH_MASK];
	218
	219	/*
	220	* Fast path: look the stack trace up without locking.
	221	* The smp_load_acquire() here pairs with smp_store_release() to
	222	* \|bucket\| below.
	223	*/
	224	found = find_stack(smp_load_acquire(bucket), trace->entries,
	225	trace->nr_entries, hash);
	226	if (found)
	227	goto exit;
	228
	229	/*
	230	* Check if the current or the next stack slab need to be initialized.
	231	* If so, allocate the memory - we won't be able to do that under the
	232	* lock.
	233	*
	234	* The smp_load_acquire() here pairs with smp_store_release() to
	235	* \|next_slab_inited\| in depot_alloc_stack() and init_stack_slab().
	236	*/
	237	if (unlikely(!smp_load_acquire(&next_slab_inited))) {
	238	/*
	239	* Zero out zone modifiers, as we don't have specific zone
	240	* requirements. Keep the flags related to allocation in atomic
	241	* contexts and I/O.
	242	*/
	243	alloc_flags &= ~GFP_ZONEMASK;
	244	alloc_flags &= (GFP_ATOMIC \| GFP_KERNEL);
87cc271d	245	alloc_flags \|= __GFP_NOWARN;
cd11016e AP	246	page = alloc_pages(alloc_flags, STACK_ALLOC_ORDER);
	247	if (page)
	248	prealloc = page_address(page);
	249	}
	250
	251	spin_lock_irqsave(&depot_lock, flags);
	252
	253	found = find_stack(*bucket, trace->entries, trace->nr_entries, hash);
	254	if (!found) {
	255	struct stack_record *new =
	256	depot_alloc_stack(trace->entries, trace->nr_entries,
	257	hash, &prealloc, alloc_flags);
	258	if (new) {
	259	new->next = *bucket;
	260	/*
	261	* This smp_store_release() pairs with
	262	* smp_load_acquire() from \|bucket\| above.
	263	*/
	264	smp_store_release(bucket, new);
	265	found = new;
	266	}
	267	} else if (prealloc) {
	268	/*
	269	* We didn't need to store this stack trace, but let's keep
	270	* the preallocated memory for the future.
	271	*/
	272	WARN_ON(!init_stack_slab(&prealloc));
	273	}
	274
	275	spin_unlock_irqrestore(&depot_lock, flags);
	276	exit:
	277	if (prealloc) {
	278	/* Nobody used this memory, ok to free it. */
	279	free_pages((unsigned long)prealloc, STACK_ALLOC_ORDER);
	280	}
	281	if (found)
	282	retval = found->handle.handle;
	283	fast_exit:
	284	return retval;
	285	}