[linux-block.git] / lib / stackdepot.c

// SPDX-License-Identifier: GPL-2.0-only
/*
 * 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 contiguous memory allocation.
 *
 * Author: Alexander Potapenko <glider@google.com>
 * Copyright (C) 2016 Google, Inc.
 *
 * Based on code by Dmitry Chernenkov.
 */

#include <linux/gfp.h>
#include <linux/jhash.h>
#include <linux/kernel.h>
#include <linux/mm.h>
#include <linux/mutex.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>
#include <linux/memblock.h>
#include <linux/kasan-enabled.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 - STACK_DEPOT_EXTRA_BITS)
#define STACK_ALLOC_SLABS_CAP 8192
#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;
		u32 extra : STACK_DEPOT_EXTRA_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[];	/* Variable-sized array of entries. */
};

static bool __stack_depot_want_early_init __initdata = IS_ENABLED(CONFIG_STACKDEPOT_ALWAYS_INIT);
static bool __stack_depot_early_init_passed __initdata;

static void *stack_slabs[STACK_ALLOC_MAX_SLABS];

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

unsigned int stack_depot_get_extra_bits(depot_stack_handle_t handle)
{
	union handle_parts parts = { .handle = handle };

	return parts.extra;
}
EXPORT_SYMBOL(stack_depot_get_extra_bits);

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;
		*prealloc = NULL;
	} else {
		/* If this is the last depot slab, do not touch the next one. */
		if (depot_index + 1 < STACK_ALLOC_MAX_SLABS) {
			stack_slabs[depot_index + 1] = *prealloc;
			*prealloc = NULL;
		}
		/*
		 * This smp_store_release pairs with smp_load_acquire() from
		 * |next_slab_inited| above and in stack_depot_save().
		 */
		smp_store_release(&next_slab_inited, 1);
	}
	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)
{
	struct stack_record *stack;
	size_t required_size = struct_size(stack, entries, size);

	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 stack_depot_save() 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;
	stack->handle.extra = 0;
	memcpy(stack->entries, entries, flex_array_size(stack, entries, size));
	depot_offset += required_size;

	return stack;
}

/* one hash table bucket entry per 16kB of memory */
#define STACK_HASH_SCALE	14
/* limited between 4k and 1M buckets */
#define STACK_HASH_ORDER_MIN	12
#define STACK_HASH_ORDER_MAX	20
#define STACK_HASH_SEED 0x9747b28c

static unsigned int stack_hash_order;
static unsigned int stack_hash_mask;

static bool stack_depot_disable;
static struct stack_record **stack_table;

static int __init is_stack_depot_disabled(char *str)
{
	int ret;

	ret = kstrtobool(str, &stack_depot_disable);
	if (!ret && stack_depot_disable) {
		pr_info("Stack Depot is disabled\n");
		stack_table = NULL;
	}
	return 0;
}
early_param("stack_depot_disable", is_stack_depot_disabled);

void __init stack_depot_want_early_init(void)
{
	/* Too late to request early init now */
	WARN_ON(__stack_depot_early_init_passed);

	__stack_depot_want_early_init = true;
}

int __init stack_depot_early_init(void)
{
	unsigned long entries = 0;

	/* This is supposed to be called only once, from mm_init() */
	if (WARN_ON(__stack_depot_early_init_passed))
		return 0;

	__stack_depot_early_init_passed = true;

	if (kasan_enabled() && !stack_hash_order)
		stack_hash_order = STACK_HASH_ORDER_MAX;

	if (!__stack_depot_want_early_init || stack_depot_disable)
		return 0;

	if (stack_hash_order)
		entries = 1UL <<  stack_hash_order;
	stack_table = alloc_large_system_hash("stackdepot",
						sizeof(struct stack_record *),
						entries,
						STACK_HASH_SCALE,
						HASH_EARLY | HASH_ZERO,
						NULL,
						&stack_hash_mask,
						1UL << STACK_HASH_ORDER_MIN,
						1UL << STACK_HASH_ORDER_MAX);

	if (!stack_table) {
		pr_err("Stack Depot hash table allocation failed, disabling\n");
		stack_depot_disable = true;
		return -ENOMEM;
	}

	return 0;
}

int stack_depot_init(void)
{
	static DEFINE_MUTEX(stack_depot_init_mutex);
	int ret = 0;

	mutex_lock(&stack_depot_init_mutex);
	if (!stack_depot_disable && !stack_table) {
		unsigned long entries;
		int scale = STACK_HASH_SCALE;

		if (stack_hash_order) {
			entries = 1UL << stack_hash_order;
		} else {
			entries = nr_free_buffer_pages();
			entries = roundup_pow_of_two(entries);

			if (scale > PAGE_SHIFT)
				entries >>= (scale - PAGE_SHIFT);
			else
				entries <<= (PAGE_SHIFT - scale);
		}

		if (entries < 1UL << STACK_HASH_ORDER_MIN)
			entries = 1UL << STACK_HASH_ORDER_MIN;
		if (entries > 1UL << STACK_HASH_ORDER_MAX)
			entries = 1UL << STACK_HASH_ORDER_MAX;

		pr_info("Stack Depot allocating hash table of %lu entries with kvcalloc\n",
				entries);
		stack_table = kvcalloc(entries, sizeof(struct stack_record *), GFP_KERNEL);
		if (!stack_table) {
			pr_err("Stack Depot hash table allocation failed, disabling\n");
			stack_depot_disable = true;
			ret = -ENOMEM;
		}
		stack_hash_mask = entries - 1;
	}
	mutex_unlock(&stack_depot_init_mutex);
	return ret;
}
EXPORT_SYMBOL_GPL(stack_depot_init);

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

/* Use our own, non-instrumented version of memcmp().
 *
 * We actually don't care about the order, just the equality.
 */
static inline
int stackdepot_memcmp(const unsigned long *u1, const unsigned long *u2,
			unsigned int n)
{
	for ( ; n-- ; u1++, u2++) {
		if (*u1 != *u2)
			return 1;
	}
	return 0;
}

/* 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 &&
		    !stackdepot_memcmp(entries, found->entries, size))
			return found;
	}
	return NULL;
}

/**
 * stack_depot_snprint - print stack entries from a depot into a buffer
 *
 * @handle:	Stack depot handle which was returned from
 *		stack_depot_save().
 * @buf:	Pointer to the print buffer
 *
 * @size:	Size of the print buffer
 *
 * @spaces:	Number of leading spaces to print
 *
 * Return:	Number of bytes printed.
 */
int stack_depot_snprint(depot_stack_handle_t handle, char *buf, size_t size,
		       int spaces)
{
	unsigned long *entries;
	unsigned int nr_entries;

	nr_entries = stack_depot_fetch(handle, &entries);
	return nr_entries ? stack_trace_snprint(buf, size, entries, nr_entries,
						spaces) : 0;
}
EXPORT_SYMBOL_GPL(stack_depot_snprint);

/**
 * stack_depot_print - print stack entries from a depot
 *
 * @stack:		Stack depot handle which was returned from
 *			stack_depot_save().
 *
 */
void stack_depot_print(depot_stack_handle_t stack)
{
	unsigned long *entries;
	unsigned int nr_entries;

	nr_entries = stack_depot_fetch(stack, &entries);
	if (nr_entries > 0)
		stack_trace_print(entries, nr_entries, 0);
}
EXPORT_SYMBOL_GPL(stack_depot_print);

/**
 * stack_depot_fetch - Fetch stack entries from a depot
 *
 * @handle:		Stack depot handle which was returned from
 *			stack_depot_save().
 * @entries:		Pointer to store the entries address
 *
 * Return: The number of trace entries for this depot.
 */
unsigned int stack_depot_fetch(depot_stack_handle_t handle,
			       unsigned long **entries)
{
	union handle_parts parts = { .handle = handle };
	void *slab;
	size_t offset = parts.offset << STACK_ALLOC_ALIGN;
	struct stack_record *stack;

	*entries = NULL;
	if (!handle)
		return 0;

	if (parts.slabindex > depot_index) {
		WARN(1, "slab index %d out of bounds (%d) for stack id %08x\n",
			parts.slabindex, depot_index, handle);
		return 0;
	}
	slab = stack_slabs[parts.slabindex];
	if (!slab)
		return 0;
	stack = slab + offset;

	*entries = stack->entries;
	return stack->size;
}
EXPORT_SYMBOL_GPL(stack_depot_fetch);

/**
 * __stack_depot_save - Save a stack trace from an array
 *
 * @entries:		Pointer to storage array
 * @nr_entries:		Size of the storage array
 * @extra_bits:		Flags to store in unused bits of depot_stack_handle_t
 * @alloc_flags:	Allocation gfp flags
 * @can_alloc:		Allocate stack slabs (increased chance of failure if false)
 *
 * Saves a stack trace from @entries array of size @nr_entries. If @can_alloc is
 * %true, is allowed to replenish the stack slab pool in case no space is left
 * (allocates using GFP flags of @alloc_flags). If @can_alloc is %false, avoids
 * any allocations and will fail if no space is left to store the stack trace.
 *
 * If the stack trace in @entries is from an interrupt, only the portion up to
 * interrupt entry is saved.
 *
 * Additional opaque flags can be passed in @extra_bits, stored in the unused
 * bits of the stack handle, and retrieved using stack_depot_get_extra_bits()
 * without calling stack_depot_fetch().
 *
 * Context: Any context, but setting @can_alloc to %false is required if
 *          alloc_pages() cannot be used from the current context. Currently
 *          this is the case from contexts where neither %GFP_ATOMIC nor
 *          %GFP_NOWAIT can be used (NMI, raw_spin_lock).
 *
 * Return: The handle of the stack struct stored in depot, 0 on failure.
 */
depot_stack_handle_t __stack_depot_save(unsigned long *entries,
					unsigned int nr_entries,
					unsigned int extra_bits,
					gfp_t alloc_flags, bool can_alloc)
{
	struct stack_record *found = NULL, **bucket;
	union handle_parts retval = { .handle = 0 };
	struct page *page = NULL;
	void *prealloc = NULL;
	unsigned long flags;
	u32 hash;

	/*
	 * If this stack trace is from an interrupt, including anything before
	 * interrupt entry usually leads to unbounded stackdepot growth.
	 *
	 * Because use of filter_irq_stacks() is a requirement to ensure
	 * stackdepot can efficiently deduplicate interrupt stacks, always
	 * filter_irq_stacks() to simplify all callers' use of stackdepot.
	 */
	nr_entries = filter_irq_stacks(entries, nr_entries);

	if (unlikely(nr_entries == 0) || stack_depot_disable)
		goto fast_exit;

	hash = hash_stack(entries, 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), entries,
			   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(can_alloc && !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);
	}

	raw_spin_lock_irqsave(&depot_lock, flags);

	found = find_stack(*bucket, entries, nr_entries, hash);
	if (!found) {
		struct stack_record *new = depot_alloc_stack(entries, nr_entries, hash, &prealloc);

		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));
	}

	raw_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.handle = found->handle.handle;
fast_exit:
	retval.extra = extra_bits;

	return retval.handle;
}
EXPORT_SYMBOL_GPL(__stack_depot_save);

/**
 * stack_depot_save - Save a stack trace from an array
 *
 * @entries:		Pointer to storage array
 * @nr_entries:		Size of the storage array
 * @alloc_flags:	Allocation gfp flags
 *
 * Context: Contexts where allocations via alloc_pages() are allowed.
 *          See __stack_depot_save() for more details.
 *
 * Return: The handle of the stack struct stored in depot, 0 on failure.
 */
depot_stack_handle_t stack_depot_save(unsigned long *entries,
				      unsigned int nr_entries,
				      gfp_t alloc_flags)
{
	return __stack_depot_save(entries, nr_entries, 0, alloc_flags, true);
}
EXPORT_SYMBOL_GPL(stack_depot_save);
Commit	Line	Data
1802d0be	1	// SPDX-License-Identifier: GPL-2.0-only
cd11016e AP	2	/*
	3	* Generic stack depot for storing stack traces.
	4	*
	5	* Some debugging tools need to save stack traces of certain events which can
	6	* be later presented to the user. For example, KASAN needs to safe alloc and
	7	* free stacks for each object, but storing two stack traces per object
	8	* requires too much memory (e.g. SLUB_DEBUG needs 256 bytes per object for
	9	* that).
	10	*
	11	* Instead, stack depot maintains a hashtable of unique stacktraces. Since alloc
	12	* and free stacks repeat a lot, we save about 100x space.
	13	* Stacks are never removed from depot, so we store them contiguously one after
9dbbc3b9	14	* another in a contiguous memory allocation.
cd11016e AP	15	*
	16	* Author: Alexander Potapenko <glider@google.com>
	17	* Copyright (C) 2016 Google, Inc.
	18	*
	19	* Based on code by Dmitry Chernenkov.
cd11016e AP	20	*/
	21
	22	#include <linux/gfp.h>
	23	#include <linux/jhash.h>
	24	#include <linux/kernel.h>
	25	#include <linux/mm.h>
2dba5eb1	26	#include <linux/mutex.h>
cd11016e AP	27	#include <linux/percpu.h>
	28	#include <linux/printk.h>
	29	#include <linux/slab.h>
	30	#include <linux/stacktrace.h>
	31	#include <linux/stackdepot.h>
	32	#include <linux/string.h>
	33	#include <linux/types.h>
e1fdc403	34	#include <linux/memblock.h>
f9987921	35	#include <linux/kasan-enabled.h>
cd11016e AP	36
	37	#define DEPOT_STACK_BITS (sizeof(depot_stack_handle_t) * 8)
	38
7c31190b	39	#define STACK_ALLOC_NULL_PROTECTION_BITS 1
cd11016e AP	40	#define STACK_ALLOC_ORDER 2 /* 'Slab' size order for stack depot, 4 pages */
	41	#define STACK_ALLOC_SIZE (1LL << (PAGE_SHIFT + STACK_ALLOC_ORDER))
	42	#define STACK_ALLOC_ALIGN 4
	43	#define STACK_ALLOC_OFFSET_BITS (STACK_ALLOC_ORDER + PAGE_SHIFT - \
	44	STACK_ALLOC_ALIGN)
7c31190b	45	#define STACK_ALLOC_INDEX_BITS (DEPOT_STACK_BITS - \
83a4f1ef AP	46	STACK_ALLOC_NULL_PROTECTION_BITS - \
83a4f1ef AP	47	STACK_ALLOC_OFFSET_BITS - STACK_DEPOT_EXTRA_BITS)
02754e0a	48	#define STACK_ALLOC_SLABS_CAP 8192
cd11016e AP	49	#define STACK_ALLOC_MAX_SLABS \
	50	(((1LL << (STACK_ALLOC_INDEX_BITS)) < STACK_ALLOC_SLABS_CAP) ? \
	51	(1LL << (STACK_ALLOC_INDEX_BITS)) : STACK_ALLOC_SLABS_CAP)
	52
	53	/* The compact structure to store the reference to stacks. */
	54	union handle_parts {
	55	depot_stack_handle_t handle;
	56	struct {
	57	u32 slabindex : STACK_ALLOC_INDEX_BITS;
	58	u32 offset : STACK_ALLOC_OFFSET_BITS;
7c31190b	59	u32 valid : STACK_ALLOC_NULL_PROTECTION_BITS;
83a4f1ef	60	u32 extra : STACK_DEPOT_EXTRA_BITS;
cd11016e AP	61	};
	62	};
	63
	64	struct stack_record {
	65	struct stack_record next; / Link in the hashtable */
	66	u32 hash; /* Hash in the hastable */
	67	u32 size; /* Number of frames in the stack */
	68	union handle_parts handle;
3a2b67e6	69	unsigned long entries[]; /* Variable-sized array of entries. */
cd11016e AP	70	};
cd11016e AP	71
a5f1783b VB	72	static bool __stack_depot_want_early_init __initdata = IS_ENABLED(CONFIG_STACKDEPOT_ALWAYS_INIT);
	73	static bool __stack_depot_early_init_passed __initdata;
	74
cd11016e AP	75	static void *stack_slabs[STACK_ALLOC_MAX_SLABS];
	76
	77	static int depot_index;
	78	static int next_slab_inited;
	79	static size_t depot_offset;
78564b94	80	static DEFINE_RAW_SPINLOCK(depot_lock);
cd11016e	81
83a4f1ef AP	82	unsigned int stack_depot_get_extra_bits(depot_stack_handle_t handle)
	83	{
	84	union handle_parts parts = { .handle = handle };
	85
	86	return parts.extra;
	87	}
	88	EXPORT_SYMBOL(stack_depot_get_extra_bits);
	89
cd11016e AP	90	static bool init_stack_slab(void **prealloc)
	91	{
	92	if (!*prealloc)
	93	return false;
	94	/*
	95	* This smp_load_acquire() pairs with smp_store_release() to
	96	* \|next_slab_inited\| below and in depot_alloc_stack().
	97	*/
	98	if (smp_load_acquire(&next_slab_inited))
	99	return true;
	100	if (stack_slabs[depot_index] == NULL) {
	101	stack_slabs[depot_index] = *prealloc;
305e519c	102	*prealloc = NULL;
cd11016e	103	} else {
305e519c AP	104	/* If this is the last depot slab, do not touch the next one. */
	105	if (depot_index + 1 < STACK_ALLOC_MAX_SLABS) {
	106	stack_slabs[depot_index + 1] = *prealloc;
	107	*prealloc = NULL;
	108	}
cd11016e AP	109	/*
cd11016e AP	110	* This smp_store_release pairs with smp_load_acquire() from
ee050dc8	111	* \|next_slab_inited\| above and in stack_depot_save().
cd11016e AP	112	*/
	113	smp_store_release(&next_slab_inited, 1);
	114	}
cd11016e AP	115	return true;
	116	}
	117
	118	/* Allocation of a new stack in raw storage */
7f2b8818 ME	119	static struct stack_record *
7f2b8818 ME	120	depot_alloc_stack(unsigned long entries, int size, u32 hash, void *prealloc)
cd11016e	121	{
cd11016e	122	struct stack_record *stack;
3a2b67e6	123	size_t required_size = struct_size(stack, entries, size);
cd11016e AP	124
	125	required_size = ALIGN(required_size, 1 << STACK_ALLOC_ALIGN);
	126
	127	if (unlikely(depot_offset + required_size > STACK_ALLOC_SIZE)) {
	128	if (unlikely(depot_index + 1 >= STACK_ALLOC_MAX_SLABS)) {
	129	WARN_ONCE(1, "Stack depot reached limit capacity");
	130	return NULL;
	131	}
	132	depot_index++;
	133	depot_offset = 0;
	134	/*
	135	* smp_store_release() here pairs with smp_load_acquire() from
ee050dc8	136	* \|next_slab_inited\| in stack_depot_save() and
cd11016e AP	137	* init_stack_slab().
	138	*/
	139	if (depot_index + 1 < STACK_ALLOC_MAX_SLABS)
	140	smp_store_release(&next_slab_inited, 0);
	141	}
	142	init_stack_slab(prealloc);
	143	if (stack_slabs[depot_index] == NULL)
	144	return NULL;
	145
	146	stack = stack_slabs[depot_index] + depot_offset;
	147
	148	stack->hash = hash;
	149	stack->size = size;
	150	stack->handle.slabindex = depot_index;
	151	stack->handle.offset = depot_offset >> STACK_ALLOC_ALIGN;
7c31190b	152	stack->handle.valid = 1;
83a4f1ef	153	stack->handle.extra = 0;
47e684aa	154	memcpy(stack->entries, entries, flex_array_size(stack, entries, size));
cd11016e AP	155	depot_offset += required_size;
	156
	157	return stack;
	158	}
	159
f9987921 VB	160	/* one hash table bucket entry per 16kB of memory */
	161	#define STACK_HASH_SCALE 14
	162	/* limited between 4k and 1M buckets */
	163	#define STACK_HASH_ORDER_MIN 12
	164	#define STACK_HASH_ORDER_MAX 20
cd11016e AP	165	#define STACK_HASH_SEED 0x9747b28c
cd11016e AP	166
f9987921 VB	167	static unsigned int stack_hash_order;
	168	static unsigned int stack_hash_mask;
	169
e1fdc403 VJ	170	static bool stack_depot_disable;
	171	static struct stack_record **stack_table;
	172
	173	static int __init is_stack_depot_disabled(char *str)
	174	{
64427985 VJ	175	int ret;
	176
	177	ret = kstrtobool(str, &stack_depot_disable);
	178	if (!ret && stack_depot_disable) {
e1fdc403 VJ	179	pr_info("Stack Depot is disabled\n");
	180	stack_table = NULL;
	181	}
	182	return 0;
	183	}
	184	early_param("stack_depot_disable", is_stack_depot_disabled);
	185
a5f1783b VB	186	void __init stack_depot_want_early_init(void)
	187	{
	188	/* Too late to request early init now */
	189	WARN_ON(__stack_depot_early_init_passed);
	190
	191	__stack_depot_want_early_init = true;
	192	}
	193
	194	int __init stack_depot_early_init(void)
	195	{
f9987921	196	unsigned long entries = 0;
a5f1783b VB	197
	198	/* This is supposed to be called only once, from mm_init() */
	199	if (WARN_ON(__stack_depot_early_init_passed))
	200	return 0;
	201
	202	__stack_depot_early_init_passed = true;
	203
f9987921 VB	204	if (kasan_enabled() && !stack_hash_order)
	205	stack_hash_order = STACK_HASH_ORDER_MAX;
	206
a5f1783b VB	207	if (!__stack_depot_want_early_init \|\| stack_depot_disable)
	208	return 0;
	209
f9987921 VB	210	if (stack_hash_order)
	211	entries = 1UL << stack_hash_order;
	212	stack_table = alloc_large_system_hash("stackdepot",
	213	sizeof(struct stack_record *),
	214	entries,
	215	STACK_HASH_SCALE,
	216	HASH_EARLY \| HASH_ZERO,
	217	NULL,
	218	&stack_hash_mask,
	219	1UL << STACK_HASH_ORDER_MIN,
	220	1UL << STACK_HASH_ORDER_MAX);
a5f1783b VB	221
	222	if (!stack_table) {
	223	pr_err("Stack Depot hash table allocation failed, disabling\n");
	224	stack_depot_disable = true;
	225	return -ENOMEM;
	226	}
	227
	228	return 0;
	229	}
	230
	231	int stack_depot_init(void)
e1fdc403	232	{
2dba5eb1	233	static DEFINE_MUTEX(stack_depot_init_mutex);
a5f1783b	234	int ret = 0;
2dba5eb1 VB	235
	236	mutex_lock(&stack_depot_init_mutex);
	237	if (!stack_depot_disable && !stack_table) {
f9987921 VB	238	unsigned long entries;
	239	int scale = STACK_HASH_SCALE;
	240
	241	if (stack_hash_order) {
	242	entries = 1UL << stack_hash_order;
	243	} else {
	244	entries = nr_free_buffer_pages();
	245	entries = roundup_pow_of_two(entries);
	246
	247	if (scale > PAGE_SHIFT)
	248	entries >>= (scale - PAGE_SHIFT);
	249	else
	250	entries <<= (PAGE_SHIFT - scale);
	251	}
	252
	253	if (entries < 1UL << STACK_HASH_ORDER_MIN)
	254	entries = 1UL << STACK_HASH_ORDER_MIN;
	255	if (entries > 1UL << STACK_HASH_ORDER_MAX)
	256	entries = 1UL << STACK_HASH_ORDER_MAX;
	257
	258	pr_info("Stack Depot allocating hash table of %lu entries with kvcalloc\n",
	259	entries);
	260	stack_table = kvcalloc(entries, sizeof(struct stack_record *), GFP_KERNEL);
a5f1783b	261	if (!stack_table) {
2dba5eb1 VB	262	pr_err("Stack Depot hash table allocation failed, disabling\n");
2dba5eb1 VB	263	stack_depot_disable = true;
a5f1783b	264	ret = -ENOMEM;
2dba5eb1	265	}
f9987921	266	stack_hash_mask = entries - 1;
e1fdc403	267	}
2dba5eb1	268	mutex_unlock(&stack_depot_init_mutex);
a5f1783b	269	return ret;
e1fdc403	270	}
2dba5eb1	271	EXPORT_SYMBOL_GPL(stack_depot_init);
cd11016e AP	272
	273	/* Calculate hash for a stack */
	274	static inline u32 hash_stack(unsigned long *entries, unsigned int size)
	275	{
	276	return jhash2((u32 *)entries,
180644f8 GS	277	array_size(size, sizeof(*entries)) / sizeof(u32),
180644f8 GS	278	STACK_HASH_SEED);
cd11016e AP	279	}
cd11016e AP	280
a571b272 AP	281	/* Use our own, non-instrumented version of memcmp().
	282	*
	283	* We actually don't care about the order, just the equality.
	284	*/
	285	static inline
	286	int stackdepot_memcmp(const unsigned long u1, const unsigned long u2,
	287	unsigned int n)
	288	{
	289	for ( ; n-- ; u1++, u2++) {
	290	if (u1 != u2)
	291	return 1;
	292	}
	293	return 0;
	294	}
	295
cd11016e AP	296	/* Find a stack that is equal to the one stored in entries in the hash */
	297	static inline struct stack_record find_stack(struct stack_record bucket,
	298	unsigned long *entries, int size,
	299	u32 hash)
	300	{
	301	struct stack_record *found;
	302
	303	for (found = bucket; found; found = found->next) {
	304	if (found->hash == hash &&
	305	found->size == size &&
a571b272	306	!stackdepot_memcmp(entries, found->entries, size))
cd11016e	307	return found;
cd11016e AP	308	}
	309	return NULL;
	310	}
	311
0f68d45e IK	312	/**
	313	* stack_depot_snprint - print stack entries from a depot into a buffer
	314	*
	315	* @handle: Stack depot handle which was returned from
	316	* stack_depot_save().
	317	* @buf: Pointer to the print buffer
	318	*
	319	* @size: Size of the print buffer
	320	*
	321	* @spaces: Number of leading spaces to print
	322	*
	323	* Return: Number of bytes printed.
	324	*/
	325	int stack_depot_snprint(depot_stack_handle_t handle, char *buf, size_t size,
	326	int spaces)
	327	{
	328	unsigned long *entries;
	329	unsigned int nr_entries;
	330
	331	nr_entries = stack_depot_fetch(handle, &entries);
	332	return nr_entries ? stack_trace_snprint(buf, size, entries, nr_entries,
	333	spaces) : 0;
	334	}
	335	EXPORT_SYMBOL_GPL(stack_depot_snprint);
	336
505be481 IK	337	/**
	338	* stack_depot_print - print stack entries from a depot
	339	*
	340	* @stack: Stack depot handle which was returned from
	341	* stack_depot_save().
	342	*
	343	*/
	344	void stack_depot_print(depot_stack_handle_t stack)
	345	{
	346	unsigned long *entries;
	347	unsigned int nr_entries;
	348
	349	nr_entries = stack_depot_fetch(stack, &entries);
	350	if (nr_entries > 0)
	351	stack_trace_print(entries, nr_entries, 0);
	352	}
	353	EXPORT_SYMBOL_GPL(stack_depot_print);
	354
c0cfc337 TG	355	/**
	356	* stack_depot_fetch - Fetch stack entries from a depot
	357	*
	358	* @handle: Stack depot handle which was returned from
	359	* stack_depot_save().
	360	* @entries: Pointer to store the entries address
	361	*
	362	* Return: The number of trace entries for this depot.
	363	*/
	364	unsigned int stack_depot_fetch(depot_stack_handle_t handle,
	365	unsigned long **entries)
cd11016e AP	366	{
cd11016e AP	367	union handle_parts parts = { .handle = handle };
69866e15	368	void *slab;
cd11016e	369	size_t offset = parts.offset << STACK_ALLOC_ALIGN;
69866e15 AP	370	struct stack_record *stack;
	371
	372	*entries = NULL;
4d4712c1 IK	373	if (!handle)
	374	return 0;
	375
69866e15 AP	376	if (parts.slabindex > depot_index) {
	377	WARN(1, "slab index %d out of bounds (%d) for stack id %08x\n",
	378	parts.slabindex, depot_index, handle);
	379	return 0;
	380	}
	381	slab = stack_slabs[parts.slabindex];
	382	if (!slab)
	383	return 0;
	384	stack = slab + offset;
cd11016e	385
c0cfc337 TG	386	*entries = stack->entries;
	387	return stack->size;
	388	}
	389	EXPORT_SYMBOL_GPL(stack_depot_fetch);
	390
cd11016e	391	/**
11ac25c6	392	* __stack_depot_save - Save a stack trace from an array
c0cfc337 TG	393	*
	394	* @entries: Pointer to storage array
	395	* @nr_entries: Size of the storage array
83a4f1ef	396	* @extra_bits: Flags to store in unused bits of depot_stack_handle_t
c0cfc337	397	* @alloc_flags: Allocation gfp flags
11ac25c6 ME	398	* @can_alloc: Allocate stack slabs (increased chance of failure if false)
	399	*
	400	* Saves a stack trace from @entries array of size @nr_entries. If @can_alloc is
	401	* %true, is allowed to replenish the stack slab pool in case no space is left
	402	* (allocates using GFP flags of @alloc_flags). If @can_alloc is %false, avoids
	403	* any allocations and will fail if no space is left to store the stack trace.
	404	*
e9400660 ME	405	* If the stack trace in @entries is from an interrupt, only the portion up to
	406	* interrupt entry is saved.
	407	*
83a4f1ef AP	408	* Additional opaque flags can be passed in @extra_bits, stored in the unused
	409	* bits of the stack handle, and retrieved using stack_depot_get_extra_bits()
	410	* without calling stack_depot_fetch().
	411	*
11ac25c6 ME	412	* Context: Any context, but setting @can_alloc to %false is required if
	413	* alloc_pages() cannot be used from the current context. Currently
	414	* this is the case from contexts where neither %GFP_ATOMIC nor
	415	* %GFP_NOWAIT can be used (NMI, raw_spin_lock).
cd11016e	416	*
11ac25c6	417	* Return: The handle of the stack struct stored in depot, 0 on failure.
cd11016e	418	*/
11ac25c6 ME	419	depot_stack_handle_t __stack_depot_save(unsigned long *entries,
11ac25c6 ME	420	unsigned int nr_entries,
83a4f1ef	421	unsigned int extra_bits,
11ac25c6	422	gfp_t alloc_flags, bool can_alloc)
cd11016e	423	{
cd11016e	424	struct stack_record found = NULL, *bucket;
83a4f1ef	425	union handle_parts retval = { .handle = 0 };
cd11016e AP	426	struct page *page = NULL;
cd11016e AP	427	void *prealloc = NULL;
c0cfc337 TG	428	unsigned long flags;
c0cfc337 TG	429	u32 hash;
cd11016e	430
e9400660 ME	431	/*
	432	* If this stack trace is from an interrupt, including anything before
	433	* interrupt entry usually leads to unbounded stackdepot growth.
	434	*
	435	* Because use of filter_irq_stacks() is a requirement to ensure
	436	* stackdepot can efficiently deduplicate interrupt stacks, always
	437	* filter_irq_stacks() to simplify all callers' use of stackdepot.
	438	*/
	439	nr_entries = filter_irq_stacks(entries, nr_entries);
	440
e1fdc403	441	if (unlikely(nr_entries == 0) \|\| stack_depot_disable)
cd11016e AP	442	goto fast_exit;
cd11016e AP	443
c0cfc337	444	hash = hash_stack(entries, nr_entries);
f9987921	445	bucket = &stack_table[hash & stack_hash_mask];
cd11016e AP	446
	447	/*
	448	* Fast path: look the stack trace up without locking.
	449	* The smp_load_acquire() here pairs with smp_store_release() to
	450	* \|bucket\| below.
	451	*/
c0cfc337 TG	452	found = find_stack(smp_load_acquire(bucket), entries,
c0cfc337 TG	453	nr_entries, hash);
cd11016e AP	454	if (found)
	455	goto exit;
	456
	457	/*
	458	* Check if the current or the next stack slab need to be initialized.
	459	* If so, allocate the memory - we won't be able to do that under the
	460	* lock.
	461	*
	462	* The smp_load_acquire() here pairs with smp_store_release() to
	463	* \|next_slab_inited\| in depot_alloc_stack() and init_stack_slab().
	464	*/
11ac25c6	465	if (unlikely(can_alloc && !smp_load_acquire(&next_slab_inited))) {
cd11016e AP	466	/*
	467	* Zero out zone modifiers, as we don't have specific zone
	468	* requirements. Keep the flags related to allocation in atomic
	469	* contexts and I/O.
	470	*/
	471	alloc_flags &= ~GFP_ZONEMASK;
	472	alloc_flags &= (GFP_ATOMIC \| GFP_KERNEL);
87cc271d	473	alloc_flags \|= __GFP_NOWARN;
cd11016e AP	474	page = alloc_pages(alloc_flags, STACK_ALLOC_ORDER);
	475	if (page)
	476	prealloc = page_address(page);
	477	}
	478
78564b94	479	raw_spin_lock_irqsave(&depot_lock, flags);
cd11016e	480
c0cfc337	481	found = find_stack(*bucket, entries, nr_entries, hash);
cd11016e	482	if (!found) {
7f2b8818 ME	483	struct stack_record *new = depot_alloc_stack(entries, nr_entries, hash, &prealloc);
7f2b8818 ME	484
cd11016e AP	485	if (new) {
	486	new->next = *bucket;
	487	/*
	488	* This smp_store_release() pairs with
	489	* smp_load_acquire() from \|bucket\| above.
	490	*/
	491	smp_store_release(bucket, new);
	492	found = new;
	493	}
	494	} else if (prealloc) {
	495	/*
	496	* We didn't need to store this stack trace, but let's keep
	497	* the preallocated memory for the future.
	498	*/
	499	WARN_ON(!init_stack_slab(&prealloc));
	500	}
	501
78564b94	502	raw_spin_unlock_irqrestore(&depot_lock, flags);
cd11016e AP	503	exit:
	504	if (prealloc) {
	505	/* Nobody used this memory, ok to free it. */
	506	free_pages((unsigned long)prealloc, STACK_ALLOC_ORDER);
	507	}
	508	if (found)
83a4f1ef	509	retval.handle = found->handle.handle;
cd11016e	510	fast_exit:
83a4f1ef AP	511	retval.extra = extra_bits;
	512
	513	return retval.handle;
cd11016e	514	}
11ac25c6 ME	515	EXPORT_SYMBOL_GPL(__stack_depot_save);
	516
	517	/**
	518	* stack_depot_save - Save a stack trace from an array
	519	*
	520	* @entries: Pointer to storage array
	521	* @nr_entries: Size of the storage array
	522	* @alloc_flags: Allocation gfp flags
	523	*
	524	* Context: Contexts where allocations via alloc_pages() are allowed.
	525	* See __stack_depot_save() for more details.
	526	*
	527	* Return: The handle of the stack struct stored in depot, 0 on failure.
	528	*/
	529	depot_stack_handle_t stack_depot_save(unsigned long *entries,
	530	unsigned int nr_entries,
	531	gfp_t alloc_flags)
	532	{
83a4f1ef	533	return __stack_depot_save(entries, nr_entries, 0, alloc_flags, true);
11ac25c6	534	}
c0cfc337	535	EXPORT_SYMBOL_GPL(stack_depot_save);