Loading...
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
45#define STACK_ALLOC_NULL_PROTECTION_BITS 1
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)
51#define STACK_ALLOC_INDEX_BITS (DEPOT_STACK_BITS - \
52 STACK_ALLOC_NULL_PROTECTION_BITS - STACK_ALLOC_OFFSET_BITS)
53#define STACK_ALLOC_SLABS_CAP 8192
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. */
59union handle_parts {
60 depot_stack_handle_t handle;
61 struct {
62 u32 slabindex : STACK_ALLOC_INDEX_BITS;
63 u32 offset : STACK_ALLOC_OFFSET_BITS;
64 u32 valid : STACK_ALLOC_NULL_PROTECTION_BITS;
65 };
66};
67
68struct 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
76static void *stack_slabs[STACK_ALLOC_MAX_SLABS];
77
78static int depot_index;
79static int next_slab_inited;
80static size_t depot_offset;
81static DEFINE_SPINLOCK(depot_lock);
82
83static 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 */
108static 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;
142 stack->handle.valid = 1;
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
154static struct stack_record *stack_table[STACK_HASH_SIZE] = {
155 [0 ... STACK_HASH_SIZE - 1] = NULL
156};
157
158/* Calculate hash for a stack */
159static 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 */
167static 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
184void 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}
195EXPORT_SYMBOL_GPL(depot_fetch_stack);
196
197/**
198 * depot_save_stack - save stack in a stack depot.
199 * @trace - the stacktrace to save.
200 * @alloc_flags - flags for allocating additional memory if required.
201 *
202 * Returns the handle of the stack struct stored in depot.
203 */
204depot_stack_handle_t depot_save_stack(struct stack_trace *trace,
205 gfp_t alloc_flags)
206{
207 u32 hash;
208 depot_stack_handle_t retval = 0;
209 struct stack_record *found = NULL, **bucket;
210 unsigned long flags;
211 struct page *page = NULL;
212 void *prealloc = NULL;
213
214 if (unlikely(trace->nr_entries == 0))
215 goto fast_exit;
216
217 hash = hash_stack(trace->entries, trace->nr_entries);
218 bucket = &stack_table[hash & STACK_HASH_MASK];
219
220 /*
221 * Fast path: look the stack trace up without locking.
222 * The smp_load_acquire() here pairs with smp_store_release() to
223 * |bucket| below.
224 */
225 found = find_stack(smp_load_acquire(bucket), trace->entries,
226 trace->nr_entries, hash);
227 if (found)
228 goto exit;
229
230 /*
231 * Check if the current or the next stack slab need to be initialized.
232 * If so, allocate the memory - we won't be able to do that under the
233 * lock.
234 *
235 * The smp_load_acquire() here pairs with smp_store_release() to
236 * |next_slab_inited| in depot_alloc_stack() and init_stack_slab().
237 */
238 if (unlikely(!smp_load_acquire(&next_slab_inited))) {
239 /*
240 * Zero out zone modifiers, as we don't have specific zone
241 * requirements. Keep the flags related to allocation in atomic
242 * contexts and I/O.
243 */
244 alloc_flags &= ~GFP_ZONEMASK;
245 alloc_flags &= (GFP_ATOMIC | GFP_KERNEL);
246 alloc_flags |= __GFP_NOWARN;
247 page = alloc_pages(alloc_flags, STACK_ALLOC_ORDER);
248 if (page)
249 prealloc = page_address(page);
250 }
251
252 spin_lock_irqsave(&depot_lock, flags);
253
254 found = find_stack(*bucket, trace->entries, trace->nr_entries, hash);
255 if (!found) {
256 struct stack_record *new =
257 depot_alloc_stack(trace->entries, trace->nr_entries,
258 hash, &prealloc, alloc_flags);
259 if (new) {
260 new->next = *bucket;
261 /*
262 * This smp_store_release() pairs with
263 * smp_load_acquire() from |bucket| above.
264 */
265 smp_store_release(bucket, new);
266 found = new;
267 }
268 } else if (prealloc) {
269 /*
270 * We didn't need to store this stack trace, but let's keep
271 * the preallocated memory for the future.
272 */
273 WARN_ON(!init_stack_slab(&prealloc));
274 }
275
276 spin_unlock_irqrestore(&depot_lock, flags);
277exit:
278 if (prealloc) {
279 /* Nobody used this memory, ok to free it. */
280 free_pages((unsigned long)prealloc, STACK_ALLOC_ORDER);
281 }
282 if (found)
283 retval = found->handle.handle;
284fast_exit:
285 return retval;
286}
287EXPORT_SYMBOL_GPL(depot_save_stack);
1// SPDX-License-Identifier: GPL-2.0-only
2/*
3 * Stack depot - a stack trace storage that avoids duplication.
4 *
5 * Internally, stack depot maintains a hash table of unique stacktraces. The
6 * stack traces themselves are stored contiguously one after another in a set
7 * of separate page allocations.
8 *
9 * Author: Alexander Potapenko <glider@google.com>
10 * Copyright (C) 2016 Google, Inc.
11 *
12 * Based on the code by Dmitry Chernenkov.
13 */
14
15#define pr_fmt(fmt) "stackdepot: " fmt
16
17#include <linux/debugfs.h>
18#include <linux/gfp.h>
19#include <linux/jhash.h>
20#include <linux/kernel.h>
21#include <linux/kmsan.h>
22#include <linux/list.h>
23#include <linux/mm.h>
24#include <linux/mutex.h>
25#include <linux/poison.h>
26#include <linux/printk.h>
27#include <linux/rculist.h>
28#include <linux/rcupdate.h>
29#include <linux/refcount.h>
30#include <linux/slab.h>
31#include <linux/spinlock.h>
32#include <linux/stacktrace.h>
33#include <linux/stackdepot.h>
34#include <linux/string.h>
35#include <linux/types.h>
36#include <linux/memblock.h>
37#include <linux/kasan-enabled.h>
38
39#define DEPOT_HANDLE_BITS (sizeof(depot_stack_handle_t) * 8)
40
41#define DEPOT_POOL_ORDER 2 /* Pool size order, 4 pages */
42#define DEPOT_POOL_SIZE (1LL << (PAGE_SHIFT + DEPOT_POOL_ORDER))
43#define DEPOT_STACK_ALIGN 4
44#define DEPOT_OFFSET_BITS (DEPOT_POOL_ORDER + PAGE_SHIFT - DEPOT_STACK_ALIGN)
45#define DEPOT_POOL_INDEX_BITS (DEPOT_HANDLE_BITS - DEPOT_OFFSET_BITS - \
46 STACK_DEPOT_EXTRA_BITS)
47#define DEPOT_POOLS_CAP 8192
48#define DEPOT_MAX_POOLS \
49 (((1LL << (DEPOT_POOL_INDEX_BITS)) < DEPOT_POOLS_CAP) ? \
50 (1LL << (DEPOT_POOL_INDEX_BITS)) : DEPOT_POOLS_CAP)
51
52/* Compact structure that stores a reference to a stack. */
53union handle_parts {
54 depot_stack_handle_t handle;
55 struct {
56 u32 pool_index : DEPOT_POOL_INDEX_BITS;
57 u32 offset : DEPOT_OFFSET_BITS;
58 u32 extra : STACK_DEPOT_EXTRA_BITS;
59 };
60};
61
62struct stack_record {
63 struct list_head hash_list; /* Links in the hash table */
64 u32 hash; /* Hash in hash table */
65 u32 size; /* Number of stored frames */
66 union handle_parts handle; /* Constant after initialization */
67 refcount_t count;
68 union {
69 unsigned long entries[CONFIG_STACKDEPOT_MAX_FRAMES]; /* Frames */
70 struct {
71 /*
72 * An important invariant of the implementation is to
73 * only place a stack record onto the freelist iff its
74 * refcount is zero. Because stack records with a zero
75 * refcount are never considered as valid, it is safe to
76 * union @entries and freelist management state below.
77 * Conversely, as soon as an entry is off the freelist
78 * and its refcount becomes non-zero, the below must not
79 * be accessed until being placed back on the freelist.
80 */
81 struct list_head free_list; /* Links in the freelist */
82 unsigned long rcu_state; /* RCU cookie */
83 };
84 };
85};
86
87static bool stack_depot_disabled;
88static bool __stack_depot_early_init_requested __initdata = IS_ENABLED(CONFIG_STACKDEPOT_ALWAYS_INIT);
89static bool __stack_depot_early_init_passed __initdata;
90
91/* Use one hash table bucket per 16 KB of memory. */
92#define STACK_HASH_TABLE_SCALE 14
93/* Limit the number of buckets between 4K and 1M. */
94#define STACK_BUCKET_NUMBER_ORDER_MIN 12
95#define STACK_BUCKET_NUMBER_ORDER_MAX 20
96/* Initial seed for jhash2. */
97#define STACK_HASH_SEED 0x9747b28c
98
99/* Hash table of stored stack records. */
100static struct list_head *stack_table;
101/* Fixed order of the number of table buckets. Used when KASAN is enabled. */
102static unsigned int stack_bucket_number_order;
103/* Hash mask for indexing the table. */
104static unsigned int stack_hash_mask;
105
106/* Array of memory regions that store stack records. */
107static void *stack_pools[DEPOT_MAX_POOLS];
108/* Newly allocated pool that is not yet added to stack_pools. */
109static void *new_pool;
110/* Number of pools in stack_pools. */
111static int pools_num;
112/* Offset to the unused space in the currently used pool. */
113static size_t pool_offset = DEPOT_POOL_SIZE;
114/* Freelist of stack records within stack_pools. */
115static LIST_HEAD(free_stacks);
116/* The lock must be held when performing pool or freelist modifications. */
117static DEFINE_RAW_SPINLOCK(pool_lock);
118
119/* Statistics counters for debugfs. */
120enum depot_counter_id {
121 DEPOT_COUNTER_REFD_ALLOCS,
122 DEPOT_COUNTER_REFD_FREES,
123 DEPOT_COUNTER_REFD_INUSE,
124 DEPOT_COUNTER_FREELIST_SIZE,
125 DEPOT_COUNTER_PERSIST_COUNT,
126 DEPOT_COUNTER_PERSIST_BYTES,
127 DEPOT_COUNTER_COUNT,
128};
129static long counters[DEPOT_COUNTER_COUNT];
130static const char *const counter_names[] = {
131 [DEPOT_COUNTER_REFD_ALLOCS] = "refcounted_allocations",
132 [DEPOT_COUNTER_REFD_FREES] = "refcounted_frees",
133 [DEPOT_COUNTER_REFD_INUSE] = "refcounted_in_use",
134 [DEPOT_COUNTER_FREELIST_SIZE] = "freelist_size",
135 [DEPOT_COUNTER_PERSIST_COUNT] = "persistent_count",
136 [DEPOT_COUNTER_PERSIST_BYTES] = "persistent_bytes",
137};
138static_assert(ARRAY_SIZE(counter_names) == DEPOT_COUNTER_COUNT);
139
140static int __init disable_stack_depot(char *str)
141{
142 return kstrtobool(str, &stack_depot_disabled);
143}
144early_param("stack_depot_disable", disable_stack_depot);
145
146void __init stack_depot_request_early_init(void)
147{
148 /* Too late to request early init now. */
149 WARN_ON(__stack_depot_early_init_passed);
150
151 __stack_depot_early_init_requested = true;
152}
153
154/* Initialize list_head's within the hash table. */
155static void init_stack_table(unsigned long entries)
156{
157 unsigned long i;
158
159 for (i = 0; i < entries; i++)
160 INIT_LIST_HEAD(&stack_table[i]);
161}
162
163/* Allocates a hash table via memblock. Can only be used during early boot. */
164int __init stack_depot_early_init(void)
165{
166 unsigned long entries = 0;
167
168 /* This function must be called only once, from mm_init(). */
169 if (WARN_ON(__stack_depot_early_init_passed))
170 return 0;
171 __stack_depot_early_init_passed = true;
172
173 /*
174 * Print disabled message even if early init has not been requested:
175 * stack_depot_init() will not print one.
176 */
177 if (stack_depot_disabled) {
178 pr_info("disabled\n");
179 return 0;
180 }
181
182 /*
183 * If KASAN is enabled, use the maximum order: KASAN is frequently used
184 * in fuzzing scenarios, which leads to a large number of different
185 * stack traces being stored in stack depot.
186 */
187 if (kasan_enabled() && !stack_bucket_number_order)
188 stack_bucket_number_order = STACK_BUCKET_NUMBER_ORDER_MAX;
189
190 /*
191 * Check if early init has been requested after setting
192 * stack_bucket_number_order: stack_depot_init() uses its value.
193 */
194 if (!__stack_depot_early_init_requested)
195 return 0;
196
197 /*
198 * If stack_bucket_number_order is not set, leave entries as 0 to rely
199 * on the automatic calculations performed by alloc_large_system_hash().
200 */
201 if (stack_bucket_number_order)
202 entries = 1UL << stack_bucket_number_order;
203 pr_info("allocating hash table via alloc_large_system_hash\n");
204 stack_table = alloc_large_system_hash("stackdepot",
205 sizeof(struct list_head),
206 entries,
207 STACK_HASH_TABLE_SCALE,
208 HASH_EARLY,
209 NULL,
210 &stack_hash_mask,
211 1UL << STACK_BUCKET_NUMBER_ORDER_MIN,
212 1UL << STACK_BUCKET_NUMBER_ORDER_MAX);
213 if (!stack_table) {
214 pr_err("hash table allocation failed, disabling\n");
215 stack_depot_disabled = true;
216 return -ENOMEM;
217 }
218 if (!entries) {
219 /*
220 * Obtain the number of entries that was calculated by
221 * alloc_large_system_hash().
222 */
223 entries = stack_hash_mask + 1;
224 }
225 init_stack_table(entries);
226
227 return 0;
228}
229
230/* Allocates a hash table via kvcalloc. Can be used after boot. */
231int stack_depot_init(void)
232{
233 static DEFINE_MUTEX(stack_depot_init_mutex);
234 unsigned long entries;
235 int ret = 0;
236
237 mutex_lock(&stack_depot_init_mutex);
238
239 if (stack_depot_disabled || stack_table)
240 goto out_unlock;
241
242 /*
243 * Similarly to stack_depot_early_init, use stack_bucket_number_order
244 * if assigned, and rely on automatic scaling otherwise.
245 */
246 if (stack_bucket_number_order) {
247 entries = 1UL << stack_bucket_number_order;
248 } else {
249 int scale = STACK_HASH_TABLE_SCALE;
250
251 entries = nr_free_buffer_pages();
252 entries = roundup_pow_of_two(entries);
253
254 if (scale > PAGE_SHIFT)
255 entries >>= (scale - PAGE_SHIFT);
256 else
257 entries <<= (PAGE_SHIFT - scale);
258 }
259
260 if (entries < 1UL << STACK_BUCKET_NUMBER_ORDER_MIN)
261 entries = 1UL << STACK_BUCKET_NUMBER_ORDER_MIN;
262 if (entries > 1UL << STACK_BUCKET_NUMBER_ORDER_MAX)
263 entries = 1UL << STACK_BUCKET_NUMBER_ORDER_MAX;
264
265 pr_info("allocating hash table of %lu entries via kvcalloc\n", entries);
266 stack_table = kvcalloc(entries, sizeof(struct list_head), GFP_KERNEL);
267 if (!stack_table) {
268 pr_err("hash table allocation failed, disabling\n");
269 stack_depot_disabled = true;
270 ret = -ENOMEM;
271 goto out_unlock;
272 }
273 stack_hash_mask = entries - 1;
274 init_stack_table(entries);
275
276out_unlock:
277 mutex_unlock(&stack_depot_init_mutex);
278
279 return ret;
280}
281EXPORT_SYMBOL_GPL(stack_depot_init);
282
283/*
284 * Initializes new stack pool, and updates the list of pools.
285 */
286static bool depot_init_pool(void **prealloc)
287{
288 lockdep_assert_held(&pool_lock);
289
290 if (unlikely(pools_num >= DEPOT_MAX_POOLS)) {
291 /* Bail out if we reached the pool limit. */
292 WARN_ON_ONCE(pools_num > DEPOT_MAX_POOLS); /* should never happen */
293 WARN_ON_ONCE(!new_pool); /* to avoid unnecessary pre-allocation */
294 WARN_ONCE(1, "Stack depot reached limit capacity");
295 return false;
296 }
297
298 if (!new_pool && *prealloc) {
299 /* We have preallocated memory, use it. */
300 WRITE_ONCE(new_pool, *prealloc);
301 *prealloc = NULL;
302 }
303
304 if (!new_pool)
305 return false; /* new_pool and *prealloc are NULL */
306
307 /* Save reference to the pool to be used by depot_fetch_stack(). */
308 stack_pools[pools_num] = new_pool;
309
310 /*
311 * Stack depot tries to keep an extra pool allocated even before it runs
312 * out of space in the currently used pool.
313 *
314 * To indicate that a new preallocation is needed new_pool is reset to
315 * NULL; do not reset to NULL if we have reached the maximum number of
316 * pools.
317 */
318 if (pools_num < DEPOT_MAX_POOLS)
319 WRITE_ONCE(new_pool, NULL);
320 else
321 WRITE_ONCE(new_pool, STACK_DEPOT_POISON);
322
323 /* Pairs with concurrent READ_ONCE() in depot_fetch_stack(). */
324 WRITE_ONCE(pools_num, pools_num + 1);
325 ASSERT_EXCLUSIVE_WRITER(pools_num);
326
327 pool_offset = 0;
328
329 return true;
330}
331
332/* Keeps the preallocated memory to be used for a new stack depot pool. */
333static void depot_keep_new_pool(void **prealloc)
334{
335 lockdep_assert_held(&pool_lock);
336
337 /*
338 * If a new pool is already saved or the maximum number of
339 * pools is reached, do not use the preallocated memory.
340 */
341 if (new_pool)
342 return;
343
344 WRITE_ONCE(new_pool, *prealloc);
345 *prealloc = NULL;
346}
347
348/*
349 * Try to initialize a new stack record from the current pool, a cached pool, or
350 * the current pre-allocation.
351 */
352static struct stack_record *depot_pop_free_pool(void **prealloc, size_t size)
353{
354 struct stack_record *stack;
355 void *current_pool;
356 u32 pool_index;
357
358 lockdep_assert_held(&pool_lock);
359
360 if (pool_offset + size > DEPOT_POOL_SIZE) {
361 if (!depot_init_pool(prealloc))
362 return NULL;
363 }
364
365 if (WARN_ON_ONCE(pools_num < 1))
366 return NULL;
367 pool_index = pools_num - 1;
368 current_pool = stack_pools[pool_index];
369 if (WARN_ON_ONCE(!current_pool))
370 return NULL;
371
372 stack = current_pool + pool_offset;
373
374 /* Pre-initialize handle once. */
375 stack->handle.pool_index = pool_index;
376 stack->handle.offset = pool_offset >> DEPOT_STACK_ALIGN;
377 stack->handle.extra = 0;
378 INIT_LIST_HEAD(&stack->hash_list);
379
380 pool_offset += size;
381
382 return stack;
383}
384
385/* Try to find next free usable entry from the freelist. */
386static struct stack_record *depot_pop_free(void)
387{
388 struct stack_record *stack;
389
390 lockdep_assert_held(&pool_lock);
391
392 if (list_empty(&free_stacks))
393 return NULL;
394
395 /*
396 * We maintain the invariant that the elements in front are least
397 * recently used, and are therefore more likely to be associated with an
398 * RCU grace period in the past. Consequently it is sufficient to only
399 * check the first entry.
400 */
401 stack = list_first_entry(&free_stacks, struct stack_record, free_list);
402 if (!poll_state_synchronize_rcu(stack->rcu_state))
403 return NULL;
404
405 list_del(&stack->free_list);
406 counters[DEPOT_COUNTER_FREELIST_SIZE]--;
407
408 return stack;
409}
410
411static inline size_t depot_stack_record_size(struct stack_record *s, unsigned int nr_entries)
412{
413 const size_t used = flex_array_size(s, entries, nr_entries);
414 const size_t unused = sizeof(s->entries) - used;
415
416 WARN_ON_ONCE(sizeof(s->entries) < used);
417
418 return ALIGN(sizeof(struct stack_record) - unused, 1 << DEPOT_STACK_ALIGN);
419}
420
421/* Allocates a new stack in a stack depot pool. */
422static struct stack_record *
423depot_alloc_stack(unsigned long *entries, unsigned int nr_entries, u32 hash, depot_flags_t flags, void **prealloc)
424{
425 struct stack_record *stack = NULL;
426 size_t record_size;
427
428 lockdep_assert_held(&pool_lock);
429
430 /* This should already be checked by public API entry points. */
431 if (WARN_ON_ONCE(!nr_entries))
432 return NULL;
433
434 /* Limit number of saved frames to CONFIG_STACKDEPOT_MAX_FRAMES. */
435 if (nr_entries > CONFIG_STACKDEPOT_MAX_FRAMES)
436 nr_entries = CONFIG_STACKDEPOT_MAX_FRAMES;
437
438 if (flags & STACK_DEPOT_FLAG_GET) {
439 /*
440 * Evictable entries have to allocate the max. size so they may
441 * safely be re-used by differently sized allocations.
442 */
443 record_size = depot_stack_record_size(stack, CONFIG_STACKDEPOT_MAX_FRAMES);
444 stack = depot_pop_free();
445 } else {
446 record_size = depot_stack_record_size(stack, nr_entries);
447 }
448
449 if (!stack) {
450 stack = depot_pop_free_pool(prealloc, record_size);
451 if (!stack)
452 return NULL;
453 }
454
455 /* Save the stack trace. */
456 stack->hash = hash;
457 stack->size = nr_entries;
458 /* stack->handle is already filled in by depot_pop_free_pool(). */
459 memcpy(stack->entries, entries, flex_array_size(stack, entries, nr_entries));
460
461 if (flags & STACK_DEPOT_FLAG_GET) {
462 refcount_set(&stack->count, 1);
463 counters[DEPOT_COUNTER_REFD_ALLOCS]++;
464 counters[DEPOT_COUNTER_REFD_INUSE]++;
465 } else {
466 /* Warn on attempts to switch to refcounting this entry. */
467 refcount_set(&stack->count, REFCOUNT_SATURATED);
468 counters[DEPOT_COUNTER_PERSIST_COUNT]++;
469 counters[DEPOT_COUNTER_PERSIST_BYTES] += record_size;
470 }
471
472 /*
473 * Let KMSAN know the stored stack record is initialized. This shall
474 * prevent false positive reports if instrumented code accesses it.
475 */
476 kmsan_unpoison_memory(stack, record_size);
477
478 return stack;
479}
480
481static struct stack_record *depot_fetch_stack(depot_stack_handle_t handle)
482{
483 const int pools_num_cached = READ_ONCE(pools_num);
484 union handle_parts parts = { .handle = handle };
485 void *pool;
486 size_t offset = parts.offset << DEPOT_STACK_ALIGN;
487 struct stack_record *stack;
488
489 lockdep_assert_not_held(&pool_lock);
490
491 if (parts.pool_index > pools_num_cached) {
492 WARN(1, "pool index %d out of bounds (%d) for stack id %08x\n",
493 parts.pool_index, pools_num_cached, handle);
494 return NULL;
495 }
496
497 pool = stack_pools[parts.pool_index];
498 if (WARN_ON(!pool))
499 return NULL;
500
501 stack = pool + offset;
502 if (WARN_ON(!refcount_read(&stack->count)))
503 return NULL;
504
505 return stack;
506}
507
508/* Links stack into the freelist. */
509static void depot_free_stack(struct stack_record *stack)
510{
511 unsigned long flags;
512
513 lockdep_assert_not_held(&pool_lock);
514
515 raw_spin_lock_irqsave(&pool_lock, flags);
516 printk_deferred_enter();
517
518 /*
519 * Remove the entry from the hash list. Concurrent list traversal may
520 * still observe the entry, but since the refcount is zero, this entry
521 * will no longer be considered as valid.
522 */
523 list_del_rcu(&stack->hash_list);
524
525 /*
526 * Due to being used from constrained contexts such as the allocators,
527 * NMI, or even RCU itself, stack depot cannot rely on primitives that
528 * would sleep (such as synchronize_rcu()) or recursively call into
529 * stack depot again (such as call_rcu()).
530 *
531 * Instead, get an RCU cookie, so that we can ensure this entry isn't
532 * moved onto another list until the next grace period, and concurrent
533 * RCU list traversal remains safe.
534 */
535 stack->rcu_state = get_state_synchronize_rcu();
536
537 /*
538 * Add the entry to the freelist tail, so that older entries are
539 * considered first - their RCU cookie is more likely to no longer be
540 * associated with the current grace period.
541 */
542 list_add_tail(&stack->free_list, &free_stacks);
543
544 counters[DEPOT_COUNTER_FREELIST_SIZE]++;
545 counters[DEPOT_COUNTER_REFD_FREES]++;
546 counters[DEPOT_COUNTER_REFD_INUSE]--;
547
548 printk_deferred_exit();
549 raw_spin_unlock_irqrestore(&pool_lock, flags);
550}
551
552/* Calculates the hash for a stack. */
553static inline u32 hash_stack(unsigned long *entries, unsigned int size)
554{
555 return jhash2((u32 *)entries,
556 array_size(size, sizeof(*entries)) / sizeof(u32),
557 STACK_HASH_SEED);
558}
559
560/*
561 * Non-instrumented version of memcmp().
562 * Does not check the lexicographical order, only the equality.
563 */
564static inline
565int stackdepot_memcmp(const unsigned long *u1, const unsigned long *u2,
566 unsigned int n)
567{
568 for ( ; n-- ; u1++, u2++) {
569 if (*u1 != *u2)
570 return 1;
571 }
572 return 0;
573}
574
575/* Finds a stack in a bucket of the hash table. */
576static inline struct stack_record *find_stack(struct list_head *bucket,
577 unsigned long *entries, int size,
578 u32 hash, depot_flags_t flags)
579{
580 struct stack_record *stack, *ret = NULL;
581
582 /*
583 * Stack depot may be used from instrumentation that instruments RCU or
584 * tracing itself; use variant that does not call into RCU and cannot be
585 * traced.
586 *
587 * Note: Such use cases must take care when using refcounting to evict
588 * unused entries, because the stack record free-then-reuse code paths
589 * do call into RCU.
590 */
591 rcu_read_lock_sched_notrace();
592
593 list_for_each_entry_rcu(stack, bucket, hash_list) {
594 if (stack->hash != hash || stack->size != size)
595 continue;
596
597 /*
598 * This may race with depot_free_stack() accessing the freelist
599 * management state unioned with @entries. The refcount is zero
600 * in that case and the below refcount_inc_not_zero() will fail.
601 */
602 if (data_race(stackdepot_memcmp(entries, stack->entries, size)))
603 continue;
604
605 /*
606 * Try to increment refcount. If this succeeds, the stack record
607 * is valid and has not yet been freed.
608 *
609 * If STACK_DEPOT_FLAG_GET is not used, it is undefined behavior
610 * to then call stack_depot_put() later, and we can assume that
611 * a stack record is never placed back on the freelist.
612 */
613 if ((flags & STACK_DEPOT_FLAG_GET) && !refcount_inc_not_zero(&stack->count))
614 continue;
615
616 ret = stack;
617 break;
618 }
619
620 rcu_read_unlock_sched_notrace();
621
622 return ret;
623}
624
625depot_stack_handle_t stack_depot_save_flags(unsigned long *entries,
626 unsigned int nr_entries,
627 gfp_t alloc_flags,
628 depot_flags_t depot_flags)
629{
630 struct list_head *bucket;
631 struct stack_record *found = NULL;
632 depot_stack_handle_t handle = 0;
633 struct page *page = NULL;
634 void *prealloc = NULL;
635 bool can_alloc = depot_flags & STACK_DEPOT_FLAG_CAN_ALLOC;
636 unsigned long flags;
637 u32 hash;
638
639 if (WARN_ON(depot_flags & ~STACK_DEPOT_FLAGS_MASK))
640 return 0;
641
642 /*
643 * If this stack trace is from an interrupt, including anything before
644 * interrupt entry usually leads to unbounded stack depot growth.
645 *
646 * Since use of filter_irq_stacks() is a requirement to ensure stack
647 * depot can efficiently deduplicate interrupt stacks, always
648 * filter_irq_stacks() to simplify all callers' use of stack depot.
649 */
650 nr_entries = filter_irq_stacks(entries, nr_entries);
651
652 if (unlikely(nr_entries == 0) || stack_depot_disabled)
653 return 0;
654
655 hash = hash_stack(entries, nr_entries);
656 bucket = &stack_table[hash & stack_hash_mask];
657
658 /* Fast path: look the stack trace up without locking. */
659 found = find_stack(bucket, entries, nr_entries, hash, depot_flags);
660 if (found)
661 goto exit;
662
663 /*
664 * Allocate memory for a new pool if required now:
665 * we won't be able to do that under the lock.
666 */
667 if (unlikely(can_alloc && !READ_ONCE(new_pool))) {
668 /*
669 * Zero out zone modifiers, as we don't have specific zone
670 * requirements. Keep the flags related to allocation in atomic
671 * contexts and I/O.
672 */
673 alloc_flags &= ~GFP_ZONEMASK;
674 alloc_flags &= (GFP_ATOMIC | GFP_KERNEL);
675 alloc_flags |= __GFP_NOWARN;
676 page = alloc_pages(alloc_flags, DEPOT_POOL_ORDER);
677 if (page)
678 prealloc = page_address(page);
679 }
680
681 raw_spin_lock_irqsave(&pool_lock, flags);
682 printk_deferred_enter();
683
684 /* Try to find again, to avoid concurrently inserting duplicates. */
685 found = find_stack(bucket, entries, nr_entries, hash, depot_flags);
686 if (!found) {
687 struct stack_record *new =
688 depot_alloc_stack(entries, nr_entries, hash, depot_flags, &prealloc);
689
690 if (new) {
691 /*
692 * This releases the stack record into the bucket and
693 * makes it visible to readers in find_stack().
694 */
695 list_add_rcu(&new->hash_list, bucket);
696 found = new;
697 }
698 }
699
700 if (prealloc) {
701 /*
702 * Either stack depot already contains this stack trace, or
703 * depot_alloc_stack() did not consume the preallocated memory.
704 * Try to keep the preallocated memory for future.
705 */
706 depot_keep_new_pool(&prealloc);
707 }
708
709 printk_deferred_exit();
710 raw_spin_unlock_irqrestore(&pool_lock, flags);
711exit:
712 if (prealloc) {
713 /* Stack depot didn't use this memory, free it. */
714 free_pages((unsigned long)prealloc, DEPOT_POOL_ORDER);
715 }
716 if (found)
717 handle = found->handle.handle;
718 return handle;
719}
720EXPORT_SYMBOL_GPL(stack_depot_save_flags);
721
722depot_stack_handle_t stack_depot_save(unsigned long *entries,
723 unsigned int nr_entries,
724 gfp_t alloc_flags)
725{
726 return stack_depot_save_flags(entries, nr_entries, alloc_flags,
727 STACK_DEPOT_FLAG_CAN_ALLOC);
728}
729EXPORT_SYMBOL_GPL(stack_depot_save);
730
731unsigned int stack_depot_fetch(depot_stack_handle_t handle,
732 unsigned long **entries)
733{
734 struct stack_record *stack;
735
736 *entries = NULL;
737 /*
738 * Let KMSAN know *entries is initialized. This shall prevent false
739 * positive reports if instrumented code accesses it.
740 */
741 kmsan_unpoison_memory(entries, sizeof(*entries));
742
743 if (!handle || stack_depot_disabled)
744 return 0;
745
746 stack = depot_fetch_stack(handle);
747 /*
748 * Should never be NULL, otherwise this is a use-after-put (or just a
749 * corrupt handle).
750 */
751 if (WARN(!stack, "corrupt handle or use after stack_depot_put()"))
752 return 0;
753
754 *entries = stack->entries;
755 return stack->size;
756}
757EXPORT_SYMBOL_GPL(stack_depot_fetch);
758
759void stack_depot_put(depot_stack_handle_t handle)
760{
761 struct stack_record *stack;
762
763 if (!handle || stack_depot_disabled)
764 return;
765
766 stack = depot_fetch_stack(handle);
767 /*
768 * Should always be able to find the stack record, otherwise this is an
769 * unbalanced put attempt (or corrupt handle).
770 */
771 if (WARN(!stack, "corrupt handle or unbalanced stack_depot_put()"))
772 return;
773
774 if (refcount_dec_and_test(&stack->count))
775 depot_free_stack(stack);
776}
777EXPORT_SYMBOL_GPL(stack_depot_put);
778
779void stack_depot_print(depot_stack_handle_t stack)
780{
781 unsigned long *entries;
782 unsigned int nr_entries;
783
784 nr_entries = stack_depot_fetch(stack, &entries);
785 if (nr_entries > 0)
786 stack_trace_print(entries, nr_entries, 0);
787}
788EXPORT_SYMBOL_GPL(stack_depot_print);
789
790int stack_depot_snprint(depot_stack_handle_t handle, char *buf, size_t size,
791 int spaces)
792{
793 unsigned long *entries;
794 unsigned int nr_entries;
795
796 nr_entries = stack_depot_fetch(handle, &entries);
797 return nr_entries ? stack_trace_snprint(buf, size, entries, nr_entries,
798 spaces) : 0;
799}
800EXPORT_SYMBOL_GPL(stack_depot_snprint);
801
802depot_stack_handle_t __must_check stack_depot_set_extra_bits(
803 depot_stack_handle_t handle, unsigned int extra_bits)
804{
805 union handle_parts parts = { .handle = handle };
806
807 /* Don't set extra bits on empty handles. */
808 if (!handle)
809 return 0;
810
811 parts.extra = extra_bits;
812 return parts.handle;
813}
814EXPORT_SYMBOL(stack_depot_set_extra_bits);
815
816unsigned int stack_depot_get_extra_bits(depot_stack_handle_t handle)
817{
818 union handle_parts parts = { .handle = handle };
819
820 return parts.extra;
821}
822EXPORT_SYMBOL(stack_depot_get_extra_bits);
823
824static int stats_show(struct seq_file *seq, void *v)
825{
826 /*
827 * data race ok: These are just statistics counters, and approximate
828 * statistics are ok for debugging.
829 */
830 seq_printf(seq, "pools: %d\n", data_race(pools_num));
831 for (int i = 0; i < DEPOT_COUNTER_COUNT; i++)
832 seq_printf(seq, "%s: %ld\n", counter_names[i], data_race(counters[i]));
833
834 return 0;
835}
836DEFINE_SHOW_ATTRIBUTE(stats);
837
838static int depot_debugfs_init(void)
839{
840 struct dentry *dir;
841
842 if (stack_depot_disabled)
843 return 0;
844
845 dir = debugfs_create_dir("stackdepot", NULL);
846 debugfs_create_file("stats", 0444, dir, NULL, &stats_fops);
847 return 0;
848}
849late_initcall(depot_debugfs_init);