1 Runtime locking correctness validator 1 Runtime locking correctness validator
2 ===================================== 2 =====================================
3 3
4 started by Ingo Molnar <mingo@redhat.com> 4 started by Ingo Molnar <mingo@redhat.com>
5 5
6 additions by Arjan van de Ven <arjan@linux.inte 6 additions by Arjan van de Ven <arjan@linux.intel.com>
7 7
8 Lock-class 8 Lock-class
9 ---------- 9 ----------
10 10
11 The basic object the validator operates upon i 11 The basic object the validator operates upon is a 'class' of locks.
12 12
13 A class of locks is a group of locks that are 13 A class of locks is a group of locks that are logically the same with
14 respect to locking rules, even if the locks ma 14 respect to locking rules, even if the locks may have multiple (possibly
15 tens of thousands of) instantiations. For exam 15 tens of thousands of) instantiations. For example a lock in the inode
16 struct is one class, while each inode has its 16 struct is one class, while each inode has its own instantiation of that
17 lock class. 17 lock class.
18 18
19 The validator tracks the 'usage state' of lock 19 The validator tracks the 'usage state' of lock-classes, and it tracks
20 the dependencies between different lock-classe 20 the dependencies between different lock-classes. Lock usage indicates
21 how a lock is used with regard to its IRQ cont 21 how a lock is used with regard to its IRQ contexts, while lock
22 dependency can be understood as lock order, wh 22 dependency can be understood as lock order, where L1 -> L2 suggests that
23 a task is attempting to acquire L2 while holdi 23 a task is attempting to acquire L2 while holding L1. From lockdep's
24 perspective, the two locks (L1 and L2) are not 24 perspective, the two locks (L1 and L2) are not necessarily related; that
25 dependency just means the order ever happened. 25 dependency just means the order ever happened. The validator maintains a
26 continuing effort to prove lock usages and dep 26 continuing effort to prove lock usages and dependencies are correct or
27 the validator will shoot a splat if incorrect. 27 the validator will shoot a splat if incorrect.
28 28
29 A lock-class's behavior is constructed by its 29 A lock-class's behavior is constructed by its instances collectively:
30 when the first instance of a lock-class is use 30 when the first instance of a lock-class is used after bootup the class
31 gets registered, then all (subsequent) instanc 31 gets registered, then all (subsequent) instances will be mapped to the
32 class and hence their usages and dependencies !! 32 class and hence their usages and dependecies will contribute to those of
33 the class. A lock-class does not go away when 33 the class. A lock-class does not go away when a lock instance does, but
34 it can be removed if the memory space of the l 34 it can be removed if the memory space of the lock class (static or
35 dynamic) is reclaimed, this happens for exampl 35 dynamic) is reclaimed, this happens for example when a module is
36 unloaded or a workqueue is destroyed. 36 unloaded or a workqueue is destroyed.
37 37
38 State 38 State
39 ----- 39 -----
40 40
41 The validator tracks lock-class usage history 41 The validator tracks lock-class usage history and divides the usage into
42 (4 usages * n STATEs + 1) categories: 42 (4 usages * n STATEs + 1) categories:
43 43
44 where the 4 usages can be: 44 where the 4 usages can be:
45 45
46 - 'ever held in STATE context' 46 - 'ever held in STATE context'
47 - 'ever held as readlock in STATE context' 47 - 'ever held as readlock in STATE context'
48 - 'ever held with STATE enabled' 48 - 'ever held with STATE enabled'
49 - 'ever held as readlock with STATE enabled' 49 - 'ever held as readlock with STATE enabled'
50 50
51 where the n STATEs are coded in kernel/locking 51 where the n STATEs are coded in kernel/locking/lockdep_states.h and as of
52 now they include: 52 now they include:
53 53
54 - hardirq 54 - hardirq
55 - softirq 55 - softirq
56 56
57 where the last 1 category is: 57 where the last 1 category is:
58 58
59 - 'ever used' 59 - 'ever used' [ == !unused ]
60 60
61 When locking rules are violated, these usage b 61 When locking rules are violated, these usage bits are presented in the
62 locking error messages, inside curlies, with a 62 locking error messages, inside curlies, with a total of 2 * n STATEs bits.
63 A contrived example:: 63 A contrived example::
64 64
65 modprobe/2287 is trying to acquire lock: 65 modprobe/2287 is trying to acquire lock:
66 (&sio_locks[i].lock){-.-.}, at: [<c02867fd 66 (&sio_locks[i].lock){-.-.}, at: [<c02867fd>] mutex_lock+0x21/0x24
67 67
68 but task is already holding lock: 68 but task is already holding lock:
69 (&sio_locks[i].lock){-.-.}, at: [<c02867fd 69 (&sio_locks[i].lock){-.-.}, at: [<c02867fd>] mutex_lock+0x21/0x24
70 70
71 71
72 For a given lock, the bit positions from left 72 For a given lock, the bit positions from left to right indicate the usage
73 of the lock and readlock (if exists), for each 73 of the lock and readlock (if exists), for each of the n STATEs listed
74 above respectively, and the character displaye 74 above respectively, and the character displayed at each bit position
75 indicates: 75 indicates:
76 76
77 === ====================================== 77 === ===================================================
78 '.' acquired while irqs disabled and not i 78 '.' acquired while irqs disabled and not in irq context
79 '-' acquired in irq context 79 '-' acquired in irq context
80 '+' acquired with irqs enabled 80 '+' acquired with irqs enabled
81 '?' acquired in irq context with irqs enab 81 '?' acquired in irq context with irqs enabled.
82 === ====================================== 82 === ===================================================
83 83
84 The bits are illustrated with an example:: 84 The bits are illustrated with an example::
85 85
86 (&sio_locks[i].lock){-.-.}, at: [<c02867fd 86 (&sio_locks[i].lock){-.-.}, at: [<c02867fd>] mutex_lock+0x21/0x24
87 |||| 87 ||||
88 ||| \-> softirq disab 88 ||| \-> softirq disabled and not in softirq context
89 || \--> acquired in s 89 || \--> acquired in softirq context
90 | \---> hardirq disab 90 | \---> hardirq disabled and not in hardirq context
91 \----> acquired in h 91 \----> acquired in hardirq context
92 92
93 93
94 For a given STATE, whether the lock is ever ac 94 For a given STATE, whether the lock is ever acquired in that STATE
95 context and whether that STATE is enabled yiel 95 context and whether that STATE is enabled yields four possible cases as
96 shown in the table below. The bit character is 96 shown in the table below. The bit character is able to indicate which
97 exact case is for the lock as of the reporting 97 exact case is for the lock as of the reporting time.
98 98
99 +--------------+-------------+-------------- 99 +--------------+-------------+--------------+
100 | | irq enabled | irq disabled 100 | | irq enabled | irq disabled |
101 +--------------+-------------+-------------- 101 +--------------+-------------+--------------+
102 | ever in irq | '?' | '-' 102 | ever in irq | '?' | '-' |
103 +--------------+-------------+-------------- 103 +--------------+-------------+--------------+
104 | never in irq | '+' | '.' 104 | never in irq | '+' | '.' |
105 +--------------+-------------+-------------- 105 +--------------+-------------+--------------+
106 106
107 The character '-' suggests irq is disabled bec 107 The character '-' suggests irq is disabled because if otherwise the
108 character '?' would have been shown instead. S !! 108 charactor '?' would have been shown instead. Similar deduction can be
109 applied for '+' too. 109 applied for '+' too.
110 110
111 Unused locks (e.g., mutexes) cannot be part of 111 Unused locks (e.g., mutexes) cannot be part of the cause of an error.
112 112
113 113
114 Single-lock state rules: 114 Single-lock state rules:
115 ------------------------ 115 ------------------------
116 116
117 A lock is irq-safe means it was ever used in a 117 A lock is irq-safe means it was ever used in an irq context, while a lock
118 is irq-unsafe means it was ever acquired with 118 is irq-unsafe means it was ever acquired with irq enabled.
119 119
120 A softirq-unsafe lock-class is automatically h 120 A softirq-unsafe lock-class is automatically hardirq-unsafe as well. The
121 following states must be exclusive: only one o 121 following states must be exclusive: only one of them is allowed to be set
122 for any lock-class based on its usage:: 122 for any lock-class based on its usage::
123 123
124 <hardirq-safe> or <hardirq-unsafe> 124 <hardirq-safe> or <hardirq-unsafe>
125 <softirq-safe> or <softirq-unsafe> 125 <softirq-safe> or <softirq-unsafe>
126 126
127 This is because if a lock can be used in irq c 127 This is because if a lock can be used in irq context (irq-safe) then it
128 cannot be ever acquired with irq enabled (irq- 128 cannot be ever acquired with irq enabled (irq-unsafe). Otherwise, a
129 deadlock may happen. For example, in the scena 129 deadlock may happen. For example, in the scenario that after this lock
130 was acquired but before released, if the conte 130 was acquired but before released, if the context is interrupted this
131 lock will be attempted to acquire twice, which 131 lock will be attempted to acquire twice, which creates a deadlock,
132 referred to as lock recursion deadlock. 132 referred to as lock recursion deadlock.
133 133
134 The validator detects and reports lock usage t 134 The validator detects and reports lock usage that violates these
135 single-lock state rules. 135 single-lock state rules.
136 136
137 Multi-lock dependency rules: 137 Multi-lock dependency rules:
138 ---------------------------- 138 ----------------------------
139 139
140 The same lock-class must not be acquired twice 140 The same lock-class must not be acquired twice, because this could lead
141 to lock recursion deadlocks. 141 to lock recursion deadlocks.
142 142
143 Furthermore, two locks can not be taken in inv 143 Furthermore, two locks can not be taken in inverse order::
144 144
145 <L1> -> <L2> 145 <L1> -> <L2>
146 <L2> -> <L1> 146 <L2> -> <L1>
147 147
148 because this could lead to a deadlock - referr 148 because this could lead to a deadlock - referred to as lock inversion
149 deadlock - as attempts to acquire the two lock 149 deadlock - as attempts to acquire the two locks form a circle which
150 could lead to the two contexts waiting for eac 150 could lead to the two contexts waiting for each other permanently. The
151 validator will find such dependency circle in 151 validator will find such dependency circle in arbitrary complexity,
152 i.e., there can be any other locking sequence 152 i.e., there can be any other locking sequence between the acquire-lock
153 operations; the validator will still find whet 153 operations; the validator will still find whether these locks can be
154 acquired in a circular fashion. 154 acquired in a circular fashion.
155 155
156 Furthermore, the following usage based lock de 156 Furthermore, the following usage based lock dependencies are not allowed
157 between any two lock-classes:: 157 between any two lock-classes::
158 158
159 <hardirq-safe> -> <hardirq-unsafe> 159 <hardirq-safe> -> <hardirq-unsafe>
160 <softirq-safe> -> <softirq-unsafe> 160 <softirq-safe> -> <softirq-unsafe>
161 161
162 The first rule comes from the fact that a hard 162 The first rule comes from the fact that a hardirq-safe lock could be
163 taken by a hardirq context, interrupting a har 163 taken by a hardirq context, interrupting a hardirq-unsafe lock - and
164 thus could result in a lock inversion deadlock 164 thus could result in a lock inversion deadlock. Likewise, a softirq-safe
165 lock could be taken by an softirq context, int 165 lock could be taken by an softirq context, interrupting a softirq-unsafe
166 lock. 166 lock.
167 167
168 The above rules are enforced for any locking s 168 The above rules are enforced for any locking sequence that occurs in the
169 kernel: when acquiring a new lock, the validat 169 kernel: when acquiring a new lock, the validator checks whether there is
170 any rule violation between the new lock and an 170 any rule violation between the new lock and any of the held locks.
171 171
172 When a lock-class changes its state, the follo 172 When a lock-class changes its state, the following aspects of the above
173 dependency rules are enforced: 173 dependency rules are enforced:
174 174
175 - if a new hardirq-safe lock is discovered, we 175 - if a new hardirq-safe lock is discovered, we check whether it
176 took any hardirq-unsafe lock in the past. 176 took any hardirq-unsafe lock in the past.
177 177
178 - if a new softirq-safe lock is discovered, we 178 - if a new softirq-safe lock is discovered, we check whether it took
179 any softirq-unsafe lock in the past. 179 any softirq-unsafe lock in the past.
180 180
181 - if a new hardirq-unsafe lock is discovered, 181 - if a new hardirq-unsafe lock is discovered, we check whether any
182 hardirq-safe lock took it in the past. 182 hardirq-safe lock took it in the past.
183 183
184 - if a new softirq-unsafe lock is discovered, 184 - if a new softirq-unsafe lock is discovered, we check whether any
185 softirq-safe lock took it in the past. 185 softirq-safe lock took it in the past.
186 186
187 (Again, we do these checks too on the basis th 187 (Again, we do these checks too on the basis that an interrupt context
188 could interrupt _any_ of the irq-unsafe or har 188 could interrupt _any_ of the irq-unsafe or hardirq-unsafe locks, which
189 could lead to a lock inversion deadlock - even 189 could lead to a lock inversion deadlock - even if that lock scenario did
190 not trigger in practice yet.) 190 not trigger in practice yet.)
191 191
192 Exception: Nested data dependencies leading to 192 Exception: Nested data dependencies leading to nested locking
193 ---------------------------------------------- 193 -------------------------------------------------------------
194 194
195 There are a few cases where the Linux kernel a 195 There are a few cases where the Linux kernel acquires more than one
196 instance of the same lock-class. Such cases ty 196 instance of the same lock-class. Such cases typically happen when there
197 is some sort of hierarchy within objects of th 197 is some sort of hierarchy within objects of the same type. In these
198 cases there is an inherent "natural" ordering 198 cases there is an inherent "natural" ordering between the two objects
199 (defined by the properties of the hierarchy), 199 (defined by the properties of the hierarchy), and the kernel grabs the
200 locks in this fixed order on each of the objec 200 locks in this fixed order on each of the objects.
201 201
202 An example of such an object hierarchy that re 202 An example of such an object hierarchy that results in "nested locking"
203 is that of a "whole disk" block-dev object and 203 is that of a "whole disk" block-dev object and a "partition" block-dev
204 object; the partition is "part of" the whole d 204 object; the partition is "part of" the whole device and as long as one
205 always takes the whole disk lock as a higher l 205 always takes the whole disk lock as a higher lock than the partition
206 lock, the lock ordering is fully correct. The 206 lock, the lock ordering is fully correct. The validator does not
207 automatically detect this natural ordering, as 207 automatically detect this natural ordering, as the locking rule behind
208 the ordering is not static. 208 the ordering is not static.
209 209
210 In order to teach the validator about this cor 210 In order to teach the validator about this correct usage model, new
211 versions of the various locking primitives wer 211 versions of the various locking primitives were added that allow you to
212 specify a "nesting level". An example call, fo 212 specify a "nesting level". An example call, for the block device mutex,
213 looks like this:: 213 looks like this::
214 214
215 enum bdev_bd_mutex_lock_class 215 enum bdev_bd_mutex_lock_class
216 { 216 {
217 BD_MUTEX_NORMAL, 217 BD_MUTEX_NORMAL,
218 BD_MUTEX_WHOLE, 218 BD_MUTEX_WHOLE,
219 BD_MUTEX_PARTITION 219 BD_MUTEX_PARTITION
220 }; 220 };
221 221
222 mutex_lock_nested(&bdev->bd_contains->bd_mut 222 mutex_lock_nested(&bdev->bd_contains->bd_mutex, BD_MUTEX_PARTITION);
223 223
224 In this case the locking is done on a bdev obj 224 In this case the locking is done on a bdev object that is known to be a
225 partition. 225 partition.
226 226
227 The validator treats a lock that is taken in s 227 The validator treats a lock that is taken in such a nested fashion as a
228 separate (sub)class for the purposes of valida 228 separate (sub)class for the purposes of validation.
229 229
230 Note: When changing code to use the _nested() 230 Note: When changing code to use the _nested() primitives, be careful and
231 check really thoroughly that the hierarchy is 231 check really thoroughly that the hierarchy is correctly mapped; otherwise
232 you can get false positives or false negatives 232 you can get false positives or false negatives.
233 233
234 Annotations 234 Annotations
235 ----------- 235 -----------
236 236
237 Two constructs can be used to annotate and che 237 Two constructs can be used to annotate and check where and if certain locks
238 must be held: lockdep_assert_held*(&lock) and 238 must be held: lockdep_assert_held*(&lock) and lockdep_*pin_lock(&lock).
239 239
240 As the name suggests, lockdep_assert_held* fam 240 As the name suggests, lockdep_assert_held* family of macros assert that a
241 particular lock is held at a certain time (and 241 particular lock is held at a certain time (and generate a WARN() otherwise).
242 This annotation is largely used all over the k 242 This annotation is largely used all over the kernel, e.g. kernel/sched/
243 core.c:: 243 core.c::
244 244
245 void update_rq_clock(struct rq *rq) 245 void update_rq_clock(struct rq *rq)
246 { 246 {
247 s64 delta; 247 s64 delta;
248 248
249 lockdep_assert_held(&rq->lock); 249 lockdep_assert_held(&rq->lock);
250 [...] 250 [...]
251 } 251 }
252 252
253 where holding rq->lock is required to safely u 253 where holding rq->lock is required to safely update a rq's clock.
254 254
255 The other family of macros is lockdep_*pin_loc 255 The other family of macros is lockdep_*pin_lock(), which is admittedly only
256 used for rq->lock ATM. Despite their limited a 256 used for rq->lock ATM. Despite their limited adoption these annotations
257 generate a WARN() if the lock of interest is " 257 generate a WARN() if the lock of interest is "accidentally" unlocked. This turns
258 out to be especially helpful to debug code wit 258 out to be especially helpful to debug code with callbacks, where an upper
259 layer assumes a lock remains taken, but a lowe 259 layer assumes a lock remains taken, but a lower layer thinks it can maybe drop
260 and reacquire the lock ("unwittingly" introduc 260 and reacquire the lock ("unwittingly" introducing races). lockdep_pin_lock()
261 returns a 'struct pin_cookie' that is then use 261 returns a 'struct pin_cookie' that is then used by lockdep_unpin_lock() to check
262 that nobody tampered with the lock, e.g. kerne 262 that nobody tampered with the lock, e.g. kernel/sched/sched.h::
263 263
264 static inline void rq_pin_lock(struct rq *rq 264 static inline void rq_pin_lock(struct rq *rq, struct rq_flags *rf)
265 { 265 {
266 rf->cookie = lockdep_pin_lock(&rq->loc 266 rf->cookie = lockdep_pin_lock(&rq->lock);
267 [...] 267 [...]
268 } 268 }
269 269
270 static inline void rq_unpin_lock(struct rq * 270 static inline void rq_unpin_lock(struct rq *rq, struct rq_flags *rf)
271 { 271 {
272 [...] 272 [...]
273 lockdep_unpin_lock(&rq->lock, rf->cook 273 lockdep_unpin_lock(&rq->lock, rf->cookie);
274 } 274 }
275 275
276 While comments about locking requirements migh 276 While comments about locking requirements might provide useful information,
277 the runtime checks performed by annotations ar 277 the runtime checks performed by annotations are invaluable when debugging
278 locking problems and they carry the same level 278 locking problems and they carry the same level of details when inspecting
279 code. Always prefer annotations when in doubt 279 code. Always prefer annotations when in doubt!
280 280
281 Proof of 100% correctness: 281 Proof of 100% correctness:
282 -------------------------- 282 --------------------------
283 283
284 The validator achieves perfect, mathematical ' 284 The validator achieves perfect, mathematical 'closure' (proof of locking
285 correctness) in the sense that for every simpl 285 correctness) in the sense that for every simple, standalone single-task
286 locking sequence that occurred at least once d 286 locking sequence that occurred at least once during the lifetime of the
287 kernel, the validator proves it with a 100% ce 287 kernel, the validator proves it with a 100% certainty that no
288 combination and timing of these locking sequen 288 combination and timing of these locking sequences can cause any class of
289 lock related deadlock. [1]_ 289 lock related deadlock. [1]_
290 290
291 I.e. complex multi-CPU and multi-task locking 291 I.e. complex multi-CPU and multi-task locking scenarios do not have to
292 occur in practice to prove a deadlock: only th 292 occur in practice to prove a deadlock: only the simple 'component'
293 locking chains have to occur at least once (an 293 locking chains have to occur at least once (anytime, in any
294 task/context) for the validator to be able to 294 task/context) for the validator to be able to prove correctness. (For
295 example, complex deadlocks that would normally 295 example, complex deadlocks that would normally need more than 3 CPUs and
296 a very unlikely constellation of tasks, irq-co 296 a very unlikely constellation of tasks, irq-contexts and timings to
297 occur, can be detected on a plain, lightly loa 297 occur, can be detected on a plain, lightly loaded single-CPU system as
298 well!) 298 well!)
299 299
300 This radically decreases the complexity of loc 300 This radically decreases the complexity of locking related QA of the
301 kernel: what has to be done during QA is to tr 301 kernel: what has to be done during QA is to trigger as many "simple"
302 single-task locking dependencies in the kernel 302 single-task locking dependencies in the kernel as possible, at least
303 once, to prove locking correctness - instead o 303 once, to prove locking correctness - instead of having to trigger every
304 possible combination of locking interaction be 304 possible combination of locking interaction between CPUs, combined with
305 every possible hardirq and softirq nesting sce 305 every possible hardirq and softirq nesting scenario (which is impossible
306 to do in practice). 306 to do in practice).
307 307
308 .. [1] 308 .. [1]
309 309
310 assuming that the validator itself is 100% 310 assuming that the validator itself is 100% correct, and no other
311 part of the system corrupts the state of t 311 part of the system corrupts the state of the validator in any way.
312 We also assume that all NMI/SMM paths [whi 312 We also assume that all NMI/SMM paths [which could interrupt
313 even hardirq-disabled codepaths] are corre 313 even hardirq-disabled codepaths] are correct and do not interfere
314 with the validator. We also assume that th 314 with the validator. We also assume that the 64-bit 'chain hash'
315 value is unique for every lock-chain in th 315 value is unique for every lock-chain in the system. Also, lock
316 recursion must not be higher than 20. 316 recursion must not be higher than 20.
317 317
318 Performance: 318 Performance:
319 ------------ 319 ------------
320 320
321 The above rules require **massive** amounts of 321 The above rules require **massive** amounts of runtime checking. If we did
322 that for every lock taken and for every irqs-e 322 that for every lock taken and for every irqs-enable event, it would
323 render the system practically unusably slow. T 323 render the system practically unusably slow. The complexity of checking
324 is O(N^2), so even with just a few hundred loc 324 is O(N^2), so even with just a few hundred lock-classes we'd have to do
325 tens of thousands of checks for every event. 325 tens of thousands of checks for every event.
326 326
327 This problem is solved by checking any given ' 327 This problem is solved by checking any given 'locking scenario' (unique
328 sequence of locks taken after each other) only 328 sequence of locks taken after each other) only once. A simple stack of
329 held locks is maintained, and a lightweight 64 329 held locks is maintained, and a lightweight 64-bit hash value is
330 calculated, which hash is unique for every loc 330 calculated, which hash is unique for every lock chain. The hash value,
331 when the chain is validated for the first time 331 when the chain is validated for the first time, is then put into a hash
332 table, which hash-table can be checked in a lo 332 table, which hash-table can be checked in a lockfree manner. If the
333 locking chain occurs again later on, the hash 333 locking chain occurs again later on, the hash table tells us that we
334 don't have to validate the chain again. 334 don't have to validate the chain again.
335 335
336 Troubleshooting: 336 Troubleshooting:
337 ---------------- 337 ----------------
338 338
339 The validator tracks a maximum of MAX_LOCKDEP_ 339 The validator tracks a maximum of MAX_LOCKDEP_KEYS number of lock classes.
340 Exceeding this number will trigger the followi 340 Exceeding this number will trigger the following lockdep warning::
341 341
342 (DEBUG_LOCKS_WARN_ON(id >= MAX_LOCKDEP 342 (DEBUG_LOCKS_WARN_ON(id >= MAX_LOCKDEP_KEYS))
343 343
344 By default, MAX_LOCKDEP_KEYS is currently set 344 By default, MAX_LOCKDEP_KEYS is currently set to 8191, and typical
345 desktop systems have less than 1,000 lock clas 345 desktop systems have less than 1,000 lock classes, so this warning
346 normally results from lock-class leakage or fa 346 normally results from lock-class leakage or failure to properly
347 initialize locks. These two problems are illu 347 initialize locks. These two problems are illustrated below:
348 348
349 1. Repeated module loading and unloading 349 1. Repeated module loading and unloading while running the validator
350 will result in lock-class leakage. Th 350 will result in lock-class leakage. The issue here is that each
351 load of the module will create a new s 351 load of the module will create a new set of lock classes for
352 that module's locks, but module unload 352 that module's locks, but module unloading does not remove old
353 classes (see below discussion of reuse 353 classes (see below discussion of reuse of lock classes for why).
354 Therefore, if that module is loaded an 354 Therefore, if that module is loaded and unloaded repeatedly,
355 the number of lock classes will eventu 355 the number of lock classes will eventually reach the maximum.
356 356
357 2. Using structures such as arrays that h 357 2. Using structures such as arrays that have large numbers of
358 locks that are not explicitly initiali 358 locks that are not explicitly initialized. For example,
359 a hash table with 8192 buckets where e 359 a hash table with 8192 buckets where each bucket has its own
360 spinlock_t will consume 8192 lock clas 360 spinlock_t will consume 8192 lock classes -unless- each spinlock
361 is explicitly initialized at runtime, 361 is explicitly initialized at runtime, for example, using the
362 run-time spin_lock_init() as opposed t 362 run-time spin_lock_init() as opposed to compile-time initializers
363 such as __SPIN_LOCK_UNLOCKED(). Failu 363 such as __SPIN_LOCK_UNLOCKED(). Failure to properly initialize
364 the per-bucket spinlocks would guarant 364 the per-bucket spinlocks would guarantee lock-class overflow.
365 In contrast, a loop that called spin_l 365 In contrast, a loop that called spin_lock_init() on each lock
366 would place all 8192 locks into a sing 366 would place all 8192 locks into a single lock class.
367 367
368 The moral of this story is that you sh 368 The moral of this story is that you should always explicitly
369 initialize your locks. 369 initialize your locks.
370 370
371 One might argue that the validator should be m 371 One might argue that the validator should be modified to allow
372 lock classes to be reused. However, if you ar 372 lock classes to be reused. However, if you are tempted to make this
373 argument, first review the code and think thro 373 argument, first review the code and think through the changes that would
374 be required, keeping in mind that the lock cla 374 be required, keeping in mind that the lock classes to be removed are
375 likely to be linked into the lock-dependency g 375 likely to be linked into the lock-dependency graph. This turns out to
376 be harder to do than to say. 376 be harder to do than to say.
377 377
378 Of course, if you do run out of lock classes, 378 Of course, if you do run out of lock classes, the next thing to do is
379 to find the offending lock classes. First, th 379 to find the offending lock classes. First, the following command gives
380 you the number of lock classes currently in us 380 you the number of lock classes currently in use along with the maximum::
381 381
382 grep "lock-classes" /proc/lockdep_stat 382 grep "lock-classes" /proc/lockdep_stats
383 383
384 This command produces the following output on 384 This command produces the following output on a modest system::
385 385
386 lock-classes: 386 lock-classes: 748 [max: 8191]
387 387
388 If the number allocated (748 above) increases 388 If the number allocated (748 above) increases continually over time,
389 then there is likely a leak. The following co 389 then there is likely a leak. The following command can be used to
390 identify the leaking lock classes:: 390 identify the leaking lock classes::
391 391
392 grep "BD" /proc/lockdep 392 grep "BD" /proc/lockdep
393 393
394 Run the command and save the output, then comp 394 Run the command and save the output, then compare against the output from
395 a later run of this command to identify the le 395 a later run of this command to identify the leakers. This same output
396 can also help you find situations where runtim 396 can also help you find situations where runtime lock initialization has
397 been omitted. 397 been omitted.
398 398
399 Recursive read locks: 399 Recursive read locks:
400 --------------------- 400 ---------------------
401 The whole of the rest document tries to prove 401 The whole of the rest document tries to prove a certain type of cycle is equivalent
402 to deadlock possibility. 402 to deadlock possibility.
403 403
404 There are three types of lockers: writers (i.e 404 There are three types of lockers: writers (i.e. exclusive lockers, like
405 spin_lock() or write_lock()), non-recursive re 405 spin_lock() or write_lock()), non-recursive readers (i.e. shared lockers, like
406 down_read()) and recursive readers (recursive 406 down_read()) and recursive readers (recursive shared lockers, like rcu_read_lock()).
407 And we use the following notations of those lo 407 And we use the following notations of those lockers in the rest of the document:
408 408
409 W or E: stands for writers (exclusive 409 W or E: stands for writers (exclusive lockers).
410 r: stands for non-recursive reade 410 r: stands for non-recursive readers.
411 R: stands for recursive readers. 411 R: stands for recursive readers.
412 S: stands for all readers (non-re 412 S: stands for all readers (non-recursive + recursive), as both are shared lockers.
413 N: stands for writers and non-rec 413 N: stands for writers and non-recursive readers, as both are not recursive.
414 414
415 Obviously, N is "r or W" and S is "r or R". 415 Obviously, N is "r or W" and S is "r or R".
416 416
417 Recursive readers, as their name indicates, ar 417 Recursive readers, as their name indicates, are the lockers allowed to acquire
418 even inside the critical section of another re 418 even inside the critical section of another reader of the same lock instance,
419 in other words, allowing nested read-side crit 419 in other words, allowing nested read-side critical sections of one lock instance.
420 420
421 While non-recursive readers will cause a self 421 While non-recursive readers will cause a self deadlock if trying to acquire inside
422 the critical section of another reader of the 422 the critical section of another reader of the same lock instance.
423 423
424 The difference between recursive readers and n 424 The difference between recursive readers and non-recursive readers is because:
425 recursive readers get blocked only by a write 425 recursive readers get blocked only by a write lock *holder*, while non-recursive
426 readers could get blocked by a write lock *wai 426 readers could get blocked by a write lock *waiter*. Considering the follow
427 example:: 427 example::
428 428
429 TASK A: TASK B: 429 TASK A: TASK B:
430 430
431 read_lock(X); 431 read_lock(X);
432 write_lock(X); 432 write_lock(X);
433 read_lock_2(X); 433 read_lock_2(X);
434 434
435 Task A gets the reader (no matter whether recu 435 Task A gets the reader (no matter whether recursive or non-recursive) on X via
436 read_lock() first. And when task B tries to ac 436 read_lock() first. And when task B tries to acquire writer on X, it will block
437 and become a waiter for writer on X. Now if re 437 and become a waiter for writer on X. Now if read_lock_2() is recursive readers,
438 task A will make progress, because writer wait 438 task A will make progress, because writer waiters don't block recursive readers,
439 and there is no deadlock. However, if read_loc 439 and there is no deadlock. However, if read_lock_2() is non-recursive readers,
440 it will get blocked by writer waiter B, and ca 440 it will get blocked by writer waiter B, and cause a self deadlock.
441 441
442 Block conditions on readers/writers of the sam 442 Block conditions on readers/writers of the same lock instance:
443 ---------------------------------------------- 443 --------------------------------------------------------------
444 There are simply four block conditions: 444 There are simply four block conditions:
445 445
446 1. Writers block other writers. 446 1. Writers block other writers.
447 2. Readers block writers. 447 2. Readers block writers.
448 3. Writers block both recursive readers a 448 3. Writers block both recursive readers and non-recursive readers.
449 4. And readers (recursive or not) don't b 449 4. And readers (recursive or not) don't block other recursive readers but
450 may block non-recursive readers (becau 450 may block non-recursive readers (because of the potential co-existing
451 writer waiters) 451 writer waiters)
452 452
453 Block condition matrix, Y means the row blocks 453 Block condition matrix, Y means the row blocks the column, and N means otherwise.
454 454
455 +---+---+---+---+ 455 +---+---+---+---+
456 | | W | r | R | !! 456 | | E | r | R |
457 +---+---+---+---+ 457 +---+---+---+---+
458 | W | Y | Y | Y | !! 458 | E | Y | Y | Y |
459 +---+---+---+---+ 459 +---+---+---+---+
460 | r | Y | Y | N | 460 | r | Y | Y | N |
461 +---+---+---+---+ 461 +---+---+---+---+
462 | R | Y | Y | N | 462 | R | Y | Y | N |
463 +---+---+---+---+ 463 +---+---+---+---+
464 464
465 (W: writers, r: non-recursive readers, 465 (W: writers, r: non-recursive readers, R: recursive readers)
466 466
467 467
468 acquired recursively. Unlike non-recursive rea 468 acquired recursively. Unlike non-recursive read locks, recursive read locks
469 only get blocked by current write lock *holder 469 only get blocked by current write lock *holders* other than write lock
470 *waiters*, for example:: 470 *waiters*, for example::
471 471
472 TASK A: TASK B: 472 TASK A: TASK B:
473 473
474 read_lock(X); 474 read_lock(X);
475 475
476 write_lock(X); 476 write_lock(X);
477 477
478 read_lock(X); 478 read_lock(X);
479 479
480 is not a deadlock for recursive read locks, as 480 is not a deadlock for recursive read locks, as while the task B is waiting for
481 the lock X, the second read_lock() doesn't nee 481 the lock X, the second read_lock() doesn't need to wait because it's a recursive
482 read lock. However if the read_lock() is non-r 482 read lock. However if the read_lock() is non-recursive read lock, then the above
483 case is a deadlock, because even if the write_ 483 case is a deadlock, because even if the write_lock() in TASK B cannot get the
484 lock, but it can block the second read_lock() 484 lock, but it can block the second read_lock() in TASK A.
485 485
486 Note that a lock can be a write lock (exclusiv 486 Note that a lock can be a write lock (exclusive lock), a non-recursive read
487 lock (non-recursive shared lock) or a recursiv 487 lock (non-recursive shared lock) or a recursive read lock (recursive shared
488 lock), depending on the lock operations used t 488 lock), depending on the lock operations used to acquire it (more specifically,
489 the value of the 'read' parameter for lock_acq 489 the value of the 'read' parameter for lock_acquire()). In other words, a single
490 lock instance has three types of acquisition d 490 lock instance has three types of acquisition depending on the acquisition
491 functions: exclusive, non-recursive read, and 491 functions: exclusive, non-recursive read, and recursive read.
492 492
493 To be concise, we call that write locks and no 493 To be concise, we call that write locks and non-recursive read locks as
494 "non-recursive" locks and recursive read locks 494 "non-recursive" locks and recursive read locks as "recursive" locks.
495 495
496 Recursive locks don't block each other, while 496 Recursive locks don't block each other, while non-recursive locks do (this is
497 even true for two non-recursive read locks). A 497 even true for two non-recursive read locks). A non-recursive lock can block the
498 corresponding recursive lock, and vice versa. 498 corresponding recursive lock, and vice versa.
499 499
500 A deadlock case with recursive locks involved 500 A deadlock case with recursive locks involved is as follow::
501 501
502 TASK A: TASK B: 502 TASK A: TASK B:
503 503
504 read_lock(X); 504 read_lock(X);
505 read_lock(Y); 505 read_lock(Y);
506 write_lock(Y); 506 write_lock(Y);
507 write_lock(X); 507 write_lock(X);
508 508
509 Task A is waiting for task B to read_unlock() 509 Task A is waiting for task B to read_unlock() Y and task B is waiting for task
510 A to read_unlock() X. 510 A to read_unlock() X.
511 511
512 Dependency types and strong dependency paths: 512 Dependency types and strong dependency paths:
513 --------------------------------------------- 513 ---------------------------------------------
514 Lock dependencies record the orders of the acq 514 Lock dependencies record the orders of the acquisitions of a pair of locks, and
515 because there are 3 types for lockers, there a 515 because there are 3 types for lockers, there are, in theory, 9 types of lock
516 dependencies, but we can show that 4 types of 516 dependencies, but we can show that 4 types of lock dependencies are enough for
517 deadlock detection. 517 deadlock detection.
518 518
519 For each lock dependency:: 519 For each lock dependency::
520 520
521 L1 -> L2 521 L1 -> L2
522 522
523 , which means lockdep has seen L1 held before 523 , which means lockdep has seen L1 held before L2 held in the same context at runtime.
524 And in deadlock detection, we care whether we 524 And in deadlock detection, we care whether we could get blocked on L2 with L1 held,
525 IOW, whether there is a locker L3 that L1 bloc 525 IOW, whether there is a locker L3 that L1 blocks L3 and L2 gets blocked by L3. So
526 we only care about 1) what L1 blocks and 2) wh 526 we only care about 1) what L1 blocks and 2) what blocks L2. As a result, we can combine
527 recursive readers and non-recursive readers fo 527 recursive readers and non-recursive readers for L1 (as they block the same types) and
528 we can combine writers and non-recursive reade 528 we can combine writers and non-recursive readers for L2 (as they get blocked by the
529 same types). 529 same types).
530 530
531 With the above combination for simplification, 531 With the above combination for simplification, there are 4 types of dependency edges
532 in the lockdep graph: 532 in the lockdep graph:
533 533
534 1) -(ER)->: 534 1) -(ER)->:
535 exclusive writer to recursive read 535 exclusive writer to recursive reader dependency, "X -(ER)-> Y" means
536 X -> Y and X is a writer and Y is 536 X -> Y and X is a writer and Y is a recursive reader.
537 537
538 2) -(EN)->: 538 2) -(EN)->:
539 exclusive writer to non-recursive 539 exclusive writer to non-recursive locker dependency, "X -(EN)-> Y" means
540 X -> Y and X is a writer and Y is 540 X -> Y and X is a writer and Y is either a writer or non-recursive reader.
541 541
542 3) -(SR)->: 542 3) -(SR)->:
543 shared reader to recursive reader 543 shared reader to recursive reader dependency, "X -(SR)-> Y" means
544 X -> Y and X is a reader (recursiv 544 X -> Y and X is a reader (recursive or not) and Y is a recursive reader.
545 545
546 4) -(SN)->: 546 4) -(SN)->:
547 shared reader to non-recursive loc 547 shared reader to non-recursive locker dependency, "X -(SN)-> Y" means
548 X -> Y and X is a reader (recursiv 548 X -> Y and X is a reader (recursive or not) and Y is either a writer or
549 non-recursive reader. 549 non-recursive reader.
550 550
551 Note that given two locks, they may have multi 551 Note that given two locks, they may have multiple dependencies between them,
552 for example:: 552 for example::
553 553
554 TASK A: 554 TASK A:
555 555
556 read_lock(X); 556 read_lock(X);
557 write_lock(Y); 557 write_lock(Y);
558 ... 558 ...
559 559
560 TASK B: 560 TASK B:
561 561
562 write_lock(X); 562 write_lock(X);
563 write_lock(Y); 563 write_lock(Y);
564 564
565 , we have both X -(SN)-> Y and X -(EN)-> Y in 565 , we have both X -(SN)-> Y and X -(EN)-> Y in the dependency graph.
566 566
567 We use -(xN)-> to represent edges that are eit 567 We use -(xN)-> to represent edges that are either -(EN)-> or -(SN)->, the
568 similar for -(Ex)->, -(xR)-> and -(Sx)-> 568 similar for -(Ex)->, -(xR)-> and -(Sx)->
569 569
570 A "path" is a series of conjunct dependency ed 570 A "path" is a series of conjunct dependency edges in the graph. And we define a
571 "strong" path, which indicates the strong depe 571 "strong" path, which indicates the strong dependency throughout each dependency
572 in the path, as the path that doesn't have two 572 in the path, as the path that doesn't have two conjunct edges (dependencies) as
573 -(xR)-> and -(Sx)->. In other words, a "strong 573 -(xR)-> and -(Sx)->. In other words, a "strong" path is a path from a lock
574 walking to another through the lock dependenci 574 walking to another through the lock dependencies, and if X -> Y -> Z is in the
575 path (where X, Y, Z are locks), and the walk f 575 path (where X, Y, Z are locks), and the walk from X to Y is through a -(SR)-> or
576 -(ER)-> dependency, the walk from Y to Z must 576 -(ER)-> dependency, the walk from Y to Z must not be through a -(SN)-> or
577 -(SR)-> dependency. 577 -(SR)-> dependency.
578 578
579 We will see why the path is called "strong" in 579 We will see why the path is called "strong" in next section.
580 580
581 Recursive Read Deadlock Detection: 581 Recursive Read Deadlock Detection:
582 ---------------------------------- 582 ----------------------------------
583 583
584 We now prove two things: 584 We now prove two things:
585 585
586 Lemma 1: 586 Lemma 1:
587 587
588 If there is a closed strong path (i.e. a stron 588 If there is a closed strong path (i.e. a strong circle), then there is a
589 combination of locking sequences that causes d 589 combination of locking sequences that causes deadlock. I.e. a strong circle is
590 sufficient for deadlock detection. 590 sufficient for deadlock detection.
591 591
592 Lemma 2: 592 Lemma 2:
593 593
594 If there is no closed strong path (i.e. strong 594 If there is no closed strong path (i.e. strong circle), then there is no
595 combination of locking sequences that could ca 595 combination of locking sequences that could cause deadlock. I.e. strong
596 circles are necessary for deadlock detection. 596 circles are necessary for deadlock detection.
597 597
598 With these two Lemmas, we can easily say a clo 598 With these two Lemmas, we can easily say a closed strong path is both sufficient
599 and necessary for deadlocks, therefore a close 599 and necessary for deadlocks, therefore a closed strong path is equivalent to
600 deadlock possibility. As a closed strong path 600 deadlock possibility. As a closed strong path stands for a dependency chain that
601 could cause deadlocks, so we call it "strong", 601 could cause deadlocks, so we call it "strong", considering there are dependency
602 circles that won't cause deadlocks. 602 circles that won't cause deadlocks.
603 603
604 Proof for sufficiency (Lemma 1): 604 Proof for sufficiency (Lemma 1):
605 605
606 Let's say we have a strong circle:: 606 Let's say we have a strong circle::
607 607
608 L1 -> L2 ... -> Ln -> L1 608 L1 -> L2 ... -> Ln -> L1
609 609
610 , which means we have dependencies:: 610 , which means we have dependencies::
611 611
612 L1 -> L2 612 L1 -> L2
613 L2 -> L3 613 L2 -> L3
614 ... 614 ...
615 Ln-1 -> Ln 615 Ln-1 -> Ln
616 Ln -> L1 616 Ln -> L1
617 617
618 We now can construct a combination of locking 618 We now can construct a combination of locking sequences that cause deadlock:
619 619
620 Firstly let's make one CPU/task get the L1 in 620 Firstly let's make one CPU/task get the L1 in L1 -> L2, and then another get
621 the L2 in L2 -> L3, and so on. After this, all 621 the L2 in L2 -> L3, and so on. After this, all of the Lx in Lx -> Lx+1 are
622 held by different CPU/tasks. 622 held by different CPU/tasks.
623 623
624 And then because we have L1 -> L2, so the hold 624 And then because we have L1 -> L2, so the holder of L1 is going to acquire L2
625 in L1 -> L2, however since L2 is already held 625 in L1 -> L2, however since L2 is already held by another CPU/task, plus L1 ->
626 L2 and L2 -> L3 are not -(xR)-> and -(Sx)-> (t 626 L2 and L2 -> L3 are not -(xR)-> and -(Sx)-> (the definition of strong), which
627 means either L2 in L1 -> L2 is a non-recursive 627 means either L2 in L1 -> L2 is a non-recursive locker (blocked by anyone) or
628 the L2 in L2 -> L3, is writer (blocking anyone 628 the L2 in L2 -> L3, is writer (blocking anyone), therefore the holder of L1
629 cannot get L2, it has to wait L2's holder to r 629 cannot get L2, it has to wait L2's holder to release.
630 630
631 Moreover, we can have a similar conclusion for 631 Moreover, we can have a similar conclusion for L2's holder: it has to wait L3's
632 holder to release, and so on. We now can prove 632 holder to release, and so on. We now can prove that Lx's holder has to wait for
633 Lx+1's holder to release, and note that Ln+1 i 633 Lx+1's holder to release, and note that Ln+1 is L1, so we have a circular
634 waiting scenario and nobody can get progress, 634 waiting scenario and nobody can get progress, therefore a deadlock.
635 635
636 Proof for necessary (Lemma 2): 636 Proof for necessary (Lemma 2):
637 637
638 Lemma 2 is equivalent to: If there is a deadlo 638 Lemma 2 is equivalent to: If there is a deadlock scenario, then there must be a
639 strong circle in the dependency graph. 639 strong circle in the dependency graph.
640 640
641 According to Wikipedia[1], if there is a deadl 641 According to Wikipedia[1], if there is a deadlock, then there must be a circular
642 waiting scenario, means there are N CPU/tasks, 642 waiting scenario, means there are N CPU/tasks, where CPU/task P1 is waiting for
643 a lock held by P2, and P2 is waiting for a loc 643 a lock held by P2, and P2 is waiting for a lock held by P3, ... and Pn is waiting
644 for a lock held by P1. Let's name the lock Px 644 for a lock held by P1. Let's name the lock Px is waiting as Lx, so since P1 is waiting
645 for L1 and holding Ln, so we will have Ln -> L 645 for L1 and holding Ln, so we will have Ln -> L1 in the dependency graph. Similarly,
646 we have L1 -> L2, L2 -> L3, ..., Ln-1 -> Ln in 646 we have L1 -> L2, L2 -> L3, ..., Ln-1 -> Ln in the dependency graph, which means we
647 have a circle:: 647 have a circle::
648 648
649 Ln -> L1 -> L2 -> ... -> Ln 649 Ln -> L1 -> L2 -> ... -> Ln
650 650
651 , and now let's prove the circle is strong: 651 , and now let's prove the circle is strong:
652 652
653 For a lock Lx, Px contributes the dependency L 653 For a lock Lx, Px contributes the dependency Lx-1 -> Lx and Px+1 contributes
654 the dependency Lx -> Lx+1, and since Px is wai 654 the dependency Lx -> Lx+1, and since Px is waiting for Px+1 to release Lx,
655 so it's impossible that Lx on Px+1 is a reader 655 so it's impossible that Lx on Px+1 is a reader and Lx on Px is a recursive
656 reader, because readers (no matter recursive o 656 reader, because readers (no matter recursive or not) don't block recursive
657 readers, therefore Lx-1 -> Lx and Lx -> Lx+1 c 657 readers, therefore Lx-1 -> Lx and Lx -> Lx+1 cannot be a -(xR)-> -(Sx)-> pair,
658 and this is true for any lock in the circle, t 658 and this is true for any lock in the circle, therefore, the circle is strong.
659 659
660 References: 660 References:
661 ----------- 661 -----------
662 [1]: https://en.wikipedia.org/wiki/Deadlock 662 [1]: https://en.wikipedia.org/wiki/Deadlock
663 [2]: Shibu, K. (2009). Intro To Embedded Syste 663 [2]: Shibu, K. (2009). Intro To Embedded Systems (1st ed.). Tata McGraw-Hill
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.