1 SPIN_LOCK_UNLOCKED and RW_LOCK_UNLOCKED defeat lockdep state tracking and
4 Please use DEFINE_SPINLOCK()/DEFINE_RWLOCK() or
5 __SPIN_LOCK_UNLOCKED()/__RW_LOCK_UNLOCKED() as appropriate for static
8 Dynamic initialization, when necessary, may be performed as
14 static int __init xxx_init(void)
16 spin_lock_init(&xxx_lock);
17 rwlock_init(&xxx_rw_lock);
21 module_init(xxx_init);
23 The following discussion is still valid, however, with the dynamic
24 initialization of spinlocks or with DEFINE_SPINLOCK, etc., used
25 instead of SPIN_LOCK_UNLOCKED.
27 -----------------------
29 On Fri, 2 Jan 1998, Doug Ledford wrote:
31 > I'm working on making the aic7xxx driver more SMP friendly (as well as
32 > importing the latest FreeBSD sequencer code to have 7895 support) and wanted
33 > to get some info from you. The goal here is to make the various routines
34 > SMP safe as well as UP safe during interrupts and other manipulating
35 > routines. So far, I've added a spin_lock variable to things like my queue
36 > structs. Now, from what I recall, there are some spin lock functions I can
37 > use to lock these spin locks from other use as opposed to a (nasty)
38 > save_flags(); cli(); stuff; restore_flags(); construct. Where do I find
39 > these routines and go about making use of them? Do they only lock on a
40 > per-processor basis or can they also lock say an interrupt routine from
41 > mucking with a queue if the queue routine was manipulating it when the
42 > interrupt occurred, or should I still use a cli(); based construct on that
45 See <asm/spinlock.h>. The basic version is:
47 spinlock_t xxx_lock = SPIN_LOCK_UNLOCKED;
52 spin_lock_irqsave(&xxx_lock, flags);
53 ... critical section here ..
54 spin_unlock_irqrestore(&xxx_lock, flags);
56 and the above is always safe. It will disable interrupts _locally_, but the
57 spinlock itself will guarantee the global lock, so it will guarantee that
58 there is only one thread-of-control within the region(s) protected by that
61 Note that it works well even under UP - the above sequence under UP
62 essentially is just the same as doing a
66 save_flags(flags); cli();
67 ... critical section ...
70 so the code does _not_ need to worry about UP vs SMP issues: the spinlocks
71 work correctly under both (and spinlocks are actually more efficient on
72 architectures that allow doing the "save_flags + cli" in one go because I
73 don't export that interface normally).
75 NOTE NOTE NOTE! The reason the spinlock is so much faster than a global
76 interrupt lock under SMP is exactly because it disables interrupts only on
77 the local CPU. The spin-lock is safe only when you _also_ use the lock
78 itself to do locking across CPU's, which implies that EVERYTHING that
79 touches a shared variable has to agree about the spinlock they want to
82 The above is usually pretty simple (you usually need and want only one
83 spinlock for most things - using more than one spinlock can make things a
84 lot more complex and even slower and is usually worth it only for
85 sequences that you _know_ need to be split up: avoid it at all cost if you
86 aren't sure). HOWEVER, it _does_ mean that if you have some code that does
89 .. critical section ..
92 and another sequence that does
94 spin_lock_irqsave(flags);
95 .. critical section ..
96 spin_unlock_irqrestore(flags);
98 then they are NOT mutually exclusive, and the critical regions can happen
99 at the same time on two different CPU's. That's fine per se, but the
100 critical regions had better be critical for different things (ie they
101 can't stomp on each other).
103 The above is a problem mainly if you end up mixing code - for example the
104 routines in ll_rw_block() tend to use cli/sti to protect the atomicity of
105 their actions, and if a driver uses spinlocks instead then you should
106 think about issues like the above..
108 This is really the only really hard part about spinlocks: once you start
109 using spinlocks they tend to expand to areas you might not have noticed
110 before, because you have to make sure the spinlocks correctly protect the
111 shared data structures _everywhere_ they are used. The spinlocks are most
112 easily added to places that are completely independent of other code (ie
113 internal driver data structures that nobody else ever touches, for
118 Lesson 2: reader-writer spinlocks.
120 If your data accesses have a very natural pattern where you usually tend
121 to mostly read from the shared variables, the reader-writer locks
122 (rw_lock) versions of the spinlocks are often nicer. They allow multiple
123 readers to be in the same critical region at once, but if somebody wants
124 to change the variables it has to get an exclusive write lock. The
125 routines look the same as above:
127 rwlock_t xxx_lock = RW_LOCK_UNLOCKED;
132 read_lock_irqsave(&xxx_lock, flags);
133 .. critical section that only reads the info ...
134 read_unlock_irqrestore(&xxx_lock, flags);
136 write_lock_irqsave(&xxx_lock, flags);
137 .. read and write exclusive access to the info ...
138 write_unlock_irqrestore(&xxx_lock, flags);
140 The above kind of lock is useful for complex data structures like linked
141 lists etc, especially when you know that most of the work is to just
142 traverse the list searching for entries without changing the list itself,
143 for example. Then you can use the read lock for that kind of list
144 traversal, which allows many concurrent readers. Anything that _changes_
145 the list will have to get the write lock.
147 Note: you cannot "upgrade" a read-lock to a write-lock, so if you at _any_
148 time need to do any changes (even if you don't do it every time), you have
149 to get the write-lock at the very beginning. I could fairly easily add a
150 primitive to create a "upgradeable" read-lock, but it hasn't been an issue
151 yet. Tell me if you'd want one.
155 Lesson 3: spinlocks revisited.
157 The single spin-lock primitives above are by no means the only ones. They
158 are the most safe ones, and the ones that work under all circumstances,
159 but partly _because_ they are safe they are also fairly slow. They are
160 much faster than a generic global cli/sti pair, but slower than they'd
161 need to be, because they do have to disable interrupts (which is just a
162 single instruction on a x86, but it's an expensive one - and on other
163 architectures it can be worse).
165 If you have a case where you have to protect a data structure across
166 several CPU's and you want to use spinlocks you can potentially use
167 cheaper versions of the spinlocks. IFF you know that the spinlocks are
168 never used in interrupt handlers, you can use the non-irq versions:
174 (and the equivalent read-write versions too, of course). The spinlock will
175 guarantee the same kind of exclusive access, and it will be much faster.
176 This is useful if you know that the data in question is only ever
177 manipulated from a "process context", ie no interrupts involved.
179 The reasons you mustn't use these versions if you have interrupts that
180 play with the spinlock is that you can get deadlocks:
184 <- interrupt comes in:
187 where an interrupt tries to lock an already locked variable. This is ok if
188 the other interrupt happens on another CPU, but it is _not_ ok if the
189 interrupt happens on the same CPU that already holds the lock, because the
190 lock will obviously never be released (because the interrupt is waiting
191 for the lock, and the lock-holder is interrupted by the interrupt and will
192 not continue until the interrupt has been processed).
194 (This is also the reason why the irq-versions of the spinlocks only need
195 to disable the _local_ interrupts - it's ok to use spinlocks in interrupts
196 on other CPU's, because an interrupt on another CPU doesn't interrupt the
197 CPU that holds the lock, so the lock-holder can continue and eventually
200 Note that you can be clever with read-write locks and interrupts. For
201 example, if you know that the interrupt only ever gets a read-lock, then
202 you can use a non-irq version of read locks everywhere - because they
203 don't block on each other (and thus there is no dead-lock wrt interrupts.
204 But when you do the write-lock, you have to use the irq-safe version.
206 For an example of being clever with rw-locks, see the "waitqueue_lock"
207 handling in kernel/sched.c - nothing ever _changes_ a wait-queue from
208 within an interrupt, they only read the queue in order to know whom to
209 wake up. So read-locks are safe (which is good: they are very common
210 indeed), while write-locks need to protect themselves against interrupts.