1Lesson 1: Spin locks
   3The most basic primitive for locking is spinlock.
   5static DEFINE_SPINLOCK(xxx_lock);
   7        unsigned long flags;
   9        spin_lock_irqsave(&xxx_lock, flags);
  10        ... critical section here ..
  11        spin_unlock_irqrestore(&xxx_lock, flags);
  13The above is always safe. It will disable interrupts _locally_, but the
  14spinlock itself will guarantee the global lock, so it will guarantee that
  15there is only one thread-of-control within the region(s) protected by that
  16lock. This works well even under UP also, so the code does _not_ need to
  17worry about UP vs SMP issues: the spinlocks work correctly under both.
  19   NOTE! Implications of spin_locks for memory are further described in:
  21     Documentation/memory-barriers.txt
  22       (5) LOCK operations.
  23       (6) UNLOCK operations.
  25The above is usually pretty simple (you usually need and want only one
  26spinlock for most things - using more than one spinlock can make things a
  27lot more complex and even slower and is usually worth it only for
  28sequences that you _know_ need to be split up: avoid it at all cost if you
  29aren't sure).
  31This is really the only really hard part about spinlocks: once you start
  32using spinlocks they tend to expand to areas you might not have noticed
  33before, because you have to make sure the spinlocks correctly protect the
  34shared data structures _everywhere_ they are used. The spinlocks are most
  35easily added to places that are completely independent of other code (for
  36example, internal driver data structures that nobody else ever touches).
  38   NOTE! The spin-lock is safe only when you _also_ use the lock itself
  39   to do locking across CPU's, which implies that EVERYTHING that
  40   touches a shared variable has to agree about the spinlock they want
  41   to use.
  45Lesson 2: reader-writer spinlocks.
  47If your data accesses have a very natural pattern where you usually tend
  48to mostly read from the shared variables, the reader-writer locks
  49(rw_lock) versions of the spinlocks are sometimes useful. They allow multiple
  50readers to be in the same critical region at once, but if somebody wants
  51to change the variables it has to get an exclusive write lock.
  53   NOTE! reader-writer locks require more atomic memory operations than
  54   simple spinlocks.  Unless the reader critical section is long, you
  55   are better off just using spinlocks.
  57The routines look the same as above:
  59   rwlock_t xxx_lock = __RW_LOCK_UNLOCKED(xxx_lock);
  61        unsigned long flags;
  63        read_lock_irqsave(&xxx_lock, flags);
  64        .. critical section that only reads the info ...
  65        read_unlock_irqrestore(&xxx_lock, flags);
  67        write_lock_irqsave(&xxx_lock, flags);
  68        .. read and write exclusive access to the info ...
  69        write_unlock_irqrestore(&xxx_lock, flags);
  71The above kind of lock may be useful for complex data structures like
  72linked lists, especially searching for entries without changing the list
  73itself.  The read lock allows many concurrent readers.  Anything that
  74_changes_ the list will have to get the write lock.
  76   NOTE! RCU is better for list traversal, but requires careful
  77   attention to design detail (see Documentation/RCU/listRCU.txt).
  79Also, you cannot "upgrade" a read-lock to a write-lock, so if you at _any_
  80time need to do any changes (even if you don't do it every time), you have
  81to get the write-lock at the very beginning.
  83   NOTE! We are working hard to remove reader-writer spinlocks in most
  84   cases, so please don't add a new one without consensus.  (Instead, see
  85   Documentation/RCU/rcu.txt for complete information.)
  89Lesson 3: spinlocks revisited.
  91The single spin-lock primitives above are by no means the only ones. They
  92are the most safe ones, and the ones that work under all circumstances,
  93but partly _because_ they are safe they are also fairly slow. They are slower
  94than they'd need to be, because they do have to disable interrupts
  95(which is just a single instruction on a x86, but it's an expensive one -
  96and on other architectures it can be worse).
  98If you have a case where you have to protect a data structure across
  99several CPU's and you want to use spinlocks you can potentially use
 100cheaper versions of the spinlocks. IFF you know that the spinlocks are
 101never used in interrupt handlers, you can use the non-irq versions:
 103        spin_lock(&lock);
 104        ...
 105        spin_unlock(&lock);
 107(and the equivalent read-write versions too, of course). The spinlock will
 108guarantee the same kind of exclusive access, and it will be much faster. 
 109This is useful if you know that the data in question is only ever
 110manipulated from a "process context", ie no interrupts involved. 
 112The reasons you mustn't use these versions if you have interrupts that
 113play with the spinlock is that you can get deadlocks:
 115        spin_lock(&lock);
 116        ...
 117                <- interrupt comes in:
 118                        spin_lock(&lock);
 120where an interrupt tries to lock an already locked variable. This is ok if
 121the other interrupt happens on another CPU, but it is _not_ ok if the
 122interrupt happens on the same CPU that already holds the lock, because the
 123lock will obviously never be released (because the interrupt is waiting
 124for the lock, and the lock-holder is interrupted by the interrupt and will
 125not continue until the interrupt has been processed). 
 127(This is also the reason why the irq-versions of the spinlocks only need
 128to disable the _local_ interrupts - it's ok to use spinlocks in interrupts
 129on other CPU's, because an interrupt on another CPU doesn't interrupt the
 130CPU that holds the lock, so the lock-holder can continue and eventually
 131releases the lock). 
 133Note that you can be clever with read-write locks and interrupts. For
 134example, if you know that the interrupt only ever gets a read-lock, then
 135you can use a non-irq version of read locks everywhere - because they
 136don't block on each other (and thus there is no dead-lock wrt interrupts. 
 137But when you do the write-lock, you have to use the irq-safe version. 
 139For an example of being clever with rw-locks, see the "waitqueue_lock" 
 140handling in kernel/sched.c - nothing ever _changes_ a wait-queue from
 141within an interrupt, they only read the queue in order to know whom to
 142wake up. So read-locks are safe (which is good: they are very common
 143indeed), while write-locks need to protect themselves against interrupts.
 145                Linus
 149Reference information:
 151For dynamic initialization, use spin_lock_init() or rwlock_init() as
 154   spinlock_t xxx_lock;
 155   rwlock_t xxx_rw_lock;
 157   static int __init xxx_init(void)
 158   {
 159        spin_lock_init(&xxx_lock);
 160        rwlock_init(&xxx_rw_lock);
 161        ...
 162   }
 164   module_init(xxx_init);
 166For static initialization, use DEFINE_SPINLOCK() / DEFINE_RWLOCK() or
 167__SPIN_LOCK_UNLOCKED() / __RW_LOCK_UNLOCKED() as appropriate.