| 1 | ===============
|
| 2 | Locking lessons
|
| 3 | ===============
|
| 4 |
|
| 5 | Lesson 1: Spin locks
|
| 6 | ====================
|
| 7 |
|
| 8 | The most basic primitive for locking is spinlock::
|
| 9 |
|
| 10 | static DEFINE_SPINLOCK(xxx_lock);
|
| 11 |
|
| 12 | unsigned long flags;
|
| 13 |
|
| 14 | spin_lock_irqsave(&xxx_lock, flags);
|
| 15 | ... critical section here ..
|
| 16 | spin_unlock_irqrestore(&xxx_lock, flags);
|
| 17 |
|
| 18 | The above is always safe. It will disable interrupts _locally_, but the
|
| 19 | spinlock itself will guarantee the global lock, so it will guarantee that
|
| 20 | there is only one thread-of-control within the region(s) protected by that
|
| 21 | lock. This works well even under UP also, so the code does _not_ need to
|
| 22 | worry about UP vs SMP issues: the spinlocks work correctly under both.
|
| 23 |
|
| 24 | NOTE! Implications of spin_locks for memory are further described in:
|
| 25 |
|
| 26 | Documentation/memory-barriers.txt
|
| 27 |
|
| 28 | (5) ACQUIRE operations.
|
| 29 |
|
| 30 | (6) RELEASE operations.
|
| 31 |
|
| 32 | The above is usually pretty simple (you usually need and want only one
|
| 33 | spinlock for most things - using more than one spinlock can make things a
|
| 34 | lot more complex and even slower and is usually worth it only for
|
| 35 | sequences that you **know** need to be split up: avoid it at all cost if you
|
| 36 | aren't sure).
|
| 37 |
|
| 38 | This is really the only really hard part about spinlocks: once you start
|
| 39 | using spinlocks they tend to expand to areas you might not have noticed
|
| 40 | before, because you have to make sure the spinlocks correctly protect the
|
| 41 | shared data structures **everywhere** they are used. The spinlocks are most
|
| 42 | easily added to places that are completely independent of other code (for
|
| 43 | example, internal driver data structures that nobody else ever touches).
|
| 44 |
|
| 45 | NOTE! The spin-lock is safe only when you **also** use the lock itself
|
| 46 | to do locking across CPU's, which implies that EVERYTHING that
|
| 47 | touches a shared variable has to agree about the spinlock they want
|
| 48 | to use.
|
| 49 |
|
| 50 | ----
|
| 51 |
|
| 52 | Lesson 2: reader-writer spinlocks.
|
| 53 | ==================================
|
| 54 |
|
| 55 | If your data accesses have a very natural pattern where you usually tend
|
| 56 | to mostly read from the shared variables, the reader-writer locks
|
| 57 | (rw_lock) versions of the spinlocks are sometimes useful. They allow multiple
|
| 58 | readers to be in the same critical region at once, but if somebody wants
|
| 59 | to change the variables it has to get an exclusive write lock.
|
| 60 |
|
| 61 | NOTE! reader-writer locks require more atomic memory operations than
|
| 62 | simple spinlocks. Unless the reader critical section is long, you
|
| 63 | are better off just using spinlocks.
|
| 64 |
|
| 65 | The routines look the same as above::
|
| 66 |
|
| 67 | rwlock_t xxx_lock = __RW_LOCK_UNLOCKED(xxx_lock);
|
| 68 |
|
| 69 | unsigned long flags;
|
| 70 |
|
| 71 | read_lock_irqsave(&xxx_lock, flags);
|
| 72 | .. critical section that only reads the info ...
|
| 73 | read_unlock_irqrestore(&xxx_lock, flags);
|
| 74 |
|
| 75 | write_lock_irqsave(&xxx_lock, flags);
|
| 76 | .. read and write exclusive access to the info ...
|
| 77 | write_unlock_irqrestore(&xxx_lock, flags);
|
| 78 |
|
| 79 | The above kind of lock may be useful for complex data structures like
|
| 80 | linked lists, especially searching for entries without changing the list
|
| 81 | itself. The read lock allows many concurrent readers. Anything that
|
| 82 | **changes** the list will have to get the write lock.
|
| 83 |
|
| 84 | NOTE! RCU is better for list traversal, but requires careful
|
| 85 | attention to design detail (see Documentation/RCU/listRCU.rst).
|
| 86 |
|
| 87 | Also, you cannot "upgrade" a read-lock to a write-lock, so if you at _any_
|
| 88 | time need to do any changes (even if you don't do it every time), you have
|
| 89 | to get the write-lock at the very beginning.
|
| 90 |
|
| 91 | NOTE! We are working hard to remove reader-writer spinlocks in most
|
| 92 | cases, so please don't add a new one without consensus. (Instead, see
|
| 93 | Documentation/RCU/rcu.rst for complete information.)
|
| 94 |
|
| 95 | ----
|
| 96 |
|
| 97 | Lesson 3: spinlocks revisited.
|
| 98 | ==============================
|
| 99 |
|
| 100 | The single spin-lock primitives above are by no means the only ones. They
|
| 101 | are the most safe ones, and the ones that work under all circumstances,
|
| 102 | but partly **because** they are safe they are also fairly slow. They are slower
|
| 103 | than they'd need to be, because they do have to disable interrupts
|
| 104 | (which is just a single instruction on a x86, but it's an expensive one -
|
| 105 | and on other architectures it can be worse).
|
| 106 |
|
| 107 | If you have a case where you have to protect a data structure across
|
| 108 | several CPU's and you want to use spinlocks you can potentially use
|
| 109 | cheaper versions of the spinlocks. IFF you know that the spinlocks are
|
| 110 | never used in interrupt handlers, you can use the non-irq versions::
|
| 111 |
|
| 112 | spin_lock(&lock);
|
| 113 | ...
|
| 114 | spin_unlock(&lock);
|
| 115 |
|
| 116 | (and the equivalent read-write versions too, of course). The spinlock will
|
| 117 | guarantee the same kind of exclusive access, and it will be much faster.
|
| 118 | This is useful if you know that the data in question is only ever
|
| 119 | manipulated from a "process context", ie no interrupts involved.
|
| 120 |
|
| 121 | The reasons you mustn't use these versions if you have interrupts that
|
| 122 | play with the spinlock is that you can get deadlocks::
|
| 123 |
|
| 124 | spin_lock(&lock);
|
| 125 | ...
|
| 126 | <- interrupt comes in:
|
| 127 | spin_lock(&lock);
|
| 128 |
|
| 129 | where an interrupt tries to lock an already locked variable. This is ok if
|
| 130 | the other interrupt happens on another CPU, but it is _not_ ok if the
|
| 131 | interrupt happens on the same CPU that already holds the lock, because the
|
| 132 | lock will obviously never be released (because the interrupt is waiting
|
| 133 | for the lock, and the lock-holder is interrupted by the interrupt and will
|
| 134 | not continue until the interrupt has been processed).
|
| 135 |
|
| 136 | (This is also the reason why the irq-versions of the spinlocks only need
|
| 137 | to disable the _local_ interrupts - it's ok to use spinlocks in interrupts
|
| 138 | on other CPU's, because an interrupt on another CPU doesn't interrupt the
|
| 139 | CPU that holds the lock, so the lock-holder can continue and eventually
|
| 140 | releases the lock).
|
| 141 |
|
| 142 | Linus
|
| 143 |
|
| 144 | ----
|
| 145 |
|
| 146 | Reference information:
|
| 147 | ======================
|
| 148 |
|
| 149 | For dynamic initialization, use spin_lock_init() or rwlock_init() as
|
| 150 | appropriate::
|
| 151 |
|
| 152 | spinlock_t xxx_lock;
|
| 153 | rwlock_t xxx_rw_lock;
|
| 154 |
|
| 155 | static int __init xxx_init(void)
|
| 156 | {
|
| 157 | spin_lock_init(&xxx_lock);
|
| 158 | rwlock_init(&xxx_rw_lock);
|
| 159 | ...
|
| 160 | }
|
| 161 |
|
| 162 | module_init(xxx_init);
|
| 163 |
|
| 164 | For static initialization, use DEFINE_SPINLOCK() / DEFINE_RWLOCK() or
|
| 165 | __SPIN_LOCK_UNLOCKED() / __RW_LOCK_UNLOCKED() as appropriate.
|