NAME
locking - kernel synchronization primitives
SYNOPSIS
All sorts of stuff to go here.
DESCRIPTION
The FreeBSD kernel is written to run across multiple CPUs and as such
requires several different synchronization primitives to allow the
developers to safely access and manipulate the many data types required.
These include:
1. Spin Mutexes
2. Sleep Mutexes
3. pool Mutexes
4. Shared-Exclusive locks
5. Reader-Writer locks
6. Read-Mostly locks
7. Turnstiles
8. Semaphores
9. Condition variables
10. Sleep/wakeup
11. Giant
12. Lockmanager locks
The primitives interact and have a number of rules regarding how they can
and can not be combined. There are too many for the average human mind
and they keep changing. (if you disagree, please write replacement text)
:-)
Some of these primitives may be used at the low (interrupt) level and
some may not.
There are strict ordering requirements and for some of the types this is
checked using the witness(4) code.
SPIN Mutexes
Mutexes are the basic primitive. You either hold it or you don’t. If
you don’t own it then you just spin, waiting for the holder (on another
CPU) to release it. Hopefully they are doing something fast. You must
not do anything that deschedules the thread while you are holding a SPIN
mutex.
Mutexes
Basically (regular) mutexes will deschedule the thread if the mutex can
not be acquired. A non-spin mutex can be considered to be equivalent to
getting a write lock on an rw_lock (see below), and in fact non-spin
mutexes and rw_locks may soon become the same thing. As in spin mutexes,
you either get it or you don’t. You may only call the sleep(9) call via
msleep() or the new mtx_sleep() variant. These will atomically drop the
mutex and reacquire it as part of waking up. This is often however a BAD
idea because it generally relies on you having such a good knowledge of
all the call graph above you and what assumptions it is making that there
are a lot of ways to make hard-to-find mistakes. For example you MUST
re-test all the assumptions you made before, all the way up the call
graph to where you got the lock. You can not just assume that mtx_sleep
can be inserted anywhere. If any caller above you has any mutex or
rwlock, your sleep, will cause a panic. If the sleep only happens rarely
it may be years before the bad code path is found.
Pool Mutexes
A variant of regular mutexes where the allocation of the mutex is handled
more by the system.
Rw_locks
Reader/writer locks allow shared access to protected data by multiple
threads, or exclusive access by a single thread. The threads with shared
access are known as readers since they should only read the protected
data. A thread with exclusive access is known as a writer since it may
modify protected data.
Although reader/writer locks look very similar to sx(9) (see below)
locks, their usage pattern is different. Reader/writer locks can be
treated as mutexes (see above and mutex(9)) with shared/exclusive
semantics. More specifically, regular mutexes can be considered to be
equivalent to a write-lock on an rw_lock. In the future this may in fact
become literally the fact. An rw_lock can be locked while holding a
regular mutex, but can not be held while sleeping. The rw_lock locks
have priority propagation like mutexes, but priority can be propagated
only to an exclusive holder. This limitation comes from the fact that
shared owners are anonymous. Another important property is that shared
holders of rw_lock can recurse, but exclusive locks are not allowed to
recurse. This ability should not be used lightly and may go away. Users
of recursion in any locks should be prepared to defend their decision
against vigorous criticism.
Rm_locks
Mostly reader locks are similar to Reader/write locks but optimized for
very infrequent writer locking. rm_lock locks implement full priority
propagation by tracking shared owners using a lock user supplied tracker
data structure.
Sx_locks
Shared/exclusive locks are used to protect data that are read far more
often than they are written. Mutexes are inherently more efficient than
shared/exclusive locks, so shared/exclusive locks should be used
prudently. The main reason for using an sx_lock is that a thread may
hold a shared or exclusive lock on an sx_lock lock while sleeping. As a
consequence of this however, an sx_lock lock may not be acquired while
holding a mutex. The reason for this is that, if one thread slept while
holding an sx_lock lock while another thread blocked on the same sx_lock
lock after acquiring a mutex, then the second thread would effectively
end up sleeping while holding a mutex, which is not allowed. The sx_lock
should be considered to be closely related to sleep(9). In fact it could
in some cases be considered a conditional sleep.
Turnstiles
Turnstiles are used to hold a queue of threads blocked on non-sleepable
locks. Sleepable locks use condition variables to implement their
queues. Turnstiles differ from a sleep queue in that turnstile queue’s
are assigned to a lock held by an owning thread. Thus, when one thread
is enqueued onto a turnstile, it can lend its priority to the owning
thread. If this sounds confusing, we need to describe it better.
Semaphores
Condition variables
Condition variables are used in conjunction with mutexes to wait for
conditions to occur. A thread must hold the mutex before calling the
cv_wait*(), functions. When a thread waits on a condition, the mutex is
atomically released before the thread is blocked, then reacquired before
the function call returns.
Giant
Giant is a special instance of a sleep lock. It has several special
characteristics.
1. It is recursive.
2. Drivers can request that Giant be locked around them, but this is
going away.
3. You can sleep while it has recursed, but other recursive locks
cannot.
4. Giant must be locked first before other locks.
5. There are places in the kernel that drop Giant and pick it back up
again. Sleep locks will do this before sleeping. Parts of the
Network or VM code may do this as well, depending on the setting of
a sysctl. This means that you cannot count on Giant keeping other
code from running if your code sleeps, even if you want it to.
Sleep/wakeup
The functions tsleep(), msleep(), msleep_spin(), pause(), wakeup(), and
wakeup_one() handle event-based thread blocking. If a thread must wait
for an external event, it is put to sleep by tsleep(), msleep(),
msleep_spin(), or pause(). Threads may also wait using one of the
locking primitive sleep routines mtx_sleep(9), rw_sleep(9), or
sx_sleep(9).
The parameter chan is an arbitrary address that uniquely identifies the
event on which the thread is being put to sleep. All threads sleeping on
a single chan are woken up later by wakeup(), often called from inside an
interrupt routine, to indicate that the resource the thread was blocking
on is available now.
Several of the sleep functions including msleep(), msleep_spin(), and the
locking primitive sleep routines specify an additional lock parameter.
The lock will be released before sleeping and reacquired before the sleep
routine returns. If priority includes the PDROP flag, then the lock will
not be reacquired before returning. The lock is used to ensure that a
condition can be checked atomically, and that the current thread can be
suspended without missing a change to the condition, or an associated
wakeup. In addition, all of the sleep routines will fully drop the Giant
mutex (even if recursed) while the thread is suspended and will reacquire
the Giant mutex before the function returns.
lockmanager locks
Largely deprecated. See the lock(9) page for more information. I don’t
know what the downsides are but I’m sure someone will fill in this part.
Usage tables.
Interaction table.
The following table shows what you can and can not do if you hold one of
the synchronization primitives discussed here: (someone who knows what
they are talking about should write this table)
You have: You want: Spin_mtx Slp_mtx sx_lock rw_lock rm_locksleep
SPIN mutex ok-1 no no no no no-3
Sleep mutex ok ok-1 no ok ok no-3
sx_lock ok ok ok-2 ok ok ok-4
rw_lock ok ok no ok-2 ok no-3
rm_lock ok ok no ok ok-2 no
*1 Recursion is defined per lock. Lock order is important.
*2 readers can recurse though writers can not. Lock order is important.
*3 There are calls atomically release this primitive when going to sleep
and reacquire it on wakeup (e.g. mtx_sleep(), rw_sleep() and
msleep_spin() ).
*4 Though one can sleep holding an sx lock, one can also use sx_sleep()
which atomically release this primitive when going to sleep and reacquire
it on wakeup.
Context mode table.
The next table shows what can be used in different contexts. At this
time this is a rather easy to remember table.
Context: Spin_mtx Slp_mtx sx_lock rw_lock rm_locksleep
interrupt: ok no no no no no
idle: ok no no no no no
SEE ALSO
condvar(9), lock(9), mtx_pool(9), mutex(9), rmlock(9), rwlock(9),
sema(9), sleep(9), sx(9), LOCK_PROFILING(9), WITNESS(9)
HISTORY
These functions appeared in BSD/OS 4.1 through FreeBSD 7.0