freebsd-skq/share/man/man9/mutex.9
John Baldwin 2c3d0ee6f4 - Document MUTEX_DECLARE and MTX_COLD
- Clean up some minor nits
2000-10-26 23:53:12 +00:00

402 lines
13 KiB
Groff

.\"
.\" Copyright (c) 1998 Berkeley Software Design, Inc. All rights reserved.
.\"
.\" Redistribution and use in source and binary forms, with or without
.\" modification, are permitted provided that the following conditions
.\" are met:
.\" 1. Redistributions of source code must retain the above copyright
.\" notice, this list of conditions and the following disclaimer.
.\" 2. Redistributions in binary form must reproduce the above copyright
.\" notice, this list of conditions and the following disclaimer in the
.\" documentation and/or other materials provided with the distribution.
.\" 3. Berkeley Software Design Inc's name may not be used to endorse or
.\" promote products derived from this software without specific prior
.\" written permission.
.\"
.\" THIS SOFTWARE IS PROVIDED BY BERKELEY SOFTWARE DESIGN INC ``AS IS'' AND
.\" ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
.\" IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
.\" ARE DISCLAIMED. IN NO EVENT SHALL BERKELEY SOFTWARE DESIGN INC BE LIABLE
.\" FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
.\" DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
.\" OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
.\" HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
.\" LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
.\" OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
.\" SUCH DAMAGE.
.\"
.\" from BSDI $Id: mutex.4,v 1.1.2.3 1998/04/27 22:53:13 ewv Exp $
.\" $FreeBSD$
.\"
.Dd April 20, 1998
.Dt MUTEX 9
.Sh NAME
.Nm mutex,
.Nm mtx_init,
.Nm mtx_enter,
.Nm mtx_try_enter,
.Nm mtx_exit,
.Nm mtx_destroy,
.Nm mtx_owned,
.Nm mtx_assert,
.Nm MUTEX_DECLARE
.Nd kernel synchronization primitives
.Sh SYNOPSIS
.Fd #include <machine/mutex.h>
.Ft int
.Fn mtx_init "struct mtx *mutex" "char *name" "flags"
.Ft void
.Fn mtx_enter "struct mtx *mutex" "int flags"
.Ft void
.Fn mtx_try_enter "struct mtx *mutex" "int flags"
.Ft void
.Fn mtx_exit "struct mtx *mutex" "int flags"
.Ft void
.Fn mtx_destroy "struct mtx *mutex"
.Ft int
.Fn mtx_owned "struct mtx *mutex"
.Ft void
.Fn mtx_assert "struct mtx *mutex" "int what"
.Fn MUTEX_DECLARE "modifiers" "name"
.Sh DESCRIPTION
Mutexes are the most basic and primary method of process synchronization.
The major design considerations for mutexes are:
.Bl -enum
.It
Acquiring and releasing uncontested mutexes should be as cheap
as possible.
.It
They must have the information and storage space to support
priority propagation.
.It
A process must be able to recursively acquire a mutex.
.El
.Pp
There are currently two flavors of mutexes, those that context switch
when they block and those that do not.
.Pp
By default mutexes will context switch when they are already held.
As a machine dependent optimization they may spin for some amount
of time before context switching.
It is important to remember that since a process may be preempted
at any time, the possible context switch introduced by acquiring a
mutex is guaranteed to not break anything that isn't already broken.
.Pp
Mutexes which do not context switch are spin mutexes.
These should only be used to protect data shared with device that
require non-preemptive interrupts, and low level scheduling code.
In most/all architectures both acquiring and releasing of a
uncontested spin mutex is more expensive than the same operation
on a non spin mutex.
In order to protect an interrupt service routine from blocking
against itself all interrupts are blocked on a processor while
holding a spin lock. It is permissible to hold multiple spin mutexes.
In this case it is a required that they be released in the opposite
order to that which they were acquired.
.Pp
Once a spin mutex has been acquired it is not permissible to acquire a
blocking mutex.
.Pp
The storage needed to implement a mutex is provided by a
.Dv struct mtx .
In general this should be treated as an opaque object and
referenced only with the mutex primitives.
.Pp
The
.Fn mtx_init
function must be used to initialize a mutex
before it can be passed to
.Fn mtx_enter .
The
.Ar name
argument is used by the witness code
to classify a mutex when doing checks of lock ordering.
The pointer passed in as the
.Ar name
is saved rather than the data it points to.
The data pointed to must remain stable
until the mutex is destroyed.
Currently the
.Ar flag
argument is unused.
In the future it will likely be at least
used to identify spin mutexes
so that debug code can verify
consistent use of a mutex.
It is not permissible to pass the same
.Ar mutex
to
.Fn mtx_init
multiple times without intervening calls to
.Fn mtx_destroy .
.Pp
The
.Fn mtx_enter
function acquires a mutual exclusion lock
on behalf of the currently running kernel thread.
If another kernel thread is holding the mutex,
the caller will be disconnected from the CPU
until the mutex is available
(i.e. it will sleep),
spin wait for the mutex,
or possibly a combination of both.
.Pp
It is possible for the same thread to recursively acquire a mutex
with no ill effects;
if recursion on a given mutex can be avoided,
faster and smaller code will usually be generated.
.Pp
The
.Fn mtx_try_enter
function is used to acquire exclusive access
to those objects protected by the mutex
pointed to by
.Ar mutex .
The
.Ar flag
argument is used to specify various options,
typically
.Dv MTX_DEF
is supplied.
If the mutex can not be immediately acquired
.Fn mtx_try_enter
will return 0,
otherwise the mutex will be acquired
and a non-zero value will be returned.
.Pp
The
.Fn mtx_exit
function releases a mutual exclusion lock;
if a higher priority thread is waiting for the mutex,
the releasing thread may be disconnected
to allow the higher priority thread to acquire the mutex and run.
.Pp
The
.Fn mtx_destroy
function is used to destroy
.Ar mutex
so the data associated with it may be freed
or otherwise overwritten.
Any mutex which is destroyed
must previously have been initialized with
.Fn mtx_init .
It is permissible to have a single hold count
on a mutex when it is destroyed.
It is not permissible to hold the mutex recursively,
or have another process blocked on the mutex
when it is destroyed.
.Pp
The
.Fn mtx_owned
function returns non-zero
if the current process holds
.Ar mutex .
If the current process does not hold
.Ar mutex
zero is returned.
.Pp
The
.Fn mtx_assert
function allows assertions to be made about
.Ar mutex .
If the assertions are not true and the kernel is compiled with
.Dv INVARIANTS
then the kernel will panic.
Currently the following assertions are supported:
.Bl -enum
.It
.Dv MA_OWNED
Assert that the current thread
holds the mutex
pointed to by the first argument.
.It
.Dv MA_NOTOWNED
Assert that the current thread
does not hold the mutex
pointed to by the first argument.
.El
.Pp
The
.Fn MUTEX_DECLARE
macro is used to declare a mutex that is initialized before
.Xr malloc 9
is operating. Unfortunately, mutex initialization may require
.Xr malloc 9 .
However, some mutexes are intialized and used before
.Xr malloc 9
can be used. Declaring these mutexes with the
.Fn MUTEX_DECLARE
macro and then using the
.Dv MTX_COLD
flag when calling
.Fn mtx_init
allows these early mutexes to be initialized and used before
.Xr malloc 9
is available.
The
.Ar modifiers
argument is a list of attributes to be applied to the mutex structure being
declared such as
.Dq static .
The
.Ar name
argument is the name of the mutex structure to declare.
.Pp
The type of a mutex is not an attribute of the mutex,
but instead a function of the
.Fa flags
argument passed to
.Fn mtx_enter
and
.Fn mtx_exit ;
this allows code to be generated for the specific mutex type
at compile time
and avoids wasting run time on the determination of lock features.
This does place on the programmer,
the burden of using matching forms of the
.Fn mtx_enter
and
.Fn mtx_exit
functions for a given mutex.
It is an error to acquire a mutex in one mode (e.g. spin)
and release it in another (e.g. default).
It is also an error to get the lock in one mode
and allow another thread to attempt to get the lock in another mode.
A good general rule is to always use a given mutex in one mode only.
.Ss The default Mutex Type
Most kernel code should use the default lock type;
the default lock type will allow the thread
to be disconnected from the CPU
if it cannot get the lock.
The machine dependent implementation
may treat the lock as a short term spin lock
under some circumstances.
However, it is always safe to use these forms of locks
in an interrupt thread
without fear of deadlock
against an interrupted thread on the same CPU.
.Ss The spin Mutex Type
A spin mutex will not relinquish the CPU
when it cannot immediately get the requested lock,
but will loop, waiting for the mutex to be released by another CPU.
This could result in deadlock
if a thread interrupted the thread which held a mutex
and then tried to acquire the mutex;
for this reason spin locks will disable all interrupts
(on the local CPU only)
by default.
.Pp
Spin locks are fairly specialized locks
that are intended to be held for very short periods of time;
their primary purpose is to protect portions of the code
that implement default (i.e. sleep) locks.
.Ss Flags
The flags passed to the
.Fn mtx_enter
and
.Fn mtx_exit
functions determine what type of mutex is being used
and also provide various options
used to generate more efficient code under certain circumstances.
.Pp
Both lock types (default and spin)
can be acquired recursively by the same thread.
This behavior can be changed with flags.
.Pp
The type of the mutex must always be specified:
.Bl -tag -width MTX_NORECURSE
.It Dv MTX_DEF
Default lock type;
will always allow the current thread to be suspended
to avoid deadlock conditions against interrupt threads.
The machine dependent implementation of this lock type
may spin for a while before suspending the current thread.
Most locks should be of this type.
.It Dv MTX_SPIN
Spin lock;
will never relinquish the CPU.
By default all interrupts are disabled on the local CPU
while any spin lock is held.
.El
.Pp
Options that modify mutex behavior:
.Bl -tag -width MTX_NORECURSE
.It Dv MTX_NORECURSE
If it is known, absolutely,
that the mutex will not be recursively acquired at this invocation
then this flag should be specified.
.Pp
If the lock is already held by the current thread,
then a kernel with
.Dv MUTEX_DEBUG
defined will panic;
without debugging enabled,
the thread may deadlock against itself
or leave the mutex in a corrupted state.
.Pp
This flag prevents generation of additional inline code
to deal with recursive lock acquisitions
and should be specified whenever possible
in the interests of efficiency.
Not specifying this flag will only cause the generated code
to be a little larger than necessary;
it will still operate correctly.
.It Dv MTX_RLIKELY
This provides a hint that it is likely that this mutex
will be held recursively at this invocation.
The actual optimization used is machine dependent;
generally, this will inline code to handle recursion
where a function call would otherwise be needed.
.Pp
This is a hint only;
leaving it out or specifying it inappropriately
will not cause any great harm other than
possibly generating less efficient code.
.It Dv MTX_TOPHALF
This option applies to spin locks only.
It indicates that the mutex is never acquired
from an interrupt thread,
so it is safe to leave interrupts enabled while holding the lock.
Since an interrupt may occur while holding the lock,
this may be detrimental to other processors
spin waiting for the lock.
Do not forget to include this option when the lock is released.
.Pp
This option should not be used in new code;
it is documented here for completeness only.
.It Dv MTX_FIRST
This option applies to spin locks only.
It indicates this is the first spin lock acquired by the thread.
No other spin locks may be held,
and the requested lock also may not be currently held.
Do not forget to include this option when the lock is released.
.It Dv MTX_NOSWITCH
When releasing a mutex,
this flag prevents a thread switch that might occur
if another higher priority thread was waiting for the mutex.
This may cause priority inversion and should be used carefully.
.Pp
This flag is used internally by the lock code.
It should not be used in general kernel code
and is documented here for completeness only.
.It Dv MTX_NOSPIN
For default locks,
this hint will prevent spinning before relinquishing the CPU.
This should be specified when it is known
that the lock will usually remain unavailable for some time
when it is not immediately available
(i.e.: coarse grained locks protecting large subsystems).
.It Dv MTX_COLD
This option is only used in
.Fn mtx_init
and is used in conjunction with mutexes declared with
.Fn MUTEX_DECLARE
to initialize mutexes that are needed before
.Xr malloc 9
is available for use.
.El
.Sh HISTORY
These
functions appeared in BSD/OS 4.1 and
.Fx 5.0 .