man/man9/mutex.9

.\"
.\" Copyright (c) 2010 The DragonFly Project.  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. Neither the name of The DragonFly Project nor the names of its
.\"    contributors may be used to endorse or promote products derived
.\"    from this software without specific, prior written permission.
.\"
.\" THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
.\" ``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 THE
.\" COPYRIGHT HOLDERS OR CONTRIBUTORS 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.
.\"
.Dd August 24, 2012
.Dt MUTEX 9
.Os
.Sh NAME
.Nm mutex ,
.Nm mtx_init ,
.Nm mtx_uninit ,
.Nm mtx_lock_sh ,
.Nm mtx_lock_sh_quick ,
.Nm mtx_lock_ex ,
.Nm mtx_lock_ex_quick ,
.Nm mtx_lock ,
.Nm mtx_spinlock ,
.Nm mtx_lock_ex_try ,
.Nm mtx_lock_sh_try ,
.Nm mtx_spinlock_try ,
.Nm mtx_downgrade ,
.Nm mtx_upgrade_try ,
.Nm mtx_unlock ,
.Nm mtx_unlock_ex ,
.Nm mtx_unlock_sh ,
.Nm mtx_spinunlock ,
.Nm mtx_islocked ,
.Nm mtx_islocked_ex ,
.Nm mtx_notlocked ,
.Nm mtx_notlocked_ex ,
.Nm mtx_owned ,
.Nm mtx_notowned ,
.Nm mtx_lockrefs ,
.Nm mtx_hold ,
.Nm mtx_drop
.Nd general blocking/spinnable mutex functions
.Sh SYNOPSIS
.In sys/globaldata.h
.In sys/mutex2.h
.Ft void
.Fn mtx_init "struct mtx *mtx"
.Ft void
.Fn mtx_uninit "struct mtx *mtx"
.Ft void
.Fn mtx_lock_sh "struct mtx *mtx" "const char *ident" "int flags" "int to"
.Ft void
.Fn mtx_lock_sh_quick "struct mtx *mtx" "const char *ident"
.Ft void
.Fn mtx_lock_ex "struct mtx *mtx" "const char *ident" "int flags" "int to"
.Ft void
.Fn mtx_lock_ex_quick "struct mtx *mtx" "const char *ident"
.Ft void
.Fn mtx_lock "struct mtx *mtx"
.Ft void
.Fn mtx_spinlock "struct mtx *mtx"
.Ft int
.Fn mtx_lock_ex_try "struct mtx *mtx"
.Ft int
.Fn mtx_lock_sh_try "struct mtx *mtx"
.Ft int
.Fn mtx_spinlock_try "struct mtx *mtx"
.Ft void
.Fn mtx_downgrade "struct mtx *mtx"
.Ft int
.Fn mtx_upgrade_try "struct mtx *mtx"
.Ft void
.Fn mtx_unlock "struct mtx *mtx"
.Ft void
.Fn mtx_unlock_ex "struct mtx *mtx"
.Ft void
.Fn mtx_unlock_sh "struct mtx *mtx"
.Ft void
.Fn mtx_spinunlock "struct mtx *mtx"
.Ft int
.Fn mtx_islocked "struct mtx *mtx"
.Ft int
.Fn mtx_islocked_ex "struct mtx *mtx"
.Ft int
.Fn mtx_notlocked "struct mtx *mtx"
.Ft int
.Fn mtx_notlocked_ex "struct mtx *mtx"
.Ft int
.Fn mtx_owned "struct mtx *mtx"
.Ft int
.Fn mtx_notowned "struct mtx *mtx"
.Ft int
.Fn mtx_lockrefs "struct mtx *mtx"
.Ft void
.Fn mtx_hold "struct mtx *mtx"
.Ft int
.Fn mtx_drop "struct mtx *mtx"
.Sh DESCRIPTION
Mutexes are used to implement mutual exclusion between threads.
Mutexes can be locked in shared or exclusive mode; they can also block
or spin the current thread when there is contention.
.Pp
Mutexes also have an associated reference count, independent of the lock.
.Pp
System-wide mutex contention statistics can be found in the
.Va kern.mtx_contention_count ,
.Va kern.mtx_collision_count ,
and
.Va kern.mtx_wakeup_count
variables.
.Va kern.mtx_contention_count
is incremented each time an attempt to acquire a mutex fails due to contention.
.Va kern.mtx_wakeup_count
is incremented each time an exclusive lock is converted to either a shared or
unlocked state an waiters for the shared state are woken.
.Pp
The mutex functions are similar to the
.Xr lockmgr 9
functions.
.Sh FUNCTIONS
The
.Fn mtx_init
function initializes a mutex to the unlocked state.
It is an error to use a mutex without initializing it.
.Pp
The
.Fn mtx_uninit
function deinitializes a mutex.
.Pp
The
.Fn mtx_lock_sh
function attempts to lock a mutex in shared mode and blocks the current
thread until it is able to do so.
The
.Fa ident
parameter is as in
.Xr tsleep 9 ,
it is a string describing the reason for a thread to be blocked.
The
.Fa flags
parameter is passed to
.Xr tsleep 9
if the thread must block; the
.Fa to
parameter is a timeout for the sleep.
The
.Fn mtx_lock_sh_quick
function is a version of
.Fn mtx_lock_sh
without flags or a timeout.
.Pp
The
.Fn mtx_lock_ex
function attempts to lock a mutex exclusively and blocks the current thread
until it is able to do so.
The
.Fa ident
and
.Fa flags
parameters are as in
.Xr tsleep 9 .
The
.Fa to
parameter is a timeout on the sleep.
The
.Fn mtx_lock_ex_quick
function is a version of
.Fn mtx_lock_ex
without flags or a timeout.
The
.Fn mtx_lock
function is a yet shorter form for exclusively locking a mutex, blocking the
current thread until acquired.
It is equivalent to mtx_lock_ex(mtx, "mtxex", 0, 0).
.Pp
The
.Fn mtx_spinlock
function attempts to lock the mutex in exclusive mode and spins until it is
able to do so.
.Pp
The
.Fn mtx_lock_ex_try
and
.Fn mtx_lock_sh_try
functions attempt to lock the mutex in exclusive or shared mode, respectively.
If they are not able to, they return
.Er EAGAIN .
The
.Fn mtx_spinlock_try
function does the same but for spin mutexes.
.Pp
The
.Fn mtx_downgrade
function converts an exclusively held lock to a shared lock.
The lock must be held by the calling thread.
If the lock is already shared, this call is a no-op.
.Pp
The
.Fn mtx_upgrade_try
function attempts to convert a shared lock to an exclusive one.
The mutex must be held by the caller in the shared state.
If the upgrade is successful, this function returns 0; otherwise, it returns
.Er EDEADLK .
.Pp
The
.Fn mtx_unlock
function releases a held mutex;
it works on both exclusive and shared mutexes.
The
.Fn mtx_unlock_ex
and
.Fn mtx_unlock_sh
functions are optimized unlock paths, used when it is known that a lock is held
exclusively or in shared state.
.Pp
The
.Fn mtx_spinunlock
function releases a held spin mutex.
.Pp
The
.Fn mtx_islocked
function returns non-zero if the mutex is locked in either shared of
exclusive state by any thread.
.Fn mtx_islocked_ex
returns non-zero if the mutex is locked exclusively by any thread.
The
.Fn mtx_notlocked
function returns non-zero if the mutex is not locked.
The
.Fn mtx_owned
function returns non-zero if the mutex is exclusively locked by the calling
thread.
The
.Fn mtx_notowned
function returns non-zero if the mutex is not exclusively locked by the
calling thread.
The
.Fn mtx_lockrefs
function returns the number of shared or exclusive locks on the mutex.
.Pp
The
.Fn mtx_hold
function increments the reference count associated with each mutex.
The reference count is independent of the lock field.
The
.Fn mtx_drop
function decrements the reference count associated with each mutex
and returns the old value of the count.
A return value of
.Sq 1
means that the current count is
.Sq 0 .
.Sh FILES
The uncontended path of the
.Nm
implementation is in
.Pa /sys/sys/mutex2.h .
The data structures are in
.Pa /sys/sys/mutex.h .
The core of the spinlock implementation is in
.Pa /sys/kern/kern_mutex.c .
.Sh SEE ALSO
.Xr crit_enter 9 ,
.Xr lockmgr 9 ,
.Xr serializer 9 ,
.Xr sleep 9 ,
.Xr spinlock 9
.Sh HISTORY
Mutexes first appeared in
.Dx 2.3 .
.Sh AUTHORS
.An -nosplit
The
.Nm
implementation was written by
.An Matthew Dillon .