1.\" 2.\" Copyright (C) 2001 Jason Evans <jasone@FreeBSD.org>. All rights reserved. 3.\" 4.\" Redistribution and use in source and binary forms, with or without 5.\" modification, are permitted provided that the following conditions 6.\" are met: 7.\" 1. Redistributions of source code must retain the above copyright 8.\" notice(s), this list of conditions and the following disclaimer as 9.\" the first lines of this file unmodified other than the possible 10.\" addition of one or more copyright notices. 11.\" 2. Redistributions in binary form must reproduce the above copyright 12.\" notice(s), this list of conditions and the following disclaimer in the 13.\" documentation and/or other materials provided with the distribution. 14.\" 15.\" THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDER(S) ``AS IS'' AND ANY 16.\" EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED 17.\" WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE 18.\" DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER(S) BE LIABLE FOR ANY 19.\" DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES 20.\" (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR 21.\" SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER 22.\" CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 23.\" LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 24.\" OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH 25.\" DAMAGE. 26.\" 27.\" $FreeBSD$ 28.\" 29.Dd February 1, 2006 30.Dt SX 9 31.Os 32.Sh NAME 33.Nm sx , 34.Nm sx_init , 35.Nm sx_destroy , 36.Nm sx_slock , 37.Nm sx_xlock , 38.Nm sx_try_slock , 39.Nm sx_try_xlock , 40.Nm sx_sunlock , 41.Nm sx_xunlock , 42.Nm sx_try_upgrade , 43.Nm sx_downgrade , 44.Nm sx_assert , 45.Nm sx_unlock , 46.Nm SX_SYSINIT 47.Nd kernel shared/exclusive lock 48.Sh SYNOPSIS 49.In sys/param.h 50.In sys/lock.h 51.In sys/sx.h 52.Ft void 53.Fn sx_init "struct sx *sx" "const char *description" 54.Ft void 55.Fn sx_destroy "struct sx *sx" 56.Ft void 57.Fn sx_slock "struct sx *sx" 58.Ft void 59.Fn sx_xlock "struct sx *sx" 60.Ft int 61.Fn sx_try_slock "struct sx *sx" 62.Ft int 63.Fn sx_try_xlock "struct sx *sx" 64.Ft void 65.Fn sx_sunlock "struct sx *sx" 66.Ft void 67.Fn sx_xunlock "struct sx *sx" 68.Ft int 69.Fn sx_try_upgrade "struct sx *sx" 70.Ft void 71.Fn sx_downgrade "struct sx *sx" 72.Ft void 73.Fn sx_assert "struct sx *sx" "int what" 74.\" 75.Ss Nm Ss utility macros 76.Fn sx_unlock "struct sx *sx" 77.Fn SX_SYSINIT "name" "struct sx *sx" "const char *description" 78.\" 79.Ss Kernel options 80.Cd "options INVARIANTS" 81.Cd "options INVARIANT_SUPPORT" 82.Sh DESCRIPTION 83Shared/exclusive locks are used to protect data that are read far more often 84than they are written. 85Mutexes are inherently more efficient than shared/exclusive locks, so 86shared/exclusive locks should be used prudently. 87.Pp 88Shared/exclusive locks are created with 89.Fn sx_init , 90where 91.Fa sx 92is a pointer to space for a 93.Vt struct sx , 94and 95.Fa description 96is a pointer to a null-terminated character string that describes the 97shared/exclusive lock. 98Shared/exclusive locks are destroyed with 99.Fn sx_destroy . 100Threads acquire and release a shared lock by calling 101.Fn sx_slock 102or 103.Fn sx_try_slock 104and 105.Fn sx_sunlock 106or 107.Fn sx_unlock . 108Threads acquire and release an exclusive lock by calling 109.Fn sx_xlock 110or 111.Fn sx_try_xlock 112and 113.Fn sx_xunlock 114or 115.Fn sx_unlock . 116A thread can attempt to upgrade a currently held shared lock to an exclusive 117lock by calling 118.Fn sx_try_upgrade . 119A thread that has an exclusive lock can downgrade it to a shared lock by 120calling 121.Fn sx_downgrade . 122.Pp 123.Fn sx_try_slock 124and 125.Fn sx_try_xlock 126will return 0 if the shared/exclusive lock cannot be acquired immediately; 127otherwise the shared/exclusive lock will be acquired and a non-zero value will 128be returned. 129.Pp 130.Fn sx_try_upgrade 131will return 0 if the shared lock cannot be upgraded to an exclusive lock 132immediately; otherwise the exclusive lock will be acquired and a non-zero value 133will be returned. 134.Pp 135When compiled with 136.Cd "options INVARIANTS" 137and 138.Cd "options INVARIANT_SUPPORT" , 139the 140.Fn sx_assert 141function tests 142.Fa sx 143for the assertions specified in 144.Fa what , 145and panics if they are not met. 146The following assertions are supported: 147.Bl -tag -width ".Dv SX_UNLOCKED" 148.It Dv SX_LOCKED 149Assert that the current thread has either a shared or an exclusive lock on the 150.Vt sx 151lock pointed to by the first argument. 152.It Dv SX_SLOCKED 153Assert that the current thread has a shared lock on the 154.Vt sx 155lock pointed to by 156the first argument. 157.It Dv SX_XLOCKED 158Assert that the current thread has an exclusive lock on the 159.Vt sx 160lock pointed to 161by the first argument. 162.It Dv SX_UNLOCKED 163Assert that the current thread has no lock on the 164.Vt sx 165lock pointed to 166by the first argument. 167.El 168.Pp 169For ease of programming, 170.Fn sx_unlock 171is provided as a macro frontend to the respective functions, 172.Fn sx_sunlock 173and 174.Fn sx_xunlock . 175Algorithms that are aware of what state the lock is in should use either 176of the two specific functions for a minor performance benefit. 177.Pp 178The 179.Fn SX_SYSINIT 180macro is used to generate a call to the 181.Fn sx_sysinit 182routine at system startup in order to initialize a given 183.Fa sx 184lock. 185The parameters are the same as 186.Fn sx_init 187but with an additional argument, 188.Fa name , 189that is used in generating unique variable names for the related 190structures associated with the lock and the sysinit routine. 191.Pp 192A thread may not hold both a shared lock and an exclusive lock on the same 193lock simultaneously; 194attempting to do so will result in deadlock. 195.Sh CONTEXT 196A thread may hold a shared or exclusive lock on an 197.Nm 198lock while sleeping. 199As a result, an 200.Nm 201lock may not be acquired while holding a mutex. 202Otherwise, if one thread slept while holding an 203.Nm 204lock while another thread blocked on the same 205.Nm 206lock after acquiring a mutex, then the second thread would effectively 207end up sleeping while holding a mutex, which is not allowed. 208.Sh SEE ALSO 209.Xr mutex 9 , 210.Xr panic 9 , 211.Xr rwlock 9 , 212.Xr sema 9 213.Sh BUGS 214Currently there is no way to assert that a lock is not held. 215This is not possible in the 216.No non- Ns Dv WITNESS 217case for asserting that this thread 218does not hold a shared lock. 219In the 220.No non- Ns Dv WITNESS 221case, the 222.Dv SX_LOCKED 223and 224.Dv SX_SLOCKED 225assertions merely check that some thread holds a shared lock. 226They do not ensure that the current thread holds a shared lock. 227