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.Dd November 11, 2017 28.Dt SX 9 29.Os 30.Sh NAME 31.Nm sx , 32.Nm sx_init , 33.Nm sx_init_flags , 34.Nm sx_destroy , 35.Nm sx_slock , 36.Nm sx_xlock , 37.Nm sx_slock_sig , 38.Nm sx_xlock_sig , 39.Nm sx_try_slock , 40.Nm sx_try_xlock , 41.Nm sx_sunlock , 42.Nm sx_xunlock , 43.Nm sx_unlock , 44.Nm sx_try_upgrade , 45.Nm sx_downgrade , 46.Nm sx_sleep , 47.Nm sx_xholder , 48.Nm sx_xlocked , 49.Nm sx_assert , 50.Nm SX_SYSINIT , 51.Nm SX_SYSINIT_FLAGS 52.Nd kernel shared/exclusive lock 53.Sh SYNOPSIS 54.In sys/param.h 55.In sys/lock.h 56.In sys/sx.h 57.Ft void 58.Fn sx_init "struct sx *sx" "const char *description" 59.Ft void 60.Fn sx_init_flags "struct sx *sx" "const char *description" "int opts" 61.Ft void 62.Fn sx_destroy "struct sx *sx" 63.Ft void 64.Fn sx_slock "struct sx *sx" 65.Ft void 66.Fn sx_xlock "struct sx *sx" 67.Ft int 68.Fn sx_slock_sig "struct sx *sx" 69.Ft int 70.Fn sx_xlock_sig "struct sx *sx" 71.Ft int 72.Fn sx_try_slock "struct sx *sx" 73.Ft int 74.Fn sx_try_xlock "struct sx *sx" 75.Ft void 76.Fn sx_sunlock "struct sx *sx" 77.Ft void 78.Fn sx_xunlock "struct sx *sx" 79.Ft void 80.Fn sx_unlock "struct sx *sx" 81.Ft int 82.Fn sx_try_upgrade "struct sx *sx" 83.Ft void 84.Fn sx_downgrade "struct sx *sx" 85.Ft int 86.Fn sx_sleep "void *chan" "struct sx *sx" "int priority" "const char *wmesg" "int timo" 87.Ft "struct thread *" 88.Fn sx_xholder "struct sx *sx" 89.Ft int 90.Fn sx_xlocked "const struct sx *sx" 91.Pp 92.Cd "options INVARIANTS" 93.Cd "options INVARIANT_SUPPORT" 94.Ft void 95.Fn sx_assert "const struct sx *sx" "int what" 96.In sys/kernel.h 97.Fn SX_SYSINIT "name" "struct sx *sx" "const char *desc" 98.Fn SX_SYSINIT_FLAGS "name" "struct sx *sx" "const char *desc" "int flags" 99.Sh DESCRIPTION 100Shared/exclusive locks are used to protect data that are read far more often 101than they are written. 102Shared/exclusive locks do not implement priority propagation like mutexes and 103reader/writer locks to prevent priority inversions, so 104shared/exclusive locks should be used prudently. 105.Pp 106Shared/exclusive locks are created with either 107.Fn sx_init 108or 109.Fn sx_init_flags 110where 111.Fa sx 112is a pointer to space for a 113.Vt struct sx , 114and 115.Fa description 116is a pointer to a null-terminated character string that describes the 117shared/exclusive lock. 118The 119.Fa opts 120argument to 121.Fn sx_init_flags 122specifies a set of optional flags to alter the behavior of 123.Fa sx . 124It contains one or more of the following flags: 125.Bl -tag -width SX_NOWITNESS 126.It Dv SX_DUPOK 127Witness should not log messages about duplicate locks being acquired. 128.It Dv SX_NOWITNESS 129Instruct 130.Xr witness 4 131to ignore this lock. 132.It Dv SX_NOPROFILE 133Do not profile this lock. 134.It Dv SX_RECURSE 135Allow threads to recursively acquire exclusive locks for 136.Fa sx . 137.It Dv SX_QUIET 138Do not log any operations for this lock via 139.Xr ktr 4 . 140.It Dv SX_NEW 141If the kernel has been compiled with 142.Cd "options INVARIANTS" , 143.Fn sx_init 144will assert that the 145.Fa sx 146has not been initialized multiple times without intervening calls to 147.Fn sx_destroy 148unless this option is specified. 149.El 150.Pp 151Shared/exclusive locks are destroyed with 152.Fn sx_destroy . 153The lock 154.Fa sx 155must not be locked by any thread when it is destroyed. 156.Pp 157Threads acquire and release a shared lock by calling 158.Fn sx_slock , 159.Fn sx_slock_sig 160or 161.Fn sx_try_slock 162and 163.Fn sx_sunlock 164or 165.Fn sx_unlock . 166Threads acquire and release an exclusive lock by calling 167.Fn sx_xlock , 168.Fn sx_xlock_sig 169or 170.Fn sx_try_xlock 171and 172.Fn sx_xunlock 173or 174.Fn sx_unlock . 175A thread can attempt to upgrade a currently held shared lock to an exclusive 176lock by calling 177.Fn sx_try_upgrade . 178A thread that has an exclusive lock can downgrade it to a shared lock by 179calling 180.Fn sx_downgrade . 181.Pp 182.Fn sx_try_slock 183and 184.Fn sx_try_xlock 185will return 0 if the shared/exclusive lock cannot be acquired immediately; 186otherwise the shared/exclusive lock will be acquired and a non-zero value will 187be returned. 188.Pp 189.Fn sx_try_upgrade 190will return 0 if the shared lock cannot be upgraded to an exclusive lock 191immediately; otherwise the exclusive lock will be acquired and a non-zero value 192will be returned. 193.Pp 194.Fn sx_slock_sig 195and 196.Fn sx_xlock_sig 197do the same as their normal versions but performing an interruptible sleep. 198They return a non-zero value if the sleep has been interrupted by a signal 199or an interrupt, otherwise 0. 200.Pp 201A thread can atomically release a shared/exclusive lock while waiting for an 202event by calling 203.Fn sx_sleep . 204For more details on the parameters to this function, 205see 206.Xr sleep 9 . 207.Pp 208When compiled with 209.Cd "options INVARIANTS" 210and 211.Cd "options INVARIANT_SUPPORT" , 212the 213.Fn sx_assert 214function tests 215.Fa sx 216for the assertions specified in 217.Fa what , 218and panics if they are not met. 219One of the following assertions must be specified: 220.Bl -tag -width ".Dv SA_UNLOCKED" 221.It Dv SA_LOCKED 222Assert that the current thread has either a shared or an exclusive lock on the 223.Vt sx 224lock pointed to by the first argument. 225.It Dv SA_SLOCKED 226Assert that the current thread has a shared lock on the 227.Vt sx 228lock pointed to by 229the first argument. 230.It Dv SA_XLOCKED 231Assert that the current thread has an exclusive lock on the 232.Vt sx 233lock pointed to 234by the first argument. 235.It Dv SA_UNLOCKED 236Assert that the current thread has no lock on the 237.Vt sx 238lock pointed to 239by the first argument. 240.El 241.Pp 242In addition, one of the following optional assertions may be included with 243either an 244.Dv SA_LOCKED , 245.Dv SA_SLOCKED , 246or 247.Dv SA_XLOCKED 248assertion: 249.Bl -tag -width ".Dv SA_NOTRECURSED" 250.It Dv SA_RECURSED 251Assert that the current thread has a recursed lock on 252.Fa sx . 253.It Dv SA_NOTRECURSED 254Assert that the current thread does not have a recursed lock on 255.Fa sx . 256.El 257.Pp 258.Fn sx_xholder 259will return a pointer to the thread which currently holds an exclusive lock on 260.Fa sx . 261If no thread holds an exclusive lock on 262.Fa sx , 263then 264.Dv NULL 265is returned instead. 266.Pp 267.Fn sx_xlocked 268will return non-zero if the current thread holds the exclusive lock; 269otherwise, it will return zero. 270.Pp 271For ease of programming, 272.Fn sx_unlock 273is provided as a macro frontend to the respective functions, 274.Fn sx_sunlock 275and 276.Fn sx_xunlock . 277Algorithms that are aware of what state the lock is in should use either 278of the two specific functions for a minor performance benefit. 279.Pp 280The 281.Fn SX_SYSINIT 282macro is used to generate a call to the 283.Fn sx_sysinit 284routine at system startup in order to initialize a given 285.Fa sx 286lock. 287The parameters are the same as 288.Fn sx_init 289but with an additional argument, 290.Fa name , 291that is used in generating unique variable names for the related 292structures associated with the lock and the sysinit routine. 293The 294.Fn SX_SYSINIT_FLAGS 295macro can similarly be used to initialize a given 296.Fa sx 297lock using 298.Fn sx_init_flags . 299.Pp 300A thread may not hold both a shared lock and an exclusive lock on the same 301lock simultaneously; 302attempting to do so will result in deadlock. 303.Sh CONTEXT 304A thread may hold a shared or exclusive lock on an 305.Nm 306lock while sleeping. 307As a result, an 308.Nm 309lock may not be acquired while holding a mutex. 310Otherwise, if one thread slept while holding an 311.Nm 312lock while another thread blocked on the same 313.Nm 314lock after acquiring a mutex, then the second thread would effectively 315end up sleeping while holding a mutex, which is not allowed. 316.Sh SEE ALSO 317.Xr lock 9 , 318.Xr locking 9 , 319.Xr mutex 9 , 320.Xr panic 9 , 321.Xr rwlock 9 , 322.Xr sema 9 323.Sh BUGS 324A kernel without 325.Dv WITNESS 326cannot assert whether the current thread does or does not hold a shared lock. 327.Dv SA_LOCKED 328and 329.Dv SA_SLOCKED 330can only assert that 331.Em any 332thread holds a shared lock. 333They cannot ensure that the current thread holds a shared lock. 334Further, 335.Dv SA_UNLOCKED 336can only assert that the current thread does not hold an exclusive lock. 337