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 November 11, 2017 30.Dt SX 9 31.Os 32.Sh NAME 33.Nm sx , 34.Nm sx_init , 35.Nm sx_init_flags , 36.Nm sx_destroy , 37.Nm sx_slock , 38.Nm sx_xlock , 39.Nm sx_slock_sig , 40.Nm sx_xlock_sig , 41.Nm sx_try_slock , 42.Nm sx_try_xlock , 43.Nm sx_sunlock , 44.Nm sx_xunlock , 45.Nm sx_unlock , 46.Nm sx_try_upgrade , 47.Nm sx_downgrade , 48.Nm sx_sleep , 49.Nm sx_xholder , 50.Nm sx_xlocked , 51.Nm sx_assert , 52.Nm SX_SYSINIT , 53.Nm SX_SYSINIT_FLAGS 54.Nd kernel shared/exclusive lock 55.Sh SYNOPSIS 56.In sys/param.h 57.In sys/lock.h 58.In sys/sx.h 59.Ft void 60.Fn sx_init "struct sx *sx" "const char *description" 61.Ft void 62.Fn sx_init_flags "struct sx *sx" "const char *description" "int opts" 63.Ft void 64.Fn sx_destroy "struct sx *sx" 65.Ft void 66.Fn sx_slock "struct sx *sx" 67.Ft void 68.Fn sx_xlock "struct sx *sx" 69.Ft int 70.Fn sx_slock_sig "struct sx *sx" 71.Ft int 72.Fn sx_xlock_sig "struct sx *sx" 73.Ft int 74.Fn sx_try_slock "struct sx *sx" 75.Ft int 76.Fn sx_try_xlock "struct sx *sx" 77.Ft void 78.Fn sx_sunlock "struct sx *sx" 79.Ft void 80.Fn sx_xunlock "struct sx *sx" 81.Ft void 82.Fn sx_unlock "struct sx *sx" 83.Ft int 84.Fn sx_try_upgrade "struct sx *sx" 85.Ft void 86.Fn sx_downgrade "struct sx *sx" 87.Ft int 88.Fn sx_sleep "void *chan" "struct sx *sx" "int priority" "const char *wmesg" "int timo" 89.Ft "struct thread *" 90.Fn sx_xholder "struct sx *sx" 91.Ft int 92.Fn sx_xlocked "const struct sx *sx" 93.Pp 94.Cd "options INVARIANTS" 95.Cd "options INVARIANT_SUPPORT" 96.Ft void 97.Fn sx_assert "const struct sx *sx" "int what" 98.In sys/kernel.h 99.Fn SX_SYSINIT "name" "struct sx *sx" "const char *desc" 100.Fn SX_SYSINIT_FLAGS "name" "struct sx *sx" "const char *desc" "int flags" 101.Sh DESCRIPTION 102Shared/exclusive locks are used to protect data that are read far more often 103than they are written. 104Shared/exclusive locks do not implement priority propagation like mutexes and 105reader/writer locks to prevent priority inversions, so 106shared/exclusive locks should be used prudently. 107.Pp 108Shared/exclusive locks are created with either 109.Fn sx_init 110or 111.Fn sx_init_flags 112where 113.Fa sx 114is a pointer to space for a 115.Vt struct sx , 116and 117.Fa description 118is a pointer to a null-terminated character string that describes the 119shared/exclusive lock. 120The 121.Fa opts 122argument to 123.Fn sx_init_flags 124specifies a set of optional flags to alter the behavior of 125.Fa sx . 126It contains one or more of the following flags: 127.Bl -tag -width SX_NOADAPTIVE 128.It Dv SX_NOADAPTIVE 129Disable adaptive spinning, rather than sleeping, for lock operations 130while an exclusive lock holder is executing on another CPU. 131Adaptive spinning is the default unless the kernel is compiled with 132.Cd "options NO_ADAPTIVE_SX" . 133.It Dv SX_DUPOK 134Witness should not log messages about duplicate locks being acquired. 135.It Dv SX_NOWITNESS 136Instruct 137.Xr witness 4 138to ignore this lock. 139.It Dv SX_NOPROFILE 140Do not profile this lock. 141.It Dv SX_RECURSE 142Allow threads to recursively acquire exclusive locks for 143.Fa sx . 144.It Dv SX_QUIET 145Do not log any operations for this lock via 146.Xr ktr 4 . 147.It Dv SX_NEW 148If the kernel has been compiled with 149.Cd "options INVARIANTS" , 150.Fn sx_init 151will assert that the 152.Fa sx 153has not been initialized multiple times without intervening calls to 154.Fn sx_destroy 155unless this option is specified. 156.El 157.Pp 158Shared/exclusive locks are destroyed with 159.Fn sx_destroy . 160The lock 161.Fa sx 162must not be locked by any thread when it is destroyed. 163.Pp 164Threads acquire and release a shared lock by calling 165.Fn sx_slock , 166.Fn sx_slock_sig 167or 168.Fn sx_try_slock 169and 170.Fn sx_sunlock 171or 172.Fn sx_unlock . 173Threads acquire and release an exclusive lock by calling 174.Fn sx_xlock , 175.Fn sx_xlock_sig 176or 177.Fn sx_try_xlock 178and 179.Fn sx_xunlock 180or 181.Fn sx_unlock . 182A thread can attempt to upgrade a currently held shared lock to an exclusive 183lock by calling 184.Fn sx_try_upgrade . 185A thread that has an exclusive lock can downgrade it to a shared lock by 186calling 187.Fn sx_downgrade . 188.Pp 189.Fn sx_try_slock 190and 191.Fn sx_try_xlock 192will return 0 if the shared/exclusive lock cannot be acquired immediately; 193otherwise the shared/exclusive lock will be acquired and a non-zero value will 194be returned. 195.Pp 196.Fn sx_try_upgrade 197will return 0 if the shared lock cannot be upgraded to an exclusive lock 198immediately; otherwise the exclusive lock will be acquired and a non-zero value 199will be returned. 200.Pp 201.Fn sx_slock_sig 202and 203.Fn sx_xlock_sig 204do the same as their normal versions but performing an interruptible sleep. 205They return a non-zero value if the sleep has been interrupted by a signal 206or an interrupt, otherwise 0. 207.Pp 208A thread can atomically release a shared/exclusive lock while waiting for an 209event by calling 210.Fn sx_sleep . 211For more details on the parameters to this function, 212see 213.Xr sleep 9 . 214.Pp 215When compiled with 216.Cd "options INVARIANTS" 217and 218.Cd "options INVARIANT_SUPPORT" , 219the 220.Fn sx_assert 221function tests 222.Fa sx 223for the assertions specified in 224.Fa what , 225and panics if they are not met. 226One of the following assertions must be specified: 227.Bl -tag -width ".Dv SA_UNLOCKED" 228.It Dv SA_LOCKED 229Assert that the current thread has either a shared or an exclusive lock on the 230.Vt sx 231lock pointed to by the first argument. 232.It Dv SA_SLOCKED 233Assert that the current thread has a shared lock on the 234.Vt sx 235lock pointed to by 236the first argument. 237.It Dv SA_XLOCKED 238Assert that the current thread has an exclusive lock on the 239.Vt sx 240lock pointed to 241by the first argument. 242.It Dv SA_UNLOCKED 243Assert that the current thread has no lock on the 244.Vt sx 245lock pointed to 246by the first argument. 247.El 248.Pp 249In addition, one of the following optional assertions may be included with 250either an 251.Dv SA_LOCKED , 252.Dv SA_SLOCKED , 253or 254.Dv SA_XLOCKED 255assertion: 256.Bl -tag -width ".Dv SA_NOTRECURSED" 257.It Dv SA_RECURSED 258Assert that the current thread has a recursed lock on 259.Fa sx . 260.It Dv SA_NOTRECURSED 261Assert that the current thread does not have a recursed lock on 262.Fa sx . 263.El 264.Pp 265.Fn sx_xholder 266will return a pointer to the thread which currently holds an exclusive lock on 267.Fa sx . 268If no thread holds an exclusive lock on 269.Fa sx , 270then 271.Dv NULL 272is returned instead. 273.Pp 274.Fn sx_xlocked 275will return non-zero if the current thread holds the exclusive lock; 276otherwise, it will return zero. 277.Pp 278For ease of programming, 279.Fn sx_unlock 280is provided as a macro frontend to the respective functions, 281.Fn sx_sunlock 282and 283.Fn sx_xunlock . 284Algorithms that are aware of what state the lock is in should use either 285of the two specific functions for a minor performance benefit. 286.Pp 287The 288.Fn SX_SYSINIT 289macro is used to generate a call to the 290.Fn sx_sysinit 291routine at system startup in order to initialize a given 292.Fa sx 293lock. 294The parameters are the same as 295.Fn sx_init 296but with an additional argument, 297.Fa name , 298that is used in generating unique variable names for the related 299structures associated with the lock and the sysinit routine. 300The 301.Fn SX_SYSINIT_FLAGS 302macro can similarly be used to initialize a given 303.Fa sx 304lock using 305.Fn sx_init_flags . 306.Pp 307A thread may not hold both a shared lock and an exclusive lock on the same 308lock simultaneously; 309attempting to do so will result in deadlock. 310.Sh CONTEXT 311A thread may hold a shared or exclusive lock on an 312.Nm 313lock while sleeping. 314As a result, an 315.Nm 316lock may not be acquired while holding a mutex. 317Otherwise, if one thread slept while holding an 318.Nm 319lock while another thread blocked on the same 320.Nm 321lock after acquiring a mutex, then the second thread would effectively 322end up sleeping while holding a mutex, which is not allowed. 323.Sh SEE ALSO 324.Xr lock 9 , 325.Xr locking 9 , 326.Xr mutex 9 , 327.Xr panic 9 , 328.Xr rwlock 9 , 329.Xr sema 9 330.Sh BUGS 331A kernel without 332.Dv WITNESS 333cannot assert whether the current thread does or does not hold a shared lock. 334.Dv SA_LOCKED 335and 336.Dv SA_SLOCKED 337can only assert that 338.Em any 339thread holds a shared lock. 340They cannot ensure that the current thread holds a shared lock. 341Further, 342.Dv SA_UNLOCKED 343can only assert that the current thread does not hold an exclusive lock. 344