freebsd-skq/share/man/man9/sx.9
attilio e05714ba70 Reverse the logic for ADAPTIVE_SX option and enable it by default.
Introduce for this operation the reverse NO_ADAPTIVE_SX option.
The flag SX_ADAPTIVESPIN to be passed to sx_init_flags(9) gets suppressed
and the new flag, offering the reversed logic, SX_NOADAPTIVE is added.

Additively implements adaptive spininning for sx held in shared mode.
The spinning limit can be handled through sysctls in order to be tuned
while the code doesn't reach the release, after which time they should
be dropped probabilly.

This change has made been necessary by recent benchmarks where it does
improve concurrency of workloads in presence of high contention
(ie. ZFS).

KPI breakage is documented by __FreeBSD_version bumping, manpage and
UPDATING updates.

Requested by:	jeff, kmacy
Reviewed by:	jeff
Tested by:	pho
2009-05-29 01:49:27 +00:00

329 lines
8.7 KiB
Groff

.\"
.\" Copyright (C) 2001 Jason Evans <jasone@FreeBSD.org>. 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(s), this list of conditions and the following disclaimer as
.\" the first lines of this file unmodified other than the possible
.\" addition of one or more copyright notices.
.\" 2. Redistributions in binary form must reproduce the above copyright
.\" notice(s), this list of conditions and the following disclaimer in the
.\" documentation and/or other materials provided with the distribution.
.\"
.\" THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDER(S) ``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 HOLDER(S) 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.
.\"
.\" $FreeBSD$
.\"
.Dd May 28, 2009
.Dt SX 9
.Os
.Sh NAME
.Nm sx ,
.Nm sx_init ,
.Nm sx_init_flags ,
.Nm sx_destroy ,
.Nm sx_slock ,
.Nm sx_xlock ,
.Nm sx_slock_sig ,
.Nm sx_xlock_sig ,
.Nm sx_try_slock ,
.Nm sx_try_xlock ,
.Nm sx_sunlock ,
.Nm sx_xunlock ,
.Nm sx_unlock ,
.Nm sx_try_upgrade ,
.Nm sx_downgrade ,
.Nm sx_sleep ,
.Nm sx_xholder ,
.Nm sx_xlocked ,
.Nm sx_assert ,
.Nm SX_SYSINIT
.Nd kernel shared/exclusive lock
.Sh SYNOPSIS
.In sys/param.h
.In sys/lock.h
.In sys/sx.h
.Ft void
.Fn sx_init "struct sx *sx" "const char *description"
.Ft void
.Fn sx_init_flags "struct sx *sx" "const char *description" "int opts"
.Ft void
.Fn sx_destroy "struct sx *sx"
.Ft void
.Fn sx_slock "struct sx *sx"
.Ft void
.Fn sx_xlock "struct sx *sx"
.Ft int
.Fn sx_slock_sig "struct sx *sx"
.Ft int
.Fn sx_xlock_sig "struct sx *sx"
.Ft int
.Fn sx_try_slock "struct sx *sx"
.Ft int
.Fn sx_try_xlock "struct sx *sx"
.Ft void
.Fn sx_sunlock "struct sx *sx"
.Ft void
.Fn sx_xunlock "struct sx *sx"
.Ft void
.Fn sx_unlock "struct sx *sx"
.Ft int
.Fn sx_try_upgrade "struct sx *sx"
.Ft void
.Fn sx_downgrade "struct sx *sx"
.Ft int
.Fn sx_sleep "void *chan" "struct sx *sx" "int priority" "const char *wmesg" "int timo"
.Ft "struct thread *"
.Fn sx_xholder "struct sx *sx"
.Ft int
.Fn sx_xlocked "struct sx *sx"
.Pp
.Cd "options INVARIANTS"
.Cd "options INVARIANT_SUPPORT"
.Ft void
.Fn sx_assert "struct sx *sx" "int what"
.In sys/kernel.h
.Fn SX_SYSINIT "name" "struct sx *sx" "const char *description"
.Sh DESCRIPTION
Shared/exclusive locks are used to protect data that are read far more often
than they are written.
Shared/exclusive locks do not implement priority propagation like mutexes and
reader/writer locks to prevent priority inversions, so
shared/exclusive locks should be used prudently.
.Pp
Shared/exclusive locks are created with either
.Fn sx_init
or
.Fn sx_init_flags
where
.Fa sx
is a pointer to space for a
.Vt struct sx ,
and
.Fa description
is a pointer to a null-terminated character string that describes the
shared/exclusive lock.
The
.Fa opts
argument to
.Fn sx_init_flags
specifies a set of optional flags to alter the behavior of
.Fa sx .
It contains one or more of the following flags:
.Bl -tag -width SX_NOADAPTIVE
.It Dv SX_NOADAPTIVE
If the kernel is not compiled with
.Cd "options NO_ADAPTIVE_SX" ,
then lock operations for
.Fa sx
will spin instead of sleeping while an exclusive lock holder is executing on
another CPU.
.It Dv SX_DUPOK
Witness should not log messages about duplicate locks being acquired.
.It Dv SX_NOWITNESS
Instruct
.Xr witness 4
to ignore this lock.
.It Dv SX_NOPROFILE
Do not profile this lock.
.It Dv SX_RECURSE
Allow threads to recursively acquire exclusive locks for
.Fa sx .
.It Dv SX_QUIET
Do not log any operations for this lock via
.Xr ktr 4 .
.El
.Pp
Shared/exclusive locks are destroyed with
.Fn sx_destroy .
The lock
.Fa sx
must not be locked by any thread when it is destroyed.
.Pp
Threads acquire and release a shared lock by calling
.Fn sx_slock ,
.Fn sx_slock_sig
or
.Fn sx_try_slock
and
.Fn sx_sunlock
or
.Fn sx_unlock .
Threads acquire and release an exclusive lock by calling
.Fn sx_xlock ,
.Fn sx_xlock_sig
or
.Fn sx_try_xlock
and
.Fn sx_xunlock
or
.Fn sx_unlock .
A thread can attempt to upgrade a currently held shared lock to an exclusive
lock by calling
.Fn sx_try_upgrade .
A thread that has an exclusive lock can downgrade it to a shared lock by
calling
.Fn sx_downgrade .
.Pp
.Fn sx_try_slock
and
.Fn sx_try_xlock
will return 0 if the shared/exclusive lock cannot be acquired immediately;
otherwise the shared/exclusive lock will be acquired and a non-zero value will
be returned.
.Pp
.Fn sx_try_upgrade
will return 0 if the shared lock cannot be upgraded to an exclusive lock
immediately; otherwise the exclusive lock will be acquired and a non-zero value
will be returned.
.Pp
.Fn sx_slock_sig
and
.Fn sx_xlock_sig
do the same as their normal versions but performing an interruptible sleep.
They return a non-zero value if the sleep has been interrupted by a signal
or an interrupt, otherwise 0.
.Pp
A thread can atomically release a shared/exclusive lock while waiting for an
event by calling
.Fn sx_sleep .
For more details on the parameters to this function,
see
.Xr sleep 9 .
.Pp
When compiled with
.Cd "options INVARIANTS"
and
.Cd "options INVARIANT_SUPPORT" ,
the
.Fn sx_assert
function tests
.Fa sx
for the assertions specified in
.Fa what ,
and panics if they are not met.
One of the following assertions must be specified:
.Bl -tag -width ".Dv SA_UNLOCKED"
.It Dv SA_LOCKED
Assert that the current thread has either a shared or an exclusive lock on the
.Vt sx
lock pointed to by the first argument.
.It Dv SA_SLOCKED
Assert that the current thread has a shared lock on the
.Vt sx
lock pointed to by
the first argument.
.It Dv SA_XLOCKED
Assert that the current thread has an exclusive lock on the
.Vt sx
lock pointed to
by the first argument.
.It Dv SA_UNLOCKED
Assert that the current thread has no lock on the
.Vt sx
lock pointed to
by the first argument.
.El
.Pp
In addition, one of the following optional assertions may be included with
either an
.Dv SA_LOCKED ,
.Dv SA_SLOCKED ,
or
.Dv SA_XLOCKED
assertion:
.Bl -tag -width ".Dv SA_NOTRECURSED"
.It Dv SA_RECURSED
Assert that the current thread has a recursed lock on
.Fa sx .
.It Dv SA_NOTRECURSED
Assert that the current thread does not have a recursed lock on
.Fa sx .
.El
.Pp
.Fn sx_xholder
will return a pointer to the thread which currently holds an exclusive lock on
.Fa sx .
If no thread holds an exclusive lock on
.Fa sx ,
then
.Dv NULL
is returned instead.
.Pp
.Fn sx_xlocked
will return non-zero if the current thread holds the exclusive lock;
otherwise, it will return zero.
.Pp
For ease of programming,
.Fn sx_unlock
is provided as a macro frontend to the respective functions,
.Fn sx_sunlock
and
.Fn sx_xunlock .
Algorithms that are aware of what state the lock is in should use either
of the two specific functions for a minor performance benefit.
.Pp
The
.Fn SX_SYSINIT
macro is used to generate a call to the
.Fn sx_sysinit
routine at system startup in order to initialize a given
.Fa sx
lock.
The parameters are the same as
.Fn sx_init
but with an additional argument,
.Fa name ,
that is used in generating unique variable names for the related
structures associated with the lock and the sysinit routine.
.Pp
A thread may not hold both a shared lock and an exclusive lock on the same
lock simultaneously;
attempting to do so will result in deadlock.
.Sh CONTEXT
A thread may hold a shared or exclusive lock on an
.Nm
lock while sleeping.
As a result, an
.Nm
lock may not be acquired while holding a mutex.
Otherwise, if one thread slept while holding an
.Nm
lock while another thread blocked on the same
.Nm
lock after acquiring a mutex, then the second thread would effectively
end up sleeping while holding a mutex, which is not allowed.
.Sh SEE ALSO
.Xr locking 9 ,
.Xr lock 9 ,
.Xr mutex 9 ,
.Xr panic 9 ,
.Xr rwlock 9 ,
.Xr sema 9
.Sh BUGS
Currently there is no way to assert that a lock is not held.
This is not possible in the
.No non- Ns Dv WITNESS
case for asserting that this thread
does not hold a shared lock.
In the
.No non- Ns Dv WITNESS
case, the
.Dv SA_LOCKED
and
.Dv SA_SLOCKED
assertions merely check that some thread holds a shared lock.
They do not ensure that the current thread holds a shared lock.