d/freebsd-skq
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
04114c5ee3
freebsd-skq/share/man/man9/lock.9
Attilio Rao 651175c9db Introduce support for adaptive spinning in lockmgr. 
Actually, as it did receive few tuning, the support is disabled by
default, but it can opt-in with the option ADAPTIVE_LOCKMGRS.
Due to the nature of lockmgrs, adaptive spinning needs to be
selectively enabled for any interested lockmgr.
The support is bi-directional, or, in other ways, it will work in both
cases if the lock is held in read or write way.  In particular, the
read path is passible of further tunning using the sysctls
debug.lockmgr.retries and debug.lockmgr.loops .  Ideally, such sysctls
should be axed or compiled out before release.

Addictionally note that adaptive spinning doesn't cope well with
LK_SLEEPFAIL.  The reason is that many (and probabilly all) consumers
of LK_SLEEPFAIL are mainly interested in knowing if the interlock was
dropped or not in order to reacquire it and re-test initial conditions.
This directly interacts with adaptive spinning because lockmgr needs
to drop the interlock while spinning in order to avoid a deadlock
(further details in the comments inside the patch).

Final note: finding someone willing to help on tuning this with
relevant workloads would be either very important and appreciated.

Tested by:	jeff, pho
Requested by:	many
2009-06-17 01:55:42 +00:00

391 lines
9.6 KiB
Groff
Raw Blame History

.\"
.\" Copyright (C) 2002 Chad David <davidc@acns.ab.ca>. 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 June 16, 2009
.Dt LOCK 9
.Os
.Sh NAME
.Nm lockinit ,
.Nm lockdestroy ,
.Nm lockmgr ,
.Nm lockmgr_args ,
.Nm lockmgr_args_rw ,
.Nm lockmgr_disown ,
.Nm lockmgr_printinfo ,
.Nm lockmgr_recursed ,
.Nm lockmgr_rw ,
.Nm lockmgr_waiters ,
.Nm lockstatus ,
.Nm lockmgr_assert
.Nd "lockmgr family of functions"
.Sh SYNOPSIS
.In sys/types.h
.In sys/lock.h
.In sys/lockmgr.h
.Ft void
.Fn lockinit "struct lock *lkp" "int prio" "const char *wmesg" "int timo" "int flags"
.Ft void
.Fn lockdestroy "struct lock *lkp"
.Ft int
.Fn lockmgr "struct lock *lkp" "u_int flags" "struct mtx *ilk"
.Ft int
.Fn lockmgr_args "struct lock *lkp" "u_int flags" "struct mtx *ilk" "const char *wmesg" "int prio" "int timo"
.Ft int
.Fn lockmgr_args_rw "struct lock *lkp" "u_int flags" "struct rwlock *ilk" "const char *wmesg" "int prio" "int timo"
.Ft void
.Fn lockmgr_disown "struct lock *lkp"
.Ft void
.Fn lockmgr_printinfo "struct lock *lkp"
.Ft int
.Fn lockmgr_recursed "struct lock *lkp"
.Ft int
.Fn lockmgr_rw "struct lock *lkp" "u_int flags" "struct rwlock *ilk"
.Ft int
.Fn lockmgr_waiters "struct lock *lkp"
.Ft int
.Fn lockstatus "struct lock *lkp"
.Pp
.Cd "options INVARIANTS"
.Cd "options INVARIANT_SUPPORT"
.Ft void
.Fn lockmgr_assert "struct lock *lkp" "int what"
.Sh DESCRIPTION
The
.Fn lockinit
function is used to initialize a lock.
It must be called before any operation can be performed on a lock.
Its arguments are:
.Bl -tag -width ".Fa wmesg"
.It Fa lkp
A pointer to the lock to initialize.
.It Fa prio
The priority passed to
.Xr sleep 9 .
.It Fa wmesg
The lock message.
This is used for both debugging output and
.Xr sleep 9 .
.It Fa timo
The timeout value passed to
.Xr sleep 9 .
.It Fa flags
The flags the lock is to be initialized with:
.Bl -tag -width ".Dv LK_CANRECURSE"
.It Dv LK_ADAPTIVE
Enable adaptive spinning for this lock if the kernel is compiled with the
ADAPTIVE_LOCKMGRS option.
.It Dv LK_CANRECURSE
Allow recursive exclusive locks.
.It Dv LK_NOPROFILE
Disable lock profiling for this lock.
.It Dv LK_NOSHARE
Allow exclusive locks only.
.It Dv LK_NOWITNESS
Instruct
.Xr witness 4
to ignore this lock.
.It Dv LK_NODUP
.Xr witness 4
should log messages about duplicate locks being acquired.
.It Dv LK_QUIET
Disable
.Xr ktr 4
logging for this lock.
.It Dv LK_TIMELOCK
Use
.Fa timo
during a sleep; otherwise, 0 is used.
.El
.El
.Pp
The
.Fn lockdestroy
function is used to destroy a lock, and while it is called in a number of
places in the kernel, it currently does nothing.
.Pp
The
.Fn lockmgr
and
.Fn lockmgr_rw
functions handle general locking functionality within the kernel, including
support for shared and exclusive locks, and recursion.
.Fn lockmgr
and
.Fn lockmgr_rw
are also able to upgrade and downgrade locks.
.Pp
Their arguments are:
.Bl -tag -width ".Fa flags"
.It Fa lkp
A pointer to the lock to manipulate.
.It Fa flags
Flags indicating what action is to be taken.
.Bl -tag -width ".Dv LK_CANRECURSE"
.It Dv LK_SHARED
Acquire a shared lock.
If an exclusive lock is currently held, it will be downgraded.
.It Dv LK_EXCLUSIVE
Acquire an exclusive lock.
If an exclusive lock is already held, and
.Dv LK_CANRECURSE
is not set, the system will
.Xr panic 9 .
.It Dv LK_DOWNGRADE
Downgrade exclusive lock to a shared lock.
Downgrading a shared lock is not permitted.
If an exclusive lock has been recursed, all references will be downgraded.
.It Dv LK_UPGRADE
Upgrade a shared lock to an exclusive lock.
If this call fails, the shared lock is lost.
During the upgrade, the shared lock could
be temporarily dropped.
Attempts to upgrade an exclusive lock will cause a
.Xr panic 9 .
.It Dv LK_RELEASE
Release the lock.
Releasing a lock that is not held can cause a
.Xr panic 9 .
.It Dv LK_DRAIN
Wait for all activity on the lock to end, then mark it decommissioned.
This is used before freeing a lock that is part of a piece of memory that is
about to be freed.
(As documented in
.In sys/lockmgr.h . )
.It Dv LK_SLEEPFAIL
Fail if operation has slept.
.It Dv LK_NOWAIT
Do not allow the call to sleep.
This can be used to test the lock.
.It Dv LK_NOWITNESS
Skip the
.Xr witness 4
checks for this instance.
.It Dv LK_CANRECURSE
Allow recursion on an exclusive lock.
For every lock there must be a release.
.It Dv LK_INTERLOCK
Unlock the interlock (which should be locked already).
.El
.It Fa ilk
An interlock mutex for controlling group access to the lock.
If
.Dv LK_INTERLOCK
is specified,
.Fn lockmgr
and
.Fn lockmgr_rw
assume
.Fa ilk
is currently owned and not recursed, and will return it unlocked.
See
.Xr mtx_assert 9 .
.El
.Pp
The
.Fn lockmgr_args
and
.Fn lockmgr_args_rw
function work like
.Fn lockmgr
and
.Fn lockmgr_rw
but accepting a
.Fa wmesg ,
.Fa timo
and
.Fa prio
on a per-instance basis.
The specified values will override the default
ones, but this can still be used passing, respectively,
.Dv LK_WMESG_DEFAULT ,
.Dv LK_PRIO_DEFAULT
and
.Dv LK_TIMO_DEFAULT .
.Pp
The
.Fn lockmgr_disown
function switches the owner from the current thread to be
.Dv LK_KERNPROC,
if the lock is already held.
.Pp
The
.Fn lockmgr_printinfo
function prints debugging information about the lock.
It is used primarily by
.Xr VOP_PRINT 9
functions.
.Pp
The
.Fn lockmgr_recursed
function returns true if the lock is recursed, 0
otherwise.
.Pp
The
.Fn lockmgr_waiters
function returns true if the lock has waiters, 0 otherwise.
.Pp
The
.Fn lockstatus
function returns the status of the lock in relation to the current thread.
.Pp
When compiled with
.Cd "options INVARIANTS"
and
.Cd "options INVARIANT_SUPPORT" ,
the
.Fn lockmgr_assert
function tests
.Fa lkp
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 KA_UNLOCKED"
.It Dv KA_LOCKED
Assert that the current thread has either a shared or an exclusive lock on the
.Vt lkp
lock pointed to by the first argument.
.It Dv KA_SLOCKED
Assert that the current thread has a shared lock on the
.Vt lkp
lock pointed to by the first argument.
.It Dv KA_XLOCKED
Assert that the current thread has an exclusive lock on the
.Vt lkp
lock pointed to by the first argument.
.It Dv KA_UNLOCKED
Assert that the current thread has no lock on the
.Vt lkp
lock pointed to by the first argument.
.El
.Pp
In addition, one of the following optional assertions can be used with
either an
.Dv KA_LOCKED ,
.Dv KA_SLOCKED ,
or
.Dv KA_XLOCKED
assertion:
.Bl -tag -width ".Dv KA_NOTRECURSED"
.It Dv KA_RECURSED
Assert that the current thread has a recursed lock on
.Fa lkp .
.It Dv KA_NOTRECURSED
Assert that the current thread does not have a recursed lock on
.Fa lkp .
.El
.Pp
.Sh RETURN VALUES
The
.Fn lockmgr
and
.Fn lockmgr_rw
functions return 0 on success and non-zero on failure.
.Pp
The
.Fn lockstatus
function returns:
.Bl -tag -width ".Dv LK_EXCLUSIVE"
.It Dv LK_EXCLUSIVE
An exclusive lock is held by the current thread.
.It Dv LK_EXCLOTHER
An exclusive lock is held by someone other than the current thread.
.It Dv LK_SHARED
A shared lock is held.
.It Li 0
The lock is not held by anyone.
.El
.Sh ERRORS
.Fn lockmgr
and
.Fn lockmgr_rw
fail if:
.Bl -tag -width Er
.It Bq Er EBUSY
.Dv LK_FORCEUPGRADE
was requested and another thread had already requested a lock upgrade.
.It Bq Er EBUSY
.Dv LK_NOWAIT
was set, and a sleep would have been required.
.It Bq Er ENOLCK
.Dv LK_SLEEPFAIL
was set and
.Fn lockmgr
or
.Fn lockmgr_rw
did sleep.
.It Bq Er EINTR
.Dv PCATCH
was set in the lock priority, and a signal was delivered during a sleep.
Note the
.Er ERESTART
error below.
.It Bq Er ERESTART
.Dv PCATCH
was set in the lock priority, a signal was delivered during a sleep,
and the system call is to be restarted.
.It Bq Er EWOULDBLOCK
a non-zero timeout was given, and the timeout expired.
.El
.Sh LOCKS
If
.Dv LK_INTERLOCK
is passed in the
.Fa flags
argument to
.Fn lockmgr
or
.Fn lockmgr_rw ,
the
.Fa ilk
must be held prior to calling
.Fn lockmgr
or
.Fn lockmgr_rw ,
and will be returned unlocked.
.Pp
Upgrade attempts that fail result in the loss of the lock that
is currently held.
Also, it is invalid to upgrade an
exclusive lock, and a
.Xr panic 9
will be the result of trying.
.Sh SEE ALSO
.Xr condvar 9 ,
.Xr locking 9 ,
.Xr mutex 9 ,
.Xr rwlock 9 ,
.Xr sleep 9 ,
.Xr sx 9 ,
.Xr mtx_assert 9 ,
.Xr panic 9 ,
.Xr VOP_PRINT 9
.Sh AUTHORS
This manual page was written by
.An Chad David Aq davidc@acns.ab.ca .
Reference in New Issue View Git Blame Copy Permalink