LOCK(9) NetBSD Kernel Developer's Manual LOCK(9)
NAME
lock, simple_lock_init, simple_lock, simple_lock_try, simple_unlock, simple_lock_freecheck, simple_lock_dump, lockinit, lockmgr, lockstatus, lockmgr_printinfo, transferlockers, spinlockinit, spinlockmgr -- kernel lock functions
SYNOPSIS
#include <sys/lock.h> void simple_lock_init(struct simplelock *slock); void simple_lock(struct simplelock *slock); int simple_lock_try(struct simplelock *slock); void simple_unlock(struct simplelock *slock); void simple_lock_freecheck(void *start, void *end); void simple_lock_dump(void); void lockinit(struct lock *lock, int prio, const char *wmesg, int timo, int flags); int lockmgr(struct lock *lock, u_int flags, struct simplelock *slock); int lockstatus(struct lock *lock); void lockmgr_printinfo(struct lock *lock); void transferlockers(struct lock *from, struct lock *to); void spinlockinit(struct lock *lock, const char *wmesg, int flags); int spinlockmgr(struct lock *lock, u_int flags, struct simplelock *slock);
DESCRIPTION
The lock functions provide synchronisation in the kernel by preventing multiple threads from simultaneously executing critical sections of code accessing shared data. A number of different locks are available: struct simplelock Provides a simple spinning mutex. A processor will busy-wait while trying to acquire a simplelock. The simplelock operations are implemented with machine-dependent locking primitives. Simplelocks are usually used only by the high-level lock manager and to protect short, critical sections of code. Simplelocks are the only locks that can be used inside an interrupt handler. For a simplelock to be used in an interrupt handler, care must be taken to disable the interrupt, acquire the lock, do any pro- cessing, release the simplelock and re-enable the interrupt. This procedure is necessary to avoid deadlock between the inter- rupt handler and other threads executing on the same processor. struct lock Provides a high-level lock supporting sleeping/spinning until the lock can be acquired. The lock manager supplies both exclu- sive-access and shared-access locks, with recursive exclusive- access locks within a single thread. It also allows upgrading a shared-access lock to an exclusive-access lock, as well as down- grading an exclusive-access lock to a shared-access lock. If the kernel option LOCKDEBUG is enabled, additional facilities are pro- vided to record additional lock information. These facilities are pro- vided to assist in determining deadlock occurrences.
FUNCTIONS
The functions which operate on simplelocks are: simple_lock_init(slock) The simplelock slock is initialised to the unlocked state. A statically allocated simplelock also can be initialised with the macro SIMPLELOCK_INITIALIZER. The effect is the same as the dynamic initialisation by a call to simple_lock_init. For exam- ple, struct simplelock slock = SIMPLELOCK_INITIALIZER; simple_lock(slock) The simplelock slock is locked. If the simplelock is held then execution will spin until the simplelock is acquired. Care must be taken to ensure that the calling thread does not already hold the simplelock. In this case, the simplelock can never be acquired. If kernel option LOCKDEBUG is enabled, a ``locking against myself'' panic will occur. simple_lock_try(slock) Try to acquire the simplelock slock without spinning. If the simplelock is held by another thread then the return value is 0. If the simplelock was acquired successfully then the return value is 1. simple_unlock(slock) The simplelock slock is unlocked. The simplelock must be locked and the calling thread must be the one that last acquired the simplelock. If the calling thread does not hold the simplelock, the simplelock will be released but the kernel behaviour is undefined. simple_lock_freecheck(start, end) Check that all simplelocks in the address range start to end are not held. If a simplelock within the range is found, the kernel enters the debugger. This function is available only with ker- nel option LOCKDEBUG. It provides a mechanism for basic simple- lock consistency checks. simple_lock_dump(void) Dump the state of all simplelocks in the kernel. This function is available only with kernel option LOCKDEBUG. The functions which operate on locks are: lockinit(lock, prio, wmesg, timo, flags) The lock lock is initialised according to the parameters pro- vided. Arguments are as follows: lock The lock. prio The thread priority when it is woken up after sleeping on the lock. wmesg A sleep message used when a thread goes to sleep wait- ing for the lock, so that the exact reason it is sleep- ing can easily be identified. timo The maximum sleep time. Used by tsleep(9). flags Flags to specify the lock behaviour permanently over the lifetime of the lock. Valid lock flags are: LK_NOWAIT Threads should not sleep when attempting to acquire the lock. LK_SLEEPFAIL Threads should sleep, then return failure when acquiring the lock. LK_CANRECURSE Threads can acquire the lock recursively. lockmgr(lock, flags, slock) Set, change or release a lock according to the parameters pro- vided. Arguments are as follows: lock The lock. slock Simplelock interlock. If the flag LK_INTERLOCK is set in flags, slock is a simplelock held by the caller. When the lock lock is acquired, the simplelock is released. If the flag LK_INTERLOCK is not set, slock is ignored. flags Flags to specify the lock request type. In addition to the flags specified above, the following flags are valid: LK_SHARED Get one of many possible shared-access locks. If a thread holding an exclusive-access lock requests a shared-access lock, the exclusive- access lock is downgraded to a shared-access lock. LK_EXCLUSIVE Stop further shared-access locks, when they are cleared, grant a pending upgrade if it exists, then grant an exclusive-access lock. Only one exclusive-access lock may exist at a time, except that a thread holding an exclu- sive-access lock may get additional exclusive- access locks if it explicitly sets the LK_CANRECURSE flag in the lock request, or if the LK_CANRECURSE flag was set when the lock was initialised. LK_UPGRADE The thread must hold a shared-access lock that it wants to have upgraded to an exclusive- access lock. Other threads may get exclusive access to the protected resource between the time that the upgrade is requested and the time that it is granted. LK_EXCLUPGRADE The thread must hold a shared-access lock that it wants to have upgraded to an exclusive- access lock. If the request succeeds, no other threads will have acquired exclusive access to the protected resource between the time that the upgrade is requested and the time that it is granted. However, if another thread has already requested an upgrade, the request will fail. LK_DOWNGRADE The thread must hold an exclusive-access lock that it wants to have downgraded to a shared- access lock. If the thread holds multiple (recursive) exclusive-access locks, they will all be downgraded to shared-access locks. LK_RELEASE Release one instance of a lock. LK_DRAIN Wait for all activity on the lock to end, then mark it decommissioned. This feature is used before freeing a lock that is part of a piece of memory that is about to be freed. LK_REENABLE Lock is to be re-enabled after drain. The LK_REENABLE flag may be set only at the release of a lock obtained by a drain. LK_SETRECURSE Other locks while we have it OK. LK_RECURSEFAIL Attempt at recursive lock fails. LK_SPIN Lock spins instead of sleeping. LK_INTERLOCK Unlock the simplelock slock when the lock is acquired. lockstatus(lock) Determine the status of lock lock. Returns one of the follow- ing: LK_EXCLUSIVE The current lwp or CPU holds an exclusive-access lock. LK_EXCLOTHER The other lwp or CPU holds an exclusive-access lock. LK_SHARED Someone holds shared-access lock. 0 Not locked. lockmgr_printinfo(lock) Print out information about state of lock lock. transferlockers(from, to) Transfer any waiting processes from lock from to lock to. spinlockinit(lock, wmesg, flags) The lock lock is initialised as a spinlock according to the parameters provided. Arguments are as follows: lock The lock. wmesg This is a simple name for lock. flags Flags to specify the lock behaviour. Valid lock flags are the same as outlined above. spinlockmgr(lock, flags, slock) Set, change or release a lock according to the parameters pro- vided. Arguments are as follows: lock The spin lock. flags Flags to specify the lock request type. Valid lock flags are the same as outlined above. slock Simplelock interlock. The simplelock slock is set by the caller. When the lock lock is acquired, the sim- plelock is released.
RETURN VALUES
Successfully acquired locks return 0. A failed lock attempt always returns a non-zero error value. No lock is held after an error return (in particular, a failed LK_UPGRADE or LK_FORCEUPGRADE will have released its shared-access lock). Locks will always succeed unless one of the following is true: [EBUSY] LK_FORCEUPGRADE is requested and some other thread has already requested a lock upgrade or LK_NOWAIT is set and a sleep would be required. [ENOLCK] LK_SLEEPFAIL is set and a sleep was done. [EINTR] PCATCH is set in lock priority and a signal arrives to interrupt a system call. [ERESTART] PCATCH is set in lock priority and a signal arrives so that the system call is restarted. [EWOULDBLOCK] Non-null lock timeout and timeout expires.
CODE REFERENCES
This section describes places within the NetBSD source tree where actual code implementing or using the locking framework can be found. All path- names are relative to /usr/src. The locking framework itself is implemented within the file sys/kern/kern_lock.c. Data structures and function prototypes for the framework are located in sys/sys/lock.h. Machine-dependent simplelock primitives are implemented within the file sys/arch/<arch>/include/lock.h.
SEE ALSO
pmap(9), spl(9), tsleep(9), uvm(9)
HISTORY
The kernel locking API first appeared in 4.4BSD-lite2. NetBSD 3.0.1 May 25, 2004 NetBSD 3.0.1
Powered by man-cgi (2024-08-26). Maintained for NetBSD by Kimmo Suominen. Based on man-cgi by Panagiotis Christias.