The NdisDprAcquireReadWriteLock function acquires a lock that the caller uses for either write or read access to the resources that are shared among driver threads.
void NdisDprAcquireReadWriteLock( PNDIS_RW_LOCK Lock, BOOLEAN fWrite, PLOCK_STATE LockState );
A pointer to an opaque variable that represents a lock. The caller can use this lock to access shared resources.
A Boolean value. If the value is TRUE, this function is provided with write access to shared resources; if the value is FALSE, this function is provided with read access.
A pointer to an opaque variable that tracks the state of the lock. This variable exists in the interval between the time the caller acquires and releases the lock. The caller must use a different variable of type LOCK_STATE for each attempt that it makes to acquire the lock from the same non-ISR driver thread.
The driver must initialize a variable of type NDIS_RW_LOCK using the NdisInitializeReadWriteLock function before the driver calls any other NdisXxxReadWriteLock function. The driver must provide resident storage for the locks it uses.
After acquiring a lock by using
NdisDprAcquireReadWriteLock, the caller must release that lock by calling the
NdisDprReleaseReadWriteLock function. To decrement the reference count of the lock, a driver must call
NdisDprReleaseReadWriteLock once for each call to NdisDprAcquireReadWriteLock.
It is safe to use both NdisDprAcquireReadWriteLock and NdisDprAcquireReadWriteLock on the same lock. However, calls must be balanced so that if the lock is acquired with NdisDprAcquireReadWriteLock, it must be released with NdisDprReleaseReadWriteLock. Likewise, if the lock is acquired with NdisAcquireReadWriteLock, it must be released with NdisReleaseReadWriteLock.
With some architectures, NdisDprAcquireReadWriteLock performs faster than NdisAcquireReadWriteLock. Drivers can use NdisDprAcquireReadWriteLock rather than NdisAcquireReadWriteLock when it is certain that the current IRQL is already DISPATCH_LEVEL. However, it is not required. The overhead of calling the KeGetCurrentIrql function is greater than the performance advantage of calling NdisDprAcquireReadWriteLock rather than NdisAcquireReadWriteLock.
To modify resources that are shared among driver threads, a driver thread must acquire a write lock. To simply monitor those resources, a driver thread must acquire a read-only lock. Read access does not require interlocked operations or contention for spin locks. Using read-only access helps to maintain good operating system and driver performance.
A driver thread should never hold a write lock for more than 25 microseconds. Holding a write lock for a prolonged period degrades both operating system and driver performance.
The driver cannot use a lock to protect resources from read or write access that its other functions
share with the
MiniportDisableInterruptEx functions. Instead, the driver must call
NdisMSynchronizeWithInterruptEx so that its
MiniportSynchronizeInterrupt function accesses such shared resources at the same DIRQL at which its
MiniportInterrupt and/or MiniportDisableInterruptEx functions do.
For more information about acquiring and releasing NDIS spin locks, see Synchronization and Notification in Network Drivers.
|Minimum supported client||Deprecated for NDIS 6.20 and later drivers, which should use NdisAcquireRWLockRead or NdisAcquireRWLockWrite instead of NdisDprAcquireReadWriteLock. Supported in NDIS 6.0 and 6.1.|
|Header||ndis.h (include Ndis.h)|
See alsoMiniportSynchronizeInterrupt NdisMSynchronizeWithInterruptEx