Locking

LockingThis document describes SMP locks and outlines the need of introducing -locks in the code. -

Locks are used to synchronize the data between threads in the kernel. They -can be also used to synchronize access to data in user side threads In SMP, -threads are executed in parallel, which means that if locks are not applied -to the code it could result in a race condition. Race conditions lead to system -crashes and data corruptions.

Locking Granularity

An -important property of a lock is its granularity. The granularity is a measure -of the amount of data the lock is protecting. There are two different granularities -for locks:

Coarse-Grained Locks enclose a large area of shared code or -multiple areas of unrelated data. The lock reduces the number of threads that -run concurrently resulting in serial execution making the code behave like -a single thread process. Coarse locks are applied in parallel. One lock can -be applied to each kernel subsystem.

The following diagram illustrates how a Coarse-Grained Lock covers -many parts of code. It not only simplifies the locking action itself but also -frees developers from having to load all the members of a code in order to -lock them. In order to get concurrency in the operating system, the operating -system must allow more than one process (or interrupt) to execute at the same -time. To do this, we divide the OS into sections and give each section a lock. -For a small number of processors, we only need a small number of locks, each -covering a large region of the OS. This model of coarse-grained locking provides -good scaling on small numbers of processors.

-Coarse-Grained Lock - - -

Fine-Grained Locks enclose a small area of code for example -a data structure. These locks are added to the code and the user must remember -to release the lock. Fine locks are error prone.

As the number of processors increases, the number of locks also increases. -The following diagram illustrates how fine locks are applied to data. In other -words, fine locks protect individual data structures or even parts of data -structures. All those locks add instructions and data. To do this, we divide -the OS into sections and divide the section into small pieces of code and -apply lock for each piece of code. Fine-grained locking can result in near -perfect scaling.

-Fine-Grained Lock - -

Type of Locks

TSpinLock is the lightest weight lock available -kernel side. If a process attempts to acquire a spinlock and one is not available, -the process will keep trying (spinning) until it can acquire the lock. Spinlocks -should be used to lock data in situations where the lock is not held for a -long time.

RFastLock is the lightest weight lock available -user side. There is no priority inheritance. This is a layer over a standard -semaphore, and only calls into the kernel side if there is contention.

RMutex is used to serialize access to a section -of re-entrant code that cannot be executed concurrently by more than one thread. -A mutex object allows one thread into a controlled section, forcing other -threads which attempt to gain access to that section to wait until the first -thread has exited from that section.

RSemaphore is used for Inter Process Communication -(IPC), they are similar in performance to RMutex. RSemaphore locks -are used when the lock must be held for a long time. These locks put the thread -into sleep mode and are used to synchronize user contexts.

-SMP - Overview - -Using Mutexes - -Using Semaphores - + + + + + +LockingThis document describes SMP locks and outlines the need of introducing +locks in the code. +

Locks are used to synchronize the data between threads in the kernel. They +can be also used to synchronize access to data in user side threads In SMP, +threads are executed in parallel, which means that if locks are not applied +to the code it could result in a race condition. Race conditions lead to system +crashes and data corruptions.

Locking Granularity

An +important property of a lock is its granularity. The granularity is a measure +of the amount of data the lock is protecting. There are two different granularities +for locks:

Coarse-Grained Locks enclose a large area of shared code or +multiple areas of unrelated data. The lock reduces the number of threads that +run concurrently resulting in serial execution making the code behave like +a single thread process. Coarse locks are applied in parallel. One lock can +be applied to each kernel subsystem.

The following diagram illustrates how a Coarse-Grained Lock covers +many parts of code. It not only simplifies the locking action itself but also +frees developers from having to load all the members of a code in order to +lock them. In order to get concurrency in the operating system, the operating +system must allow more than one process (or interrupt) to execute at the same +time. To do this, we divide the OS into sections and give each section a lock. +For a small number of processors, we only need a small number of locks, each +covering a large region of the OS. This model of coarse-grained locking provides +good scaling on small numbers of processors.

+Coarse-Grained Lock + + +

Fine-Grained Locks enclose a small area of code for example +a data structure. These locks are added to the code and the user must remember +to release the lock. Fine locks are error prone.

As the number of processors increases, the number of locks also increases. +The following diagram illustrates how fine locks are applied to data. In other +words, fine locks protect individual data structures or even parts of data +structures. All those locks add instructions and data. To do this, we divide +the OS into sections and divide the section into small pieces of code and +apply lock for each piece of code. Fine-grained locking can result in near +perfect scaling.

+Fine-Grained Lock + +

Type of Locks

TSpinLock is the lightest weight lock available +kernel side. If a process attempts to acquire a spinlock and one is not available, +the process will keep trying (spinning) until it can acquire the lock. Spinlocks +should be used to lock data in situations where the lock is not held for a +long time.

RFastLock is the lightest weight lock available +user side. There is no priority inheritance. This is a layer over a standard +semaphore, and only calls into the kernel side if there is contention.

RMutex is used to serialize access to a section +of re-entrant code that cannot be executed concurrently by more than one thread. +A mutex object allows one thread into a controlled section, forcing other +threads which attempt to gain access to that section to wait until the first +thread has exited from that section.

RSemaphore is used for Inter Process Communication +(IPC), they are similar in performance to RMutex. RSemaphore locks +are used when the lock must be held for a long time. These locks put the thread +into sleep mode and are used to synchronize user contexts.

+SMP - Overview + +Using Mutexes + +Using Semaphores +

\ No newline at end of file