Re: consistent read and LIO question

From: Mladen Gogala <>
Date: Mon, 27 Apr 2009 22:03:17 +0000 (UTC)
Message-ID: <gt5a34$p9c$>

On Sun, 26 Apr 2009 08:22:05 -0700, lsllcm wrote:

> What is the difference "cache buffer chains latch" and "buffer handle"?
> Are they same?

No. Buffer handle is a pointer to the buffer, along with the status information. Buffer handles form structures called "linked lists". In order to manipulate a linked list, you have to acquire a lock of a sort. The qualification "of a sort" is necessary because there are no real locks and certainly no queues. There is a shared memory structure that the proces which needs to manipulate a list of buffers needs to "own". The underlying mechanism is called a spin lock and is based on atomic instruction called "test and set lock" (TSL). In short, CPU tests a memory location and checks if its value is zero. If it is, it sets it to 1. If it isn't, the instruction returns status that it was unsuccessful. The process then spins in an idle loop, spending certain amount of iterations before it retries. I can only recommend Andrew Tannenbaum's "Modern Operating Systems" and the chapter on synchronization mechanisms and critical sections. The book is more then 20 years old, but is still modern. Given the fact that Tannenbaum's book is very old, it's probably hard to come by. Here is a decent explanation of a spin lock: Cache chain latches are spin locks that protect linked lists of buffer handles.

Received on Mon Apr 27 2009 - 17:03:17 CDT

Original text of this message