Read-copy-update and interrupt latency

Posted Jan 15, 2004 10:51 UTC (Thu) by oak (guest, #2786)
Parent article: Read-copy-update and interrupt latency

What's the effect of RCU code on single CPU machines?

to post comments

Read-copy-update and interrupt latency

Posted Jan 16, 2004 23:32 UTC (Fri) by PaulMcKenney (✭ supporter ✭, #9624) [Link]

It defers deletion, similar to SMP kernels. For an example of why this is needed, consider an interrupt handler (perhaps softirq) that deletes an element from a data structure that can be referenced from mainline code. The interrupt handler then cannot immediately free up the element, since the interrupted mainline code might well still be referencing it.

There are cases where it -is- safe to immediately free things up in an UP kernel, and Dipankar has some code that permits this in some cases as well.