Re: [PATCH] KVM guest: prevent tracing recursion with kvmclock

[Steven Rostedt <rostedt@goodmis.org>]

On Tue, 2011-11-15 at 15:45 +0200, Avi Kivity wrote:

> What about function traces?  Will any noninlined calls cause the same
> problem?
> 

Both function_graph and function tracing have recursion protection
around the users of the call. function tracing is much lighter weight
than function_graph tracing. The problem here was that function_graph
tracing does some accounting, and calls the clock handler outside the
recursion protection. I need to add recursion protection around the
accounting too. But I still want to warn when it happens as recursion
protection protects from crashing, but recursion is still bad because it
still goes through a bit of arch specific setup before the recursion is
detected, slowing function_graph tracing down much more than function
tracing when it recurses.

-- Steve