From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Date: Fri, 23 Mar 2007 09:09:29 -0500 From: Anton Blanchard To: Paul Mackerras Subject: Re: [patch 06/10] Add notify die hooks and remove some redundant debugger hooks Message-ID: <20070323140929.GA17198@kryten> References: <20070321013810.404636000@samba.org> <20070321013825.283024000@samba.org> <17923.46762.363994.307295@cargo.ozlabs.ibm.com> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii In-Reply-To: <17923.46762.363994.307295@cargo.ozlabs.ibm.com> Cc: linuxppc-dev@ozlabs.org List-Id: Linux on PowerPC Developers Mail List List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Hi, > As far as I can see, at the moment there is just one thing that gets > registered to be called by die_notify, and that is > kprobe_exceptions_notify. It doesn't do anything with DIE_OOPS or > DIE_MACHINE_CHECK. Keep reading :) Later on I remove all the debugger() indirect branches. > I would rather get rid of die_notify and have a kprobe_notify instead > that just directly calls kprobe_exceptions_notify. In fact having > separate kprobe routines for the different events would be even > better. > > It's not like our cpus are particularly good at indirect functions > calls... ;( True and we are removing a set of indirect branches that exist already (debugger). Id prefer not to have 2 sets of duplicate hooks for all this stuff. Keep in mind with xmon and kprobes disabled these calls are fast. I verified it in sim after Christophs 300 cycle rant, a pessimistic estimate is 20 cycles on power5, including function call overhead. Anton