From mboxrd@z Thu Jan 1 00:00:00 1970 Date: Tue, 22 Apr 2008 12:21:52 +0100 From: Mel Gorman Subject: Re: [PATCH 1/4] Add a basic debugging framework for memory initialisation Message-ID: <20080422112151.GB30798@csn.ul.ie> References: <20080417000624.18399.35041.sendpatchset@skynet.skynet.ie> <20080417000644.18399.66175.sendpatchset@skynet.skynet.ie> <20080421151405.GI5474@elte.hu> MIME-Version: 1.0 Content-Type: text/plain; charset=iso-8859-15 Content-Disposition: inline In-Reply-To: <20080421151405.GI5474@elte.hu> Sender: owner-linux-mm@kvack.org Return-Path: To: Ingo Molnar Cc: linux-mm@kvack.org, linux-kernel@vger.kernel.org List-ID: On (21/04/08 17:14), Ingo Molnar didst pronounce: > > * Mel Gorman wrote: > > > +config DEBUG_MEMORY_INIT > > + bool "Debug memory initialisation" > > + depends on DEBUG_KERNEL > > + help > > + Enable this to turn on debug checks during memory initialisation. By > > + default, sanity checks will be made on the memory model and > > + information provided by the architecture. What level of checking > > + made and verbosity during boot can be set with the > > + mminit_debug_level= command-line option. > > + > > + If unsure, say N > > should be "default y" - and perhaps only disable-able on > CONFIG_EMBEDDED. Ok, that would be something like the following? bool "Debug memory initialisation" if DEBUG_KERNEL && EMBEDDED depends on DEBUG_KERNEL default !EMBEDDED This will slow up boot slightly on debug kernels as the additional checks are made. It'll remain to be seen as to whether this is a problem for people or not. I doubt it'll be noticed. > We generally want such bugs to pop up as soon as > possible, and the sanity checks should only go away if someone > specifically aims for lowest system footprint. > Seems fair and it's the second time this has been suggested (off-list reviewer again). The only potential gotcha is if a sanity check is introduced that is itself broken. It should be very obvious when this type of bug occurs though. > the default loglevel for debug printouts might deserve another debug > option - but the core checks should always be included, and _errors_ > should always be printed out. > I'll replace mminit_debug_level with mminit_loglevel that determines whether information is printed at KERN_DEBUG or not. This matches what other similar debug-frameworks are doing. I'll make sure errors always get printed. Thanks -- Mel Gorman Part-time Phd Student Linux Technology Center University of Limerick IBM Dublin Software Lab -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@kvack.org. For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: email@kvack.org