From: John Levon Subject: Re: [RFC] bootloader improvements - pygrub-kernel-config Date: Fri, 10 Nov 2006 21:10:43 +0000 Message-ID: <20061110211043.GA7003@totally.trollied.org> References: <20061109192017.GA16487@totally.trollied.org> <20061109192240.GC16717@totally.trollied.org> <20061110201428.GH19736@redhat.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Return-path: Content-Disposition: inline In-Reply-To: <20061110201428.GH19736@redhat.com> List-Unsubscribe: , List-Post: List-Help: List-Subscribe: , Sender: xen-devel-bounces@lists.xensource.com Errors-To: xen-devel-bounces@lists.xensource.com To: "Daniel P. Berrange" Cc: xen-devel@lists.xensource.com List-Id: xen-devel@lists.xenproject.org On Fri, Nov 10, 2006 at 08:14:28PM +0000, Daniel P. Berrange wrote: > I'm not sure this is a good idea - its essentially completely inverting > the current semantics for bootloader/kernel params in the config files. > Currently a 'kernel' parameter will always override a 'bootloader' param, > but with this change a 'bootloader' param will always override a 'kernel' > param. I can see that both approaches have their merits, I don't think > we can ever pick one prioritization rule which satisfies everyone here > and thus for sake of compatability we should keep the current semantics > for kernel/booloader prioritization in the configs. Taken in concert with the other patch, the 'kernel' choice will always be respected if it's set. > > self.kernel = get_cfg("kernel") > > + self.ramdisk = get_cfg("ramdisk", '') > > + self.boot_kernel = get_cfg("boot_kernel") > > + if not self.boot_kernel: > > + self.boot_kernel = self.kernel > > + self.boot_ramdisk = get_cfg("boot_ramdisk") > > + if not self.boot_ramdisk: > > + self.boot_ramdisk = self.ramdisk > > + > > This is also changing the semantics of the SEXPR exposed by XenD - previously > the ('vmlinux') and ('initrd') parameters would refer to the live kernel > in the guest, but now you'd have to look at the boot_XXX variants instead. > > If we did want to change the prioritzation such that bootloader overrides > any explicit kernel/initrd, then I think we should just have the values > from bootloader directly overwrite the existing named SXPR fields rather > than adding new boot_XX fields. See the other patch. We could make it so kernel/initrd are always the 'live' ones, than we have cfg_kernel/cfg_initrd or something, but I don't really see the advantage in that, and the disadvantage is reasonably obvious: if I want to lookup to see "what kernel is this", I'd have to check first cfg_kernel then kernel. With boot_*, I can always look in 'kernel' and know that this is the one that the .py config chose. regards john