From mboxrd@z Thu Jan 1 00:00:00 1970 From: Matt Fleming Subject: Re: [PATCH 2/2] efi: Capsule update support Date: Tue, 14 Oct 2014 22:46:33 +0100 Message-ID: <20141014214633.GD14343@console-pimps.org> References: <1412692951-25478-1-git-send-email-matt@console-pimps.org> <1412692951-25478-2-git-send-email-matt@console-pimps.org> <20141010182846.GA10588@pd.tnic> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Return-path: Content-Disposition: inline In-Reply-To: <20141010182846.GA10588-fF5Pk5pvG8Y@public.gmane.org> Sender: linux-efi-owner-u79uwXL29TY76Z2rM5mHXA@public.gmane.org To: Borislav Petkov Cc: linux-efi-u79uwXL29TY76Z2rM5mHXA@public.gmane.org, linux-kernel-u79uwXL29TY76Z2rM5mHXA@public.gmane.org, Matt Fleming , Leif Lindholm , "Kweh, Hock Leong" List-Id: linux-efi@vger.kernel.org On Fri, 10 Oct, at 08:28:47PM, Borislav Petkov wrote: > > You have efi_capsule_update() vs efi_update_capsule(). Maybe change the > names a bit more for differentiation. Or prepend the workhorse doing all > the work with "__" or so... Yeah, I really didn't come up with a great naming scheme here. I'll fix that. > > + > > + for (i = 0; i < nr_block_pgs; i++) { > > + block_pgs[i] = alloc_page(GFP_KERNEL); > > Maybe alloc_pages() once we verify that it actually gives phys. contig. > memory and maybe also try to do it outside of the locked region. I don't > know if it would matter to drop the locks though as capsule updating is > not something you do pretty often. I'd hope! Actually, I'm not bothered about getting physically contiguous memory because we pass a scatter gather list to the firmware anyway. What I was looking for was to avoid doing high order allocations when we don't really need them (lots of low order allocs are fine). Right, allocating under the lock isn't a great idea. I'll take a look at reworking this to do the allocation up front. -- Matt Fleming, Intel Open Source Technology Center