From mboxrd@z Thu Jan 1 00:00:00 1970 From: Joerg Roedel Subject: Re: [PATCH 35/62] x86/sev-es: Setup per-cpu GHCBs for the runtime handler Date: Wed, 12 Feb 2020 16:16:13 +0100 Message-ID: <20200212151613.GC22063@8bytes.org> References: <20200211135256.24617-1-joro@8bytes.org> <20200211135256.24617-36-joro@8bytes.org> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Return-path: Content-Disposition: inline In-Reply-To: Sender: linux-kernel-owner@vger.kernel.org To: Andy Lutomirski Cc: X86 ML , "H. Peter Anvin" , Dave Hansen , Peter Zijlstra , Thomas Hellstrom , Jiri Slaby , Dan Williams , Tom Lendacky , Juergen Gross , Kees Cook , LKML , kvm list , Linux Virtualization , Joerg Roedel List-Id: virtualization@lists.linuxfoundation.org On Tue, Feb 11, 2020 at 02:46:11PM -0800, Andy Lutomirski wrote: > On Tue, Feb 11, 2020 at 5:53 AM Joerg Roedel wrote: > > +/* Runtime GHCBs */ > > +static DEFINE_PER_CPU_DECRYPTED(struct ghcb, ghcb_page) __aligned(PAGE_SIZE); > > Hmm. This is a largeish amount of memory on large non-SEV-ES systems. > Maybe store a pointer instead? It would be even better if it could be > DEFINE_PER_CPU like this but be discarded if we don't need it, but I > don't think we have the infrastructure for that. Yeah, discarding is not easily possible right now, but I changed it to only store a pointer and allocating the pages only when running as an SEV-ES guest. Regards, Joerg