> -----Original Message----- > From: Jan Beulich <jbeul...@suse.com> > Sent: 06 March 2020 13:44 > To: Paul Durrant <xadimg...@gmail.com> > Cc: pdurr...@amzn.com; 'Stefano Stabellini' <sstabell...@kernel.org>; 'Julien > Grall' <jul...@xen.org>; > 'Wei Liu' <w...@xen.org>; 'Konrad Rzeszutek Wilk' <konrad.w...@oracle.com>; > 'Andrew Cooper' > <andrew.coop...@citrix.com>; Durrant, Paul <pdurr...@amazon.co.uk>; 'Ian > Jackson' > <ian.jack...@eu.citrix.com>; 'George Dunlap' <george.dun...@citrix.com>; 'Tim > Deegan' <t...@xen.org>; > 'Tamas K Lengyel' <ta...@tklengyel.com>; xen-devel@lists.xenproject.org; > 'Roger Pau Monné' > <roger....@citrix.com> > Subject: Re: [PATCH v3 5/6] mm: add 'is_special_page' macro... > > On 06.03.2020 13:35, Paul Durrant wrote: > >> -----Original Message----- > >> From: Xen-devel <xen-devel-boun...@lists.xenproject.org> On Behalf Of Jan > >> Beulich > >> Sent: 06 March 2020 12:20 > >> To: pdurr...@amzn.com > >> Cc: Stefano Stabellini <sstabell...@kernel.org>; Julien Grall > >> <jul...@xen.org>; Wei Liu > <w...@xen.org>; > >> Konrad Rzeszutek Wilk <konrad.w...@oracle.com>; Andrew Cooper > >> <andrew.coop...@citrix.com>; Paul > >> Durrant <pdurr...@amazon.com>; Ian Jackson <ian.jack...@eu.citrix.com>; > >> George Dunlap > >> <george.dun...@citrix.com>; Tim Deegan <t...@xen.org>; Tamas K Lengyel > >> <ta...@tklengyel.com>; xen- > >> de...@lists.xenproject.org; Roger Pau Monné <roger....@citrix.com> > >> Subject: Re: [Xen-devel] [PATCH v3 5/6] mm: add 'is_special_page' macro... > >> > >> On 05.03.2020 13:45, pdurr...@amzn.com wrote: > >>> --- a/xen/arch/x86/mm/shadow/common.c > >>> +++ b/xen/arch/x86/mm/shadow/common.c > >>> @@ -2087,19 +2087,22 @@ static int sh_remove_all_mappings(struct domain > >>> *d, mfn_t gmfn, gfn_t gfn) > >>> * The qemu helper process has an untyped mapping of this dom's > >>> RAM > >>> * and the HVM restore program takes another. > >>> * Also allow one typed refcount for > >>> - * - Xen heap pages, to match share_xen_page_with_guest(), > >>> - * - ioreq server pages, to match prepare_ring_for_helper(). > >>> + * - special pages, which are explicitly referenced and mapped by > >>> + * Xen. > >>> + * - ioreq server pages, which may be special pages or normal > >>> + * guest pages with an extra reference taken by > >>> + * prepare_ring_for_helper(). > >>> */ > >>> if ( !(shadow_mode_external(d) > >>> && (page->count_info & PGC_count_mask) <= 3 > >>> && ((page->u.inuse.type_info & PGT_count_mask) > >>> - == (is_xen_heap_page(page) || > >>> + == (is_special_page(page) || > >>> (is_hvm_domain(d) && is_ioreq_server_page(d, > >>> page))))) ) > >> > >> Shouldn't you delete most of this line, after the previous patch > >> converted the ioreq server pages to PGC_extra ones? > > > > I thought that too originally but then I realise we still have to > > cater for the 'legacy' emulators that still require IOREQ server > > pages to be mapped through the p2m, in which case they will not > > be PGC_extra pages. > > Oh, indeed. (I don't suppose we can ever do away with this legacy > mechanism?)
It's tricky because it would either mean breaking older (pre resource-mapping) QEMUs, or allowing the toolstack to allocate the 'special' pages with an extra flag to make them PGC_extra. > > >> Also I notice this construct is used by x86 code only - is there > >> a particular reason it doesn't get placed in an x86 header (at > >> least for the time being)? > > > > PGC_extra pages are common so maybe it is better off defined here > > so it is available to ARM code? > > To be honest, my question was mainly based on me being puzzled that > Arm (or common) code doesn't need any such adjustment. As a result > I'm wondering whether that's just "luck" (in which case I'd agree > the placement could remain as is), or whether there's a deeper > reason behind that, largely guaranteeing Arm would also never need > it. > I'll have a chat with Julien and see what he thinks. Paul _______________________________________________ Xen-devel mailing list Xen-devel@lists.xenproject.org https://lists.xenproject.org/mailman/listinfo/xen-devel
