From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1755634AbZEMDGX (ORCPT ); Tue, 12 May 2009 23:06:23 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1755117AbZEMDF5 (ORCPT ); Tue, 12 May 2009 23:05:57 -0400 Received: from adsl-99-35-133-201.dsl.pltn13.sbcglobal.net ([99.35.133.201]:36205 "EHLO abulafia.goop.org" rhost-flags-OK-OK-OK-FAIL) by vger.kernel.org with ESMTP id S1754663AbZEMDF4 (ORCPT ); Tue, 12 May 2009 23:05:56 -0400 From: Jeremy Fitzhardinge To: Ingo Molnar Cc: the arch/x86 maintainers , Linux Kernel Mailing List , Xen-devel , Ian Campbell , Jeremy Fitzhardinge Subject: [PATCH 10/16] xen: clear reserved bits in l3 entries given in the initial pagetables Date: Tue, 12 May 2009 14:28:32 -0700 Message-Id: <1242163718-2934-11-git-send-email-jeremy@goop.org> X-Mailer: git-send-email 1.6.0.6 In-Reply-To: <1242163718-2934-1-git-send-email-jeremy@goop.org> References: <1242163718-2934-1-git-send-email-jeremy@goop.org> Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org From: Ian Campbell In native PAE, the only flag that may be legitimately set in an L3 entry is Present. When Xen grafts the top-level PAE L3 pagetable entries into the L4 pagetable, it must also set the other permissions flags so that the mapped pages are actually accessible. However, due to a bug in the hypervisor, it validates update to the L3 entries as formal PAE entries, so it will refuse to validate these entries with the extra bits requires for 4-level pagetables. This patch simply masks the entries back to the bare PAE level, leaving Xen to add whatever bits it feels are necessary. [ Impact: workaround Xen bug in 32-on-64 dom0 ] Signed-off-by: Ian Campbell Signed-off-by: Jeremy Fitzhardinge --- arch/x86/xen/mmu.c | 15 +++++++++++++++ 1 files changed, 15 insertions(+), 0 deletions(-) diff --git a/arch/x86/xen/mmu.c b/arch/x86/xen/mmu.c index 05cfd30..85d9d18 100644 --- a/arch/x86/xen/mmu.c +++ b/arch/x86/xen/mmu.c @@ -1835,6 +1835,7 @@ __init pgd_t *xen_setup_kernel_pagetable(pgd_t *pgd, unsigned long max_pfn) { pmd_t *kernel_pmd; + int i; max_pfn_mapped = PFN_DOWN(__pa(xen_start_info->pt_base) + xen_start_info->nr_pt_frames * PAGE_SIZE + @@ -1846,6 +1847,20 @@ __init pgd_t *xen_setup_kernel_pagetable(pgd_t *pgd, xen_map_identity_early(level2_kernel_pgt, max_pfn); memcpy(swapper_pg_dir, pgd, sizeof(pgd_t) * PTRS_PER_PGD); + + /* + * When running a 32 bit domain 0 on a 64 bit hypervisor a + * pinned L3 (such as the initial pgd here) contains bits + * which are reserved in the PAE layout but not in the 64 bit + * layout. Unfortunately some versions of the hypervisor + * (incorrectly) validate compat mode guests against the PAE + * layout and hence will not allow such a pagetable to be + * pinned by the guest. Therefore we mask off only the PFN and + * Present bits of the supplied L3. + */ + for (i = 0; i < PTRS_PER_PGD; i++) + swapper_pg_dir[i].pgd &= (PTE_PFN_MASK | _PAGE_PRESENT); + set_pgd(&swapper_pg_dir[KERNEL_PGD_BOUNDARY], __pgd(__pa(level2_kernel_pgt) | _PAGE_PRESENT)); -- 1.6.0.6