[PATCH] xen: do not set NX bit when making initial pagetables readonly

[Ian Campbell]

__supported_pte_mask has not been correctly configured at this point
and Xen prevents us from using the NX bit if the hardware does not
support it. Some BIOSes seem to offer the option to disable NX.

Signed-off-by: Ian Campbell <ian.campbell(a)citrix.com&gt;
Cc: Mark McLoughlin <markmc(a)redhat.com&gt;
Cc: Jon Swanson <jswanson(a)valuecommerce.co.jp&gt;
Cc: fedora-virt(a)redhat.com
Cc: Jeremy Fitzhardinge <jeremy(a)goop.org&gt;
Cc: Ingo Molnar <mingo(a)elte.hu&gt;
---
 arch/x86/xen/enlighten.c |   24 +++++++++++++-----------
 1 files changed, 13 insertions(+), 11 deletions(-)

diff --git a/arch/x86/xen/enlighten.c b/arch/x86/xen/enlighten.c
index bea2152..e705bdf 100644
--- a/arch/x86/xen/enlighten.c
+++ b/arch/x86/xen/enlighten.c
@@ -53,6 +53,8 @@
 #include "mmu.h"
 #include "multicalls.h"
 
+#define _KERNPG_TABLE_RO __pgprot(_KERNPG_TABLE & ~_PAGE_RW)
+
 EXPORT_SYMBOL_GPL(hypercall_page);
 
 DEFINE_PER_CPU(struct vcpu_info *, xen_vcpu);
@@ -1487,9 +1489,9 @@ static __init void xen_map_identity_early(pmd_t *pmd, unsigned long
max_pfn)
 	}
 
 	for (pteidx = 0; pteidx < ident_pte; pteidx += PTRS_PER_PTE)
-		set_page_prot(&level1_ident_pgt[pteidx], PAGE_KERNEL_RO);
+		set_page_prot(&level1_ident_pgt[pteidx], _KERNPG_TABLE_RO);
 
-	set_page_prot(pmd, PAGE_KERNEL_RO);
+	set_page_prot(pmd, _KERNPG_TABLE_RO);
 }
 
 #ifdef CONFIG_X86_64
@@ -1543,12 +1545,12 @@ static __init pgd_t *xen_setup_kernel_pagetable(pgd_t *pgd,
 	xen_map_identity_early(level2_ident_pgt, max_pfn);
 
 	/* Make pagetable pieces RO */
-	set_page_prot(init_level4_pgt, PAGE_KERNEL_RO);
-	set_page_prot(level3_ident_pgt, PAGE_KERNEL_RO);
-	set_page_prot(level3_kernel_pgt, PAGE_KERNEL_RO);
-	set_page_prot(level3_user_vsyscall, PAGE_KERNEL_RO);
-	set_page_prot(level2_kernel_pgt, PAGE_KERNEL_RO);
-	set_page_prot(level2_fixmap_pgt, PAGE_KERNEL_RO);
+	set_page_prot(init_level4_pgt, _KERNPG_TABLE_RO);
+	set_page_prot(level3_ident_pgt, _KERNPG_TABLE_RO);
+	set_page_prot(level3_kernel_pgt, _KERNPG_TABLE_RO);
+	set_page_prot(level3_user_vsyscall, _KERNPG_TABLE_RO);
+	set_page_prot(level2_kernel_pgt, _KERNPG_TABLE_RO);
+	set_page_prot(level2_fixmap_pgt, _KERNPG_TABLE_RO);
 
 	/* Pin down new L4 */
 	pin_pagetable_pfn(MMUEXT_PIN_L4_TABLE,
@@ -1597,9 +1599,9 @@ static __init pgd_t *xen_setup_kernel_pagetable(pgd_t *pgd,
 	set_pgd(&swapper_pg_dir[KERNEL_PGD_BOUNDARY],
 			__pgd(__pa(level2_kernel_pgt) | _PAGE_PRESENT));
 
-	set_page_prot(level2_kernel_pgt, PAGE_KERNEL_RO);
-	set_page_prot(swapper_pg_dir, PAGE_KERNEL_RO);
-	set_page_prot(empty_zero_page, PAGE_KERNEL_RO);
+	set_page_prot(level2_kernel_pgt, _KERNPG_TABLE_RO);
+	set_page_prot(swapper_pg_dir, _KERNPG_TABLE_RO);
+	set_page_prot(empty_zero_page, _KERNPG_TABLE_RO);
 
 	pin_pagetable_pfn(MMUEXT_UNPIN_TABLE, PFN_DOWN(__pa(pgd)));
 
-- 
1.5.6.5

[Ian Campbell]

I meant to CC lkml and stable@ (some of the original reports are against
2.6.27) and fat-fingered the to: line.

A few previous threads on the subject:
http://lists.xensource.com/archives/html/xen-devel/2009-01/msg00176.html
http://www.redhat.com/archives/fedora-xen/2009-January/msg00022.html
http://www.redhat.com/archives/fedora-virt/2009-January/msg00013.html

Ian.

On Fri, 2009-01-30 at 18:19 +0000, Ian Campbell wrote:
>  __supported_pte_mask has not been correctly configured at this point
>  and Xen prevents us from using the NX bit if the hardware does not
>  support it. Some BIOSes seem to offer the option to disable NX.
>  
>  Signed-off-by: Ian Campbell <ian.campbell(a)citrix.com&gt;
>  Cc: Mark McLoughlin <markmc(a)redhat.com&gt;
>  Cc: Jon Swanson <jswanson(a)valuecommerce.co.jp&gt;
>  Cc: fedora-virt(a)redhat.com
>  Cc: Jeremy Fitzhardinge <jeremy(a)goop.org&gt;
>  Cc: Ingo Molnar <mingo(a)elte.hu&gt;
>  ---
>   arch/x86/xen/enlighten.c |   24 +++++++++++++-----------
>   1 files changed, 13 insertions(+), 11 deletions(-)
>  
>  diff --git a/arch/x86/xen/enlighten.c b/arch/x86/xen/enlighten.c
>  index bea2152..e705bdf 100644
>  --- a/arch/x86/xen/enlighten.c
>  +++ b/arch/x86/xen/enlighten.c
>  @@ -53,6 +53,8 @@
>   #include "mmu.h"
>   #include "multicalls.h"
>   
>  +#define _KERNPG_TABLE_RO __pgprot(_KERNPG_TABLE & ~_PAGE_RW)
>  +
>   EXPORT_SYMBOL_GPL(hypercall_page);
>   
>   DEFINE_PER_CPU(struct vcpu_info *, xen_vcpu);
>  @@ -1487,9 +1489,9 @@ static __init void xen_map_identity_early(pmd_t *pmd, unsigned long
> max_pfn)
>   	}
>   
>   	for (pteidx = 0; pteidx < ident_pte; pteidx += PTRS_PER_PTE)
>  -		set_page_prot(&level1_ident_pgt[pteidx], PAGE_KERNEL_RO);
>  +		set_page_prot(&level1_ident_pgt[pteidx], _KERNPG_TABLE_RO);
>   
>  -	set_page_prot(pmd, PAGE_KERNEL_RO);
>  +	set_page_prot(pmd, _KERNPG_TABLE_RO);
>   }
>   
>   #ifdef CONFIG_X86_64
>  @@ -1543,12 +1545,12 @@ static __init pgd_t *xen_setup_kernel_pagetable(pgd_t *pgd,
>   	xen_map_identity_early(level2_ident_pgt, max_pfn);
>   
>   	/* Make pagetable pieces RO */
>  -	set_page_prot(init_level4_pgt, PAGE_KERNEL_RO);
>  -	set_page_prot(level3_ident_pgt, PAGE_KERNEL_RO);
>  -	set_page_prot(level3_kernel_pgt, PAGE_KERNEL_RO);
>  -	set_page_prot(level3_user_vsyscall, PAGE_KERNEL_RO);
>  -	set_page_prot(level2_kernel_pgt, PAGE_KERNEL_RO);
>  -	set_page_prot(level2_fixmap_pgt, PAGE_KERNEL_RO);
>  +	set_page_prot(init_level4_pgt, _KERNPG_TABLE_RO);
>  +	set_page_prot(level3_ident_pgt, _KERNPG_TABLE_RO);
>  +	set_page_prot(level3_kernel_pgt, _KERNPG_TABLE_RO);
>  +	set_page_prot(level3_user_vsyscall, _KERNPG_TABLE_RO);
>  +	set_page_prot(level2_kernel_pgt, _KERNPG_TABLE_RO);
>  +	set_page_prot(level2_fixmap_pgt, _KERNPG_TABLE_RO);
>   
>   	/* Pin down new L4 */
>   	pin_pagetable_pfn(MMUEXT_PIN_L4_TABLE,
>  @@ -1597,9 +1599,9 @@ static __init pgd_t *xen_setup_kernel_pagetable(pgd_t *pgd,
>   	set_pgd(&swapper_pg_dir[KERNEL_PGD_BOUNDARY],
>   			__pgd(__pa(level2_kernel_pgt) | _PAGE_PRESENT));
>   
>  -	set_page_prot(level2_kernel_pgt, PAGE_KERNEL_RO);
>  -	set_page_prot(swapper_pg_dir, PAGE_KERNEL_RO);
>  -	set_page_prot(empty_zero_page, PAGE_KERNEL_RO);
>  +	set_page_prot(level2_kernel_pgt, _KERNPG_TABLE_RO);
>  +	set_page_prot(swapper_pg_dir, _KERNPG_TABLE_RO);
>  +	set_page_prot(empty_zero_page, _KERNPG_TABLE_RO);
>   
>   	pin_pagetable_pfn(MMUEXT_UNPIN_TABLE, PFN_DOWN(__pa(pgd)));
>

[Jeremy Fitzhardinge]

Ian Campbell wrote:
>  __supported_pte_mask has not been correctly configured at this point
>  and Xen prevents us from using the NX bit if the hardware does not
>  support it. Some BIOSes seem to offer the option to disable NX.
>     
Could we sniff EFER and update __supported_pte_mask accordingly?

    J

[Ian Campbell]

On Fri, 2009-01-30 at 10:56 -0800, Jeremy Fitzhardinge wrote:
>  Ian Campbell wrote:
>  > __supported_pte_mask has not been correctly configured at this point
>  > and Xen prevents us from using the NX bit if the hardware does not
>  > support it. Some BIOSes seem to offer the option to disable NX.
>  >   
>  Could we sniff EFER and update __supported_pte_mask accordingly? 
Perhaps, we might also have to handle the various noexec= command line
options? I don't suppose they matter so much in a guest though.

The equivalent native seems to use _KERNPG_TABLE as well (e.g.
head_64.S) -- is there something later on which comes along and tries to
apply the NX bit to those pages which didn't get it at start of day?

Ian.

[Virgil]

Hi list,

Thanks for all your efforts in trying to resolve the NX CPU capability
issue.

Turns out that in our circumstance, IBM xSeries 336 machines had
Irwindale? CPUs. Intel documentation says this was one of the first CPUs
to have Execute Disable (XD).

Herein lies some misunderstanding. NX really means XD that is:
"No Execute" is "Execute Disable" in IntelSpeak. Armed with this
knowledge I went into the bios and "Enabled Execute Disable".

The NX capability is now in the /proc/cpuinfo listing. The F10 kernel
works on the F8 host in 64 bit mode.

For IBM xSeries 336 owners, Execute Disable was Disabled by default (on
delivery).

Hope this helps someone out there.
Cheers
Phill.

On Fri, 2009-01-30 at 19:07 +0000, Ian Campbell wrote:
>  On Fri, 2009-01-30 at 10:56 -0800, Jeremy Fitzhardinge wrote:
>  > Ian Campbell wrote:
>  > > __supported_pte_mask has not been correctly configured at this point
>  > > and Xen prevents us from using the NX bit if the hardware does not
>  > > support it. Some BIOSes seem to offer the option to disable NX.
>  > >   
>  > Could we sniff EFER and update __supported_pte_mask accordingly?
>  
>  Perhaps, we might also have to handle the various noexec= command line
>  options? I don't suppose they matter so much in a guest though.
>  
>  The equivalent native seems to use _KERNPG_TABLE as well (e.g.
>  head_64.S) -- is there something later on which comes along and tries to
>  apply the NX bit to those pages which didn't get it at start of day?
>  
>  Ian.
>  
>  _______________________________________________
>  Fedora-virt mailing list
>  Fedora-virt(a)redhat.com
>  https://www.redhat.com/mailman/listinfo/fedora-virt
>