[PATCH 0/4] paravirt_ops-64 compile fixes

[PATCH 0/4] paravirt_ops-64 compile fixes

[Eduardo Habkost]

This series contain fixes to make the paravirt_ops code compile and boot
on x86_64.

This is a follow-up for the previous series from Glauber.

[PATCH 1/4] Add missing semicolon to SWAPGS macro

[Eduardo Habkost]

Trivial compile fix.

Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
---
 include/asm-x86/paravirt.h |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)

diff --git a/include/asm-x86/paravirt.h b/include/asm-x86/paravirt.h
index 52bcd9d..62cecb7 100644
--- a/include/asm-x86/paravirt.h
+++ b/include/asm-x86/paravirt.h
@@ -1329,7 +1329,7 @@ static inline unsigned long __raw_local_irq_save(void)
 #else
 #define SWAPGS								\
 	PARA_SITE(PARA_PATCH(pv_cpu_ops, PV_CPU_swapgs), CLBR_NONE,	\
-		  PV_SAVE_REGS						\
+		  PV_SAVE_REGS;						\
 		  call *pv_cpu_ops+PV_CPU_swapgs;			\
 		  PV_RESTORE_REGS					\
 		 )
-- 
1.5.3.4

[PATCH 2/4] include/asm-x86/paravirt.h: x86_64 mmu operations

[Eduardo Habkost]

Add .set_pgd field to pv_mmu_ops.

Implement pud_val(), __pud(), set_pgd(), pud_clear(), pgd_clear().

pud_clear() and pgd_clear() are implemented simply using set_pud()
and set_pmd(). They don't have a field at pv_mmu_ops.

Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
---
 include/asm-x86/paravirt.h |   55 ++++++++++++++++++++++++++++++++++++++++++++
 1 files changed, 55 insertions(+), 0 deletions(-)

diff --git a/include/asm-x86/paravirt.h b/include/asm-x86/paravirt.h
index 62cecb7..9f9ff57 100644
--- a/include/asm-x86/paravirt.h
+++ b/include/asm-x86/paravirt.h
@@ -259,6 +259,8 @@ struct pv_mmu_ops {
 #if PAGETABLE_LEVELS == 4
 	pudval_t (*pud_val)(pud_t);
 	pud_t (*make_pud)(pudval_t pud);
+
+	void (*set_pgd)(pgd_t *pudp, pgd_t pgdval);
 #endif	/* PAGETABLE_LEVELS == 4 */
 #endif	/* PAGETABLE_LEVELS >= 3 */
 
@@ -1065,6 +1067,59 @@ static inline void set_pud(pud_t *pudp, pud_t pud)
 		PVOP_VCALL2(pv_mmu_ops.set_pud, pudp,
 			    val);
 }
+#if PAGETABLE_LEVELS == 4
+static inline pud_t __pud(pudval_t val)
+{
+	pudval_t ret;
+
+	if (sizeof(pudval_t) > sizeof(long))
+		ret = PVOP_CALL2(pudval_t, pv_mmu_ops.make_pud,
+				 val, (u64)val >> 32);
+	else
+		ret = PVOP_CALL1(pudval_t, pv_mmu_ops.make_pud,
+				 val);
+
+	return (pud_t) { ret };
+}
+
+static inline pudval_t pud_val(pud_t pud)
+{
+	pudval_t ret;
+
+	if (sizeof(pudval_t) > sizeof(long))
+		ret =  PVOP_CALL2(pudval_t, pv_mmu_ops.pud_val,
+				  pud.pud, (u64)pud.pud >> 32);
+	else
+		ret =  PVOP_CALL1(pudval_t, pv_mmu_ops.pud_val,
+				  pud.pud);
+
+	return ret;
+}
+
+static inline void set_pgd(pgd_t *pgdp, pgd_t pgd)
+{
+	pgdval_t val = native_pgd_val(pgd);
+
+	if (sizeof(pgdval_t) > sizeof(long))
+		PVOP_VCALL3(pv_mmu_ops.set_pgd, pgdp,
+			    val, (u64)val >> 32);
+	else
+		PVOP_VCALL2(pv_mmu_ops.set_pgd, pgdp,
+			    val);
+}
+
+static inline void pgd_clear(pgd_t *pgdp)
+{
+	set_pgd(pgdp, __pgd(0));
+}
+
+static inline void pud_clear(pud_t *pudp)
+{
+	set_pud(pudp, __pud(0));
+}
+
+#endif	/* PAGETABLE_LEVELS == 4 */
+
 #endif	/* PAGETABLE_LEVELS >= 3 */
 
 #ifdef CONFIG_X86_PAE
-- 
1.5.3.4

[PATCH 3/4] Don't set pagetable_setup_{start,done} hooks on x86_64

[Eduardo Habkost]

paravirt_pagetable_setup_{start,done}() are not used (yet) under x86_64,
and native_pagetable_setup_{start,done}() don't exist on x86_64. So they
don't need to be set.

Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
---
 arch/x86/kernel/paravirt.c |    2 ++
 1 files changed, 2 insertions(+), 0 deletions(-)

diff --git a/arch/x86/kernel/paravirt.c b/arch/x86/kernel/paravirt.c
index c67d331..37f38b7 100644
--- a/arch/x86/kernel/paravirt.c
+++ b/arch/x86/kernel/paravirt.c
@@ -367,8 +367,10 @@ struct pv_apic_ops pv_apic_ops = {
 };
 
 struct pv_mmu_ops pv_mmu_ops = {
+#ifndef CONFIG_X86_64
 	.pagetable_setup_start = native_pagetable_setup_start,
 	.pagetable_setup_done = native_pagetable_setup_done,
+#endif
 
 	.read_cr2 = native_read_cr2,
 	.write_cr2 = native_write_cr2,
-- 
1.5.3.4

[PATCH 4/4] Fill in missing pv_mmu_ops entries for PAGETABLE_LEVELS >= 3

[Eduardo Habkost]

This finally makes paravirt-ops able to compile and boot under x86_64.

Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
---
 arch/x86/kernel/paravirt.c |   11 +++++++++--
 1 files changed, 9 insertions(+), 2 deletions(-)

diff --git a/arch/x86/kernel/paravirt.c b/arch/x86/kernel/paravirt.c
index 37f38b7..075962c 100644
--- a/arch/x86/kernel/paravirt.c
+++ b/arch/x86/kernel/paravirt.c
@@ -398,16 +398,23 @@ struct pv_mmu_ops pv_mmu_ops = {
 	.kmap_atomic_pte = kmap_atomic,
 #endif
 
+#if PAGETABLE_LEVELS >= 3
 #ifdef CONFIG_X86_PAE
 	.set_pte_atomic = native_set_pte_atomic,
 	.set_pte_present = native_set_pte_present,
-	.set_pud = native_set_pud,
 	.pte_clear = native_pte_clear,
 	.pmd_clear = native_pmd_clear,
-
+#endif
+	.set_pud = native_set_pud,
 	.pmd_val = native_pmd_val,
 	.make_pmd = native_make_pmd,
+
+#if PAGETABLE_LEVELS == 4
+	.pud_val = native_pud_val,
+	.make_pud = native_make_pud,
+	.set_pgd = native_set_pgd,
 #endif
+#endif /* PAGETABLE_LEVELS >= 3 */
 
 	.pte_val = native_pte_val,
 	.pgd_val = native_pgd_val,
-- 
1.5.3.4

Re: [PATCH 0/4] paravirt_ops-64 compile fixes

[Ingo Molnar]

* Eduardo Habkost <ehabkost@redhat.com> wrote:

> This series contain fixes to make the paravirt_ops code compile and 
> boot on x86_64.
> 
> This is a follow-up for the previous series from Glauber.

thanks Eduardo, i've applied your fixes.

could you perhaps send the missing bits that add the missing 64-bit 
Kconfig entries? Even if the 64-bit side is not functional yet as a real 
guest, it would help us find the build bugs and any runtime regressions 
PARAVIRT might cause on the native 64-bit side of the kernel.

Or is it now just a matter of removing all the depends-on X86_32 bits in 
arch/x86/Kconfig?

	Ingo

Re: [PATCH 0/4] paravirt_ops-64 compile fixes

[Eduardo Pereira Habkost]

On Tue, Jan 22, 2008 at 01:02:07PM +0100, Ingo Molnar wrote:
> 
> * Eduardo Habkost <ehabkost@redhat.com> wrote:
> 
> > This series contain fixes to make the paravirt_ops code compile and 
> > boot on x86_64.
> > 
> > This is a follow-up for the previous series from Glauber.
> 
> thanks Eduardo, i've applied your fixes.
> 
> could you perhaps send the missing bits that add the missing 64-bit 
> Kconfig entries? Even if the 64-bit side is not functional yet as a real 
> guest, it would help us find the build bugs and any runtime regressions 
> PARAVIRT might cause on the native 64-bit side of the kernel.
> 
> Or is it now just a matter of removing all the depends-on X86_32 bits in 
> arch/x86/Kconfig?

Removing the depends won't be enough because CONFIG_PARAVIRT won't be
selected unless one of Xen, VMI, or lguest is selected. (PARAVIRT is
not visible on menuconfig, only PARAVIRT_GUEST, that doesn't enable any
code).

When hacking, I remove the 32-bit depends and add "select PARAVIRT" to
"menuconfig PARAVIRT_GUEST".

I think neither of the three guest implementations are ready for 64-bit
yet. I am working on Xen and it is far from being ready for inclusion. I
don't know about VMI and lguest.

If this is desirable for broader testing, we can make config PARAVIRT
visible and selectable on menuconfig, even when there is no guest
implementation being enabled. Should I do it?

Maybe should this force-paravirt mode be a kernel debugging option?

-- 
Eduardo

Re: [PATCH 0/4] paravirt_ops-64 compile fixes

[Jeremy Fitzhardinge]

Eduardo Pereira Habkost wrote:
> If this is desirable for broader testing, we can make config PARAVIRT
> visible and selectable on menuconfig, even when there is no guest
> implementation being enabled. Should I do it?
>   

Yes please.

> Maybe should this force-paravirt mode be a kernel debugging option?
>   

No, just a plain CONFIG_PARAVIRT to turn it on and off independently of 
everything else.

    J

[PATCH 1/2] Allow enabling PARAVIRT without any guest implementation

[Eduardo Pereira Habkost]

On Tue, Jan 22, 2008 at 09:55:41AM -0800, Jeremy Fitzhardinge wrote:
> Eduardo Pereira Habkost wrote:
> >If this is desirable for broader testing, we can make config PARAVIRT
> >visible and selectable on menuconfig, even when there is no guest
> >implementation being enabled. Should I do it?
> >  
> 
> Yes please.
> 

This will allow people to enable the paravirt_ops code even when no
guest support is enabled, for broader testing of the paravirt_ops code.

Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
---
 arch/x86/Kconfig |   18 +++++++++---------
 1 files changed, 9 insertions(+), 9 deletions(-)

diff --git a/arch/x86/Kconfig b/arch/x86/Kconfig
index 4e910d8..715bbcd 100644
--- a/arch/x86/Kconfig
+++ b/arch/x86/Kconfig
@@ -325,15 +325,6 @@ config SCHED_NO_NO_OMIT_FRAME_POINTER
 
 	  If in doubt, say "Y".
 
-config PARAVIRT
-	bool
-	depends on X86_32 && !(X86_VISWS || X86_VOYAGER)
-	help
-	  This changes the kernel so it can modify itself when it is run
-	  under a hypervisor, potentially improving performance significantly
-	  over full virtualization.  However, when run without a hypervisor
-	  the kernel is theoretically slower and slightly larger.
-
 menuconfig PARAVIRT_GUEST
 	bool "Paravirtualized guest support"
 	depends on X86_32
@@ -359,6 +350,15 @@ config VMI
 
 source "arch/x86/lguest/Kconfig"
 
+config PARAVIRT
+	bool "Enable paravirtualization code"
+	depends on X86_32 && !(X86_VISWS || X86_VOYAGER)
+	help
+	  This changes the kernel so it can modify itself when it is run
+	  under a hypervisor, potentially improving performance significantly
+	  over full virtualization.  However, when run without a hypervisor
+	  the kernel is theoretically slower and slightly larger.
+
 endif
 
 config ACPI_SRAT
-- 
1.5.3.4

-- 
Eduardo

[PATCH 2/2] Remove depends on X86_32 from PARAVIRT & PARAVIRT_GUEST

[Eduardo Pereira Habkost]

With this, the paravirt_ops code can be enabled on x86_64 also.

Each guest implementation (Xen, VMI, lguest) now depends on X86_32. The
dependencies can be dropped for each one when they start to support
x86_64.

Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
---
 arch/x86/Kconfig        |    4 ++--
 arch/x86/lguest/Kconfig |    1 +
 arch/x86/xen/Kconfig    |    1 +
 3 files changed, 4 insertions(+), 2 deletions(-)

diff --git a/arch/x86/Kconfig b/arch/x86/Kconfig
index 715bbcd..43d535a 100644
--- a/arch/x86/Kconfig
+++ b/arch/x86/Kconfig
@@ -327,7 +327,6 @@ config SCHED_NO_NO_OMIT_FRAME_POINTER
 
 menuconfig PARAVIRT_GUEST
 	bool "Paravirtualized guest support"
-	depends on X86_32
 	help
 	  Say Y here to get to see options related to running Linux under
 	  various hypervisors.  This option alone does not add any kernel code.
@@ -341,6 +340,7 @@ source "arch/x86/xen/Kconfig"
 config VMI
 	bool "VMI Guest support"
 	select PARAVIRT
+	depends on X86_32
 	depends on !(X86_VISWS || X86_VOYAGER)
 	help
 	  VMI provides a paravirtualized interface to the VMware ESX server
@@ -352,7 +352,7 @@ source "arch/x86/lguest/Kconfig"
 
 config PARAVIRT
 	bool "Enable paravirtualization code"
-	depends on X86_32 && !(X86_VISWS || X86_VOYAGER)
+	depends on !(X86_VISWS || X86_VOYAGER)
 	help
 	  This changes the kernel so it can modify itself when it is run
 	  under a hypervisor, potentially improving performance significantly
diff --git a/arch/x86/lguest/Kconfig b/arch/x86/lguest/Kconfig
index 19626ac..964dfa3 100644
--- a/arch/x86/lguest/Kconfig
+++ b/arch/x86/lguest/Kconfig
@@ -1,6 +1,7 @@
 config LGUEST_GUEST
 	bool "Lguest guest support"
 	select PARAVIRT
+	depends on X86_32
 	depends on !X86_PAE
 	depends on !(X86_VISWS || X86_VOYAGER)
 	select VIRTIO
diff --git a/arch/x86/xen/Kconfig b/arch/x86/xen/Kconfig
index fbfa55c..4d5f264 100644
--- a/arch/x86/xen/Kconfig
+++ b/arch/x86/xen/Kconfig
@@ -5,6 +5,7 @@
 config XEN
 	bool "Xen guest support"
 	select PARAVIRT
+	depends on X86_32
 	depends on X86_CMPXCHG && X86_TSC && !NEED_MULTIPLE_NODES && !(X86_VISWS || X86_VOYAGER)
 	help
 	  This is the Linux Xen port.  Enabling this will allow the
-- 
1.5.3.4

Re: [PATCH 0/4] paravirt_ops-64 compile fixes

[Ingo Molnar]

* Jeremy Fitzhardinge <jeremy@goop.org> wrote:

>> Maybe should this force-paravirt mode be a kernel debugging option?
>
> No, just a plain CONFIG_PARAVIRT to turn it on and off independently 
> of everything else.

it should be a debugging option in the way that it's dependent on 
KERNEL_DEBUG and KERNEL_EXPERIMENTAL.

i dont think you shold worry about this - i think 64-bit guest support 
for specific hypervisors could be added even after the merge window, if 
it's reasonably isolated (which i'd expect it to be). It clearly cannot 
break anything else so it's the same category as new driver or new 
hardware support - more relaxed integration rules. Then we can remove 
the KERNEL_DEBUG and KERNEL_EXPERIMENTAL dependency as well.

	Ingo

Re: [PATCH 0/4] paravirt_ops-64 compile fixes

[Jeremy Fitzhardinge]

Ingo Molnar wrote:
> * Jeremy Fitzhardinge <jeremy@goop.org> wrote:
>
>   
>>> Maybe should this force-paravirt mode be a kernel debugging option?
>>>       
>> No, just a plain CONFIG_PARAVIRT to turn it on and off independently 
>> of everything else.
>>     
>
> it should be a debugging option in the way that it's dependent on 
> KERNEL_DEBUG and KERNEL_EXPERIMENTAL.
>   

EXPERIMENTAL certainly, but not DEBUG.  Sure, CONFIG_PARAVIRT is 
somewhat useless if you don't also configure Xen/lguest/etc, but running 
a paravirt system on bare hardware will probably be the most common way 
a paravirt kernel gets run (a distro will always enable it and compile 
in whatever hypervisor support they want, but most people will end up 
just running it on bare hardware).  Therefore, we should try to be as 
sure as possible that PARAVIRT works well (ie, indistinguishable from 
non-PARAVIRT) on bare hardware.

> i dont think you shold worry about this - i think 64-bit guest support 
> for specific hypervisors could be added even after the merge window, if 
> it's reasonably isolated (which i'd expect it to be). It clearly cannot 
> break anything else so it's the same category as new driver or new 
> hardware support - more relaxed integration rules. Then we can remove 
> the KERNEL_DEBUG and KERNEL_EXPERIMENTAL dependency as well.

Good, I'm glad you feel that way.

    J