* [PATCH 1/2] PV Hugetlb support - Hypervisor portion
2008-07-24 13:45 [PATCH 0/2] Add hugetlb support for PV Dave McCracken
@ 2008-07-24 13:48 ` Dave McCracken
2008-07-24 13:51 ` Dave McCracken
2008-07-24 13:50 ` [PATCH 2/2] PV hugetlb support - kernel portion Dave McCracken
2008-07-24 14:23 ` [PATCH 0/2] Add hugetlb support for PV Jan Beulich
2 siblings, 1 reply; 6+ messages in thread
From: Dave McCracken @ 2008-07-24 13:48 UTC (permalink / raw)
To: xen-devel
[-- Attachment #1: Type: text/plain, Size: 69 bytes --]
Here is the hypervisor portion of hugetlb support for PV machines.
[-- Attachment #2: linux-xen-hpage-02.patch --]
[-- Type: text/x-diff, Size: 3193 bytes --]
--- linux-2.6.18-xen/./fs/Kconfig 2008-07-17 09:54:14.000000000 -0500
+++ linux-hpage/./fs/Kconfig 2008-07-23 09:31:06.000000000 -0500
@@ -870,7 +870,7 @@ config TMPFS
config HUGETLBFS
bool "HugeTLB file system support"
depends X86 || IA64 || PPC64 || SPARC64 || SUPERH || BROKEN
- depends on !XEN
+# depends on !XEN
help
hugetlbfs is a filesystem backing for HugeTLB pages, based on
ramfs. For architectures that support it, say Y here and read
--- linux-2.6.18-xen/./include/asm-x86_64/mach-xen/asm/page.h 2008-07-17 09:54:18.000000000 -0500
+++ linux-hpage/./include/asm-x86_64/mach-xen/asm/page.h 2008-07-23 09:31:06.000000000 -0500
@@ -62,6 +62,8 @@
#define HPAGE_MASK (~(HPAGE_SIZE - 1))
#define HUGETLB_PAGE_ORDER (HPAGE_SHIFT - PAGE_SHIFT)
+#define ARCH_HAS_SETCLEAR_HUGE_PTE
+
#ifdef __KERNEL__
#ifndef __ASSEMBLY__
--- linux-2.6.18-xen/./include/asm-x86_64/mach-xen/asm/pgtable.h 2008-07-17 09:54:18.000000000 -0500
+++ linux-hpage/./include/asm-x86_64/mach-xen/asm/pgtable.h 2008-07-23 09:31:06.000000000 -0500
@@ -261,6 +261,12 @@ static inline unsigned long pud_bad(pud_
set_pte((ptep), (pteval)); \
} while (0)
+static inline void set_huge_pte_at(struct mm_struct *mm, unsigned long addr, pte_t *ptep, pte_t pteval) {
+ if ((mm != current->mm && mm != &init_mm) ||
+ HYPERVISOR_update_va_mapping(addr, pteval, 0))
+ set_pmd((pmd_t *)ptep, (pmd_t){__pte_val(pteval)});
+}
+
#define pte_none(x) (!(x).pte)
#define pte_present(x) ((x).pte & (_PAGE_PRESENT | _PAGE_PROTNONE))
#define pte_clear(mm,addr,xp) do { set_pte_at(mm, addr, xp, __pte(0)); } while (0)
@@ -296,6 +302,19 @@ static inline pte_t ptep_get_and_clear(s
return pte;
}
+static inline pte_t huge_ptep_get_and_clear(struct mm_struct *mm, unsigned long addr, pte_t *ptep)
+{
+ pte_t pte = *ptep;
+ if (!pte_none(pte)) {
+ if ((mm != &init_mm) ||
+ HYPERVISOR_update_va_mapping(addr, __pte(0), 0)) {
+ pte = *ptep;
+ set_pmd((pmd_t *)ptep, __pmd(0));
+ }
+ }
+ return pte;
+}
+
static inline pte_t ptep_get_and_clear_full(struct mm_struct *mm, unsigned long addr, pte_t *ptep, int full)
{
if (full) {
--- linux-2.6.18-xen/./arch/x86_64/mm/pageattr-xen.c 2008-07-17 09:54:10.000000000 -0500
+++ linux-hpage/./arch/x86_64/mm/pageattr-xen.c 2008-07-23 09:31:06.000000000 -0500
@@ -62,6 +62,8 @@ static void _pin_lock(struct mm_struct *
if (pmd_none(*pmd))
continue;
+ if (pte_huge(*(pte_t *)pmd))
+ continue;
ptl = pte_lockptr(0, pmd);
if (lock)
spin_lock(ptl);
--- linux-2.6.18-xen/./mm/hugetlb.c 2008-07-17 09:54:19.000000000 -0500
+++ linux-hpage/./mm/hugetlb.c 2008-07-23 09:31:06.000000000 -0500
@@ -294,12 +294,14 @@ static pte_t make_huge_pte(struct vm_are
int writable)
{
pte_t entry;
+ pgprot_t pgprot;
+ pgprot = __pgprot(pgprot_val(vma->vm_page_prot) | _PAGE_PRESENT);
if (writable) {
entry =
- pte_mkwrite(pte_mkdirty(mk_pte(page, vma->vm_page_prot)));
+ pte_mkwrite(pte_mkdirty(mk_pte(page, pgprot)));
} else {
- entry = pte_wrprotect(mk_pte(page, vma->vm_page_prot));
+ entry = pte_wrprotect(mk_pte(page, pgprot));
}
entry = pte_mkyoung(entry);
entry = pte_mkhuge(entry);
[-- Attachment #3: Type: text/plain, Size: 138 bytes --]
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel
^ permalink raw reply [flat|nested] 6+ messages in thread* [PATCH 1/2] PV Hugetlb support - Hypervisor portion
2008-07-24 13:48 ` [PATCH 1/2] PV Hugetlb support - Hypervisor portion Dave McCracken
@ 2008-07-24 13:51 ` Dave McCracken
0 siblings, 0 replies; 6+ messages in thread
From: Dave McCracken @ 2008-07-24 13:51 UTC (permalink / raw)
To: xen-devel
[-- Attachment #1: Type: text/plain, Size: 243 bytes --]
On Thursday 24 July 2008, Dave McCracken wrote:
> Here is the hypervisor portion of hugetlb support for PV machines.
Oops. Here's really the hypervisor portion. The first one was a copy of the
kernel portion.
Dave McCracken
Oracle Corp.
[-- Attachment #2: xen-hpage-02.patch --]
[-- Type: text/x-diff, Size: 1923 bytes --]
--- xen-unstable/./xen/include/asm-x86/x86_32/page.h 2008-07-17 09:49:27.000000000 -0500
+++ xen-hpage/./xen/include/asm-x86/x86_32/page.h 2008-07-23 09:30:40.000000000 -0500
@@ -112,7 +112,7 @@ extern unsigned int PAGE_HYPERVISOR_NOCA
* Disallow unused flag bits plus PAT/PSE, PCD, PWT and GLOBAL.
* Permit the NX bit if the hardware supports it.
*/
-#define BASE_DISALLOW_MASK (0xFFFFF198U & ~_PAGE_NX)
+#define BASE_DISALLOW_MASK (0xFFFFF118U & ~_PAGE_NX)
#define L1_DISALLOW_MASK (BASE_DISALLOW_MASK | _PAGE_GNTTAB)
#define L2_DISALLOW_MASK (BASE_DISALLOW_MASK)
--- xen-unstable/./xen/include/asm-x86/x86_64/page.h 2008-07-17 09:49:27.000000000 -0500
+++ xen-hpage/./xen/include/asm-x86/x86_64/page.h 2008-07-23 09:30:40.000000000 -0500
@@ -109,7 +109,7 @@ typedef l4_pgentry_t root_pgentry_t;
* Permit the NX bit if the hardware supports it.
* Note that range [62:52] is available for software use on x86/64.
*/
-#define BASE_DISALLOW_MASK (0xFF800198U & ~_PAGE_NX)
+#define BASE_DISALLOW_MASK (0xFF800118U & ~_PAGE_NX)
#define L1_DISALLOW_MASK (BASE_DISALLOW_MASK | _PAGE_GNTTAB)
#define L2_DISALLOW_MASK (BASE_DISALLOW_MASK)
--- xen-unstable/./xen/arch/x86/mm.c 2008-07-17 09:49:27.000000000 -0500
+++ xen-hpage/./xen/arch/x86/mm.c 2008-07-23 09:30:40.000000000 -0500
@@ -755,6 +755,9 @@ get_page_from_l2e(
MEM_LOG("Bad L2 flags %x", l2e_get_flags(l2e) & L2_DISALLOW_MASK);
return 0;
}
+ if ( l2e_get_flags(l2e) & _PAGE_PSE )
+ return 1;
+
rc = get_page_and_type_from_pagenr(l2e_get_pfn(l2e), PGT_l1_page_table, d);
if ( unlikely(!rc) )
@@ -948,6 +951,7 @@ void put_page_from_l1e(l1_pgentry_t l1e,
static void put_page_from_l2e(l2_pgentry_t l2e, unsigned long pfn)
{
if ( (l2e_get_flags(l2e) & _PAGE_PRESENT) &&
+ (!(l2e_get_flags(l2e) & _PAGE_PSE)) &&
(l2e_get_pfn(l2e) != pfn) )
put_page_and_type(l2e_get_page(l2e));
}
[-- Attachment #3: Type: text/plain, Size: 138 bytes --]
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel
^ permalink raw reply [flat|nested] 6+ messages in thread
* [PATCH 2/2] PV hugetlb support - kernel portion
2008-07-24 13:45 [PATCH 0/2] Add hugetlb support for PV Dave McCracken
2008-07-24 13:48 ` [PATCH 1/2] PV Hugetlb support - Hypervisor portion Dave McCracken
@ 2008-07-24 13:50 ` Dave McCracken
2008-07-24 14:23 ` [PATCH 0/2] Add hugetlb support for PV Jan Beulich
2 siblings, 0 replies; 6+ messages in thread
From: Dave McCracken @ 2008-07-24 13:50 UTC (permalink / raw)
To: xen-devel
[-- Attachment #1: Type: text/plain, Size: 114 bytes --]
Here is the patch against the kernel that enables hugetlb support for PV
machines.
Dave McCracken
Oracle Corp.
[-- Attachment #2: linux-xen-hpage-02.patch --]
[-- Type: text/x-diff, Size: 3193 bytes --]
--- linux-2.6.18-xen/./fs/Kconfig 2008-07-17 09:54:14.000000000 -0500
+++ linux-hpage/./fs/Kconfig 2008-07-23 09:31:06.000000000 -0500
@@ -870,7 +870,7 @@ config TMPFS
config HUGETLBFS
bool "HugeTLB file system support"
depends X86 || IA64 || PPC64 || SPARC64 || SUPERH || BROKEN
- depends on !XEN
+# depends on !XEN
help
hugetlbfs is a filesystem backing for HugeTLB pages, based on
ramfs. For architectures that support it, say Y here and read
--- linux-2.6.18-xen/./include/asm-x86_64/mach-xen/asm/page.h 2008-07-17 09:54:18.000000000 -0500
+++ linux-hpage/./include/asm-x86_64/mach-xen/asm/page.h 2008-07-23 09:31:06.000000000 -0500
@@ -62,6 +62,8 @@
#define HPAGE_MASK (~(HPAGE_SIZE - 1))
#define HUGETLB_PAGE_ORDER (HPAGE_SHIFT - PAGE_SHIFT)
+#define ARCH_HAS_SETCLEAR_HUGE_PTE
+
#ifdef __KERNEL__
#ifndef __ASSEMBLY__
--- linux-2.6.18-xen/./include/asm-x86_64/mach-xen/asm/pgtable.h 2008-07-17 09:54:18.000000000 -0500
+++ linux-hpage/./include/asm-x86_64/mach-xen/asm/pgtable.h 2008-07-23 09:31:06.000000000 -0500
@@ -261,6 +261,12 @@ static inline unsigned long pud_bad(pud_
set_pte((ptep), (pteval)); \
} while (0)
+static inline void set_huge_pte_at(struct mm_struct *mm, unsigned long addr, pte_t *ptep, pte_t pteval) {
+ if ((mm != current->mm && mm != &init_mm) ||
+ HYPERVISOR_update_va_mapping(addr, pteval, 0))
+ set_pmd((pmd_t *)ptep, (pmd_t){__pte_val(pteval)});
+}
+
#define pte_none(x) (!(x).pte)
#define pte_present(x) ((x).pte & (_PAGE_PRESENT | _PAGE_PROTNONE))
#define pte_clear(mm,addr,xp) do { set_pte_at(mm, addr, xp, __pte(0)); } while (0)
@@ -296,6 +302,19 @@ static inline pte_t ptep_get_and_clear(s
return pte;
}
+static inline pte_t huge_ptep_get_and_clear(struct mm_struct *mm, unsigned long addr, pte_t *ptep)
+{
+ pte_t pte = *ptep;
+ if (!pte_none(pte)) {
+ if ((mm != &init_mm) ||
+ HYPERVISOR_update_va_mapping(addr, __pte(0), 0)) {
+ pte = *ptep;
+ set_pmd((pmd_t *)ptep, __pmd(0));
+ }
+ }
+ return pte;
+}
+
static inline pte_t ptep_get_and_clear_full(struct mm_struct *mm, unsigned long addr, pte_t *ptep, int full)
{
if (full) {
--- linux-2.6.18-xen/./arch/x86_64/mm/pageattr-xen.c 2008-07-17 09:54:10.000000000 -0500
+++ linux-hpage/./arch/x86_64/mm/pageattr-xen.c 2008-07-23 09:31:06.000000000 -0500
@@ -62,6 +62,8 @@ static void _pin_lock(struct mm_struct *
if (pmd_none(*pmd))
continue;
+ if (pte_huge(*(pte_t *)pmd))
+ continue;
ptl = pte_lockptr(0, pmd);
if (lock)
spin_lock(ptl);
--- linux-2.6.18-xen/./mm/hugetlb.c 2008-07-17 09:54:19.000000000 -0500
+++ linux-hpage/./mm/hugetlb.c 2008-07-23 09:31:06.000000000 -0500
@@ -294,12 +294,14 @@ static pte_t make_huge_pte(struct vm_are
int writable)
{
pte_t entry;
+ pgprot_t pgprot;
+ pgprot = __pgprot(pgprot_val(vma->vm_page_prot) | _PAGE_PRESENT);
if (writable) {
entry =
- pte_mkwrite(pte_mkdirty(mk_pte(page, vma->vm_page_prot)));
+ pte_mkwrite(pte_mkdirty(mk_pte(page, pgprot)));
} else {
- entry = pte_wrprotect(mk_pte(page, vma->vm_page_prot));
+ entry = pte_wrprotect(mk_pte(page, pgprot));
}
entry = pte_mkyoung(entry);
entry = pte_mkhuge(entry);
[-- Attachment #3: Type: text/plain, Size: 138 bytes --]
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel
^ permalink raw reply [flat|nested] 6+ messages in thread* Re: [PATCH 0/2] Add hugetlb support for PV
2008-07-24 13:45 [PATCH 0/2] Add hugetlb support for PV Dave McCracken
2008-07-24 13:48 ` [PATCH 1/2] PV Hugetlb support - Hypervisor portion Dave McCracken
2008-07-24 13:50 ` [PATCH 2/2] PV hugetlb support - kernel portion Dave McCracken
@ 2008-07-24 14:23 ` Jan Beulich
2008-07-24 14:50 ` Jeremy Fitzhardinge
2 siblings, 1 reply; 6+ messages in thread
From: Jan Beulich @ 2008-07-24 14:23 UTC (permalink / raw)
To: Dave McCracken; +Cc: xen-devel
>>> Dave McCracken <dcm@mccr.org> 24.07.08 15:45 >>>
>
>Here is a set of small patches that enables hugetlb on PV machines. They are
>against xen-unstable and linux-2.6.18-xen.
>
>The patches are specifically for x86_64.
>
>I originally had these patches working back in May, but recently for some
>reason my machine now refuses to boot the baseline xen-unstable, so I've been
>unable to verify them. However, they should still work.
>
>Comments, bug reports, and suggestions are welcome.
I must be missing something, but I can't see how you guarantee the
super pages to be contiguous in machine memory. And even if you did,
the hypervisor would still need to do accounting on all the 4k portions
a super page consists of when it gets mapped/unmapped.
Jan
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: [PATCH 0/2] Add hugetlb support for PV
2008-07-24 14:23 ` [PATCH 0/2] Add hugetlb support for PV Jan Beulich
@ 2008-07-24 14:50 ` Jeremy Fitzhardinge
0 siblings, 0 replies; 6+ messages in thread
From: Jeremy Fitzhardinge @ 2008-07-24 14:50 UTC (permalink / raw)
To: Jan Beulich; +Cc: Dave McCracken, xen-devel
Jan Beulich wrote:
>>>> Dave McCracken <dcm@mccr.org> 24.07.08 15:45 >>>
>>>>
>> Here is a set of small patches that enables hugetlb on PV machines. They are
>> against xen-unstable and linux-2.6.18-xen.
>>
>> The patches are specifically for x86_64.
>>
>> I originally had these patches working back in May, but recently for some
>> reason my machine now refuses to boot the baseline xen-unstable, so I've been
>> unable to verify them. However, they should still work.
>>
>> Comments, bug reports, and suggestions are welcome.
>>
>
> I must be missing something, but I can't see how you guarantee the
> super pages to be contiguous in machine memory. And even if you did,
> the hypervisor would still need to do accounting on all the 4k portions
> a super page consists of when it gets mapped/unmapped.
>
Yeah, I'm pretty sure there has to be a lot more to it than that.
J
^ permalink raw reply [flat|nested] 6+ messages in thread