[PATCH 1/2] powerpc: add hugepagesz boot-time parameter

[PATCH 1/2] powerpc: add hugepagesz boot-time parameter

[Jon Tollefson]

This patch adds the hugepagesz boot-time parameter for ppc64 that lets 
you pick the size for your huge pages.  The choices available are 64K 
and 16M.  It defaults to 16M (previously the only choice) if nothing or 
an invalid choice is specified.  Tested 64K huge pages with the 
libhugetlbfs 1.2 release with its 'make func' and 'make stress' test 
invocations.

This patch requires the patch posted by Mel Gorman that adds 
HUGETLB_PAGE_SIZE_VARIABLE; "[PATCH] Fix boot problem with iSeries 
lacking hugepage support" on 2007-11-15.

Signed-off-by: Jon Tollefson <kniht@linux.vnet.ibm.com>
---

 Documentation/kernel-parameters.txt |    1 
 arch/powerpc/mm/hash_utils_64.c     |   11 +--------
 arch/powerpc/mm/hugetlbpage.c       |   41 ++++++++++++++++++++++++++++++++++++
 include/asm-powerpc/mmu-hash64.h    |    1 
 mm/hugetlb.c                        |    1 
 5 files changed, 46 insertions(+), 9 deletions(-)

diff --git a/Documentation/kernel-parameters.txt b/Documentation/kernel-parameters.txt
index 33121d6..2fc1fb8 100644
--- a/Documentation/kernel-parameters.txt
+++ b/Documentation/kernel-parameters.txt
@@ -685,6 +685,7 @@ and is between 256 and 4096 characters. It is defined in the file
 			See Documentation/isdn/README.HiSax.
 
 	hugepages=	[HW,X86-32,IA-64] Maximal number of HugeTLB pages.
+	hugepagesz=	[HW,IA-64,PPC] The size of the HugeTLB pages.
 
 	i8042.direct	[HW] Put keyboard port into non-translated mode
 	i8042.dumbkbd	[HW] Pretend that controller can only read data from
diff --git a/arch/powerpc/mm/hash_utils_64.c b/arch/powerpc/mm/hash_utils_64.c
index f09730b..afc044c 100644
--- a/arch/powerpc/mm/hash_utils_64.c
+++ b/arch/powerpc/mm/hash_utils_64.c
@@ -368,18 +368,11 @@ static void __init htab_init_page_sizes(void)
 	 * on what is available
 	 */
 	if (mmu_psize_defs[MMU_PAGE_16M].shift)
-		mmu_huge_psize = MMU_PAGE_16M;
+		set_huge_psize(MMU_PAGE_16M);
 	/* With 4k/4level pagetables, we can't (for now) cope with a
 	 * huge page size < PMD_SIZE */
 	else if (mmu_psize_defs[MMU_PAGE_1M].shift)
-		mmu_huge_psize = MMU_PAGE_1M;
-
-	/* Calculate HPAGE_SHIFT and sanity check it */
-	if (mmu_psize_defs[mmu_huge_psize].shift > MIN_HUGEPTE_SHIFT &&
-	    mmu_psize_defs[mmu_huge_psize].shift < SID_SHIFT)
-		HPAGE_SHIFT = mmu_psize_defs[mmu_huge_psize].shift;
-	else
-		HPAGE_SHIFT = 0; /* No huge pages dude ! */
+		set_huge_psize(MMU_PAGE_1M);
 #endif /* CONFIG_HUGETLB_PAGE */
 }
 
diff --git a/arch/powerpc/mm/hugetlbpage.c b/arch/powerpc/mm/hugetlbpage.c
index 71efb38..f4632ad 100644
--- a/arch/powerpc/mm/hugetlbpage.c
+++ b/arch/powerpc/mm/hugetlbpage.c
@@ -24,6 +24,9 @@
 #include <asm/cputable.h>
 #include <asm/spu.h>
 
+#define HPAGE_SHIFT_64K	16
+#define HPAGE_SHIFT_16M	24
+
 #define NUM_LOW_AREAS	(0x100000000UL >> SID_SHIFT)
 #define NUM_HIGH_AREAS	(PGTABLE_RANGE >> HTLB_AREA_SHIFT)
 
@@ -526,6 +529,44 @@ repeat:
 	return err;
 }
 
+void set_huge_psize(int psize)
+{
+	/* Check that it is a page size supported by the hardware and
+	 * that it fits within pagetable limits. */
+	if (mmu_psize_defs[psize].shift && mmu_psize_defs[psize].shift < SID_SHIFT &&
+		(mmu_psize_defs[psize].shift > MIN_HUGEPTE_SHIFT ||
+			mmu_psize_defs[psize].shift == HPAGE_SHIFT_64K)) {
+		HPAGE_SHIFT = mmu_psize_defs[psize].shift;
+		mmu_huge_psize = psize;
+	} else
+		HPAGE_SHIFT = 0;
+}
+
+static int __init hugepage_setup_sz(char *str)
+{
+	unsigned long long size;
+
+	size = memparse(str, &str);
+
+	int shift = __ffs(size);
+	int mmu_psize = -1;
+	switch (shift) {
+	case HPAGE_SHIFT_64K:
+		mmu_psize = MMU_PAGE_64K;
+		break;
+	case HPAGE_SHIFT_16M:
+		mmu_psize = MMU_PAGE_16M;
+		break;
+	}
+	if (mmu_psize >= 0 && mmu_psize_defs[mmu_psize].shift)
+		set_huge_psize(mmu_psize);
+	else
+		printk(KERN_WARNING "Invalid huge page size specified(%i)\n", size);
+
+	return 1;
+}
+__setup("hugepagesz=", hugepage_setup_sz);
+
 static void zero_ctor(struct kmem_cache *cache, void *addr)
 {
 	memset(addr, 0, kmem_cache_size(cache));
diff --git a/include/asm-powerpc/mmu-hash64.h b/include/asm-powerpc/mmu-hash64.h
index 82328de..f35c945 100644
--- a/include/asm-powerpc/mmu-hash64.h
+++ b/include/asm-powerpc/mmu-hash64.h
@@ -277,6 +277,7 @@ extern int hash_huge_page(struct mm_struct *mm, unsigned long access,
 extern int htab_bolt_mapping(unsigned long vstart, unsigned long vend,
 			     unsigned long pstart, unsigned long mode,
 			     int psize, int ssize);
+extern void set_huge_psize(int psize);
 
 extern void htab_initialize(void);
 extern void htab_initialize_secondary(void);
diff --git a/mm/hugetlb.c b/mm/hugetlb.c
index 6121b57..055d232 100644
--- a/mm/hugetlb.c
+++ b/mm/hugetlb.c
@@ -422,6 +422,7 @@ static int __init hugetlb_init(void)
 			break;
 	}
 	max_huge_pages = free_huge_pages = nr_huge_pages = i;
+	printk(KERN_INFO "HugeTLB page size: %ld bytes\n", HPAGE_SIZE);
 	printk("Total HugeTLB memory allocated, %ld\n", free_huge_pages);
 	return 0;
 }

Re: [PATCH 1/2] powerpc: add hugepagesz boot-time parameter

[Arnd Bergmann]

On Wednesday 28 November 2007, Jon Tollefson wrote:
> This patch adds the hugepagesz boot-time parameter for ppc64 that lets=20
> you pick the size for your huge pages. =A0The choices available are 64K=20
> and 16M. =A0It defaults to 16M (previously the only choice) if nothing or=
=20
> an invalid choice is specified. =A0Tested 64K huge pages with the=20
> libhugetlbfs 1.2 release with its 'make func' and 'make stress' test=20
> invocations.

How hard would it be to add the 1MB page size that some CPUs support
as well? On systems with small physical memory like the PS3, that
sounds very useful to me.

	Arnd <><

Re: [PATCH 1/2] powerpc: add hugepagesz boot-time parameter

[Mel Gorman]

On (28/11/07 08:26), Arnd Bergmann didst pronounce:
> On Wednesday 28 November 2007, Jon Tollefson wrote:
> > This patch adds the hugepagesz boot-time parameter for ppc64 that lets 
> > you pick the size for your huge pages.  The choices available are 64K 
> > and 16M.  It defaults to 16M (previously the only choice) if nothing or 
> > an invalid choice is specified.  Tested 64K huge pages with the 
> > libhugetlbfs 1.2 release with its 'make func' and 'make stress' test 
> > invocations.
> 
> How hard would it be to add the 1MB page size that some CPUs support
> as well? On systems with small physical memory like the PS3, that
> sounds very useful to me.
> 

Does the PS3 support 1M pages in hardware? When I last looked, the magic
ibm,segment-page-sizes file that described the supported pagesizes was
missing from the device tree. In this situation, the default sizes
become 4K and 16M because no other ones are advertised.

-- 
Mel Gorman
Part-time Phd Student                          Linux Technology Center
University of Limerick                         IBM Dublin Software Lab

Re: [PATCH 1/2] powerpc: add hugepagesz boot-time parameter

[Arnd Bergmann]

On Wednesday 28 November 2007, Mel Gorman wrote:
> On (28/11/07 08:26), Arnd Bergmann didst pronounce:
> > On Wednesday 28 November 2007, Jon Tollefson wrote:
> > > This patch adds the hugepagesz boot-time parameter for ppc64 that let=
s=20
> > > you pick the size for your huge pages. =A0The choices available are 6=
4K=20
> > > and 16M. =A0It defaults to 16M (previously the only choice) if nothin=
g or=20
> > > an invalid choice is specified. =A0Tested 64K huge pages with the=20
> > > libhugetlbfs 1.2 release with its 'make func' and 'make stress' test=
=20
> > > invocations.
> >=20
> > How hard would it be to add the 1MB page size that some CPUs support
> > as well? On systems with small physical memory like the PS3, that
> > sounds very useful to me.
> >=20
>=20
> Does the PS3 support 1M pages in hardware? When I last looked, the magic
> ibm,segment-page-sizes file that described the supported pagesizes was
> missing from the device tree. In this situation, the default sizes
> become 4K and 16M because no other ones are advertised.

I think you can select the page size using a hypercall on the PS3.
The CPU supports any two of (64k, 1M, 16M) simultaneously.

	Arnd <><

Re: [PATCH 1/2] powerpc: add hugepagesz boot-time parameter

[Geoff Levand]

Arnd Bergmann wrote:
> On Wednesday 28 November 2007, Mel Gorman wrote:
>> On (28/11/07 08:26), Arnd Bergmann didst pronounce:
>> > On Wednesday 28 November 2007, Jon Tollefson wrote:
>> > > This patch adds the hugepagesz boot-time parameter for ppc64 that lets 
>> > > you pick the size for your huge pages.  The choices available are 64K 
>> > > and 16M.  It defaults to 16M (previously the only choice) if nothing or 
>> > > an invalid choice is specified.  Tested 64K huge pages with the 
>> > > libhugetlbfs 1.2 release with its 'make func' and 'make stress' test 
>> > > invocations.
>> > 
>> > How hard would it be to add the 1MB page size that some CPUs support
>> > as well? On systems with small physical memory like the PS3, that
>> > sounds very useful to me.
>> > 
>> 
>> Does the PS3 support 1M pages in hardware? When I last looked, the magic
>> ibm,segment-page-sizes file that described the supported pagesizes was
>> missing from the device tree. In this situation, the default sizes
>> become 4K and 16M because no other ones are advertised.
> 
> I think you can select the page size using a hypercall on the PS3.
> The CPU supports any two of (64k, 1M, 16M) simultaneously.

The PS3's hypervisor allows you to create the lpar's virtual address
space with two page sizes.  I currently have this hard coded to 64K
and 16M.  Within the address space you create memory regions.  I have
the hot-plug memory mapped in as a single region that is hard coded
as 16M pages.

The current HV implementation only allows you to create an htab of
at maximum 1M, so that influences how you can configure page sizes.

Just as a note, since the amount of memory available to the Linux lpar
is not a multiple of 16M, some of the available hot-plug memory is not
mapped into the address space.  For firmware 1.90, 8M is unused.  I
was thinking to create another region for that memory and put things
like the storage bounce buffers in there.  Maybe I'll use 1M pages
for that memory.

-Geoff

Re: [PATCH 1/2] powerpc: add hugepagesz boot-time parameter

[Randy Dunlap]

On Tue, 27 Nov 2007 23:03:10 -0600 Jon Tollefson wrote:

> This patch adds the hugepagesz boot-time parameter for ppc64 that lets 
> you pick the size for your huge pages.  The choices available are 64K 
> and 16M.  It defaults to 16M (previously the only choice) if nothing or 
> an invalid choice is specified.  Tested 64K huge pages with the 
> libhugetlbfs 1.2 release with its 'make func' and 'make stress' test 
> invocations.
> 
> This patch requires the patch posted by Mel Gorman that adds 
> HUGETLB_PAGE_SIZE_VARIABLE; "[PATCH] Fix boot problem with iSeries 
> lacking hugepage support" on 2007-11-15.
> 
> Signed-off-by: Jon Tollefson <kniht@linux.vnet.ibm.com>
> ---
> 
>  Documentation/kernel-parameters.txt |    1 
>  arch/powerpc/mm/hash_utils_64.c     |   11 +--------
>  arch/powerpc/mm/hugetlbpage.c       |   41 ++++++++++++++++++++++++++++++++++++
>  include/asm-powerpc/mmu-hash64.h    |    1 
>  mm/hugetlb.c                        |    1 
>  5 files changed, 46 insertions(+), 9 deletions(-)
> 
> diff --git a/Documentation/kernel-parameters.txt b/Documentation/kernel-parameters.txt
> index 33121d6..2fc1fb8 100644
> --- a/Documentation/kernel-parameters.txt
> +++ b/Documentation/kernel-parameters.txt
> @@ -685,6 +685,7 @@ and is between 256 and 4096 characters. It is defined in the file
>  			See Documentation/isdn/README.HiSax.
>  
>  	hugepages=	[HW,X86-32,IA-64] Maximal number of HugeTLB pages.
> +	hugepagesz=	[HW,IA-64,PPC] The size of the HugeTLB pages.

Any chance of spelling it as "hugepagesize" so that it's a little
less cryptic and more difficult to typo as "hugepages"?
(i.e., less confusion between them)


>  
>  	i8042.direct	[HW] Put keyboard port into non-translated mode
>  	i8042.dumbkbd	[HW] Pretend that controller can only read data from

Thanks.
---
~Randy

Re: [PATCH 1/2] powerpc: add hugepagesz boot-time parameter

[Nish Aravamudan]

On 11/28/07, Randy Dunlap <randy.dunlap@oracle.com> wrote:
> On Tue, 27 Nov 2007 23:03:10 -0600 Jon Tollefson wrote:
>
> > This patch adds the hugepagesz boot-time parameter for ppc64 that lets
> > you pick the size for your huge pages.  The choices available are 64K
> > and 16M.  It defaults to 16M (previously the only choice) if nothing or
> > an invalid choice is specified.  Tested 64K huge pages with the
> > libhugetlbfs 1.2 release with its 'make func' and 'make stress' test
> > invocations.
> >
> > This patch requires the patch posted by Mel Gorman that adds
> > HUGETLB_PAGE_SIZE_VARIABLE; "[PATCH] Fix boot problem with iSeries
> > lacking hugepage support" on 2007-11-15.
> >
> > Signed-off-by: Jon Tollefson <kniht@linux.vnet.ibm.com>
> > ---
> >
> >  Documentation/kernel-parameters.txt |    1
> >  arch/powerpc/mm/hash_utils_64.c     |   11 +--------
> >  arch/powerpc/mm/hugetlbpage.c       |   41 ++++++++++++++++++++++++++++++++++++
> >  include/asm-powerpc/mmu-hash64.h    |    1
> >  mm/hugetlb.c                        |    1
> >  5 files changed, 46 insertions(+), 9 deletions(-)
> >
> > diff --git a/Documentation/kernel-parameters.txt b/Documentation/kernel-parameters.txt
> > index 33121d6..2fc1fb8 100644
> > --- a/Documentation/kernel-parameters.txt
> > +++ b/Documentation/kernel-parameters.txt
> > @@ -685,6 +685,7 @@ and is between 256 and 4096 characters. It is defined in the file
> >                       See Documentation/isdn/README.HiSax.
> >
> >       hugepages=      [HW,X86-32,IA-64] Maximal number of HugeTLB pages.
> > +     hugepagesz=     [HW,IA-64,PPC] The size of the HugeTLB pages.
>
> Any chance of spelling it as "hugepagesize" so that it's a little
> less cryptic and more difficult to typo as "hugepages"?
> (i.e., less confusion between them)

It already exists as hugepagesz= for IA64. Changing it to hugepagesize
would either make ppc be different than IA64, or require keeping both
so as to make IA64 setups continue working as before?

Thanks,
Nish

Re: [PATCH 1/2] powerpc: add hugepagesz boot-time parameter

[Randy Dunlap]

Nish Aravamudan wrote:
> On 11/28/07, Randy Dunlap <randy.dunlap@oracle.com> wrote:
>> On Tue, 27 Nov 2007 23:03:10 -0600 Jon Tollefson wrote:
>>
>>> This patch adds the hugepagesz boot-time parameter for ppc64 that lets
>>> you pick the size for your huge pages.  The choices available are 64K
>>> and 16M.  It defaults to 16M (previously the only choice) if nothing or
>>> an invalid choice is specified.  Tested 64K huge pages with the
>>> libhugetlbfs 1.2 release with its 'make func' and 'make stress' test
>>> invocations.
>>>
>>> This patch requires the patch posted by Mel Gorman that adds
>>> HUGETLB_PAGE_SIZE_VARIABLE; "[PATCH] Fix boot problem with iSeries
>>> lacking hugepage support" on 2007-11-15.
>>>
>>> Signed-off-by: Jon Tollefson <kniht@linux.vnet.ibm.com>
>>> ---
>>>
>>>  Documentation/kernel-parameters.txt |    1
>>>  arch/powerpc/mm/hash_utils_64.c     |   11 +--------
>>>  arch/powerpc/mm/hugetlbpage.c       |   41 ++++++++++++++++++++++++++++++++++++
>>>  include/asm-powerpc/mmu-hash64.h    |    1
>>>  mm/hugetlb.c                        |    1
>>>  5 files changed, 46 insertions(+), 9 deletions(-)
>>>
>>> diff --git a/Documentation/kernel-parameters.txt b/Documentation/kernel-parameters.txt
>>> index 33121d6..2fc1fb8 100644
>>> --- a/Documentation/kernel-parameters.txt
>>> +++ b/Documentation/kernel-parameters.txt
>>> @@ -685,6 +685,7 @@ and is between 256 and 4096 characters. It is defined in the file
>>>                       See Documentation/isdn/README.HiSax.
>>>
>>>       hugepages=      [HW,X86-32,IA-64] Maximal number of HugeTLB pages.
>>> +     hugepagesz=     [HW,IA-64,PPC] The size of the HugeTLB pages.
>> Any chance of spelling it as "hugepagesize" so that it's a little
>> less cryptic and more difficult to typo as "hugepages"?
>> (i.e., less confusion between them)
> 
> It already exists as hugepagesz= for IA64. Changing it to hugepagesize
> would either make ppc be different than IA64, or require keeping both
> so as to make IA64 setups continue working as before?

Oh, but it wasn't in Doc/kernel-parameters.txt ?  :(

OK, just leave it as is, I think.

Thanks,
-- 
~Randy

Re: [PATCH 1/2] powerpc: add hugepagesz boot-time parameter

[Nish Aravamudan]

On 11/28/07, Randy Dunlap <randy.dunlap@oracle.com> wrote:
> Nish Aravamudan wrote:
> > On 11/28/07, Randy Dunlap <randy.dunlap@oracle.com> wrote:
> >> On Tue, 27 Nov 2007 23:03:10 -0600 Jon Tollefson wrote:
> >>
> >>> This patch adds the hugepagesz boot-time parameter for ppc64 that lets
> >>> you pick the size for your huge pages.  The choices available are 64K
> >>> and 16M.  It defaults to 16M (previously the only choice) if nothing or
> >>> an invalid choice is specified.  Tested 64K huge pages with the
> >>> libhugetlbfs 1.2 release with its 'make func' and 'make stress' test
> >>> invocations.
> >>>
> >>> This patch requires the patch posted by Mel Gorman that adds
> >>> HUGETLB_PAGE_SIZE_VARIABLE; "[PATCH] Fix boot problem with iSeries
> >>> lacking hugepage support" on 2007-11-15.
> >>>
> >>> Signed-off-by: Jon Tollefson <kniht@linux.vnet.ibm.com>
> >>> ---
> >>>
> >>>  Documentation/kernel-parameters.txt |    1
> >>>  arch/powerpc/mm/hash_utils_64.c     |   11 +--------
> >>>  arch/powerpc/mm/hugetlbpage.c       |   41 ++++++++++++++++++++++++++++++++++++
> >>>  include/asm-powerpc/mmu-hash64.h    |    1
> >>>  mm/hugetlb.c                        |    1
> >>>  5 files changed, 46 insertions(+), 9 deletions(-)
> >>>
> >>> diff --git a/Documentation/kernel-parameters.txt b/Documentation/kernel-parameters.txt
> >>> index 33121d6..2fc1fb8 100644
> >>> --- a/Documentation/kernel-parameters.txt
> >>> +++ b/Documentation/kernel-parameters.txt
> >>> @@ -685,6 +685,7 @@ and is between 256 and 4096 characters. It is defined in the file
> >>>                       See Documentation/isdn/README.HiSax.
> >>>
> >>>       hugepages=      [HW,X86-32,IA-64] Maximal number of HugeTLB pages.
> >>> +     hugepagesz=     [HW,IA-64,PPC] The size of the HugeTLB pages.
> >> Any chance of spelling it as "hugepagesize" so that it's a little
> >> less cryptic and more difficult to typo as "hugepages"?
> >> (i.e., less confusion between them)
> >
> > It already exists as hugepagesz= for IA64. Changing it to hugepagesize
> > would either make ppc be different than IA64, or require keeping both
> > so as to make IA64 setups continue working as before?
>
> Oh, but it wasn't in Doc/kernel-parameters.txt ?  :(

Nope :( I wonder how many other kernel parameters are in the same
boat? Where's an RPJDay-script when you need it?

> OK, just leave it as is, I think.

Yeah, I guess that's probably easiest...unfortunately.

-Nish