Mailing List Archive: [RFC/PATCH] Shared Page Tables [0/2]

[RFC/PATCH] Shared Page Tables [0/2]

Apr 10, 2006, 9:13 AM

Post #1 of 5 (223 views)

Here's a new cut of the shared page table patch. I divided it into
two patches. The first one just fleshes out the
pxd_page/pxd_page_kernel macros across the architectures. The
second one is the main patch.

This version of the patch should address the concerns Hugh raised.
Hugh, I'd appreciate your feedback again. Did I get everything?

These patches apply against 2.6.17-rc1.

Dave McCracken

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

Re: [RFC/PATCH] Shared Page Tables [0/2] [ In reply to ]

clameter at sgi

Apr 10, 2006, 10:22 AM

Post #2 of 5 (219 views)

Permalink

On Mon, 10 Apr 2006, Dave McCracken wrote:

> Here's a new cut of the shared page table patch. I divided it into
> two patches. The first one just fleshes out the
> pxd_page/pxd_page_kernel macros across the architectures. The
> second one is the main patch.
>
> This version of the patch should address the concerns Hugh raised.
> Hugh, I'd appreciate your feedback again. Did I get everything?
>
> These patches apply against 2.6.17-rc1.

Could you break out the locking changes to huge pages?
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

Re: [RFC/PATCH] Shared Page Tables [0/2] [ In reply to ]

dmccr at us

Apr 10, 2006, 1:11 PM

Post #3 of 5 (222 views)

Permalink

--On Monday, April 10, 2006 10:22:34 -0700 Christoph Lameter
<clameter@sgi.com> wrote:

>> Here's a new cut of the shared page table patch. I divided it into
>> two patches. The first one just fleshes out the
>> pxd_page/pxd_page_kernel macros across the architectures. The
>> second one is the main patch.
>> (...)
>
> Could you break out the locking changes to huge pages?

The lock changes to hugetlb are only to support sharing of pmd pages when
they contain hugetlb pages. They just substitute the struct page lock for
the page_table_lock, and are only about 30 lines of code. Is this really
worth separating out?

Dave McCracken

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

Re: [RFC/PATCH] Shared Page Tables [0/2] [ In reply to ]

clameter at sgi

Apr 10, 2006, 1:20 PM

Post #4 of 5 (219 views)

Permalink

On Mon, 10 Apr 2006, Dave McCracken wrote:

> The lock changes to hugetlb are only to support sharing of pmd pages when
> they contain hugetlb pages. They just substitute the struct page lock for
> the page_table_lock, and are only about 30 lines of code. Is this really
> worth separating out?

Ia64 does not use pmd pages for huge pages. It relies instead on a
separate region. I wonder if this works on IA64.

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

Re: [RFC/PATCH] Shared Page Tables [0/2] [ In reply to ]

dmccr at us

Apr 10, 2006, 1:27 PM

Post #5 of 5 (223 views)

Permalink

--On Monday, April 10, 2006 13:20:59 -0700 Christoph Lameter
<clameter@sgi.com> wrote:

>> The lock changes to hugetlb are only to support sharing of pmd pages when
>> they contain hugetlb pages. They just substitute the struct page lock
>> for the page_table_lock, and are only about 30 lines of code. Is this
>> really worth separating out?
>
> Ia64 does not use pmd pages for huge pages. It relies instead on a
> separate region. I wonder if this works on IA64.

Sharing of hugetlb page tables is enabled on a per-architecture basis, so
if ia64 doesn't use pmd pages we shouldn't try to enable it. If it's not
enabled all the locking in hugetlb resolves to using page_table_lock, so
the original semantics will be preserved.

Dave McCracken

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/