Nishanth,
Thanks, i will probably migrate to 2.6.27 then.
Another doubt i had was, how libhugetlbfs allocates hugepages for
application's
data/text segment without any kernel component in "fork/execv/elfload" path ?
Could
you help me in understanding the flow ?
I checked the file "hugeedit" file, it modifies the program header of
text/data
segment and sets "PF_LINUX_HUGETLB" bit in pflags. But who reads this finally ?
Does the kernel architecture port of hugetlbfs has to read this field and
allocate a page
accordingly ?
Thanks,
Mehul.
There are only 10 kind of people in the world ... those who knows binary and
those who do not......
--- On Mon, 12/29/08, Nishanth Aravamudan <n...@us.ibm.com> wrote:
From: Nishanth Aravamudan <n...@us.ibm.com>
Subject: Re: [Libhugetlbfs-devel] Lib hugetlbfs
To: "mehul vora" <mehu...@yahoo.com>
Cc: libhugetlbfs-devel@lists.sourceforge.net, a...@us.ibm.com
Date: Monday, December 29, 2008, 10:34 AM
On 25.12.2008 [06:01:26 -0800], mehul vora wrote:
> Hi Guys,
>
> Marry X-MAS to you ALL.
>
> I am re-starting the long paused libhugetlb porting for MIPS.
>
> One quick question - Is there any additional driver/kernel
> component to support libhugetlbfs that i may have to port for MIPS ?
>
> I have already ported the basic hugetlb (except MAP_PRIVATE) for
> MIPS and i can allocate MAP_SHARED hugepages thru mmap/shmem. I am
> using linux 2.6.21 and hugetlbfs 2.1 release.
Well, I would use current mainline if you are going to use such a recent
libhugetlbfs. If you run the `make func` portion of the testsuite
against your patched kernel (2.6.28) based, you might see some FAILs.
Those are functionality gaps between your MIPS port and the expected
mainline features. Some may be impossible to fix (not sure of the
limitations of hugepages on MIPS) and will need exclusions in the
testsuite to skip in MIPS, but only if they really can't be supported,
not if they simply are unsupported right now. The `make stress` tests
will help verify that there are no locking/race issues, generally, as
they stress the various hugepage allocation/freeing paths throughout the
kernel.
Good luck,
Nish
--
Nishanth Aravamudan <n...@us.ibm.com>
IBM Linux Technology Center
------------------------------------------------------------------------------
_______________________________________________
Libhugetlbfs-devel mailing list
Libhugetlbfs-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/libhugetlbfs-devel
