Re: [patch] remove hugetlb syscalls

David Mosberger-Tang (davidm@mostang.com)
Thu, 14 Nov 2002 13:46:26 -0800


>>>>> On Thu, 14 Nov 2002 13:38:22 -0800, William Lee Irwin III <wli@holomorphy.com> said:

Bill> On Thu, 14 Nov 2002 12:34:11 -0800, William Lee Irwin III
Bill> said:
William> (3) ->f_op->mmap() will hand -EINVAL back to userspace
William> instead of automatically placing the vma, for explicit and
William> 0 start adresses

Bill> On Thu, Nov 14, 2002 at 01:31:21PM -0800, David Mosberger-Tang
Bill> wrote:
>> This sounds like a receipe for creating unportable programs
>> because the alignment constraints will be different from one
>> platform to another. (Adding gethugepagesize() in libc would
>> alleviate the problem but it wouldn't solve it completely.)
>> Overall, it does sound to me that the hugetlbfs is so specialized
>> that it would be much cleaner to provide a separate interface at
>> the user-level. That would leave more flexibility should the
>> implementation change over time (which it well might). --david

Bill> Okay, that's a serious problem. But it's easy to fix; I'll
Bill> just call the hugepage vma placement functions that are also
Bill> used by the syscalls.

Sounds good to me.

Thanks,

--david
-
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/