David Mosberger-Tang wrote:
> But that's excactly the point. The hugepage interface returns a
> different kind of virtual memory. There are tons of programs out
> there using mmap(). If such a program gets fed a path to the
> hugepagefs, it might end up with huge pages without knowing anything
> about huge pages. For the most part, that might work fine, but it
> could lead to subtle failures.
Yeah, that was one of Linus's points about the syscalls, in a private
email. I mentioned how the new syscalls were in poor taste, when
existing syscalls would work fine, and he flamed me right back ;-)
One of his main points to me was exactly what you are elucidating:
there are subtle differences between normal pages and superpages that
are exposed to userland, and we should make that explicit [with the
syscalls] rather than hide it [with hugetlbfs/mmap/etc.]. So I think
this is further indication Linus has a very valid point ;-)
However, that said, I think hugetlbfs will almost always get used in
preference to the syscalls, so leaving them in may be more a statement
of technical correctness/cleanliness than anything else.
[tangent warning]
This whole hugetlb affair is unfortunately pretty ugly, and this thread
is just one component of that. All these discussions occurred off-list,
and it's _still_ a political football. Sigh. I just hope that the
furor dies down soon, that smart technical [apolitical] decisions are
made, and future discussions are at least CC'd to lkml.
Jeff
-
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/
This archive was generated by hypermail 2b29 : Fri Nov 15 2002 - 22:00:34 EST