On 20.03.2007 [15:45:20 -0700], Nishanth Aravamudan wrote:
> The semantics of hugetlb_slice_{start,end} are ... difficult to say the
> least. IA64's virtual address limitation makes it difficult to say what
> the "correct" behavior is for valid addresses. It is virtually (no pun
> intended) impossible to indicate a passed in address is not valid for
> hugepage mappings, period, without breaking some other semantics.
>
> The two users of this functionality, currently, morecore and
> brk_near_huge, don't even want to know this level of detail. They want
> to simply know the next valid hint address for a hugepage mmap(), given
> a particular address. This is distinct from where the next slice starts,
> as well, so that's not a good fit. Instead, give them exactly what they
> want. IA64 is almost the same as all the other archs, then, in that
> addresses in the hugepage region are aligned up to the hugepage size.
> But if the address is below the hugepage region, we simply return the
> start of the region.
Bah this isn't quite right...
Both morecore and brk_near_huge know they are in a smallpage slice (the
current sbrk()), so they want the next slice that is usable for
hugepages. Looks like we do want a hugetlb_next_slice_start() function.
I'll respin this tomorrow.
Thanks,
Nish
--
Nishanth Aravamudan <[EMAIL PROTECTED]>
IBM Linux Technology Center
-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your
opinions on IT & business topics through brief surveys-and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV
_______________________________________________
Libhugetlbfs-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/libhugetlbfs-devel
