>but 4M real-memory mappings raise other serious issues like how do we
>allocate/deallocate (swap?) those monsters.
Swappable memory is exactly what big pages aren't good for, so it seems
best not to use them for that.
There are several ways to deal with this though:
1. Don't swap the big pages. Bad if there are many of them.
2. swap the entire 4M page. Not very efficient.
3. swap parts of the 4M page, and mark the entire page as gone.
Requires extra bookkeping, 4k pages seems the preferred way. 4. Convert
the 4M page to 1024 4k pages if we ever need to swap it out.
The conversion take time of course, but it should be small compared
to disk io.
This would offer the best of both worlds - TLB-efficient 4M pages
for those that need them as long as there is memory enough, and
efficient swapping when necessary.
Helge Hafting
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/