Re: [PATCH] anobjrmap 1/6 objrmap

[William Lee Irwin III]

On Thu, Mar 18, 2004 at 11:21:07PM +0000, Hugh Dickins wrote:
> First of six patches implementing full object-based rmap over 2.6.5-rc1,
> reviving my anonmm method to compare against Andrea's anon_vma method.
> I've not yet implemented Linus' early-COW solution to the mremap move
> issue, that will follow; handling of non-linear obj vmas also to follow.
> Sorry, not yet checked against wli's tree, he may have some fixes to it.

It would actually take serious rereading to verify that what issues I'd
fixed weren't ones I introduced myself. In that set of patches,
anobjrmap appeared alongside a page allocator rewrite, a top-down vma
allocation policy for i386, an arch/i386/mm/pgtable.c rewrite, wrapping
every modification to userspace ptes to track statistics wanted by
/proc/, highpmd, something that RCU'd inode->i_mmap{,_shared} missing
the needed smp_read_barrier_depends() calls, and using wrappers around
rwlocks to allow mapping->page_lock to be configured as an rwlock or
spinlock at compile-time thrown in for good measure, so there isn't
much of a way to rule out my own hacks. There was even experimental
junk at some point e.g. to remove files_lock in addition to a fair
number of other quetionable/buggy patches I dumped instead of debugging.

The story of that tree is too tortuous and sad to tell. I'll put up a
new tree with a substantially different emphasis, comprised of
completely different patches, when I have enough material to warrant it.


-- wli
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/