>
> Ok, this is _exactly_ why I didn't check the mtime in the first place,
> and just to explain WHY this patch is bad bad bad, I'll walk you through
> it.
>
Linus, you missed something. Please, look at my patch again.
Your comments are not about it!
The only correct note is about race condition between getattr and write,
but it is the cause that I wrote the comment on NFSv3!
(Well, maybe this comment is pretty obscure.)
1. It does not stop old invalidation scheme.
I never touch read_cache_mtime/jiffies, so that
if mtime was changed, not depending who changed it,
cache is flushed.
2. It does not trash cache. Really, it is the main idea
of the patch and you missed it.
3. It is correct, I bet 8)8)
Alexey Kuznetsov.