On Sat, 2 Sep 2000, Alexander Viro wrote:
> +
> + /*
> + * So truncate in the middle of a hole not on a block boundary will
> + * allocate a block. BFD. Everything is still consistent, so trying
> + * to be smart is not worth the trouble.
> + */
You don't actually have to be smart.
There's a really simple way to avoid this: compare the thing you're going
to zero out against zero before you memset() it to zero. If it was already
zero, you just unlock the page and release.
Downside: you have to read the page in, in order to compare. However, this
turns out to be a non-issue: if there were no blocks on that page (like a
forward truncate) this ends up generating no IO anyway. If there _were_
blocks on the page, it's likely that one of them was the one we're going
to partially clear anyway, so the "prepare_write()" would have had to read
it in anyway.
Upside: you don't need to mark it dirty unnecessarily, assuming you just
extended the file. No extra allocation, and no extra write to disk.
Comments? Basically the "grab_cache_page()" would be a "read_cache_page()"
instead with all the wait-on-page etc stuff.
Linus
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Thu Sep 07 2000 - 21:00:15 EST