Re: partial reiser4 review comments

[Christoph Hellwig]

On Sun, Aug 06, 2006 at 06:38:34PM +0400, Alexander Zarochentsev wrote:
> > > - reiser4_readpages() shouldn't need to clean up the remaining
> > > pages on *pages.  read_cache_pages() does that now.
> >
> > Without looking at the code I remember someone from the Namesys
> > people told me they could use plain mpage_readpages now.  Anything
> > still blocking using that function?
> 
> reiser4 tries to reduce number of tree lookups. better, if there would 
> be one tree lookup for one readpages call.

Right now mpage_redpages does one get_block per extent.  I think it's
pretty messy do do one block allocator call that can return multiple
extents because that leads into a lot of complexity for a rather
questionable gain.(XFS on IRIX does that)

> 
> what we are currently doing in a not-yet-submitted patch (below), i 
> don't see how it can be done by mpage_readpages.
> 
> +struct uf_readpages_context {
> +	lock_handle lh;
> +	coord_t coord;
> +};

I must admit that standalone code snipplet doesn't really tell me a lot.
Do you mean the possibility to pass around a filesystem-defined structure
to multiple allocator calls?  I'm pretty sure can add that, I though it
would be useful multiple times in the past but always found ways around
it.

> BTW, read_cache_page() and mpage_readpages are similar, I guess the 
> second can be rewritten using the first one, no?

Do you mean read_cache_page() or read_cache_pages() ?

(Yeah, it really bad that we have two functions sounding the same but doing
thing quite differently..)

read_cache_pages() could probably be folded into mpages_readpages by allowing
it to give an additional readpage callback similar to how mpage_writepages
can either do real direct to bio or be used as an interator over writepages
calls.

read_cache_page() is quite different from read_cache_pages() and
mpages_readpages in that it is synchronous and waits for the read to complete
before returning.
-
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/