Re: [PATCH] PAGE_SIZE IO for RAW (RAW VARY)

[Andrea Arcangeli <andrea@suse.de>]

On Wed, Jan 09, 2002 at 03:48:17PM -0800, Badari Pulavarty wrote:
> Alan,
> 
> > 
> > > If it is not reasonable to fix all the brokern drivers,
> > > how about making this configurable (to do variable size IO) ?
> > > Do you favour this solution ?
> > 
> > We have hardware that requires aligned power of two for writes (ie 4K on
> > 4K boundaries only). The 3ware is one example Jeff Merkey found
> > 
> 
> emm.. come to think of it, I can easily (2 line) change my patch to
> do 512 byte buffer heads till we get PAGE alignment and then start
> issuing 4K IO buffer heads. What do you think ? will this work ? 

you should make sure the buffer is naturally aligned with the b_size (so
you emulate the change of blocksize in a filesystem), if it wasn't the
case you should fix it.

> 
> And also, do you know any low level drivers Ben mentioning:
> 
> > low level drivers, some of which assume that 
> > all buffer heads within a request have the same block size.
> 
> Is it still true for 2.4 ? 

yes, you simply need to call submit_bh or to use separate ll_rw_blocks
when you change b_size.

> 
> Regards,
> Badari


Andrea