Re: /pub/linux/kernel/people/hedrick/ide-2.5.32

[Andrew Morton <akpm@zip.com.au>]

I'm a wee bit reluctant to shrink these settings globally,
because it will have a small impact on people's ongoing
evaluation of 2.5 on high-end SCSI hardware.

Could you please set BIO_MAX_SECTORS to 64 within the IDE
patch, and add a BIG FAT COMMENT?


Andre Hedrick wrote:
> 
> Andrew,
> 
> I am just now getting back to crawling speeds in the 2.5 tree.
> I can only comment on what works and what Viro tells me.
> But if you think it needs to be globally set, until the updates from Jens
> arrive, please send to Linus.  I am running my stuff by Viro and company.
> 
> Cheers,
> 
> On Thu, 29 Aug 2002, Andrew Morton wrote:
> 
> > Andre Hedrick wrote:
> > >
> > > ...
> > > There is one more thing to fix.
> > >
> > > ./fs/mpage.c
> > >
> > > /*
> > >  * The largest-sized BIO which this code will assemble, in bytes.  Set this
> > >  * to PAGE_CACHE_SIZE if your drivers are broken.
> > >  */
> > > #define MPAGE_BIO_MAX_SIZE 32768        //BIO_MAX_SIZE
> > >
> > > This is confirmed with Al Viro and was required to make things sane!
> >
> > You'll need to do the same thing to fs/direct-io.c:DIO_BIO_MAX_SIZE
> > in that case.
> >
> > I'd suggest that you just go in and change BIO_MAX_SECTORS
> > to 64.   Or 32 if you happen to be using a qlogic controller :(
> >
> > So everything's broken in there - a hardwired constant doesn't
> > cut it.   Jens is cooking up an `add_page_to_bio()' API which
> > will do the right thing based upon q->max_sectors.  But that
> > is not yet available.
> >
> 
> Andre Hedrick
> LAD Storage Consulting Group