[Drbd-dev] [PATCH v5 08/12] block: Introduce new bio_split()

Kent Overstreet koverstreet at google.com
Thu Aug 9 09:33:34 CEST 2012


On Thu, Aug 09, 2012 at 12:22:17AM -0700, Tejun Heo wrote:
> On Wed, Aug 08, 2012 at 06:39:23PM -0700, Kent Overstreet wrote:
> > On Wed, Aug 08, 2012 at 04:05:32PM -0700, Tejun Heo wrote:
> > > One more thing.
> > > 
> > > On Mon, Aug 06, 2012 at 03:08:37PM -0700, Kent Overstreet wrote:
> > > > +	if (bio_integrity(bio)) {
> > > > +		bio_integrity_clone(ret, bio, gfp, bs);
> > > > +		bio_integrity_trim(ret, 0, bio_sectors(ret));
> > > > +		bio_integrity_trim(bio, bio_sectors(ret), bio_sectors(bio));
> > > 
> > > Is this equivalent to bio_integrity_split() performance-wise?
> > 
> > Strictly speaking, no. But it has the advantage of being drastically
> > simpler - and the only one only worked for single page bios so I
> > would've had to come up with something new from scratch, and as
> > confusing as the integrity stuff is I wouldn't trust the result.
> 
> There's already bio_integrity_split() and you're actively dropping it.

Because it only works for single page bios, AFAICT. I'd have to start
from scratch.

> > I'm skeptical that it's going to matter in practice given how much
> > iteration is done elsewhere in the course of processing a bio and given
> > that this stuff isn't used with high end SSDs...
> 
> If you think the active dropping is justified, please let the change
> and justification clearly stated.  You're burying the active change in
> two separate patches without even mentioning it or cc'ing people who
> care about bio-integrity (Martin K. Petersen). 

Not intentionally, he isn't in MAINTAINERS so get_maintainers.pl missed
it and it slipped by while I was looking for people to CC. Added him.


More information about the drbd-dev mailing list