[dm-devel] [PATCH 0/22] Immutable biovecs, block layer changes

Kent Overstreet kmo at daterainc.com
Tue Sep 24 19:19:54 UTC 2013


On Tue, Sep 24, 2013 at 09:20:14AM -0400, Mike Snitzer wrote:
> On Tue, Sep 24 2013 at  7:00am -0400,
> Christoph Hellwig <hch at infradead.org> wrote:
> 
> > Just curious, what's the state of the remaining immutable bio work?
> 
> Hey Christoph,
> 
> Have you been over the patchset?  Looks sane to you?
> 
> Given how disruptive this patchset is to the block layer I'm wondering
> how painful this change will be in combination with Jens' blk-mq
> changes.  I'd prefer to see blk-mq before immutable biovecs; but I have
> my own selfish reasons for that.
> 
> I'm also concerned about DM regressions given that a lot of DM code to
> handle splitting a bio that spans target boundaries is ripped out, see:
> https://patchwork.kernel.org/patch/2840657/

Oh good, you had seen that :)

Are you worried about performance regressions or actual bugs?

I _have_ explicitly tested that code path, but I don't know what corner
cases I may have missed so it definitely needs your eyeballs and testing
too.

As for performance - this'll be more efficient than what the dm code was
doing before. Especially when the multipage bvec patches go in.

> But that is just handwaving/FUD at this point cause I haven't put the
> time to looking at this patchset close enough to feel comfortable.  I
> can make reviewing this patchset a priority this week though.
> 
> Kent, have you rebased this patchset at all?  Do you have a git tree I
> can clone to save the pain of pulling these patches out my my mbox,
> etc.

Yeah: git://evilpiepirate.org/~kent/linux-bcache.git for-jens




More information about the dm-devel mailing list