lkml.org 
[lkml]   [2006]   [Jun]   [27]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
From
SubjectRe: [Suspend2][ 0/9] Extents support.
Date
Hi.

On Tuesday 27 June 2006 17:05, Jens Axboe wrote:
> On Tue, Jun 27 2006, Nigel Cunningham wrote:
> > On Tuesday 27 June 2006 15:36, Jens Axboe wrote:
> > > On Tue, Jun 27 2006, Nigel Cunningham wrote:
> > > > On Tuesday 27 June 2006 07:20, Rafael J. Wysocki wrote:
> > > > > On Monday 26 June 2006 18:54, Nigel Cunningham wrote:
> > > > > > Add Suspend2 extent support. Extents are used for storing the
> > > > > > lists of blocks to which the image will be written, and are
> > > > > > stored in the image header for use at resume time.
> > > > >
> > > > > Could you please put all of the changes in kernel/power/extents.c
> > > > > into one patch?  It's quite difficult to review them now, at least
> > > > > for me.
> > > >
> > > > I spent a long time splitting them up because I was asked in previous
> > > > iterations to break them into manageable chunks. How about if I were
> > > > to email you the individual files off line, so as to not send the
> > > > same amount again?
> > >
> > > Managable chunks means logical changes go together, one function per
> > > diff is really extreme and unreviewable. Support for extents is one
> > > logical change, so it's one patch. Unless of course you have to do some
> > > preparatory patches, then you'd do those separately.
> > >
> > > I must admit I thought you were kidding when I read through this
> > > extents patch series, having a single patch just adding includes!
> >
> > Sorry for fluffing it up. I'm pretty inexperienced, but I'm trying to
> > follow CodingStyle and all the other advice. If I'd known I'd
> > misunderstood what was wanted, I probably could have submitted this
> > months ago. Oh well. Live and learn. What would you have me do at this
> > point?
>
> Split up your patches differently, and not in so many steps. Ideally
> each step should work and compile, with each introducing some sort of
> functionality. Each patch should be reviewable on its own.

The difficulty I have there is that suspending to disk doesn't seem to me to
be something where you can add a bit at a time like that. I do have proc
entries that allow you to say "Just freeze the processes and prepare the
metadata, then free it and exit" (freezer_test) and "Do everything but
actually writing the image and doing the atomic copy, then exit
(test_filter_speed), for diagnosing problems and tuning the configuration,
but if I start were to start again with nothing, I'd only write the dynamic
pageflags code to start with and submit it (giving you lib/dyn_pageflags.c
and kernel/power/pageflags.c), then the refrigerator changes and the extent
code and so on. I guess what I'm trying to say is that I'm not mutating
swsusp into suspend2 here, and I don't think I can. Suspend2 is a
reimplementation of swsusp, not a series of incremental modifications. It
uses completely different methods for writing the image, storing the metadata
and so on. Until recently, the only thing it shared with swsusp was the
refrigerator and driver model calls, and even now the sharing of lowlevel
code is only a tiny fraction of all that is done.

Could I ask what might be a dumb question in this regard - why isn't Reiser4
going through the same process? Is it an indication that I shouldn't have
submitted these patches and should have just asked Andrew to take Suspend2
into mm, or is there something different between Reiser4 and Suspend2 that
I'm missing?

Regards,

Nigel

--
See http://www.suspend2.net for Howtos, FAQs, mailing
lists, wiki and bugzilla info.
[unhandled content-type:application/pgp-signature]
\
 
 \ /
  Last update: 2006-06-27 09:41    [from the cache]
©2003-2014 Jasper Spaans. Advertise on this site