Enable UFS2, logging by default in sysinst
Andrew Doran
Any comments? Please Cc: me, I am not subscribed.
Thanks,
Andrew
--
Posted automagically by a mail2news gateway at muc.de e.V.
Please direct questions, flames, donations, etc. to news-...@muc.de
Bernd Ernesti
> http://www.netbsd.org/~ad/sysinst.diff
>
> Any comments? Please Cc: me, I am not subscribed.
I'm not commenting on the patch itself, but to the functional changes in it.
IMHO only UFS2 should be enabled, keeping logging off by default for
the moment.
Wabpl is too new to have it enable by default.
Bernd
Andrew Doran
> On Sun, Feb 22, 2009 at 11:58:08AM +0000, Andrew Doran wrote:
> > http://www.netbsd.org/~ad/sysinst.diff
> >
> > Any comments? Please Cc: me, I am not subscribed.
>
> I'm not commenting on the patch itself, but to the functional changes in it.
>
> IMHO only UFS2 should be enabled, keeping logging off by default for
> the moment.
> Wabpl is too new to have it enable by default.
I understand, but at what point would you declare it stable? Bear in mind
that the change applies to -current, not 5.0.
We already know that wapbl is more reliable than softdep, and if we wait as
long as we did for UFS2 to say "yay verily this is good", we'd be shooting
ourselves in the foot by not exploiting a feature that's both useful to
users and useful to the project as a differentiator.
Hauke Fath
>IMHO only UFS2 should be enabled, keeping logging off by default for
>the moment.
>Wabpl is too new to have it enable by default.
Since this is about -current, we might as well have as many people as
possible bang on the logging code.
hauke
--
"It's never straight up and down" (DEVO)
Alan Barrett
> We already know that wapbl is more reliable than softdep, and if we wait as
> long as we did for UFS2 to say "yay verily this is good", we'd be shooting
> ourselves in the foot by not exploiting a feature that's both useful to
> users and useful to the project as a differentiator.
I suggest putting it in a sysinst menu:
For netbsd-5: softdep/wapbl/neither, default neither.
For current: wapbl yes/no, default yes.
--apb (Alan Barrett)
Bernd Ernesti
> On Sun, Feb 22, 2009 at 01:18:48PM +0100, Bernd Ernesti wrote:
>
> > >
> > > Any comments? Please Cc: me, I am not subscribed.
> >
> >
> > IMHO only UFS2 should be enabled, keeping logging off by default for
> > the moment.
> > Wabpl is too new to have it enable by default.
>
> that the change applies to -current, not 5.0.
I'm not against to add it, but it was my understanding that you and Joerg (?)
are working on getting it in a more stable state.
I saw your mail, about the ffs fixes, which looks promising, but I don't know
about what Joerg is doing.
What about enabling it one or two weeks after your patches made it into -current
and a few more people have a chance to test -current by then?
> We already know that wapbl is more reliable than softdep, and if we wait as
> long as we did for UFS2 to say "yay verily this is good", we'd be shooting
> ourselves in the foot by not exploiting a feature that's both useful to
> users and useful to the project as a differentiator.
Hmmm, I had other experience with wapbl in 5.99.7, that forced me to disable
it (pr 40472).
Bernd
Joerg Sonnenberger
> On Sun, Feb 22, 2009 at 03:32:05PM +0000, Andrew Doran wrote:
> > On Sun, Feb 22, 2009 at 01:18:48PM +0100, Bernd Ernesti wrote:
> >
> > > On Sun, Feb 22, 2009 at 11:58:08AM +0000, Andrew Doran wrote:
> > > > http://www.netbsd.org/~ad/sysinst.diff
> > > >
> > > > Any comments? Please Cc: me, I am not subscribed.
> > >
> > > I'm not commenting on the patch itself, but to the functional changes in it.
> > >
> > > IMHO only UFS2 should be enabled, keeping logging off by default for
> > > the moment.
> > > Wabpl is too new to have it enable by default.
> >
> > I understand, but at what point would you declare it stable? Bear in mind
> > that the change applies to -current, not 5.0.
>
> I'm not against to add it, but it was my understanding that you and Joerg (?)
> are working on getting it in a more stable state.
> I saw your mail, about the ffs fixes, which looks promising, but I don't know
> about what Joerg is doing.
I have tested the public patches from ad@ (slightly older), pooka@
(applied) and myself. Neither of that seems to fix all the performance
issues and all the dead lock conditions, but the all of them improve the
situation and at least the combination of pooka's changes with ad's
recent patch seem to make the timing range small enough that it can't be
hit during the parallel bulk builds.
Joerg
Bernd Ernesti
[..]
> I have tested the public patches from ad@ (slightly older), pooka@
> (applied) and myself. Neither of that seems to fix all the performance
> issues and all the dead lock conditions, but the all of them improve the
> situation and at least the combination of pooka's changes with ad's
> recent patch seem to make the timing range small enough that it can't be
> hit during the parallel bulk builds.
Sounds good enough to enable it by default.
Bernd
Perry E. Metzger
Catching up with old email...
Andrew Doran <a...@netbsd.org> writes:
> http://www.netbsd.org/~ad/sysinst.diff
>
> Any comments? Please Cc: me, I am not subscribed.
I think this should go in. It is for -current, and if WAPBL isn't
stable by the 6.0 release NetBSD is in bigger trouble.
Perry
--
Perry E. Metzger pe...@piermont.com
Martin Husemann
> I think this should go in. It is for -current, and if WAPBL isn't
> stable by the 6.0 release NetBSD is in bigger trouble.
I don't mind the LOG part, but booting from FFSv2 is not supported on all
archs.
Martin
Perry E. Metzger
Martin Husemann <mar...@duskware.de> writes:
> On Mon, Mar 23, 2009 at 03:44:25PM -0400, Perry E. Metzger wrote:
>> stable by the 6.0 release NetBSD is in bigger trouble.
>
> archs.
One could just turn it on for the platforms where it worked. Of
course, v2 boot is probably important on all arches if only because
new disks are getting past ffsv1 size.
Andrew Doran
> On Mon, Mar 23, 2009 at 03:44:25PM -0400, Perry E. Metzger wrote:
> > stable by the 6.0 release NetBSD is in bigger trouble.
>
> archs.
That's poor because ufs2 was integrated six years ago. Hopefully the magic
smoke will have come out the back of these machines by 2038, or NetBSD
*really* will be in big trouble.
Is there a list somewhere that you know of?
Martin Husemann
> That's poor because ufs2 was integrated six years ago.
Sure, but only in the last year people started claiming that it has
any real world advantages.
On some machines it is pretty hard to arrange, since firmware knows
about FFSv1 and loads the kernel directly, but does not provide good
enough (or working) callbacks for a standalone bootloader.
Of course we can always go with a small boot partition on those archs.
> Is there a list somewhere that you know of?
I know of sparc64 (bootblock is in forth so can't call libsa ffsv2
support, but otherwise just a SMOP) and shark (firmware like above).
Everything with bootblocks in C and no too hard size constraints should
be easy to fix if there are others.
Martin