Skip site navigation (1)Skip section navigation (2)
Date:      Fri, 2 Nov 2012 20:57:01 +0100
From:      Mateusz Guzik <mjguzik@gmail.com>
To:        Gary Palmer <gpalmer@freebsd.org>
Cc:        Bas Smeelen <b.smeelen@ose.nl>, freebsd-stable@freebsd.org
Subject:   Re: [patch] Re: SU+J on 9.1-RC2 ISO
Message-ID:  <20121102195701.GA496@dft-labs.eu>
In-Reply-To: <20121102185929.GA24320@in-addr.com>
References:  <5093F934.7050306@ose.nl> <5093FD3D.3080201@ateamsystems.com> <1351876381.2657.1.camel@mjakubik.localdomain> <50940276.5030306@ateamsystems.com> <50940C20.3090409@ose.nl> <50940E40.3090709@ose.nl> <5094112C.2070102@ose.nl> <20121102184131.GB22755@dft-labs.eu> <20121102185929.GA24320@in-addr.com>

next in thread | previous in thread | raw e-mail | index | archive | help
On Fri, Nov 02, 2012 at 02:59:29PM -0400, Gary Palmer wrote:
> On Fri, Nov 02, 2012 at 07:41:31PM +0100, Mateusz Guzik wrote:
> > On Fri, Nov 02, 2012 at 07:30:04PM +0100, Bas Smeelen wrote:
> > > On 11/02/2012 07:17 PM, Bas Smeelen wrote:
> > > > On 11/02/2012 07:08 PM, Bas Smeelen wrote:
> > > >> On 11/02/2012 06:27 PM, Adam Strohl wrote:
> > > >>> On 11/3/2012 0:13, Mike Jakubik wrote:
> > > >>>> You can disable SU+J after installing, though it would be nice if the
> > > >>>> installer gave you a choice.
> > > >>> This assumes that you know about this flaw, which most people do not.
> > > >>>
> > > >>> I didn't until I discovered it by panic-ing a perfectly fine running
> > > >>> server.  Getting burned by a known bug like this shouldn't be "SOP"
> > > >>> for users of FreeBSD.
> > > >>>
> > > >>> If anything it should be turned off by default, and people can turn it
> > > >>> on if they want given the landmine it plants.  If they know how to
> > > >>> turn it on they're much more likely to be aware of the issue.
> > > >>>
> > > >>>
> > > >>>
> > > To sum it up
> > > SU+J should be turned off by default because of
> > > 1. It does not work with dumping a live system e.g. snapshot
> > > 2. it is not recommended for SSD installs
> > > 3. "Smart" admins can turn it on if they want
> > > 
> > > root@sys:/usr/src/usr.sbin/bsdinstall/partedit # diff -u gpart_ops.c 
> > > gpart_ops.cnew
> > > --- gpart_ops.c    2012-08-06 01:54:33.000000000 +0200
> > > +++ gpart_ops.cnew    2012-11-02 19:07:45.000000000 +0100
> > > @@ -90,8 +90,8 @@
> > >               {"SU", "Softupdates",
> > >                   "Enable softupdates (default)", 1 },
> > >               {"SUJ", "Softupdates journaling",
> > > -                "Enable file system journaling (default - "
> > > -                "turn off for SSDs)", 1 },
> > > +                "Disable file system journaling (default - "
> > > +                "turn on for adventurish admins)", 0 },
> > >               {"TRIM", "Enable SSD TRIM support",
> > >                   "Enable TRIM support, useful on solid-state drives",
> > >                   0 },
> > > 
> > > Please comment, then I can file a PR or not
> > 
> > As was noted in my another mail, the kernel will no longer crash when an
> > attempt to take a snapshot is made. Also AFAIR SUJ can be disabled
> > later.
> > 
> > Given that I prefer the following:
> > 
> > diff --git a/usr.sbin/bsdinstall/partedit/gpart_ops.c b/usr.sbin/bsdinstall/partedit/gpart_ops.c
> > index 479365a..80296c2 100644
> > --- a/usr.sbin/bsdinstall/partedit/gpart_ops.c
> > +++ b/usr.sbin/bsdinstall/partedit/gpart_ops.c
> > @@ -91,7 +91,7 @@ newfs_command(const char *fstype, char *command, int use_default)
> >  			    "Enable softupdates (default)", 1 },
> >  			{"SUJ", "Softupdates journaling",
> >  			    "Enable file system journaling (default - "
> > -			    "turn off for SSDs)", 1 },
> > +			    "turn off for SSDs or if you use snapshots)", 1 },
> >  			{"TRIM", "Enable SSD TRIM support",
> >  			    "Enable TRIM support, useful on solid-state drives",
> >  			    0 },
> > 
> > http://people.freebsd.org/~mjg/patches/suj-snapshot-comment.diff
> 
> How many people realise that snapshots are needed for dump based backups
> (and other related features)?
> 

I thought this is a common knowledge.

When you run dump without -L on a live filesystem, it advises you to re-run
with -L option. And -L description in the man page clearly states that it
uses snapshots.

The comment can be updated to mention that snapshots are used by dump
with -L.

I don't have a strong opinion on this subject, just wanted to correct
Adam on his claims regarding panics with snapshots and while here I
provided a patch that is more likely to get accepted since it does not
change any defaults (and I hoped it would address concerns about admins
installing new systems with SUJ unaware that it does not work with
snapshots).

-- 
Mateusz Guzik <mjguzik gmail.com>



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20121102195701.GA496>