Skip site navigation (1)Skip section navigation (2)
Date:      Sun, 24 Apr 2011 09:14:03 -0500
From:      Nathan Whitehorn <nwhitehorn@freebsd.org>
To:        Robert Watson <rwatson@FreeBSD.org>
Cc:        svn-src-head@freebsd.org, Alexander Best <arundel@freebsd.org>, Alexander Motin <mav@FreeBSD.org>, src-committers@freebsd.org, svn-src-all@freebsd.org
Subject:   Re: svn commit: r220983 - head
Message-ID:  <4DB4302B.5050501@freebsd.org>
In-Reply-To: <alpine.BSF.2.00.1104241148410.36270@fledge.watson.org>
References:  <201104240923.p3O9N8QG025386@svn.freebsd.org> <20110424095736.GA92492@freebsd.org> <alpine.BSF.2.00.1104241148410.36270@fledge.watson.org>

next in thread | previous in thread | raw e-mail | index | archive | help
On 04/24/11 05:49, Robert Watson wrote:
>
> On Sun, 24 Apr 2011, Alexander Best wrote:
>
>>> ==============================================================================
>>>
>>> --- head/UPDATING Sun Apr 24 08:58:58 2011 (r220982)
>>> +++ head/UPDATING Sun Apr 24 09:23:08 2011 (r220983)
>>> @@ -28,8 +28,9 @@ NOTE TO PEOPLE WHO THINK THAT FreeBSD 9.
>>> removed and replaced by respective CAM drivers. If you are using
>>> ATA device names in /etc/fstab or other places, make sure to update
>>> them respectively (adX -> adaY, acdX -> cdY, afdX -> daY, astX -> saY,
>>> - where 'Y's are the sequential numbers for each type in order of
>>> - detection, unless configured otherwise with tunables, see cam(4)).
>>> + where 'Y's are the sequential numbers starting from zero for each type
>>> + in order of detection, unless configured otherwise with tunables,
>>> + see cam(4)).
>>
>> maybe an extra sentence mentioning the fact that {geom,ufs}-labels and
>> ufsids are preferred over device nodes nowadays would be a good idea?
>
> If this is true, that's what our installer needs to do. Does our
> installer do that?
>
> (In general, a reasonable definition of "preferred" is "what the
> installer does" -- if that's not the case, it's a serious bug.)

Our installer does not do that. I outlined the reasons in another 
earlier email, but the summary is that labels seem to be somewhat 
fragile. Gpart supports a partition-generic labeling interface for three 
schemes: PC98, GPT, and APM. The installer now supports setting labels 
on all of these. The appearance of labeled geom providers happens in a 
different module, glabel, which relies on a totally parallel parse of 
the partition map. As a result, labeled partitions appearing in /dev 
works only on GPT volumes, and even then, due to endianness bugs, only 
on little-endian platforms. Glabel is also not guaranteed available by 
gpart, and I'm a little dubious that reparses happen at the correct time 
with 100% reliability. The other, more minor, difficulty, is that it is 
somewhat difficult to get the labeled provider name from the gpart 
provider (you have to try to cross-correlate names, and there may be 
many label providers to choose from for the same partition).

If gpart were modified to provide label providers itself based on its 
generic labeling mechanism, and to provide the name of those providers 
in its config XML, using labels by default in the installer would be 
trivial and I would have it in the tree within 5 minutes of getting that 
commit mail. Unfortunately, I don't have time to do the requisite gpart 
modifications (which probably aren't that hard) at this time.
-Nathan



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?4DB4302B.5050501>