svn commit: r256343 - in head/usr.sbin/bsdinstall: . scripts

[Allan Jude]

On 2013-10-15 12:28, Teske, Devin wrote:
> On Oct 15, 2013, at 8:23 AM, Allan Jude wrote:
>
>> On 2013-10-15 11:00, Teske, Devin wrote:
>>> CC'ing Allan Jude for feedback (I didn't have a hand in the
>>> selection of those values)
>>>
>>> -- 
>>> Devin
>>>
>>>
>>> On Oct 15, 2013, at 7:50 AM, Bryan Drewery wrote:
>>>
>>>
>>>> On Fri, Oct 11, 2013 at 08:41:35PM +0000, Devin Teske wrote:
>>>>
>>>>> Author: dteske
>>>>>
>>>>> Date: Fri Oct 11 20:41:35 2013
>>>>>
>>>>> New Revision: 256343
>>>>>
>>>>>>> URL: http://svnweb.freebsd.org/changeset/base/256343
>>>>>>>
>>>>>>>
>>>>> Log:
>>>>>
>>>>>  Add zfsboot module as an option for automatic configuration. Default is
>>>>>
>>>>>  to run interactively but it can be scripted too (optinally completely
>>>>>
>>>>>  non-interactive). Currently supports GELI and all ZFS vdev types. Also
>>>>>
>>>>>  performs validation on selections/settings providing error messages if
>>>>>
>>>>>  necessary, explaining (in plain language) what the issue is. Currently
>>>>>
>>>>>  the auto partitioning of naked disks only supports GPT and MBR (VTOC8
>>>>>
>>>>>  pending for sparc64), so is only available for i386/amd64 install.
>>>>>
>>>>>>>  Submitted by:	Allan Jude <freebsd at allanjude.com>
>>>>> , myself
>>>>>
>>>>>>>  Reviewed by:	Allan Jude <freebsd at allanjude.com>
>>>>>>>
>>>>>  Approved by:	re (glebius)
>>>>>
>>>>>>>
>>>>> Added:
>>>>>
>>>>>  head/usr.sbin/bsdinstall/scripts/zfsboot   (contents, props changed)
>>>>>
>>>>> Modified:
>>>>>
>>>>>  head/usr.sbin/bsdinstall/bsdinstall
>>>>>
>>>>>  head/usr.sbin/bsdinstall/scripts/Makefile
>>>>>
>>>>>  head/usr.sbin/bsdinstall/scripts/auto
>>>>>
>>>>>  head/usr.sbin/bsdinstall/scripts/config
>>>>>
>>>>>>>
>>>>> +# Should we use gnop(8) to configure a transparent mapping to 4K sectors?
>>>>>
>>>>> +#
>>>>>
>>>>> +: ${ZFSBOOT_GNOP_4K_FORCE_ALIGN:=1}
>>>>>
>>>>>
>>>>>
>>>> I believe this trick is no longer needed after r254591.
>>>>
>> This may be the case, I am not sure if r254591 'always' fixes the problem, maybe someone else can weight in with a more definite yes/no
>>
> I'll hold off on doing anything with 4K sectors gnop hack until
> we get confirmation.
>
>
>
>
>
>>>>> *** DIFF OUTPUT TRUNCATED AT 1000 LINES ***
>>>>>
>>>>>
>>>> ...
>>>>
>>>>> +# Default name for the boot environment parent dataset
>>>>>
>>>>> +#
>>>>>
>>>>> +: ${ZFSBOOT_BEROOT_NAME:=bootenv}
>>>>>
>>>>>
>>>> ...
>>>>
>>>>> +       # Set bootfs property
>>>>>
>>>>> +       zpool set bootfs="$poolname/$ZFSBOOT_BEROOT_NAME/$ZFSBOOT_BOOTFS_NAME" \
>>>>>
>>>>> +               "$poolname" || return $FAILURE
>>>>>
>>>>>
>>>>>
>>>> Can you please make ZFSBOOT_BEROOT_NAME be ROOT by default? This is what
>>>>
>>>> sysutils/beadm, OpenSolaris/Illumos beadm, and I believe PC-BSD all expect
>>>>
>>>>> and use as well.
>> I suppose. I dislike having a dataset in all caps, and with a not exactly descriptive name. beadm-devel (used in PCBSD) supports figuring out where whatever ZFSBOOT_BEROOT_NAME you used, but since sysutils/beadm does not, POLA suggests we change the value back to ROOT
>>
> So we have consensus?
>
> I need to know today if we're going to get it into 10.0-BETA2 (3-day MFC cycle).

Lets change ZFSBOOT_BEROOT_NAME to 'ROOT' and remove the /usr/local and
/var/db/pkg datasets entirely (so they become part of the BE). Based on
IRC discussion with bdrewery and Shawn Webb

-- 
Allan Jude