Skip site navigation (1)Skip section navigation (2)
Date:      Fri, 11 Mar 2016 13:00:32 +0300
From:      Andrey Fesenko <f0andrey@gmail.com>
To:        krad <kraduk@gmail.com>
Cc:        =?UTF-8?Q?Trond_Endrest=C3=B8l?= <Trond.Endrestol@fagskolen.gjovik.no>,  "freebsd-hackers@freebsd.org" <freebsd-hackers@freebsd.org>, freebsd-current <freebsd-current@freebsd.org>
Subject:   Re: EFI zfs loader and beadm?
Message-ID:  <CA%2BK5SrMDTwLCisvYrKeC7Kvj9aACdHg70QHbU74jfznKzpbPTg@mail.gmail.com>
In-Reply-To: <CALfReyenvWqM9-Xd41eOo29eH2gJrv1qSExWMHX6AFpbxx0eGg@mail.gmail.com>
References:  <CA%2BK5SrPZ-esGHU=rggSgOFiyy4NqOWhP31rfO4S=P9=m10pEbw@mail.gmail.com> <56E045C8.1040000@vangyzen.net> <CA%2BK5SrPhwH4%2BtSHs9d=xg1mzBbojtWwXhMqFeL0MdKmAJywKTg@mail.gmail.com> <CALfReydcsb-MhaTYAwMeEkJm0QGgUsgSweUC%2BObOFKSv-G4QYw@mail.gmail.com> <CA%2BK5SrOuah4j_U_qPpnfqPo5nhSF7DKwYR63wiW85Qj0bbOiHA@mail.gmail.com> <CALfReyceEMiN7czQ9wU8oBTC706jdL9JnoED6r3g%2BhaH30J%2BQw@mail.gmail.com> <CA%2BK5SrPAhtt3cao1-i54z6dDpdFh7UXLqD7WPBs2jhYxb6EAKA@mail.gmail.com> <CALfReyfoxgzTk%2Be9EMdheE=0TCP0g_eVDr_wajyrz1AiGoPWNQ@mail.gmail.com> <CA%2BK5SrNKtqcc79ipgx7hsRvM7n5Z-WUNDAzzeugizimTTMT_=Q@mail.gmail.com> <alpine.BSF.2.20.1603102014210.1014@mail.fig.ol.no> <CALfReyenvWqM9-Xd41eOo29eH2gJrv1qSExWMHX6AFpbxx0eGg@mail.gmail.com>

next in thread | previous in thread | raw e-mail | index | archive | help
On Fri, Mar 11, 2016 at 10:52 AM, krad <kraduk@gmail.com> wrote:
> Its also worth pointing that if you decouple the userland from the kernel
> files you no longer have a boot environment, as all the basic stuff to bo=
ot
> the os isn't contained within the bootfs. I'm still struggling to see why
> /boot needs to be on a different dataset, its just a bit of a linuxism.
>
> On 10 March 2016 at 19:23, Trond Endrest=C3=B8l
> <Trond.Endrestol@fagskolen.gjovik.no> wrote:
>>
>> On Thu, 10 Mar 2016 18:38+0300, Andrey Fesenko wrote:
>>
>> > On Thu, Mar 10, 2016 at 6:11 PM, krad <kraduk@gmail.com> wrote:
>> > > As Eric said you cant have /boot on a separate dataset as the whole
>> > > loader
>> > > bootstrap isnt designed too look for it on the dataset defined by
>> > > bootfs.
>> > > Remember no other datasets are mounted at that stage of the bootstra=
p.
>> > >
>> > > You could maybe bodge something by manually playing around with the
>> > > bootfs
>> > > property, symlinks and rootfs variables in the loader.conf.  But why
>> > > would
>> > > you want to do this? It's more work and non standard, and will break=
 a
>> > > lot?
>> > >
>> > >
>> > >
>> > > On 10 March 2016 at 12:11, Andrey Fesenko <f0andrey@gmail.com> wrote=
:
>> > >>
>> > >> On Thu, Mar 10, 2016 at 2:55 PM, krad <kraduk@gmail.com> wrote:
>> > >> > presumably it boots now?
>> > >> >
>> > >> > On 10 March 2016 at 11:01, Andrey Fesenko <f0andrey@gmail.com>
>> > >> > wrote:
>> > >> >>
>> > >> >> On Thu, Mar 10, 2016 at 1:49 PM, krad <kraduk@gmail.com> wrote:
>> > >> >> > Make sure you are running the latest snapshot of current or 10=
.3
>> > >> >> > as
>> > >> >> > well, as
>> > >> >> > the MFC commits were in early February for 10-stable
>> > >> >> >
>> > >> >> >>
>> > >> >> >> If remove efiwpool/ROOT/init/boot and copy his content on
>> > >> >> >> efiwpool/ROOT/init my scheme work fine too.
>> > >> >> >> /usr /var /home and other included in BE for consistent boot
>> > >> >> >> system
>> > >> >> >> (CURRENT world may not boot with kernel other rev),  and old
>> > >> >> >> home
>> > >> >> >> snapshot sometimes useful for backup/restore
>> > >> >> >> _______________________________________________
>> > >> >>
>> > >> >> % uname -a
>> > >> >> FreeBSD x220.efi.local 11.0-CURRENT FreeBSD 11.0-CURRENT #0
>> > >> >> r296548:
>> > >> >> Wed Mar  9 01:16:17 MSK 2016
>> > >> >> root@des.local:/usr/obj/usr/src/sys/X220  amd64
>> > >> >
>> > >> >
>> > >>
>> > >> My current working config
>> > >> % mount
>> > >> ....
>> > >>
>> > >> This work fine, booted, beadm create new env, activate them, see bo=
ot
>> > >> menu and select BE.
>> > >>
>> > >> % beadm list
>> > >> BE    Active Mountpoint  Space Created
>> > >> init  -      -          420.7M 2016-03-09 02:57
>> > >> init0 NR     /           35.9G 2016-03-10 05:00
>> > >>
>> > >> If i'm add separate dataset for /boot (efiwpool/ROOT/init0/boot)
>> > >> system not booted, efi loader (first stage) see only my pool, not
>> > >> found /boot/loader.efi
>> > >
>> > >
>> >
>> > It probably does not matter, as bootfs have snapshots (BE), just
>> > wanted to make it more clear (having taken significant mountpoint
>> > /boot, /usr, /var... in zfs dataset) and was surprised why the system
>> > does not boot
>> >
>> > It is clear that as long as the functionality is experimental and
>> > under development, but would like to see where the full instructions
>> > on its implementation / restrictions, at least as early as has been
>> > described https://wiki.freebsd.org/RootOnZFS
>>
>> If you keep /boot as a separate dataset/filesystem, with
>> efiwpool/ROOT/init0/boot as the given bootfs, then boot1.efi will not
>> see a /boot directory inside that dataset. The files and directories
>> from /boot will be presented as living in /, the local root directory
>> of that dataset.
>>
>> You could create a /boot/boot symlink pointing to . (dot), but it's
>> better to let /boot be part of the regular boot environment, pretty
>> similar to what you would find on a UFS system using a separate root
>> filesystem.
>>

why?
On the contrary, my scheme involves the entire system as a single BE
(including home)

> linuxism
Is unlikely, to rather an attempt to use all the features ZFS. ZFS
also many possible settings for each dataset/FS (compression, atime,
exec......) Probably for /boot many of them unnecessary, but such
separate extra snapshots, can would not be harmful



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?CA%2BK5SrMDTwLCisvYrKeC7Kvj9aACdHg70QHbU74jfznKzpbPTg>