[OmniOS-discuss] In release-151020 solaris10 brand is broken

[Dan McDonald]

> On Nov 21, 2016, at 12:23 AM, Илья Кулагин <kiv at kiv.pp.ru> wrote:
> 
> Thank you very much Dan: with patched module, zone boots, our legacy
> system (based on ancient oracle10) starts, database mounts and works.
> 
> I even did not need modunload -- because release module was not loaded :)
> 
> By the way. New lx with centos6 - I think - will be a good replacement
> for old solaris10, for our purpose.

Thanks for the quick test.

The PROPER solution is to:

1.) Have Joyent make sure they/I didn't miss anything ELSE.  This code was broken by partially backing out a change made mid-stream during Joyent's LX work.  The two commits of relevance on the illumos-omnios tree are:

commit 5312cd77c26a98281973239359f5c5cd5dd132b5
Author: DJ Hoffman <dj.hoffman at joyent.com>
Date:   Mon Jul 7 18:06:37 2014 +0000

    OS-3149 lx brand always sends SIGCHLD to parent processes, regardless of how clone was invoked
    OS-2887 lxbrand add WALL, WCLONE, WNOTHREAD support to waitid
    Reviewed by: Jerry Jelinek <jerry.jelinek at joyent.com>
    Reviewed by: Robert Mustacchi <rm at joyent.com>

and the backout:

commit 3bdacdb34e9adb4e92016ea49a7b9d9d3ca5e305
Author: Patrick Mooney <patrick.f.mooney at gmail.com>
Date:   Tue Mar 31 21:57:35 2015 +0000

    OS-4129 lxbrand should not abuse p_brand_data for storing exit signal
    Reviewed by: Jerry Jelinek <jerry.jelinek at joyent.com>
    Reviewed by: Joshua M. Clulow <jmc at joyent.com>



2.) Once those are confirmed, ship replacement packages.  The GOOD news is that I think we only need to ship brand/s10 and brand/sn1.  No reboot should be needed for those guys (unless IPS gets something wrong).



Thanks for the help!
Dan