Bug 6297

Summary: Define install target scope
Product: [Build System, Metadata & Runtime] OE-Core Reporter: Darren Hart <dvhart>
Component: deploymentAssignee: Scott Rifenbark <srifenbark>
Status: RESOLVED FIXED QA Contact: Costin <costin.c.constantin>
Severity: enhancement    
Priority: Medium+ CC: alexandru.c.georgescu, blloyd, bluelightning, corneliux.stoicescu, mark.hatle, seebs, sgw, srifenbark
Version: 1.7   
Target Milestone: 2.1   
Hardware: x86   
OS: Multiple   
Whiteboard: 18 Feb 2016: RESOLVED
OS type for building Yocto: --- Type of Regression: ---
Verified: Documentation change: Done (doc changes complete)
Attachments:
Description Flags
Installer scope
none
Installer scope none

Description Darren Hart 2014-05-08 15:04:00 UTC
We have a very simple installer in oe-core for grub and syslinux boot loaders. It has a number of limitations and specific requirements people often run up against. We need to define the scope for the installer in order to be able to triage bugs filed against it.
Comment 1 Saul Wold 2014-07-01 17:46:20 UTC
Working with different boot loaders, 
Race with Sata vs USB drives (sda vs sdb when booting from USB)
UUID - initial ram disk 
Multiple Kernel - not in scope
Live Image -> cpio root / kernel/ initramfs
Rootfs has installer scripts (basic and efi)
basic partition (msdos or gpt) 
   efi will create 2 parts (system & root)
   Mount rootfs looped and copy from rootfs -> root partition
   magic for boot loader and mbr.
Wic as possible installer on target?
 - Create a couple of simple install targets
 - Use plugins to gummiboot target

Release note or white paper fo definition of how to boot and deploy with the 
varitey of booters

Review Scope
 - Minimal User interaction (install to disk ? (select from list if available)
 - On target install for Automated QA testing
 - Rootfs based
   - possible plugin arch to install other pacakges such as ptest. 
 - Rootfs partitioning only (system, rootfs, swap)
 - Sata vs USB 
   - initramdisk to enable UUID?
   - udev rules
   - Check with Robert on what WR does
 - Reuse WIC where possible, what are the requirements for wic (python, 
libraries and tools)
   - Maybe it does not make sense
   - How big wic vs c-i-minimal
   - Can we have initramfs (could install any target image)

Boot off some media
Document steps to  
Implement Step
Comment 2 Blloyd 2014-07-03 15:25:59 UTC
With "multiple kernel - not in scope", are you referring to being able to have two different kernels in the final image and selecting a different one each boot right?  This is not covering the need of scripts to be able to patch/replace a kernel image with a newer one?
And with several boot technologies the kernel file ends up in multiple locations.

Upgrading of all parts of a system is a necessary part of any long term deployment because any component could have a security breaking issue that needs to be removed, therefore the installer may only need to have one bootable setup, but systems produced by the installer need to be able to reliably find all parts of their boot chain so they can update all their components after the installer is done.
Comment 3 Saul Wold 2014-07-16 14:39:20 UTC
Add documentation to installer area in initrd
Comment 4 Stephen K Jolley 2014-07-17 15:07:32 UTC
*** Bug 4105 has been marked as a duplicate of this bug. ***
Comment 5 Saul Wold 2015-04-29 17:10:13 UTC
Scott, please use the information here to start working on a white paper, I am 
a lousy copy writer, but can spend some time talking with you about this.
Comment 6 Mark Hatle 2015-04-29 17:15:32 UTC
We have someone working on a generic interface for on-target kernel package (and subsequent bootloader setup).

I'll see if he can be connected into this project.  (Our intent is to submit this as part of the YP 1.9 development.)
Comment 7 Scott Rifenbark 2015-05-04 21:13:03 UTC
Adding 1.9 M4 as milestone and 5 "perfect" days for this.  This of course assumes I would have all the perfect information with perfect timing, none of which would happen.

Scott
Comment 8 Scott Rifenbark 2015-05-11 21:20:29 UTC
Hi, 

I had a brief chat with Saul on the phone and got a clearer idea of what Comment 2 is driving at.  I will create a working draft for this paper based on Saul's outline and my understanding of it.  We will iterate from there.

Scott
Comment 9 Scott Rifenbark 2015-05-29 19:04:33 UTC
Created attachment 2532 [details]
Installer scope

Whitepaper to help define the scope of the simple bootloader we have in oe-core for grub and syslinux.
Comment 10 Scott Rifenbark 2015-05-29 19:05:23 UTC
Created attachment 2533 [details]
Installer scope

PDF version of the file.
Comment 11 Scott Rifenbark 2015-05-29 19:08:38 UTC
Saul, 

I attached a couple files as a first pass for this whitepaper on the installer scope.  I made a lot of guesses and assumptions on much of it.  I tried to "bucket" the information into some areas based on our previous phone discussion.  I attached the original text from your Comment 1 so you can see the transition from that to what I have.  

We need to work more on this and expand things so that it is readable by normal people.

Thanks, 
Scott
Comment 12 Scott Rifenbark 2015-09-22 14:52:48 UTC
So this paper can be seen from the link in Comment 10.  I have received no feedback on it since putting it out there.  Leaving everything as is.  Not sure where we are with the state of this.

Scott
Comment 13 Scott Rifenbark 2016-02-09 22:37:08 UTC
No feedback for a long time.  Setting to RESOLVED and putting doc flag to "done."

Scott
Comment 14 Scott Rifenbark 2016-02-18 15:42:04 UTC
Hi, 

See https://wiki.yoctoproject.org/wiki/Installer_Scope for a new wiki page on this.  I set the bug's doc flag to "done."

Scott