[syslinux] regression: relocatable kernels on a chromebook

[Ady]

> On Sat, 7 Feb 2015, Ady via Syslinux wrote:
> > Thank you for this meaningful report. Ideally, I would suggest 
> > performing a similar test (at least with the same kernel built with all 
> > the above "config_*=y" settings) with official pre-built Syslinux 
> > versions 4.07 and 3.86 (remembering that all Syslinux-related files, 
> > including c32 modules, if being used, shall match the same version of 
> > the bootloader).
> 
> You're welcome :-) I'll post a patch to fix this problem as a follow-up to 
> my other patch. I've confirmed 4.07 and 5.10 are unaffected both by 
> running the pre-built versions from kernel.org and by inspecting their 
> code. Pre-built 6.01 is unaffected, while 6.02 and 6.03 are.
> 
 
Thank you for the reports, the tests and the patch.

There have been discussions and/or reports somewhat related to this 
matter during 2013 (Mar-Dec), and reports about failing to boot some 
kernels with Syslinux 6.03 (2014Q4).

The topics involved:

_ Kernels with no protected-mode code (e.g. grub's kernel lnxboot.img 
initrd=core.img)
_ bzImage and zImage files
_ memdisk
_ grldr, plop(bm)...
_ memtest (and alternatives), DBAN, several hardware testers...
_ gPXE/iPXE...
_ Older kernels and/or kernels for older machines (e.g. PuppyLinux and 
family and derivatives; I could mention several others)
_ Non-relocatable and relocatable kernels
_ Linux kernels prior to version 3.3 and/or with similar "older 
(EFISTUB) standards", "supposedly" bootable in BIOS and/or (U)EFI 
systems at the time (e.g. Slackware 14.0 / 14.1, Debian 7 Wheezy, RHEL 
6.x / 7 )
_ BIOS and/or (U)EFI systems

As a remainder and FWIW, from prior discussions (2014Q4) we have:
http://www.syslinux.org/wiki/index.php/Common_Problems#Linux_EFI_kernels

Yet, "fallback" method(s) should be allowed (in some way; with and/or 
without user intervention, and/or with/without printing information for 
the user). This feature is currently not present in Syslinux 6.03.

Additionally, I would tend to think that whichever the conditions / 
standards / rules to "correctly" boot kernels in UEFI systems, there 
should not be an impediment to boot kernels in BIOS systems.

In Syslinux 6.03, we seem to have some kernels not bootable in (some) 
UEFI systems, possibly because of some conventions (which were not yet 
fully in-use in prior kernels) and some kernels not bootable in (some) 
BIOS machines. Ouch.

I cannot comment on the code itself, but the reported behaviors seem to 
be consistent.

Hopefully relevant developers (Peter? Matt? Others? Anyone?) can take a 
look at the patch [1] and comment about it here. Hopefully there won't 
be regressions.

[1]:
[syslinux] [PATCH] load_linux: relocate protected-mode code as intended
 http://www.syslinux.org/archives/2015-February/023209.html 


TIA,
Ady.
 
> _______________________________________________
> Syslinux mailing list
> Submissions to Syslinux at zytor.com
> Unsubscribe or set options at:
> http://www.zytor.com/mailman/listinfo/syslinux
>