Syslinux - Jan 2014 - USB boot problems on Gigabyte GA-M55Plus-S3G

Hi,

(mixing statements from various mails)

Ronald F. Guilmette wrote:
> To repeat, I already do know exactly how to set or unset the bootable
> flag.
Sorry for not getting this right on the first reading.

I currently understand that a set active/bootable flag prevents the
USB stick from showing up in BIOS, and that without that flag it
begins to boot but fails with files not found.

> After that, if anyone wants to produce a test image that I can simply
> download and directly dd to a USB stick,
> [other mail]
> Well, I now have a _minimal_ install of ArchLinux
If you installed Archlinux from an ISO image, then this is probably
a test candidate for USB stick. 

E.g. archlinux-2013.11.01-dual.iso contains an isohybrid MBR
which makes it suitable for dd'ing on the whole device (not
into a partition or slice of the device).
On my olde FreeBSD 8, the whole stick appears as /dev/da0.

After copying to stick, /sbin/fdisk shows 2 MBR partitions:
Number 1 has the active/bootable flag set. So you will have
to unset it on the stick.
Partition number 2 is for EFI booting and embedded inside
partition 1. Some people frown on that layout. If there is
suspicion that your mainboard hates it, then you may delete
the second partition. It marks the content of a single file
in the image (/EFI/archiso/efiboot.img) which you will not
need with PC-BIOS.

Another candidate image would be
 
http://cdimage.debian.org/debian-cd/current/i386/iso-cd/debian-7.3.0-i386-netinst.iso
which would have the advantage of not containing EFI/GPT stuff,
that might introduce more confusion. (I could not figure out
whether your mad board claims to be capable of EFI.)

> I, as an ordinary end-luser, shouldn't
> have to employ specializsed install tools,
>From the user point of view you would indeed be best off if you
returned the board to the seller and demand a different one.
Its firmware is ill. No doubt. Its reaction on the active/boot
flag clearly indicates it.

But we bystanders are interested in learning how to succeed
even with this miserable hardware. One never knows when one
of our users steps into the same puddle.
So we use you as probe, as long as you are willing to boldly
go where no human has gone before.


Have a nice day :)

Thomas

In message <16222645049513725239 at scdbackup.webframe.org>, 
"Thomas Schmitt" <scdbackup at gmx.net> wrote:
>I currently understand that a set active/bootable flag prevents the
>USB stick from showing up in BIOS,
yes.
>and that without that flag it
>begins to boot but fails with files not found.
Well, for this case, it has appeared to me to fail in various different ways.
>> After that, if anyone wants to produce a test image that I can simply
>> download and directly dd to a USB stick,
>> [other mail]
>> Well, I now have a _minimal_ install of ArchLinux
>
>If you installed Archlinux from an ISO image, then this is probably
>a test candidate for USB stick. 
OK, I have the file "archlinux-2014.01.05-dual.iso" and I have just
now
dd'd it to a 8GB Patriot brand stick I have here.  (I used bs=1m.  I'm
not sure if that is the a proper block size to use or not when doing
stuff like this, but I believe it is.)

Just for the sake of clarity, now, fdisk on my FreeBSD system reports
this for the stick:

=========================================================================*******
Working on device /dev/da1 *******
parameters extracted from in-core disklabel are:
cylinders=973 heads=255 sectors/track=63 (16065 blks/cyl)

parameters to be used for BIOS calculations are:
cylinders=973 heads=255 sectors/track=63 (16065 blks/cyl)

Media sector size is 512
Warning: BIOS sector numbering starts with sector 1
Information from DOS bootblock is:
The data for partition 1 is:
sysid 0 (0000),(unused)
    start 0, size 1081344 (528 Meg), flag 80 (active)
        beg: cyl 0/ head 0/ sector 1;
        end: cyl 527/ head 63/ sector 32
The data for partition 2 is:
sysid 239 (0xef),(EFI System Partition)
    start 248, size 63488 (31 Meg), flag 0
        beg: cyl 1023/ head 254/ sector 63;
        end: cyl 1023/ head 254/ sector 63
The data for partition 3 is:
<UNUSED>
The data for partition 4 is:
<UNUSED>

=========================================================================
Humm... well... son-of-a-gun!!  This stick *does* show up in the
Boot Priority list on the GA-M55Plus-S3G, and ... wait for it...
it *does* boot up to the Archlinux start screen!

Yipee!  *Something* actually works!

OK, so now what?

Now that I know that this motherboard isn't totally hopeless, I'm a bit
more willing to try to provide more help to you guys, as you work to iron
out whatever problems have prevented Clonezilla and UBCD and OpenELEC from
booting on this motherboard.

>> I, as an ordinary end-luser, shouldn't
>> have to employ specializsed install tools,
>
>>From the user point of view you would indeed be best off if you
>returned the board to the seller and demand a different one.
hehe

Now, *that* isn't going to happen.  I first purchased this board at
least several years ago.
>But we bystanders are interested in learning how to succeed
>even with this miserable hardware.
I understand, am sympathetic, and want to be helpful, if I can be.  But
there have to be limits.  I only have so much time that I can invest
in this.
>One never knows when one
>of our users steps into the same puddle.
>So we use you as probe, as long as you are willing to boldly
>go where no human has gone before.
Yes.

Am 16.01.2014 08:17, schrieb Thomas Schmitt:
> Another candidate image would be 
>
http://cdimage.debian.org/debian-cd/current/i386/iso-cd/debian-7.3.0-i386-netinst.iso
> which would have the advantage of not containing EFI/GPT stuff, that 
> might introduce more confusion. (I could not figure out whether your 
> mad board claims to be capable of EFI.)
The board first appeared mid 2006, it is even AM2 and not AM2+. As far 
as I remember the first EFI mainboards in the PC market (not Mac, not 
servers) appeared late 2007, but often were limited to some kind of "CSM 
only mode" and cannot be switched to EFI boot. Since early 2010 nearly 
every freshly released mainboard/notebook seems to support EFI (at least 
without Secure Boot), albeit before Windows 8 they usually were 
configured to CSM as default.

I've seen some totally broken EFI implementations originating from 
between 2008 and 2010, for example early TianoCore implementations that 
do not allow to be configured as "CSM only" or "EFI only"
and get
totally confused if a USB thumb drive contains GPT, compat MBR and UEFI 
as well as legacy boot partitions (a layout that works flawlessly on 
modern UEFI implementations and old BIOSes).
> From the user point of view you would indeed be best off if you 
> returned the board to the seller and demand a different one. Its 
> firmware is ill. No doubt. Its reaction on the active/boot flag 
> clearly indicates it. But we bystanders are interested in learning how 
> to succeed even with this miserable hardware. 
To fix broken firmware I would first suggest to check if a BIOS update 
is available and then reset the BIOS to it's defaults.

As a bystander I am taking notes for every piece of ill hardware. 
Between 1% and 5% of all computers out there seem not to behave as 
expected when booting from removable media. I am preparing bootable live 
systems for magazines and customers like botfree.eu  on a monthly base. 
Those might sum up to a few hundred thousand or a bit more than a 
million "cover mount" DVDs per year I prepare. 1% of a million is the 
usual noise, more than 5% means extra work. So I am trying to keep track 
of even the weirdest bugs.

Regards,
Mattias

-- 
Mattias Schlenker - Redaktion + EDV-Beratung + Linux-CD/DVD-Konzepte
August-Bebel-Str. 74 - 04275 LEIPZIG - GERMANY

Bitte fuer geschaeftliche Telefonate vorzugsweise die VoIP-Telefonnummer
+49 341 39290767 verwenden, da ich diese aufs Mobiltelefon routen kann!

Hi,
> OK, I have the file "archlinux-2014.01.05-dual.iso" and I have
just now
> dd'd it to a 8GB Patriot brand stick I have here.  (I used bs=1m. 
I'm
> not sure if that is the a proper block size to use or not when doing
> stuff like this, but I believe it is.)
With a data file as source, the dd block size should not matter.
It may influence the speed of the copy run.

> it *does* boot up to the Archlinux start screen!
> Yipee!  *Something* actually works!
Quality filesystem: ISO 9660. :)))

>    start 0, size 1081344 (528 Meg), flag 80 (active)
Did you unset the "active" flag before your experiment ?
It is the same as "bootable".

> OK, so now what?
Did the failures happen with the same USB stick as was
successful with the ISO image ?

Your previous failures were probably all with variants of the
filesystem FAT. The error messages you showed were about
missing files. So maybe the firmware has a problem with FAT.
It should not think that much about partition content.
But as it is susipicious to be mad anyway ...

> I first purchased this board at least several years ago.
That makes it unlikely we have to deal with too much EFI
in the firmware. (EFI knows about FAT.)

Nevertheless, the successful ISO has a GPT partition table
additionally to the MBR partition table. (That's why it is
called "dual".iso)
Did you try debian-7.3.0-i386-netinst.iso which has only MBR ?
Maybe it fails. That would be great.


My only other idea would be to install some system in a
non-FAT read-write filesystem. Like ext2.
That could strenghten or weaken the suspicion about FAT
partitions.

Any proposals from the other bystanders ?
Goal is to bisect the disk characteristics until we have
some single property that triggers failure resp. success.


Have a nice day :)

Thomas

In message <52D846EC.2060303 at mattiasschlenker.de>, 
Mattias Schlenker <ms at mattiasschlenker.de> wrote:
>Am 16.01.2014 08:17, schrieb Thomas Schmitt:
>> Another candidate image would be 
>>
http://cdimage.debian.org/debian-cd/current/i386/iso-cd/debian-7.3.0-i386-ne
>tinst.iso 
I just happened to have a copy of this file also already, so I dd'd it to
the same stick, and volia!  This image also boots with no problems on
the GA-M55Plus-S3G.

Just another data point.
>The board first appeared mid 2006, it is even AM2 and not AM2+.
Yes.  To my shame, I admit that this is a fairly old board.

I hope that non of you guys will hate me for not confessing this sooner,
and I hope that you won't be mad at me for asking for help relating to
such an old board.


Regards,
rfg