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

In message <1767645779554124797 at scdbackup.webframe.org>, 
"Thomas Schmitt" <scdbackup at gmx.net> wrote:
>Hi,
>
>> > debian-7.3.0-i386-netinst.iso
>> This image also boots with no problems on the GA-M55Plus-S3G.
>
>So isohybrid boots with and without additional GPT.
>It boots with partition start 0 and with partition start at 32 kB.
OK.  If you say so.  (Hey!  What do I know?)
>Everything points to FAT or the individual USB stick.
>
>Can you confirm that the same USB stick works with isohybrid
>and fails with the hard-disk-style setups ?
I would like to try, if doing so will be of any help.  But I'm sorry
to say that I don't even understand the question.  If you want me to
do more (or another) test, please give me precice instructions and
I'll do my best to carry them out and report results.

(For clarity, I used my 8GB Patriot XT USB stick for the two tests I
did with (1) the "archlinux-2014.01.05-dual.iso" file and (2) the
"debian-7.3.0-i386-netinst.iso" file.)

I have several brands and sizes of USB sticks to play with here,
specifically:

      2GB Sandisk Blade
      4GB Sandisk Blade (3 of these)
      4GB Transcend JetFlash Model 530 (2 of these)
      8GB Patriot XT
      16GB Adata S102 (USB 3.0/2.0)
      32GB Patroit Axle

I can use any or all of these for testing.  Just tell me which ones
you want me to use.
>> Yes.  To my shame, I admit that this is a fairly old board.
>
>Old or not. It exists and offers resistance.
Thank you for your kind understanding.


Regards,
rfg

Hi,
> > So isohybrid boots with and without additional GPT.
> > It boots with partition start 0 and with partition start at 32 kB.
> OK.  If you say so.  (Hey!  What do I know?)
These are parameters which might influence bootability.
Now they are unlikely to play a role in the problem.

> > Can you confirm that the same USB stick works with isohybrid
> > and fails with the hard-disk-style setups ?
> I'm sorry to say that I don't even understand the question. 
Put Clonezilla or another failure candidate onto the 8 GB Patriot
stick that worked fine with isohybrid or Ady's image.
Check whether it fails with Clonezilla.

If it does fail, then this kills the theory about individual
problems between your board and a particular USB stick.
Then only the ideas of Mattias Schlenker remain: Some mishap
while the USB stick gets partitioned.

If you cannot produce a failure with Clonezilla on that stick,
then find a stick that fails with Clonezilla, and try this one
with isohybrid images or Ady's image.
If these images fail too, then the stick is to blame.


Have a nice day :)

Thomas

In message <2455764567543797146 at scdbackup.webframe.org>, 
"Thomas Schmitt" <scdbackup at gmx.net> wrote:
>... then find a stick that fails with Clonezilla, and try this one
>with isohybrid images or Ady's image.
I can do this test.  And have.

I still have the 2GB Sandisk Cruzer Blade which had Clonezilla 2.2.0 on
it at the time I started this thread.  I just now re-verified that (a)
it boots to Clonezilla just fine on two other systems I have here
(laptop and HTPC), and I verified also (b) that it still cannot be
booted by the GA-M55Plus-S3G (and further, that the GA-M55Plus-S3G
still does not even list it in it's Boot Priority list).

After re-verifying the above information, I have taken that 2GB Sandisk
Cruzer Blade and I have dd'd Ady's test.img file to it, using bs=1m as
before.

It boots up and prints "Hello, world, ..." on the GA-M55Plus-S3G.

For the sake of completness, I also tried the same stick after dd'ing
to it the file "archlinux-2014.01.05-dual.iso".

This also boots up with no difficulty on the GA-M55Plus-S3G.


Regards,
rfg

In message <2455764567543797146 at scdbackup.webframe.org>, 
"Thomas Schmitt" <scdbackup at gmx.net> wrote:
>> > Can you confirm that the same USB stick works with isohybrid
>> > and fails with the hard-disk-style setups ?
>> I'm sorry to say that I don't even understand the question. 
>
>Put Clonezilla or another failure candidate onto the 8 GB Patriot
>stick that worked fine with isohybrid or Ady's image.
>Check whether it fails with Clonezilla.
Please be specific.  For Clonezilla, there are many different versions
available on the Clonezilla web site at the present time (and I gather
that they may use various different versions of Syslinux, so this could
be important).

The version that I had on a USB stick (and that failed booting) at
the time I began this thread (2.2.0) is apparently no longer even
available from the Clonezilla web site... at least not obviously so.

Also, the install instructions for Cloenzilla suggest a total of four (4)
different ways to install the thing using MS Windows, and three (3) different
install methods if the person is using Linux instead.  Which one did you
want me to use?  (It seems that this may perhaps make a difference too.)

Furthermore, and separately, regardless of the install method (Linux or
Windows) the Clonezilla install instructions say that one must first
*begin* with a USB stick that has already had a FAT file system placed
onto it.  This of course implies that the stick must already have had
an MBR partition table _and_ a partition created on it by using
*something* to do those two steps.  What tool would you like me to
use to create and/or format the prerequsite partition?  (Keep in mind
that I have already been warned away from using GParted in this thread.)

Other "failure candidates" include (1) Ultimate Boot CD and (2)
OpenELEC,
but in these cases also, the versions I have/had on sticks at the time
I started this thread may perhaps no longer even be available from the
respective web sites (or maybe just not obviously so).  If you want me
to try one of these two packages, I will do so, but please tell me
which one you prefer, what version I should try, and the exact install
procedure you wish me to use.  (I greatly prefer ones that DO NOT rely
on Windows at all, if possible.  Thank you.)

I await your direction.


Regards,
rfg

Hi,

me:
> > Check whether it fails with Clonezilla.
Ronald F. Guilmette:
> Please be specific.  For Clonezilla, there are many different versions
Any failure on the same stick which also succeeds is enough
to kill the theory of bad relation between USB stick and
mainboard.
You have proven this by the success and failure with the 2 GB
stick.

So the problem must be either in partitioning or in the filesystem
format. That's out of my personal expertise, though.

> Also, the install instructions for Cloenzilla suggest a total of four (4)
> different ways to install the thing using MS Windows, and three (3)
> different install methods if the person is using Linux instead.
If there is one which promises to lead to LBA adresseing,
that would be the most hopeful candidate.

> I await your direction.
I lack the skills for explicit directions how to explore
the plausible suspicions of Ady and Peter:

Ady:
> I also "forced" a FAT32 LBA filesystem,
Ronald F. Guilmette:
>     *) with respect to Ultimate Boot CD (5.2.7), the answer is
"CHS".
>     *) with respect to OpenELEC (3.2.3, I think), the answer is
"CHS".
H. Peter Anvin:
> OK, it is "CHS"... which most likely means this is a geometry
problem.
But i can give some background info:

CHS means Cylinder/Head/Sector addressing method.

The Master Boot Record (MBR) at the beginning of the stick has
a partition table with block addresses in two alternative forms.
A CHS triple of three numbers where up to 63 sectors are one head,
and up to 255 (or 256) heads are one cylinder. Alltogether 24 bit,
which can address at most 8 Gigabyte, but often less.
Then there is Logical Block Address (LBA), a linear counting of
blocks with 32 bit length. This allows to address 2 Terabyte.

There are FAT filesystems which operate via CHS addresses,
and such which use LBA. (Visualized by Gparted's "LBA flag",
i assume.)
The successful isohybrid images cannot be used via CHS.

The current best guess is that your mainboard does not like
some or all variants of CHS but works with LBA.
Whether your stick's FAT uses CHS or LBA is decided when you
partition the stick and format the empty FAT filesystem.

H. Peter Anvin:
> find the "magic
> geometry" that works on that machine, which may be 64/32 for example.
> However, in CHS mode 1024x64x32 is only 1 GiB,
So Peter proposes to try 32 sectors per head and 64 heads per
cylinder when you partiton the USB stick. The partitioning
program should offer you an opportunity to set these factors.

Maybe it suffices to format the partitioned stick to a FAT
filesystem variant which uses LBA.

(As said, i am no expert in partitioning, formatting and FATing.)


Have a nice day :)

Thomas

> 
> Ady:
> > I also "forced" a FAT32 LBA filesystem,
> Ronald F. Guilmette:
> >     *) with respect to Ultimate Boot CD (5.2.7), the answer is
"CHS".
> >     *) with respect to OpenELEC (3.2.3, I think), the answer is
"CHS".
> H. Peter Anvin:
> > OK, it is "CHS"... which most likely means this is a
geometry problem.
> 
 
Just to clarify, the two "CHS" messages were seen when booting with 
the initially-failing USB drives (and my guess is that the third USB 
drive that by now has been overwritten was also in "CHS" mode).

_Later_ tests, including the ones with my test.img, are not using CHS 
mode and, more importantly, they are successfully booting the system.

> The current best guess is that your mainboard does not like
> some or all variants of CHS but works with LBA.
> Whether your stick's FAT uses CHS or LBA is decided when you
> partition the stick and format the empty FAT filesystem.
> 
 
There are several reasons for the initial failure to boot; one of the 
possible ones being some potential mix-up and incompatibility with 
some CHS values.
> H. Peter Anvin:
> > find the "magic
> > geometry" that works on that machine, which may be 64/32 for
example.
> > However, in CHS mode 1024x64x32 is only 1 GiB,
> 
> So Peter proposes to try 32 sectors per head and 64 heads per
> cylinder when you partiton the USB stick. The partitioning
> program should offer you an opportunity to set these factors.
> 
> Maybe it suffices to format the partitioned stick to a FAT
> filesystem variant which uses LBA.
> 
 
Let's try to avoid special or specific CHS values. Giving the recent 
successful tests, they seem to be not necessary. We are already 
seeing success when using adequate (and commonly-used) Nx255x63 
values in the MBR, in the VBR, formatting as FAT32-LBA with starting 
offset of 2048 sectors. I would suggest keeping this format.

Regards,
Ady.

In message <73645861681468542 at scdbackup.webframe.org>, 
"Thomas Schmitt" <scdbackup at gmx.net> wrote:
>me:
>> > Check whether it fails with Clonezilla.
>Ronald F. Guilmette:
>> Please be specific.  For Clonezilla, there are many different versions
>
>Any failure on the same stick which also succeeds is enough
>to kill the theory of bad relation between USB stick and
>mainboard.
>You have proven this by the success and failure with the 2 GB
>stick.
OK.  I believe that I understand what you just said.  (Actually, I for
one never even contemplated the possibility of any sort of bad inter-
action between a specific USB stick and a specific motherboard.  Do
such instances ever actually crop up, in practice?  I would guess
not.)
>So the problem must be either in partitioning or in the filesystem
>format. That's out of my personal expertise, though.
Mine too.
>> I await your direction.
>
>I lack the skills for explicit directions how to explore
>the plausible suspicions of Ady and Peter:
Very well.  I await detailed instructions from one or both of them.
>CHS means Cylinder/Head/Sector addressing method.
I knew that, actually.
>The Master Boot Record (MBR) at the beginning of the stick has
>a partition table with block addresses in two alternative forms.
Hummm... so the world has not entirely done away with the kludgy,
antiquated and anachronistic old C/H/S addressing scheme just yet,
eh?

Too bad.  We should be rid of it once and for all.  (As I understand
it, it has been AT LEAST more than a decade and a half... or maybe
two... since the notion of physical/addressable cylinders had any
relevance to any actual products on the market.)
>There are FAT filesystems which operate via CHS addresses,
Still??  Who uses them?  Why do they continue to use them?

(I guess that there will always be some hold-out who is still running
MS-DOS 5.0 or Windows 3.1.  But I don't think that any of the rest of
us should provide any support or encouragement to them.)
>So Peter proposes to try 32 sectors per head and 64 heads per...
If Peter desires me to run another test, he will need to give me
detailed instructions on what exactly he wants done. I will then
attempt to follow those instructions to the best of my ability, but
my hope is that I will only need to fetch files, dd them to sticks,
and then try booting them.


Regards,
rfg