Windows 8.1 on USB not booting

[edborg]

I start this topic to further explore the issue reported here:

 

I have installed three instances of Win8.x 32b to three primary partitions of an USB HD with the WinNTSetup 3.6.1 utility.

 

Whilst Win8 pro on partition1 works fine, Win8.1 Enterprise on partitions 2 and 3 do not.

Installations reported success but reboot looped to a blinking underscore.

 

All installations were made from the same laptop (2 GB RAM) from WinXP (the working one) and from WinXP and Win8 (the other two).

 

The USB HD is a Toshiba 1 TB with three primary partitions NTFS and a few logical volumes.

MBR and bootsectors of the primary partitions (read with HDHacker) are enclosed.

 

Apparently it is a bootsector problem, though only happens for the Win8.1 installations on partitions 2 and 3, not for Win8 (not even when it was installed on partitions 2 and 3).

Some attempts to rewrite the bootsectors with different versions of bootsect.exe have been made, to no avail.

 

Any suggestion is welcome.

 

edborg

[jaclaz]

 

 

MBR and bootsectors of the primary partitions (read with HDHacker) are enclosed.

 

Where?

 

How exactly you are testing (choosing) to boot form first, second or third partition?

 

jaclaz

[jaclaz]

I see, there was somehow a double post, the file is attached to his duplicated one:

http://www.msfn.org/board/topic/172145-windows-81-on-usb-not-booting/?p=1081140

 

I'll have  a look.

 

jaclaz

[Tripredacus]

Well it "was"

 

Does this work?

http://www.msfn.org/board/index.php?app=core&module=attach&section=attach&attach_id=38469

[jaclaz]

At first sight there are no issues, BUT there are two other things that may come into play.

 

1) is the computer BIOS fully LBA48 compatible on the USB bus?

2) is the actual HD external case an USB bridge fully compatible with LBA48?

 

Can you reset the first partition to "active" and add to it a grldr added through BOOT.INI?

http://reboot.pro/topic/19730-dmde-basic-disk-imaging-test-and-results/?p=184060

 

jaclaz

[edborg]

 

 

 

MBR and bootsectors of the primary partitions (read with HDHacker) are enclosed.

 

Where?

 

How exactly you are testing (choosing) to boot form first, second or third partition?

 

jaclaz

 

Forum links are broken 

I can no longer see my own post with attachments that were downloaded 

 

To answer your question, I open Computer Management / Disk Management in Windows and make partition active, then boot

 

edborg

[jaclaz]

To be fair, it's mainly your fault.

 

You posted this one at 12:17 Zulu and *somehow* this has no attachment, then (probably the board software or your connection "stalled") you re-posted it at 12:19 Zulu, this time with the attachment or however this time the attachment "went through".

 

However no problem I got it (and it is stil available in the link Tripredacus posted).

 

See if you can manage to boot to grub4dos (see previous post).

 

jaclaz

[edborg]

At first sight there are no issues, BUT there are two other things that may come into play.

 

1) is the computer BIOS fully LBA48 compatible on the USB bus?

2) is the actual HD external case an USB bridge fully compatible with LBA48?

 

Can you reset the first partition to "active" and add to it a grldr added through BOOT.INI?

http://reboot.pro/topic/19730-dmde-basic-disk-imaging-test-and-results/?p=184060

 

jaclaz

Don't know how to check those compatibilities (unless you guide me).

BIOS is ASUS 900 ACPI BIOS Revision 0501

HD is Toshiba USB 1 TB

 

Not sure I understand the purpose of your other question.

I've reset the first partition to active, added grldr, a menu.lst, a boot.ini to load grldr from bootmgr... and booted.

It boots fine to either Win8 (the working one on partition1) or to the menu.lst of GRUB4DOS.

 

Is this helpful?

 

edborg

[edborg]

To be fair, it's mainly your fault.

 

You posted this one at 12:17 Zulu and *somehow* this has no attachment, then (probably the board software or your connection "stalled") you re-posted it at 12:19 Zulu, this time with the attachment or however this time the attachment "went through".

 

However no problem I got it (and it is stil available in the link Tripredacus posted).

 

See if you can manage to boot to grub4dos (see previous post).

 

jaclaz

Well, to be fair, no.

I saw my post AND its attachment, and linked to it in the other thread.

Then all of a sudden it disappeared ... and reappeared later at a different address.

 

If you don't believe me look links in the other thread

 

May be the upload went wrong (not my fault  ) and LATER the board software cleaned up duplicated posts

Edited June 27, 2014 by edborg

[jaclaz]

Sure, the board software (in this case) is called "Tripredacus" :rofl:

 

 

 

Not sure I understand the purpose of your other question.

I've reset the first partition to active, added grldr, a menu.lst, a boot.ini to load grldr from bootmgr... and booted.

It boots fine to either Win8 (the working one on partition1) or to the menu.lst of GRUB4DOS.

 

Is this helpful?

Sure it is.

Boot.

Choose the grub4dos entry.

Press "c" to get to the grub4dos grub> prompt.

 

Start testing commands (the booted USB external disk should be hd0, i.e. first boot device, you can check this, once you are in grub4dos by simply running the root command without parameters):

root

[ENTER]

and the geometry command   

geometry

[ENTER]

Then try loading the install on second partition:

root (hd0,1)

[ENTER]

chainloader +1

[ENTER]

The above will chainload from grub4dos the bootsector of second partition.

What happens?

 

Try:

root (hd0,1)

[ENTER]

chainloader /bootmgr

[ENTER]

The above will attempt to boot the install on second partition bypassing the PBR of the second partition.

 

What happens?

 

Repeat with the third partition (that should be (hd0,2) ) and report.

 

jaclaz

Edited June 27, 2014 by jaclaz

[edborg]

Thanks for your help!

I can't do anything right now.

I shall let you know when done.

edborg

Sent from my phone using Tapatalk

[Tripredacus]

 

To be fair, it's mainly your fault.

 

You posted this one at 12:17 Zulu and *somehow* this has no attachment, then (probably the board software or your connection "stalled") you re-posted it at 12:19 Zulu, this time with the attachment or however this time the attachment "went through".

 

However no problem I got it (and it is stil available in the link Tripredacus posted).

 

See if you can manage to boot to grub4dos (see previous post).

 

jaclaz

Well, to be fair, no.

I saw my post AND its attachment, and linked to it in the other thread.

Then all of a sudden it disappeared ... and reappeared later at a different address.

 

If you don't believe me look links in the other thread

 

May be the upload went wrong (not my fault  ) and LATER the board software cleaned up duplicated posts

 

 

Your duplicate post (the one with the attachment) still exists. That was how I was able to find the link to post to where your attachment resides. It could still be possible that this thread could be merged into that other post, but it was unknown to me (at least) that when the duplicate topic issue happens, what would happen if an attachment were specified. Anyways, the reason why that particular duplicate topic was deleted (unknowing of its contents) was that *this* topic had replies and the other did not.

[edborg]

I did the tests and found inconsistencies in the geometry.

 

This made me recall that similar isssues (of which I had forgot) had already been discussed at lenght a few years ago here

 

At the time it was an incompatibility in partitioning between XP and Vista; this time it might again be a problem due to working/partitioning with different OSes (WinXP, Win8, Win8.1)

 

In any case, these are the results:

 

root

FileSystem type is NTFS, partition type 0x07

 

geometry

drive 0x80 (LBA): C/H/S=12161/255/63, Sector count/size=1953520065/512

Partition num: 0, active, FileSystem type is NTFS, partition type 0x07

Partition num: 1, FileSystem type is NTFS, partition type 0x07

Partition num: 2,

Fatal! Inconsistent data read from (0x80) xxxxxxxxxxxxxxx+2 (some 30 lines)

Error 25: Disk read error

 

root (hd0,1)

FileSystem type is NTFS, partition type 0x07

 

chainloader+1

prompt

 

chainloader /bootmgr

Fatal! Inconsistent data read from (0x80) 314568838+2

Error 16: Inconsisten filesystem structure

 

similarly for (hd0,2)

 

edborg

[edborg]

After refreshing my memory re-reading the above mentioned thread, I realize that inconsistencies and incompatibilities may occur when working with theoretically (at least) backward compatible OSes, even if used correctly.

 

During the various attempts to install Win8.1 on USB HD I may have used more than one OS and/or software to partition, format, write bootrecords, etc.

I can't recall the EXACT sequence of actions.

 

I have partitioned with XP Disk Management, formatted partitions, eliminated and recreated single partitions (without repartitioning the whole disk) checked/edited MBR/PBR with Bootice, edited PBR with various versions of bootsect.exe from the building utilities used (fujianabc, WinNTSetup), and even other more recent versions downloaded from newer WAIKs, etc. I have set partitions active from XP or Win8 Disk Management, etc.

This has always been done following a rational approach and not randomly just to try and see what happens, but may still have caused the geometry inconsistencies on my USB HD.

 

This issue has only emerged with Win8.1, not Win8, nor Win7 (even if my behaviour has always been consistent with the three of them), which had led me to think of a Win8.1 incompatibility, that however has been excluded.

 

At this point in time I think i will repartition the whole USB HD and start from scracth with Win8.1 to partition1 without installing other OSes first, to see whether such problems recur .

 

To reduce the risk of future problems I sum up my present understanding.

Please confirm or correct what stated below:

 

1. MBR is written by the partition utility (and this has never been a problem because the HD has always booted at least from one partition)

2. PBR is written by the parttion utility, changed by Bootice, changed by bootsect.exe (but only for the code part, not for the data part?); so data should remain as it was set when partitioning

3. Various bootsect.exe versions exist, that change the PBR differently (how?), but in backward compatible way, so that the most recent one should be the best choice (?). If so, WinNTSetup might update its bootsect to the 8.1 version

4. The code in bootsector (PBR) loads a bootloader/manager (bootmgr in case of Win8.x) that load the OS. Bootmgr is OS specific and requires the use of an OS specific bootsect.exe

5. Different ways of making a partition active should give the same result (?)

6. I preferably work from XP where I have the most utilities, ease of access and familiarity. However some utilities (say latest bootsect.exe) refuse to run from there, so I can't completely avoid mixing OSes.

7. What else?

 

edborg

Edited June 28, 2014 by edborg

[jaclaz]

Yes, but you should (could) try to pinpoint the actual issue.

When you return to the prompt here:

 

root (hd0,1)
FileSystem type is NTFS, partition type 0x07

chainloader+1
prompt

 

 

have you tried entering:

boot

[ENTER]

 

Explanation:

IF the issue is connected with LBA48:

First partition is entirely within 128 Gb and works

Second partition begins before the 128 Gb limit and may work nonetheless

Third partition is entirely after the 128 Gb limit and will not work

 

Before anything else, apart trying to boot from second partition with grub4dos as above, would be to simply delete the third partition (AND the following Extended) and try again.

The grub4dos geometry command output is more or less saying that the first and second partition are fine, but anything beyond them is an issue and this may affect the normal booting. 

 

Your next test should be to re-do from scratch, making the first two partitions BOTH within the 128 Gb and the third outside it.

And, this time, try to install one instance of the 8.1 to the first partition, the 8.0 to the second and the other instance of 8.1 on the third.

This might help to establish if there is a specific 8.1 incompatibility (which cannot possibly yet be ruled out) or if the issue is *something* else.

 

What we are trying to establish is whether there was an incompatibility of some kind in the mix of tools you used, so this time, please jolt down the exact steps (and tools) you use, the alternative being the mentioned issues with LBA48.

 

Over the years it has been observed that some motherboards, while having a BIOS that supports LBA48 sized internal disks alright, have a USB stack in the BIOS that tops at LBA28 addresses, and as well it has been observed that some external USB hard disk cases had a controller that was compatible only with LBA28 addresses.

 

Regarding your listed points number:

1. Yes, but still the MBR CODE may somehow be incompatible with 48 buit LBA addressing (extremely rare, but possible) or, just as an example (which does not apply in your case) you need a particular MBR code to boot a BitLocker volume, I believe.

2. Yes, any program changing the bootsector will change just the code but not the data.

3. The point of the thread I pointed you too http://www.msfn.org/board/topic/171749-bootsectexe-various-versions-compared/ was EXACTLY that of establishing WHICH version of bootsect.exe did WHAT, as you can see they are categorized and the latest (Win 8.1 one has NOTHING different from a few previous version if not the fact that it won't run under XP), i.e. using the "C" type or the "D" type has the SAME results.

4. Not really-really. The PBR code is simply an hardcoded way to load BOOTMGR, grub4dos can chainload it bypassing the PBR, even if there are some differences in the PBR code, usually any PBR code invoking BOOTMGR will do.

5. NO, this is exactly what was found several years ago http://reboot.pro/topic/9897-vistawin7-versus-xp-partitioning-issue/#entry85947 what is needed to set a Primary partition active consists of writing a 0x80 to a byte in the corresponding entry in the partition table in the MBR and make sure that all other entries have the corresponding bytes set to 0x00 but for whatever reasons, the XP disk manager re-checks and "corrects" entries in the Extended Partition (and/or in its EMBR's), in that occasion Primary partitions were NOT affected AND that hapened only if the disk was "Mb aligned", this one is seemingly "Cylinder aligned" (unless this was the result of your "mixing" partitioning tools). So *any* MBR/partitioning tool may do something else besides changing the 0x80's and 0x00's and this is the reason why at the time it was all in all suggested that you do NOT use the XP manager to change the active status of a parititon.

6.Re-read point #3 and given link attentively

7. N/A

 

You will need some patience, and 

 

jaclaz