XP dislikes MiB alignment

[pointertovoid]

A tired hello to everybody...

After many failures, trials and angryness, it seems that:

• The XP installation doesn't work on a disk that has been formatted with its volumes aligned on MiB boundaries.
• But W2k installs and repairs under the same conditions when XP is absent.
• Adding XP on a preexisting MiB-aligned volume prevents W2k already installed on other volumes to boot.
• Once the XP installer has bricked something, the W2k installer can't repair it easily (or I didn't find how).
• It's even possible that XP or its installer damages the MiB-aligned volumes.
• I didn't try if only the boot volume has alignment conditions, nor if it's a matter of alignment or of distance to the disk's beginning.

Conditions of observation:

• X25-E 64GB disks on ich10r on Ga-ep45-ud3r mobo.
• Ahci mode. Intel XP drivers v8.7, BWC's improved drivers for W2k.
• ich10r drivers provided by F6 floppy mostly, but sometimes nLited in an install CD.
• Tried original (pressed) CD, improved burnt CD and DVD, several diskettes that had worked the day before.
• Partition table and volume formatting by GPartEd 0.8.0.5, but also 0.12.0.2, 0.5.1.1 and by the XP install disk.
• Four volumes of 10GB, 10GB, 10GB and 34GB. All volumes were main if created by GPartEd - Windows does it differently.

XP installation succeeded when creating all the volumes by

• The XP install CD
• GPartEd 0.5.1.1 with the option "align on cylinders" and zero (0) MB before
• Even as the (MS-DOS) partition table had been created by GPartEd 0.8.0.5

It failed when creating all the volumes

• By GPartEd 0.8.0.5 with the option "align on MiB" and one (1) MB before
• By GPartEd 0.12.0.2 with the option "align on MiB" and one (1) MB before.
  Note that 0.12.0.2 can align on cylinders but imposes 1MB before while 0.5.1.1 permits 0MB.
• Even if booting on the install CD, the XP part already on the HDD can't start.

<rant>
W2k didn't fail on that, but the more recent XP even bricks what W2k does well.
Trying several disks, diskettes, Windows and CDs, formatting tools, checking the Ram, Cpu, host, disks... took me one day, and I sacrificed an installation that had cost me an other day on an other disk to check if the hardware was sound.
</rant>

Maybe this serves to someone else, at least.

[jaclaz]

Well, no offence intended, but you provided no actual report that you checked the results of using either GPartEd 0.8.0.5 or GPartEd 0.12.0.2 "properly" or that the programs actually produced "proper" partitionings.

Xp (actually disk manager) will make a mess of any disk where there are Logical volumes aligned to Mb (this is known), it is very possible that the same happens during install, so you should provide a copy of the MBR partition table (aligned to the cylinder) and of the MBR partition table (aligned to the Mb) or the exact partitioning scheme you used.

I never saw any problem with primary partitions, so my guess is that in your setup one or more logical volumes inside Extended are involved.

The extended partition/logical volume issue may also happen post install as just changing the active status of any primary partition will botch all logical volumes inside extended, JFYI:
http://reboot.pro/topic/9897-vistawin7-versus-xp-partitioning-issue/

http://www.dcr.net/~w-clayton/Vista/DisappearingPartitions/DisappearingPartitions.htm
 

There are also other factors that may come into play depending on the SP level of the XP (big LBA), but I don't think this is the case.

jaclaz
 

[HarryTri]

I have installed Windows XP on a MiB aligned VHD on Virtual Box and then applied the SP2 and SP3 with no problem. I don't know if the fact that it was a virtual machine makes any difference.

[dencorso]

No. But the fact you surely installed it in a primary partition does.

[jaclaz]

To clear the matter:
1) Primary partitions are UNaffected
2) ONLY Logical volumes inside Extended are affected[1]

In a multiboot system with both XP and Windows 2000 it is "standard" and actually historically advised to use logical volumes to store the operating systems, and this is very likely what the OP had.

jaclaz

[1] Technically it is due to the fact that disk manager *somehow* has hardcoded 62 sectors gap between EMBR and EVBR of each volume, and even when something unrelated (like changing the active status of a primary partition) it automagically re-calculates addresses, thus breaking the EMBR chain and making ALL logical volumes incorrectly mapped or not mapped at all.
 

[pointertovoid]

Unfortunately I didn't take detailed information of the partition schemes (...1 full day of reboot, reinstall). What I know for sure:

• The Xp install Cd makes, on an empty disk, one primary volume and one secondary volume with logical volumes in it (3 in my case) - much like the Ms-Dos diskettes and Win95 boot diskettes do.
• With every version of GPartEd, I defined everytime 4 primary volumes, with "boot" flag on the lowest one defined first. Definitely sure.
• The physical disk (X25-E) is 64GB big, so the big Lba shouldn't hopefully interfere.

And my uncertain hypothesis relying on a vague supposition: that the Xp install Cd makes assumptions that do not belong to the partition standard, while W2k and GPartEd stick to the flexibility of the standard. For instance, that nothing imposes the sector 63.

[jaclaz]

3 hours ago, pointertovoid said:

 What I know for sure:

• The Xp install Cd makes, on an empty disk, one primary volume and one secondary volume with logical volumes in it (3 in my case) - much like the Ms-Dos diskettes and Win95 boot diskettes do.

No, it doesn't normally.

Maybe you have a "custom" or "oem" install CD that works as you describe, but a normal XP install CD proposes to make a single, huge, primary partition, as large as possible, respecting the cylinder end "boundary".

And AFAICR the Win2K behaves exactly the same.

Whether the Win2K disk manager (and the subsystem beneath it) is "smarter" than the XP one I cannot say, but most probably it is .

GParted (but it may depend on specific version) defaults (if I remember correctly) to the "new" 1 Mb alignment.

jaclaz
 

[pointertovoid]

I should have told: one pimary and one extended volumes because I defined four volumes.

On the first attempt, the partitions had been made by GPartEd, aligned on MiB I believe, and used succesfully by W2k for installations and months of use, but bricked by the Xp install disk - so I'd suppose that Xp's installation does something worse than W2k's installation, possibly not placing the boot sector where the preexisting partition tells.

[jaclaz]

Then you didn't have all primary partitions, as expected.

And - again - the issue is ONLY with logical volumes inside extended partition, and only if some "featrures" of the disk manager (and also possibly diskpart, but that would need to be tested) are used, (it is entirely possible that the same failed routine that rewrites wrongly the EPBR chain of logical volumes inside extended is used during install without the customer intervention).

I know it is confusing, but a partition can be primary OR extended.

There is NO such thing as an "extended volume", all volumes are "logical volumes".

A primary partition "contains" (actually "is") a logical volume.

There is NO "empty space" between the start address of the primary partition and the start of the logical volumes inside it, you may say that there is a 0 offset before the VBR (volume boot record, what is commony called bootsector) of the logical volume inside the primary partition, or, if you prefer, that the VBR is the first sector inside a primary partition.

An extended partition can contain one or more logical volumes.

The first sector of an extended partition is an EPBR, very similar to a MBR, containing partition address entries.

Of the four available "slots", only two are used, the first one points to the next logical volume inside the extended partition and the second points to the following next EPBR (if any), and the scheme continues for each added logical volume.

On a cylinder/head aligned partitioning the gap between the EMBR and the VBR is 62 sectors (the relative offset of the VBR is 63, i.e. 62+1 which is the EPBR).

On a Mb aligned partitioning the gap between the EMBR and the VBR is 2047 sectors (the relative offset of the VBR is 2048, i.e. 2047+1 which is the EPBR).

The defect in the Dsk Manager of Xp is that in some instances it "decides" to rewrite the 2048 value of the offset replacing it with 63, obviously making all logical volumes inside extended not anymore accessible/findable.

Primary partitions are UNaffected by this.

jaclaz

 

 

 

[pointertovoid]

I confirm I made only primary partitions with all versions of GPartEd. The logical volumes in a secondary partition were only made by the Windows installation Cd because I had no choice.

----------

New observations:

I made the same operation

• Four primary partitions by GPartEd, this time v12.0.2, asking for no alignment at all, but putting the first partition after 1MB for having no choice.
• Installed W2k on the first, then Xp on the second partition.
• But this time on a WD5000AAKX at an (half broken) ich7 rather than on an X25-E at an ich10r.
• This first edition of ich7 know no Ahci so I gave no driver through the F6 diskette process. Access at Sata/3000 speed with Pata commands and no Ncq.

and it worked without any difficulty. To my best knowledge, my ich10r is sound. I used the Intel driver v8.7.0.1007 for Xp, normally it shouldn't matter.

Edited October 23, 2016 by pointertovoid

[jaclaz]

I will try to re-type this very slowly.

Primary partitions are NOT affected by the bug with Cylinder vs. Mb alignment, ONLY logical volumes inside extended may be.

jaclaz

[pointertovoid]

Glad to hear that. But I confirm that the Xp install bugged with four primary partitions, and that I'm positive about that.
Maybe something else went wrong, but I'm certain about my observations.

[jaclaz]

Well, then it is another bug , different from the mentioned one, possibly limited to the install routine, or maybe triggered by having all four MBR partition table slots "full" with primaries .

Lots of people dual boot XP and Vista  or 7 using two primary partitions (one for each operating system) using the Mb alignment and something similar has never been reported AFAIK, and - on the other hand - MS has a dedicated tool to create a Mb aligned (actually a 4 Kb which is all that is needed, but however non-cylinder aligned) partition:
https://technet.microsoft.com/en-us/library/bb643096(v=exchg.80).aspx

jaclaz
 

[HarryTri]

On Δευτέρα, 24 Οκτωβρίου 2016 at 0:31 AM, pointertovoid said:

I confirm I made only primary partitions with all versions of GPartEd. The logical volumes in a secondary partition were only made by the Windows installation Cd because I had no choice.

You made four primary partitions? Then how the logical volumes were created since they can only be created into an extended partition? A new partition can't have been created, since in BIOS mode you can create four primary partitions or three primary partitions and an extended one but not more than that. Maybe the tools you used did something different than what you expected?
Also what do you mean by "I had no choice"? The Windows installation CD refused to install on any of the primary partitions (if it is possible)?
Sorry if I ask to much but I am a bit confused and I'm afraid that others who will read the topic will be confused too.

On Παρασκευή, 7 Οκτωβρίου 2016 at 1:06 PM, pointertovoid said:

The Xp install Cd makes, on an empty disk, one primary volume and one secondary volume with logical volumes in it (3 in my case) - much like the Ms-Dos diskettes and Win95 boot diskettes do.

That's the most confusing for me. You mean you can't just make a primary partition? Or more primary partitions one after the other in the non-allocated space of the disk (which I was able to do if I remember correctly)? What kind of Xp install Cd is this (an OEM or modified one)?

Edited October 29, 2016 by HarryTri

[pointertovoid]

On 29/10/2016 at 11:18 AM, jaclaz said:

Well, then it is another bug , different from the mentioned one, possibly limited to the install routine, or maybe triggered by having all four MBR partition table slots "full" with primaries .

While I'm sure about my observations, with a different disk on a different southbridge the same operation ran smoothly, so I don't know what to think.