I am desperately trying to find a way to install Windows XP on a USB Hard Drive, in such a manner that it will be able to boot up MANY computers, rather than just one. Even better, I would like to make it so I could even copy this installation onto a second, third, forth, etc, USB HD as backups, and that these backups could then also be used, if needed to boot up MANY computers.

The purpose of this would be so that if my computer crashes, it would be no major inconvenience, because I could just connect my USB drive to another computer, and continue whatever projects I was working on, without having to waste a lot of time. Also, if it was my USB drive that crashed, then that would also be no major inconvenience, because I could just go to my most recent USB backup and continue working from there.

I know that there are those on this board with the kind of expertize to do this, but I am not one of them. I am trying REALLY hard, though, to figure this out as best I can. In my mind, this would be pure OS nirvanna!

So is there ANYONE on here willing to take this on, or anyone, like me, who thinks this would be FANTASTIC to have?

yeah, think thats FANTASTIC

Good luck!

PS, are you the guy that invented the wheel as well?

If im not mistaken (which I am most often ) I believe this very topic has been often touched on a ton of time.

I would point out trying the search function on this board as I KNOW its been discussed.

Here is one post I would start at.
Here is another!
Here is the Busy'st one I think!

MrXP

besides... while convenient, it is VERY slow. You'd be better off using removable HDs or basic portable apps on a thumb drive.


In a corporate environment, a PC can be imaged in 10 minutes or so

I think its too hard a job myself, ive settled for something easier:

Searching for the actual Holy Grail


Some people want to make things harder than they have to be.........

After many days of research, despair, trial and error, I finally succeeded in building both a full WinXP and a BartPE (+XPE) on three different USB supports, tested on three different computers.
It doesn't mean at all that it works on all USB supports, nor all computers, but still it's encouraging.

This I did only thanks to the great guidance and help from experts like Dietmar, sisal, jaclaz and others contributing in Dietmar's thread.
That thread, which includes at least four different approaches with many variants, representing different attempts to solve various Bios and hardware problems, is now so full of information (47 pages today), that it's quite difficult for a newcomer, and also for who has read it all, to extract what is needed to get to an overall conclusion.

To this end, I was trying to divide the problem into pieces (USB format, BIOS recognition, USB reset from Windows, USB drivers, CriticalDeviceDataBase, and so on) and get an understanding of what the key (un)success factors in each stage of the XP USB booting process are.

Starting from USB format, I've followed indications in the thread and tried a number of tools, most of which are proprietary and only work on their own supports. Amongst those of general use these seem the most appropriate:

MS Format + MKBT by Bart (this one worked fine on my USB sticks/disk)
HPUSBF by HP
MakeBootFat by Andrea Mazzoleni (tutorial by jaclaz)

Although they all work, they give different results on different sticks. It would be useful to understand in what they differ to take the best out of them and put together an "universal" XP formatting approach.
I do not have myself the necessary level of expertise, but others have it, joined to the capability of clarifying complex things (jaclaz are you there?).
May be others reading this thread might be ready to experiment.
edborg

P.S. Suggestions/contributions on the other "pieces" of the problem are welcomed as well

It's not necessarily VERY slow; at least in my experience, about the same as HD.

Yep, PRESENT!

OK, following the results of checking the error that Dietmar had in the other thread about the different bootsectors, I already posted my new finding about USB formatting (pardon me the BOLD CAPITAL):
IF WIN2K/XP FINDS A VALID MBR ON A MASS STORAGE DEVICE, IT "TRUSTS" THIS MBR'S CONTENTS AND FORMATS ACCORDINGLY!

So (in the case of one big partition, as usual on USB sticks) all you need is a valid MBR, then you can format the stick "normally" under disk management, of course the boot sector will invoke NTLDR.

I already posted a (actually rather complex) howto here:
http://www.911cd.net/forums//index.php?sho...4181&st=894
The (good) news is that I wrote a small batch file to "create" an MBR for a stick from the one already on your Hard Disk(s) or on a HD image, this effectively gets rid of the "problems" with non english versions of Win2K/XP; get it here:

http://home.graffiti.net/jaclaz:graffiti.n...BXP/MakeMBR.zip
It needs the DSFOK package, downloadable from here:
http://members.ozemail.com.au/~nulifetv/freezip/freeware/

(you need just the dsfo.exe and dsfi.exe files from the package in the same directory of the MakeMBR.cmd and USBstick.dat)

No instruction is needed, it should be self-explanatory, only (rather obvious) need is to unplug and re-plug USB device before attempting formatting with Disk Management.

With reference to the .xls worksheet referenced in the above said post, and downloadable here:
http://home.graffiti.net/jaclaz:graffiti.n.../PTtableUSB.zip
I put for the moment just the "safe" table of formats, not the MAXimum capacity one.

The settings have already been tested by me for sizes 128MB and 512MB, theoretically it should work for all other sizes.

As far as I could see there are no problems in using it, just to make it a little bit safe, writing a MBR to \\.\PHYSICALDRIVE0 (the boot drive) is only possible with command line parameters and an additional "Force0" parameter.

If anyone is willing to test it for:
256MB
1GB
2GB
and report, I would be grateful and go on making another .dat file with the MAXimum settings.

As always, bug reports, suggestions, comments and whatever are welcome.

jaclaz

P.S.: Don't forget the "Careware" License that applies to ALL my small files :

Thanks for replying to the call so quickly ... and for the solution already worked out!
I had missed your previous post , but I'll have a look at it in detail asap.

EDIT
I couldn't wait experimenting!
I've tested on a 256MB stick (USB-FDD), getting the MBR from the HD and from another 1GB stick (USB-HDD).
The CAREWARE script works well and it's quite straightforward, saving a lot of time in comparison with the lengthy manual procedure (which I did not test yet).
The PTtableUSB.xls, even if it's made no longer necessary by the script, gives a lot of info to "those who want to know" everything of a pendrive.
Congratulations!
edborg
P.S. With the script is it possible to write any working MBR, regardless the type/size/partitions of the source to any target?
PP.SS. My "256MB" stick is still seen by Windows as 247MB, 259739648 byte

Sorry, Edborg, in the haste of uploding, I uploaded a "test" version with a "jump" to :end NOT REMmed out.

All you saw was just what the program was "supposed" to do, but it actually did NOTHING!

I have now uploaded the "working" version.



Yep, what it actually does is:
1)get first 440 bytes from beginning of Source
2)extract from USBstick.dat a "padding" file formed by 6 bytes set to 00
3)extract from USBstick.dat the "chosen size" partition entry (16 bytes)
4)extract from USBstick.dat a "padding" file formed by 16 bytes set to 00
5)extract from USBstick.dat a "Magic signature" file formed by 2 bytes 55 AA
copy 1)+2)+3)+4)+4)+4)+5) to a single 512 bytes file
copy the 512 bytes file to beginning of Destination

jaclaz

P.S.: The USBstick.dat can be made manually/edited, I only put an MD5 signature to the USBstick.dat as to make sure that the data file is not "unwantingly" corrupted
The volume inside an image made of 256,000 Kbytes, aka 262,144,000 bytes results using my batch (with windows explorer right click -> properties, after formatting):
FAT16 254,685,184 bytes, 0 bytes occupied, 254,685,184 free
FAT32 253,941,760 bytes, 2048 bytes occupied, 253,939,712 free
NTFS 254,950,912 bytes, 3,064,320 bytes occupied, 251,886,592

Using the MAXimum capacity I obtain:
FAT16 261,836,800 bytes, 0 bytes occupied, 261,836,800 free
FAT32 261,074,944 bytes, 2048 bytes occupied, 261,072,896 free
NTFS 262,111,232 bytes, 3,131,392 bytes occupied, 258,979,840 free

By using a "fictional" 32 sectors geometry, making partition start at CHS 0/0/33 instead of the 0/1/1 of the 63 one you can recover theoretically yet 31*512 bytes =15,872 bytes,
but probably due to some other constraints, results are:
FAT16 261,869,568 bytes, 0 bytes occupied, 261,869,568 free
FAT32 261,105,664 bytes, 2048 bytes occupied, 261,103,616 free
NTFS 262,142,976 bytes, 3,131,392 bytes occupied, 259,011,584 free

I re-downloaded the full version, and it works fine ... even when it writes to the stick!
The reason why the "test" version seemed to work as well is that the utility doesn't format the stick and doesn't destroy any data, so the stick worked ... as before!

Unfortunately my satisfaction ended very soon, as another problem appeared on my 256MB pendrive, although this time it may have nothing to do with the format, but probably with subsequent steps of the booting process.
These are the facts:
After running MakeMBR.cmd I copied the XP system files: boot.ini, ntldr, and (Dietmar's) ntdetect.com.
I then booted and got the error "INF txtsetup.sif ... missing" which I interpreted positively, as it was the same message I got for another stick that, after copying the BartPE files, booted regularly.
But in this case, after copying the BartPE files I got another (new) error message "File \minint\system32\biosinfo.inf could not be loaded. The error code is 1...".
Needless to say that the BartPE build was OK and required file was there.
Any ideas?
edborg

I suspect that that error is more due to motherboard than to stick.
If you search on the board for
\minint\system32\biosinfo.inf

you will find a few posts of people having this kind of problem with motherboards with Intel 845 or 865 chips.

However from what you write it seems like you used a "setupldr" renamed to "NTLDR" as per BartPE install.

The "missing txtsetup.sif" message, like seeing the boot.ini choices, if you use a "real" NTLDR (the one that reads "boot.ini") should mean that the MBR/Boot record part of the process is OK.

jaclaz

My txtsetup.sif dosn't contain anymore. Setupldr.bin dosn't seek biosinfo.inf anymore.

BartPE boot from USB hard disk \minint mode.

Besides building PE, listening to music and browsing sites I trust, I find it a good idea to use vmware for everything else. Especially using BartPE. It's a great way to do some research without getting killed in a drive-by download. x_X'

Most modern computers are like that. Hell Longhorn is like that on this computer! It's faaa-st! o.o'

My server isn't like that. The box connected to this one is a legacy Super-7 with an ATA-33 IDE controller and a mix of AGP/PCI/ISA peripherals. I'm just lucky that I didn't get stuck with some crap that can't even boot a CD. It was what I started with in 2000 and I had to fall back to using this slow thing a few months after building my P4 because it liked killing weak 4xx watt PSUs. It takes 45 minutes to install Windows XP on that machine and 60 minutes to install Windows Server 2003 Enterprise. Imagine having that to rely on and it suddenly gets nuked for God knows what reason in the middle of working on a report.

EDIT: Without a recovery solution! o_O'

Mine does (built with new PEbuilder 3.1.9)! What's wrong with it?
Could you pls explain?
Thanks
edborg

Small correction might be necessary AFTER using the makeMBR.cmd and Windows formatting.
At least on one system the 128 Mb key did not boot.

To make it boot I had to edit (with Beeblebrox) the Partition table and change Partition type from 06 (BIGFAT) to 0E (FAT16 LBA mapped).

It's however a minor problem that can be "automated" as well, once it is found to be recurring.


jaclaz

But how comes that other sticks/disk boot regularly on this notebook (an Intel 852/855 chipset)?
Right!

This is good news! I then did some more tests:
NTLDR BartPE --------> biosinfo.inf error
NTLDR standard ------> boot.ini menu OK --------> hal.dll corrupted error (no more biosinfo.inf!)
Does this mean something?
edborg
P.S. When you're fed up with replying to all my questions, just say so!

I had already made some tests changing from ID06 to ID0E, but noticed no difference in the error messages.

This also suprises me, because from previous tests conducted on two sticks with MS Format + mkbt (ID06) and HP Format (ID0E) I had got different results on two sticks (Seitec 256MB and Verbatim 1GB), that led me to believe that the partition type was fundamental!

With MS Format (ID06): "disk error" on Seitec (OK on Verbatim)
With HP Format (ID0E): OK, but subsequent error as described, on Seitec (not tested on Verbatim).
Apparently something else, in addition to the partition type, makes the difference.
edborg
EDIT: I realize now that one thing I din't do exactly as per your instructions. I didn't re-format the stick, as I thought that changing the partition type, or even the entire MBR wouldn't affect format.

@edborg

I opened file USB:\i386\txtsetup.sif and deleted both lines.

Well, the idea was to start from a stick that has been ENTIRELY "wiped".

Simple method:

fsz 256MBstick.img 262144000
(the above will create a file size 262,144,000 bytes filled with 0 bytes)
dsfi \\.\PHYSICALDRIVEx 0 0 256MBstick.img
(the above will copy the said image to the x drive make sure you use the RIGHT number



Yes it means that the stick boots ALLRIGHT.

Actually the error with the standard NTLDR should be something like "missing file" in [windows directory]\system32\ntoskernel.exe
IF the NTDETECT.COM is present in root directory
and something like "NTDETECT failed" if it is not present.

Try the following to make sure:
1)Copy to the stick ONLY boot.ini and NTLDR
2)try booting from it, error should be "Ntdetect failed"
3)add to the stick NTDETECT.COM
4)error should be the NTOSKERNEL one

The error you are having usually is attributed to a wrong entry in boot.ini
http://www.kellys-korner-xp.com/xp_haldll_missing.htm

make sure yours is


jaclaz

I did it too and, as "Setupldr.bin dosn't seek biosinfo.inf anymore", got no biosinfo.inf error.
But I got instead: "The entry "ntoskrnl.exe" in the [SourceDisksFiles] sextion of the INF file is corrupt or missing..."
Should I now delete Ntoskrnl as well? ... and so on with the other files?

@jaclaz
THANKS FOR YOUR PATIENCE!
I followed exactly your instructions:
1. wiped the stick with fsz/dsfi
2. checked that it had been filled with 00
3. copied only your boot.ini and standard ntldr from XP SP1 (changed WINNT to WINDOWS)
4. booted ... and got immediate endless loops (no message at all)
5. added standard ntdetect.com (from XP SP1)
6. rebooted and got: "Windows could not start because the following file is missing or corrupt: <Windows root>\system32\hal.dll. Please re-install a copy of the above file."
I then made a few other combinations with Dietmar's ntdetect, and found no difference.

At this point, in my view the responsible of the failure must be this stick (my other stick and disk work both perfectly).
They had both been formatted with MS Format + MBKT (using a script I created modifying Bart's pe2usb.cmd)
As I don't dare to put them at risk with further experiments , I think I'll have to give up now.
edborg

I sincerely hope that BETWEEN 2. and 3. you:
2a) ran the makeMBR.CMD
2b) unplugged the stick
2c) re-plugged the stick
2d) double clicked on it and, when Windows said "Unreadable disk, do you want to format it?" chose to do so, formatting as fat 16



However if you get till there, it should mean that BOTH MBR and bootsector are OK.
(but I quite frankly doubt that the problem is the stick itself)
It is possible that something in "my" method is eluding us.

Did you try the method you succesfully used on your OTHER stick on THIS stick?
It would be interesting to know if there is any difference.

jaclaz

Hi all,

I do not know, whether this is important but:

I build by hand a MBR for my Buffalo 2Gbyte USB stick.

It booted fine and Xp installed itself on it without any changes of that MBR or other problems via F6 methode.

This USB stick works on an AMD (Siemens D1607) and on my Shuttle Barebone SB61G2 PC.

And now comes that, I do not understand: On my Asus P4C800-E Deluxe motherboard is
now "NTLDR not found" shown.
So this message must belong in this special case simple to the Bios of the Asus motherboard.
If I use the HP-Format utility, this message is there not be shown (?!) but the MBR has a wrong CHS structur and you cant install XP on it.

Nice to hear from you
Dietmar

Yep, the problem "should be" this:
1) BIOS can read the disk as CHS or LBA (it depends on motherboard/BIOSes, not on sticks)
2) Windows 2k/XP will ALWAYS get the geometry of the stick as CHS (but actually uses LBA) with 255 Heads and 63 Sectors
3) BIOS (to be more exact DOS/Win9x that "relies" on BIOS for hard disk information), tend to use the "elder" geometry of 16 Heads 63 Sectors

So, in a nutshell, a "correctly" formatted stick, meaning with 255H/63S, geometry while being "good" for Windows 2K/XP is NOT good for all BIOSes.

If the BIOS uses LBA access, the CHS is alltogether useless and the key will boot OK.

If it uses CHS access, it will not boot correctly, and you have to use the "trick" referred to in my pages, to use a CHS that is correspondent to the geometry the BIOS uses and an LBA that is correct even with the geometry Win2k/XP "likes".

It is also possible that sticks up to a certain size (probably 512Mb or 1GB) are seen by BIOS in a "different" way than smaller ones.

Most probably the SETUP of XP finds this and aborts the install.

@edborg
The geometry for a 256MB stick (as seen in beeblebrox) "should" be:
255H/63S
0-06-80-0-1-1-30-254-63-63-497952

16H/63S
0-06-80-0-1-1-494-0-63-63-497952

jaclaz

Hi jaclaz,

is the first behavior with the CHS problem, that you described, the same for USB sticks and USB harddisks?

The second behavior, that large USB sticks are shown to most Bioses in a different way than small ones, happens not to USB harddisks.
I tested that with a lot of USB harddisks and computers.

One thing is another on a USB harddisk: You can choose LBA detection on a lot of Bioses, because a harddisk is shown not as a removable device.

Nice to hear from you
Dietmar

I'm still waiting on a method that can allow a PC to boot a external usb hard drive even if the BIOS doesn't support USB booting. To boot up BartPE/WinPE from a external drive enclosure or even a thumb drive.

MrXP

The idea of using VMWare has popped up a number of times in relation to the idea of installing WindowXP on a usb drive in a way which would allow the USB HD to boot up on many different computers. I'm not sure if this would work, but here's the idea as I understand it:

First install some version of Linux on the USB HD

Then install and run VMWare for Linux

Then install WindowsXP on the Virtual Machine

Does anyone know if this would allow that USB HD to boot up on many other computers, and also, would the WindowsXP that was installed in the Virtual Machine ALSO boot up (from the Linux on that USB HD) on other machines as well? Also, would the WindowsXP in this case be able to use various peripherals for which there are no Linux Drivers (such as most USB modems, etc)?

If this would work, then this seems like one of the best and easiest solutions to the problem.
Does anyone know if it would?

Yes, of course! I only mentioned in detail what I did before and after formatting the stick.

Yes, I did it since the beginning! It's only after failing with THIS stick (Seitec) that I started looking for other methods.
As already posted above, the utility which gave me good results with my Verbatim 1GB stick is Microsoft Format + MKBT from Bart (the one used by pe2usb.cmd when installing BartPE on USB).
The very same method resulted in "Disk error. Press any key to restart" with Seitec.

If I use on Seitec the HP Format utility (not tested on Verbatim), I get no disk error, and booting starts, but it ends very soon with the already reported errors.
This happens on my Asus notebook and might be similar to what Dietmar mentions above.
To find out whether the mainboard/bios is (also) responsible, I could repeat exactly the same experiments on my desktop when I have the time.
It might help understand once and for all at which level (motherboard/bios/stick/utility) problems arise.

edborg
P.S. This is only a matter of curiosity, as I don't need that stick anymore, but simply want to understand.
So, if it's of interest to others, I won't give up!

@Dietmar
To help better understand the booting process, could you explain a bit (in concept) what your modified NTDETECT.COM does, as compared to the standard one?
Thanks

edborg

@JVision
Thanks for initiating this thread.
I'm not able to reply to your last question, but...
In addition to becoming crazy to try and make all my (and other people's) USB sticks/disks bootable (whith some problems, as you can see from these posts) , I also have a couple of sticks/disks that work!
They work on all the computers where I've tested them, including those which give problems with the other stick.
There are only three for now (not so many), but I plan to extend the test to other notebooks in the near future and post results.
In those cases when the booting problems are solved there might be other problems due to the different hardware (CriticalDeviceDatabase), another area where Dietmar has already done a lot of research.

edborg

Hi Edborg,

I have been warned by two people, not to document the Hexcode of an original Microsoft driver.
Even I am not sure, whether I understand all, whats going on at boottime.
But if you know IDA pro...

Those two people do not belong to Microsoft.

And I believe, that this is a very good warning:
In America sits now a man in prison for 2 years, because he did something like.

Nice to hear from you
Dietmar

Well, as I see it, (and I mught be wrong, mind you) the problem is related to the so-called Hard-Disk "barriers":
http://www.storagereview.com/guide2000/ref.../bios/size.html

Shortly:
1) "Original" BIOSes used "pure" CHS addressing that posed a barrier at 528 Mbytes
2) Soon was adopted the ECHS addressing that shifted that limit to 2,11 Gbytes
3) Later LBA was adopted that shifted the barrier beyoind actual limits

I did not try (I will as soon as I have time, following your hint) to see if anything is different in BIOS "sensing" with the drive as "Fixed" or "Removable".

However the vast range of reports available on this board and on others lead me to think that more or less EVERY PC/BIOS manufacturer solved the problem in a different way.

So, if I am right, ANYTHING with a size bigger than 2,11 Gbytes HAS to be addressed as LBA, whilst anything with size below 528 Mbytes is probably seen as CHS, and ANYTHING between those limits can be seen as either:
1) ECHS mode:
http://www.storagereview.com/guide2000/ref.../modesECHS.html
2) LBA or LBA assist mode:
http://www.dewassoc.com/kbase/hard_drives/...mitations_2.htm
Particularly the LBA assist mode has NOT been implemented by manufacturers in a coherent or consant way.



ONCE the Win2k/XP drivers take control on the device, LBA is used, though for certain operations, CHS has still a role, at least for smaller sizes.

As an example after I use the said MakeMBR.cmd on a "real" USB stick, the formatting is done on a Nx255x63 geometry, whilst if I do the same on a disk IMAGE mounted by Ken Kato's VDK driver (which by default assigns a geometry of Nx64x32) I have later to manually correct the geometry of the bootsector.

If I open the same IMAGE with Winhex, it will be assigned by defaikt a CHS Geometry of Nx128x63 (that can be changed manually.

To check if the same happens to hard disks, you would need to test, in a USB case, an elder 500 Mbytes hard disk, and maybe, the translation would be generated by the hard disk controller inside the external case and not by BIOS.


jaclaz

So how has this been managed then? I have read the posts, and I find nothing but instructions to make boot sectors! Please tell me if I am wrong... or if nobody has suceeded yet.

I know maybe this thread is drifting away...
....but lefty01 has posted here a nice USB stick Formatting tool that is the HP one stripped of unneded code:
http://www.911cd.net/forums//index.php?showtopic=16154


It seems like things are getting hot about this, have a look at this (of course the files must be on some BOOTABLE device, i.e. a floppy for machines that do not support USB booting):
http://64.124.13.3/hacks/USB_Boot_using_GRUB.html
and at this thread here:
http://www.911cd.net/forums//index.php?sho...l=grub&st=0

to the above you add info in this post by thibeact:
http://www.911cd.net/forums//index.php?sho...0806&st=228
and you should get a fairly "universal" stick/disk

jaclaz