Jump to content











Photo
- - - - -

Community OFA NVMe 1.5 Storport Miniport for Windows Server 2003 R2 SP2

nvme ofa storport 2003

  • Please log in to reply
324 replies to this topic

#126 Dietmar

Dietmar

    Frequent Member

  • Advanced user
  • 236 posts
  •  
    Afghanistan

Posted 01 August 2018 - 07:21 PM

Hi dencorso,

when you want to write back the hive system to the XP.IMG,

first you have to mount it.

And at this point the problem appears, because this XP in XP.IMG has the same signatur and checksum.

Other USB sticks or harddrives you can connect, I tried.

The trick has to be, to put back hive system from nvme RD400 to XP.IMG without mounting it (or without underlying XP gets notice of that).

With Shadow Copy it can be done, but how without mounting?!

This would work

Dietmar

 

PS: You can copy the hive system back to the USB stick, where the XP.IMG stays on.

       But later you have to put it back into XP.IMG, mounted on other XP.



#127 Dietmar

Dietmar

    Frequent Member

  • Advanced user
  • 236 posts
  •  
    Afghanistan

Posted 01 August 2018 - 08:10 PM

Yessa, I got it :) !

I noticed, that the kicker USB stick without grub4dos is recogniced as removable device,

it gets in Mounted Devices in hive system on the nvme disk RD400

as device with D:\ another MBR signatur there, but keeps the signature from the harddisk in real,

as you can see with Winhex, really crazy, but then crashes always.

 

Now I did the following:

After boot I connect few min later the kicker USB stick again and shut it down at once after it gets letter D:\, before crash :lol: happens.

Then I copy from other XP this hive system from the RD 400 to the kicker USB stick.

And now it is really crazy: This kicker USB stick is not recogniced from RD400 disk,

it is just there from the beginning, with another MBR signature, but has in real the signature from RD400.

 

So, now Shadow Copy can work,

because the kicker USB stick is always mounted as D:\

Dietmar



#128 cdob

cdob

    Gold Member

  • Expert
  • 1449 posts

Posted 01 August 2018 - 08:12 PM

With the "new" RD400 the "Kansas City Shuffle"
is now rocksolid as before and very fast.

That's good news.
 

But it does not like, when the kickstarter on USB stick gets mounted.
I think, because XP does not know, how to handle 2 devices with the same signatur

How do you mount the kickstarter?
The MBR, signature and checksum is valid already. No need to mount the whole hard disk image.
Use imdisk and mount the volume only: MBR (and signature) is not included.

#129 Dietmar

Dietmar

    Frequent Member

  • Advanced user
  • 236 posts
  •  
    Afghanistan

Posted 01 August 2018 - 08:20 PM

Hi cdob,

the USB kickstarter has now no need to be mounted.

So funny, without Grub4dos it starts the nvme disk RD400,

than, because not usb friendly, loses contact,

but then it is not again recogniced by XP, it is just there,

because all its entries in registry are there.

This is really funny,

but works

Dietmar

 

kick.png



#130 Dietmar

Dietmar

    Frequent Member

  • Advanced user
  • 236 posts
  •  
    Afghanistan

Posted 01 August 2018 - 09:15 PM

I forget to tell,

that the MBR from the USB kicker stick and from the nvme harddisk

I make 100% identic. Dont know, if this has to be, but now XP has no chance to decide,

something is other than in its normal life!

Very funny, Aomei tells now, that the disks are identic but

XP shows its real size

Dietmar

 

PS: It is rockstable, survives reboot, remove USB stick, reconnect, extract from registry, just all.



#131 Dietmar

Dietmar

    Frequent Member

  • Advanced user
  • 236 posts
  •  
    Afghanistan

Posted 01 August 2018 - 11:37 PM

Now I have put everything on NTFS for Shadow Copy.

Still works

Dietmar



#132 Dietmar

Dietmar

    Frequent Member

  • Advanced user
  • 236 posts
  •  
    Afghanistan

Posted 02 August 2018 - 09:42 AM

Hi all,

until now I was not able to copy anything with Shadow Copy.

Strange, I look, whether it depends on Admin rights,

check everything 10 times.

It starts, build the shadow image from C:\ as B:\

but then hangs. For to understand, what is going on,

I will now set up a new quick NTFS XP, lasts only 5 min from USB stick.

Here is my batch file.

Something wrong there?

 

@echo off
dosdev.exe B: %1
dir B:\
robocopy.exe B:\WINDOWS\system32\config\system D:\WINDOWS\system32\config\system
pause

 

This I store as

 

clone_C2F.cmd  on C:\

 

In commandline on C:\   I start it with

 

vscsc.exe -exec=clone_C2F.cmd C:

 

The files vscsc.exe, dosdev.exe, robocopy.exe are in root C:\.

 

It depends not on "system" or that it is still there before on D:\ or if you use copy instead of robocopy.

I try to copy a simple *.txt file, always the same message "Looking for.." then comes the MBR name of old C:\.

hangs

 

Dietmar



#133 Wonko the Sane

Wonko the Sane

    The Finder

  • Advanced user
  • 15149 posts
  • Location:The Outside of the Asylum (gate is closed)
  •  
    Italy

Posted 02 August 2018 - 10:06 AM

A USB stick is normally seen as "removable" by the Windows XP.

As such the MountedDevices keys for the drive and drive letter in Registry use NOT the MBR signature, but rather a string like:

 

\??\STORAGE#RemovableMedia#7&201264be&0&RM#{53f5630d-b6bf-11d0-94f2-00a0c91efb8b}

so there is no direct conflict with the Disk Signature, but it is still possible that a whole range of programs and/or tools/subsystem still "see" the Disk Signature and use it.

 

:duff:

Wonko



#134 Dietmar

Dietmar

    Frequent Member

  • Advanced user
  • 236 posts
  •  
    Afghanistan

Posted 02 August 2018 - 12:21 PM

I changed after start the USB kickstarter against another USB stick.

This one gets the letter E:

But Shadow Copy hangs at exact the same place (query the shadow image..)

Dietmar



#135 Dietmar

Dietmar

    Frequent Member

  • Advanced user
  • 236 posts
  •  
    Afghanistan

Posted 02 August 2018 - 02:14 PM

On XP on NTFS harddisk this following batchfile works, stored as clone_C2F.cmd on C:\

 

@echo off
dosdev X: %1
robocopy X:\Windows\system32\config D:\Windows\system32\config system
DOSDEV /D X:
pause

 

Starting from commandline with

 

vscsc -exec=clone_C2F.cmd C:

 

Now I look, when nvme XP starts with grub4dos, if Shadow Copy there works

Dietmar



#136 Dietmar

Dietmar

    Frequent Member

  • Advanced user
  • 236 posts
  •  
    Afghanistan

Posted 02 August 2018 - 02:41 PM

And again the Sandisc Ultra 64Gb USB stick does not like to get a single file with 16 GB to be copied on it,

brrr..

Dietmar



#137 Dietmar

Dietmar

    Frequent Member

  • Advanced user
  • 236 posts
  •  
    Afghanistan

Posted 02 August 2018 - 03:13 PM

grub4dos does not help.

It stops exact at the same place:

Shadow Copy build new Shadow from C:\ and then asks for its own, new build shadow X:\ (with arcname),

there it hangs always

Dietmar



#138 Dietmar

Dietmar

    Frequent Member

  • Advanced user
  • 236 posts
  •  
    Afghanistan

Posted 02 August 2018 - 05:06 PM

Hi jaclaz,

you where right.

I copied the content of the nvme RD 400 to a real NTFS harddisk

and connect there the USB kickstarter disk.

Oh, I cannot open a single file on the USB kickstarter from the harddisk,

no matter in which slot it is!

Always is told 0 Bytes, you have no right to open it :P .

And when I do a Shadow Copy,

it answers: The folder Windows is blocked.

When I use another USB stick, at once the Shadow Copy of hive system happens

Dietmar



#139 Dietmar

Dietmar

    Frequent Member

  • Advanced user
  • 236 posts
  •  
    Afghanistan

Posted 03 August 2018 - 12:58 PM

Hi all,

again the Toshiba RD400 needs about 10 min,

for to write a 16 GB continues file on it with Winhex,

meaning that the underlying harddisk black WD 2TB is

about 10 times faster!

This behavior remembers me a lot to USB sticks

Dietmar

 

PS: But no errors in 16GB on RD400, this is nice.



#140 Dietmar

Dietmar

    Frequent Member

  • Advanced user
  • 236 posts
  •  
    Afghanistan

Posted 04 August 2018 - 12:44 PM

Hi all,

I think I found the reason,

why XP does not like to start from Samsung 950 pro (with own Bios).

It is the fault of ntldr from XP (this happens not on NT4 and also not with the bootloader from Reactos).

 

NTLDR from XP sees a different geometry of the harddisk as Bios from 950 pro via INT13 sees!

And this case gives the crazy message "ntldr not found".

 

Is there another version of ntldr in the world, which does not decide, if this is the correct numbers of heads..etc.

on harddisk, may be one, which works only with lba?

Have a nice day

Dietmar

 

EDIT: Is there a program, with which you can find out, how the geometry of the harddisk is,

          that Bios from 950 pro tells to ntldr via INT13?



#141 Wonko the Sane

Wonko the Sane

    The Finder

  • Advanced user
  • 15149 posts
  • Location:The Outside of the Asylum (gate is closed)
  •  
    Italy

Posted 04 August 2018 - 01:38 PM

What about the grub4dos geometry command?

AFAIK it should show what the INT13 provides.

 

Is it possible that the issue is with CHS/LBA unbalanced values?

 

There should be no issues with having only LBA values in the MBR partition table, however.

 

:duff:

Wonko



#142 Dietmar

Dietmar

    Frequent Member

  • Advanced user
  • 236 posts
  •  
    Afghanistan

Posted 04 August 2018 - 02:09 PM

Hi jaclaz,

I came to the idea,

when I read

 

http://blog.clemens....ndows_3170.html

 

He did a hack, looking for 4 bytes

 

0F823A00 and changed them against 90909090  for NTFS and

0F824A00 and changed them against 90909090  for Fat32

 

this I am trying just now to fake all missunderstandings in geometrie in ntldr.

I am sure, that I found the reason, because the message "ntldr not found" happens very short after

this 4 bytes, as you can see with WinHex.

 

Also I will try your idea with grub4dos geometry command.

 

Very clear, why this is the reason: The manufacturer Samsung gives in BIOS via INT13 fixed

unknown values for its never changing, underlying nvme harddisk 950 pro in its rom Bios.

As long as the OS loader has no problem with this, may be "strange geometrie", it boots.

If ntldr has a problem with the geometrie values from this 950 pro BIOS,

it just gives the message "ntldr not found"

Dietmar

 

EDIT: Hack 90909090 does not work on 950 pro.



#143 Dietmar

Dietmar

    Frequent Member

  • Advanced user
  • 236 posts
  •  
    Afghanistan

Posted 04 August 2018 - 03:07 PM

hi jaclaz

 

grub>geometry

drive 0x80(LBA): C/H/S=2089/255/63, Sector Count/Size=33559785/512

    Partition num: 0, active, Filesystem type is fat32, partition type 0x0C

 

Grub in mbr works, but ntldr not,

Dietmar

 

EDIT: This are the values from mbr I think. So, no real information what Bios from 950 pro tells.



#144 Wonko the Sane

Wonko the Sane

    The Finder

  • Advanced user
  • 15149 posts
  • Location:The Outside of the Asylum (gate is closed)
  •  
    Italy

Posted 04 August 2018 - 03:12 PM

I don't think you found anything connected.

The patch for the NTFS and FAT32 bootsector code is ONLY related to the code used in the bootsector.

If you use the grub4dos and chainload /ntldr that piece of code is never used, as you are bypassing the bootsector code altogether.

 

If I recall correctly we already booted (with cdob) some XP's without any CHS data in the MBR partition table, only LBA values.

 

:duff:

Wonko



#145 Dietmar

Dietmar

    Frequent Member

  • Advanced user
  • 236 posts
  •  
    Afghanistan

Posted 04 August 2018 - 03:31 PM

Hi jaclaz,

do you have a working mbr for XP for this

"If I recall correctly we already booted (with cdob) some XP's without any CHS data in the MBR partition table, only LBA values."

Dietmar



#146 Wonko the Sane

Wonko the Sane

    The Finder

  • Advanced user
  • 15149 posts
  • Location:The Outside of the Asylum (gate is closed)
  •  
    Italy

Posted 04 August 2018 - 05:56 PM

Hi jaclaz,

do you have a working mbr for XP for this

"If I recall correctly we already booted (with cdob) some XP's without any CHS data in the MBR partition table, only LBA values."

Dietmar

I don't think the MBR code was modified.

SImply 00 the CHS table bytes of the entry, it should work. :unsure:

 

Or try using the modified MBR code (for very different reasons) used here:

http://reboot.pro/to...e-4#entry186449

I believe the above suits better the case than the seconf version here:

http://reboot.pro/to...o-gpt/?p=186471

with that the used by BIOS/code partition table entry is "moved" in such a position that you can effectively boot (first part of booting, i.e. real mode) a volume which is not in the "real" partition table (or it could be the same but with different CHS values).

 

:duff:

Wonko



#147 dencorso

dencorso

    Frequent Member

  • Advanced user
  • 142 posts
  •  
    Brazil

Posted 07 August 2018 - 06:02 AM

Nowadays, my main rig is an Asus P8Z68-V LX, powered by an i7 3770k and having 16 GiB RAM. I decide to try Kai's Community OFA NVMe drivers in it and so got a no-name M.2 to PCI-e adapter and a brand-new 250 GB Saumsung 970 EVO. After installing that, the BIOS remained fully ignorant of it, as expected, but Windows 7 SP1 x32 detected it all right, and works well with it with either v. 1.3 or v. 1.5 of Kai's drivers, or with Samsung's own v. v3.0.0.1802 WHQL driver (and it's recognized by Samsung's Magician, which also tells me it's genuine it has no problems). Once I was satisfied with that I've used diskpsrt to convert the drive from GPT to MBR, then formatted it as a single 256GB FAT-32 primary partition (to get it as plain vanilla as possible for XP SP3). Then I rebooted the machine into XP SP3 and it detected the PCI-e device promptly (just as Win 7 SP1 had done), and accepted Kai's Community OFA NVMe driver v. 1.3 (I had already put storport.sys v. 5.2.3790.4173) into the drivers folder. However, after installing the driver on XP I get a yellow bang, with the infamous "This device cannot start. (Code 10)". Same result with Kai's driver v. 1.5, and also when replacing storport.sys v. 5.2.3790.4173 by v. 5.2.3790.4163 GDR (from KB943295). At this point, I confess I'm baffled. The 970 EVO cannot be that much different from the 960 EVO Kai's used while developping his mods, now, or can it? Any ideas on what may be going wrong will surely be much appreciated. 



#148 Dietmar

Dietmar

    Frequent Member

  • Advanced user
  • 236 posts
  •  
    Afghanistan

Posted 07 August 2018 - 10:55 AM

Hi dencorso,

 

be sure, that you have 5.2.3790.4173 ver. storport.sys in windows\system32\drivers.

 

Take a look at the Bios of the Asus P8Z68-V LX.

For this download the UEFITool.exe

from https://www.win-raid...-UEFI-BIOS.html

wether your BIOS supports legacy nvme,

then you will find three files there:

Nvme

NvmeSmm

NVMEINT13

 

You can do this with a downloaded BIOS from Internet for your motherboard.

If you do not find this 3 files, no INT13 boot XP SP3 from nvme disk is possible on the board.

In this case, only the "Kansas City Shuffle" methode

works,

have a nice day

Dietmar

 

PS: Without any support from BIOS, some motherboards recognice the nvme disk as codevice, some not.

       Make a try to change the PCIe slot for the graphikcard with PCIe slot for the nvme disk.

      

       You can also make a try to integrate NvmExpressDxe_4.ffs to the BIOS from the Website from Fernando.

       Be very careful with that!



#149 Wonko the Sane

Wonko the Sane

    The Finder

  • Advanced user
  • 15149 posts
  • Location:The Outside of the Asylum (gate is closed)
  •  
    Italy

Posted 07 August 2018 - 11:38 AM

Only to (hopefully) contribute to the communication, dencorso never stated he actually attempted to boot from the new thingy, he attempted to access it as a secondary drive from an already booted XP SP3[1].

 

There is NO need whatsoever of any BIOS level support for a Mass Storage device accessed as secondary drive (not the boot/system one) this seems a much more "normal" :

 

"This device cannot start. (Code 10)"

 

:duff:

Wonko

 

 

[1] which - with no offence whatever implied to dencorso - very likely it is not a fresh installed XP SP3, but rather an existing installations where hundreds or possibly more drivers, programs and what not installed and removed, registry hacks, original or modified updates, etc. has been run.



#150 Dietmar

Dietmar

    Frequent Member

  • Advanced user
  • 236 posts
  •  
    Afghanistan

Posted 07 August 2018 - 11:44 AM

I remember,

that on my x58 board it was also not possible under XP

to have a nvme disk as codevice, same message (Code 10).

Dietmar

 

PS: I noticed, that some PCIe slots are connected with each other.

       This can also give the "This device cannot start. (Code 10)" message.

 

EDIT: Asus P8Z68-V LX

Der PCIe x16_2-Steckplatz teilt die Bandbreite mit dem PCIe x1_1- und dem PCIe x1_2-Steckplatz.






1 user(s) are reading this topic

0 members, 1 guests, 0 anonymous users