Get Windows XP x86 to recognize more than 4Gb with PAE?

[jaclaz 1,218]

I found another compatability issue with XP PAE.

I have USB logic analyser and using driver and software from ...

 

Basically you are confirming that the MS choice of removing PAE from XP was - though a perverted one - valid   , what happened at the time, more or less was:

Since the hardware, but mainly the Third Party's software and particularly drivers suck (and suck big)^[1] we remove the feature, limiting it to Server 2003(and ONLY on the Enterprise/Datacenter editions and - but senselessly limited to 8 Gb - on the lesser known "Windows Storage Server 2003") where both hardware and drivers are better checked, i.e. there is more money per unit involved... 

... since we are MS we will put the blame on the Third Parties and start senselessly pushing 64 bit computing, creating to both the end users and to the Third Parties involved, even those that were involved in the writing of those sucking drivers, every kind of new problem/issue.

This has the not-so trifling advantage that when we will starting pushing for a new (artificially made incompatible) disk partitioning structure and a new (artificially made incompatible) firmware, people won't be shocked by it, as they will by that time be used to all kind of arrogant, senseless or plainly stupid move we will make.

And of course, as soon as the issues caused by these two unneeded changes will start to fade away because people will either get used to them or find workarounds, we will make a couple new Operating Systems that will be made artificially incompatible even visually with anything else....

 

jaclaz 

 

 

 

^[1] ... and maybe the fact that we changed three times in three years the understructure and the driver models, each time failing to provide complete documentation or to provide real-world working sampled more complex than hello-world-type simplified and largely malfunctioning sample drivers is marginally  involved in this ...

[xrayer 4]

Since the hardware, but mainly the Third Party's software and particularly drivers suck (and suck big)^[1] we remove the feature, limiting it to Server 2003(and ONLY on the Enterprise/Datacenter editions and - but senselessly limited to 8 Gb - on the lesser known "Windows Storage Server 2003") where both hardware and drivers are better checked, i.e. there is more money per unit involved...

 

But in this case no 3rd party driver is involved (I don't count the plain inf file), winusb.sys is from microsoft same as low level usb*.sys drivers. So they give to other driver writers a bad example...

I tried with hidusb.sys from w2k3 but no change, as I expected this driver is used for HID devices, log. analyser is not a HID class.

[jaclaz 1,218]

 

But in this case no 3rd party driver is involved (I don't count the plain inf file), winusb.sys is from microsoft same as low level usb*.sys drivers. So they give to other driver writers a bad example...

 

 

Yep , that's the idea of note ^[1] in my post.

 

I know nothing or very little about drivers, mind you, but I have seen too many failed attempts to write proper drivers and too many subsequent revisions/versions of what should normally be a "plain enough" piece of code, and I suspect that - besides the complexity of the NT structure - there is a large amount of misssing or even plainly wrong documentation.

 

jaclaz

[dencorso 1,083]

 I suspect that - besides the complexity of the NT structure - there is a large amount of misssing or even plainly wrong documentation.

 

That surely was the case in .VxD times (even the LE EXE format is, up-to-now, very poorly documented), then there were at least two installments of WDM drivers (the second of which is the one actually used in XP), and when that was becoming better understood and documented they've moved on to yet another model in 7+, besides the user-mode drivers they've been pushing onto us for sometime, which are backwards compatible to XP, sort of.

[jaclaz 1,218]

... on to yet another model in 7+

 

Wasn't the main "switch" in Vista ?

 

http://en.wikipedia.org/wiki/Windows_Driver_Model#Criticism

 

jaclaz

[dencorso 1,083]

Well... 

 

AFAIK, there are, from the oldest to the newest:
1. Old static .VxDs (usually using the .386, not the .VxD extension), from Win 3.0 times.
2. Windows NT Driver Model, from NT incept time.
3. Newer .VxDs (usually using .VxD or .PDR extensions), from Win 95 up to ME (inclusive).
4. 9x/ME WDM drivers (usually using .SYS or .MPD extensions), from Win 95 OSR 2.x up to ME (inclusive).
5. NT-type WDM drivers (usually using .SYS or .MPD extensions), from Win 2k onwards.
6. KMDF drivers, portable backwards to 2k, with add-on installation required, native for Vista+
7. UMDF drivers, portable backwards to XP, with add-on installation required, native for Vista+
8. Yet another driver model being pushed now, IIRR, from 7 onwards, the ostensible reason for Intel producing USB 3.0 drivers suposedly impossible to backport... but I may be wrong and this further model may just be a bad dream conjured by my imagination.

Although WDM was initially intended to be binary and source compatible between Win 9x and Win 2k, this almost never has been the case, so specific drivers developed for one OS may sometimes be coerced to work on some other OS, although most times even that is not feasible. WDM drivers are designed to be forward-compatible, so that a WDM driver ought to run on a version of Windows newer than the one the driver was initially written for, but doing that would mean that the driver cannot take advantage of any new features introduced with the new windows version. OTOH, WDM drivers are generally not intended to be backward-compatible, so that a WDM rarely can run as-is on an earlier Windows version that the one it was intended for, unless wrapped or shimmed for that (a case-by-case solution, which sometimes works).

[jaclaz 1,218]

3. Newer .VxDs (usually using .VxD or .PDR extensions), from Win 95 up to ME (inclusive).

4. 9x/ME WDM drivers (usually using .SYS or .MPD extensions), from Win 95 OSR 2.x up to ME (inclusive).

5. NT-type WDM drivers (usually using .SYS or .MPD extensions), from Win 2k onwards.

Yep , these three were the three in three years time I was referring to, but I seem to remember that there were also some slight differences for Windows 2003 drivers, possibly introduced with SP1 or SP2 (but cannot really say for sure), while I am pretty sure that for some reasons (but cannot really remember the specifics) I did keep some Windows 2000 Server installs because there were not available drivers for the 2003 and the 2k one's did not work (and thus went "forward compatibility" of the WDM's).

 

8. Yet another driver model being pushed now, IIRR, from 7 onwards, the ostensible reason for Intel producing USB 3.0 drivers suposedly impossible to backport... but I may be wrong and this further model may just be a bad dream conjured by my imagination.

To be fair I believe that - just like UEFI/GPT and their artificial or senseless incompatibilities - the main culprit in this case is more Intel than MS (but of course I may well be wrong).

 

jaclaz

[xrayer 4]

One of my friend using PAE patch reported me another problem with HP Scannjet 3800 scanner - when command to scanning is sent it doesn't start and locks the scanner application that cannot be closed any way. Also it cause some corruption to Windows that cannot be restarted/shutdown normal way but freeze at last turning off screen. Again, when booted with original unpatched kernel it works fine. Seems similar issue like with my logic analyzer.

So XP64G patch has only limited usage and it would need some improvement that could make some workaround for that faulty drivers to not load them above 4GB and make sure they don't need handle any address >4GB (using some buffers in "low" mem), etc. simply so they can run same as on nonPAE system. I really don't know how complex modification would be needed...

[dencorso 1,083]

Fact is that the mods discussed in this thread remain unreliable.

 

OTOH, simply activating /PAE on boot.ini and adding a Gavotte ramdisk using PAE remains a good option for those using XP x86 and having more than 3.2 GiB RAM. If the resulting ramdisk is big enough, after setting DisablePagingExecutive=1, one can even put the pagefile in it, besides things like the Temporary Internet Files...

 

BTW: Gavotte Ramdisk Version History and Read Me in English.

[noric 0]

So XP64G patch has only limited usage and it would need some improvement that could make some workaround for that faulty drivers to not load them above 4GB and make sure they don't need handle any address >4GB (using some buffers in "low" mem), etc. simply so they can run same as on nonPAE system. I really don't know how complex modification would be needed...

Some time ago I stumbled upon this set of files that supposedly need to be replaced. I don't remember if I found it on a chinese site or on overclock.net. I haven't used it up to now, but here you are.

Edited January 26, 2015 by noric

[xrayer 4]

Some time ago I stumbled upon this set of files that supposedly need to be replaced. I don't remember if I found it on a chinese site or on overclock.net. I haven't used it up to now, but here you are.

 

 

Thanks but this are the same files included in XP64G patch (from w2k3 server) that I have already replaced...

I have no idea what usb releated drivers should I replace more. Maybe I could install W2k3 server to a VM and then copy all *.sys drivers to XP to try if something will change...

[noric 0]

Thanks but this are the same files included in XP64G patch (from w2k3 server) that I have already replaced...

I have no idea what usb releated drivers should I replace more. Maybe I could install W2k3 server to a VM and then copy all *.sys drivers to XP to try if something will change...

 

Did you replace any other file, other than those I uploaded yesterday?

 

However, I've just noticed that some of my usb ports don't work (nothing happens when I plug) with 2k3 usbport.sys and usbhub.sys. This is with or without PAE. The only way to get them working is to restore xp usbport.sys and usbhub.sys.

I'll try to replace all the files I uploaded (up to now I had replaced only those two .sys) and see if it fixes my issue.

 

Copying all 2k3 .sys is another option, indeed...

Edited January 27, 2015 by noric

[dencorso 1,083]

The real problem are not the drivers... the real problem is the currently known hal patches are not enough.

Until someone finds what else is missing, the 2k3 to workstation conversion remains a more reliable option:

 

How to convert your Win 2k3... to a Workstation!

[xrayer 4]

Yes, may be an option. For me, the advantage of 4GB PAE is far much less than time spent by fresh instal of OS, apps, drivers, etc. I'm running my current XP installation for about 8 years while I'm patching and improving it, I never needed make clear install during that time bacuse of my backup and testing system. And because I didn't radically changed my HW setup - I did just some upgrades that doesn't affected comaptability like CPU, RAM, SSD that boosted system a lot. Still giving enough power for common work this days...

[noric 0]

Yes, same for me. I improved my xp installation through the years and PAE is a great chance to keep it alive and avoid hassle. Ram limit is the only downside I see in my xp.

Edited January 28, 2015 by noric