Coreboot and SeaBIOS

[Mevi]

I'm calling it a night. The current Realtek NIC driver is loaded, no exclamation mark, but windows says no connections..... Network diagnostics don't work in safe mode and the lack of the user profile and whatever other services don't start mean I can't use explorer, control panel and a whole lot else in safe mode. I'll switch OS tomorrow and see if I can get you some more bluescreens.

[inaxeon]

I'm calling it a night. The current Realtek NIC driver is loaded, no exclamation mark, but windows says no connections..... Network diagnostics don't work in safe mode and the lack of the user profile and whatever other services don't start mean I can't use explorer, control panel and a whole lot else in safe mode. I'll switch OS tomorrow and see if I can get you some more bluescreens.

I did notice something interesting there last night. The event viewer has a description of the drivers problem - Bad MAC address. Linux complains about this too. That comes from the SROM.

There's definitely an SROM fitted, It's U17. Assuming the thing is the correct size, we just have to reprogram it with the correct blob. I've never seen any tool which can do this but there is some code floating around which can reprogram older realtek chips. I could probably update this for the 8111C once I get time.

Have you got a Joggler which is known to have the ethernet port working windows? If so, what does Win 7 do?

[roobarb!]

You probably already know this, but the Joggler uses the wifi adapter's MAC address for both the wired and wireless interfaces (it just applies the MAC in software to the wired interface). I'd wondered why Linux didn't see a proper address for it, but I guess it's just not there in that SROM. Would be mighty handy if it did give a proper address.

[inaxeon]

That is an awesome way of getting out of the $5000 fee for a block of MAC addresses.

On my Joggler, the SROM appears to be blank. So why the hell did they fit it BOM cost was obviously not a consideration when designing the Joggler

[Mevi]

Can't the NIC be reprogrammed from DOS? I recall seeing a DOS utility on the Realtek site for the PCIe family chip.

I'll fire up the other corebooted Joggler which does have a working NIC and let you know what happens.

EDIT: the Joggler I've been using with w7 has the WLAN removed.... Hmm

[inaxeon]

Can't the NIC be reprogrammed from DOS? I recall seeing a DOS utility on the Realtek site for the PCIe family chip.

I'll fire up the other corebooted Joggler which does have a working NIC and let you know what happens.

EDIT: the Joggler I've been using with w7 has the WLAN removed.... Hmm

Don't worry about missing WLAN. This MAC address sharing scheme roobarb! mentioned wouldn't be the slightest bit applicable on anything other than the factory OS - because it needs a hacked realtek driver to work.

There is a utility for reprogramming the Boot ROM isn't the same as the SROM.

SROM (Typically 128-256 bytes in size): Stores MAC address, serial number, and any other special configuration i.e. exotic PHYs attached.
BootROM (Typically 16-128K in size): Stores a boot agent such as Etherboot, PXE etc. We don't have one of those on the Joggler.

SROM Programming tools are generally closely guarded and require a Non Disclosure Agreement to obtain. I have managed to get these tools out of manufacturers in the past, by pretending to be a company about to buy millions of chips...

This is what an SROM programmer looks like:

[pete]

Great news on the NIC card stuff there inaxeon!

That whole NIC card issue is prevalent with about 50% of my Jogglers and the working 50% are hit or miss. I have though removed almost every one of the WLAN cards in my Jogglers. I still get this on/off connectivity issue with the wireless depending on how much work the Joggler is doing. I saw the same with the Chumby's chrooted and running on the Arm 800Mhz CPU and their basic OS.

A wireless test here was streaming HD TV via MythTV wired and wirelessly (big difference).

This messes with my direction of using POE/Gb connected Jogglers.

[Mevi]

For anyone wanting to try XP pro, usbooting a plain vanilla Xp just results in a black screen right after seabios boots from hdd. I also tried my build 3 which just results in flashing crazy coloured scrolling exclamation and question marks. Screen nonsense. It's ntdetect that is the culprit here i think as I tried original and hacked ntldr, ntdetect and VGA.sys.

[inaxeon]

For anyone wanting to try XP pro, usbooting a plain vanilla Xp just results in a black screen right after seabios boots from hdd. I also tried my build 3 which just results in flashing crazy coloured scrolling exclamation and question marks. Screen nonsense. It's ntdetect that is the culprit here i think as I tried original and hacked ntldr, ntdetect and VGA.sys.

Oh yeah, XP is a real pain. NTDETECT locks its self solid trying to probe the non existant i8042. Dan mentioned that if i8042prt.sys wasn't present, it may not attempt to do this. You can't use the hacks for the XOJ process, they're not right for this.

This is the most interesting tidbit of information I've seen on this problem: http://www.itlisting.org/3-windows-xp/b ... 00b80.aspx

[pete]

inaxeon

You got me looking now at my CF XPE builds to try (pile sitting on the shelf these days) from yesteryear.

Off the topic a bit seems too that automotive HU's are going the CE route run on anything lately versus .... geez even my replacement Tivo satellite DVR is using a CE base.

[danfoshizzle]

For anyone wanting to try XP pro, usbooting a plain vanilla Xp just results in a black screen right after seabios boots from hdd. I also tried my build 3 which just results in flashing crazy coloured scrolling exclamation and question marks. Screen nonsense. It's ntdetect that is the culprit here i think as I tried original and hacked ntldr, ntdetect and VGA.sys.

Oh yeah, XP is a real pain. NTDETECT locks its self solid trying to probe the non existant i8042. Dan mentioned that if i8042prt.sys wasn't present, it may not attempt to do this. You can't use the hacks for the XOJ process, they're not right for this.

This is the most interesting tidbit of information I've seen on this problem: http://www.itlisting.org/3-windows-xp/b ... 00b80.aspx

Yeah I spent a good few hours researching late the other night in a coffee addled brainstorm, but that is all I can come with, if it does not work though, the only way out would probably be patching ntdetect, which is by no means easy, as a friend of mine said, its more likely to break stuff patching the .sys than actually fixing it.

[Mevi]

Deleting i8042prt.sys does not fix this problem - still locks up. I8042prt.sys is referenced a whole ton of times in the xp registry.

[inaxeon]

Deleting i8042prt.sys does not fix this problem - still locks up. I8042prt.sys is referenced a whole ton of times in the xp registry.

I think it works like this (here's some pseudo code ):

(In NTDETECT):

If (Probe_PS2_Keyboard_Behind_i8042() = TRUE)
Add_Driver_To_Load("i8042prt.sys")
Else
// Don't load i8042prt.sys

Probe_PS2_Keyboard_Behind_i8042() {
// Directly write to i8042 ports 0x60, 0x61 and 0x64
If Keyboard_Exists
return TRUE
Else
return FALSE
}


So long as NTDETECT doesn't think there's a PS/2 keyboard attached, it wont ever load the driver to talk to the i8042, so if we hack ntdetect, there won't be any interaction with our non existant i8042.

I don't think it'd take more than a day to do this simple hack, but it needs someone who's not afraid of x86 machine code. Because NTDETECT is talking directly to the i8042, finding the culprit code wouldn't be difficult.

[pete]

You have a base of seabios testers here inaxeon. The only difference I see is the pata port thing on some of the Jogglers being used.

Just tell us what to do?

[danfoshizzle]

Found this

Windows XP supports operating without a keyboard, mouse, or monitor attached to the system. On an ACPI-enabled system, Windows XP supports operating without a legacy 8042 keyboard controller. Windows XP also includes support for systems without VGA/display hardware.

Here:http://www.osronline.com/ddkx/appendix/ ... 5_3984.htm

EDIT: Apparently something called dietmars ntdetect may help, it patches ntdetect for usb booting but may also work for i8042. But this is an open source thread so I wont share the one I just made, but read up how to do it here: http://www.msfn.org/board/topic/112630- ... ntry738009

Will try it later on when I have some time lol

[inaxeon]

On an ACPI-enabled system, Windows XP supports operating without a legacy 8042 keyboard controller.

That is incredibly interesting. If true, we can theoretically pull this off. I don't have a descriptor for the i8042 in the ACPI at present, Perhaps you've got to put some stuff in there to explicitly say it isn't there? Or perhaps we're stuck on something else now.

If anyone's got spare time, boot a copy of XP using NTDETECT.CHK and ensure it's still the keyboard it's stuck on.

[danfoshizzle]

Thinking ACPI will have to be told its not there so it does not try to call it, just brainstorming here but its the only way I can help, coding is not something I am very good at so if I can throw ideas at yah maybe one will help lol

EDIT: HaHa, copy and paste coder, thats what my mate used to call me lol

[pete]

I remember some tweaking in early versions of XP SP3 boot stuff cuz of the new SATA HD's. Nothing in the BIOS at the time and a checking in the XP boot would kill the install. So it was changing the NTDETECT circumventing the checking which allowed booting. Just a few years back.

[Mevi]

I tried disabling the i8042prt.sys driver in the registry: http://windowstechissues.wordpress.com/ ... d-on-boot/

Doesn't work! I'll have a go at booting with ntdetect.chk. I'll update in a bit.

[Mevi]