[LinuxBIOS] No keyboard.

[Feng, Libo]

Hi, all,

Today, I almost finish LinuxBIOS on Tyan s3992. A text-mode RedHat is booted, and a "Hello, World" is written and run. Thank a lot of people here. The mistake I made is building LinuxBIOS on a FC machine and testing it on another RH machine, many machine check panics arose.

The one thing left is that keyboard is not enabled in Linux. It seems it is not correctly configured in LinuxBIOS. I have to use a serial port and terminal to access the machine. Where the keyboard is enabled in LinuxBIOS? And maybe easier, Linux has some kind of application like Windows to configure keyboard or not?

Best Regards

丰立波 Feng Libo @ AMD  Ext: 20906
Mobile Phone: [phone]
Office Phone: [phone]

[ron minnich]

On 8/2/07, Feng, Libo  wrote:
> Hi, all,
>
> Today, I almost finish LinuxBIOS on Tyan s3992. A text-mode RedHat is
> booted, and a "Hello, World" is written and run. Thank a lot of people here.
> The mistake I made is building LinuxBIOS on a FC machine and testing it on
> another RH machine, many machine check panics arose.
>
> The one thing left is that keyboard is not enabled in Linux. It seems it is
> not correctly configured in LinuxBIOS. I have to use a serial port and
> terminal to access the machine. Where the keyboard is enabled in LinuxBIOS?
> And maybe easier, Linux has some kind of application like Windows to
> configure keyboard or not?

This is great news!

What chip is the keyboard connected to? Then we can see how to enable it.

ron

[Peter Stuge]

Hi,

On Thu, Aug 02, 2007 at 05:21:05PM +0800, Feng, Libo wrote:
> Today, I almost finish LinuxBIOS on Tyan s3992. A text-mode RedHat
> is booted, and a "Hello, World" is written and run.

Excellent! :)
> Thank a lot of people here. The mistake I made is building
> LinuxBIOS on a FC machine and testing it on another RH machine,
> many machine check panics arose.

The GNU toolchain (gcc and binutils) in the FC system may be unable
to build LB correctly. Can you tell us what version of gcc and
binutils you were using in the FC system that produced bad results,
and what version you have used to produce a working result?
> The one thing left is that keyboard is not enabled in Linux. It
> seems it is not correctly configured in LinuxBIOS. I have to use a
> serial port and terminal to access the machine. Where the keyboard
> is enabled in LinuxBIOS?

That depends on what the keyboard connects to. Is it a USB keyboard
or a PS/2 keyboard? Whatever it is, could you also try the other?
Also see Ron's question about which chip the keyboard is connected
to.
> And maybe easier, Linux has some kind of application like Windows
> to configure keyboard or not?

Not always an application, but there are configuration files in most
if not all Linux distributions for setting up the keymap. That
doesn't help if the kernel can't see the keyboard however. Feel free
to send a serial port capture from a full boot with LB to the mailing
list, we'll look at it and see if we can find something pointing to
the error.

//Peter

[Feng, Libo]

Hi,

The Super IO is nsc-87417, there is a directory for the chip. I have enabled the keyboard device in mainboard Config.lb, as below,

Device pnp 2e.6 on
    io 0x60 = 0x60
    io 0x62 = 0x64
    irq 0x70 = 1
End
However, in super/nsc/pc87417/superio.c, the keyboard pnp_info is {&ops, PC87417_KBCK, PNP_IO0 | PNP_IO1 | PNP_IRQ0, {0x7f8, 0}, {0x7f8, 4}}, is it correct? I think it should be PNP_IRQ1, {0x60, 0} and {0x60, 4}, I changed it, but still no keyboard. What should I do next?

Best Regards

??? Feng Libo @ AMD  Ext: 20906
Mobile Phone: [phone]
Office Phone: [phone]

-----Original Message-----
From: ron minnich [mailto:[e-mail]] 
Sent: Thursday, August 02, 2007 11:32 PM
To: Feng, Libo
Cc: [e-mail]
Subject: Re: [LinuxBIOS] No keyboard.

On 8/2/07, Feng, Libo  wrote:
> Hi, all,
>
> Today, I almost finish LinuxBIOS on Tyan s3992. A text-mode RedHat is 
> booted, and a "Hello, World" is written and run. Thank a lot of people here.
> The mistake I made is building LinuxBIOS on a FC machine and testing 
> it on another RH machine, many machine check panics arose.
>
> The one thing left is that keyboard is not enabled in Linux. It seems 
> it is not correctly configured in LinuxBIOS. I have to use a serial 
> port and terminal to access the machine. Where the keyboard is enabled in LinuxBIOS?
> And maybe easier, Linux has some kind of application like Windows to 
> configure keyboard or not?

This is great news!

What chip is the keyboard connected to? Then we can see how to enable it.

ron

[Feng, Libo]

Hi,

I built LinuxBIOS on Linux version 2.6.11-1.1369_FC4smp and gcc version 4.0.0 20050525, then tested LinuxBIOS on Linux version 2.6.9-22 Elsmp and gcc version 3.4.4 20050721. There were always many kind of machine check panics.

Now I build and test on Linux version 2.6.9-22 Elsmp and gcc version 3.4.4 20050721.

I have tried both USB and PS/2 keyboard, neither can work. I changed some code, it still doesn't work. Maybe I miss something.
In Linux, kudzu can do some hardware detect, but don't work in this case.
The attached is a log, maybe you can help me find some clue. Thank you. We just want to gain some BIOS experience, so the LinuxBIOS version is very old.

Best Regards

??? Feng Libo @ AMD  Ext: 20906
Mobile Phone: [phone]
Office Phone: [phone]

-----Original Message-----
From: [e-mail] [mailto:[e-mail]] On Behalf Of Peter Stuge
Sent: Friday, August 03, 2007 4:43 AM
To: [e-mail]
Subject: Re: [LinuxBIOS] No keyboard.

Hi,

On Thu, Aug 02, 2007 at 05:21:05PM +0800, Feng, Libo wrote:
> Today, I almost finish LinuxBIOS on Tyan s3992. A text-mode RedHat is 
> booted, and a "Hello, World" is written and run.

Excellent! :)
> Thank a lot of people here. The mistake I made is building LinuxBIOS 
> on a FC machine and testing it on another RH machine, many machine 
> check panics arose.

The GNU toolchain (gcc and binutils) in the FC system may be unable to build LB correctly. Can you tell us what version of gcc and binutils you were using in the FC system that produced bad results, and what version you have used to produce a working result?
> The one thing left is that keyboard is not enabled in Linux. It seems 
> it is not correctly configured in LinuxBIOS. I have to use a serial 
> port and terminal to access the machine. Where the keyboard is enabled 
> in LinuxBIOS?

That depends on what the keyboard connects to. Is it a USB keyboard or a PS/2 keyboard? Whatever it is, could you also try the other?
Also see Ron's question about which chip the keyboard is connected to.
> And maybe easier, Linux has some kind of application like Windows to 
> configure keyboard or not?

Not always an application, but there are configuration files in most if not all Linux distributions for setting up the keymap. That doesn't help if the kernel can't see the keyboard however. Feel free to send a serial port capture from a full boot with LB to the mailing list, we'll look at it and see if we can find something pointing to the error.

//Peter

[Peter Stuge]

On Fri, Aug 03, 2007 at 06:34:23PM +0800, Feng, Libo wrote:
> I built LinuxBIOS on Linux version 2.6.11-1.1369_FC4smp and gcc
> version 4.0.0 20050525, then tested LinuxBIOS on Linux version
> 2.6.9-22 Elsmp and gcc version 3.4.4 20050721. There were always
> many kind of machine check panics.
>
> Now I build and test on Linux version 2.6.9-22 Elsmp and gcc
> version 3.4.4 20050721.

Kernel versions should not matter, but both kernels are patched
(RedHat?) so we can't be sure. I don't trust patched kernels.

gcc version is the same on both the non-working system and the
currently working system. Can you also check the version of binutils
on old and new system? (e.g. by running ld --version or nm --version)
> I have tried both USB and PS/2 keyboard, neither can work.
> From the serial output I see that it is impossible for a USB keyboard

to work. More below. USB and PS/2 are quite different code paths in
the kernel, until one of them is working there is surely some generic
problem, quite possibly in LB.
> I changed some code, it still doesn't work. Maybe I miss something.
> In Linux, kudzu can do some hardware detect, but don't work in this
> case.

Can you get more details from kudzu about why it fails? Perhaps by
running it manually after the system has booted? Perhaps by looking
in the syslog files? I don't know much about RedHat so I can't give
much advice on kudzu. (On the other hand, kudzu itself is not very
important, having the hardware work is important, and then kudzu will
be happy too.)
> The attached is a log, maybe you can help me find some clue. Thank
> you.

Great - see some comments below.
> We just want to gain some BIOS experience, so the LinuxBIOS
> version is very old.

This can be both good and bad. If you started from a known working LB
version then there's no big problem, but if a lot of changes and
bugfixes have gone into LB since the version you are using, it could
be worthwhile to try also a new version.

Comments on the log:
> PCI: pci_scan_bus for bus 00
> PCI: 00:01.0 [1166/0036] enabled

HT1000 PCI/PCI-X bridge
> PCI: 00:02.0 [1166/0205] enabled

HT1000 Legacy South Bridge
> PCI: 00:02.1 [1166/0214] enabled

HT1000 Legacy IDE controller
> PCI: 00:02.2 [1166/0234] enabled

HT1000 LPC Bridge
> PCI: 00:02.3 [1166/0238] enabled
> PCI: 00:02.4 [1166/0235] enabled
> PCI: 00:02.5 [1166/0235] enabled
> PCI: 00:02.6 [1166/0235] enabled

Not in my pci.ids.
> PCI: 00:03.0 [1166/0223] enabled
> PCI: 00:03.1 [1166/0223] enabled
> PCI: 00:03.2 [1166/0223] enabled

3x HT1000 USB Controller
> PCI: 00:04.0 [8086/1229] enabled

82557/8/9 [Ethernet Pro 100]

[..]
> Reading resources...

[..]
> PCI: 00:01.0 24 <- [0xfffffffffff00000 - 0xffffffffffefffff] bus 01 prefmem

[..]
> PCI: 00:00.0 missing read_resources
> PCI: 00:01.0 missing read_resources
> PCI: 00:02.0 missing read_resources
> PCI: 00:03.0 missing read_resources
> PCI: 00:04.0 missing read_resources
> PCI: 00:05.0 missing read_resources
> PCI: 00:00.0 missing read_resources
> PCI: 00:01.0 missing read_resources
> PCI: 00:01.1 missing read_resources
> PCI: 00:01.2 missing read_resources
> PCI: 00:01.3 missing read_resources
> PCI: 00:01.4 missing read_resources
> PCI: 00:01.5 missing read_resources
> PCI: 00:01.6 missing read_resources
> PCI: 00:02.0 missing read_resources
> PCI: 00:02.1 missing read_resources
> PCI: 00:02.2 missing read_resources
> PCI: 00:03.0 missing read_resources
> PCI: 00:00.0 missing read_resources
> PCI: 00:01.0 missing read_resources
> PCI: 00:02.0 missing read_resources
> PCI: 00:03.0 missing read_resources
> PCI: 00:04.0 missing read_resources
> PCI: 00:05.0 missing read_resources
> PCI: 00:00.0 missing read_resources
> PCI: 00:01.0 missing read_resources
> PCI: 00:01.1 missing read_resources
> PCI: 00:01.2 missing read_resources
> PCI: 00:01.3 missing read_resources
> PCI: 00:01.4 missing read_resources
> PCI: 00:01.5 missing read_resources
> PCI: 00:01.6 missing read_resources
> PCI: 00:02.0 missing read_resources
> PCI: 00:02.1 missing read_resources
> PCI: 00:02.2 missing read_resources
> PCI: 00:03.0 missing read_resources
> PCI: 00:00.0 missing read_resources
> PCI: 00:01.0 missing read_resources
> PCI: 00:02.0 missing read_resources
> PCI: 00:03.0 missing read_resources
> PCI: 00:04.0 missing read_resources
> PCI: 00:05.0 missing read_resources
> PCI: 00:00.0 missing read_resources
> PCI: 00:01.0 missing read_resources
> PCI: 00:01.1 missing read_resources
> PCI: 00:01.2 missing read_resources
> PCI: 00:01.3 missing read_resources
> PCI: 00:01.4 missing read_resources
> PCI: 00:01.5 missing read_resources
> PCI: 00:01.6 missing read_resources
> PCI: 00:02.0 missing read_resources
> PCI: 00:02.1 missing read_resources
> PCI: 00:02.2 missing read_resources
> PCI: 00:03.0 missing read_resources
> Done reading resources.

[..]
> Setting resources...

[..]
> PCI: 00:00.0 missing read_resources
> PCI: 00:01.0 missing read_resources
> PCI: 00:02.0 missing read_resources
> PCI: 00:03.0 missing read_resources
> PCI: 00:04.0 missing read_resources
> PCI: 00:05.0 missing read_resources

[..]

Does this matter, anyone? If there's no read_resources() op (or even
no dev->ops) there's no fallback to generic routines in the code -
are the devices likely to work anyway?

[..]
> PCI: 00:02.0 90 <- [0x0000003040 - 0x000000304f] io
> PCI: 00:02.1 10 <- [0x0000003060 - 0x0000003067] io
> PCI: 00:02.1 14 <- [0x0000003080 - 0x0000003083] io
> PCI: 00:02.1 18 <- [0x0000003070 - 0x0000003077] io
> PCI: 00:02.1 1c <- [0x0000003090 - 0x0000003093] io
> PCI: 00:02.1 20 <- [0x0000003050 - 0x000000305f] io
> PCI: 00:02.1 64 <- [0x00000030a0 - 0x00000030a3] io
> PCI: 00:02.3 10 <- [0x00fc250000 - 0x00fc250fff] mem
> PCI: 00:02.4 10 <- [0x00fc251000 - 0x00fc251fff] mem
> PCI: 00:02.5 10 <- [0x00fc252000 - 0x00fc252fff] mem
> PCI: 00:02.6 10 <- [0x00fc253000 - 0x00fc253fff] mem

Some resources are being assigned either way..

[..]
> Done setting resources.
> Done allocating resources.
> Enabling resources...

[..]
> PCI: 00:01.0 bridge ctrl <- 0003
> PCI: 00:01.0 cmd <- 147

[..]
> PCI: 00:02.0 cmd <- 147
> PCI: 00:02.1 cmd <- 141
> PCI: 00:02.2 cmd <- 144
> PCI: 00:02.3 cmd <- 146
> PCI: 00:02.4 cmd <- 146
> PCI: 00:02.5 cmd <- 146
> PCI: 00:02.6 cmd <- 146
> PCI: 00:03.0 cmd <- 143
> PCI: 00:03.1 cmd <- 143
> PCI: 00:03.2 cmd <- 143
> PCI: 00:04.0 cmd <- 143

[..]

And some command seems to be sent, but..
> PCI: 00:00.0 missing enable_resources
> PCI: 00:01.0 missing enable_resources
> PCI: 00:02.0 missing enable_resources
> PCI: 00:03.0 missing enable_resources
> PCI: 00:04.0 missing enable_resources
> PCI: 00:05.0 missing enable_resources
> PCI: 00:00.0 missing enable_resources
> PCI: 00:01.0 missing enable_resources
> PCI: 00:01.1 missing enable_resources
> PCI: 00:01.2 missing enable_resources
> PCI: 00:01.3 missing enable_resources
> PCI: 00:01.4 missing enable_resources
> PCI: 00:01.5 missing enable_resources
> PCI: 00:01.6 missing enable_resources
> PCI: 00:02.0 missing enable_resources
> PCI: 00:02.1 missing enable_resources
> PCI: 00:02.2 missing enable_resources
> PCI: 00:03.0 missing enable_resources

Here we go again about missing *_resources op or even dev->ops.

Is it just that none is needed for HT1000 or can this be the cause of
the problems?
> NB: Function 3 Misc Control.. done.
> PCI: 00:02.0 init
> PCI: 00:02.1 init
> PCI: 00:02.2 init
> RTC Init
> Invalid CMOS LB checksum
> PCI: 00:02.3 init
> PCI: 00:02.4 init
> PCI: 00:02.5 init
> PCI: 00:02.6 init
> PCI: 00:03.0 init
> PCI: 00:03.1 init
> PCI: 00:03.2 init
> PCI: 00:04.0 init

[..]
> Devices initialized

So most of the devices are initialized, but that may not mean much.
> FILO version 0.5 ([e-mail]) Sat Jul  7 02:41:14 CST 2007
> menu: hda1:/grub/grub.conf

What about FILO - have you tried using a PS/2 keyboard there?
(Compile FILO with a timeout for the boot prompt.) Until that works,
there is not much reason to boot Linux. (Quicker cycles)

[..]
> Linux version 2.6.9-22.ELsmp ([e-mail])
> (gcc version 3.4.4 20050721 (Red Hat 3.4.4-2)) #1 SMP Mon Sep 19
> 18:00:54 EDT 2005

This kernel is almost two years old, you are ignoring the last 1/5
part of the lifetime of the Linux project.

I would strongly suggest building a recent kernel.org version of
Linux (2.6.21.6 or 2.6.22.1) without modules and making sure it is
configured according to the hardware you're working with. If you want
the list to review your configuration, please feel free to send the
.config created by make menuconfig. (or make xconfig or make qconfig)

[..]
> usbcore: registered new driver hiddev
> usbcore: registered new driver usbhid
> drivers/usb/input/hid-core.c: v2.0:USB HID core driver
> mice: PS/2 mouse device common for all mice

No USB host controllers are found by Linux so it is impossible for a
USB keyboard to work. The HID driver is loaded but it only depends on
the USB core. Maybe the host controller drivers aren't enabled in
this kernel, maybe the PCI devices don't respond properly (LB's
fault) or maybe there are bugs in the HCI drivers. Too many unknowns.

There are at least two interesting things you could try:

1. Investigate the {read,enable}_resources messages from LB for
HT1000 and fix the problem if there is in fact a problem. Best case
this makes everything work. A good first test is if a PS/2 keyboard
works in FILO.

2. Upgrade to the latest kernel, correctly configured for the system,
and see what works/does not work. Perhaps that can give clues about
what is wrong in LB. (if there is something more than 1)

(Could be useful to try 2 before 1 as well.)

//Peter

[yhlu]

please check if bcm5785 bcm5780 is in the first ht link in your MB Config.lb.
Or send out your MB Config.lb.

YH

[Feng, Libo]

Hi, Mr. Lu,

You are quit right. I commented out two statements in Config.lb as you suggest, the keyboard can work now. The attached is the modified file, a BTDC tag indicates where I modified. Thank you a lot.

Best Regards

??? Feng Libo @ AMD  Ext: 20906
Mobile Phone: [phone]
Office Phone: [phone]

-----Original Message-----
From: yhlu [mailto:[e-mail]] 
Sent: Saturday, August 04, 2007 1:40 AM
To: Feng, Libo
Cc: Peter Stuge; [e-mail]
Subject: Re: [LinuxBIOS] No keyboard.

please check if bcm5785 bcm5780 is in the first ht link in your MB Config.lb.
Or send out your MB Config.lb.

YH