a pre-release of merging loongson patchs to linux-2.6.29.1

hi, all

these days, I am working on merging loongson patchs to linux-2.6.29.1, the fuloong(2f) & yeeloong source code have been completely merged to an 2f directory, most of the 2e & 2f source code have been merged except the irq.c & reset.c, fixup-loongson2e.c & fixup-loongson2f.c.

the current directory architecture is as following:

$ tree arch/mips/loongson/
arch/mips/loongson/
|-- 2e
|   |-- Makefile
|   |-- irq.c
|   `-- reset.c
|-- 2f
|   |-- 8250.c
|   |-- Makefile
|   |-- clock.c
|   |-- cs5536.h
|   |-- cs5536_pci.h
|   |-- cs5536_vsm.c
|   |-- irq.c
|   |-- mfgpt.c
|   |-- mipsdha.c
|   |-- pcireg.h
|   `-- reset.c
|-- Kconfig
`-- common
    |-- Makefile
    |-- bonito-irq.c
    |-- dbg_io.c
    |-- mem.c
    |-- pci.c
    |-- prom.c
    `-- setup.c

$ tree arch/mips/include/asm/mach-loongson/
arch/mips/include/asm/mach-loongson/
|-- cpu-feature-overrides.h
|-- dma-coherence.h
|-- mc146818rtc.h
|-- mem.h
|-- pci.h
|-- prom.h
`-- war.h

$ ls arch/mips/pci/*loongson*
arch/mips/pci/fixup-loongson2e.c  arch/mips/pci/fixup-loongson2f.c

a current version is released to git://dev.lemote.com/rt4ls.git,
(for avoid creating another git repository for it, i just use my RT_PREEMPT git tree instead, so, it may be very big :-( )

$ git clone git://dev.lemote.com/rt4ls.git
$ git checkout linux-2.6.29-stable-loongson --track
origin/linux-2.6.29-stable-loongson

TODO:

* clean up some of the commits carefully, especially the following
files:
arch/mips/pci/fixup-loongson2* arch/mips/loongson/2f/cs5536*
irq.c, reset.c

* try to merge the left files if possible.
* update it to the latest mainline kernel and push it in.

best regards,
Wu Zhangjin

Wu Zhangjin
DSLab, Lanzhou University, China
www.lemote.com, Jiangsu Province, China

Cool ! :)

[...]

You meant :

git checkout -b linux-2.6.29-stable-loongson --track origin/linux-2.6.29-stable-loongson

don't you ?

I can not compile it :

  CC      arch/mips/loongson/common/bonito-irq.o
  CC      arch/mips/loongson/common/mem.o
  CC      arch/mips/loongson/common/dbg_io.o

cc1: warnings being treated as errors arch/mips/loongson/common/dbg_io.c: In function �prom_printf arch/mips/loongson/common/dbg_io.c:178: error: the frame size of 1040 bytes is larger than 1024 bytes
make[1]: *** [arch/mips/loongson/common/dbg_io.o] Error 1
make: *** [arch/mips/loongson/common] Error 2

I have used the config file from here :
arch/mips/configs/yeeloong2f_defconfig

Cheers and thanks for your work !

Erwan

Yes, thanks for correcting this error, forget to add the -b option.

which gcc compiler do you use? gcc 4.4? i just try to compile it in gcc 4.4, get the same error. but i only test it in gcc 4.3(cross-compiler) before, only a few "trivial" warnings.

a new branch for gcc 4.4 will be created as quickly as i can.

really thanks for your encouragement, welcome to take part in, lots of source code lines need to be cleaned up.

Yeap :

gcc version 4.4.0-pre9999 built 20090212 (Gentoo SVN ebuild) rev. 144120 ()

I would be happy to help with the code but my C skills are need the "hello world" stuff :) I can help with testing and writing documentations.

Cheers
Erwan

a new branch for gcc 4.4 is created as linux-2.6.29-stable-loongson-gcc4.4, welcome to pull it.

$ git clone git://dev.lemote.com/rt4ls.git
$ git checkout -b linux-2.6.29-stable-loongson-gcc4.4 --track origin/linux-2.6.29-stable-loongson-gcc4.4

if you have cloned it, just update it and then checkout the branch

$ git pull
$ git checkout -b linux-2.6.29-stable-loongson-gcc4.4 --track origin/linux-2.6.29-stable-loongson-gcc4.4

thanks all the same :-)

Hello,

The kernel is compiling fine now. I have been using it for a few hours.

I don't understand why some stuff are not included in arch/mips/configs/yeeloong2f_defconfig,
for example the sound chip or the v4l stuff for the webcam.

I also had to :
- add INPUT_EDEV to get the mouse and the keyboard under X (1.6.99.1).
- compile external wifi modules from

http://www.lemote.com/upfiles/wifi/rtl8187B_linux_26.1049.1215.2008_release2.tar.gz

to get proper Wifi performances (or I had to sit on the access point)
- get and compile the ec_module stuff from git
- had the PPP and bluetooth support for my 3G connection that's works now with wvdial (thanks robert :))

The box is also complaining when it boots and try to set time :

xiwen ~ (n32) # hwclock --debug
hwclock from util-linux-ng 2.14.2
hwclock: Open of /dev/rtc failed, errno=2: No such file or directory. No usable clock interface found.
Cannot access the Hardware Clock via any known method.

I have seen some patches related to system clock in the git tree but i don't understand.

What has to be included in the config tree to get suspend/hibernate and cpu_freq working ?
It seems to be linked to the clocking nope ?

Cheers
Erwan

thanks for your work :-)

I just updated the default kernel configuration file for loongson2f based machines, hope it can help you :-)

the previous kernel configuration file not include the "Real Time Clock", so, no /dev/rtc there, so sorry :-)

* try the following configuration options:

Machine selection --->
[*] Using cs5536's MFGPT as system clock

Power management options --->
[*] Power Management support [*] Suspend to RAM and standby [*] Hibernation (aka 'suspend to disk') (/dev/hda3) Default resume partition

CPU Frequency scaling --->
[*] CPU Frequency scaling
[*] Loongson-2F CPU Frequency driver

* basic user manual(from www.lemote.com):

1. install a shell script

# apt-get install hibernate

2. modify the configuration file /etc/hibernate/common.conf

* find the "UnloadModules" section, modify it like this

UnloadModules r8187 usbhid ohci_hcd ehci_hcd

remove the # before "LoadModules auto"

* modify the "network" section

DownInterfaces eth0
UpInterfaces auto

* modify the "hardware_tweaks" section

remove the # before "FullSpeedCPU yes"

3. prepare a swap partition, by default, it is configured in kernel as /dev/hda3

change it to yours swap partition in kernel or configure it via /sys/power/resume, for example:

# fdisk -l | grep swap | cut -d' ' -f1
/dev/sda5
# ls -l /dev/sda5
brw-rw---- 1 root disk 8, 5 2009-04-10 10:26 /dev/sda5
# echo 8:5 > /sys/power/resume

4. resume

pass an argument "resume=/dev/hdaX" to kernel, /dev/hdaX is your swap partition.

5. try STD

# hibernate-disk

Yeap you have include many things now, it takes more time to compile but at least some basic stuff are not missing anymore :)

Does everybody has the same issue, I mean bad performances when using the rtl8187 module that is included in the kernel ?

Cool, i could even break my box from the CLI now :

Flash flash device: 80000 at 1fc00000
flash device: Found 1 x8 devices at 0x0 in 8-bit bank Amd/Fujitsu Extended Query Table at 0x0040
number of CFI chips: 1
cfi_cmdset_0002: Disabling erase-suspend-program due to code brokenness. Creating 1 MTD partitions on "flash device":
0x000000000000-0x000000080000 : "Bootloader"
pmon flash device initialized

:)

Yeah i'm stupid i should have config that...

emerge hibernate-script and configure the system like you have said. The box is reacting now.

There's no hibernate-disk on gentoo, just hibernate or hibernate-ram.

Starting suspend at Sun Apr 26 17:56:52 CEST 2009
hibernate: [01] Executing CheckLastResume ...
hibernate: [01] Executing CheckRunlevel ...
hibernate: [01] Executing LockFileGet ...
hibernate: [01] Executing NewKernelFileCheck ...
hibernate: [10] Executing EnsureSysfsPowerStateCapable ...
hibernate: [11] Executing XHacksSuspendHook1 ...
hibernate: [59] Executing RemountXFSBootRO ...
hibernate: [60] Executing NetworkStop ... Bringing down interface eth0
* Bringing down interface eth0
* Stopping ifplugd on eth0... [ ok ]
* Removing addresses
Bringing down interface wlan0
* Bringing down interface wlan0
* Stopping dhcpcd on wlan0... [ ok ]
* Removing addresses
hibernate: [89] Executing SaveKernelModprobe ... Saved /proc/sys/kernel/modprobe is /sbin/modprobe
hibernate: [90] Executing ModulesUnload ... Unloading module r8187...Removing modules with rmmod.

Unloading module usbhid...not loaded.
Unloading module ohci_hcd...
Unloading module ehci_hcd...not loaded.
Unloading module r8187...not loaded.
Unloading module usbhid...not loaded.
Unloading module ohci_hcd...not loaded.
Unloading module ehci_hcd...not loaded.
hibernate: [91] Executing ModulesUnloadBlacklist ... Unloading blacklisted modules listed /etc/hibernate/blacklisted-modules Module version for ipw2100 is
Module version for ipw2200 is
Module version for snd_bt_sco is
Module version for ndiswrapper is
Unloading blacklisted module uvcvideo (and dependencies) Unloading uvcvideo ...
hibernate: [91] Executing ModulesUnloadBlacklist ... Unloading blacklisted modules listed /etc/hibernate/blacklisted-modules Module version for ipw2100 is
Module version for ipw2200 is
Module version for snd_bt_sco is
Module version for ndiswrapper is
hibernate: [95] Executing XHacksSuspendHook2 ...
xhacks: changing console from 2 to 15
hibernate: [98] Executing CheckRunlevel ...
hibernate: [98] Executing FullSpeedCPUSuspend ...
Switched to performance, with min freq at 797000
hibernate: [99] Executing DoSysfsPowerStateSuspend ...
hibernate: Activating sysfs power state disk ...

But when the machine is rebooting, I can see :

hibernate: Activating sysfs power state disk...
+ '[' -n '' ']'
+ /bin/echo -n disk

and even switch from one vt to another but I cannot type anything.

The only solution is too reboot... Maybe I got to read more man 5 hibernate.conf :)

If I try the hibernate-ram script, no more chance :

xiwen ~ (n32) # hibernate-ram
hibernate-ram: Trying method in sysfs-ram.conf...
hibernate-ram: Including configuration from common.conf
hibernate-ram: No suitable suspend methods were found on your machine.
hibernate-ram: You need to install a kernel with support for suspending to
hibernate-ram: disk or RAM and reboot, then try again.

or

xiwen ~ (n32) # echo -n mem > /sys/power/state
bash: echo: write error: No such device

Now, let's talk about the cpufreq stuff :)

cpufrequtils 005: cpufreq-info (C) Dominik Brodowski 2004-2006 Report errors and bugs to [email protected], please. analyzing CPU 0:
driver: loongson2f
CPUs which need to switch frequency at the same time: 0 hardware limits: 199 MHz - 797 MHz
available frequency steps: 199 MHz, 299 MHz, 398 MHz, 498 MHz, 598
MHz, 697 MHz, 797 MHz
available cpufreq governors: conservative, ondemand, userspace,
powersave, performance
current policy: frequency should be within 199 MHz and 797 MHz.
The governor "ondemand" may decide which speed to use within this range.
current CPU frequency is 797 MHz (asserted by call to hardware). cpufreq stats: 199 MHz:0.00%, 299 MHz:0.00%, 398 MHz:0.00%, 498
MHz:0.00%, 598 MHz:0.00%, 697 MHz:0.00%, 797 MHz:100.00%

It seems to be working but it's weird. When I start X and gnome (cpufreq applet). I can see that's the system is using the ondemand performance but is stuck at 797 Mhz if I don't do anything. If i start working (yeah it's happening sometimes), the frequency is moving from 199Mhz to 797Mhz.

The other thing which is weird is that I don't have this problem with the Loonux stock kernel.

Btw, I have heard of an overcloking module ? is that working ? where's the source code ? I want to reach the speed of 900 Mhz :)

What about the power button ? Even on Loonux, the box is not asking for shutdown when I press it.

Do you know where I can find the source code of the binary for the osd stuff to work (/usr/bin/fnkey) ?

Sorry for this big mail, all this question (more or less stupid) :)

Cheers
r1

Hello,

I've switched from the ondemand governor to the conservative one and it seems to be working properly.

I have also noticed that the fans seems to be running at full speed all the time. Any comment, on this issue ?

Is there a way to query the sensors ?

Cheers
r1

Yes, the conservative governor is better than the ondemand.

we test that it only save power about 1~2w when the cpufreq is min.
So the temperature will be reduced a little.
when the temp is the some section,
eg. the 60°C and 65°C are the same section when the section is 60°C~70°C， the speed of fans will be the same.

Please insmod ec_ftd which is one of the ec modules.
And then type "cat /proc/ft" , the output: "1.31 0x01 4705 0x00 56"

the 4705 means the speed of fans, 56 stands for the temperature

the ec_modules source URL:
http://dev.lemote.com/code/ec_module

git URL:
git://dev.lemote.com/ec_module.git

---------------------------------------------------------
Hongbing,Hu (Software Department)
Tel:    0512-52308631
E-mail:	[email protected]
MSN:	[[email protected]]
JiangSu Lemote Corp. Ltd.
MengLan, Yushan, Changshu, JiangSu Province, China
---------------------------------------------------------