On README.p4b?

Mark D. Studebaker mds at paradyne.com
Sat Jun 28 21:14:00 CEST 2003


It's bits 8 and 3.
You've found some bugs. Thanks.
I'll fix the code and the docs in CVS.

Yoshifumi R. Shimizu wrote:
> Dear Developers of the lm_sensor project:
> 
>       My name is Yoshifumi Shimizu.  I have a question about the README.p4b,
> which is on the SMBus problem of ASUS Pentimum4 motherbaords.
> 
>       In the "README.p4b" file, it says:
> 
> ----------------------------------------------------------------------------
> 
>>It turnes off (!) the bits number 8 and 3 in the LPC register of the ICH2.
>>This you can also try as root with setpci and pcitweak.
>>First do a:
>>
>>setpci -d 8086:2440 f2.w
>>
>>You will get an hex number VAL - in my case 148 (Hex)-, where you have to 
>>erase the two bits - in my case you get 40 (Hex).
>>Then do a:
>>
>>setpci -d 8086:2440 f2.w=VAL
>>(snip)
> 
> ----------------------------------------------------------------------------
> 
> while in the "p4b_smbus.c"
> 
> ----------------------------------------------------------------------------
> 
>>/* 
>>* Checks whether SMBus is enabled and turns it on in case they are not. 
>>* It's done by clearing Bit 8 and 4 in i801 config space F2h, PCI-Device 0x8086:0x2440(ICH2)/0x24c0(ICH4)
>>*/
>>static int
>>i801smbus_enable(struct pci_dev *dev){
>>	u16  val   = 0;
>>
>>	pci_read_config_word(dev, 0xF2, &val);
>>	DBG("i801smbus: i801smbus config byte reading 0x%X.\n", val);
>>	if (val & 0x008) {
>>		pci_write_config_word(dev, 0xF2, val & 0x77);
>>		pci_read_config_word(dev, 0xF2, &val);
>>(snip)
> 
> ----------------------------------------------------------------------------
> 
> So it seems that the content of two are conflicting:
> 
>    README.p4b        --->   bits 8 and 3 (counting from 0)
>    p4b_smbus.c       --->   bits 7 and 3 (counting from 0)
> 
> which is correct?   This is my first question.  The second is, of course,
> related to the first:  If what is written in "p4b_smbus.c" is correct,
> then the data is byte, so that it should be:
> 
>         pci_write_config_byte(dev, 0xF2, val & 0x77);
>                          ^^^^
> 
>         pci_read_config_byte(dev, 0xF2, &val);
>                         ^^^^
> in stead of "***_word()".  And then, in the README.p4b, one has to change?:
> 
>         setpci -d 8086:2440 f2.b=VAL
>                               ^^
> (for this, I'm not sure.  Is it correct?  I am not a linux user!  Please,
> let me know if it's correct).  If what is written in "README.p4b" is
> correct, on the other hand, then the data is word, but, of course, then
> the mask should be changed:
> 
>         pci_write_config_word(dev, 0xF2, val & 0xFEF7);
>                                                ^^^^^^
> What is correct?
> 
> 
>      I am developing a motherboard monitor program for FreeBSD
> (it also runs on Linux box, though), named "xmbmon/mbmon", which is
> a userland program and not of device driver type like lm_sensors.
> 
>    http://www.nt.phys.kyushu-u.ac.jp/shimizu/download/download.html
> 
> Your project is a great help for developing the program. I appreciate
> it so much.
> 
>      Now, many users of ASUS Pentium4 motherboards had the same
> problem when using "xmbmon/mbmon", so I want to write down a kind of
> FAQ for it.  So I looked into the your project files.  But, because of
> the inconsistencies mentioned above, I cannot write down.  I cannot
> either check by myself, since I don't have any ASUS Pen4 motherboards,
> so I hope to clear out what is correct by asking you.  Thank you.
> 
>      With best regards,                   Yoshifumi R. Shimizu
> 



More information about the lm-sensors mailing list