headers
Bruce Allen | 1 Mar 2006 17:14
Picon
Favicon

HT1000 support (see lm_sensors ticket 2031, new chips page)

I'd like to get lm_sensors support for a system with an HT1000 chip.
References are on http://www2.lm-sensors.nu/~lm78/newdrivers.html

-- I've rebuilt a stock 2.6.15.4 kernel using the HT1000 patch referenced
    on the newdrivers page:
    http://khali.linux-fr.org/devel/i2c/linux-2.6/i2c-piix4-add-ht1000-support.patch

-- I then built and installed sysfsutils-2.0.0

-- I then built and installed lm_sensors-2.10.0.

-- I loaded the i2c_piix4 module using modprobe.

-- I mounted sysfs on /sysfs

However sensors-detect does not find ANYTHING.  Not even eeprom!  Am I 
doing something stupid?

(Note: for kicks I also went into the lm_sensors source and 'hacked' the 
SERVERWORKS CSB6 0x0203 PCI-id to be 0x0205 both the sensors-detect script 
and in the i2c_piix4 code contained there. It didn't help.)

One other possibility is that because I have a BMC card plugged in to the 
motherboard (SuperMicro H8SSL-i) the BMC card is not allowing access to 
the devices from lm_sensors.

Cheers,
 	Bruce

_______________________________________________
(Continue reading)

Permalink | Reply |
headers
Martin Devera | 1 Mar 2006 21:23
Picon
Favicon

Re: HT1000 support (see lm_sensors ticket 2031, new chips page)

> One other possibility is that because I have a BMC card plugged in to 
> the motherboard (SuperMicro H8SSL-i) the BMC card is not allowing access 
> to the devices from lm_sensors.

it works well for me for the same MB, only loaded patched piix4 and then
adm1026 module. Then even supermicro's superodoctor started to work on 
the bow :) But I can imagine that it may become unavailable after BMC is 
plugged in. I havn't bought BMC yet but I want to.

By the way, Jean, is it possible to have two masters on one i2c bus ? 
How are priorities handled then ?

devik

_______________________________________________
lm-sensors mailing list
lm-sensors <at> lm-sensors.org
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors
Permalink | Reply |
headers
Mark M. Hoffman | 2 Mar 2006 05:45

Re: HT1000 support (see lm_sensors ticket 2031, new chips page)

Hi Martin:

* Martin Devera <devik <at> cdi.cz> [2006-03-01 21:23:54 +0100]:
> > One other possibility is that because I have a BMC card plugged in to 
> > the motherboard (SuperMicro H8SSL-i) the BMC card is not allowing access 
> > to the devices from lm_sensors.
> 
> it works well for me for the same MB, only loaded patched piix4 and then
> adm1026 module. Then even supermicro's superodoctor started to work on 
> the bow :) But I can imagine that it may become unavailable after BMC is 
> plugged in. I havn't bought BMC yet but I want to.
> 
> By the way, Jean, is it possible to have two masters on one i2c bus ? 
> How are priorities handled then ?

The i2c bus is multi-master capable.  There is no defined priority scheme.
However, there is an arbitration protocol that determines the outcome in
case two or more masters begin a transfer at the same time.  In a nutshell:
the transfer that is *not* corrupted wins.  It's simple and (IMHO) elegant.

I refer you to the I2C bus specification for more details, esp section 8.2:

http://www.semiconductors.philips.com/acrobat/literature/9398/39340011.pdf

If a Linux host I2C master loses arbitration, the transfer fails and an
error status is returned to the slave device driver.  Many (but not all)
hwmon drivers will retry the transfer several times before giving up.

Regards,
(Continue reading)

Permalink | Reply |
headers
Bruce Allen | 1 Mar 2006 21:53
Picon
Favicon

Re: HT1000 support (see lm_sensors ticket 2031, new chips page)

>> One other possibility is that because I have a BMC card plugged in to the 
>> motherboard (SuperMicro H8SSL-i) the BMC card is not allowing access to the 
>> devices from lm_sensors.
>
> it works well for me for the same MB, only loaded patched piix4 and then
> adm1026 module. Then even supermicro's superodoctor started to work on the 
> bow :) But I can imagine that it may become unavailable after BMC is plugged 
> in. I havn't bought BMC yet but I want to.

I loaded the adm1026 module by hand:

[root <at> nemo-slave0017 ~]# lsmod | egrep '(i2c|adm)'
adm1026                46736  0
i2c_piix4              14480  0
adm1025                27156  0
hwmon_vid               7040  2 adm1026,adm1025
i2c_isa                10624  0
i2c_dev                17408  0
i2c_core               30720  6 adm1026,i2c_piix4,adm1025,eeprom,i2c_isa,i2c_dev

I have:
[root <at> nemo-slave0017 ~]# ls /sysfs
block  bus  class  devices  firmware  kernel  module  power
[root <at> nemo-slave0017 ~]# ls /sys
block  bus  class  devices  firmware  kernel  module  power

sensors-detect reports: "Sorry, no chips were detected."

[root <at> nemo-slave0017 ~]# sensors
Can't access procfs/sysfs file
(Continue reading)

Permalink | Reply |
headers
Martin Devera | 1 Mar 2006 21:57
Picon
Favicon

Re: HT1000 support (see lm_sensors ticket 2031, new chips page)

> I loaded the adm1026 module by hand:
> 
> [root <at> nemo-slave0017 ~]# lsmod | egrep '(i2c|adm)'
> adm1026                46736  0
> i2c_piix4              14480  0
> adm1025                27156  0
> hwmon_vid               7040  2 adm1026,adm1025
> i2c_isa                10624  0
> i2c_dev                17408  0
> i2c_core               30720  6 
> adm1026,i2c_piix4,adm1025,eeprom,i2c_isa,i2c_dev
> 
> I have:
> [root <at> nemo-slave0017 ~]# ls /sysfs
> block  bus  class  devices  firmware  kernel  module  power
> [root <at> nemo-slave0017 ~]# ls /sys
> block  bus  class  devices  firmware  kernel  module  power
> 

send
dmesg|tail -n 30
and try
ls /sys/class
(there is typicaly hwmon class, if yes, list it too)

_______________________________________________
lm-sensors mailing list
lm-sensors <at> lm-sensors.org
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors
(Continue reading)

Permalink | Reply |
headers
Bruce Allen | 1 Mar 2006 23:07
Picon
Favicon

Re: HT1000 support (see lm_sensors ticket 2031, new chips page)

>> I loaded the adm1026 module by hand:
>> 
>> [root <at> nemo-slave0017 ~]# lsmod | egrep '(i2c|adm)'
>> adm1026                46736  0
>> i2c_piix4              14480  0
>> adm1025                27156  0
>> hwmon_vid               7040  2 adm1026,adm1025
>> i2c_isa                10624  0
>> i2c_dev                17408  0
>> i2c_core               30720  6 
>> adm1026,i2c_piix4,adm1025,eeprom,i2c_isa,i2c_dev
>> 
>> I have:
>> [root <at> nemo-slave0017 ~]# ls /sysfs
>> block  bus  class  devices  firmware  kernel  module  power
>> [root <at> nemo-slave0017 ~]# ls /sys
>> block  bus  class  devices  firmware  kernel  module  power
>> 
>
> send
> dmesg|tail -n 30

[root <at> nemo-slave0017 ~]# modprobe i2c_piix4
[root <at> nemo-slave0017 ~]# modprobe adm1026
[root <at> nemo-slave0017 ~]# dmesg | tail -n 30
md: ... autorun DONE.
device-mapper: 4.4.0-ioctl (2005-01-12) initialised: dm-devel <at> redhat.com
cdrom: open failed.
cdrom: open failed.
EXT3 FS on sda3, internal journal
(Continue reading)

Permalink | Reply |
headers
Martin Devera | 1 Mar 2006 23:10
Picon
Favicon

Re: HT1000 support (see lm_sensors ticket 2031, new chips page)

>  IPMI kcs interface initialized
> ipmi device interface
> eth0: no IPv6 routers present

interestingly there is no info from piix4 module...

>> and try
>> ls /sys/class
>> (there is typicaly hwmon class, if yes, list it too)
> 
> [root <at> nemo-slave0017 ~]# ls /sys/class/
> cpuid  firmware  graphics  hwmon  i2c-adapter  i2c-dev  input  ipmi  mem 
> misc  msr  net  pci_bus  printer  scsi_device  scsi_host  tty  usb 
> usb_device  usb_host  vc
> [root <at> nemo-slave0017 ~]# ls /sys/class/hwmon/
> [root <at> nemo-slave0017 ~]# ls /sys/class/i2c-adapter/
> i2c-9191
> [root <at> nemo-slave0017 ~]# ls /sys/class/i2c-dev/
> [root <at> nemo-slave0017 ~]#
> 

Hmm, it really seems as if BMC "disconnected" i2c bus from HT1000 bus. 
It would be interesting to try with BMC withdrawn. I'll test once I get 
the BMC.
devik

_______________________________________________
lm-sensors mailing list
lm-sensors <at> lm-sensors.org
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors
(Continue reading)

Permalink | Reply |
headers
Bruce Allen | 2 Mar 2006 03:12
Picon
Favicon

Re: HT1000 support (see lm_sensors ticket 2031, new chips page)

>>  IPMI kcs interface initialized
>> ipmi device interface
>> eth0: no IPv6 routers present
>
> interestingly there is no info from piix4 module...
>
>>> and try
>>> ls /sys/class
>>> (there is typicaly hwmon class, if yes, list it too)
>> 
>> [root <at> nemo-slave0017 ~]# ls /sys/class/
>> cpuid  firmware  graphics  hwmon  i2c-adapter  i2c-dev  input  ipmi  mem 
>> misc  msr  net  pci_bus  printer  scsi_device  scsi_host  tty  usb 
>> usb_device  usb_host  vc
>> [root <at> nemo-slave0017 ~]# ls /sys/class/hwmon/
>> [root <at> nemo-slave0017 ~]# ls /sys/class/i2c-adapter/
>> i2c-9191
>> [root <at> nemo-slave0017 ~]# ls /sys/class/i2c-dev/
>> [root <at> nemo-slave0017 ~]#
>> 
>
> Hmm, it really seems as if BMC "disconnected" i2c bus from HT1000 bus. It 
> would be interesting to try with BMC withdrawn. I'll test once I get the BMC.

I'll give this a try tomorrow also.  It's easy to yank out the BMC.

Cheers,
 	Bruce

_______________________________________________
(Continue reading)

Permalink | Reply |
headers
Jean Delvare | 1 Mar 2006 20:19
Gravatar

Re: HT1000 support (see lm_sensors ticket 2031, new chips page)

Hi Bruce,

> I'd like to get lm_sensors support for a system with an HT1000 chip.
> References are on http://www2.lm-sensors.nu/~lm78/newdrivers.html
> 
> -- I've rebuilt a stock 2.6.15.4 kernel using the HT1000 patch referenced
>     on the newdrivers page:
>     http://khali.linux-fr.org/devel/i2c/linux-2.6/i2c-piix4-add-ht1000-support.patch
> 
> -- I then built and installed sysfsutils-2.0.0
> 
> -- I then built and installed lm_sensors-2.10.0.
> 
> -- I loaded the i2c_piix4 module using modprobe.
> 
> -- I mounted sysfs on /sysfs
> 
> However sensors-detect does not find ANYTHING.  Not even eeprom!  Am I 
> doing something stupid?

The HT-1000 is (from our simplified point of view) an SMBus master. It
may give you access to I2C/SMBus hardware monitoring chips, but isn't
itself a hardware monitoring chip.

> (Note: for kicks I also went into the lm_sensors source and 'hacked' the 
> SERVERWORKS CSB6 0x0203 PCI-id to be 0x0205 both the sensors-detect script 
> and in the i2c_piix4 code contained there. It didn't help.)

The i2c-piix4 driver found in the lm_sensors is a Linux 2.4 driver, so
it is of no use to you.
(Continue reading)

Permalink | Reply |
headers
Bruce Allen | 1 Mar 2006 22:03
Picon
Favicon

Re: HT1000 support (see lm_sensors ticket 2031, new chips page)

>> -- I've rebuilt a stock 2.6.15.4 kernel using the HT1000 patch referenced
>>     on the newdrivers page:
>>     http://khali.linux-fr.org/devel/i2c/linux-2.6/i2c-piix4-add-ht1000-support.patch
>>
>> -- I then built and installed sysfsutils-2.0.0
>>
>> -- I then built and installed lm_sensors-2.10.0.
>>
>> -- I loaded the i2c_piix4 module using modprobe.
>>
>> -- I mounted sysfs on /sysfs
>>
>> However sensors-detect does not find ANYTHING.  Not even eeprom!  Am I
>> doing something stupid?
>
> The HT-1000 is (from our simplified point of view) an SMBus master. It
> may give you access to I2C/SMBus hardware monitoring chips, but isn't
> itself a hardware monitoring chip.

OK.  I tried loading the adm1026 module that Martin uses -- no go.
>
>> (Note: for kicks I also went into the lm_sensors source and 'hacked' the
>> SERVERWORKS CSB6 0x0203 PCI-id to be 0x0205 both the sensors-detect script
>> and in the i2c_piix4 code contained there. It didn't help.)
>
> The i2c-piix4 driver found in the lm_sensors is a Linux 2.4 driver, so
> it is of no use to you.

OK.  In any case I am loading the i2c-piix4 modules from the (modified) 
2.6.15 tree.
(Continue reading)

Permalink | Reply |

Gmane