[lm-sensors] HT-1000 smbus

[lm-sensors] HT-1000 smbus

[Martin Devera]

Hello,
I've read ticket of someone requesting driver for subj.
I have Supermicro MB witj ADM1026 connected via HT-1000
and looking for its smbus driver too.
I might be able to write it but I'd like to ask for hints
how can one get DS from Broadcom :) Maybe someone from
lm-sensors has "better influence" when asking for it ?

I already tried PIIX4/older serverworks method but it no
longer works :( Also I can't find other driver/sw to reverse
engineer. I feel a bit lost ;-)

devik

----------------------------
Martin Devera, CDI computers
Linux kernel  HTB maintainer

[lm-sensors] HT-1000 smbus

[Rudolf Marek]

Martin Devera wrote:
> Hello,
> I've read ticket of someone requesting driver for subj.
> I have Supermicro MB witj ADM1026 connected via HT-1000
> and looking for its smbus driver too.
> I might be able to write it but I'd like to ask for hints
> how can one get DS from Broadcom :) Maybe someone from
> lm-sensors has "better influence" when asking for it ?

NBo influence sorry.

Well I have _NO_ time for this until end of next week, but I was curious and tired from studying so I have looked :)

here are the results:
ftp://ftp.supermicro.com/utility/Supero_Doctor_III/32-Bit_Edition/AllSuperD.ini

This software has an interesting file smbus.sys

   1051d:       3d 66 11 00 02          cmp    $0x2001166,%eax
   10522:       0f 84 9b 00 00 00       je     0x105c3
   10528:       3d 66 11 01 02          cmp    $0x2011166,%eax
   1052d:       0f 84 90 00 00 00       je     0x105c3
   10533:       3d 66 11 05 02          cmp    $0x2051166,%eax
   10538:       0f 84 85 00 00 00       je     0x105c3
   1053e:       3d 86 80 83 24          cmp    $0x24838086,%eax
   10543:       74 73                   je     0x105b8


HT1000 legacy southbridge should be same as:
0200	OSB4 South Bridge
0201	CSB5 South Bridge

let confirm by:
Please provide output of

lspci -v -v -v
and
lspci -n

http://pci-ids.ucw.cz/iii/?i\x1166

> I already tried PIIX4/older serverworks method but it no
> longer works :( Also I can't find other driver/sw to reverse
> engineer. I feel a bit lost ;-)


Maybe just adding the line with:
        { PCI_DEVICE(0x1166, 0x205),
          .driver_data = 0 },

Will do the trick. Assuming you have the "HT1000 Legacy South Bridge" If not we must find out where is the PCI id located and start from that point.

regards
Rudolf

[lm-sensors] HT-1000 smbus

[Martin Devera]

Hi Rudolf,

Rudolf Marek wrote:
> Martin Devera wrote:
>> Hello,
> Well I have _NO_ time for this until end of next week, but I was curious and tired from studying so I have looked :)

thanks to you and your curiosity :-)

> here are the results:
> ftp://ftp.supermicro.com/utility/Supero_Doctor_III/32-Bit_Edition/AllSuperD.ini
> 
> This software has an interesting file smbus.sys
> 
>    1051d:       3d 66 11 00 02          cmp    $0x2001166,%eax
>    10522:       0f 84 9b 00 00 00       je     0x105c3
>    10528:       3d 66 11 01 02          cmp    $0x2011166,%eax
>    1052d:       0f 84 90 00 00 00       je     0x105c3
>    10533:       3d 66 11 05 02          cmp    $0x2051166,%eax
>    10538:       0f 84 85 00 00 00       je     0x105c3
>    1053e:       3d 86 80 83 24          cmp    $0x24838086,%eax
>    10543:       74 73                   je     0x105b8
> 
> 
> HT1000 legacy southbridge should be same as:
> 0200	OSB4 South Bridge
> 0201	CSB5 South Bridge

excelent catch. I already looked into SuperoDoctor, but was not able to 
find this part. Hovewer I already assumed that it might be compatible 
and tested it yesterday (more below).

> let confirm by:
> Please provide output of
> 
> lspci -v -v -v
> and
> lspci -n
00:02.0 Host bridge: Broadcom HT1000 Legacy South Bridge
         Subsystem: Broadcom: Unknown device 0201
         Control: I/O+ Mem+ BusMaster+ SpecCycle- MemWINV- VGASnoop- 
ParErr- Stepping- SERR+ Fast B2B-
         Status: Cap- 66Mhz- UDF- FastB2B- ParErr- DEVSEL=medium 
 >TAbort- <TAbort- <MAbort- >SERR - <PERR-
         Latency: 64

-n will tell you that it IS 1166:0205.

> Maybe just adding the line with:
>         { PCI_DEVICE(0x1166, 0x205),
>           .driver_data = 0 },
> 
> Will do the trick. Assuming you have the "HT1000 Legacy South Bridge" If not we must find out where is the PCI id located and start from that point.

this is exactly what I did:
piix4_smbus 0000:00:02.0: Found 0000:00:02.0 device
piix4_smbus 0000:00:02.0: Unusual config register value
piix4_smbus 0000:00:02.0: Try using fix_hstcfg=1 if you experience problems
piix4_smbus 0000:00:02.0: Illegal Interrupt configuration (or code out 
of date)!
piix4_smbus 0000:00:02.0: SMBREV = 0x0
piix4_smbus 0000:00:02.0: SMBA = 0x580

after using fix_hstcfg=1, machine freeze...
If you will have again a long time (just interested, what faculty and 
year, FEL ?) I will not leave the effort un-appreciated ;-)
Going to dig into smbus.sys more for now.

Martin

[lm-sensors] HT-1000 smbus

[Rudolf Marek]

Hi,

> 00:02.0 Host bridge: Broadcom HT1000 Legacy South Bridge
>          Subsystem: Broadcom: Unknown device 0201
>          Control: I/O+ Mem+ BusMaster+ SpecCycle- MemWINV- VGASnoop- 
> ParErr- Stepping- SERR+ Fast B2B-
>          Status: Cap- 66Mhz- UDF- FastB2B- ParErr- DEVSEL=medium 
>  >TAbort- <TAbort- <MAbort- >SERR - <PERR-
>          Latency: 64
> 
> -n will tell you that it IS 1166:0205.

Hmm subsytem ID is 201 maybe it will say that it is in fact CSB5?

> this is exactly what I did:
> piix4_smbus 0000:00:02.0: Found 0000:00:02.0 device
> piix4_smbus 0000:00:02.0: Unusual config register value
> piix4_smbus 0000:00:02.0: Try using fix_hstcfg=1 if you experience problems
> piix4_smbus 0000:00:02.0: Illegal Interrupt configuration (or code out 
> of date)!
> piix4_smbus 0000:00:02.0: SMBREV = 0x0
> piix4_smbus 0000:00:02.0: SMBA = 0x580
> 
> after using fix_hstcfg=1, machine freeze...
> If you will have again a long time (just interested, what faculty and 
> year, FEL ?) 

Yes diagnose is FEL. One week left ;)

>I will not leave the effort un-appreciated ;-)
> Going to dig into smbus.sys more for now.

Good. I must study for final exams...

If you find anything I can take a look.

regards
Rudolf

[lm-sensors] HT-1000 smbus

[Martin Devera]

> 
>> I will not leave the effort un-appreciated ;-)
>> Going to dig into smbus.sys more for now.
> 
> Good. I must study for final exams...

good luck :) I just found it was working all the time only I was fooled
by I2C warnings so I even didn't test it. Now after I successfully 
disasembled and commented whole smbus.sys I found that it is really
working :-) Well dumb me ... but at least nice rev-engineering experience.
I owe you a beer, if you have some time, say, at Monday :)

Patch is attached. Is it enough to post it here to be considered ?

Martin
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: ht1000-smb.diff
Url: http://lists.lm-sensors.org/pipermail/lm-sensors/attachments/20060214/865c4149/ht1000-smb.pl

[lm-sensors] HT-1000 smbus

[Jean Delvare]

Hi Martin,

> Patch is attached. Is it enough to post it here to be considered ?

Well the fix is quite simple so yes I'll consider it. However there are
a few things I'd like to see added before I accept the patch.

Firstly, please include a Signed-off-by line (see
Documentation/SubmittingPatches for details.)

Secondly, some minimalistic Kconfig and documentation updates would be
needed to mention the HT-1000.

Also, I do agree that the warning about SMBHSTCFG triggers in many
cases where it shouldn't and doing what it suggests then breaks things
more than it helps. We might discard the warning as long as the
documentation is clear about the module parameter which "fixes" the
problem for people who have it (a couple boards only as far as I know -
and I consider it a BIOS bug.) However, this should be made a different
patch as this isn't related to the HT-1000.

Thanks,
-- 
Jean Delvare