Hi Rudolf,
On Wed, Nov 30, 2005 at 11:11:00PM +0100, Rudolf Marek wrote:
> Never give up, never surrender :) Lets focus on SMI. This should disable
> the SMI from this chip. So we can at least eliminate this source.
> OVT# is also unused so this should point if you have HW or SW (bus driver)
> problem.
That's the spirit :-)
> Please can you try with this sequence, which is same except of new first
> line.
>
> Quoting Jean:
>
> You can use the i2cset program to write to the W83792D chip directly. You
> can try using the following commands in your script, without loading the
> w83792d driver. Let us know if it hangs of not.
>
> i2cset -y 0 0x2c 0x40 0x1 # DISABLE SMI
> i2cset -y 0 0x2c 0x2b 0xff # in0_max
> ...
> This should completly disable the SMI. Lets see how it works :)
Ok, I've tried this and a few other things. First of all, I can't shut
off SMI which seems really odd.
The first time a ran this sequence I got this result and the machine
did not hang:
oden:~# ./set_safe_direct_wo_smi.sh
+ i2cset -y 0 0x2c 0x40 0x1
No size specified (using byte-data access)
Warning - data mismatch - wrote 0x01, read back 0x06
+ i2cset -y 0 0x2c 0x2b 0xff
No size specified (using byte-data access)
Value 0xff written, readback matched
+ i2cset -y 0 0x2c 0x2c 0x00
[all the remaining writes succeeded]
The second time I got this result and the machine did still not hang:
oden:~# ./set_safe_direct_wo_smi.sh
+ i2cset -y 0 0x2c 0x40 0x1
No size specified (using byte-data access)
Warning - data mismatch - wrote 0x01, read back 0x03
+ i2cset -y 0 0x2c 0x2b 0xff
No size specified (using byte-data access)
Value 0xff written, readback matched
+ i2cset -y 0 0x2c 0x2c 0x00
[all the remaining writes succeeded]
Finally the third time the machine just got this far and then hung:
oden:~# ./set_safe_direct_wo_smi.sh
+ i2cset -y 0 0x2c 0x40 0x1
No size specified (using byte-data access)
So according to the datasheet writing 0x01 to this register should
shut off EN_SMI, but that doesn't seem to work since the readback
still shows 0x03 (should be 0x01, right?). I've been trying all sorts
of things to disable this EN_SMI bit but I can't seem to do it.
For example, writing 0x80 should re-init the whole chip:
oden:~# i2cset -y 0 0x2c 0x40 0x80
No size specified (using byte-data access)
Warning - data mismatch - wrote 0x80, read back 0x03
But that still has the EN_SMI bit set??? It shouldn't according to my
datasheet at least, that bit should be reset to '0'...
I've also played around in the BIOS for a while, there seems to be
some king of "MiniBMC" functionality in the BIOS that is logging
overtemp events someplace. That log was filled with events since BIOS
enabled all fan monitors by default. I've shut off all those monitors
and even the whole MiniBMC funtionality but I still get the same hangs
when writing the sensor limits. I honestly don't know if there is some
other device on this bus that is communicating with the w82792d chip
to read these levels as well that might be interfering here. Since
this MiniBMC thing might be of interest I actually wrote Gigabyte
support about this whole issue, I'm keeping my fingers crossed but I
doubt that the question will reach the appropriate people at Gigabyte.
One more thing I observed;
oden:~# i2cdump -y 0 0x2c
No size specified (using byte-data access)
0 1 2 3 4 5 6 7 8 9 a b c d e f 0123456789abcdef
00: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 ................
10: 00 00 64 64 bf 00 00 63 00 ff 00 00 f5 00 8f 80 ..dd?..c....?.??
20: d3 d4 b9 b9 b8 b8 cd 15 ff ff ff ff 00 96 64 d7 ????????.....?d?
30: af d9 b1 d9 b1 d7 af ff 00 25 00 f0 f0 f0 2f 39 ???????..%.???/9
40: 03 00 00 00 00 ff ff 11 2c 13 40 c3 51 ff 80 5c ?......?,?@?Q.?\
50: ff ff ff ff ff ff ff ff 7a 60 ff 11 11 c1 05 7f ........z`.?????
60: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ................
70: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ................
80: 01 8f 01 8f 00 00 00 00 11 11 ff ff 3c 3c 0a 0a ????....??..<<??
90: 00 00 00 01 8f ff 00 00 11 ff 3c 00 00 01 01 ff ...??...?.<..??.
a0: 01 01 01 8f 8f 8f 8f ff 3e e2 00 e0 ff ff 00 00 ???????.>?.?....
b0: d3 cb ff ff ff 00 e2 b9 ff ff ff f0 f0 f0 ff ff ??....??...???..
c0: f2 80 02 39 00 3c 00 ff 2c 00 00 5d 00 60 00 ff ???9.<..,..].`..
d0: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ................
e0: 8b 8b 8b 28 3c 50 28 3c 50 28 3c 50 ff ff ff ff ???(<P(<P(<P....
f0: ff ff 80 ff 00 01 00 00 ff 02 ff 00 00 00 00 ff ..?..?...?......
Look at index 49, it says "13". According to the datasheet I have a
revision C chip should have "12". Is this a newer chip then what the
datasheet covers? In this case there might be some confusion on
disabling SMI...
Yuan, could you offer some help and guidance as to why the EN_SMI bit
can't be turned off and what this revision "13" chip is? My
motherboard is a Gigabyte GA-4MXSV by the way.
Regards
--
Daniel Nilsson
