it8718 supports 5 fans

[Paul.Vriens@…]

Hi,

Apparently the it8718f supports upto 5 fans. My mainboard (Gigabyte GA-P35-DS4) has 5 fan connectors and I actually have 3 connected. One of them on a connector which isn't covered by the 3 fans that the it87 covers.

I'm not sure this is the right place as the kernel module also doesn't cover 5 fans.

Cheers,

Paul

[ticket]

Ouch, the summary should read 'it8718' not 'it8717'.

[ticket]

And I think 'Enhancement' should be 'Enhancement needed' :-), so it's a 'defect'.

[khali]

Summary fixed.

There is no bug in the driver, you are asking for an additional feature, so "enhancement" is the right category and "defect" isn't.

Note that on some motherboards, the additional fans are not handled by the IT8716F or IT8718F but by a second hardware monitoring chip on the SMBus. You should make sure this isn't the case for you as well.

If you only have 3 fans, a quick workaround would be to connect them to the 3 fan headers which are already supported by the it87 driver.

[ticket]

The workaround is of course possible, although I hate workarounds (and I have to buy another cable, because it's too short to reach the other connector). I'm not sure that this motherboard uses the 8718 for all fans, but it's a wild guess. Is there way to test this (without opening the case). What are the 'problems' if the it87 driver was extended to 5 fans (maybe only for the it8718)?

[khali]

Replying to ticket:

I'm not sure that this motherboard uses the 8718 for all fans, but it's a wild guess. Is there way to test this (without opening the case).

A register dump would tell. Assuming that your chip's hardware monitoring function lives at 0x290, the command would be:

isadump 0x295 0x296

If your address is different, adjust the command.

What are the 'problems' if the it87 driver was extended to 5 fans (maybe only for the it8718)?

No particular problem, just that I don't know when I (or someone else) will have the time to implement support for the two additional fans.

The IT8716F and late revisions of the IT8712F support up to 5 fans too.

[ticket]

Here is the output for 'isadump 0x295 0x296'

     0  1  2  3  4  5  6  7  8  9  a  b  c  d  e  f
00: 13 10 10 00 ff ff 00 37 ff 87 54 09 17 e3 ff 62 
10: ff ff ff 73 d7 81 80 7f 02 ff 01 ff ff ff ff ff 
20: 48 76 cd b5 00 03 03 c2 c2 22 1c fe 80 f0 6c 6c 
30: ff 00 ff 00 ff 00 ff 00 ff 00 ff 00 ff 00 ff 00 
40: 7f 7f 7f 7f 7f 7f 5f 74 2d 40 9c 22 ff ff ff ff 
50: ff 2a 7f 7f 7f 50 fd fd 90 fd 37 12 60 00 00 00 
60: 00 14 7f 23 90 03 ff ff 00 14 41 23 90 03 ff ff 
70: 00 14 41 23 90 03 ff ff ff ff ff ff ff ff ff ff 
80: 19 02 00 00 ff ff 00 00 00 00 00 00 00 00 02 00 
90: 41 ff 22 00 41 7f 22 00 ff ff ff ff ff ff ff ff 
a0: 00 00 00 00 00 00 00 ff ff ff ff ff ff ff ff ff 
b0: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff 
c0: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff 
d0: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff 
e0: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff 
f0: ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff 

It doesn't seem trivial to add the 2 fans to the driver as (from the datasheet) it appears to not be in line with the way the other 3 are treated. Kernel-hacking is not one of my specialties.

[khali]

The dump shows that fan4 is in use (1256 RPM) and fan5 isn't.

Adding support shouldn't be too difficult, the it87 driver in 2.6.19 and later already supports 16-bit fan values. The only difficulty is that the fan4 and fan5 pins can be used for other functions, so we can't just create fan4 and fan5 if the chip supports them. We need to make sure they are really available. And for the IT8712F, we also need to check the chip revision first.

[ticket]

Ok, in my case that must mean (Gigabyte GA-P35-DS4):

CPU_FAN: fan1
SYS_FAN2 or NB_FAN: fan2
PWR_FAN: fan3
SYS_FAN1: fan4
NB_FAN or SYS_FAN2: fan5

If needed I can check by changing some cables.

[khali]

Is the BIOS reporting 5 fan speeds? I'm asking because your fan5 is disabled in the IT8718F, so I'd be quite surprised it the BIOS was able to report it.

[khali]

User-space support for it87 fan4 and fan5

[khali]

I attached one kernel patch and one user-space patch. I can't test them but they should give you your missing fan speed reading, with low limit and alarm. Please try it and report.

[ticket]

Thanks, I will see if I can test them :-). You're right btw fan5 (NB_FAN) is not reported in the BIOS.

[khali]

North bridge fans are often two-wire fans, so they can't be monitored :-(

[ticket]

I've tried to patch against current kernel GIT but it fails. Most of the lines you change in you're patch have 'data' instead of 'client'. For example part of you're patch:

@@ -555,7 +555,7 @@ static ssize_t set_fan_min(struct device
        }

        data->fan_min[nr] = FAN_TO_REG(val, DIV_FROM_REG(data->fan_div[nr]));
-       it87_write_value(data, IT87_REG_FAN_MIN(nr), data->fan_min[nr]);
+       it87_write_value(data, IT87_REG_FAN_MIN[nr], data->fan_min[nr]);
        mutex_unlock(&data->update_lock);
        return count;

where 'my' it87.c contains:

        data->fan_min[nr] = FAN_TO_REG(val, DIV_FROM_REG(data->fan_div[nr]));
        it87_write_value(client, IT87_REG_FAN_MIN(nr), data->fan_min[nr]);
        mutex_unlock(&data->update_lock);
        return count;

Am I missing something?

[khali]

My patch is on top of the latest -mm, not the latest Linus' tree. Sorry for not telling you. So you're missing these two patches to the it87 driver:

http://lm-sensors.org/kernel?p=kernel/mhoffman/hwmon-2.6.git;a=commitdiff;h=6d8cc2c5f3b69870c49419a992d415a87d8b3536

http://lm-sensors.org/kernel?p=kernel/mhoffman/hwmon-2.6.git;a=commitdiff;h=6e38cccf5557263b3a23b37132e5977118b999c6

Apply these first, and then my patch, and it should work.

[ticket]

I patched and build the kernel. I've also downloaded the latest snapshot of lm_sensors and added you're patch. I do have a fan4 now (and as expected no fan5), but with a speed of -1, so something is missing. I'll have another look in the code.

[khali]

Support for fan4 and fan5 in it87

[khali]

Sorry, one change was missing. I've updated the patch, please try again.

[ticket]

That's exactly what I've just changed :-). Now compiling.

[ticket]

Yep, that did the trick. Thanks very much for your fast patches. I'm as always amazed by the fact how fast things can go in an 'open' world.

Now I just have to wait to wait till everything ends up in the main F7 packages :-).

Thanks again.

[khali]

User-space part was committed to lm-sensors SVN (changeset r4565) so it'll be in 2.10.4.

Kernel patch was submitted for inclusion:
 http://lists.lm-sensors.org/pipermail/lm-sensors/2007-July/020303.html

If you insist on showing your appreciation, I can point you to my wishlist :)
 http://khali.linux-fr.org/wishlist.html

[ticket]

(I'm not sure if I should re-open, so sorry if I shouldn't). I've been watching the current kernel (2.6.13-rc5 and the latest one from Andrew Morton 2.6.23-rc4-mm1). The patch doesn't seem to be in. Any idea?

[khali]

The problem is that the hwmon subsystem maintainer did not pick my patch yet. Presumably he is waiting for someone to review it, but nobody volunteered to do so. I've just sent it again, let's see what happens.

[khali]

My patch was accepted by the hwmon subsystem maintainer and is staging in -mm now.