For the temperature sensing drivers we have *intentionally* avoided doing
writes to i2c devices.
On many machines, ACPI/SMI retain control of these devices, and make
changes behind OpenBSD's back without us knowing what is going on. We
don't know if this will break a machine, do to their changes depending
up on a specific configuration. The breakage could be subtle and very
weird. This is a policy we established for all these drivers, and we've
avoided creating too many exceptions to it because if the exception
becomes the rule, subtle interactions could create harm.
As a result, all sensor drivers for these chips only do 'sensing',
meaning they only do reads. (OK, a few drivers do writes to "enable"
them, if we discover that the system has not yet enabled them, then it
means there is no BIOS driving behind our back; this situation feels
more safe, then OpenBSD is in control).
That said, your argument is for macppc. I would not argue against a
macppc-only configuration change, especially if it is clearly from a
"chip not configured state"), but I don't know what that code would look
like, whether it is an #ifdef or run-time comparison of the machine
architecture and model.
jon@elytron.openbsd.amsterdam wrote:
> Hello. It's been about 6 months since I tweaked adt(4) to control
> the temperature target that the cooling system uses to regulate
> itself, following the source code of G4FanControl which may be found
> around online.
>
> This makes it possible to power the fans on earlier and thus improve
> thermals and, hopefully, lifespan of the system which -given these
> macppc machines are quite old and increasingly rare- I would argue
> is worth preserving. I myself find it quite comfortable at 46 degC
>
> Perhaps it would be better to add a sysctl to control the temperature
> target? This diff is a fully functional proof of concept I guess,
> would really appreciate your thoughts.
>
> All the best
>
> Index: dev/i2c/adt7460.c
> ===================================================================
> RCS file: /cvs/src/sys/dev/i2c/adt7460.c,v
> retrieving revision 1.22
> diff -u -p -r1.22 adt7460.c
> --- dev/i2c/adt7460.c 6 Apr 2022 18:59:28 -0000 1.22
> +++ dev/i2c/adt7460.c 23 Nov 2023 16:23:48 -0000
> @@ -34,6 +34,9 @@
> #define ADT7460_REM1_TEMP 0x25
> #define ADT7460_LOCAL_TEMP 0x26
> #define ADT7460_REM2_TEMP 0x27
> +#define ADT7460_REM1_TEMP_TRG 0x68
> +#define ADT7460_LOCAL_TEMP_TRG 0x67
> +#define ADT7460_REM2_TEMP_TRG 0x69
> #define ADT7460_TACH1L 0x28
> #define ADT7460_TACH1H 0x29
> #define ADT7460_TACH2L 0x2a
> @@ -193,6 +196,16 @@ adt_attach(struct device *parent, struct
> sc->sc_addr, &cmd, sizeof cmd, &data, sizeof data, 0)) {
> iic_release_bus(sc->sc_tag, 0);
> printf(": cannot set control register\n");
> + return;
> + }
> + }
> +
> + for (cmd = ADT7460_LOCAL_TEMP_TRG; cmd <= ADT7460_REM2_TEMP_TRG; cmd++){
> + data = 46; /* temperature target in deg C*/
> + if (iic_exec(sc->sc_tag, I2C_OP_WRITE_WITH_STOP,
> + sc->sc_addr, &cmd, sizeof cmd, &data, sizeof data, 0)) {
> + iic_release_bus(sc->sc_tag, 0);
> + printf(": cannot set temperature target register\n");
> return;
> }
> }
>
