[LinuxBIOS] alix 1c is stuck on this problem

[ron minnich]

it always turns off in 4 seconds.

I have been told to look at this:
6.18.3.14 PM Fail-Safe Delay and Enable (PM_FSD)
PMS I/O Offset     40h
Type               R/W
Reset Value        00000000h
Reads always return the value written, except for RSVD bits [29:20].

But I can't see that we're setting it anywhere. Any good ideas welcome :-)

ron

[Jordan Crouse]

On 15/10/07 09:11 -0700, ron minnich wrote:
> it always turns off in 4 seconds.
>
> I have been told to look at this:
> 6.18.3.14 PM Fail-Safe Delay and Enable (PM_FSD)
> PMS I/O Offset     40h
> Type               R/W
> Reset Value        00000000h
> Reads always return the value written, except for RSVD bits [29:20].
>
> But I can't see that we're setting it anywhere. Any good ideas welcome :-)

What are you trying to do?

Jordan

[ron minnich]

On 10/15/07, Jordan Crouse  wrote:
> What are you trying to do?

I load linuxbios, and alix1c runs for 3-4 seconds and turns off. I am
convinced it is something in PM hardware, just don't know what.

Here are the settings we make
static void cs5536_setup_power_button(void)
{
    /*      Power Button Setup */
    outl(0x40020000, PMS_IO_BASE + 0x40);

/* setup WORK_AUX/GPIO24, it is the external signal for 5536
     * vsb_work_aux controls all voltage rails except Vstandby & Vmem.
     * We need to enable, OUT_AUX1 and OUTPUT_ENABLE in this order.
     * If WORK_AUX/GPIO24 is not enabled then soft-off will not work.
     */
    outl(GPIOH_24_SET, GPIO_IO_BASE + GPIOH_OUT_AUX1_SELECT);
    outl(GPIOH_24_SET, GPIO_IO_BASE + GPIOH_OUTPUT_ENABLE);

}

ron

[Marc Jones]

ron minnich wrote:
> On 10/15/07, Jordan Crouse  wrote:
> > What are you trying to do?
>
> I load linuxbios, and alix1c runs for 3-4 seconds and turns off. I am
> convinced it is something in PM hardware, just don't know what.
>
> Here are the settings we make
> static void cs5536_setup_power_button(void)
> {
>    /*      Power Button Setup */
>    outl(0x40020000, PMS_IO_BASE + 0x40);
>
> /* setup WORK_AUX/GPIO24, it is the external signal for 5536
>     * vsb_work_aux controls all voltage rails except Vstandby & Vmem.
>     * We need to enable, OUT_AUX1 and OUTPUT_ENABLE in this order.
>     * If WORK_AUX/GPIO24 is not enabled then soft-off will not work.
>     */
>    outl(GPIOH_24_SET, GPIO_IO_BASE + GPIOH_OUT_AUX1_SELECT);
>    outl(GPIOH_24_SET, GPIO_IO_BASE + GPIOH_OUTPUT_ENABLE);
>
> }
>
> ron

Does this platform have a power button? You could be getting the four 
second ACPI soft-off if the power button signal was tied to on.

Marc

[ron minnich]

On 10/15/07, Marc Jones  wrote:
> Does this platform have a power button? You could be getting the four
> second ACPI soft-off if the power button signal was tied to on.

it has an unconnected power  button.

Call it the Case of the Phantom Power Button.

Time to check once more for a jumper, I suppose.

ron

[Tom Sylla]

ron minnich wrote:
> On 10/15/07, Jordan Crouse  wrote:
> > What are you trying to do?
>
> I load linuxbios, and alix1c runs for 3-4 seconds and turns off. I am
> convinced it is something in PM hardware, just don't know what.
>
> Here are the settings we make
> static void cs5536_setup_power_button(void)
> {
>    /*      Power Button Setup */
>    outl(0x40020000, PMS_IO_BASE + 0x40);

I don't understand. You *are* doing that write to FSD? That is causing 
your problem. The 0x40020000 enables the power button fail-safe for 
exactly 4 seconds. I would think it should be 0x80000000 for a system 
with no power button. (lock the power button disabled)

In another mail, you mentioned the the power button is unconnected. That 
isn't valid, the input needs a button and pullup, or it needs to be 
grounded (see the "skip" feature in the datasheet)

[ron minnich]

Thanks tom, you are right re power button. Don't know how I missed
this. BUT, this setup is done in the early startup code for all 5536
parts. I need to add a Config.lb entry into ALL mainboards that use
the 5536 to allow people to set this value. I don't know why we
decided to set the value so early, it can wait and be done in
gcc-compiled code.

more later.

ron