[PATCH] ARM: omap_hwmod: Add a new state to handle hwmods left enabled at init

[Cousson, Benoit]

Hi Kevin and Rajendra,

On 11/18/2011 7:44 AM, Rajendra Nayak wrote:
> Hi Kevin,
>
> On Friday 18 November 2011 01:05 AM, Kevin Hilman wrote:
>> Rajendra Nayak<rnayak at ti.com> writes:
>>
>>> A hwmod with a 'HWMOD_INIT_NO_IDLE' flag set, is left in
>>> enabled state by the hwmod framework post the initial setup.
>>> Once a real user of the device (a driver) tries to enable it
>>> at a later point, the hmwod framework throws a WARN() about
>>> the device being already in enabled state.
>>>
>>> Fix this by introducing a new state '_HWMOD_STATE_ENABLED_AT_INIT'
>>> to identify such devices/hwmods, so nothing but just a state
>>> change to '_HWMOD_STATE_ENABLED' can be done when the device/hwmod
>>> is requested to be enabled (the first time) by its driver/user.
>>>
>>> A good example of a such a device is an UART used as debug console.
>>> The UART module needs to be kept enabled through the boot, until the
>>> UART driver takes control of it, for debug prints to appear on
>>> the console.
>>
>> Nice. This is indeed much cleaner than what we're doing in the UART
>> code. However...
>>
>>> Signed-off-by: Rajendra Nayak<rnayak at ti.com>
>>> ---
>>> arch/arm/mach-omap2/omap_hwmod.c | 16 ++++++++++++++--
>>> arch/arm/plat-omap/include/plat/omap_hwmod.h | 1 +
>>> 2 files changed, 15 insertions(+), 2 deletions(-)
>>>
>>> diff --git a/arch/arm/mach-omap2/omap_hwmod.c
>>> b/arch/arm/mach-omap2/omap_hwmod.c
>>> index d7f4623..7d94cc3 100644
>>> --- a/arch/arm/mach-omap2/omap_hwmod.c
>>> +++ b/arch/arm/mach-omap2/omap_hwmod.c
>>> @@ -1441,6 +1441,17 @@ static int _enable(struct omap_hwmod *oh)
>>>
>>> pr_debug("omap_hwmod: %s: enabling\n", oh->name);
>>>
>>> + /*
>>> + * hwmods' with HWMOD_INIT_NO_IDLE flag set, are left
>>
>> (no the ' isn't necessary)
>>
>>> + * in enabled state at init.
>>> + * Now that someone is really trying to enable them,
>>> + * just update the state.
>>> + */
>>> + if (oh->_state == _HWMOD_STATE_ENABLED_AT_INIT) {
>>> + oh->_state = _HWMOD_STATE_ENABLED;
>>> + return 0;
>>> + }
>>
>> ...this subtly changes the behavior, at least compared to how the UART
>> code handles this today.
>>
>> One thing that this doesn't do that the current UART code does is ensure
>> that the IP is actually in a state that can properly idle after this is
>> done.
>>
>> For example, if the bootloader is dumb (most are) and has configured the
>> UARTs in mode that prevents idle (e.g. no-idle mode), then these UARTs
>> will never allow the SoC to hit low power states.

That will not happen because we already did the complete enable sequence 
during setup phase that will set the sysconfig properly.
So the device is properly initialized, we just skipped the idle phase 
because we wanted to keep the device operational for the console.

>> So, what's really needed is not just a return here, but an _idle() and
>> then continue so we know that the HW is in a state that we know can
>> idle from here on out.

We are in the enable call, so the driver / omap_device does expect the 
IP to be operational at the moment, we cannot put it into idle.

> I don't know if I completely understand what you are saying.
> The above change is in the _enable() call which would be triggered
> as part of runtime call from the serial driver to enable the module.
> (Since the module is already enabled, you just change state to suggest
> its now enabled by its user, rather than 'left enabled' at init)
> There would be another call from the serial driver to idle it when
> a corresponding hwmod _idle() call would idle the device.
> I don't seem to understand why would a call to _idle() be needed
> in an _enable() call.

Yep, I agree. I think Kevin's point is due to the previous sequence used 
by the uart:

+	omap_hwmod_idle(od->hwmods[0]);
+	omap_hwmod_enable(od->hwmods[0]);

But that sequence is not needed for my point of view. The device is 
already properly enabled. I don't think it was even needed in the 
previous code. The hwmod was already enabled properly, only the 
omap_device was disabled. So the original omap_console_hwmod_enable 
should just do nothing.

Regards,
Benoit