RE: [PATCH 2/2] ARM: at91/pm: add ultra Low-power mode 1(ULP1) support

From: Yang, Wenyou
Date: Thu Oct 22 2015 - 21:54:13 EST


Hi Alexandre,

Sorry for late answer.

> -----Original Message-----
> From: Alexandre Belloni [mailto:alexandre.belloni@xxxxxxxxxxxxxxxxxx]
> Sent: 2015年10月15日 21:22
> To: Yang, Wenyou
> Cc: Ferre, Nicolas; Jean-Christophe Plagniol-Villard; Russell King; Stephen Boyd;
> Michael Turquette; linux-arm-kernel@xxxxxxxxxxxxxxxxxxx; linux-
> kernel@xxxxxxxxxxxxxxx; linux-clk@xxxxxxxxxxxxxxx
> Subject: Re: [PATCH 2/2] ARM: at91/pm: add ultra Low-power mode 1(ULP1)
> support
>
> Hello Wenyou,
>
> On 15/10/2015 at 11:41:07 +0800, Wenyou Yang wrote :
> > +#define ULP0_MODE 0x00
> > +#define ULP1_MODE 0x11
>
> Those are not used.
>
> > +static void at91_config_ulp1_wkup_source(void)
> > +{
> > + if (at91_pmc_read(AT91_PMC_VERSION) >= SAMA5D2_PMC_VERSION)
> {
> > + at91_pmc_write(AT91_PMC_FSMR, AT91_PMC_RTCAL |
> > + AT91_PMC_FSTT9 |
> > + AT91_PMC_FSTT8 |
> > + AT91_PMC_FSTT7 |
> > + AT91_PMC_FSTT6 |
> > + AT91_PMC_FSTT5 |
> > + AT91_PMC_FSTT4 |
> > + AT91_PMC_FSTT3 |
> > + AT91_PMC_FSTT2 |
> > + AT91_PMC_FSTT0);
> > +
> > + at91_pmc_write(AT91_PMC_FSPR, 0);
>
> Shouldn't those be configured from irq_set_wake() in the aic driver instead of
> activating all the wakeup sources?
> I think you need to get the PMC syscon from the aic5 driver and implement a new
> irq_set_wake function when you can find the sam5d2 pmc.
I understand these code should not be here, but I don't think those should be configured from irq_set_wake().
As said in datasheet, those configuration is to trigger a fast restart signal to the PMC. I think they should not be related to irq.

>
> The PMC syscon is introduced with that series:
> http://lists.infradead.org/pipermail/linux-arm-kernel/2015-October/376493.html
>
> > @@ -140,6 +165,10 @@ static void at91_pm_suspend(suspend_state_t state)
> > pm_data |= (state == PM_SUSPEND_MEM) ?
> > AT91_PM_MODE(AT91_PM_SLOW_CLOCK) : 0;
> >
> > + pm_data |= ((state == PM_SUSPEND_MEM) &&
> > + (at91_pmc_read(AT91_PMC_VERSION) >=
> SAMA5D2_PMC_VERSION)) ?
> > + AT91_PM_ULP(AT91_PM_ULP1_MODE) : 0;
> > +
>
> I would prefer not relying on the AT91_PMC_VERSION. Also, you probably
> shouldn't read the register each time you go to suspend.
> Finally, this could be better written by testing state only once.
Yes, it is not good.
>
>
> > flush_cache_all();
> > outer_disable();
> >
> > diff --git a/arch/arm/mach-at91/pm.h b/arch/arm/mach-at91/pm.h index
> > 3fcf881..2e76745 100644
> > --- a/arch/arm/mach-at91/pm.h
> > +++ b/arch/arm/mach-at91/pm.h
> > @@ -39,4 +39,11 @@ extern void __iomem *at91_ramc_base[];
> >
> > #define AT91_PM_SLOW_CLOCK 0x01
> >
> > +#define AT91_PM_ULP_OFFSET 5
> > +#define AT91_PM_ULP_MASK 0x03
> > +#define AT91_PM_ULP(x) (((x) & AT91_PM_ULP_MASK) <<
> AT91_PM_ULP_OFFSET)
> > +
> > +#define AT91_PM_ULP0_MODE 0x00
> > +#define AT91_PM_ULP1_MODE 0x01
> > +
> > #endif
> > diff --git a/arch/arm/mach-at91/pm_suspend.S
> > b/arch/arm/mach-at91/pm_suspend.S index 825347b..543c430 100644
> > --- a/arch/arm/mach-at91/pm_suspend.S
> > +++ b/arch/arm/mach-at91/pm_suspend.S
> > @@ -41,6 +41,15 @@ tmp2 .req r5
> > .endm
> >
> > /*
> > + * Wait for main oscillator selection is done */
> > + .macro wait_moscsels
> > +1: ldr tmp1, [pmc, #AT91_PMC_SR]
> > + tst tmp1, #AT91_PMC_MOSCSELS
> > + beq 1b
> > + .endm
> > +
> > +/*
> > * Wait until PLLA has locked.
> > */
> > .macro wait_pllalock
> > @@ -99,6 +108,10 @@ ENTRY(at91_pm_suspend_in_sram)
> > and r0, r0, #AT91_PM_MODE_MASK
> > str r0, .pm_mode
> >
> > + lsr r0, r3, #AT91_PM_ULP_OFFSET
> > + and r0, r0, #AT91_PM_ULP_MASK
> > + str r0, .ulp_mode
> > +
> > /* Active the self-refresh mode */
> > mov r0, #SRAMC_SELF_FRESH_ACTIVE
> > bl at91_sramc_self_refresh
> > @@ -107,6 +120,14 @@ ENTRY(at91_pm_suspend_in_sram)
> > tst r0, #AT91_PM_SLOW_CLOCK
> > beq standby_mode
> >
> > + ldr r0, .ulp_mode
> > + tst r0, #AT91_PM_ULP1_MODE
> > + beq ulp0_mode
> > +
> > +ulp1_mode:
> > + bl at91_pm_ulp1_mode
> > + b pm_exit
> > +
> > ulp0_mode:
> > bl at91_pm_ulp0_mode
> > b pm_exit
> > @@ -313,6 +334,94 @@ ENTRY(at91_pm_ulp0_mode)
> > mov pc, lr
> > ENDPROC(at91_pm_ulp0_mode)
> >
> > +/*
> > + * void at91_pm_ulp1_mode(void)
> > + *
> > + */
> > +ENTRY(at91_pm_ulp1_mode)
> > + ldr pmc, .pmc_base
> > +
> > + /* Save PMC_MCKR config */
> > + ldr tmp1, [pmc, #AT91_PMC_MCKR]
> > + str tmp1, .saved_mckr
> > +
> > + /* Switch the master clock source to main clock */
> > + bic tmp1, tmp1, #AT91_PMC_CSS
> > + orr tmp1, tmp1, #AT91_PMC_CSS_MAIN
> > + str tmp1, [pmc, #AT91_PMC_MCKR]
> > +
> > + wait_mckrdy
> > +
> > + /* Save PLLA config, then and disable PLLA */
> > + ldr tmp1, [pmc, #AT91_CKGR_PLLAR]
> > + str tmp1, .saved_pllar
> > +
> > + bic tmp1, tmp1, #AT91_PMC3_MUL
> > + str tmp1, [pmc, #AT91_CKGR_PLLAR]
> > +
> > + /* Switch main clock to 12-MHz RC oscillator */
> > + ldr tmp1, [pmc, #AT91_CKGR_MOR]
> > + bic tmp1, tmp1, #AT91_PMC_MOSCSEL
> > + bic tmp1, tmp1, #AT91_PMC_KEY_MASK
> > + orr tmp1, tmp1, #AT91_PMC_KEY
> > + bic tmp1, tmp1, #(7 << 4)
> > + str tmp1, [pmc, #AT91_CKGR_MOR]
> > +
> > + wait_moscsels
> > +
> > + /* Disable the main oscillator */
> > + ldr tmp1, [pmc, #AT91_CKGR_MOR]
> > + bic tmp1, tmp1, #AT91_PMC_MOSCEN
> > + bic tmp1, tmp1, #AT91_PMC_KEY_MASK
> > + orr tmp1, tmp1, #AT91_PMC_KEY
> > + bic tmp1, tmp1, #(7 << 4)
> > + str tmp1, [pmc, #AT91_CKGR_MOR]
> > +
> > + /* Enter the ULP1 mode by setting WAITMODE bit in CKGR_MOR */
> > + ldr tmp1, [pmc, #AT91_CKGR_MOR]
> > + orr tmp1, tmp1, #AT91_PMC_WAITMODE
> > + bic tmp1, tmp1, #AT91_PMC_KEY_MASK
> > + orr tmp1, tmp1, #AT91_PMC_KEY
> > + bic tmp1, tmp1, #(7 << 4)
> > + str tmp1, [pmc, #AT91_CKGR_MOR]
> > +
> > + wait_mckrdy
> > +
> > + /* Enable the main oscillator */
> > + ldr tmp1, [pmc, #AT91_CKGR_MOR]
> > + orr tmp1, tmp1, #AT91_PMC_MOSCEN
> > + bic tmp1, tmp1, #AT91_PMC_KEY_MASK
> > + orr tmp1, tmp1, #AT91_PMC_KEY
> > + bic tmp1, tmp1, #(7 << 4)
> > + str tmp1, [pmc, #AT91_CKGR_MOR]
> > +
> > + wait_moscrdy
> > +
> > + /* Switch main clock to the main oscillator */
> > + ldr tmp1, [pmc, #AT91_CKGR_MOR]
> > + orr tmp1, tmp1, #AT91_PMC_MOSCSEL
> > + bic tmp1, tmp1, #AT91_PMC_KEY_MASK
> > + orr tmp1, tmp1, #AT91_PMC_KEY
> > + bic tmp1, tmp1, #(7 << 4)
> > + str tmp1, [pmc, #AT91_CKGR_MOR]
> > +
> > + wait_moscsels
> > +
> > + /* Restore PLLA config */
> > + ldr tmp1, .saved_pllar
> > + str tmp1, [pmc, #AT91_CKGR_PLLAR]
> > +
> > + wait_pllalock
> > +
> > + /* Restore PMC_MCKR config */
> > + ldr tmp1, .saved_mckr
> > + str tmp1, [pmc, #AT91_PMC_MCKR]
> > +
> > + wait_mckrdy
> > +
> > + mov pc, lr
> > +ENDPROC(at91_pm_ulp1_mode)
> > +
>
> This makes a lot of duplication anyway, why not having separate code for
> ULP0 and ULP1 that would be copied in SRAM? This would avoid the test
> on .ulp_mode
Thank you for your suggestion. I agree to have separate code for ULP0 and ULP1 code.
But I still think there are a lot a duplication even so.
Anyway, I will prepare a patch as your suggestion.

>
> --
> Alexandre Belloni, Free Electrons
> Embedded Linux, Kernel and Android engineering http://free-electrons.com

Best Regards,
Wenyou Yang
N?叉??y??b??千v??藓{.n???{?赙zXФ?塄}?财??j:+v???赙zZ+€?zf"?????i????ア??璀??撷f?^j谦y??@A?囤?0鹅h??i