>-----Original Message----- >From: ext Vishwanath Sripathy [mailto:vishwanath.bs@xxxxxx] >Sent: 19 January, 2011 08:06 >To: Kristo Tero (Nokia-MS/Tampere); linux-omap@xxxxxxxxxxxxxxx >Cc: Paul Walmsley; Kevin Hilman >Subject: RE: [PATCH] OMAP3: CPUIdle: prevent CORE from going off if >doing so would reset an active clockdomain > >Tero, > >> -----Original Message----- >> From: Vishwanath Sripathy [mailto:vishwanath.bs@xxxxxx] >> Sent: Wednesday, January 19, 2011 10:09 AM >> To: Tero Kristo; linux-omap@xxxxxxxxxxxxxxx >> Cc: Paul Walmsley; Kevin Hilman >> Subject: RE: [PATCH] OMAP3: CPUIdle: prevent CORE from going off if >> doing so would reset an active clockdomain >> >> Tero, >> >> > -----Original Message----- >> > From: linux-omap-owner@xxxxxxxxxxxxxxx [mailto:linux-omap- >> > owner@xxxxxxxxxxxxxxx] On Behalf Of Tero Kristo >> > Sent: Tuesday, January 18, 2011 3:18 PM >> > To: linux-omap@xxxxxxxxxxxxxxx >> > Cc: Paul Walmsley; Kevin Hilman >> > Subject: [PATCH] OMAP3: CPUIdle: prevent CORE from going off if >> doing >> > so would reset an active clockdomain >> > >> > On OMAP3 SoCs, if the CORE powerdomain enters off-mode, many >> other >> > parts of the chip will be reset. If those parts of the chip are >busy, >> > the reset will disrupt them, causing unpredictable and generally >> > undesirable results. >> If some parts of the chip are busy, then how can Core domain enter off >> state? The necessary condition for Core to enter low power state is >that >> all the clock domains (including DSS, CAM, IVA, USB, PER etc) should >> have >> idled. Doesn't it mean that all the modules have idled and asserted >> idleack when Core is entering off state? >Besides these, Core off should reset the modules which are only in Core >domain. It should not impact other power domains. Also Core domain >modules >which are reset will restore their context when Core comes out of off >mode. So why are you saying that "If those parts of the chip are busy, >the reset will disrupt them, causing unpredictable and generally >undesirable results."? Core off issues reset to peripheral domains when it wakes up, this is somehow (badly) visible in TRM (look for COREDOMAINWKUP_RST.) When this reset happens, the peripheral domain shows its reset status as being high, but the powerdomain itself has not entered off (previous state can be e.g. RET), thus its context will not be restored. Also, another justification for this patch is to prevent unnecessary core-off when the device can't really idle at all => prevents unnecessary overhead inside omap_sram_idle path. > >Vishwa >> >> Vishwa >> > >> > Signed-off-by: Tero Kristo <tero.kristo@xxxxxxxxx> >> > Cc: Paul Walmsley <paul@xxxxxxxxx> >> > Cc: Kevin Hilman <khilman@xxxxxxxxxxxxxxxxxxx> >> > --- >> > arch/arm/mach-omap2/cpuidle34xx.c | 40 >> > +++++++++++++++++++++++++++++++++++- >> > 1 files changed, 38 insertions(+), 2 deletions(-) >> > >> > diff --git a/arch/arm/mach-omap2/cpuidle34xx.c b/arch/arm/mach- >> > omap2/cpuidle34xx.c >> > index f3e043f..d31b858 100644 >> > --- a/arch/arm/mach-omap2/cpuidle34xx.c >> > +++ b/arch/arm/mach-omap2/cpuidle34xx.c >> > @@ -61,7 +61,7 @@ struct omap3_processor_cx { >> > struct omap3_processor_cx >> > omap3_power_states[OMAP3_MAX_STATES]; >> > struct omap3_processor_cx current_cx_state; >> > struct powerdomain *mpu_pd, *core_pd, *per_pd; >> > -struct powerdomain *cam_pd; >> > +struct powerdomain *cam_pd, *dss_pd, *iva2_pd, *sgx_pd, >> *usb_pd; >> > >> > /* >> > * The latencies/thresholds for various C states have >> > @@ -235,7 +235,7 @@ static int omap3_enter_idle_bm(struct >> > cpuidle_device *dev, >> > { >> > struct cpuidle_state *new_state = next_valid_state(dev, state); >> > u32 core_next_state, per_next_state = 0, per_saved_state = 0; >> > - u32 cam_state; >> > + u32 cam_state, dss_state, iva2_state, sgx_state, usb_state; >> > struct omap3_processor_cx *cx; >> > int ret; >> > >> > @@ -256,6 +256,8 @@ static int omap3_enter_idle_bm(struct >> > cpuidle_device *dev, >> > * its own code. >> > */ >> > >> > + /* XXX Add CORE-active check here */ >> > + >> > /* >> > * Prevent idle completely if CAM is active. >> > * CAM does not have wakeup capability in OMAP3. >> > @@ -275,6 +277,36 @@ static int omap3_enter_idle_bm(struct >> > cpuidle_device *dev, >> > (core_next_state > PWRDM_POWER_RET)) >> > per_next_state = PWRDM_POWER_RET; >> > >> > + /* XXX Add prevent-PER-off check here */ >> > + >> > + /* >> > + * If we are attempting CORE off, check if any other >> > powerdomains >> > + * are at retention or higher. CORE off causes chipwide reset >> > which >> > + * would reset these domains also. >> > + */ >> > + if (core_next_state == PWRDM_POWER_OFF) { >> > + iva2_state = pwrdm_read_pwrst(iva2_pd); >> > + sgx_state = pwrdm_read_pwrst(sgx_pd); >> > + usb_state = pwrdm_read_pwrst(usb_pd); >> > + dss_state = pwrdm_read_pwrst(dss_pd); >> > + >> > + if (cam_state > PWRDM_POWER_OFF || >> > + dss_state > PWRDM_POWER_OFF || >> > + iva2_state > PWRDM_POWER_OFF || >> > + per_next_state > PWRDM_POWER_OFF || >> > + sgx_state > PWRDM_POWER_OFF || >> > + usb_state > PWRDM_POWER_OFF) >> > + core_next_state = PWRDM_POWER_RET; >> > + } >> > + >> > + /* Fallback to new target core/mpu state */ >> > + while (cx->core_state < core_next_state) { >> > + state--; >> > + cx = cpuidle_get_statedata(state); >> > + } >> > + >> > + new_state = state; >> > + >> > /* Are we changing PER target state? */ >> > if (per_next_state != per_saved_state) >> > pwrdm_set_next_pwrst(per_pd, per_next_state); >> > @@ -489,6 +521,10 @@ int __init omap3_idle_init(void) >> > core_pd = pwrdm_lookup("core_pwrdm"); >> > per_pd = pwrdm_lookup("per_pwrdm"); >> > cam_pd = pwrdm_lookup("cam_pwrdm"); >> > + dss_pd = pwrdm_lookup("dss_pwrdm"); >> > + iva2_pd = pwrdm_lookup("iva2_pwrdm"); >> > + sgx_pd = pwrdm_lookup("sgx_pwrdm"); >> > + usb_pd = pwrdm_lookup("usbhost_pwrdm"); >> > >> > omap_init_power_states(); >> > cpuidle_register_driver(&omap3_idle_driver); >> > -- >> > 1.7.1 >> > >> > -- >> > To unsubscribe from this list: send the line "unsubscribe linux- >omap" >in >> > the body of a message to majordomo@xxxxxxxxxxxxxxx >> > More majordomo info at http://vger.kernel.org/majordomo-info.html -- To unsubscribe from this list: send the line "unsubscribe linux-omap" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
