diff mbox

[v4,4/6] ARM: rockchip: add basic smp support for rk3288

Message ID 1413231122-1770-5-git-send-email-kever.yang@rock-chips.com (mailing list archive)
State New, archived
Headers show

Commit Message

Kever Yang Oct. 13, 2014, 8:12 p.m. UTC
This patch add basic rk3288 smp support.

Only cortex-A9 need invalid L1, A7/A12/A15/A17 should not invalid L1, since
for A7/A12/A15, the invalidation would be taken as clean and invalidate.

If you use the software manual invalidation instead of hardware invalidation
(assert l1/l2rstdisable during reset) after reset, there is tiny change that
some cachelines would be in dirty and valid state after reset(since the ram
content would be random value after reset), then the unexpected clean might
lead to system crash.

It is a known issue for the A12/A17 MPCore multiprocessor that the active
processors might be stalled when the individual processor is powered down,
we can avoid this prolbem by softreset the processor before power it down.

Signed-off-by: Kever Yang <kever.yang@rock-chips.com>
---

Changes in v4:
- merge patch "fix up rk3288 smp cpu hotplug" into this patch

Changes in v3:
- use one ops and secondary_starup for all rockchip SOCs
- pick back the power domain operation for cpu hotplug

Changes in v2:
- use rk3288_boot_secondary instead ofsmp_boot_secondary
- discards the power domain operation
- handle the per cpu starup when actived by 'sev'

 arch/arm/mach-rockchip/headsmp.S |   5 +-
 arch/arm/mach-rockchip/platsmp.c | 120 +++++++++++++++++++++++++++++++--------
 2 files changed, 100 insertions(+), 25 deletions(-)

Comments

Doug Anderson Oct. 14, 2014, 8:24 p.m. UTC | #1
Kever,

On Mon, Oct 13, 2014 at 1:12 PM, Kever Yang <kever.yang@rock-chips.com> wrote:
> +       /*
> +        * We need to soft reset the cpu when we turn off the cpu power domain,
> +        * or else the active processors might be stalled when the individual
> +        * processor is powered down.
> +        */
> +       if (read_cpuid_part_number() != ARM_CPU_PART_CORTEX_A9) {

I haven't done a full review of this patch, but it seems unlikely that
your uses of read_cpuid_part_number() and read_cpuid_part() in this
patch are correct.  You use both functions and in both cases compare
the results to ARM_CPU_PART_CORTEX_A9.
Heiko Stübner Oct. 14, 2014, 9:23 p.m. UTC | #2
Am Dienstag, 14. Oktober 2014, 13:24:03 schrieb Doug Anderson:
> Kever,
> 
> On Mon, Oct 13, 2014 at 1:12 PM, Kever Yang <kever.yang@rock-chips.com> 
wrote:
> > +       /*
> > +        * We need to soft reset the cpu when we turn off the cpu power
> > domain, +        * or else the active processors might be stalled when
> > the individual +        * processor is powered down.
> > +        */
> > +       if (read_cpuid_part_number() != ARM_CPU_PART_CORTEX_A9) {
> 
> I haven't done a full review of this patch, but it seems unlikely that
> your uses of read_cpuid_part_number() and read_cpuid_part() in this
> patch are correct.  You use both functions and in both cases compare
> the results to ARM_CPU_PART_CORTEX_A9.

I think read_cpuid_part() would be the correct one, as it does

	read_cpuid_id() & ARM_CPU_PART_MASK

which in turn should mask the correct parts of the cpuid to match against 
ARM_CPU_PART_CORTEX_A9 [0].



[0] http://lxr.free-electrons.com/source/arch/arm/include/asm/cputype.h#L77
Kever Yang Oct. 14, 2014, 9:50 p.m. UTC | #3
Heiko,

On 10/14/2014 02:23 PM, Heiko Stübner wrote:
> Am Dienstag, 14. Oktober 2014, 13:24:03 schrieb Doug Anderson:
>> Kever,
>>
>> On Mon, Oct 13, 2014 at 1:12 PM, Kever Yang <kever.yang@rock-chips.com>
> wrote:
>>> +       /*
>>> +        * We need to soft reset the cpu when we turn off the cpu power
>>> domain, +        * or else the active processors might be stalled when
>>> the individual +        * processor is powered down.
>>> +        */
>>> +       if (read_cpuid_part_number() != ARM_CPU_PART_CORTEX_A9) {
>> I haven't done a full review of this patch, but it seems unlikely that
>> your uses of read_cpuid_part_number() and read_cpuid_part() in this
>> patch are correct.  You use both functions and in both cases compare
>> the results to ARM_CPU_PART_CORTEX_A9.
> I think read_cpuid_part() would be the correct one, as it does
You are right, read_cpuid_part() is correct one on kernel next,
I mix up 3.14 kernel and next tree, only read_cpuid_part_number() is 
available
in 3.14 kernel.

I will correct it in my next version, any other changes needed for new 
version?

- Kever.
Russell King - ARM Linux Oct. 14, 2014, 11:36 p.m. UTC | #4
On Tue, Oct 14, 2014 at 11:23:40PM +0200, Heiko Stübner wrote:
> Am Dienstag, 14. Oktober 2014, 13:24:03 schrieb Doug Anderson:
> > Kever,
> > 
> > On Mon, Oct 13, 2014 at 1:12 PM, Kever Yang <kever.yang@rock-chips.com> 
> wrote:
> > > +       /*
> > > +        * We need to soft reset the cpu when we turn off the cpu power
> > > domain, +        * or else the active processors might be stalled when
> > > the individual +        * processor is powered down.
> > > +        */
> > > +       if (read_cpuid_part_number() != ARM_CPU_PART_CORTEX_A9) {
> > 
> > I haven't done a full review of this patch, but it seems unlikely that
> > your uses of read_cpuid_part_number() and read_cpuid_part() in this
> > patch are correct.  You use both functions and in both cases compare
> > the results to ARM_CPU_PART_CORTEX_A9.
> 
> I think read_cpuid_part() would be the correct one, as it does
> 
> 	read_cpuid_id() & ARM_CPU_PART_MASK
> 
> which in turn should mask the correct parts of the cpuid to match against 
> ARM_CPU_PART_CORTEX_A9 [0].

The clue is the deprecated message.  Clearly the original author did not
even build test their code, or if they did, they ignored the warnings.
Either way, that's a fairly worrying issue which raises the question:
how well was this code tested in the first place?

Note that the above quoted if() evaluates to always-false.
Russell King - ARM Linux Oct. 14, 2014, 11:37 p.m. UTC | #5
On Tue, Oct 14, 2014 at 02:50:07PM -0700, Kever Yang wrote:
> Heiko,
> 
> On 10/14/2014 02:23 PM, Heiko Stübner wrote:
> >Am Dienstag, 14. Oktober 2014, 13:24:03 schrieb Doug Anderson:
> >>Kever,
> >>
> >>On Mon, Oct 13, 2014 at 1:12 PM, Kever Yang <kever.yang@rock-chips.com>
> >wrote:
> >>>+       /*
> >>>+        * We need to soft reset the cpu when we turn off the cpu power
> >>>domain, +        * or else the active processors might be stalled when
> >>>the individual +        * processor is powered down.
> >>>+        */
> >>>+       if (read_cpuid_part_number() != ARM_CPU_PART_CORTEX_A9) {
> >>I haven't done a full review of this patch, but it seems unlikely that
> >>your uses of read_cpuid_part_number() and read_cpuid_part() in this
> >>patch are correct.  You use both functions and in both cases compare
> >>the results to ARM_CPU_PART_CORTEX_A9.
> >I think read_cpuid_part() would be the correct one, as it does
> You are right, read_cpuid_part() is correct one on kernel next,
> I mix up 3.14 kernel and next tree, only read_cpuid_part_number() is
> available
> in 3.14 kernel.
> 
> I will correct it in my next version, any other changes needed for new
> version?

You need to at the _very_ _minimum_ build test your code against a
recent kernel, and preferably test it to make sure that it works as
you intend.

Developing on such an old kernel version, and hoping that the code is
going to be correct for later kernels isn't a nice idea.
Kever Yang Oct. 15, 2014, 6:49 a.m. UTC | #6
Russell,

On 10/14/2014 04:37 PM, Russell King - ARM Linux wrote:
> On Tue, Oct 14, 2014 at 02:50:07PM -0700, Kever Yang wrote:
>> Heiko,
>>
>> On 10/14/2014 02:23 PM, Heiko Stübner wrote:
>>> Am Dienstag, 14. Oktober 2014, 13:24:03 schrieb Doug Anderson:
>>>> Kever,
>>>>
>>>> On Mon, Oct 13, 2014 at 1:12 PM, Kever Yang <kever.yang@rock-chips.com>
>>> wrote:
>>>>> +       /*
>>>>> +        * We need to soft reset the cpu when we turn off the cpu power
>>>>> domain, +        * or else the active processors might be stalled when
>>>>> the individual +        * processor is powered down.
>>>>> +        */
>>>>> +       if (read_cpuid_part_number() != ARM_CPU_PART_CORTEX_A9) {
>>>> I haven't done a full review of this patch, but it seems unlikely that
>>>> your uses of read_cpuid_part_number() and read_cpuid_part() in this
>>>> patch are correct.  You use both functions and in both cases compare
>>>> the results to ARM_CPU_PART_CORTEX_A9.
>>> I think read_cpuid_part() would be the correct one, as it does
>> You are right, read_cpuid_part() is correct one on kernel next,
>> I mix up 3.14 kernel and next tree, only read_cpuid_part_number() is
>> available
>> in 3.14 kernel.
>>
>> I will correct it in my next version, any other changes needed for new
>> version?
> You need to at the _very_ _minimum_ build test your code against a
> recent kernel, and preferably test it to make sure that it works as
> you intend.
Thanks for you advice, I'll be more careful next time.

I'm sorry for not test this patch with rk3188(A9 based) board,
I do have test it on top of kernel next with my rk3288 evb(not A9 based),
and I didn't get any warning for read_cpuid_part_number()
during building the image.

- Kever
>
> Developing on such an old kernel version, and hoping that the code is
> going to be correct for later kernels isn't a nice idea.
>
diff mbox

Patch

diff --git a/arch/arm/mach-rockchip/headsmp.S b/arch/arm/mach-rockchip/headsmp.S
index 73206e3..46c22de 100644
--- a/arch/arm/mach-rockchip/headsmp.S
+++ b/arch/arm/mach-rockchip/headsmp.S
@@ -16,7 +16,10 @@ 
 #include <linux/init.h>
 
 ENTRY(rockchip_secondary_startup)
-	bl	v7_invalidate_l1
+	mrc	p15, 0, r0, c0, c0, 0	@ read main ID register
+	ldr	r1, =0x00000c09		@ Cortex-A9 primary part number
+	teq	r0, r1
+	beq	v7_invalidate_l1
 	b	secondary_startup
 ENDPROC(rockchip_secondary_startup)
 
diff --git a/arch/arm/mach-rockchip/platsmp.c b/arch/arm/mach-rockchip/platsmp.c
index 57b53b3..866579e 100644
--- a/arch/arm/mach-rockchip/platsmp.c
+++ b/arch/arm/mach-rockchip/platsmp.c
@@ -22,6 +22,8 @@ 
 #include <linux/regmap.h>
 #include <linux/mfd/syscon.h>
 
+#include <linux/reset.h>
+#include <linux/cpu.h>
 #include <asm/cacheflush.h>
 #include <asm/cp15.h>
 #include <asm/smp_scu.h>
@@ -53,11 +55,47 @@  static int pmu_power_domain_is_on(int pd)
 	return !(val & BIT(pd));
 }
 
+struct reset_control *rockchip_get_core_reset(int cpu)
+{
+	struct device *dev = get_cpu_device(cpu);
+	struct device_node *np;
+
+	/* The cpu device is only available after the initial core bringup */
+	if (dev)
+		np = dev->of_node;
+	else
+		np = of_get_cpu_node(cpu, 0);
+
+	return of_reset_control_get(np, NULL);
+}
+
 static int pmu_set_power_domain(int pd, bool on)
 {
 	u32 val = (on) ? 0 : BIT(pd);
 	int ret;
 
+	/*
+	 * We need to soft reset the cpu when we turn off the cpu power domain,
+	 * or else the active processors might be stalled when the individual
+	 * processor is powered down.
+	 */
+	if (read_cpuid_part_number() != ARM_CPU_PART_CORTEX_A9) {
+		struct reset_control *rstc = rockchip_get_core_reset(pd);
+
+		if (IS_ERR(rstc)) {
+			pr_err("%s: could not get reset control for core %d\n",
+			       __func__, pd);
+			return PTR_ERR(rstc);
+		}
+
+		if (on)
+			reset_control_deassert(rstc);
+		else
+			reset_control_assert(rstc);
+
+		reset_control_put(rstc);
+	}
+
 	ret = regmap_update_bits(pmu, PMU_PWRDN_CON, BIT(pd), val);
 	if (ret < 0) {
 		pr_err("%s: could not update power domain\n", __func__);
@@ -84,6 +122,8 @@  static int pmu_set_power_domain(int pd, bool on)
 static int __cpuinit rockchip_boot_secondary(unsigned int cpu,
 					     struct task_struct *idle)
 {
+	int ret;
+
 	if (!sram_base_addr || !pmu) {
 		pr_err("%s: sram or pmu missing for cpu boot\n", __func__);
 		return -ENXIO;
@@ -96,7 +136,26 @@  static int __cpuinit rockchip_boot_secondary(unsigned int cpu,
 	}
 
 	/* start the core */
-	return pmu_set_power_domain(0 + cpu, true);
+	ret = pmu_set_power_domain(0 + cpu, true);
+	if (ret < 0)
+		return ret;
+
+	if (read_cpuid_part() != ARM_CPU_PART_CORTEX_A9) {
+		/* We communicate with the bootrom to active the cpus other
+		 * than cpu0, after a blob of initialize code, they will
+		 * stay at wfe state, once they are actived, they will check
+		 * the mailbox:
+		 * sram_base_addr + 4: 0xdeadbeaf
+		 * sram_base_addr + 8: start address for pc
+		 * */
+		udelay(10);
+		writel(virt_to_phys(rockchip_secondary_startup),
+			sram_base_addr + 8);
+		writel(0xDEADBEAF, sram_base_addr + 4);
+		dsb_sev();
+	}
+
+	return 0;
 }
 
 /**
@@ -129,8 +188,6 @@  static int __init rockchip_smp_prepare_sram(struct device_node *node)
 		return -EINVAL;
 	}
 
-	sram_base_addr = of_iomap(node, 0);
-
 	/* set the boot function for the sram code */
 	rockchip_boot_fn = virt_to_phys(rockchip_secondary_startup);
 
@@ -204,40 +261,55 @@  static void __init rockchip_smp_prepare_cpus(unsigned int max_cpus)
 	struct device_node *node;
 	unsigned int i;
 
-	node = of_find_compatible_node(NULL, NULL, "arm,cortex-a9-scu");
-	if (!node) {
-		pr_err("%s: missing scu\n", __func__);
-		return;
-	}
-
-	scu_base_addr = of_iomap(node, 0);
-	if (!scu_base_addr) {
-		pr_err("%s: could not map scu registers\n", __func__);
-		return;
-	}
-
 	node = of_find_compatible_node(NULL, NULL, "rockchip,rk3066-smp-sram");
 	if (!node) {
 		pr_err("%s: could not find sram dt node\n", __func__);
 		return;
 	}
 
-	if (rockchip_smp_prepare_sram(node))
+	sram_base_addr = of_iomap(node, 0);
+	if (!sram_base_addr) {
+		pr_err("%s: could not map sram registers\n", __func__);
 		return;
+	}
 
 	if (rockchip_smp_prepare_pmu())
 		return;
 
-	/* enable the SCU power domain */
-	pmu_set_power_domain(PMU_PWRDN_SCU, true);
+	if (read_cpuid_part() == ARM_CPU_PART_CORTEX_A9) {
+		if (rockchip_smp_prepare_sram(node))
+			return;
 
-	/*
-	 * While the number of cpus is gathered from dt, also get the number
-	 * of cores from the scu to verify this value when booting the cores.
-	 */
-	ncores = scu_get_core_count(scu_base_addr);
+		/* enable the SCU power domain */
+		pmu_set_power_domain(PMU_PWRDN_SCU, true);
+
+		node = of_find_compatible_node(NULL, NULL, "arm,cortex-a9-scu");
+		if (!node) {
+			pr_err("%s: missing scu\n", __func__);
+			return;
+		}
 
-	scu_enable(scu_base_addr);
+		scu_base_addr = of_iomap(node, 0);
+		if (!scu_base_addr) {
+			pr_err("%s: could not map scu registers\n", __func__);
+			return;
+		}
+
+		/*
+		 * While the number of cpus is gathered from dt, also get the
+		 * number of cores from the scu to verify this value when
+		 * booting the cores.
+		 */
+		ncores = scu_get_core_count(scu_base_addr);
+		pr_err("%s: ncores %d\n", __func__, ncores);
+
+		scu_enable(scu_base_addr);
+	} else {
+		unsigned int l2ctlr;
+
+		asm ("mrc p15, 1, %0, c9, c0, 2\n" : "=r" (l2ctlr));
+		ncores = ((l2ctlr >> 24) & 0x3) + 1;
+	}
 
 	/* Make sure that all cores except the first are really off */
 	for (i = 1; i < ncores; i++)