[PATCH v3] ARM: rockchip: smp: enable secondary rk3066a/rk3188 CPUs

[PATCH v3] ARM: rockchip: smp: enable secondary rk3066a/rk3188 CPUs

[Johan Jonker via B4 Relay]

From: Johan Jonker <jbx6244@gmail.com>

Early Rockchip SoCs like rk3066a and rk3188 had boot loaders
that remapped the boot ROM and the internal memory. Currently
U-boot makes use of a common board file without remapping
resulting in the secondary CPUs power up hanging in a wait loop.

Make the various boot loader methods work by:

* add code that disables remapping on both rk3066 and rk3188
* change to use the same boot ROM path for all socs in rockchip_boot_secondary()
* drop the rockchip_smp_prepare_sram() part

Signed-off-by: Johan Jonker <jbx6244@gmail.com>
---

Changed V3:
remove include
use regmap_write
---
 arch/arm/mach-rockchip/Makefile   |   2 +-
 arch/arm/mach-rockchip/core.h     |  10 ----
 arch/arm/mach-rockchip/headsmp.S  |  16 ------
 arch/arm/mach-rockchip/platsmp.c  | 105 +++++++++++++++-----------------------
 arch/arm/mach-rockchip/rockchip.c |   1 -
 5 files changed, 42 insertions(+), 92 deletions(-)

diff --git a/arch/arm/mach-rockchip/Makefile b/arch/arm/mach-rockchip/Makefile
index 62596d5deb88..d05123a7b711 100644
--- a/arch/arm/mach-rockchip/Makefile
+++ b/arch/arm/mach-rockchip/Makefile
@@ -3,4 +3,4 @@ CFLAGS_platsmp.o := -march=armv7-a
 
 obj-$(CONFIG_ARCH_ROCKCHIP) += rockchip.o
 obj-$(CONFIG_PM_SLEEP) += pm.o sleep.o
-obj-$(CONFIG_SMP) += headsmp.o platsmp.o
+obj-$(CONFIG_SMP) += platsmp.o
diff --git a/arch/arm/mach-rockchip/core.h b/arch/arm/mach-rockchip/core.h
deleted file mode 100644
index fabcaa66607c..000000000000
--- a/arch/arm/mach-rockchip/core.h
+++ /dev/null
@@ -1,10 +0,0 @@
-/* SPDX-License-Identifier: GPL-2.0-or-later */
-/*
- * Copyright (c) 2013 MundoReader S.L.
- * Author: Heiko Stuebner <heiko@sntech.de>
- */
-
-extern char rockchip_secondary_trampoline;
-extern char rockchip_secondary_trampoline_end;
-
-extern unsigned long rockchip_boot_fn;
diff --git a/arch/arm/mach-rockchip/headsmp.S b/arch/arm/mach-rockchip/headsmp.S
deleted file mode 100644
index 37a7ea524a16..000000000000
--- a/arch/arm/mach-rockchip/headsmp.S
+++ /dev/null
@@ -1,16 +0,0 @@
-/* SPDX-License-Identifier: GPL-2.0-or-later */
-/*
- * Copyright (c) 2013 MundoReader S.L.
- * Author: Heiko Stuebner <heiko@sntech.de>
- */
-#include <linux/linkage.h>
-#include <linux/init.h>
-
-ENTRY(rockchip_secondary_trampoline)
-	ldr	pc, 1f
-ENDPROC(rockchip_secondary_trampoline)
-	.globl	rockchip_boot_fn
-rockchip_boot_fn:
-1:	.space	4
-
-ENTRY(rockchip_secondary_trampoline_end)
diff --git a/arch/arm/mach-rockchip/platsmp.c b/arch/arm/mach-rockchip/platsmp.c
index f659d894bfae..5f3ca7c0cfe4 100644
--- a/arch/arm/mach-rockchip/platsmp.c
+++ b/arch/arm/mach-rockchip/platsmp.c
@@ -21,17 +21,21 @@
 #include <asm/smp_plat.h>
 #include <asm/mach/map.h>
 
-#include "core.h"
-
 static void __iomem *scu_base_addr;
 static void __iomem *sram_base_addr;
 static int ncores;
 
+#define RK3066_GRF_SOC_CON0	0x0150
+#define RK3188_GRF_SOC_CON0	0x00a0
+
+#define NOC_REMAP		BIT(12)
+
 #define PMU_PWRDN_CON		0x08
 #define PMU_PWRDN_ST		0x0c
 
 #define PMU_PWRDN_SCU		4
 
+static struct regmap *grf;
 static struct regmap *pmu;
 static int has_pmu = true;
 static struct reset_control *cpu_rstc[4];
@@ -133,66 +137,21 @@ static int rockchip_boot_secondary(unsigned int cpu, struct task_struct *idle)
 	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 activated, they will check
-		 * the mailbox:
-		 * sram_base_addr + 4: 0xdeadbeaf
-		 * sram_base_addr + 8: start address for pc
-		 * The cpu0 need to wait the other cpus other than cpu0 entering
-		 * the wfe state.The wait time is affected by many aspects.
-		 * (e.g: cpu frequency, bootrom frequency, sram frequency, ...)
-		 */
-		mdelay(1); /* ensure the cpus other than cpu0 to startup */
-
-		writel(__pa_symbol(secondary_startup), sram_base_addr + 8);
-		writel(0xDEADBEAF, sram_base_addr + 4);
-		dsb_sev();
-	}
-
-	return 0;
-}
-
-/**
- * rockchip_smp_prepare_sram - populate necessary sram block
- * Starting cores execute the code residing at the start of the on-chip sram
- * after power-on. Therefore make sure, this sram region is reserved and
- * big enough. After this check, copy the trampoline code that directs the
- * core to the real startup code in ram into the sram-region.
- * @node: mmio-sram device node
- */
-static int __init rockchip_smp_prepare_sram(struct device_node *node)
-{
-	unsigned int trampoline_sz = &rockchip_secondary_trampoline_end -
-					    &rockchip_secondary_trampoline;
-	struct resource res;
-	unsigned int rsize;
-	int ret;
-
-	ret = of_address_to_resource(node, 0, &res);
-	if (ret < 0) {
-		pr_err("%s: could not get address for node %pOF\n",
-		       __func__, node);
-		return ret;
-	}
-
-	rsize = resource_size(&res);
-	if (rsize < trampoline_sz) {
-		pr_err("%s: reserved block with size 0x%x is too small for trampoline size 0x%x\n",
-		       __func__, rsize, trampoline_sz);
-		return -EINVAL;
-	}
-
-	/* set the boot function for the sram code */
-	rockchip_boot_fn = __pa_symbol(secondary_startup);
-
-	/* copy the trampoline to sram, that runs during startup of the core */
-	memcpy_toio(sram_base_addr, &rockchip_secondary_trampoline, trampoline_sz);
-	flush_cache_all();
-	outer_clean_range(0, trampoline_sz);
-
+	/*
+	* 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 activated, they will check
+	* the mailbox:
+	* sram_base_addr + 4: 0xdeadbeaf
+	* sram_base_addr + 8: start address for pc
+	* The cpu0 need to wait the other cpus other than cpu0 entering
+	* the wfe state.The wait time is affected by many aspects.
+	* (e.g: cpu frequency, bootrom frequency, sram frequency, ...)
+	*/
+	mdelay(1); /* ensure the cpus other than cpu0 to startup */
+
+	writel(__pa_symbol(secondary_startup), sram_base_addr + 8);
+	writel(0xDEADBEAF, sram_base_addr + 4);
 	dsb_sev();
 
 	return 0;
@@ -320,9 +279,27 @@ static void __init rockchip_smp_prepare_cpus(unsigned int max_cpus)
 	for (i = 1; i < ncores; i++)
 		pmu_set_power_domain(0 + i, false);
 
+	/* Make sure that the remap function is disabled */
 	if (read_cpuid_part() == ARM_CPU_PART_CORTEX_A9) {
-		if (rockchip_smp_prepare_sram(node)) {
-			of_node_put(node);
+		unsigned int reg;
+		int ret;
+
+		if (of_machine_is_compatible("rockchip,rk3066a")) {
+			reg = RK3066_GRF_SOC_CON0;
+			grf = syscon_regmap_lookup_by_compatible("rockchip,rk3066-grf");
+		} else {
+			reg = RK3188_GRF_SOC_CON0;
+			grf = syscon_regmap_lookup_by_compatible("rockchip,rk3188-grf");
+		}
+
+		if (IS_ERR(grf)) {
+			pr_err("%s: could not find grf dt node\n", __func__);
+			return;
+		}
+
+		ret = regmap_write(grf, reg, NOC_REMAP << 16);
+		if (ret < 0) {
+			pr_err("%s: could not update remap function\n", __func__);
 			return;
 		}
 	}
diff --git a/arch/arm/mach-rockchip/rockchip.c b/arch/arm/mach-rockchip/rockchip.c
index 82102fbabf92..c2b919ac99bc 100644
--- a/arch/arm/mach-rockchip/rockchip.c
+++ b/arch/arm/mach-rockchip/rockchip.c
@@ -14,7 +14,6 @@
 #include <linux/clocksource.h>
 #include <asm/mach/arch.h>
 #include <asm/mach/map.h>
-#include "core.h"
 #include "pm.h"
 
 #define RK3288_TIMER6_7_PHYS 0xff810000

---
base-commit: 9d80aa4617b32f5054c5aa471d06b66704854935
change-id: 20261017-cpu-c8f61c6466f9

Best regards,
-- 
Johan Jonker <jbx6244@gmail.com>

Re: [PATCH v3] ARM: rockchip: smp: enable secondary rk3066a/rk3188 CPUs

[boogie]

Hello Johan

Thanks for the clean patch!

Tested on my Lifetab md98486 tablet, everything seems to work fine with 
rockchip TPL (30_LPDDR2_300MHz_DDR3_300MHz_20130517.bin) + UBOOT SPL 
(almost latest 2026.xx) booted from nand idb.

[    0.169171] smp: Bringing up secondary CPUs ...
[    0.176761] CPU1: thread -1, cpu 1, socket 0, mpidr 80000001
[    0.177211] CPU1: Spectre v2: using BPIALL workaround
[    0.189310] smp: Brought up 1 node, 2 CPUs
[    0.193902] SMP: Total of 2 processors activated (96.00 BogoMIPS).
[    0.200728] CPU: All CPU(s) started in SVC mode.


[alarm@alarm ~]$ cat /proc/cpuinfo
processor       : 0
model name      : ARMv7 Processor rev 0 (v7l)
BogoMIPS        : 101.76
Features        : half thumb fastmult vfp edsp thumbee neon vfpv3 tls 
vfpd32
CPU implementer : 0x41
CPU architecture: 7
CPU variant     : 0x3
CPU part        : 0xc09
CPU revision    : 0

processor       : 1
model name      : ARMv7 Processor rev 0 (v7l)
BogoMIPS        : 101.76
Features        : half thumb fastmult vfp edsp thumbee neon vfpv3 tls 
vfpd32
CPU implementer : 0x41
CPU architecture: 7
CPU variant     : 0x3
CPU part        : 0xc09
CPU revision    : 0

Hardware        : Rockchip (Device Tree)
Revision        : 0000
Serial          : 0000000000000000
[alarm@alarm ~]$ uname -a
Linux alarm 7.1.0-ARCH+ #22 SMP PREEMPT Sat Oct 17 22:11:35 CEST 2026 
armv7l GNU/Linux


Tested-by: Hüseyin BIYIK <boogiepop@gmx.com>

On 10/17/26 15:27, Johan Jonker via B4 Relay wrote:
> From: Johan Jonker <jbx6244@gmail.com>
> 
> Early Rockchip SoCs like rk3066a and rk3188 had boot loaders
> that remapped the boot ROM and the internal memory. Currently
> U-boot makes use of a common board file without remapping
> resulting in the secondary CPUs power up hanging in a wait loop.
> 
> Make the various boot loader methods work by:
> 
> * add code that disables remapping on both rk3066 and rk3188
> * change to use the same boot ROM path for all socs in rockchip_boot_secondary()
> * drop the rockchip_smp_prepare_sram() part
> 
> Signed-off-by: Johan Jonker <jbx6244@gmail.com>
> ---
> 
> Changed V3:
> remove include
> use regmap_write
> ---
>   arch/arm/mach-rockchip/Makefile   |   2 +-
>   arch/arm/mach-rockchip/core.h     |  10 ----
>   arch/arm/mach-rockchip/headsmp.S  |  16 ------
>   arch/arm/mach-rockchip/platsmp.c  | 105 +++++++++++++++-----------------------
>   arch/arm/mach-rockchip/rockchip.c |   1 -
>   5 files changed, 42 insertions(+), 92 deletions(-)
> 
> diff --git a/arch/arm/mach-rockchip/Makefile b/arch/arm/mach-rockchip/Makefile
> index 62596d5deb88..d05123a7b711 100644
> --- a/arch/arm/mach-rockchip/Makefile
> +++ b/arch/arm/mach-rockchip/Makefile
> @@ -3,4 +3,4 @@ CFLAGS_platsmp.o := -march=armv7-a
>   
>   obj-$(CONFIG_ARCH_ROCKCHIP) += rockchip.o
>   obj-$(CONFIG_PM_SLEEP) += pm.o sleep.o
> -obj-$(CONFIG_SMP) += headsmp.o platsmp.o
> +obj-$(CONFIG_SMP) += platsmp.o
> diff --git a/arch/arm/mach-rockchip/core.h b/arch/arm/mach-rockchip/core.h
> deleted file mode 100644
> index fabcaa66607c..000000000000
> --- a/arch/arm/mach-rockchip/core.h
> +++ /dev/null
> @@ -1,10 +0,0 @@
> -/* SPDX-License-Identifier: GPL-2.0-or-later */
> -/*
> - * Copyright (c) 2013 MundoReader S.L.
> - * Author: Heiko Stuebner <heiko@sntech.de>
> - */
> -
> -extern char rockchip_secondary_trampoline;
> -extern char rockchip_secondary_trampoline_end;
> -
> -extern unsigned long rockchip_boot_fn;
> diff --git a/arch/arm/mach-rockchip/headsmp.S b/arch/arm/mach-rockchip/headsmp.S
> deleted file mode 100644
> index 37a7ea524a16..000000000000
> --- a/arch/arm/mach-rockchip/headsmp.S
> +++ /dev/null
> @@ -1,16 +0,0 @@
> -/* SPDX-License-Identifier: GPL-2.0-or-later */
> -/*
> - * Copyright (c) 2013 MundoReader S.L.
> - * Author: Heiko Stuebner <heiko@sntech.de>
> - */
> -#include <linux/linkage.h>
> -#include <linux/init.h>
> -
> -ENTRY(rockchip_secondary_trampoline)
> -	ldr	pc, 1f
> -ENDPROC(rockchip_secondary_trampoline)
> -	.globl	rockchip_boot_fn
> -rockchip_boot_fn:
> -1:	.space	4
> -
> -ENTRY(rockchip_secondary_trampoline_end)
> diff --git a/arch/arm/mach-rockchip/platsmp.c b/arch/arm/mach-rockchip/platsmp.c
> index f659d894bfae..5f3ca7c0cfe4 100644
> --- a/arch/arm/mach-rockchip/platsmp.c
> +++ b/arch/arm/mach-rockchip/platsmp.c
> @@ -21,17 +21,21 @@
>   #include <asm/smp_plat.h>
>   #include <asm/mach/map.h>
>   
> -#include "core.h"
> -
>   static void __iomem *scu_base_addr;
>   static void __iomem *sram_base_addr;
>   static int ncores;
>   
> +#define RK3066_GRF_SOC_CON0	0x0150
> +#define RK3188_GRF_SOC_CON0	0x00a0
> +
> +#define NOC_REMAP		BIT(12)
> +
>   #define PMU_PWRDN_CON		0x08
>   #define PMU_PWRDN_ST		0x0c
>   
>   #define PMU_PWRDN_SCU		4
>   
> +static struct regmap *grf;
>   static struct regmap *pmu;
>   static int has_pmu = true;
>   static struct reset_control *cpu_rstc[4];
> @@ -133,66 +137,21 @@ static int rockchip_boot_secondary(unsigned int cpu, struct task_struct *idle)
>   	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 activated, they will check
> -		 * the mailbox:
> -		 * sram_base_addr + 4: 0xdeadbeaf
> -		 * sram_base_addr + 8: start address for pc
> -		 * The cpu0 need to wait the other cpus other than cpu0 entering
> -		 * the wfe state.The wait time is affected by many aspects.
> -		 * (e.g: cpu frequency, bootrom frequency, sram frequency, ...)
> -		 */
> -		mdelay(1); /* ensure the cpus other than cpu0 to startup */
> -
> -		writel(__pa_symbol(secondary_startup), sram_base_addr + 8);
> -		writel(0xDEADBEAF, sram_base_addr + 4);
> -		dsb_sev();
> -	}
> -
> -	return 0;
> -}
> -
> -/**
> - * rockchip_smp_prepare_sram - populate necessary sram block
> - * Starting cores execute the code residing at the start of the on-chip sram
> - * after power-on. Therefore make sure, this sram region is reserved and
> - * big enough. After this check, copy the trampoline code that directs the
> - * core to the real startup code in ram into the sram-region.
> - * @node: mmio-sram device node
> - */
> -static int __init rockchip_smp_prepare_sram(struct device_node *node)
> -{
> -	unsigned int trampoline_sz = &rockchip_secondary_trampoline_end -
> -					    &rockchip_secondary_trampoline;
> -	struct resource res;
> -	unsigned int rsize;
> -	int ret;
> -
> -	ret = of_address_to_resource(node, 0, &res);
> -	if (ret < 0) {
> -		pr_err("%s: could not get address for node %pOF\n",
> -		       __func__, node);
> -		return ret;
> -	}
> -
> -	rsize = resource_size(&res);
> -	if (rsize < trampoline_sz) {
> -		pr_err("%s: reserved block with size 0x%x is too small for trampoline size 0x%x\n",
> -		       __func__, rsize, trampoline_sz);
> -		return -EINVAL;
> -	}
> -
> -	/* set the boot function for the sram code */
> -	rockchip_boot_fn = __pa_symbol(secondary_startup);
> -
> -	/* copy the trampoline to sram, that runs during startup of the core */
> -	memcpy_toio(sram_base_addr, &rockchip_secondary_trampoline, trampoline_sz);
> -	flush_cache_all();
> -	outer_clean_range(0, trampoline_sz);
> -
> +	/*
> +	* 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 activated, they will check
> +	* the mailbox:
> +	* sram_base_addr + 4: 0xdeadbeaf
> +	* sram_base_addr + 8: start address for pc
> +	* The cpu0 need to wait the other cpus other than cpu0 entering
> +	* the wfe state.The wait time is affected by many aspects.
> +	* (e.g: cpu frequency, bootrom frequency, sram frequency, ...)
> +	*/
> +	mdelay(1); /* ensure the cpus other than cpu0 to startup */
> +
> +	writel(__pa_symbol(secondary_startup), sram_base_addr + 8);
> +	writel(0xDEADBEAF, sram_base_addr + 4);
>   	dsb_sev();
>   
>   	return 0;
> @@ -320,9 +279,27 @@ static void __init rockchip_smp_prepare_cpus(unsigned int max_cpus)
>   	for (i = 1; i < ncores; i++)
>   		pmu_set_power_domain(0 + i, false);
>   
> +	/* Make sure that the remap function is disabled */
>   	if (read_cpuid_part() == ARM_CPU_PART_CORTEX_A9) {
> -		if (rockchip_smp_prepare_sram(node)) {
> -			of_node_put(node);
> +		unsigned int reg;
> +		int ret;
> +
> +		if (of_machine_is_compatible("rockchip,rk3066a")) {
> +			reg = RK3066_GRF_SOC_CON0;
> +			grf = syscon_regmap_lookup_by_compatible("rockchip,rk3066-grf");
> +		} else {
> +			reg = RK3188_GRF_SOC_CON0;
> +			grf = syscon_regmap_lookup_by_compatible("rockchip,rk3188-grf");
> +		}
> +
> +		if (IS_ERR(grf)) {
> +			pr_err("%s: could not find grf dt node\n", __func__);
> +			return;
> +		}
> +
> +		ret = regmap_write(grf, reg, NOC_REMAP << 16);
> +		if (ret < 0) {
> +			pr_err("%s: could not update remap function\n", __func__);
>   			return;
>   		}
>   	}
> diff --git a/arch/arm/mach-rockchip/rockchip.c b/arch/arm/mach-rockchip/rockchip.c
> index 82102fbabf92..c2b919ac99bc 100644
> --- a/arch/arm/mach-rockchip/rockchip.c
> +++ b/arch/arm/mach-rockchip/rockchip.c
> @@ -14,7 +14,6 @@
>   #include <linux/clocksource.h>
>   #include <asm/mach/arch.h>
>   #include <asm/mach/map.h>
> -#include "core.h"
>   #include "pm.h"
>   
>   #define RK3288_TIMER6_7_PHYS 0xff810000
> 
> ---
> base-commit: 9d80aa4617b32f5054c5aa471d06b66704854935
> change-id: 20261017-cpu-c8f61c6466f9
> 
> Best regards,