plat: imx8mq: Add workaround code for ERR11171 on imx8mq

This new workaround takes advantage of the per core IMR registers in GPC in
order to unmask the IRQ0, still generated by the 12bit in IOMUX_GPR register
(which now remains always set), so it can only wake up one core at the time.
Also, this entire workaround has now been moved here in TF-A, allowing the
kernel side to be minimal.

Another advantage this workaround brings is the removal of the 50us delay
(which was necessary before in gic_raise_softirq in kernel) by allowing the
core that is waking up to mask his own IRQ0 in the suspend finish callback.

One important change here is the way the cores are woken up in
dram_dvfs_handler. Since the wake up mechanism has changed from asserting the
12th bit in IOMUX_GPR and leaving the IMR1 1st bit on for each core to exactly
the reverse, that is, leaving the IOMUX_GPR 12th bit always set and then
masking/unmasking the IMR1 1st bit for each independent core, we need to use
the imx_gpc_core_wake to wake up the cores.

Also, the 50us udelay is moved to TF-A (inside imx_pwr_domain_off) from kernel
(gic_raise_softirq), since the new cpuidle workaround does not need it in order
to clean the IOMUX_GPC 12bit. For now, the udelay seems to be still needed
in order to delay the affinity info OFF for the dying core. This is something
that needs further investigation.

Signed-off-by: Abel Vesa <abel.vesa@nxp.com>
Signed-off-by: Jacky Bai <ping.bai@nxp.com>

1 files changed, 206 insertions, 12 deletions

diff --git a/plat/imx/imx8m/imx8mq/gpc.c b/plat/imx/imx8m/imx8mq/gpc.c
index 9b6e9951..c15c8739 100644
--- a/plat/imx/imx8m/imx8mq/gpc.c
+++ b/plat/imx/imx8m/imx8mq/gpc.c
@@ -8,16 +8,183 @@
 #include <stdint.h>
 #include <stdbool.h>
 
+#include <arch_helpers.h>
 #include <common/debug.h>
 #include <common/runtime_svc.h>
 #include <drivers/delay_timer.h>
 #include <lib/mmio.h>
 #include <lib/psci/psci.h>
+#include <lib/spinlock.h>
 #include <platform_def.h>
+#include <plat/common/platform.h>
 #include <services/std_svc.h>
 
 #include <gpc.h>
 
+#define FSL_SIP_CONFIG_GPC_MASK		0x00
+#define FSL_SIP_CONFIG_GPC_UNMASK	0x01
+#define FSL_SIP_CONFIG_GPC_SET_WAKE	0x02
+#define FSL_SIP_CONFIG_GPC_PM_DOMAIN	0x03
+#define FSL_SIP_CONFIG_GPC_SET_AFF	0x04
+#define FSL_SIP_CONFIG_GPC_CORE_WAKE	0x05
+
+static uint32_t gpc_saved_imrs[16];
+static uint32_t gpc_wake_irqs[4];
+static uint32_t gpc_imr_offset[] = {
+	IMX_GPC_BASE + IMR1_CORE0_A53,
+	IMX_GPC_BASE + IMR1_CORE1_A53,
+	IMX_GPC_BASE + IMR1_CORE2_A53,
+	IMX_GPC_BASE + IMR1_CORE3_A53,
+	IMX_GPC_BASE + IMR1_CORE0_M4,
+};
+
+spinlock_t gpc_imr_lock[4];
+
+static void gpc_imr_core_spin_lock(unsigned int core_id)
+{
+	spin_lock(&gpc_imr_lock[core_id]);
+}
+
+static void gpc_imr_core_spin_unlock(unsigned int core_id)
+{
+	spin_unlock(&gpc_imr_lock[core_id]);
+}
+
+static void gpc_save_imr_lpm(unsigned int core_id, unsigned int imr_idx)
+{
+	uint32_t reg = gpc_imr_offset[core_id] + imr_idx * 4;
+
+	gpc_imr_core_spin_lock(core_id);
+
+	gpc_saved_imrs[core_id + imr_idx * 4] = mmio_read_32(reg);
+	mmio_write_32(reg, ~gpc_wake_irqs[imr_idx]);
+
+	gpc_imr_core_spin_unlock(core_id);
+}
+
+static void gpc_restore_imr_lpm(unsigned int core_id, unsigned int imr_idx)
+{
+	uint32_t reg = gpc_imr_offset[core_id] + imr_idx * 4;
+	uint32_t val = gpc_saved_imrs[core_id + imr_idx * 4];
+
+	gpc_imr_core_spin_lock(core_id);
+
+	mmio_write_32(reg, val);
+
+	gpc_imr_core_spin_unlock(core_id);
+}
+
+/*
+ * On i.MX8MQ, only in system suspend mode, the A53 cluster can
+ * enter LPM mode and shutdown the A53 PLAT power domain. So LPM
+ * wakeup only used for system suspend. when system enter suspend,
+ * any A53 CORE can be the last core to suspend the system, But
+ * the LPM wakeup can only use the C0's IMR to wakeup A53 cluster
+ * from LPM, so save C0's IMRs before suspend, restore back after
+ * resume.
+ */
+void imx_set_sys_wakeup(unsigned int last_core, bool pdn)
+{
+	unsigned int imr, core;
+
+	if (pdn)
+		for (imr = 0; imr < 4; imr++)
+			for (core = 0; core < 4; core++)
+				gpc_save_imr_lpm(core, imr);
+	else
+		for (imr = 0; imr < 4; imr++)
+			for (core = 0; core < 4; core++)
+				gpc_restore_imr_lpm(core, imr);
+}
+
+static void imx_gpc_hwirq_mask(unsigned int hwirq)
+{
+	uintptr_t reg;
+	unsigned int val;
+
+	gpc_imr_core_spin_lock(0);
+	reg = gpc_imr_offset[0] + (hwirq / 32) * 4;
+	val = mmio_read_32(reg);
+	val |= 1 << hwirq % 32;
+	mmio_write_32(reg, val);
+	gpc_imr_core_spin_unlock(0);
+}
+
+static void imx_gpc_hwirq_unmask(unsigned int hwirq)
+{
+	uintptr_t reg;
+	unsigned int val;
+
+	gpc_imr_core_spin_lock(0);
+	reg = gpc_imr_offset[0] + (hwirq / 32) * 4;
+	val = mmio_read_32(reg);
+	val &= ~(1 << hwirq % 32);
+	mmio_write_32(reg, val);
+	gpc_imr_core_spin_unlock(0);
+}
+
+static void imx_gpc_set_wake(uint32_t hwirq, unsigned int on)
+{
+	uint32_t mask, idx;
+
+	mask = 1 << hwirq % 32;
+	idx = hwirq / 32;
+	gpc_wake_irqs[idx] = on ? gpc_wake_irqs[idx] | mask :
+				 gpc_wake_irqs[idx] & ~mask;
+}
+
+static void imx_gpc_mask_irq0(uint32_t core_id, uint32_t mask)
+{
+	gpc_imr_core_spin_lock(core_id);
+	if (mask)
+		mmio_setbits_32(gpc_imr_offset[core_id], 1);
+	else
+		mmio_clrbits_32(gpc_imr_offset[core_id], 1);
+	dsb();
+	gpc_imr_core_spin_unlock(core_id);
+}
+
+void imx_gpc_core_wake(uint32_t cpumask)
+{
+	for (int i = 0; i < 4; i++)
+		if (cpumask & (1 << i))
+			imx_gpc_mask_irq0(i, false);
+}
+
+void imx_gpc_set_a53_core_awake(uint32_t core_id)
+{
+	imx_gpc_mask_irq0(core_id, true);
+}
+
+static void imx_gpc_set_affinity(uint32_t hwirq, unsigned cpu_idx)
+{
+	uintptr_t reg;
+	unsigned int val;
+
+	/*
+	 * using the mask/unmask bit as affinity function.unmask the
+	 * IMR bit to enable IRQ wakeup for this core.
+	 */
+	gpc_imr_core_spin_lock(cpu_idx);
+	reg = gpc_imr_offset[cpu_idx] + (hwirq / 32) * 4;
+	val = mmio_read_32(reg);
+	val &= ~(1 << hwirq % 32);
+	mmio_write_32(reg, val);
+	gpc_imr_core_spin_unlock(cpu_idx);
+
+	/* clear affinity of other core */
+	for (int i = 0; i < 4; i++) {
+		if (cpu_idx != i) {
+			gpc_imr_core_spin_lock(i);
+			reg = gpc_imr_offset[i] + (hwirq / 32) * 4;
+			val = mmio_read_32(reg);
+			val |= (1 << hwirq % 32);
+			mmio_write_32(reg, val);
+			gpc_imr_core_spin_unlock(i);
+		}
+	}
+}
+
 /* use wfi power down the core */
 void imx_set_cpu_pwr_off(unsigned int core_id)
 {
@@ -125,26 +292,53 @@ void imx_set_cluster_powerdown(unsigned int last_core, uint8_t power_state)
 	}
 }
 
+int imx_gpc_handler(uint32_t smc_fid,
+			  u_register_t x1,
+			  u_register_t x2,
+			  u_register_t x3)
+{
+	switch(x1) {
+	case FSL_SIP_CONFIG_GPC_CORE_WAKE:
+		imx_gpc_core_wake(x2);
+		break;
+	case FSL_SIP_CONFIG_GPC_SET_WAKE:
+		imx_gpc_set_wake(x2, x3);
+		break;
+	case FSL_SIP_CONFIG_GPC_MASK:
+		imx_gpc_hwirq_mask(x2);
+		break;
+	case FSL_SIP_CONFIG_GPC_UNMASK:
+		imx_gpc_hwirq_unmask(x2);
+		break;
+	case FSL_SIP_CONFIG_GPC_SET_AFF:
+		imx_gpc_set_affinity(x2, x3);
+		break;
+	default:
+		return SMC_UNK;
+	}
+
+	return 0;
+}
+
 void imx_gpc_init(void)
 {
 	uint32_t val;
-	int i;
+	int i, j;
+
 	/* mask all the interrupt by default */
-	for (i = 0; i < 4; i++) {
-		mmio_write_32(IMX_GPC_BASE + IMR1_CORE0_A53 + i * 4, ~0x0);
-		mmio_write_32(IMX_GPC_BASE + IMR1_CORE1_A53 + i * 4, ~0x0);
-		mmio_write_32(IMX_GPC_BASE + IMR1_CORE2_A53 + i * 4, ~0x0);
-		mmio_write_32(IMX_GPC_BASE + IMR1_CORE3_A53 + i * 4, ~0x0);
-		mmio_write_32(IMX_GPC_BASE + IMR1_CORE0_M4 + i * 4, ~0x0);
-	}
+	for (i = 0; i < 4; i++)
+		for (j = 0; j < 5; j++)
+			mmio_write_32(gpc_imr_offset[j] + i * 4, ~0x0);
+
 	/* Due to the hardware design requirement, need to make
 	 * sure GPR interrupt(#32) is unmasked during RUN mode to
 	 * avoid entering DSM mode by mistake.
 	 */
-	mmio_write_32(IMX_GPC_BASE + IMR1_CORE0_A53, 0xFFFFFFFE);
-	mmio_write_32(IMX_GPC_BASE + IMR1_CORE1_A53, 0xFFFFFFFE);
-	mmio_write_32(IMX_GPC_BASE + IMR1_CORE2_A53, 0xFFFFFFFE);
-	mmio_write_32(IMX_GPC_BASE + IMR1_CORE3_A53, 0xFFFFFFFE);
+	for (i = 0; i < 4; i++)
+		mmio_write_32(gpc_imr_offset[i], ~0x1);
+
+	/* leave the IOMUX_GPC bit 12 on for core wakeup */
+	mmio_setbits_32(IMX_IOMUX_GPR_BASE + 0x4, 1 << 12);
 
 	/* use external IRQs to wakeup C0~C3 from LPM */
 	val = mmio_read_32(IMX_GPC_BASE + LPCR_A53_BSC);