[Skiboot] [PATCH 24/24] cpu: Support setting HID[RADIX] and set it by default on P9

[Benjamin Herrenschmidt]

This adds new opal_reinit_cpus() flags to setup radix or hash
mode in HID[8] on POWER9.

By default HID[8] will be set. On P9 DD1.0, Linux will change
it as needed. On P9 DD2.0 hash works in radix mode (radix is
really "dual" mode) so KVM won't break and existing kernels
will work.

Newer kernels built for hash will call this to clear the HID bit
and thus get the full size of the TLB as an optimization.

Signed-off-by: Benjamin Herrenschmidt <benh at kernel.crashing.org>
---
 core/cpu.c          | 38 ++++++++++++++++++++++++++++++++++++++
 core/init.c         |  3 +++
 include/cpu.h       |  1 +
 include/opal-api.h  |  9 +++++++++
 include/processor.h |  1 +
 5 files changed, 52 insertions(+)

diff --git a/core/cpu.c b/core/cpu.c
index 31f5c92..75c7008 100644
--- a/core/cpu.c
+++ b/core/cpu.c
@@ -49,10 +49,12 @@ unsigned int cpu_max_pir;
 struct cpu_thread *boot_cpu;
 static struct lock reinit_lock = LOCK_UNLOCKED;
 static bool hile_supported;
+static bool radix_supported;
 static unsigned long hid0_hile;
 static unsigned long hid0_attn;
 static bool pm_enabled;
 static bool current_hile_mode;
+static bool current_radix_mode;
 
 unsigned long cpu_secondary_start __force_data = 0;
 
@@ -700,6 +702,7 @@ void init_boot_cpu(void)
 	case PVR_TYPE_P9:
 		proc_gen = proc_gen_p9;
 		hile_supported = true;
+		radix_supported = true;
 		hid0_hile = SPR_HID0_POWER9_HILE;
 		hid0_attn = SPR_HID0_POWER9_ENABLE_ATTN;
 		break;
@@ -1060,11 +1063,27 @@ static int64_t cpu_change_all_hid0(struct hid0_change_req *req)
 	return OPAL_SUCCESS;
 }
 
+void cpu_set_radix_mode(void)
+{
+	struct hid0_change_req req;
+
+	if (!radix_supported)
+		return;
+
+	req.clr_bits = 0;
+	req.set_bits = SPR_HID0_POWER9_RADIX;
+	cleanup_global_tlb();
+	current_radix_mode = true;
+	cpu_change_all_hid0(&req);
+}
+
 void cpu_fast_reboot_complete(void)
 {
 	/* Fast reboot will have cleared HID0:HILE */
 	current_hile_mode = false;
 
+	/* On P9, restore radix mode */
+	cpu_set_radix_mode();
 }
 
 static int64_t opal_reinit_cpus(uint64_t flags)
@@ -1129,6 +1148,25 @@ static int64_t opal_reinit_cpus(uint64_t flags)
 		}
 	}
 
+	/* If MMU mode change is supported */
+	if (radix_supported &&
+	    (flags & (OPAL_REINIT_CPUS_MMU_HASH |
+		      OPAL_REINIT_CPUS_MMU_RADIX))) {
+		bool radix = !!(flags & OPAL_REINIT_CPUS_MMU_RADIX);
+
+		flags &= ~(OPAL_REINIT_CPUS_MMU_HASH |
+			   OPAL_REINIT_CPUS_MMU_RADIX);
+		if (radix != current_radix_mode) {
+			if (radix)
+				req.set_bits |= SPR_HID0_POWER9_RADIX;
+			else
+				req.clr_bits |= SPR_HID0_POWER9_RADIX;
+
+			cleanup_global_tlb();
+			current_radix_mode = radix;
+		}
+	}
+
 	/* Apply HID bits changes if any */
 	if (req.set_bits || req.clr_bits)
 		cpu_change_all_hid0(&req);
diff --git a/core/init.c b/core/init.c
index 244a78f..9ffbc72 100644
--- a/core/init.c
+++ b/core/init.c
@@ -1045,6 +1045,9 @@ void __noreturn __nomcount main_cpu_entry(const void *fdt)
 
 	prd_register_reserved_memory();
 
+	/* On P9, switch to radix mode by default */
+	cpu_set_radix_mode();
+
 	load_and_boot_kernel(false);
 }
 
diff --git a/include/cpu.h b/include/cpu.h
index 1d19c20..fd3acf7 100644
--- a/include/cpu.h
+++ b/include/cpu.h
@@ -270,6 +270,7 @@ extern unsigned long __attrconst cpu_stack_top(unsigned int pir);
 extern void cpu_idle_job(void);
 extern void cpu_idle_delay(unsigned long delay, unsigned long min_pm);
 
+extern void cpu_set_radix_mode(void);
 extern void cpu_fast_reboot_complete(void);
 
 #endif /* __CPU_H */
diff --git a/include/opal-api.h b/include/opal-api.h
index 2713490..8ffdd4b 100644
--- a/include/opal-api.h
+++ b/include/opal-api.h
@@ -1014,6 +1014,15 @@ struct OpalIoPhb4ErrorData {
 enum {
 	OPAL_REINIT_CPUS_HILE_BE	= (1 << 0),
 	OPAL_REINIT_CPUS_HILE_LE	= (1 << 1),
+
+	/* These two define the base MMU mode of the host on P9
+	 *
+	 * On P9 Nimbus DD2.0 and Cumlus (and later), KVM can still
+	 * create hash guests in "radix" mode with care (full core
+	 * switch only).
+	 */
+	OPAL_REINIT_CPUS_MMU_HASH	= (1 << 2),
+	OPAL_REINIT_CPUS_MMU_RADIX	= (1 << 3),
 };
 
 typedef struct oppanel_line {
diff --git a/include/processor.h b/include/processor.h
index 5906b86..af3fd2b 100644
--- a/include/processor.h
+++ b/include/processor.h
@@ -169,6 +169,7 @@
 #define SPR_HID0_POWER9_HILE		PPC_BIT(4)
 #define SPR_HID0_POWER8_ENABLE_ATTN	PPC_BIT(31)
 #define SPR_HID0_POWER9_ENABLE_ATTN	PPC_BIT(3)
+#define SPR_HID0_POWER9_RADIX		PPC_BIT(8)
 
 /* PVR bits */
 #define SPR_PVR_TYPE			0xffff0000
-- 
2.9.4