[PATCH 1/2] arm64: traps: Support for registering SError hooks

Add the ability for drivers to register SError hooks to be run
on a fatal SError interrupt. This allows printing of system-specific
error information to aid with debugging.

Signed-off-by: Mikko Perttunen <mperttunen@nvidia.com>
---
 arch/arm64/include/asm/traps.h |  6 ++++++
 arch/arm64/kernel/traps.c      | 20 ++++++++++++++++++++
 2 files changed, 26 insertions(+)

diff --git a/arch/arm64/include/asm/traps.h b/arch/arm64/include/asm/traps.h
index 54f32a0675df..054fecfa22f0 100644
--- a/arch/arm64/include/asm/traps.h
+++ b/arch/arm64/include/asm/traps.h

@@ -22,8 +22,14 @@ struct undef_hook {
 	int (*fn)(struct pt_regs *regs, u32 instr);
 };
 
+struct serror_hook {
+	struct list_head node;
+	void (*fn)(void);
+};
+
 void register_undef_hook(struct undef_hook *hook);
 void unregister_undef_hook(struct undef_hook *hook);
+void register_serror_hook(struct serror_hook *hook);
 void force_signal_inject(int signal, int code, unsigned long address, unsigned int err);
 void arm64_notify_segfault(unsigned long addr);
 void arm64_force_sig_fault(int signo, int code, unsigned long far, const char *str);

diff --git a/arch/arm64/kernel/traps.c b/arch/arm64/kernel/traps.c
index 273066279bb5..02dbaab71ea3 100644
--- a/arch/arm64/kernel/traps.c
+++ b/arch/arm64/kernel/traps.c

@@ -890,8 +890,23 @@ void panic_bad_stack(struct pt_regs *regs, unsigned int esr, unsigned long far)
 }
 #endif
 
+static LIST_HEAD(serror_hook);
+static DEFINE_RAW_SPINLOCK(serror_lock);
+
+void register_serror_hook(struct serror_hook *hook)
+{
+	unsigned long flags;
+
+	raw_spin_lock_irqsave(&serror_lock, flags);
+	list_add(&hook->node, &serror_hook);
+	raw_spin_unlock_irqrestore(&serror_lock, flags);
+}
+
 void __noreturn arm64_serror_panic(struct pt_regs *regs, u32 esr)
 {
+	struct serror_hook *hook;
+	unsigned long flags;
+
 	console_verbose();
 
 	pr_crit("SError Interrupt on CPU%d, code 0x%08x -- %s\n",

@@ -899,6 +914,11 @@ void __noreturn arm64_serror_panic(struct pt_regs *regs, u32 esr)
 	if (regs)
 		__show_regs(regs);
 
+	raw_spin_lock_irqsave(&serror_lock, flags);
+	list_for_each_entry(hook, &serror_hook, node)
+		hook->fn();
+	raw_spin_unlock_irqrestore(&serror_lock, flags);
+
 	nmi_panic(regs, "Asynchronous SError Interrupt");
 
 	cpu_park_loop();

-- 
2.30.1


_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

Add a driver to hook into SError interrupts and print machine check
status for debugging. Status information is retrieved via SMC. This
is supported by upstream ARM Trusted Firmware.

Signed-off-by: Mikko Perttunen <mperttunen@nvidia.com>
---
 drivers/soc/tegra/Makefile       |  1 +
 drivers/soc/tegra/ari-tegra186.c | 78 ++++++++++++++++++++++++++++++++
 2 files changed, 79 insertions(+)
 create mode 100644 drivers/soc/tegra/ari-tegra186.c

diff --git a/drivers/soc/tegra/Makefile b/drivers/soc/tegra/Makefile
index 9c809c1814bd..054e862b63d8 100644
--- a/drivers/soc/tegra/Makefile
+++ b/drivers/soc/tegra/Makefile

@@ -7,3 +7,4 @@ obj-$(CONFIG_SOC_TEGRA_PMC) += pmc.o
 obj-$(CONFIG_SOC_TEGRA_POWERGATE_BPMP) += powergate-bpmp.o
 obj-$(CONFIG_SOC_TEGRA20_VOLTAGE_COUPLER) += regulators-tegra20.o
 obj-$(CONFIG_SOC_TEGRA30_VOLTAGE_COUPLER) += regulators-tegra30.o
+obj-$(CONFIG_ARCH_TEGRA_186_SOC) += ari-tegra186.o

diff --git a/drivers/soc/tegra/ari-tegra186.c b/drivers/soc/tegra/ari-tegra186.c
new file mode 100644
index 000000000000..51b95af1dc78
--- /dev/null
+++ b/drivers/soc/tegra/ari-tegra186.c

@@ -0,0 +1,78 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * Copyright (c) 2021, NVIDIA CORPORATION.  All rights reserved.
+ */
+
+#include <linux/arm-smccc.h>
+#include <linux/kernel.h>
+#include <linux/of.h>
+
+#include <asm/traps.h>
+
+#define SMC_SIP_INVOKE_MCE			0xc2ffff00
+#define MCE_SMC_READ_MCA			12
+
+#define MCA_ARI_CMD_RD_SERR			1
+
+#define MCA_ARI_RW_SUBIDX_STAT			1
+#define SERR_STATUS_VAL				BIT_ULL(63)
+
+#define MCA_ARI_RW_SUBIDX_ADDR			2
+#define MCA_ARI_RW_SUBIDX_MSC1			3
+#define MCA_ARI_RW_SUBIDX_MSC2			4
+
+static const char * const bank_names[] = {
+	"SYS:DPMU", "ROC:IOB", "ROC:MCB", "ROC:CCE", "ROC:CQX", "ROC:CTU",
+};
+
+static void read_uncore_mca(u8 cmd, u8 idx, u8 subidx, u8 inst, u64 *data)
+{
+	struct arm_smccc_res res;
+
+	arm_smccc_smc(SMC_SIP_INVOKE_MCE | MCE_SMC_READ_MCA,
+		      ((u64)inst << 24) | ((u64)idx << 16) |
+			      ((u64)subidx << 8) | ((u64)cmd << 0),
+		      0, 0, 0, 0, 0, 0, &res);
+
+	*data = res.a2;
+}
+
+static void tegra186_ari_process_serror(void)
+{
+	u64 status;
+	int i;
+
+	for (i = 0; i < ARRAY_SIZE(bank_names); i++) {
+		read_uncore_mca(MCA_ARI_CMD_RD_SERR, i, MCA_ARI_RW_SUBIDX_STAT,
+				0, &status);
+
+		if (status & SERR_STATUS_VAL) {
+			u64 addr, misc1, misc2;
+
+			read_uncore_mca(MCA_ARI_CMD_RD_SERR, i,
+					MCA_ARI_RW_SUBIDX_ADDR, 0, &addr);
+			read_uncore_mca(MCA_ARI_CMD_RD_SERR, i,
+					MCA_ARI_RW_SUBIDX_MSC1, 0, &misc1);
+			read_uncore_mca(MCA_ARI_CMD_RD_SERR, i,
+					MCA_ARI_RW_SUBIDX_MSC2, 0, &misc2);
+
+			pr_crit("Machine Check Error in %s\n"
+				"  status=0x%llx addr=0x%llx\n"
+				"  msc1=0x%llx msc2=0x%llx\n",
+				bank_names[i], status, addr, misc1, misc2);
+		}
+	}
+}
+
+static struct serror_hook tegra186_ari_serror_hook = {
+	.fn = tegra186_ari_process_serror,
+};
+
+static int __init tegra186_ari_init(void)
+{
+	if (of_machine_is_compatible("nvidia,tegra186"))
+		register_serror_hook(&tegra186_ari_serror_hook);
+
+	return 0;
+}
+early_initcall(tegra186_ari_init);

-- 
2.30.1


_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

On Thu, Jun 17, 2021 at 01:40:52PM +0300, Mikko Perttunen wrote:

Add the ability for drivers to register SError hooks to be run
on a fatal SError interrupt. This allows printing of system-specific
error information to aid with debugging.

Signed-off-by: Mikko Perttunen <mperttunen@nvidia.com>

Can't you dump information unconditionally in a panic notifier?

That wouldn't require any new infrastructure.

Thanks,
Mark.

_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

On 6/17/21 2:46 PM, Mark Rutland wrote:

On Thu, Jun 17, 2021 at 01:40:52PM +0300, Mikko Perttunen wrote:

quoted

Add the ability for drivers to register SError hooks to be run
on a fatal SError interrupt. This allows printing of system-specific
error information to aid with debugging.

Signed-off-by: Mikko Perttunen <mperttunen@nvidia.com>

Can't you dump information unconditionally in a panic notifier?

That wouldn't require any new infrastructure.

That does indeed work also, thanks for the tip! Will change to use those.

Thanks,
Mikko

Thanks,
Mark.

quoted

---
  arch/arm64/include/asm/traps.h |  6 ++++++
  arch/arm64/kernel/traps.c      | 20 ++++++++++++++++++++
  2 files changed, 26 insertions(+)

diff --git a/arch/arm64/include/asm/traps.h b/arch/arm64/include/asm/traps.h
index 54f32a0675df..054fecfa22f0 100644
--- a/arch/arm64/include/asm/traps.h
+++ b/arch/arm64/include/asm/traps.h

@@ -22,8 +22,14 @@ struct undef_hook {
  	int (*fn)(struct pt_regs *regs, u32 instr);
  };
  
+struct serror_hook {
+	struct list_head node;
+	void (*fn)(void);
+};
+
  void register_undef_hook(struct undef_hook *hook);
  void unregister_undef_hook(struct undef_hook *hook);
+void register_serror_hook(struct serror_hook *hook);
  void force_signal_inject(int signal, int code, unsigned long address, unsigned int err);
  void arm64_notify_segfault(unsigned long addr);
  void arm64_force_sig_fault(int signo, int code, unsigned long far, const char *str);

diff --git a/arch/arm64/kernel/traps.c b/arch/arm64/kernel/traps.c
index 273066279bb5..02dbaab71ea3 100644
--- a/arch/arm64/kernel/traps.c
+++ b/arch/arm64/kernel/traps.c

@@ -890,8 +890,23 @@ void panic_bad_stack(struct pt_regs *regs, unsigned int esr, unsigned long far)
  }
  #endif
  
+static LIST_HEAD(serror_hook);
+static DEFINE_RAW_SPINLOCK(serror_lock);
+
+void register_serror_hook(struct serror_hook *hook)
+{
+	unsigned long flags;
+
+	raw_spin_lock_irqsave(&serror_lock, flags);
+	list_add(&hook->node, &serror_hook);
+	raw_spin_unlock_irqrestore(&serror_lock, flags);
+}
+
  void __noreturn arm64_serror_panic(struct pt_regs *regs, u32 esr)
  {
+	struct serror_hook *hook;
+	unsigned long flags;
+
  	console_verbose();
  
  	pr_crit("SError Interrupt on CPU%d, code 0x%08x -- %s\n",

@@ -899,6 +914,11 @@ void __noreturn arm64_serror_panic(struct pt_regs *regs, u32 esr)
  	if (regs)
  		__show_regs(regs);
  
+	raw_spin_lock_irqsave(&serror_lock, flags);
+	list_for_each_entry(hook, &serror_hook, node)
+		hook->fn();
+	raw_spin_unlock_irqrestore(&serror_lock, flags);
+
  	nmi_panic(regs, "Asynchronous SError Interrupt");
  
  	cpu_park_loop();

-- 
2.30.1

_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel