[PATCH v3 0/1] Add L1 and L2 error detection for A53 and A57

[PATCH v3 0/1] Add L1 and L2 error detection for A53 and A57

[Sascha Hauer]

Hi all,

It's been a while since I last sent this, so here's an update.

Like Rob suggested I dropped using a virtual device tree node
to attach the driver to, so from the original three patches series
is only one patch left now. I think I have integrated all review
feedback I got last time, please have a look.

One thing that I find unfortunate in the driver is that I have to
register the "cpu" hierarchy under edac for all CPUs. Only the
CPUs supported by this driver actually count anything. On JUNO for
example we have four cortex-a53 cores (which are supported by this
driver) and four cortex-a72 cores (which are not supported).
Nevertheless we create the directory hierarchy for all eight cores.
I don't know how I should work around this, but maybe James has an
idea.

As usual, feedback welcome.

Sascha

Changes since v2:
- drop usage of virtual dt node (Robh)
- use read_sysreg_s instead of open coded variant (James Morse)
- separate error retrieving from error reporting
- use smp_call_function_single rather than smp_call_function_single_async
- make driver single instance and register all 'cpu' hierarchy up front once

Changes since v1:
- Split dt-binding into separate patch
- Sort local function variables in reverse-xmas tree order
- drop unnecessary comparison and make variable bool

Sascha Hauer (1):
  drivers/edac: Add L1 and L2 error detection for A53 and A57

 drivers/edac/Kconfig              |   6 +
 drivers/edac/Makefile             |   1 +
 drivers/edac/cortex_arm64_l1_l2.c | 218 ++++++++++++++++++++++++++++++
 3 files changed, 225 insertions(+)
 create mode 100644 drivers/edac/cortex_arm64_l1_l2.c

-- 
2.20.1

[PATCH 1/1] drivers/edac: Add L1 and L2 error detection for A53 and A57

[Sascha Hauer]

The Cortex A53 and A57 cores have error detection capabilities for the
L1/L2 Caches, this patch adds a driver for them.

Unfortunately there is no robust way to inject errors into the caches,
so this driver doesn't contain any code to actually test it. It has
been tested though with code taken from an older version of this driver
found here: https://lkml.org/lkml/2018/3/14/1203. For reasons stated
in this thread the error injection code is not suitable for mainline,
so it is removed from the driver.

Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
---
 drivers/edac/Kconfig              |   6 +
 drivers/edac/Makefile             |   1 +
 drivers/edac/cortex_arm64_l1_l2.c | 218 ++++++++++++++++++++++++++++++
 3 files changed, 225 insertions(+)
 create mode 100644 drivers/edac/cortex_arm64_l1_l2.c

diff --git a/drivers/edac/Kconfig b/drivers/edac/Kconfig
index 81c42664f21b..116ad56534a4 100644
--- a/drivers/edac/Kconfig
+++ b/drivers/edac/Kconfig
@@ -539,4 +539,10 @@ config EDAC_DMC520
 	  Support for error detection and correction on the
 	  SoCs with ARM DMC-520 DRAM controller.
 
+config EDAC_CORTEX_ARM64_L1_L2
+	tristate "ARM Cortex A57/A53"
+	depends on ARM64
+	help
+	  Support for L1/L2 cache error detection on ARM Cortex A57 and A53.
+
 endif # EDAC
diff --git a/drivers/edac/Makefile b/drivers/edac/Makefile
index 464d3d8d850a..a324fbc1693e 100644
--- a/drivers/edac/Makefile
+++ b/drivers/edac/Makefile
@@ -89,3 +89,4 @@ obj-$(CONFIG_EDAC_QCOM)			+= qcom_edac.o
 obj-$(CONFIG_EDAC_ASPEED)		+= aspeed_edac.o
 obj-$(CONFIG_EDAC_BLUEFIELD)		+= bluefield_edac.o
 obj-$(CONFIG_EDAC_DMC520)		+= dmc520_edac.o
+obj-$(CONFIG_EDAC_CORTEX_ARM64_L1_L2)	+= cortex_arm64_l1_l2.o
diff --git a/drivers/edac/cortex_arm64_l1_l2.c b/drivers/edac/cortex_arm64_l1_l2.c
new file mode 100644
index 000000000000..3b1e2f3ccab6
--- /dev/null
+++ b/drivers/edac/cortex_arm64_l1_l2.c
@@ -0,0 +1,218 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Cortex A57 and A53 EDAC L1 and L2 cache error detection
+ *
+ * Copyright (c) 2020 Pengutronix, Sascha Hauer <s.hauer@pengutronix.de>
+ *
+ * Based on Code from:
+ * Copyright (c) 2018, NXP Semiconductor
+ * Author: York Sun <york.sun@nxp.com>
+ *
+ */
+
+#include <linux/module.h>
+#include <linux/of_platform.h>
+#include <linux/of_device.h>
+#include <linux/bitfield.h>
+#include <asm/smp_plat.h>
+
+#include "edac_module.h"
+
+#define DRVNAME				"cortex-arm64-edac"
+
+#define CPUMERRSR_EL1_RAMID		GENMASK(30, 24)
+
+#define CPUMERRSR_EL1_VALID		BIT(31)
+#define CPUMERRSR_EL1_FATAL		BIT(63)
+
+#define L1_I_TAG_RAM			0x00
+#define L1_I_DATA_RAM			0x01
+#define L1_D_TAG_RAM			0x08
+#define L1_D_DATA_RAM			0x09
+#define L1_D_DIRTY_RAM			0x14
+#define TLB_RAM				0x18
+
+#define L2MERRSR_EL1_VALID		BIT(31)
+#define L2MERRSR_EL1_FATAL		BIT(63)
+
+struct merrsr {
+	u64 cpumerr;
+	u64 l2merr;
+};
+
+#define MESSAGE_SIZE 64
+
+#define SYS_CPUMERRSR_EL1			sys_reg(3, 1, 15, 2, 2)
+#define SYS_L2MERRSR_EL1			sys_reg(3, 1, 15, 2, 3)
+
+static struct cpumask compat_mask;
+
+static void report_errors(struct edac_device_ctl_info *edac_ctl, int cpu,
+			  struct merrsr *merrsr)
+{
+	char msg[MESSAGE_SIZE];
+	u64 cpumerr = merrsr->cpumerr;
+	u64 l2merr = merrsr->l2merr;
+
+	if (cpumerr & CPUMERRSR_EL1_VALID) {
+		const char *str;
+		bool fatal = cpumerr & CPUMERRSR_EL1_FATAL;
+
+		switch (FIELD_GET(CPUMERRSR_EL1_RAMID, cpumerr)) {
+		case L1_I_TAG_RAM:
+			str = "L1-I Tag RAM";
+			break;
+		case L1_I_DATA_RAM:
+			str = "L1-I Data RAM";
+			break;
+		case L1_D_TAG_RAM:
+			str = "L1-D Tag RAM";
+			break;
+		case L1_D_DATA_RAM:
+			str = "L1-D Data RAM";
+			break;
+		case L1_D_DIRTY_RAM:
+			str = "L1 Dirty RAM";
+			break;
+		case TLB_RAM:
+			str = "TLB RAM";
+			break;
+		default:
+			str = "unknown";
+			break;
+		}
+
+		snprintf(msg, MESSAGE_SIZE, "%s %s error(s) on CPU %d",
+			 str, fatal ? "fatal" : "correctable", cpu);
+
+		if (fatal)
+			edac_device_handle_ue(edac_ctl, cpu, 0, msg);
+		else
+			edac_device_handle_ce(edac_ctl, cpu, 0, msg);
+	}
+
+	if (l2merr & L2MERRSR_EL1_VALID) {
+		bool fatal = l2merr & L2MERRSR_EL1_FATAL;
+
+		snprintf(msg, MESSAGE_SIZE, "L2 %s error(s) on CPU %d",
+			 fatal ? "fatal" : "correctable", cpu);
+		if (fatal)
+			edac_device_handle_ue(edac_ctl, cpu, 1, msg);
+		else
+			edac_device_handle_ce(edac_ctl, cpu, 1, msg);
+	}
+}
+
+static void read_errors(void *data)
+{
+	struct merrsr *merrsr = data;
+
+	merrsr->cpumerr = read_sysreg_s(SYS_CPUMERRSR_EL1);
+	write_sysreg_s(0, SYS_CPUMERRSR_EL1);
+	merrsr->l2merr = read_sysreg_s(SYS_L2MERRSR_EL1);
+	write_sysreg_s(0, SYS_L2MERRSR_EL1);
+}
+
+static void cortex_arm64_edac_check(struct edac_device_ctl_info *edac_ctl)
+{
+	struct merrsr merrsr;
+	int cpu;
+
+	for_each_cpu_and(cpu, cpu_online_mask, &compat_mask) {
+		smp_call_function_single(cpu, read_errors, &merrsr, true);
+		report_errors(edac_ctl, cpu, &merrsr);
+	}
+}
+
+static int cortex_arm64_edac_probe(struct platform_device *pdev)
+{
+	struct edac_device_ctl_info *edac_ctl;
+	struct device *dev = &pdev->dev;
+	int rc;
+
+	edac_ctl = edac_device_alloc_ctl_info(0, "cpu",
+					      num_possible_cpus(), "L", 2, 1, NULL, 0,
+					      edac_device_alloc_index());
+	if (!edac_ctl)
+		return -ENOMEM;
+
+	edac_ctl->edac_check = cortex_arm64_edac_check;
+	edac_ctl->dev = dev;
+	edac_ctl->mod_name = dev_name(dev);
+	edac_ctl->dev_name = dev_name(dev);
+	edac_ctl->ctl_name = DRVNAME;
+	dev_set_drvdata(dev, edac_ctl);
+
+	rc = edac_device_add_device(edac_ctl);
+	if (rc)
+		goto out_dev;
+
+	return 0;
+
+out_dev:
+	edac_device_free_ctl_info(edac_ctl);
+
+	return rc;
+}
+
+static int cortex_arm64_edac_remove(struct platform_device *pdev)
+{
+	struct edac_device_ctl_info *edac_ctl = dev_get_drvdata(&pdev->dev);
+
+	edac_device_del_device(edac_ctl->dev);
+	edac_device_free_ctl_info(edac_ctl);
+
+	return 0;
+}
+
+static const struct of_device_id cortex_arm64_edac_of_match[] = {
+	{ .compatible = "arm,cortex-a53" },
+	{ .compatible = "arm,cortex-a57" },
+	{}
+};
+MODULE_DEVICE_TABLE(of, cortex_arm64_edac_of_match);
+
+static struct platform_driver cortex_arm64_edac_driver = {
+	.probe = cortex_arm64_edac_probe,
+	.remove = cortex_arm64_edac_remove,
+	.driver = {
+		.name = DRVNAME,
+	},
+};
+
+static int __init cortex_arm64_edac_driver_init(void)
+{
+	struct device_node *np;
+	int cpu;
+	struct platform_device *pdev;
+	int err;
+
+	for_each_possible_cpu(cpu) {
+		np = of_get_cpu_node(cpu, NULL);
+
+		if (of_match_node(cortex_arm64_edac_of_match, np))
+			cpumask_set_cpu(cpu, &compat_mask);
+	}
+
+	if (cpumask_empty(&compat_mask))
+		return 0;
+
+	err = platform_driver_register(&cortex_arm64_edac_driver);
+        if (err)
+                return err;
+
+	pdev = platform_device_register_simple(DRVNAME, -1, NULL, 0);
+	if (IS_ERR(pdev)) {
+		pr_err("failed to register cortex arm64 edac device\n");
+		platform_driver_unregister(&cortex_arm64_edac_driver);
+		return PTR_ERR(pdev);
+	}
+
+	return 0;
+}
+
+device_initcall(cortex_arm64_edac_driver_init);
+
+MODULE_LICENSE("GPL v2");
+MODULE_AUTHOR("Sascha Hauer <s.hauer@pengutronix.de>");
+MODULE_DESCRIPTION("Cortex A57 and A53 L1 and L2 cache EDAC driver");
-- 
2.20.1

Re: [PATCH 1/1] drivers/edac: Add L1 and L2 error detection for A53 and A57

[Marc Zyngier]

On 2021-01-07 10:38, Sascha Hauer wrote:
> The Cortex A53 and A57 cores have error detection capabilities for the
> L1/L2 Caches, this patch adds a driver for them.
> 
> Unfortunately there is no robust way to inject errors into the caches,
> so this driver doesn't contain any code to actually test it. It has
> been tested though with code taken from an older version of this driver
> found here: https://lkml.org/lkml/2018/3/14/1203. For reasons stated
> in this thread the error injection code is not suitable for mainline,
> so it is removed from the driver.
> 
> Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
> ---
>  drivers/edac/Kconfig              |   6 +
>  drivers/edac/Makefile             |   1 +
>  drivers/edac/cortex_arm64_l1_l2.c | 218 ++++++++++++++++++++++++++++++
>  3 files changed, 225 insertions(+)
>  create mode 100644 drivers/edac/cortex_arm64_l1_l2.c
> 
> diff --git a/drivers/edac/Kconfig b/drivers/edac/Kconfig
> index 81c42664f21b..116ad56534a4 100644
> --- a/drivers/edac/Kconfig
> +++ b/drivers/edac/Kconfig
> @@ -539,4 +539,10 @@ config EDAC_DMC520
>  	  Support for error detection and correction on the
>  	  SoCs with ARM DMC-520 DRAM controller.
> 
> +config EDAC_CORTEX_ARM64_L1_L2
> +	tristate "ARM Cortex A57/A53"
> +	depends on ARM64
> +	help
> +	  Support for L1/L2 cache error detection on ARM Cortex A57 and A53.
> +
>  endif # EDAC
> diff --git a/drivers/edac/Makefile b/drivers/edac/Makefile
> index 464d3d8d850a..a324fbc1693e 100644
> --- a/drivers/edac/Makefile
> +++ b/drivers/edac/Makefile
> @@ -89,3 +89,4 @@ obj-$(CONFIG_EDAC_QCOM)			+= qcom_edac.o
>  obj-$(CONFIG_EDAC_ASPEED)		+= aspeed_edac.o
>  obj-$(CONFIG_EDAC_BLUEFIELD)		+= bluefield_edac.o
>  obj-$(CONFIG_EDAC_DMC520)		+= dmc520_edac.o
> +obj-$(CONFIG_EDAC_CORTEX_ARM64_L1_L2)	+= cortex_arm64_l1_l2.o
> diff --git a/drivers/edac/cortex_arm64_l1_l2.c
> b/drivers/edac/cortex_arm64_l1_l2.c
> new file mode 100644
> index 000000000000..3b1e2f3ccab6
> --- /dev/null
> +++ b/drivers/edac/cortex_arm64_l1_l2.c
> @@ -0,0 +1,218 @@
> +// SPDX-License-Identifier: GPL-2.0
> +/*
> + * Cortex A57 and A53 EDAC L1 and L2 cache error detection
> + *
> + * Copyright (c) 2020 Pengutronix, Sascha Hauer 
> <s.hauer@pengutronix.de>
> + *
> + * Based on Code from:
> + * Copyright (c) 2018, NXP Semiconductor
> + * Author: York Sun <york.sun@nxp.com>
> + *
> + */
> +
> +#include <linux/module.h>
> +#include <linux/of_platform.h>
> +#include <linux/of_device.h>
> +#include <linux/bitfield.h>
> +#include <asm/smp_plat.h>
> +
> +#include "edac_module.h"
> +
> +#define DRVNAME				"cortex-arm64-edac"
> +
> +#define CPUMERRSR_EL1_RAMID		GENMASK(30, 24)
> +
> +#define CPUMERRSR_EL1_VALID		BIT(31)
> +#define CPUMERRSR_EL1_FATAL		BIT(63)
> +
> +#define L1_I_TAG_RAM			0x00
> +#define L1_I_DATA_RAM			0x01
> +#define L1_D_TAG_RAM			0x08
> +#define L1_D_DATA_RAM			0x09
> +#define L1_D_DIRTY_RAM			0x14
> +#define TLB_RAM				0x18
> +
> +#define L2MERRSR_EL1_VALID		BIT(31)
> +#define L2MERRSR_EL1_FATAL		BIT(63)
> +
> +struct merrsr {
> +	u64 cpumerr;
> +	u64 l2merr;
> +};
> +
> +#define MESSAGE_SIZE 64
> +
> +#define SYS_CPUMERRSR_EL1			sys_reg(3, 1, 15, 2, 2)
> +#define SYS_L2MERRSR_EL1			sys_reg(3, 1, 15, 2, 3)

Oh, IMPDEF registers...

This means that a kernel with this driver running in a virtualized
environment is likely to just take an UNDEF and die. Not great.

Also, how does it work if EL3 implements the same functionality?

         M.
-- 
Who you jivin' with that Cosmik Debris?

Re: [PATCH 1/1] drivers/edac: Add L1 and L2 error detection for A53 and A57

[Mark Rutland]

Hi Sascha,

On Thu, Jan 07, 2021 at 11:38:19AM +0100, Sascha Hauer wrote:
> The Cortex A53 and A57 cores have error detection capabilities for the
> L1/L2 Caches, this patch adds a driver for them.
> 
> Unfortunately there is no robust way to inject errors into the caches,
> so this driver doesn't contain any code to actually test it. It has
> been tested though with code taken from an older version of this driver
> found here: https://lkml.org/lkml/2018/3/14/1203. For reasons stated
> in this thread the error injection code is not suitable for mainline,
> so it is removed from the driver.
> 
> Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
> ---
>  drivers/edac/Kconfig              |   6 +
>  drivers/edac/Makefile             |   1 +
>  drivers/edac/cortex_arm64_l1_l2.c | 218 ++++++++++++++++++++++++++++++
>  3 files changed, 225 insertions(+)
>  create mode 100644 drivers/edac/cortex_arm64_l1_l2.c

[...]

> +#define SYS_CPUMERRSR_EL1			sys_reg(3, 1, 15, 2, 2)
> +#define SYS_L2MERRSR_EL1			sys_reg(3, 1, 15, 2, 3)

Just to note -- these registers are IMPLEMENTATION DEFINED, and so may
be trapped by a hypervisor or messed around with arbitrarily by FW.

> +static void read_errors(void *data)
> +{
> +	struct merrsr *merrsr = data;
> +
> +	merrsr->cpumerr = read_sysreg_s(SYS_CPUMERRSR_EL1);
> +	write_sysreg_s(0, SYS_CPUMERRSR_EL1);
> +	merrsr->l2merr = read_sysreg_s(SYS_L2MERRSR_EL1);
> +	write_sysreg_s(0, SYS_L2MERRSR_EL1);
> +}

... and so here the MRS instructions may behave as UNDEFINED in some
cases. Foer example, KVM would trap these via HCR_EL2.TIDCP, and emulate
these as UNDEFINED.

> +static const struct of_device_id cortex_arm64_edac_of_match[] = {
> +	{ .compatible = "arm,cortex-a53" },
> +	{ .compatible = "arm,cortex-a57" },
> +	{}
> +};
> +MODULE_DEVICE_TABLE(of, cortex_arm64_edac_of_match);

... and so it's not safe to probe this based on the CPU compatible
string alone. If we really need to support this, it will need its own
binding (to describe that the reigsters are safe to access and behave in
a useful way).

Thanks,
Mark.

Re: [PATCH 1/1] drivers/edac: Add L1 and L2 error detection for A53 and A57

[Sascha Hauer]

On Thu, Jan 07, 2021 at 11:15:20AM +0000, Mark Rutland wrote:
> Hi Sascha,
> 
> On Thu, Jan 07, 2021 at 11:38:19AM +0100, Sascha Hauer wrote:
> > The Cortex A53 and A57 cores have error detection capabilities for the
> > L1/L2 Caches, this patch adds a driver for them.
> > 
> > Unfortunately there is no robust way to inject errors into the caches,
> > so this driver doesn't contain any code to actually test it. It has
> > been tested though with code taken from an older version of this driver
> > found here: https://lkml.org/lkml/2018/3/14/1203. For reasons stated
> > in this thread the error injection code is not suitable for mainline,
> > so it is removed from the driver.
> > 
> > Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
> > ---
> >  drivers/edac/Kconfig              |   6 +
> >  drivers/edac/Makefile             |   1 +
> >  drivers/edac/cortex_arm64_l1_l2.c | 218 ++++++++++++++++++++++++++++++
> >  3 files changed, 225 insertions(+)
> >  create mode 100644 drivers/edac/cortex_arm64_l1_l2.c
> 
> [...]
> 
> > +#define SYS_CPUMERRSR_EL1			sys_reg(3, 1, 15, 2, 2)
> > +#define SYS_L2MERRSR_EL1			sys_reg(3, 1, 15, 2, 3)
> 
> Just to note -- these registers are IMPLEMENTATION DEFINED, and so may
> be trapped by a hypervisor or messed around with arbitrarily by FW.
> 
> > +static void read_errors(void *data)
> > +{
> > +	struct merrsr *merrsr = data;
> > +
> > +	merrsr->cpumerr = read_sysreg_s(SYS_CPUMERRSR_EL1);
> > +	write_sysreg_s(0, SYS_CPUMERRSR_EL1);
> > +	merrsr->l2merr = read_sysreg_s(SYS_L2MERRSR_EL1);
> > +	write_sysreg_s(0, SYS_L2MERRSR_EL1);
> > +}
> 
> ... and so here the MRS instructions may behave as UNDEFINED in some
> cases. Foer example, KVM would trap these via HCR_EL2.TIDCP, and emulate
> these as UNDEFINED.
> 
> > +static const struct of_device_id cortex_arm64_edac_of_match[] = {
> > +	{ .compatible = "arm,cortex-a53" },
> > +	{ .compatible = "arm,cortex-a57" },
> > +	{}
> > +};
> > +MODULE_DEVICE_TABLE(of, cortex_arm64_edac_of_match);
> 
> ... and so it's not safe to probe this based on the CPU compatible
> string alone. If we really need to support this, it will need its own
> binding (to describe that the reigsters are safe to access and behave in
> a useful way).

Rob objected to adding dedicated nodes to attach the driver to. That
leaves adding a property like "edac-enabled" or similar to the CPU nodes
when edac is desired. Would that be ok?

Sascha

-- 
Pengutronix e.K.                           |                             |
Steuerwalder Str. 21                       | http://www.pengutronix.de/  |
31137 Hildesheim, Germany                  | Phone: +49-5121-206917-0    |
Amtsgericht Hildesheim, HRA 2686           | Fax:   +49-5121-206917-5555 |