platform/mellanox: Add bootctl driver for Mellanox BlueField Soc

This commit adds the bootctl platform driver for Mellanox BlueField
Soc, which queries secure state and controls the eMMC boot partition
swapping by sending SMC calls to ATF running at EL3.

Below are the sequences of typical use case.

  1. User requests boot partition swapping, which could be on-demand or
     during boot-image upgrade via UEFI capsule;

  2. This bootctl driver handles the request and sends SMC call
     to ATF. ATF programs register BREADCRUMB0 which has value
     preserved during warm reset. It also programs eMMC to swap
     the boot partition;

  3. After software reset (rebooting), ATF BL1 (BootRom) checks
     register BREADCRUMB0 and enable watchdog if configured;

  4. If booting fails, the watchdog timer will trigger rebooting.
     In such case, ATF Boot ROM will switch the boot partition
     back to the previous one. This is a robust feature and used
     to prevent failure during boot partition upgrade.

Reviewed-by: Vadim Pasternak <vadimp@mellanox.com>
Signed-off-by: Liming Sun <lsun@mellanox.com>
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>

Documentation/ABI/testing/sysfs-platform-mellanox-bootctl

@@ -0,0 +1,58 @@
+What:		/sys/bus/platform/devices/MLNXBF04:00/driver/lifecycle_state
+Date:		Oct 2019
+KernelVersion:	5.5
+Contact:	"Liming Sun <lsun@mellanox.com>"
+Description:
+		The Life-cycle state of the SoC, which could be one of the
+		following values.
+		  Production - Production state and can be updated to secure
+		  GA Secured - Secure chip and not able to change state
+		  GA Non-Secured - Non-Secure chip and not able to change state
+		  RMA - Return Merchandise Authorization
+
+What:		/sys/bus/platform/devices/MLNXBF04:00/driver/post_reset_wdog
+Date:		Oct 2019
+KernelVersion:	5.5
+Contact:	"Liming Sun <lsun@mellanox.com>"
+Description:
+		The watchdog setting in seconds for the next booting. It's used
+		to reboot the chip and recover it to the old state if the new
+		boot partition fails.
+
+What:		/sys/bus/platform/devices/MLNXBF04:00/driver/reset_action
+Date:		Oct 2019
+KernelVersion:	5.5
+Contact:	"Liming Sun <lsun@mellanox.com>"
+Description:
+		The source of the boot stream for the next reset. It could be
+		one of the following values.
+		  external - boot from external source (USB or PCIe)
+		  emmc - boot from the onchip eMMC
+		  emmc_legacy - boot from the onchip eMMC in legacy (slow) mode
+
+What:		/sys/bus/platform/devices/MLNXBF04:00/driver/second_reset_action
+Date:		Oct 2019
+KernelVersion:	5.5
+Contact:	"Liming Sun <lsun@mellanox.com>"
+Description:
+		Update the source of the boot stream after next reset. It could
+		be one of the following values and will be applied after next
+		reset.
+		  external - boot from external source (USB or PCIe)
+		  emmc - boot from the onchip eMMC
+		  emmc_legacy - boot from the onchip eMMC in legacy (slow) mode
+		  swap_emmc - swap the primary / secondary boot partition
+		  none - cancel the action
+
+What:		/sys/bus/platform/devices/MLNXBF04:00/driver/secure_boot_fuse_state
+Date:		Oct 2019
+KernelVersion:	5.5
+Contact:	"Liming Sun <lsun@mellanox.com>"
+Description:
+		The state of eFuse versions with the following values.
+		  InUse - burnt, valid and currently in use
+		  Used - burnt and valid
+		  Free - not burnt and free to use
+		  Skipped - not burnt but not free (skipped)
+		  Wasted - burnt and invalid
+		  Invalid - not burnt but marked as valid (error state).

MAINTAINERS

@@ -10432,6 +10432,7 @@ M:	Darren Hart <dvhart@infradead.org>
 M:	Vadim Pasternak <vadimp@mellanox.com>
 L:	platform-driver-x86@vger.kernel.org
 S:	Supported
+F:	Documentation/ABI/testing/sysfs-platform-mellanox-bootctl
 F:	drivers/platform/mellanox/
 F:	include/linux/platform_data/mlxreg.h
 

drivers/platform/mellanox/Kconfig

@@ -44,4 +44,16 @@ config MLXBF_TMFIFO
           platform driver support for the TmFifo which supports console
           and networking based on the virtio framework.
 
+config MLXBF_BOOTCTL
+	tristate "Mellanox BlueField Firmware Boot Control driver"
+	depends on ARM64
+	depends on ACPI
+	help
+          The Mellanox BlueField firmware implements functionality to
+          request swapping the primary and alternate eMMC boot partition,
+          and to set up a watchdog that can undo that swap if the system
+          does not boot up correctly. This driver provides sysfs access
+          to the userspace tools, to be used in conjunction with the eMMC
+          device driver to do necessary initial swap of the boot partition.
+
 endif # MELLANOX_PLATFORM

drivers/platform/mellanox/Makefile

@@ -3,6 +3,7 @@
 # Makefile for linux/drivers/platform/mellanox
 # Mellanox Platform-Specific Drivers
 #
+obj-$(CONFIG_MLXBF_BOOTCTL)	+= mlxbf-bootctl.o
 obj-$(CONFIG_MLXBF_TMFIFO)	+= mlxbf-tmfifo.o
 obj-$(CONFIG_MLXREG_HOTPLUG)	+= mlxreg-hotplug.o
 obj-$(CONFIG_MLXREG_IO) += mlxreg-io.o

drivers/platform/mellanox/mlxbf-bootctl.c

@@ -0,0 +1,321 @@
+// SPDX-License-Identifier: GPL-2.0+
+/*
+ * Mellanox boot control driver
+ *
+ * This driver provides a sysfs interface for systems management
+ * software to manage reset-time actions.
+ *
+ * Copyright (C) 2019 Mellanox Technologies
+ */
+
+#include <linux/acpi.h>
+#include <linux/arm-smccc.h>
+#include <linux/module.h>
+#include <linux/platform_device.h>
+
+#include "mlxbf-bootctl.h"
+
+#define MLXBF_BOOTCTL_SB_SECURE_MASK		0x03
+#define MLXBF_BOOTCTL_SB_TEST_MASK		0x0c
+
+#define MLXBF_SB_KEY_NUM			4
+
+/* UUID used to probe ATF service. */
+static const char *mlxbf_bootctl_svc_uuid_str =
+	"89c036b4-e7d7-11e6-8797-001aca00bfc4";
+
+struct mlxbf_bootctl_name {
+	u32 value;
+	const char *name;
+};
+
+static struct mlxbf_bootctl_name boot_names[] = {
+	{ MLXBF_BOOTCTL_EXTERNAL, "external" },
+	{ MLXBF_BOOTCTL_EMMC, "emmc" },
+	{ MLNX_BOOTCTL_SWAP_EMMC, "swap_emmc" },
+	{ MLXBF_BOOTCTL_EMMC_LEGACY, "emmc_legacy" },
+	{ MLXBF_BOOTCTL_NONE, "none" },
+};
+
+static const char * const mlxbf_bootctl_lifecycle_states[] = {
+	[0] = "Production",
+	[1] = "GA Secured",
+	[2] = "GA Non-Secured",
+	[3] = "RMA",
+};
+
+/* ARM SMC call which is atomic and no need for lock. */
+static int mlxbf_bootctl_smc(unsigned int smc_op, int smc_arg)
+{
+	struct arm_smccc_res res;
+
+	arm_smccc_smc(smc_op, smc_arg, 0, 0, 0, 0, 0, 0, &res);
+
+	return res.a0;
+}
+
+/* Return the action in integer or an error code. */
+static int mlxbf_bootctl_reset_action_to_val(const char *action)
+{
+	int i;
+
+	for (i = 0; i < ARRAY_SIZE(boot_names); i++)
+		if (sysfs_streq(boot_names[i].name, action))
+			return boot_names[i].value;
+
+	return -EINVAL;
+}
+
+/* Return the action in string. */
+static const char *mlxbf_bootctl_action_to_string(int action)
+{
+	int i;
+
+	for (i = 0; i < ARRAY_SIZE(boot_names); i++)
+		if (boot_names[i].value == action)
+			return boot_names[i].name;
+
+	return "invalid action";
+}
+
+static ssize_t post_reset_wdog_show(struct device *dev,
+				    struct device_attribute *attr, char *buf)
+{
+	int ret;
+
+	ret = mlxbf_bootctl_smc(MLXBF_BOOTCTL_GET_POST_RESET_WDOG, 0);
+	if (ret < 0)
+		return ret;
+
+	return sprintf(buf, "%d\n", ret);
+}
+
+static ssize_t post_reset_wdog_store(struct device *dev,
+				     struct device_attribute *attr,
+				     const char *buf, size_t count)
+{
+	unsigned long value;
+	int ret;
+
+	ret = kstrtoul(buf, 10, &value);
+	if (ret)
+		return ret;
+
+	ret = mlxbf_bootctl_smc(MLXBF_BOOTCTL_SET_POST_RESET_WDOG, value);
+	if (ret < 0)
+		return ret;
+
+	return count;
+}
+
+static ssize_t mlxbf_bootctl_show(int smc_op, char *buf)
+{
+	int action;
+
+	action = mlxbf_bootctl_smc(smc_op, 0);
+	if (action < 0)
+		return action;
+
+	return sprintf(buf, "%s\n", mlxbf_bootctl_action_to_string(action));
+}
+
+static int mlxbf_bootctl_store(int smc_op, const char *buf, size_t count)
+{
+	int ret, action;
+
+	action = mlxbf_bootctl_reset_action_to_val(buf);
+	if (action < 0)
+		return action;
+
+	ret = mlxbf_bootctl_smc(smc_op, action);
+	if (ret < 0)
+		return ret;
+
+	return count;
+}
+
+static ssize_t reset_action_show(struct device *dev,
+				 struct device_attribute *attr, char *buf)
+{
+	return mlxbf_bootctl_show(MLXBF_BOOTCTL_GET_RESET_ACTION, buf);
+}
+
+static ssize_t reset_action_store(struct device *dev,
+				  struct device_attribute *attr,
+				  const char *buf, size_t count)
+{
+	return mlxbf_bootctl_store(MLXBF_BOOTCTL_SET_RESET_ACTION, buf, count);
+}
+
+static ssize_t second_reset_action_show(struct device *dev,
+					struct device_attribute *attr,
+					char *buf)
+{
+	return mlxbf_bootctl_show(MLXBF_BOOTCTL_GET_SECOND_RESET_ACTION, buf);
+}
+
+static ssize_t second_reset_action_store(struct device *dev,
+					 struct device_attribute *attr,
+					 const char *buf, size_t count)
+{
+	return mlxbf_bootctl_store(MLXBF_BOOTCTL_SET_SECOND_RESET_ACTION, buf,
+				   count);
+}
+
+static ssize_t lifecycle_state_show(struct device *dev,
+				    struct device_attribute *attr, char *buf)
+{
+	int lc_state;
+
+	lc_state = mlxbf_bootctl_smc(MLXBF_BOOTCTL_GET_TBB_FUSE_STATUS,
+				     MLXBF_BOOTCTL_FUSE_STATUS_LIFECYCLE);
+	if (lc_state < 0)
+		return lc_state;
+
+	lc_state &=
+		MLXBF_BOOTCTL_SB_TEST_MASK | MLXBF_BOOTCTL_SB_SECURE_MASK;
+
+	/*
+	 * If the test bits are set, we specify that the current state may be
+	 * due to using the test bits.
+	 */
+	if (lc_state & MLXBF_BOOTCTL_SB_TEST_MASK) {
+		lc_state &= MLXBF_BOOTCTL_SB_SECURE_MASK;
+
+		return sprintf(buf, "%s(test)\n",
+			       mlxbf_bootctl_lifecycle_states[lc_state]);
+	}
+
+	return sprintf(buf, "%s\n", mlxbf_bootctl_lifecycle_states[lc_state]);
+}
+
+static ssize_t secure_boot_fuse_state_show(struct device *dev,
+					   struct device_attribute *attr,
+					   char *buf)
+{
+	int burnt, valid, key, key_state, buf_len = 0, upper_key_used = 0;
+	const char *status;
+
+	key_state = mlxbf_bootctl_smc(MLXBF_BOOTCTL_GET_TBB_FUSE_STATUS,
+				      MLXBF_BOOTCTL_FUSE_STATUS_KEYS);
+	if (key_state < 0)
+		return key_state;
+
+	/*
+	 * key_state contains the bits for 4 Key versions, loaded from eFuses
+	 * after a hard reset. Lower 4 bits are a thermometer code indicating
+	 * key programming has started for key n (0000 = none, 0001 = version 0,
+	 * 0011 = version 1, 0111 = version 2, 1111 = version 3). Upper 4 bits
+	 * are a thermometer code indicating key programming has completed for
+	 * key n (same encodings as the start bits). This allows for detection
+	 * of an interruption in the progamming process which has left the key
+	 * partially programmed (and thus invalid). The process is to burn the
+	 * eFuse for the new key start bit, burn the key eFuses, then burn the
+	 * eFuse for the new key complete bit.
+	 *
+	 * For example 0000_0000: no key valid, 0001_0001: key version 0 valid,
+	 * 0011_0011: key 1 version valid, 0011_0111: key version 2 started
+	 * programming but did not complete, etc. The most recent key for which
+	 * both start and complete bit is set is loaded. On soft reset, this
+	 * register is not modified.
+	 */
+	for (key = MLXBF_SB_KEY_NUM - 1; key >= 0; key--) {
+		burnt = key_state & BIT(key);
+		valid = key_state & BIT(key + MLXBF_SB_KEY_NUM);
+
+		if (burnt && valid)
+			upper_key_used = 1;
+
+		if (upper_key_used) {
+			if (burnt)
+				status = valid ? "Used" : "Wasted";
+			else
+				status = valid ? "Invalid" : "Skipped";
+		} else {
+			if (burnt)
+				status = valid ? "InUse" : "Incomplete";
+			else
+				status = valid ? "Invalid" : "Free";
+		}
+		buf_len += sprintf(buf + buf_len, "%d:%s ", key, status);
+	}
+	buf_len += sprintf(buf + buf_len, "\n");
+
+	return buf_len;
+}
+
+static DEVICE_ATTR_RW(post_reset_wdog);
+static DEVICE_ATTR_RW(reset_action);
+static DEVICE_ATTR_RW(second_reset_action);
+static DEVICE_ATTR_RO(lifecycle_state);
+static DEVICE_ATTR_RO(secure_boot_fuse_state);
+
+static struct attribute *mlxbf_bootctl_attrs[] = {
+	&dev_attr_post_reset_wdog.attr,
+	&dev_attr_reset_action.attr,
+	&dev_attr_second_reset_action.attr,
+	&dev_attr_lifecycle_state.attr,
+	&dev_attr_secure_boot_fuse_state.attr,
+	NULL
+};
+
+ATTRIBUTE_GROUPS(mlxbf_bootctl);
+
+static const struct acpi_device_id mlxbf_bootctl_acpi_ids[] = {
+	{"MLNXBF04", 0},
+	{}
+};
+
+MODULE_DEVICE_TABLE(acpi, mlxbf_bootctl_acpi_ids);
+
+static bool mlxbf_bootctl_guid_match(const guid_t *guid,
+				     const struct arm_smccc_res *res)
+{
+	guid_t id = GUID_INIT(res->a0, res->a1, res->a1 >> 16,
+			      res->a2, res->a2 >> 8, res->a2 >> 16,
+			      res->a2 >> 24, res->a3, res->a3 >> 8,
+			      res->a3 >> 16, res->a3 >> 24);
+
+	return guid_equal(guid, &id);
+}
+
+static int mlxbf_bootctl_probe(struct platform_device *pdev)
+{
+	struct arm_smccc_res res = { 0 };
+	guid_t guid;
+	int ret;
+
+	/* Ensure we have the UUID we expect for this service. */
+	arm_smccc_smc(MLXBF_BOOTCTL_SIP_SVC_UID, 0, 0, 0, 0, 0, 0, 0, &res);
+	guid_parse(mlxbf_bootctl_svc_uuid_str, &guid);
+	if (!mlxbf_bootctl_guid_match(&guid, &res))
+		return -ENODEV;
+
+	/*
+	 * When watchdog is used, it sets boot mode to MLXBF_BOOTCTL_SWAP_EMMC
+	 * in case of boot failures. However it doesn't clear the state if there
+	 * is no failure. Restore the default boot mode here to avoid any
+	 * unnecessary boot partition swapping.
+	 */
+	ret = mlxbf_bootctl_smc(MLXBF_BOOTCTL_SET_RESET_ACTION,
+				MLXBF_BOOTCTL_EMMC);
+	if (ret < 0)
+		dev_warn(&pdev->dev, "Unable to reset the EMMC boot mode\n");
+
+	return 0;
+}
+
+static struct platform_driver mlxbf_bootctl_driver = {
+	.probe = mlxbf_bootctl_probe,
+	.driver = {
+		.name = "mlxbf-bootctl",
+		.groups = mlxbf_bootctl_groups,
+		.acpi_match_table = mlxbf_bootctl_acpi_ids,
+	}
+};
+
+module_platform_driver(mlxbf_bootctl_driver);
+
+MODULE_DESCRIPTION("Mellanox boot control driver");
+MODULE_LICENSE("GPL v2");
+MODULE_AUTHOR("Mellanox Technologies");

drivers/platform/mellanox/mlxbf-bootctl.h

@@ -0,0 +1,103 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+/*
+ * Copyright (c) 2019, Mellanox Technologies. All rights reserved.
+ */
+
+#ifndef __MLXBF_BOOTCTL_H__
+#define __MLXBF_BOOTCTL_H__
+
+/*
+ * Request that the on-chip watchdog be enabled, or disabled, after
+ * the next chip soft reset. This call does not affect the current
+ * status of the on-chip watchdog. If non-zero, the argument
+ * specifies the watchdog interval in seconds. If zero, the watchdog
+ * will not be enabled after the next soft reset. Non-zero errors are
+ * returned as documented below.
+ */
+#define MLXBF_BOOTCTL_SET_POST_RESET_WDOG	0x82000000
+
+/*
+ * Query the status which has been requested for the on-chip watchdog
+ * after the next chip soft reset. Returns the interval as set by
+ * MLXBF_BOOTCTL_SET_POST_RESET_WDOG.
+ */
+#define MLXBF_BOOTCTL_GET_POST_RESET_WDOG	0x82000001
+
+/*
+ * Request that a specific boot action be taken at the next soft
+ * reset. By default, the boot action is set by external chip pins,
+ * which are sampled on hard reset. Note that the boot action
+ * requested by this call will persist on subsequent resets unless
+ * this service, or the MLNX_SET_SECOND_RESET_ACTION service, is
+ * invoked. See below for the available MLNX_BOOT_xxx parameter
+ * values. Non-zero errors are returned as documented below.
+ */
+#define MLXBF_BOOTCTL_SET_RESET_ACTION		0x82000002
+
+/*
+ * Return the specific boot action which will be taken at the next
+ * soft reset. Returns the reset action (see below for the parameter
+ * values for MLXBF_BOOTCTL_SET_RESET_ACTION).
+ */
+#define MLXBF_BOOTCTL_GET_RESET_ACTION		0x82000003
+
+/*
+ * Request that a specific boot action be taken at the soft reset
+ * after the next soft reset. For a specified valid boot mode, the
+ * effect of this call is identical to that of invoking
+ * MLXBF_BOOTCTL_SET_RESET_ACTION after the next chip soft reset; in
+ * particular, after that reset, the action for the now next reset can
+ * be queried with MLXBF_BOOTCTL_GET_RESET_ACTION and modified with
+ * MLXBF_BOOTCTL_SET_RESET_ACTION. You may also specify the parameter as
+ * MLNX_BOOT_NONE, which is equivalent to specifying that no call to
+ * MLXBF_BOOTCTL_SET_RESET_ACTION be taken after the next chip soft reset.
+ * This call does not affect the action to be taken at the next soft
+ * reset. Non-zero errors are returned as documented below.
+ */
+#define MLXBF_BOOTCTL_SET_SECOND_RESET_ACTION	0x82000004
+
+/*
+ * Return the specific boot action which will be taken at the soft
+ * reset after the next soft reset; this will be one of the valid
+ * actions for MLXBF_BOOTCTL_SET_SECOND_RESET_ACTION.
+ */
+#define MLXBF_BOOTCTL_GET_SECOND_RESET_ACTION	0x82000005
+
+/*
+ * Return the fuse status of the current chip. The caller should specify
+ * with the second argument if the state of the lifecycle fuses or the
+ * version of secure boot fuse keys left should be returned.
+ */
+#define MLXBF_BOOTCTL_GET_TBB_FUSE_STATUS	0x82000006
+
+/* Reset eMMC by programming the RST_N register. */
+#define MLXBF_BOOTCTL_SET_EMMC_RST_N		0x82000007
+
+#define MLXBF_BOOTCTL_GET_DIMM_INFO		0x82000008
+
+/* SMC function IDs for SiP Service queries */
+#define MLXBF_BOOTCTL_SIP_SVC_CALL_COUNT	0x8200ff00
+#define MLXBF_BOOTCTL_SIP_SVC_UID		0x8200ff01
+#define MLXBF_BOOTCTL_SIP_SVC_VERSION		0x8200ff03
+
+/* ARM Standard Service Calls version numbers */
+#define MLXBF_BOOTCTL_SVC_VERSION_MAJOR		0x0
+#define MLXBF_BOOTCTL_SVC_VERSION_MINOR		0x2
+
+/* Number of svc calls defined. */
+#define MLXBF_BOOTCTL_NUM_SVC_CALLS 12
+
+/* Valid reset actions for MLXBF_BOOTCTL_SET_RESET_ACTION. */
+#define MLXBF_BOOTCTL_EXTERNAL	0 /* Not boot from eMMC */
+#define MLXBF_BOOTCTL_EMMC	1 /* From primary eMMC boot partition */
+#define MLNX_BOOTCTL_SWAP_EMMC	2 /* Swap eMMC boot partitions and reboot */
+#define MLXBF_BOOTCTL_EMMC_LEGACY	3 /* From primary eMMC in legacy mode */
+
+/* Valid arguments for requesting the fuse status. */
+#define MLXBF_BOOTCTL_FUSE_STATUS_LIFECYCLE	0 /* Return lifecycle status. */
+#define MLXBF_BOOTCTL_FUSE_STATUS_KEYS	1 /* Return secure boot key status */
+
+/* Additional value to disable the MLXBF_BOOTCTL_SET_SECOND_RESET_ACTION. */
+#define MLXBF_BOOTCTL_NONE	0x7fffffff /* Don't change next boot action */
+
+#endif /* __MLXBF_BOOTCTL_H__ */