powerpc/powernv: Add opal-prd channel

This change adds a char device to access the "PRD" (processor runtime
diagnostics) channel to OPAL firmware.

Includes contributions from Vaidyanathan Srinivasan, Neelesh Gupta &
Vishal Kulkarni.

Signed-off-by: Neelesh Gupta <neelegup@linux.vnet.ibm.com>
Signed-off-by: Jeremy Kerr <jk@ozlabs.org>
Acked-by: Stewart Smith <stewart@linux.vnet.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>

arch/powerpc/include/asm/opal-api.h

@@ -153,7 +153,8 @@
 #define OPAL_FLASH_READ				110
 #define OPAL_FLASH_WRITE			111
 #define OPAL_FLASH_ERASE			112
-#define OPAL_LAST				112
+#define OPAL_PRD_MSG				113
+#define OPAL_LAST				113
 
 /* Device tree flags */
 
@@ -359,6 +360,7 @@ enum opal_msg_type {
 	OPAL_MSG_SHUTDOWN,		/* params[0] = 1 reboot, 0 shutdown */
 	OPAL_MSG_HMI_EVT,
 	OPAL_MSG_DPO,
+	OPAL_MSG_PRD,
 	OPAL_MSG_TYPE_MAX,
 };
 
@@ -681,6 +683,23 @@ typedef struct oppanel_line {
 	__be64 line_len;
 } oppanel_line_t;
 
+enum opal_prd_msg_type {
+	OPAL_PRD_MSG_TYPE_INIT = 0,	/* HBRT --> OPAL */
+	OPAL_PRD_MSG_TYPE_FINI,		/* HBRT/kernel --> OPAL */
+	OPAL_PRD_MSG_TYPE_ATTN,		/* HBRT <-- OPAL */
+	OPAL_PRD_MSG_TYPE_ATTN_ACK,	/* HBRT --> OPAL */
+	OPAL_PRD_MSG_TYPE_OCC_ERROR,	/* HBRT <-- OPAL */
+	OPAL_PRD_MSG_TYPE_OCC_RESET,	/* HBRT <-- OPAL */
+};
+
+struct opal_prd_msg_header {
+	uint8_t		type;
+	uint8_t		pad[1];
+	__be16		size;
+};
+
+struct opal_prd_msg;
+
 /*
  * SG entries
  *

arch/powerpc/include/asm/opal.h

@@ -194,6 +194,7 @@ int64_t opal_ipmi_recv(uint64_t interface, struct opal_ipmi_msg *msg,
 		uint64_t *msg_len);
 int64_t opal_i2c_request(uint64_t async_token, uint32_t bus_id,
 			 struct opal_i2c_request *oreq);
+int64_t opal_prd_msg(struct opal_prd_msg *msg);
 
 int64_t opal_flash_read(uint64_t id, uint64_t offset, uint64_t buf,
 		uint64_t size, uint64_t token);

arch/powerpc/include/uapi/asm/opal-prd.h

@@ -0,0 +1,58 @@
+/*
+ * OPAL Runtime Diagnostics interface driver
+ * Supported on POWERNV platform
+ *
+ * (C) Copyright IBM 2015
+ *
+ * Author: Vaidyanathan Srinivasan <svaidy at linux.vnet.ibm.com>
+ * Author: Jeremy Kerr <jk@ozlabs.org>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2, or (at your option)
+ * any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ */
+
+#ifndef _UAPI_ASM_POWERPC_OPAL_PRD_H_
+#define _UAPI_ASM_POWERPC_OPAL_PRD_H_
+
+#include <linux/types.h>
+
+/**
+ * The version of the kernel interface of the PRD system. This describes the
+ * interface available for the /dev/opal-prd device. The actual PRD message
+ * layout and content is private to the firmware <--> userspace interface, so
+ * is not covered by this versioning.
+ *
+ * Future interface versions are backwards-compatible; if a later kernel
+ * version is encountered, functionality provided in earlier versions
+ * will work.
+ */
+#define OPAL_PRD_KERNEL_VERSION		1
+
+#define OPAL_PRD_GET_INFO		_IOR('o', 0x01, struct opal_prd_info)
+#define OPAL_PRD_SCOM_READ		_IOR('o', 0x02, struct opal_prd_scom)
+#define OPAL_PRD_SCOM_WRITE		_IOW('o', 0x03, struct opal_prd_scom)
+
+#ifndef __ASSEMBLY__
+
+struct opal_prd_info {
+	__u64	version;
+	__u64	reserved[3];
+};
+
+struct opal_prd_scom {
+	__u64	chip;
+	__u64	addr;
+	__u64	data;
+	__s64	rc;
+};
+
+#endif /* __ASSEMBLY__ */
+
+#endif /* _UAPI_ASM_POWERPC_OPAL_PRD_H */

arch/powerpc/platforms/powernv/Kconfig

@@ -19,3 +19,10 @@ config PPC_POWERNV
 	select CPU_FREQ_GOV_CONSERVATIVE
 	select PPC_DOORBELL
 	default y
+
+config OPAL_PRD
+	tristate 'OPAL PRD driver'
+	depends on PPC_POWERNV
+	help
+	  This enables the opal-prd driver, a facility to run processor
+	  recovery diagnostics on OpenPower machines

arch/powerpc/platforms/powernv/Makefile

@@ -9,3 +9,4 @@ obj-$(CONFIG_EEH)	+= eeh-powernv.o
 obj-$(CONFIG_PPC_SCOM)	+= opal-xscom.o
 obj-$(CONFIG_MEMORY_FAILURE)	+= opal-memory-errors.o
 obj-$(CONFIG_TRACEPOINTS)	+= opal-tracepoints.o
+obj-$(CONFIG_OPAL_PRD)	+= opal-prd.o

arch/powerpc/platforms/powernv/opal-prd.c

@@ -0,0 +1,444 @@
+/*
+ * OPAL Runtime Diagnostics interface driver
+ * Supported on POWERNV platform
+ *
+ * Copyright IBM Corporation 2015
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ */
+
+#define pr_fmt(fmt) "opal-prd: " fmt
+
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/platform_device.h>
+#include <linux/miscdevice.h>
+#include <linux/fs.h>
+#include <linux/of.h>
+#include <linux/of_address.h>
+#include <linux/poll.h>
+#include <linux/mm.h>
+#include <linux/slab.h>
+#include <asm/opal-prd.h>
+#include <asm/opal.h>
+#include <asm/io.h>
+#include <asm/uaccess.h>
+
+
+struct opal_prd_msg_queue_item {
+	struct opal_prd_msg_header	msg;
+	struct list_head		list;
+};
+
+static struct device_node *prd_node;
+static LIST_HEAD(opal_prd_msg_queue);
+static DEFINE_SPINLOCK(opal_prd_msg_queue_lock);
+static DECLARE_WAIT_QUEUE_HEAD(opal_prd_msg_wait);
+static atomic_t prd_usage;
+
+static bool opal_prd_range_is_valid(uint64_t addr, uint64_t size)
+{
+	struct device_node *parent, *node;
+	bool found;
+
+	if (addr + size < addr)
+		return false;
+
+	parent = of_find_node_by_path("/reserved-memory");
+	if (!parent)
+		return false;
+
+	found = false;
+
+	for_each_child_of_node(parent, node) {
+		uint64_t range_addr, range_size, range_end;
+		const __be32 *addrp;
+		const char *label;
+
+		addrp = of_get_address(node, 0, &range_size, NULL);
+
+		range_addr = of_read_number(addrp, 2);
+		range_end = range_addr + range_size;
+
+		label = of_get_property(node, "ibm,prd-label", NULL);
+
+		/* PRD ranges need a label */
+		if (!label)
+			continue;
+
+		if (range_end <= range_addr)
+			continue;
+
+		if (addr >= range_addr && addr + size <= range_end) {
+			found = true;
+			of_node_put(node);
+			break;
+		}
+	}
+
+	of_node_put(parent);
+	return found;
+}
+
+static int opal_prd_open(struct inode *inode, struct file *file)
+{
+	/*
+	 * Prevent multiple (separate) processes from concurrent interactions
+	 * with the FW PRD channel
+	 */
+	if (atomic_xchg(&prd_usage, 1) == 1)
+		return -EBUSY;
+
+	return 0;
+}
+
+/*
+ * opal_prd_mmap - maps firmware-provided ranges into userspace
+ * @file: file structure for the device
+ * @vma: VMA to map the registers into
+ */
+
+static int opal_prd_mmap(struct file *file, struct vm_area_struct *vma)
+{
+	size_t addr, size;
+	int rc;
+
+	pr_devel("opal_prd_mmap(0x%016lx, 0x%016lx, 0x%lx, 0x%lx)\n",
+			vma->vm_start, vma->vm_end, vma->vm_pgoff,
+			vma->vm_flags);
+
+	addr = vma->vm_pgoff << PAGE_SHIFT;
+	size = vma->vm_end - vma->vm_start;
+
+	/* ensure we're mapping within one of the allowable ranges */
+	if (!opal_prd_range_is_valid(addr, size))
+		return -EINVAL;
+
+	vma->vm_page_prot = __pgprot(pgprot_val(phys_mem_access_prot(file,
+						vma->vm_pgoff,
+						 size, vma->vm_page_prot))
+					| _PAGE_SPECIAL);
+
+	rc = remap_pfn_range(vma, vma->vm_start, vma->vm_pgoff, size,
+			vma->vm_page_prot);
+
+	return rc;
+}
+
+static bool opal_msg_queue_empty(void)
+{
+	unsigned long flags;
+	bool ret;
+
+	spin_lock_irqsave(&opal_prd_msg_queue_lock, flags);
+	ret = list_empty(&opal_prd_msg_queue);
+	spin_unlock_irqrestore(&opal_prd_msg_queue_lock, flags);
+
+	return ret;
+}
+
+static unsigned int opal_prd_poll(struct file *file,
+		struct poll_table_struct *wait)
+{
+	poll_wait(file, &opal_prd_msg_wait, wait);
+
+	if (!opal_msg_queue_empty())
+		return POLLIN | POLLRDNORM;
+
+	return 0;
+}
+
+static ssize_t opal_prd_read(struct file *file, char __user *buf,
+		size_t count, loff_t *ppos)
+{
+	struct opal_prd_msg_queue_item *item;
+	unsigned long flags;
+	ssize_t size, err;
+	int rc;
+
+	/* we need at least a header's worth of data */
+	if (count < sizeof(item->msg))
+		return -EINVAL;
+
+	if (*ppos)
+		return -ESPIPE;
+
+	item = NULL;
+
+	for (;;) {
+
+		spin_lock_irqsave(&opal_prd_msg_queue_lock, flags);
+		if (!list_empty(&opal_prd_msg_queue)) {
+			item = list_first_entry(&opal_prd_msg_queue,
+					struct opal_prd_msg_queue_item, list);
+			list_del(&item->list);
+		}
+		spin_unlock_irqrestore(&opal_prd_msg_queue_lock, flags);
+
+		if (item)
+			break;
+
+		if (file->f_flags & O_NONBLOCK)
+			return -EAGAIN;
+
+		rc = wait_event_interruptible(opal_prd_msg_wait,
+				!opal_msg_queue_empty());
+		if (rc)
+			return -EINTR;
+	}
+
+	size = be16_to_cpu(item->msg.size);
+	if (size > count) {
+		err = -EINVAL;
+		goto err_requeue;
+	}
+
+	rc = copy_to_user(buf, &item->msg, size);
+	if (rc) {
+		err = -EFAULT;
+		goto err_requeue;
+	}
+
+	kfree(item);
+
+	return size;
+
+err_requeue:
+	/* eep! re-queue at the head of the list */
+	spin_lock_irqsave(&opal_prd_msg_queue_lock, flags);
+	list_add(&item->list, &opal_prd_msg_queue);
+	spin_unlock_irqrestore(&opal_prd_msg_queue_lock, flags);
+	return err;
+}
+
+static ssize_t opal_prd_write(struct file *file, const char __user *buf,
+		size_t count, loff_t *ppos)
+{
+	struct opal_prd_msg_header hdr;
+	ssize_t size;
+	void *msg;
+	int rc;
+
+	size = sizeof(hdr);
+
+	if (count < size)
+		return -EINVAL;
+
+	/* grab the header */
+	rc = copy_from_user(&hdr, buf, sizeof(hdr));
+	if (rc)
+		return -EFAULT;
+
+	size = be16_to_cpu(hdr.size);
+
+	msg = kmalloc(size, GFP_KERNEL);
+	if (!msg)
+		return -ENOMEM;
+
+	rc = copy_from_user(msg, buf, size);
+	if (rc) {
+		size = -EFAULT;
+		goto out_free;
+	}
+
+	rc = opal_prd_msg(msg);
+	if (rc) {
+		pr_warn("write: opal_prd_msg returned %d\n", rc);
+		size = -EIO;
+	}
+
+out_free:
+	kfree(msg);
+
+	return size;
+}
+
+static int opal_prd_release(struct inode *inode, struct file *file)
+{
+	struct opal_prd_msg_header msg;
+
+	msg.size = cpu_to_be16(sizeof(msg));
+	msg.type = OPAL_PRD_MSG_TYPE_FINI;
+
+	opal_prd_msg((struct opal_prd_msg *)&msg);
+
+	atomic_xchg(&prd_usage, 0);
+
+	return 0;
+}
+
+static long opal_prd_ioctl(struct file *file, unsigned int cmd,
+		unsigned long param)
+{
+	struct opal_prd_info info;
+	struct opal_prd_scom scom;
+	int rc = 0;
+
+	switch (cmd) {
+	case OPAL_PRD_GET_INFO:
+		memset(&info, 0, sizeof(info));
+		info.version = OPAL_PRD_KERNEL_VERSION;
+		rc = copy_to_user((void __user *)param, &info, sizeof(info));
+		if (rc)
+			return -EFAULT;
+		break;
+
+	case OPAL_PRD_SCOM_READ:
+		rc = copy_from_user(&scom, (void __user *)param, sizeof(scom));
+		if (rc)
+			return -EFAULT;
+
+		scom.rc = opal_xscom_read(scom.chip, scom.addr,
+				(__be64 *)&scom.data);
+		scom.data = be64_to_cpu(scom.data);
+		pr_devel("ioctl SCOM_READ: chip %llx addr %016llx data %016llx rc %lld\n",
+				scom.chip, scom.addr, scom.data, scom.rc);
+
+		rc = copy_to_user((void __user *)param, &scom, sizeof(scom));
+		if (rc)
+			return -EFAULT;
+		break;
+
+	case OPAL_PRD_SCOM_WRITE:
+		rc = copy_from_user(&scom, (void __user *)param, sizeof(scom));
+		if (rc)
+			return -EFAULT;
+
+		scom.rc = opal_xscom_write(scom.chip, scom.addr, scom.data);
+		pr_devel("ioctl SCOM_WRITE: chip %llx addr %016llx data %016llx rc %lld\n",
+				scom.chip, scom.addr, scom.data, scom.rc);
+
+		rc = copy_to_user((void __user *)param, &scom, sizeof(scom));
+		if (rc)
+			return -EFAULT;
+		break;
+
+	default:
+		rc = -EINVAL;
+	}
+
+	return rc;
+}
+
+static const struct file_operations opal_prd_fops = {
+	.open		= opal_prd_open,
+	.mmap		= opal_prd_mmap,
+	.poll		= opal_prd_poll,
+	.read		= opal_prd_read,
+	.write		= opal_prd_write,
+	.unlocked_ioctl	= opal_prd_ioctl,
+	.release	= opal_prd_release,
+	.owner		= THIS_MODULE,
+};
+
+static struct miscdevice opal_prd_dev = {
+	.minor		= MISC_DYNAMIC_MINOR,
+	.name		= "opal-prd",
+	.fops		= &opal_prd_fops,
+};
+
+/* opal interface */
+static int opal_prd_msg_notifier(struct notifier_block *nb,
+		unsigned long msg_type, void *_msg)
+{
+	struct opal_prd_msg_queue_item *item;
+	struct opal_prd_msg_header *hdr;
+	struct opal_msg *msg = _msg;
+	unsigned long flags;
+	int size;
+
+	if (msg_type != OPAL_MSG_PRD)
+		return 0;
+
+	/* Calculate total size of the item we need to store. The 'size' field
+	 * in the header includes the header itself. */
+	hdr = (void *)msg->params;
+	size = (sizeof(*item) - sizeof(item->msg)) + be16_to_cpu(hdr->size);
+
+	item = kzalloc(size, GFP_ATOMIC);
+	if (!item)
+		return -ENOMEM;
+
+	memcpy(&item->msg, msg->params, size);
+
+	spin_lock_irqsave(&opal_prd_msg_queue_lock, flags);
+	list_add_tail(&item->list, &opal_prd_msg_queue);
+	spin_unlock_irqrestore(&opal_prd_msg_queue_lock, flags);
+
+	wake_up_interruptible(&opal_prd_msg_wait);
+
+	return 0;
+}
+
+static struct notifier_block opal_prd_event_nb = {
+	.notifier_call	= opal_prd_msg_notifier,
+	.next		= NULL,
+	.priority	= 0,
+};
+
+static int opal_prd_probe(struct platform_device *pdev)
+{
+	int rc;
+
+	if (!pdev || !pdev->dev.of_node)
+		return -ENODEV;
+
+	/* We should only have one prd driver instance per machine; ensure
+	 * that we only get a valid probe on a single OF node.
+	 */
+	if (prd_node)
+		return -EBUSY;
+
+	prd_node = pdev->dev.of_node;
+
+	rc = opal_message_notifier_register(OPAL_MSG_PRD, &opal_prd_event_nb);
+	if (rc) {
+		pr_err("Couldn't register event notifier\n");
+		return rc;
+	}
+
+	rc = misc_register(&opal_prd_dev);
+	if (rc) {
+		pr_err("failed to register miscdev\n");
+		opal_message_notifier_unregister(OPAL_MSG_PRD,
+				&opal_prd_event_nb);
+		return rc;
+	}
+
+	return 0;
+}
+
+static int opal_prd_remove(struct platform_device *pdev)
+{
+	misc_deregister(&opal_prd_dev);
+	opal_message_notifier_unregister(OPAL_MSG_PRD, &opal_prd_event_nb);
+	return 0;
+}
+
+static const struct of_device_id opal_prd_match[] = {
+	{ .compatible = "ibm,opal-prd" },
+	{ },
+};
+
+static struct platform_driver opal_prd_driver = {
+	.driver = {
+		.name		= "opal-prd",
+		.owner		= THIS_MODULE,
+		.of_match_table	= opal_prd_match,
+	},
+	.probe	= opal_prd_probe,
+	.remove	= opal_prd_remove,
+};
+
+module_platform_driver(opal_prd_driver);
+
+MODULE_DEVICE_TABLE(of, opal_prd_match);
+MODULE_DESCRIPTION("PowerNV OPAL runtime diagnostic driver");
+MODULE_LICENSE("GPL");

arch/powerpc/platforms/powernv/opal-wrappers.S

@@ -296,3 +296,4 @@ OPAL_CALL(opal_i2c_request,			OPAL_I2C_REQUEST);
 OPAL_CALL(opal_flash_read,			OPAL_FLASH_READ);
 OPAL_CALL(opal_flash_write,			OPAL_FLASH_WRITE);
 OPAL_CALL(opal_flash_erase,			OPAL_FLASH_ERASE);
+OPAL_CALL(opal_prd_msg,				OPAL_PRD_MSG);

arch/powerpc/platforms/powernv/opal.c

@@ -711,9 +711,10 @@ static int __init opal_init(void)
 		opal_msglog_init();
 	}
 
-	/* Initialize platform devices: IPMI backend & flash interface */
+	/* Initialize platform devices: IPMI backend, PRD & flash interface */
 	opal_pdev_init(opal_node, "ibm,opal-ipmi");
 	opal_pdev_init(opal_node, "ibm,opal-flash");
+	opal_pdev_init(opal_node, "ibm,opal-prd");
 
 	return 0;
 }
@@ -752,6 +753,7 @@ EXPORT_SYMBOL_GPL(opal_ipmi_recv);
 EXPORT_SYMBOL_GPL(opal_flash_read);
 EXPORT_SYMBOL_GPL(opal_flash_write);
 EXPORT_SYMBOL_GPL(opal_flash_erase);
+EXPORT_SYMBOL_GPL(opal_prd_msg);
 
 /* Convert a region of vmalloc memory to an opal sg list */
 struct opal_sg_list *opal_vmalloc_to_sg_list(void *vmalloc_addr,