WATCHDOG: Add watchdog driver for OCTEON SOCs

The OCTEON is a MIPS64 based SOC family with an on chip watchdog unit.

The driver is split into two source files one for the C code and one
for assembly.  Assembly is needed to handle the NMI and then print the
machine state before the reboot is triggered.

Signed-off-by: David Daney <ddaney@caviumnetworks.com>
Cc: Wim Van Sebroeck <wim@iguana.be>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Russell King <rmk+kernel@arm.linux.org.uk>
Cc: Tony Lindgren <tony@atomide.com>
Cc: Marc Zyngier <maz@misterjones.org>
Cc: Thierry Reding <thierry.reding@avionic-design.de>
Cc: Sam Ravnborg <sam@ravnborg.org>
To: linux-mips@linux-mips.org
Cc: linux-kernel@vger.kernel.org,
Patchwork: https://patchwork.linux-mips.org/patch/1503/
Signed-off-by: Wim Van Sebroeck <wim@iguana.be>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>

 create mode 100644 drivers/watchdog/octeon-wdt-main.c
 create mode 100644 drivers/watchdog/octeon-wdt-nmi.S
diff --git a/drivers/watchdog/Kconfig b/drivers/watchdog/Kconfig
index afcfacc..b04b184 100644
--- a/drivers/watchdog/Kconfig
+++ b/drivers/watchdog/Kconfig
@@ -875,6 +875,24 @@
 	help
 	  Hardware driver for the built-in watchdog timer on TXx9 MIPS SoCs.
 
+config OCTEON_WDT
+	tristate "Cavium OCTEON SOC family Watchdog Timer"
+	depends on CPU_CAVIUM_OCTEON
+	default y
+	select EXPORT_UASM if OCTEON_WDT = m
+	help
+	  Hardware driver for OCTEON's on chip watchdog timer.
+	  Enables the watchdog for all cores running Linux. It
+	  installs a NMI handler and pokes the watchdog based on an
+	  interrupt.  On first expiration of the watchdog, the
+	  interrupt handler pokes it.  The second expiration causes an
+	  NMI that prints a message. The third expiration causes a
+	  global soft reset.
+
+	  When userspace has /dev/watchdog open, no poking is done
+	  from the first interrupt, it is then only poked when the
+	  device is written.
+
 # PARISC Architecture
 
 # POWERPC Architecture
diff --git a/drivers/watchdog/Makefile b/drivers/watchdog/Makefile
index 72f3e20..e30289a 100644
--- a/drivers/watchdog/Makefile
+++ b/drivers/watchdog/Makefile
@@ -114,6 +114,8 @@
 obj-$(CONFIG_SIBYTE_WDOG) += sb_wdog.o
 obj-$(CONFIG_AR7_WDT) += ar7_wdt.o
 obj-$(CONFIG_TXX9_WDT) += txx9wdt.o
+obj-$(CONFIG_OCTEON_WDT) += octeon-wdt.o
+octeon-wdt-y := octeon-wdt-main.o octeon-wdt-nmi.o
 
 # PARISC Architecture
 
diff --git a/drivers/watchdog/octeon-wdt-main.c b/drivers/watchdog/octeon-wdt-main.c
new file mode 100644
index 0000000..2a41017
--- /dev/null
+++ b/drivers/watchdog/octeon-wdt-main.c
@@ -0,0 +1,745 @@
+/*
+ * Octeon Watchdog driver
+ *
+ * Copyright (C) 2007, 2008, 2009, 2010 Cavium Networks
+ *
+ * Some parts derived from wdt.c
+ *
+ *	(c) Copyright 1996-1997 Alan Cox <alan@lxorguk.ukuu.org.uk>,
+ *						All Rights Reserved.
+ *
+ *	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 of the License, or (at your option) any later version.
+ *
+ *	Neither Alan Cox nor CymruNet Ltd. admit liability nor provide
+ *	warranty for any of this software. This material is provided
+ *	"AS-IS" and at no charge.
+ *
+ *	(c) Copyright 1995    Alan Cox <alan@lxorguk.ukuu.org.uk>
+ *
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ *
+ * The OCTEON watchdog has a maximum timeout of 2^32 * io_clock.
+ * For most systems this is less than 10 seconds, so to allow for
+ * software to request longer watchdog heartbeats, we maintain software
+ * counters to count multiples of the base rate.  If the system locks
+ * up in such a manner that we can not run the software counters, the
+ * only result is a watchdog reset sooner than was requested.  But
+ * that is OK, because in this case userspace would likely not be able
+ * to do anything anyhow.
+ *
+ * The hardware watchdog interval we call the period.  The OCTEON
+ * watchdog goes through several stages, after the first period an
+ * irq is asserted, then if it is not reset, after the next period NMI
+ * is asserted, then after an additional period a chip wide soft reset.
+ * So for the software counters, we reset watchdog after each period
+ * and decrement the counter.  But for the last two periods we need to
+ * let the watchdog progress to the NMI stage so we disable the irq
+ * and let it proceed.  Once in the NMI, we print the register state
+ * to the serial port and then wait for the reset.
+ *
+ * A watchdog is maintained for each CPU in the system, that way if
+ * one CPU suffers a lockup, we also get a register dump and reset.
+ * The userspace ping resets the watchdog on all CPUs.
+ *
+ * Before userspace opens the watchdog device, we still run the
+ * watchdogs to catch any lockups that may be kernel related.
+ *
+ */
+
+#include <linux/miscdevice.h>
+#include <linux/interrupt.h>
+#include <linux/watchdog.h>
+#include <linux/cpumask.h>
+#include <linux/bitops.h>
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/string.h>
+#include <linux/delay.h>
+#include <linux/cpu.h>
+#include <linux/smp.h>
+#include <linux/fs.h>
+
+#include <asm/mipsregs.h>
+#include <asm/uasm.h>
+
+#include <asm/octeon/octeon.h>
+
+/* The count needed to achieve timeout_sec. */
+static unsigned int timeout_cnt;
+
+/* The maximum period supported. */
+static unsigned int max_timeout_sec;
+
+/* The current period.  */
+static unsigned int timeout_sec;
+
+/* Set to non-zero when userspace countdown mode active */
+static int do_coundown;
+static unsigned int countdown_reset;
+static unsigned int per_cpu_countdown[NR_CPUS];
+
+static cpumask_t irq_enabled_cpus;
+
+#define WD_TIMO 60			/* Default heartbeat = 60 seconds */
+
+static int heartbeat = WD_TIMO;
+module_param(heartbeat, int, S_IRUGO);
+MODULE_PARM_DESC(heartbeat,
+	"Watchdog heartbeat in seconds. (0 < heartbeat, default="
+				__MODULE_STRING(WD_TIMO) ")");
+
+static int nowayout = WATCHDOG_NOWAYOUT;
+module_param(nowayout, int, S_IRUGO);
+MODULE_PARM_DESC(nowayout,
+	"Watchdog cannot be stopped once started (default="
+				__MODULE_STRING(WATCHDOG_NOWAYOUT) ")");
+
+static unsigned long octeon_wdt_is_open;
+static char expect_close;
+
+static u32 __initdata nmi_stage1_insns[64];
+/* We need one branch and therefore one relocation per target label. */
+static struct uasm_label __initdata labels[5];
+static struct uasm_reloc __initdata relocs[5];
+
+enum lable_id {
+	label_enter_bootloader = 1
+};
+
+/* Some CP0 registers */
+#define K0		26
+#define C0_CVMMEMCTL 11, 7
+#define C0_STATUS 12, 0
+#define C0_EBASE 15, 1
+#define C0_DESAVE 31, 0
+
+void octeon_wdt_nmi_stage2(void);
+
+static void __init octeon_wdt_build_stage1(void)
+{
+	int i;
+	int len;
+	u32 *p = nmi_stage1_insns;
+#ifdef CONFIG_HOTPLUG_CPU
+	struct uasm_label *l = labels;
+	struct uasm_reloc *r = relocs;
+#endif
+
+	/*
+	 * For the next few instructions running the debugger may
+	 * cause corruption of k0 in the saved registers. Since we're
+	 * about to crash, nobody probably cares.
+	 *
+	 * Save K0 into the debug scratch register
+	 */
+	uasm_i_dmtc0(&p, K0, C0_DESAVE);
+
+	uasm_i_mfc0(&p, K0, C0_STATUS);
+#ifdef CONFIG_HOTPLUG_CPU
+	uasm_il_bbit0(&p, &r, K0, ilog2(ST0_NMI), label_enter_bootloader);
+#endif
+	/* Force 64-bit addressing enabled */
+	uasm_i_ori(&p, K0, K0, ST0_UX | ST0_SX | ST0_KX);
+	uasm_i_mtc0(&p, K0, C0_STATUS);
+
+#ifdef CONFIG_HOTPLUG_CPU
+	uasm_i_mfc0(&p, K0, C0_EBASE);
+	/* Coreid number in K0 */
+	uasm_i_andi(&p, K0, K0, 0xf);
+	/* 8 * coreid in bits 16-31 */
+	uasm_i_dsll_safe(&p, K0, K0, 3 + 16);
+	uasm_i_ori(&p, K0, K0, 0x8001);
+	uasm_i_dsll_safe(&p, K0, K0, 16);
+	uasm_i_ori(&p, K0, K0, 0x0700);
+	uasm_i_drotr_safe(&p, K0, K0, 32);
+	/*
+	 * Should result in: 0x8001,0700,0000,8*coreid which is
+	 * CVMX_CIU_WDOGX(coreid) - 0x0500
+	 *
+	 * Now ld K0, CVMX_CIU_WDOGX(coreid)
+	 */
+	uasm_i_ld(&p, K0, 0x500, K0);
+	/*
+	 * If bit one set handle the NMI as a watchdog event.
+	 * otherwise transfer control to bootloader.
+	 */
+	uasm_il_bbit0(&p, &r, K0, 1, label_enter_bootloader);
+	uasm_i_nop(&p);
+#endif
+
+	/* Clear Dcache so cvmseg works right. */
+	uasm_i_cache(&p, 1, 0, 0);
+
+	/* Use K0 to do a read/modify/write of CVMMEMCTL */
+	uasm_i_dmfc0(&p, K0, C0_CVMMEMCTL);
+	/* Clear out the size of CVMSEG	*/
+	uasm_i_dins(&p, K0, 0, 0, 6);
+	/* Set CVMSEG to its largest value */
+	uasm_i_ori(&p, K0, K0, 0x1c0 | 54);
+	/* Store the CVMMEMCTL value */
+	uasm_i_dmtc0(&p, K0, C0_CVMMEMCTL);
+
+	/* Load the address of the second stage handler */
+	UASM_i_LA(&p, K0, (long)octeon_wdt_nmi_stage2);
+	uasm_i_jr(&p, K0);
+	uasm_i_dmfc0(&p, K0, C0_DESAVE);
+
+#ifdef CONFIG_HOTPLUG_CPU
+	uasm_build_label(&l, p, label_enter_bootloader);
+	/* Jump to the bootloader and restore K0 */
+	UASM_i_LA(&p, K0, (long)octeon_bootloader_entry_addr);
+	uasm_i_jr(&p, K0);
+	uasm_i_dmfc0(&p, K0, C0_DESAVE);
+#endif
+	uasm_resolve_relocs(relocs, labels);
+
+	len = (int)(p - nmi_stage1_insns);
+	pr_debug("Synthesized NMI stage 1 handler (%d instructions).\n", len);
+
+	pr_debug("\t.set push\n");
+	pr_debug("\t.set noreorder\n");
+	for (i = 0; i < len; i++)
+		pr_debug("\t.word 0x%08x\n", nmi_stage1_insns[i]);
+	pr_debug("\t.set pop\n");
+
+	if (len > 32)
+		panic("NMI stage 1 handler exceeds 32 instructions, was %d\n", len);
+}
+
+static int cpu2core(int cpu)
+{
+#ifdef CONFIG_SMP
+	return cpu_logical_map(cpu);
+#else
+	return cvmx_get_core_num();
+#endif
+}
+
+static int core2cpu(int coreid)
+{
+#ifdef CONFIG_SMP
+	return cpu_number_map(coreid);
+#else
+	return 0;
+#endif
+}
+
+/**
+ * Poke the watchdog when an interrupt is received
+ *
+ * @cpl:
+ * @dev_id:
+ *
+ * Returns
+ */
+static irqreturn_t octeon_wdt_poke_irq(int cpl, void *dev_id)
+{
+	unsigned int core = cvmx_get_core_num();
+	int cpu = core2cpu(core);
+
+	if (do_coundown) {
+		if (per_cpu_countdown[cpu] > 0) {
+			/* We're alive, poke the watchdog */
+			cvmx_write_csr(CVMX_CIU_PP_POKEX(core), 1);
+			per_cpu_countdown[cpu]--;
+		} else {
+			/* Bad news, you are about to reboot. */
+			disable_irq_nosync(cpl);
+			cpumask_clear_cpu(cpu, &irq_enabled_cpus);
+		}
+	} else {
+		/* Not open, just ping away... */
+		cvmx_write_csr(CVMX_CIU_PP_POKEX(core), 1);
+	}
+	return IRQ_HANDLED;
+}
+
+/* From setup.c */
+extern int prom_putchar(char c);
+
+/**
+ * Write a string to the uart
+ *
+ * @str:        String to write
+ */
+static void octeon_wdt_write_string(const char *str)
+{
+	/* Just loop writing one byte at a time */
+	while (*str)
+		prom_putchar(*str++);
+}
+
+/**
+ * Write a hex number out of the uart
+ *
+ * @value:      Number to display
+ * @digits:     Number of digits to print (1 to 16)
+ */
+static void octeon_wdt_write_hex(u64 value, int digits)
+{
+	int d;
+	int v;
+	for (d = 0; d < digits; d++) {
+		v = (value >> ((digits - d - 1) * 4)) & 0xf;
+		if (v >= 10)
+			prom_putchar('a' + v - 10);
+		else
+			prom_putchar('0' + v);
+	}
+}
+
+const char *reg_name[] = {
+	"$0", "at", "v0", "v1", "a0", "a1", "a2", "a3",
+	"a4", "a5", "a6", "a7", "t0", "t1", "t2", "t3",
+	"s0", "s1", "s2", "s3", "s4", "s5", "s6", "s7",
+	"t8", "t9", "k0", "k1", "gp", "sp", "s8", "ra"
+};
+
+/**
+ * NMI stage 3 handler. NMIs are handled in the following manner:
+ * 1) The first NMI handler enables CVMSEG and transfers from
+ * the bootbus region into normal memory. It is careful to not
+ * destroy any registers.
+ * 2) The second stage handler uses CVMSEG to save the registers
+ * and create a stack for C code. It then calls the third level
+ * handler with one argument, a pointer to the register values.
+ * 3) The third, and final, level handler is the following C
+ * function that prints out some useful infomration.
+ *
+ * @reg:    Pointer to register state before the NMI
+ */
+void octeon_wdt_nmi_stage3(u64 reg[32])
+{
+	u64 i;
+
+	unsigned int coreid = cvmx_get_core_num();
+	/*
+	 * Save status and cause early to get them before any changes
+	 * might happen.
+	 */
+	u64 cp0_cause = read_c0_cause();
+	u64 cp0_status = read_c0_status();
+	u64 cp0_error_epc = read_c0_errorepc();
+	u64 cp0_epc = read_c0_epc();
+
+	/* Delay so output from all cores output is not jumbled together. */
+	__delay(100000000ull * coreid);
+
+	octeon_wdt_write_string("\r\n*** NMI Watchdog interrupt on Core 0x");
+	octeon_wdt_write_hex(coreid, 1);
+	octeon_wdt_write_string(" ***\r\n");
+	for (i = 0; i < 32; i++) {
+		octeon_wdt_write_string("\t");
+		octeon_wdt_write_string(reg_name[i]);
+		octeon_wdt_write_string("\t0x");
+		octeon_wdt_write_hex(reg[i], 16);
+		if (i & 1)
+			octeon_wdt_write_string("\r\n");
+	}
+	octeon_wdt_write_string("\terr_epc\t0x");
+	octeon_wdt_write_hex(cp0_error_epc, 16);
+
+	octeon_wdt_write_string("\tepc\t0x");
+	octeon_wdt_write_hex(cp0_epc, 16);
+	octeon_wdt_write_string("\r\n");
+
+	octeon_wdt_write_string("\tstatus\t0x");
+	octeon_wdt_write_hex(cp0_status, 16);
+	octeon_wdt_write_string("\tcause\t0x");
+	octeon_wdt_write_hex(cp0_cause, 16);
+	octeon_wdt_write_string("\r\n");
+
+	octeon_wdt_write_string("\tsum0\t0x");
+	octeon_wdt_write_hex(cvmx_read_csr(CVMX_CIU_INTX_SUM0(coreid * 2)), 16);
+	octeon_wdt_write_string("\ten0\t0x");
+	octeon_wdt_write_hex(cvmx_read_csr(CVMX_CIU_INTX_EN0(coreid * 2)), 16);
+	octeon_wdt_write_string("\r\n");
+
+	octeon_wdt_write_string("*** Chip soft reset soon ***\r\n");
+}
+
+static void octeon_wdt_disable_interrupt(int cpu)
+{
+	unsigned int core;
+	unsigned int irq;
+	union cvmx_ciu_wdogx ciu_wdog;
+
+	core = cpu2core(cpu);
+
+	irq = OCTEON_IRQ_WDOG0 + core;
+
+	/* Poke the watchdog to clear out its state */
+	cvmx_write_csr(CVMX_CIU_PP_POKEX(core), 1);
+
+	/* Disable the hardware. */
+	ciu_wdog.u64 = 0;
+	cvmx_write_csr(CVMX_CIU_WDOGX(core), ciu_wdog.u64);
+
+	free_irq(irq, octeon_wdt_poke_irq);
+}
+
+static void octeon_wdt_setup_interrupt(int cpu)
+{
+	unsigned int core;
+	unsigned int irq;
+	union cvmx_ciu_wdogx ciu_wdog;
+
+	core = cpu2core(cpu);
+
+	/* Disable it before doing anything with the interrupts. */
+	ciu_wdog.u64 = 0;
+	cvmx_write_csr(CVMX_CIU_WDOGX(core), ciu_wdog.u64);
+
+	per_cpu_countdown[cpu] = countdown_reset;
+
+	irq = OCTEON_IRQ_WDOG0 + core;
+
+	if (request_irq(irq, octeon_wdt_poke_irq,
+			IRQF_DISABLED, "octeon_wdt", octeon_wdt_poke_irq))
+		panic("octeon_wdt: Couldn't obtain irq %d", irq);
+
+	cpumask_set_cpu(cpu, &irq_enabled_cpus);
+
+	/* Poke the watchdog to clear out its state */
+	cvmx_write_csr(CVMX_CIU_PP_POKEX(core), 1);
+
+	/* Finally enable the watchdog now that all handlers are installed */
+	ciu_wdog.u64 = 0;
+	ciu_wdog.s.len = timeout_cnt;
+	ciu_wdog.s.mode = 3;	/* 3 = Interrupt + NMI + Soft-Reset */
+	cvmx_write_csr(CVMX_CIU_WDOGX(core), ciu_wdog.u64);
+}
+
+static int octeon_wdt_cpu_callback(struct notifier_block *nfb,
+					   unsigned long action, void *hcpu)
+{
+	unsigned int cpu = (unsigned long)hcpu;
+
+	switch (action) {
+	case CPU_DOWN_PREPARE:
+		octeon_wdt_disable_interrupt(cpu);
+		break;
+	case CPU_ONLINE:
+	case CPU_DOWN_FAILED:
+		octeon_wdt_setup_interrupt(cpu);
+		break;
+	default:
+		break;
+	}
+	return NOTIFY_OK;
+}
+
+static void octeon_wdt_ping(void)
+{
+	int cpu;
+	int coreid;
+
+	for_each_online_cpu(cpu) {
+		coreid = cpu2core(cpu);
+		cvmx_write_csr(CVMX_CIU_PP_POKEX(coreid), 1);
+		per_cpu_countdown[cpu] = countdown_reset;
+		if ((countdown_reset || !do_coundown) &&
+		    !cpumask_test_cpu(cpu, &irq_enabled_cpus)) {
+			/* We have to enable the irq */
+			int irq = OCTEON_IRQ_WDOG0 + coreid;
+			enable_irq(irq);
+			cpumask_set_cpu(cpu, &irq_enabled_cpus);
+		}
+	}
+}
+
+static void octeon_wdt_calc_parameters(int t)
+{
+	unsigned int periods;
+
+	timeout_sec = max_timeout_sec;
+
+
+	/*
+	 * Find the largest interrupt period, that can evenly divide
+	 * the requested heartbeat time.
+	 */
+	while ((t % timeout_sec) != 0)
+		timeout_sec--;
+
+	periods = t / timeout_sec;
+
+	/*
+	 * The last two periods are after the irq is disabled, and
+	 * then to the nmi, so we subtract them off.
+	 */
+
+	countdown_reset = periods > 2 ? periods - 2 : 0;
+	heartbeat = t;
+	timeout_cnt = ((octeon_get_clock_rate() >> 8) * timeout_sec) >> 8;
+}
+
+static int octeon_wdt_set_heartbeat(int t)
+{
+	int cpu;
+	int coreid;
+	union cvmx_ciu_wdogx ciu_wdog;
+
+	if (t <= 0)
+		return -1;
+
+	octeon_wdt_calc_parameters(t);
+
+	for_each_online_cpu(cpu) {
+		coreid = cpu2core(cpu);
+		cvmx_write_csr(CVMX_CIU_PP_POKEX(coreid), 1);
+		ciu_wdog.u64 = 0;
+		ciu_wdog.s.len = timeout_cnt;
+		ciu_wdog.s.mode = 3;	/* 3 = Interrupt + NMI + Soft-Reset */
+		cvmx_write_csr(CVMX_CIU_WDOGX(coreid), ciu_wdog.u64);
+		cvmx_write_csr(CVMX_CIU_PP_POKEX(coreid), 1);
+	}
+	octeon_wdt_ping(); /* Get the irqs back on. */
+	return 0;
+}
+
+/**
+ *	octeon_wdt_write:
+ *	@file: file handle to the watchdog
+ *	@buf: buffer to write (unused as data does not matter here
+ *	@count: count of bytes
+ *	@ppos: pointer to the position to write. No seeks allowed
+ *
+ *	A write to a watchdog device is defined as a keepalive signal. Any
+ *	write of data will do, as we we don't define content meaning.
+ */
+
+static ssize_t octeon_wdt_write(struct file *file, const char __user *buf,
+				size_t count, loff_t *ppos)
+{
+	if (count) {
+		if (!nowayout) {
+			size_t i;
+
+			/* In case it was set long ago */
+			expect_close = 0;
+
+			for (i = 0; i != count; i++) {
+				char c;
+				if (get_user(c, buf + i))
+					return -EFAULT;
+				if (c == 'V')
+					expect_close = 1;
+			}
+		}
+		octeon_wdt_ping();
+	}
+	return count;
+}
+
+/**
+ *	octeon_wdt_ioctl:
+ *	@file: file handle to the device
+ *	@cmd: watchdog command
+ *	@arg: argument pointer
+ *
+ *	The watchdog API defines a common set of functions for all
+ *	watchdogs according to their available features. We only
+ *	actually usefully support querying capabilities and setting
+ *	the timeout.
+ */
+
+static long octeon_wdt_ioctl(struct file *file, unsigned int cmd,
+			     unsigned long arg)
+{
+	void __user *argp = (void __user *)arg;
+	int __user *p = argp;
+	int new_heartbeat;
+
+	static struct watchdog_info ident = {
+		.options =		WDIOF_SETTIMEOUT|
+					WDIOF_MAGICCLOSE|
+					WDIOF_KEEPALIVEPING,
+		.firmware_version =	1,
+		.identity =		"OCTEON",
+	};
+
+	switch (cmd) {
+	case WDIOC_GETSUPPORT:
+		return copy_to_user(argp, &ident, sizeof(ident)) ? -EFAULT : 0;
+	case WDIOC_GETSTATUS:
+	case WDIOC_GETBOOTSTATUS:
+		return put_user(0, p);
+	case WDIOC_KEEPALIVE:
+		octeon_wdt_ping();
+		return 0;
+	case WDIOC_SETTIMEOUT:
+		if (get_user(new_heartbeat, p))
+			return -EFAULT;
+		if (octeon_wdt_set_heartbeat(new_heartbeat))
+			return -EINVAL;
+		/* Fall through. */
+	case WDIOC_GETTIMEOUT:
+		return put_user(heartbeat, p);
+	default:
+		return -ENOTTY;
+	}
+}
+
+/**
+ *	octeon_wdt_open:
+ *	@inode: inode of device
+ *	@file: file handle to device
+ *
+ *	The watchdog device has been opened. The watchdog device is single
+ *	open and on opening we do a ping to reset the counters.
+ */
+
+static int octeon_wdt_open(struct inode *inode, struct file *file)
+{
+	if (test_and_set_bit(0, &octeon_wdt_is_open))
+		return -EBUSY;
+	/*
+	 *	Activate
+	 */
+	octeon_wdt_ping();
+	do_coundown = 1;
+	return nonseekable_open(inode, file);
+}
+
+/**
+ *	octeon_wdt_release:
+ *	@inode: inode to board
+ *	@file: file handle to board
+ *
+ *	The watchdog has a configurable API. There is a religious dispute
+ *	between people who want their watchdog to be able to shut down and
+ *	those who want to be sure if the watchdog manager dies the machine
+ *	reboots. In the former case we disable the counters, in the latter
+ *	case you have to open it again very soon.
+ */
+
+static int octeon_wdt_release(struct inode *inode, struct file *file)
+{
+	if (expect_close) {
+		do_coundown = 0;
+		octeon_wdt_ping();
+	} else {
+		pr_crit("octeon_wdt: WDT device closed unexpectedly.  WDT will not stop!\n");
+	}
+	clear_bit(0, &octeon_wdt_is_open);
+	expect_close = 0;
+	return 0;
+}
+
+static const struct file_operations octeon_wdt_fops = {
+	.owner		= THIS_MODULE,
+	.llseek		= no_llseek,
+	.write		= octeon_wdt_write,
+	.unlocked_ioctl	= octeon_wdt_ioctl,
+	.open		= octeon_wdt_open,
+	.release	= octeon_wdt_release,
+};
+
+static struct miscdevice octeon_wdt_miscdev = {
+	.minor	= WATCHDOG_MINOR,
+	.name	= "watchdog",
+	.fops	= &octeon_wdt_fops,
+};
+
+static struct notifier_block octeon_wdt_cpu_notifier = {
+	.notifier_call = octeon_wdt_cpu_callback,
+};
+
+
+/**
+ * Module/ driver initialization.
+ *
+ * Returns Zero on success
+ */
+static int __init octeon_wdt_init(void)
+{
+	int i;
+	int ret;
+	int cpu;
+	u64 *ptr;
+
+	/*
+	 * Watchdog time expiration length = The 16 bits of LEN
+	 * represent the most significant bits of a 24 bit decrementer
+	 * that decrements every 256 cycles.
+	 *
+	 * Try for a timeout of 5 sec, if that fails a smaller number
+	 * of even seconds,
+	 */
+	max_timeout_sec = 6;
+	do {
+		max_timeout_sec--;
+		timeout_cnt = ((octeon_get_clock_rate() >> 8) * max_timeout_sec) >> 8;
+	} while (timeout_cnt > 65535);
+
+	BUG_ON(timeout_cnt == 0);
+
+	octeon_wdt_calc_parameters(heartbeat);
+
+	pr_info("octeon_wdt: Initial granularity %d Sec.\n", timeout_sec);
+
+	ret = misc_register(&octeon_wdt_miscdev);
+	if (ret) {
+		pr_err("octeon_wdt: cannot register miscdev on minor=%d (err=%d)\n",
+			WATCHDOG_MINOR, ret);
+		goto out;
+	}
+
+	/* Build the NMI handler ... */
+	octeon_wdt_build_stage1();
+
+	/* ... and install it. */
+	ptr = (u64 *) nmi_stage1_insns;
+	for (i = 0; i < 16; i++) {
+		cvmx_write_csr(CVMX_MIO_BOOT_LOC_ADR, i * 8);
+		cvmx_write_csr(CVMX_MIO_BOOT_LOC_DAT, ptr[i]);
+	}
+	cvmx_write_csr(CVMX_MIO_BOOT_LOC_CFGX(0), 0x81fc0000);
+
+	cpumask_clear(&irq_enabled_cpus);
+
+	for_each_online_cpu(cpu)
+		octeon_wdt_setup_interrupt(cpu);
+
+	register_hotcpu_notifier(&octeon_wdt_cpu_notifier);
+out:
+	return ret;
+}
+
+/**
+ * Module / driver shutdown
+ */
+static void __exit octeon_wdt_cleanup(void)
+{
+	int cpu;
+
+	misc_deregister(&octeon_wdt_miscdev);
+
+	unregister_hotcpu_notifier(&octeon_wdt_cpu_notifier);
+
+	for_each_online_cpu(cpu) {
+		int core = cpu2core(cpu);
+		/* Disable the watchdog */
+		cvmx_write_csr(CVMX_CIU_WDOGX(core), 0);
+		/* Free the interrupt handler */
+		free_irq(OCTEON_IRQ_WDOG0 + core, octeon_wdt_poke_irq);
+	}
+	/*
+	 * Disable the boot-bus memory, the code it points to is soon
+	 * to go missing.
+	 */
+	cvmx_write_csr(CVMX_MIO_BOOT_LOC_CFGX(0), 0);
+}
+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Cavium Networks <support@caviumnetworks.com>");
+MODULE_DESCRIPTION("Cavium Networks Octeon Watchdog driver.");
+module_init(octeon_wdt_init);
+module_exit(octeon_wdt_cleanup);
diff --git a/drivers/watchdog/octeon-wdt-nmi.S b/drivers/watchdog/octeon-wdt-nmi.S
new file mode 100644
index 0000000..8a900a5
--- /dev/null
+++ b/drivers/watchdog/octeon-wdt-nmi.S
@@ -0,0 +1,64 @@
+/*
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * Copyright (C) 2007 Cavium Networks
+ */
+#include <asm/asm.h>
+#include <asm/regdef.h>
+
+#define SAVE_REG(r)	sd $r, -32768+6912-(32-r)*8($0)
+
+        NESTED(octeon_wdt_nmi_stage2, 0, sp)
+	.set 	push
+	.set 	noreorder
+	.set 	noat
+	/* Save all registers to the top CVMSEG. This shouldn't
+	 * corrupt any state used by the kernel. Also all registers
+	 * should have the value right before the NMI. */
+	SAVE_REG(0)
+	SAVE_REG(1)
+	SAVE_REG(2)
+	SAVE_REG(3)
+	SAVE_REG(4)
+	SAVE_REG(5)
+	SAVE_REG(6)
+	SAVE_REG(7)
+	SAVE_REG(8)
+	SAVE_REG(9)
+	SAVE_REG(10)
+	SAVE_REG(11)
+	SAVE_REG(12)
+	SAVE_REG(13)
+	SAVE_REG(14)
+	SAVE_REG(15)
+	SAVE_REG(16)
+	SAVE_REG(17)
+	SAVE_REG(18)
+	SAVE_REG(19)
+	SAVE_REG(20)
+	SAVE_REG(21)
+	SAVE_REG(22)
+	SAVE_REG(23)
+	SAVE_REG(24)
+	SAVE_REG(25)
+	SAVE_REG(26)
+	SAVE_REG(27)
+	SAVE_REG(28)
+	SAVE_REG(29)
+	SAVE_REG(30)
+	SAVE_REG(31)
+	/* Set the stack to begin right below the registers */
+	li	sp, -32768+6912-32*8
+	/* Load the address of the third stage handler */
+	dla	a0, octeon_wdt_nmi_stage3
+	/* Call the third stage handler */
+	jal	a0
+	/* a0 is the address of the saved registers */
+	 move	a0, sp
+	/* Loop forvever if we get here. */
+1:	b	1b
+	nop
+	.set pop
+	END(octeon_wdt_nmi_stage2)