Add qemu 2.4.0
[kvmfornfv.git] / qemu / hw / watchdog / wdt_i6300esb.c
diff --git a/qemu/hw/watchdog/wdt_i6300esb.c b/qemu/hw/watchdog/wdt_i6300esb.c
new file mode 100644 (file)
index 0000000..cfa2b1b
--- /dev/null
@@ -0,0 +1,470 @@
+/*
+ * Virtual hardware watchdog.
+ *
+ * Copyright (C) 2009 Red Hat Inc.
+ *
+ * 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.
+ *
+ * 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.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, see <http://www.gnu.org/licenses/>.
+ *
+ * By Richard W.M. Jones (rjones@redhat.com).
+ */
+
+#include <inttypes.h>
+
+#include "qemu-common.h"
+#include "qemu/timer.h"
+#include "sysemu/watchdog.h"
+#include "hw/hw.h"
+#include "hw/pci/pci.h"
+
+/*#define I6300ESB_DEBUG 1*/
+
+#ifdef I6300ESB_DEBUG
+#define i6300esb_debug(fs,...) \
+    fprintf(stderr,"i6300esb: %s: "fs,__func__,##__VA_ARGS__)
+#else
+#define i6300esb_debug(fs,...)
+#endif
+
+/* PCI configuration registers */
+#define ESB_CONFIG_REG  0x60            /* Config register                   */
+#define ESB_LOCK_REG    0x68            /* WDT lock register                 */
+
+/* Memory mapped registers (offset from base address) */
+#define ESB_TIMER1_REG  0x00            /* Timer1 value after each reset     */
+#define ESB_TIMER2_REG  0x04            /* Timer2 value after each reset     */
+#define ESB_GINTSR_REG  0x08            /* General Interrupt Status Register */
+#define ESB_RELOAD_REG  0x0c            /* Reload register                   */
+
+/* Lock register bits */
+#define ESB_WDT_FUNC    (0x01 << 2)   /* Watchdog functionality            */
+#define ESB_WDT_ENABLE  (0x01 << 1)   /* Enable WDT                        */
+#define ESB_WDT_LOCK    (0x01 << 0)   /* Lock (nowayout)                   */
+
+/* Config register bits */
+#define ESB_WDT_REBOOT  (0x01 << 5)   /* Enable reboot on timeout          */
+#define ESB_WDT_FREQ    (0x01 << 2)   /* Decrement frequency               */
+#define ESB_WDT_INTTYPE (0x11 << 0)   /* Interrupt type on timer1 timeout  */
+
+/* Reload register bits */
+#define ESB_WDT_RELOAD  (0x01 << 8)    /* prevent timeout                   */
+
+/* Magic constants */
+#define ESB_UNLOCK1     0x80            /* Step 1 to unlock reset registers  */
+#define ESB_UNLOCK2     0x86            /* Step 2 to unlock reset registers  */
+
+/* Device state. */
+struct I6300State {
+    PCIDevice dev;
+    MemoryRegion io_mem;
+
+    int reboot_enabled;         /* "Reboot" on timer expiry.  The real action
+                                 * performed depends on the -watchdog-action
+                                 * param passed on QEMU command line.
+                                 */
+    int clock_scale;            /* Clock scale. */
+#define CLOCK_SCALE_1KHZ 0
+#define CLOCK_SCALE_1MHZ 1
+
+    int int_type;               /* Interrupt type generated. */
+#define INT_TYPE_IRQ 0          /* APIC 1, INT 10 */
+#define INT_TYPE_SMI 2
+#define INT_TYPE_DISABLED 3
+
+    int free_run;               /* If true, reload timer on expiry. */
+    int locked;                 /* If true, enabled field cannot be changed. */
+    int enabled;                /* If true, watchdog is enabled. */
+
+    QEMUTimer *timer;           /* The actual watchdog timer. */
+
+    uint32_t timer1_preload;    /* Values preloaded into timer1, timer2. */
+    uint32_t timer2_preload;
+    int stage;                  /* Stage (1 or 2). */
+
+    int unlock_state;           /* Guest writes 0x80, 0x86 to unlock the
+                                 * registers, and we transition through
+                                 * states 0 -> 1 -> 2 when this happens.
+                                 */
+
+    int previous_reboot_flag;   /* If the watchdog caused the previous
+                                 * reboot, this flag will be set.
+                                 */
+};
+
+typedef struct I6300State I6300State;
+
+#define TYPE_WATCHDOG_I6300ESB_DEVICE "i6300esb"
+#define WATCHDOG_I6300ESB_DEVICE(obj) \
+    OBJECT_CHECK(I6300State, (obj), TYPE_WATCHDOG_I6300ESB_DEVICE)
+
+/* This function is called when the watchdog has either been enabled
+ * (hence it starts counting down) or has been keep-alived.
+ */
+static void i6300esb_restart_timer(I6300State *d, int stage)
+{
+    int64_t timeout;
+
+    if (!d->enabled)
+        return;
+
+    d->stage = stage;
+
+    if (d->stage <= 1)
+        timeout = d->timer1_preload;
+    else
+        timeout = d->timer2_preload;
+
+    if (d->clock_scale == CLOCK_SCALE_1KHZ)
+        timeout <<= 15;
+    else
+        timeout <<= 5;
+
+    /* Get the timeout in units of ticks_per_sec.
+     *
+     * ticks_per_sec is typically 10^9 == 0x3B9ACA00 (30 bits), with
+     * 20 bits of user supplied preload, and 15 bits of scale, the
+     * multiply here can exceed 64-bits, before we divide by 33MHz, so
+     * we use a higher-precision intermediate result.
+     */
+    timeout = muldiv64(get_ticks_per_sec(), timeout, 33000000);
+
+    i6300esb_debug("stage %d, timeout %" PRIi64 "\n", d->stage, timeout);
+
+    timer_mod(d->timer, qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL) + timeout);
+}
+
+/* This is called when the guest disables the watchdog. */
+static void i6300esb_disable_timer(I6300State *d)
+{
+    i6300esb_debug("timer disabled\n");
+
+    timer_del(d->timer);
+}
+
+static void i6300esb_reset(DeviceState *dev)
+{
+    PCIDevice *pdev = PCI_DEVICE(dev);
+    I6300State *d = WATCHDOG_I6300ESB_DEVICE(pdev);
+
+    i6300esb_debug("I6300State = %p\n", d);
+
+    i6300esb_disable_timer(d);
+
+    /* NB: Don't change d->previous_reboot_flag in this function. */
+
+    d->reboot_enabled = 1;
+    d->clock_scale = CLOCK_SCALE_1KHZ;
+    d->int_type = INT_TYPE_IRQ;
+    d->free_run = 0;
+    d->locked = 0;
+    d->enabled = 0;
+    d->timer1_preload = 0xfffff;
+    d->timer2_preload = 0xfffff;
+    d->stage = 1;
+    d->unlock_state = 0;
+}
+
+/* This function is called when the watchdog expires.  Note that
+ * the hardware has two timers, and so expiry happens in two stages.
+ * If d->stage == 1 then we perform the first stage action (usually,
+ * sending an interrupt) and then restart the timer again for the
+ * second stage.  If the second stage expires then the watchdog
+ * really has run out.
+ */
+static void i6300esb_timer_expired(void *vp)
+{
+    I6300State *d = vp;
+
+    i6300esb_debug("stage %d\n", d->stage);
+
+    if (d->stage == 1) {
+        /* What to do at the end of stage 1? */
+        switch (d->int_type) {
+        case INT_TYPE_IRQ:
+            fprintf(stderr, "i6300esb_timer_expired: I would send APIC 1 INT 10 here if I knew how (XXX)\n");
+            break;
+        case INT_TYPE_SMI:
+            fprintf(stderr, "i6300esb_timer_expired: I would send SMI here if I knew how (XXX)\n");
+            break;
+        }
+
+        /* Start the second stage. */
+        i6300esb_restart_timer(d, 2);
+    } else {
+        /* Second stage expired, reboot for real. */
+        if (d->reboot_enabled) {
+            d->previous_reboot_flag = 1;
+            watchdog_perform_action(); /* This reboots, exits, etc */
+            i6300esb_reset(&d->dev.qdev);
+        }
+
+        /* In "free running mode" we start stage 1 again. */
+        if (d->free_run)
+            i6300esb_restart_timer(d, 1);
+    }
+}
+
+static void i6300esb_config_write(PCIDevice *dev, uint32_t addr,
+                                  uint32_t data, int len)
+{
+    I6300State *d = WATCHDOG_I6300ESB_DEVICE(dev);
+    int old;
+
+    i6300esb_debug("addr = %x, data = %x, len = %d\n", addr, data, len);
+
+    if (addr == ESB_CONFIG_REG && len == 2) {
+        d->reboot_enabled = (data & ESB_WDT_REBOOT) == 0;
+        d->clock_scale =
+            (data & ESB_WDT_FREQ) != 0 ? CLOCK_SCALE_1MHZ : CLOCK_SCALE_1KHZ;
+        d->int_type = (data & ESB_WDT_INTTYPE);
+    } else if (addr == ESB_LOCK_REG && len == 1) {
+        if (!d->locked) {
+            d->locked = (data & ESB_WDT_LOCK) != 0;
+            d->free_run = (data & ESB_WDT_FUNC) != 0;
+            old = d->enabled;
+            d->enabled = (data & ESB_WDT_ENABLE) != 0;
+            if (!old && d->enabled) /* Enabled transitioned from 0 -> 1 */
+                i6300esb_restart_timer(d, 1);
+            else if (!d->enabled)
+                i6300esb_disable_timer(d);
+        }
+    } else {
+        pci_default_write_config(dev, addr, data, len);
+    }
+}
+
+static uint32_t i6300esb_config_read(PCIDevice *dev, uint32_t addr, int len)
+{
+    I6300State *d = WATCHDOG_I6300ESB_DEVICE(dev);
+    uint32_t data;
+
+    i6300esb_debug ("addr = %x, len = %d\n", addr, len);
+
+    if (addr == ESB_CONFIG_REG && len == 2) {
+        data =
+            (d->reboot_enabled ? 0 : ESB_WDT_REBOOT) |
+            (d->clock_scale == CLOCK_SCALE_1MHZ ? ESB_WDT_FREQ : 0) |
+            d->int_type;
+        return data;
+    } else if (addr == ESB_LOCK_REG && len == 1) {
+        data =
+            (d->free_run ? ESB_WDT_FUNC : 0) |
+            (d->locked ? ESB_WDT_LOCK : 0) |
+            (d->enabled ? ESB_WDT_ENABLE : 0);
+        return data;
+    } else {
+        return pci_default_read_config(dev, addr, len);
+    }
+}
+
+static uint32_t i6300esb_mem_readb(void *vp, hwaddr addr)
+{
+    i6300esb_debug ("addr = %x\n", (int) addr);
+
+    return 0;
+}
+
+static uint32_t i6300esb_mem_readw(void *vp, hwaddr addr)
+{
+    uint32_t data = 0;
+    I6300State *d = vp;
+
+    i6300esb_debug("addr = %x\n", (int) addr);
+
+    if (addr == 0xc) {
+        /* The previous reboot flag is really bit 9, but there is
+         * a bug in the Linux driver where it thinks it's bit 12.
+         * Set both.
+         */
+        data = d->previous_reboot_flag ? 0x1200 : 0;
+    }
+
+    return data;
+}
+
+static uint32_t i6300esb_mem_readl(void *vp, hwaddr addr)
+{
+    i6300esb_debug("addr = %x\n", (int) addr);
+
+    return 0;
+}
+
+static void i6300esb_mem_writeb(void *vp, hwaddr addr, uint32_t val)
+{
+    I6300State *d = vp;
+
+    i6300esb_debug("addr = %x, val = %x\n", (int) addr, val);
+
+    if (addr == 0xc && val == 0x80)
+        d->unlock_state = 1;
+    else if (addr == 0xc && val == 0x86 && d->unlock_state == 1)
+        d->unlock_state = 2;
+}
+
+static void i6300esb_mem_writew(void *vp, hwaddr addr, uint32_t val)
+{
+    I6300State *d = vp;
+
+    i6300esb_debug("addr = %x, val = %x\n", (int) addr, val);
+
+    if (addr == 0xc && val == 0x80)
+        d->unlock_state = 1;
+    else if (addr == 0xc && val == 0x86 && d->unlock_state == 1)
+        d->unlock_state = 2;
+    else {
+        if (d->unlock_state == 2) {
+            if (addr == 0xc) {
+                if ((val & 0x100) != 0)
+                    /* This is the "ping" from the userspace watchdog in
+                     * the guest ...
+                     */
+                    i6300esb_restart_timer(d, 1);
+
+                /* Setting bit 9 resets the previous reboot flag.
+                 * There's a bug in the Linux driver where it sets
+                 * bit 12 instead.
+                 */
+                if ((val & 0x200) != 0 || (val & 0x1000) != 0) {
+                    d->previous_reboot_flag = 0;
+                }
+            }
+
+            d->unlock_state = 0;
+        }
+    }
+}
+
+static void i6300esb_mem_writel(void *vp, hwaddr addr, uint32_t val)
+{
+    I6300State *d = vp;
+
+    i6300esb_debug ("addr = %x, val = %x\n", (int) addr, val);
+
+    if (addr == 0xc && val == 0x80)
+        d->unlock_state = 1;
+    else if (addr == 0xc && val == 0x86 && d->unlock_state == 1)
+        d->unlock_state = 2;
+    else {
+        if (d->unlock_state == 2) {
+            if (addr == 0)
+                d->timer1_preload = val & 0xfffff;
+            else if (addr == 4)
+                d->timer2_preload = val & 0xfffff;
+
+            d->unlock_state = 0;
+        }
+    }
+}
+
+static const MemoryRegionOps i6300esb_ops = {
+    .old_mmio = {
+        .read = {
+            i6300esb_mem_readb,
+            i6300esb_mem_readw,
+            i6300esb_mem_readl,
+        },
+        .write = {
+            i6300esb_mem_writeb,
+            i6300esb_mem_writew,
+            i6300esb_mem_writel,
+        },
+    },
+    .endianness = DEVICE_LITTLE_ENDIAN,
+};
+
+static const VMStateDescription vmstate_i6300esb = {
+    .name = "i6300esb_wdt",
+    /* With this VMSD's introduction, version_id/minimum_version_id were
+     * erroneously set to sizeof(I6300State), causing a somewhat random
+     * version_id to be set for every build. This eventually broke
+     * migration.
+     *
+     * To correct this without breaking old->new migration for older
+     * versions of QEMU, we've set version_id to a value high enough
+     * to exceed all past values of sizeof(I6300State) across various
+     * build environments, and have reset minimum_version_id to 1,
+     * since this VMSD has never changed and thus can accept all past
+     * versions.
+     *
+     * For future changes we can treat these values as we normally would.
+     */
+    .version_id = 10000,
+    .minimum_version_id = 1,
+    .fields = (VMStateField[]) {
+        VMSTATE_PCI_DEVICE(dev, I6300State),
+        VMSTATE_INT32(reboot_enabled, I6300State),
+        VMSTATE_INT32(clock_scale, I6300State),
+        VMSTATE_INT32(int_type, I6300State),
+        VMSTATE_INT32(free_run, I6300State),
+        VMSTATE_INT32(locked, I6300State),
+        VMSTATE_INT32(enabled, I6300State),
+        VMSTATE_TIMER_PTR(timer, I6300State),
+        VMSTATE_UINT32(timer1_preload, I6300State),
+        VMSTATE_UINT32(timer2_preload, I6300State),
+        VMSTATE_INT32(stage, I6300State),
+        VMSTATE_INT32(unlock_state, I6300State),
+        VMSTATE_INT32(previous_reboot_flag, I6300State),
+        VMSTATE_END_OF_LIST()
+    }
+};
+
+static void i6300esb_realize(PCIDevice *dev, Error **errp)
+{
+    I6300State *d = WATCHDOG_I6300ESB_DEVICE(dev);
+
+    i6300esb_debug("I6300State = %p\n", d);
+
+    d->timer = timer_new_ns(QEMU_CLOCK_VIRTUAL, i6300esb_timer_expired, d);
+    d->previous_reboot_flag = 0;
+
+    memory_region_init_io(&d->io_mem, OBJECT(d), &i6300esb_ops, d,
+                          "i6300esb", 0x10);
+    pci_register_bar(&d->dev, 0, 0, &d->io_mem);
+    /* qemu_register_coalesced_mmio (addr, 0x10); ? */
+}
+
+static WatchdogTimerModel model = {
+    .wdt_name = "i6300esb",
+    .wdt_description = "Intel 6300ESB",
+};
+
+static void i6300esb_class_init(ObjectClass *klass, void *data)
+{
+    DeviceClass *dc = DEVICE_CLASS(klass);
+    PCIDeviceClass *k = PCI_DEVICE_CLASS(klass);
+
+    k->config_read = i6300esb_config_read;
+    k->config_write = i6300esb_config_write;
+    k->realize = i6300esb_realize;
+    k->vendor_id = PCI_VENDOR_ID_INTEL;
+    k->device_id = PCI_DEVICE_ID_INTEL_ESB_9;
+    k->class_id = PCI_CLASS_SYSTEM_OTHER;
+    dc->reset = i6300esb_reset;
+    dc->vmsd = &vmstate_i6300esb;
+    set_bit(DEVICE_CATEGORY_MISC, dc->categories);
+}
+
+static const TypeInfo i6300esb_info = {
+    .name          = TYPE_WATCHDOG_I6300ESB_DEVICE,
+    .parent        = TYPE_PCI_DEVICE,
+    .instance_size = sizeof(I6300State),
+    .class_init    = i6300esb_class_init,
+};
+
+static void i6300esb_register_types(void)
+{
+    watchdog_add_model(&model);
+    type_register_static(&i6300esb_info);
+}
+
+type_init(i6300esb_register_types)