aboutsummaryrefslogtreecommitdiffstats
path: root/hw/misc/macio/pmu.c
diff options
context:
space:
mode:
Diffstat (limited to 'hw/misc/macio/pmu.c')
-rw-r--r--hw/misc/macio/pmu.c871
1 files changed, 871 insertions, 0 deletions
diff --git a/hw/misc/macio/pmu.c b/hw/misc/macio/pmu.c
new file mode 100644
index 000000000..eb39c6469
--- /dev/null
+++ b/hw/misc/macio/pmu.c
@@ -0,0 +1,871 @@
+/*
+ * QEMU PowerMac PMU device support
+ *
+ * Copyright (c) 2016 Benjamin Herrenschmidt, IBM Corp.
+ * Copyright (c) 2018 Mark Cave-Ayland
+ *
+ * Based on the CUDA device by:
+ *
+ * Copyright (c) 2004-2007 Fabrice Bellard
+ * Copyright (c) 2007 Jocelyn Mayer
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the "Software"), to deal
+ * in the Software without restriction, including without limitation the rights
+ * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+ * copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
+ * THE SOFTWARE.
+ */
+
+#include "qemu/osdep.h"
+#include "qemu-common.h"
+#include "hw/ppc/mac.h"
+#include "hw/qdev-properties.h"
+#include "migration/vmstate.h"
+#include "hw/input/adb.h"
+#include "hw/irq.h"
+#include "hw/misc/mos6522.h"
+#include "hw/misc/macio/gpio.h"
+#include "hw/misc/macio/pmu.h"
+#include "qapi/error.h"
+#include "qemu/timer.h"
+#include "sysemu/runstate.h"
+#include "qapi/error.h"
+#include "qemu/cutils.h"
+#include "qemu/log.h"
+#include "qemu/module.h"
+#include "trace.h"
+
+
+/* Bits in B data register: all active low */
+#define TACK 0x08 /* Transfer request (input) */
+#define TREQ 0x10 /* Transfer acknowledge (output) */
+
+/* PMU returns time_t's offset from Jan 1, 1904, not 1970 */
+#define RTC_OFFSET 2082844800
+
+#define VIA_TIMER_FREQ (4700000 / 6)
+
+static void via_update_irq(PMUState *s)
+{
+ MOS6522PMUState *mps = MOS6522_PMU(&s->mos6522_pmu);
+ MOS6522State *ms = MOS6522(mps);
+
+ bool new_state = !!(ms->ifr & ms->ier & (SR_INT | T1_INT | T2_INT));
+
+ if (new_state != s->via_irq_state) {
+ s->via_irq_state = new_state;
+ qemu_set_irq(s->via_irq, new_state);
+ }
+}
+
+static void via_set_sr_int(void *opaque)
+{
+ PMUState *s = opaque;
+ MOS6522PMUState *mps = MOS6522_PMU(&s->mos6522_pmu);
+ MOS6522State *ms = MOS6522(mps);
+ MOS6522DeviceClass *mdc = MOS6522_GET_CLASS(ms);
+
+ mdc->set_sr_int(ms);
+}
+
+static void pmu_update_extirq(PMUState *s)
+{
+ if ((s->intbits & s->intmask) != 0) {
+ macio_set_gpio(s->gpio, 1, false);
+ } else {
+ macio_set_gpio(s->gpio, 1, true);
+ }
+}
+
+static void pmu_adb_poll(void *opaque)
+{
+ PMUState *s = opaque;
+ ADBBusState *adb_bus = &s->adb_bus;
+ int olen;
+
+ if (!(s->intbits & PMU_INT_ADB)) {
+ olen = adb_poll(adb_bus, s->adb_reply, adb_bus->autopoll_mask);
+ trace_pmu_adb_poll(olen);
+
+ if (olen > 0) {
+ s->adb_reply_size = olen;
+ s->intbits |= PMU_INT_ADB | PMU_INT_ADB_AUTO;
+ pmu_update_extirq(s);
+ }
+ }
+}
+
+static void pmu_one_sec_timer(void *opaque)
+{
+ PMUState *s = opaque;
+
+ trace_pmu_one_sec_timer();
+
+ s->intbits |= PMU_INT_TICK;
+ pmu_update_extirq(s);
+ s->one_sec_target += 1000;
+
+ timer_mod(s->one_sec_timer, s->one_sec_target);
+}
+
+static void pmu_cmd_int_ack(PMUState *s,
+ const uint8_t *in_data, uint8_t in_len,
+ uint8_t *out_data, uint8_t *out_len)
+{
+ if (in_len != 0) {
+ qemu_log_mask(LOG_GUEST_ERROR,
+ "PMU: INT_ACK command, invalid len: %d want: 0\n",
+ in_len);
+ return;
+ }
+
+ /* Make appropriate reply packet */
+ if (s->intbits & PMU_INT_ADB) {
+ if (!s->adb_reply_size) {
+ qemu_log_mask(LOG_GUEST_ERROR,
+ "Odd, PMU_INT_ADB set with no reply in buffer\n");
+ }
+
+ memcpy(out_data + 1, s->adb_reply, s->adb_reply_size);
+ out_data[0] = s->intbits & (PMU_INT_ADB | PMU_INT_ADB_AUTO);
+ *out_len = s->adb_reply_size + 1;
+ s->intbits &= ~(PMU_INT_ADB | PMU_INT_ADB_AUTO);
+ s->adb_reply_size = 0;
+ } else {
+ out_data[0] = s->intbits;
+ s->intbits = 0;
+ *out_len = 1;
+ }
+
+ pmu_update_extirq(s);
+}
+
+static void pmu_cmd_set_int_mask(PMUState *s,
+ const uint8_t *in_data, uint8_t in_len,
+ uint8_t *out_data, uint8_t *out_len)
+{
+ if (in_len != 1) {
+ qemu_log_mask(LOG_GUEST_ERROR,
+ "PMU: SET_INT_MASK command, invalid len: %d want: 1\n",
+ in_len);
+ return;
+ }
+
+ trace_pmu_cmd_set_int_mask(s->intmask);
+ s->intmask = in_data[0];
+
+ pmu_update_extirq(s);
+}
+
+static void pmu_cmd_set_adb_autopoll(PMUState *s, uint16_t mask)
+{
+ ADBBusState *adb_bus = &s->adb_bus;
+
+ trace_pmu_cmd_set_adb_autopoll(mask);
+
+ if (mask) {
+ adb_set_autopoll_mask(adb_bus, mask);
+ adb_set_autopoll_enabled(adb_bus, true);
+ } else {
+ adb_set_autopoll_enabled(adb_bus, false);
+ }
+}
+
+static void pmu_cmd_adb(PMUState *s,
+ const uint8_t *in_data, uint8_t in_len,
+ uint8_t *out_data, uint8_t *out_len)
+{
+ int len, adblen;
+ uint8_t adb_cmd[255];
+
+ if (in_len < 2) {
+ qemu_log_mask(LOG_GUEST_ERROR,
+ "PMU: ADB PACKET, invalid len: %d want at least 2\n",
+ in_len);
+ return;
+ }
+
+ *out_len = 0;
+
+ if (!s->has_adb) {
+ trace_pmu_cmd_adb_nobus();
+ return;
+ }
+
+ /* Set autopoll is a special form of the command */
+ if (in_data[0] == 0 && in_data[1] == 0x86) {
+ uint16_t mask = in_data[2];
+ mask = (mask << 8) | in_data[3];
+ if (in_len != 4) {
+ qemu_log_mask(LOG_GUEST_ERROR,
+ "PMU: ADB Autopoll requires 4 bytes, got %d\n",
+ in_len);
+ return;
+ }
+
+ pmu_cmd_set_adb_autopoll(s, mask);
+ return;
+ }
+
+ trace_pmu_cmd_adb_request(in_len, in_data[0], in_data[1], in_data[2],
+ in_data[3], in_data[4]);
+
+ *out_len = 0;
+
+ /* Check ADB len */
+ adblen = in_data[2];
+ if (adblen > (in_len - 3)) {
+ qemu_log_mask(LOG_GUEST_ERROR,
+ "PMU: ADB len is %d > %d (in_len -3)...erroring\n",
+ adblen, in_len - 3);
+ len = -1;
+ } else if (adblen > 252) {
+ qemu_log_mask(LOG_GUEST_ERROR, "PMU: ADB command too big!\n");
+ len = -1;
+ } else {
+ /* Format command */
+ adb_cmd[0] = in_data[0];
+ memcpy(&adb_cmd[1], &in_data[3], in_len - 3);
+ len = adb_request(&s->adb_bus, s->adb_reply + 2, adb_cmd, in_len - 2);
+
+ trace_pmu_cmd_adb_reply(len);
+ }
+
+ if (len > 0) {
+ /* XXX Check this */
+ s->adb_reply_size = len + 2;
+ s->adb_reply[0] = 0x01;
+ s->adb_reply[1] = len;
+ } else {
+ /* XXX Check this */
+ s->adb_reply_size = 1;
+ s->adb_reply[0] = 0x00;
+ }
+
+ s->intbits |= PMU_INT_ADB;
+ pmu_update_extirq(s);
+}
+
+static void pmu_cmd_adb_poll_off(PMUState *s,
+ const uint8_t *in_data, uint8_t in_len,
+ uint8_t *out_data, uint8_t *out_len)
+{
+ ADBBusState *adb_bus = &s->adb_bus;
+
+ if (in_len != 0) {
+ qemu_log_mask(LOG_GUEST_ERROR,
+ "PMU: ADB POLL OFF command, invalid len: %d want: 0\n",
+ in_len);
+ return;
+ }
+
+ if (s->has_adb) {
+ adb_set_autopoll_enabled(adb_bus, false);
+ }
+}
+
+static void pmu_cmd_shutdown(PMUState *s,
+ const uint8_t *in_data, uint8_t in_len,
+ uint8_t *out_data, uint8_t *out_len)
+{
+ if (in_len != 4) {
+ qemu_log_mask(LOG_GUEST_ERROR,
+ "PMU: SHUTDOWN command, invalid len: %d want: 4\n",
+ in_len);
+ return;
+ }
+
+ *out_len = 1;
+ out_data[0] = 0;
+
+ if (in_data[0] != 'M' || in_data[1] != 'A' || in_data[2] != 'T' ||
+ in_data[3] != 'T') {
+
+ qemu_log_mask(LOG_GUEST_ERROR,
+ "PMU: SHUTDOWN command, Bad MATT signature\n");
+ return;
+ }
+
+ qemu_system_shutdown_request(SHUTDOWN_CAUSE_GUEST_SHUTDOWN);
+}
+
+static void pmu_cmd_reset(PMUState *s,
+ const uint8_t *in_data, uint8_t in_len,
+ uint8_t *out_data, uint8_t *out_len)
+{
+ if (in_len != 0) {
+ qemu_log_mask(LOG_GUEST_ERROR,
+ "PMU: RESET command, invalid len: %d want: 0\n",
+ in_len);
+ return;
+ }
+
+ qemu_system_reset_request(SHUTDOWN_CAUSE_GUEST_RESET);
+}
+
+static void pmu_cmd_get_rtc(PMUState *s,
+ const uint8_t *in_data, uint8_t in_len,
+ uint8_t *out_data, uint8_t *out_len)
+{
+ uint32_t ti;
+
+ if (in_len != 0) {
+ qemu_log_mask(LOG_GUEST_ERROR,
+ "PMU: GET_RTC command, invalid len: %d want: 0\n",
+ in_len);
+ return;
+ }
+
+ ti = s->tick_offset + (qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL)
+ / NANOSECONDS_PER_SECOND);
+ out_data[0] = ti >> 24;
+ out_data[1] = ti >> 16;
+ out_data[2] = ti >> 8;
+ out_data[3] = ti;
+ *out_len = 4;
+}
+
+static void pmu_cmd_set_rtc(PMUState *s,
+ const uint8_t *in_data, uint8_t in_len,
+ uint8_t *out_data, uint8_t *out_len)
+{
+ uint32_t ti;
+
+ if (in_len != 4) {
+ qemu_log_mask(LOG_GUEST_ERROR,
+ "PMU: SET_RTC command, invalid len: %d want: 4\n",
+ in_len);
+ return;
+ }
+
+ ti = (((uint32_t)in_data[0]) << 24) + (((uint32_t)in_data[1]) << 16)
+ + (((uint32_t)in_data[2]) << 8) + in_data[3];
+
+ s->tick_offset = ti - (qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL)
+ / NANOSECONDS_PER_SECOND);
+}
+
+static void pmu_cmd_system_ready(PMUState *s,
+ const uint8_t *in_data, uint8_t in_len,
+ uint8_t *out_data, uint8_t *out_len)
+{
+ /* Do nothing */
+}
+
+static void pmu_cmd_get_version(PMUState *s,
+ const uint8_t *in_data, uint8_t in_len,
+ uint8_t *out_data, uint8_t *out_len)
+{
+ *out_len = 1;
+ *out_data = 1; /* ??? Check what Apple does */
+}
+
+static void pmu_cmd_power_events(PMUState *s,
+ const uint8_t *in_data, uint8_t in_len,
+ uint8_t *out_data, uint8_t *out_len)
+{
+ if (in_len < 1) {
+ qemu_log_mask(LOG_GUEST_ERROR,
+ "PMU: POWER EVENTS command, invalid len %d, want at least 1\n",
+ in_len);
+ return;
+ }
+
+ switch (in_data[0]) {
+ /* Dummies for now */
+ case PMU_PWR_GET_POWERUP_EVENTS:
+ *out_len = 2;
+ out_data[0] = 0;
+ out_data[1] = 0;
+ break;
+ case PMU_PWR_SET_POWERUP_EVENTS:
+ case PMU_PWR_CLR_POWERUP_EVENTS:
+ break;
+ case PMU_PWR_GET_WAKEUP_EVENTS:
+ *out_len = 2;
+ out_data[0] = 0;
+ out_data[1] = 0;
+ break;
+ case PMU_PWR_SET_WAKEUP_EVENTS:
+ case PMU_PWR_CLR_WAKEUP_EVENTS:
+ break;
+ default:
+ qemu_log_mask(LOG_GUEST_ERROR,
+ "PMU: POWER EVENTS unknown subcommand 0x%02x\n",
+ in_data[0]);
+ }
+}
+
+static void pmu_cmd_get_cover(PMUState *s,
+ const uint8_t *in_data, uint8_t in_len,
+ uint8_t *out_data, uint8_t *out_len)
+{
+ /* Not 100% sure here, will have to check what a real Mac
+ * returns other than byte 0 bit 0 is LID closed on laptops
+ */
+ *out_len = 1;
+ *out_data = 0x00;
+}
+
+static void pmu_cmd_download_status(PMUState *s,
+ const uint8_t *in_data, uint8_t in_len,
+ uint8_t *out_data, uint8_t *out_len)
+{
+ /* This has to do with PMU firmware updates as far as I can tell.
+ *
+ * We return 0x62 which is what OpenPMU expects
+ */
+ *out_len = 1;
+ *out_data = 0x62;
+}
+
+static void pmu_cmd_read_pmu_ram(PMUState *s,
+ const uint8_t *in_data, uint8_t in_len,
+ uint8_t *out_data, uint8_t *out_len)
+{
+ if (in_len < 3) {
+ qemu_log_mask(LOG_GUEST_ERROR,
+ "PMU: READ_PMU_RAM command, invalid len %d, expected 3\n",
+ in_len);
+ return;
+ }
+
+ qemu_log_mask(LOG_GUEST_ERROR,
+ "PMU: Unsupported READ_PMU_RAM, args: %02x %02x %02x\n",
+ in_data[0], in_data[1], in_data[2]);
+
+ *out_len = 0;
+}
+
+/* description of commands */
+typedef struct PMUCmdHandler {
+ uint8_t command;
+ const char *name;
+ void (*handler)(PMUState *s,
+ const uint8_t *in_args, uint8_t in_len,
+ uint8_t *out_args, uint8_t *out_len);
+} PMUCmdHandler;
+
+static const PMUCmdHandler PMUCmdHandlers[] = {
+ { PMU_INT_ACK, "INT ACK", pmu_cmd_int_ack },
+ { PMU_SET_INTR_MASK, "SET INT MASK", pmu_cmd_set_int_mask },
+ { PMU_ADB_CMD, "ADB COMMAND", pmu_cmd_adb },
+ { PMU_ADB_POLL_OFF, "ADB POLL OFF", pmu_cmd_adb_poll_off },
+ { PMU_RESET, "REBOOT", pmu_cmd_reset },
+ { PMU_SHUTDOWN, "SHUTDOWN", pmu_cmd_shutdown },
+ { PMU_READ_RTC, "GET RTC", pmu_cmd_get_rtc },
+ { PMU_SET_RTC, "SET RTC", pmu_cmd_set_rtc },
+ { PMU_SYSTEM_READY, "SYSTEM READY", pmu_cmd_system_ready },
+ { PMU_GET_VERSION, "GET VERSION", pmu_cmd_get_version },
+ { PMU_POWER_EVENTS, "POWER EVENTS", pmu_cmd_power_events },
+ { PMU_GET_COVER, "GET_COVER", pmu_cmd_get_cover },
+ { PMU_DOWNLOAD_STATUS, "DOWNLOAD STATUS", pmu_cmd_download_status },
+ { PMU_READ_PMU_RAM, "READ PMGR RAM", pmu_cmd_read_pmu_ram },
+};
+
+static void pmu_dispatch_cmd(PMUState *s)
+{
+ unsigned int i;
+
+ /* No response by default */
+ s->cmd_rsp_sz = 0;
+
+ for (i = 0; i < ARRAY_SIZE(PMUCmdHandlers); i++) {
+ const PMUCmdHandler *desc = &PMUCmdHandlers[i];
+
+ if (desc->command != s->cmd) {
+ continue;
+ }
+
+ trace_pmu_dispatch_cmd(desc->name);
+ desc->handler(s, s->cmd_buf, s->cmd_buf_pos,
+ s->cmd_rsp, &s->cmd_rsp_sz);
+
+ if (s->rsplen != -1 && s->rsplen != s->cmd_rsp_sz) {
+ trace_pmu_debug_protocol_string("QEMU internal cmd resp mismatch!");
+ } else {
+ trace_pmu_debug_protocol_resp_size(s->cmd_rsp_sz);
+ }
+
+ return;
+ }
+
+ trace_pmu_dispatch_unknown_cmd(s->cmd);
+
+ /* Manufacture fake response with 0's */
+ if (s->rsplen == -1) {
+ s->cmd_rsp_sz = 0;
+ } else {
+ s->cmd_rsp_sz = s->rsplen;
+ memset(s->cmd_rsp, 0, s->rsplen);
+ }
+}
+
+static void pmu_update(PMUState *s)
+{
+ MOS6522PMUState *mps = &s->mos6522_pmu;
+ MOS6522State *ms = MOS6522(mps);
+ ADBBusState *adb_bus = &s->adb_bus;
+
+ /* Only react to changes in reg B */
+ if (ms->b == s->last_b) {
+ return;
+ }
+ s->last_b = ms->b;
+
+ /* Check the TREQ / TACK state */
+ switch (ms->b & (TREQ | TACK)) {
+ case TREQ:
+ /* This is an ack release, handle it and bail out */
+ ms->b |= TACK;
+ s->last_b = ms->b;
+
+ trace_pmu_debug_protocol_string("handshake: TREQ high, setting TACK");
+ return;
+ case TACK:
+ /* This is a valid request, handle below */
+ break;
+ case TREQ | TACK:
+ /* This is an idle state */
+ return;
+ default:
+ /* Invalid state, log and ignore */
+ trace_pmu_debug_protocol_error(ms->b);
+ return;
+ }
+
+ /* If we wanted to handle commands asynchronously, this is where
+ * we would delay the clearing of TACK until we are ready to send
+ * the response
+ */
+
+ /* We have a request, handshake TACK so we don't stay in
+ * an invalid state. If we were concurrent with the OS we
+ * should only do this after we grabbed the SR but that isn't
+ * a problem here.
+ */
+
+ trace_pmu_debug_protocol_clear_treq(s->cmd_state);
+
+ ms->b &= ~TACK;
+ s->last_b = ms->b;
+
+ /* Act according to state */
+ switch (s->cmd_state) {
+ case pmu_state_idle:
+ if (!(ms->acr & SR_OUT)) {
+ trace_pmu_debug_protocol_string("protocol error! "
+ "state idle, ACR reading");
+ break;
+ }
+
+ s->cmd = ms->sr;
+ via_set_sr_int(s);
+ s->cmdlen = pmu_data_len[s->cmd][0];
+ s->rsplen = pmu_data_len[s->cmd][1];
+ s->cmd_buf_pos = 0;
+ s->cmd_rsp_pos = 0;
+ s->cmd_state = pmu_state_cmd;
+
+ adb_autopoll_block(adb_bus);
+ trace_pmu_debug_protocol_cmd(s->cmd, s->cmdlen, s->rsplen);
+ break;
+
+ case pmu_state_cmd:
+ if (!(ms->acr & SR_OUT)) {
+ trace_pmu_debug_protocol_string("protocol error! "
+ "state cmd, ACR reading");
+ break;
+ }
+
+ if (s->cmdlen == -1) {
+ trace_pmu_debug_protocol_cmdlen(ms->sr);
+
+ s->cmdlen = ms->sr;
+ if (s->cmdlen > sizeof(s->cmd_buf)) {
+ trace_pmu_debug_protocol_cmd_toobig(s->cmdlen);
+ }
+ } else if (s->cmd_buf_pos < sizeof(s->cmd_buf)) {
+ s->cmd_buf[s->cmd_buf_pos++] = ms->sr;
+ }
+
+ via_set_sr_int(s);
+ break;
+
+ case pmu_state_rsp:
+ if (ms->acr & SR_OUT) {
+ trace_pmu_debug_protocol_string("protocol error! "
+ "state resp, ACR writing");
+ break;
+ }
+
+ if (s->rsplen == -1) {
+ trace_pmu_debug_protocol_cmd_send_resp_size(s->cmd_rsp_sz);
+
+ ms->sr = s->cmd_rsp_sz;
+ s->rsplen = s->cmd_rsp_sz;
+ } else if (s->cmd_rsp_pos < s->cmd_rsp_sz) {
+ trace_pmu_debug_protocol_cmd_send_resp(s->cmd_rsp_pos, s->rsplen);
+
+ ms->sr = s->cmd_rsp[s->cmd_rsp_pos++];
+ }
+
+ via_set_sr_int(s);
+ break;
+ }
+
+ /* Check for state completion */
+ if (s->cmd_state == pmu_state_cmd && s->cmdlen == s->cmd_buf_pos) {
+ trace_pmu_debug_protocol_string("Command reception complete, "
+ "dispatching...");
+
+ pmu_dispatch_cmd(s);
+ s->cmd_state = pmu_state_rsp;
+ }
+
+ if (s->cmd_state == pmu_state_rsp && s->rsplen == s->cmd_rsp_pos) {
+ trace_pmu_debug_protocol_cmd_resp_complete(ms->ier);
+
+ adb_autopoll_unblock(adb_bus);
+ s->cmd_state = pmu_state_idle;
+ }
+}
+
+static uint64_t mos6522_pmu_read(void *opaque, hwaddr addr, unsigned size)
+{
+ PMUState *s = opaque;
+ MOS6522PMUState *mps = &s->mos6522_pmu;
+ MOS6522State *ms = MOS6522(mps);
+
+ addr = (addr >> 9) & 0xf;
+ return mos6522_read(ms, addr, size);
+}
+
+static void mos6522_pmu_write(void *opaque, hwaddr addr, uint64_t val,
+ unsigned size)
+{
+ PMUState *s = opaque;
+ MOS6522PMUState *mps = &s->mos6522_pmu;
+ MOS6522State *ms = MOS6522(mps);
+
+ addr = (addr >> 9) & 0xf;
+ mos6522_write(ms, addr, val, size);
+}
+
+static const MemoryRegionOps mos6522_pmu_ops = {
+ .read = mos6522_pmu_read,
+ .write = mos6522_pmu_write,
+ .endianness = DEVICE_BIG_ENDIAN,
+ .impl = {
+ .min_access_size = 1,
+ .max_access_size = 1,
+ },
+};
+
+static bool pmu_adb_state_needed(void *opaque)
+{
+ PMUState *s = opaque;
+
+ return s->has_adb;
+}
+
+static const VMStateDescription vmstate_pmu_adb = {
+ .name = "pmu/adb",
+ .version_id = 1,
+ .minimum_version_id = 1,
+ .needed = pmu_adb_state_needed,
+ .fields = (VMStateField[]) {
+ VMSTATE_UINT8(adb_reply_size, PMUState),
+ VMSTATE_BUFFER(adb_reply, PMUState),
+ VMSTATE_END_OF_LIST()
+ }
+};
+
+static const VMStateDescription vmstate_pmu = {
+ .name = "pmu",
+ .version_id = 1,
+ .minimum_version_id = 1,
+ .fields = (VMStateField[]) {
+ VMSTATE_STRUCT(mos6522_pmu.parent_obj, PMUState, 0, vmstate_mos6522,
+ MOS6522State),
+ VMSTATE_UINT8(last_b, PMUState),
+ VMSTATE_UINT8(cmd, PMUState),
+ VMSTATE_UINT32(cmdlen, PMUState),
+ VMSTATE_UINT32(rsplen, PMUState),
+ VMSTATE_UINT8(cmd_buf_pos, PMUState),
+ VMSTATE_BUFFER(cmd_buf, PMUState),
+ VMSTATE_UINT8(cmd_rsp_pos, PMUState),
+ VMSTATE_UINT8(cmd_rsp_sz, PMUState),
+ VMSTATE_BUFFER(cmd_rsp, PMUState),
+ VMSTATE_UINT8(intbits, PMUState),
+ VMSTATE_UINT8(intmask, PMUState),
+ VMSTATE_UINT32(tick_offset, PMUState),
+ VMSTATE_TIMER_PTR(one_sec_timer, PMUState),
+ VMSTATE_INT64(one_sec_target, PMUState),
+ VMSTATE_END_OF_LIST()
+ },
+ .subsections = (const VMStateDescription * []) {
+ &vmstate_pmu_adb,
+ NULL
+ }
+};
+
+static void pmu_reset(DeviceState *dev)
+{
+ PMUState *s = VIA_PMU(dev);
+
+ /* OpenBIOS needs to do this? MacOS 9 needs it */
+ s->intmask = PMU_INT_ADB | PMU_INT_TICK;
+ s->intbits = 0;
+
+ s->cmd_state = pmu_state_idle;
+}
+
+static void pmu_realize(DeviceState *dev, Error **errp)
+{
+ PMUState *s = VIA_PMU(dev);
+ SysBusDevice *sbd;
+ ADBBusState *adb_bus = &s->adb_bus;
+ struct tm tm;
+
+ if (!sysbus_realize(SYS_BUS_DEVICE(&s->mos6522_pmu), errp)) {
+ return;
+ }
+
+ /* Pass IRQ from 6522 */
+ sbd = SYS_BUS_DEVICE(s);
+ sysbus_pass_irq(sbd, SYS_BUS_DEVICE(&s->mos6522_pmu));
+
+ qemu_get_timedate(&tm, 0);
+ s->tick_offset = (uint32_t)mktimegm(&tm) + RTC_OFFSET;
+ s->one_sec_timer = timer_new_ms(QEMU_CLOCK_VIRTUAL, pmu_one_sec_timer, s);
+ s->one_sec_target = qemu_clock_get_ms(QEMU_CLOCK_VIRTUAL) + 1000;
+ timer_mod(s->one_sec_timer, s->one_sec_target);
+
+ if (s->has_adb) {
+ qbus_init(&s->adb_bus, sizeof(s->adb_bus), TYPE_ADB_BUS,
+ dev, "adb.0");
+ adb_register_autopoll_callback(adb_bus, pmu_adb_poll, s);
+ }
+}
+
+static void pmu_init(Object *obj)
+{
+ SysBusDevice *d = SYS_BUS_DEVICE(obj);
+ PMUState *s = VIA_PMU(obj);
+
+ object_property_add_link(obj, "gpio", TYPE_MACIO_GPIO,
+ (Object **) &s->gpio,
+ qdev_prop_allow_set_link_before_realize,
+ 0);
+
+ object_initialize_child(obj, "mos6522-pmu", &s->mos6522_pmu,
+ TYPE_MOS6522_PMU);
+
+ memory_region_init_io(&s->mem, obj, &mos6522_pmu_ops, s, "via-pmu",
+ 0x2000);
+ sysbus_init_mmio(d, &s->mem);
+}
+
+static Property pmu_properties[] = {
+ DEFINE_PROP_BOOL("has-adb", PMUState, has_adb, true),
+ DEFINE_PROP_END_OF_LIST()
+};
+
+static void pmu_class_init(ObjectClass *oc, void *data)
+{
+ DeviceClass *dc = DEVICE_CLASS(oc);
+
+ dc->realize = pmu_realize;
+ dc->reset = pmu_reset;
+ dc->vmsd = &vmstate_pmu;
+ device_class_set_props(dc, pmu_properties);
+ set_bit(DEVICE_CATEGORY_BRIDGE, dc->categories);
+}
+
+static const TypeInfo pmu_type_info = {
+ .name = TYPE_VIA_PMU,
+ .parent = TYPE_SYS_BUS_DEVICE,
+ .instance_size = sizeof(PMUState),
+ .instance_init = pmu_init,
+ .class_init = pmu_class_init,
+};
+
+static void mos6522_pmu_portB_write(MOS6522State *s)
+{
+ MOS6522PMUState *mps = container_of(s, MOS6522PMUState, parent_obj);
+ PMUState *ps = container_of(mps, PMUState, mos6522_pmu);
+
+ if ((s->pcr & 0xe0) == 0x20 || (s->pcr & 0xe0) == 0x60) {
+ s->ifr &= ~CB2_INT;
+ }
+ s->ifr &= ~CB1_INT;
+
+ via_update_irq(ps);
+ pmu_update(ps);
+}
+
+static void mos6522_pmu_portA_write(MOS6522State *s)
+{
+ MOS6522PMUState *mps = container_of(s, MOS6522PMUState, parent_obj);
+ PMUState *ps = container_of(mps, PMUState, mos6522_pmu);
+
+ if ((s->pcr & 0x0e) == 0x02 || (s->pcr & 0x0e) == 0x06) {
+ s->ifr &= ~CA2_INT;
+ }
+ s->ifr &= ~CA1_INT;
+
+ via_update_irq(ps);
+}
+
+static void mos6522_pmu_reset(DeviceState *dev)
+{
+ MOS6522State *ms = MOS6522(dev);
+ MOS6522PMUState *mps = container_of(ms, MOS6522PMUState, parent_obj);
+ PMUState *s = container_of(mps, PMUState, mos6522_pmu);
+ MOS6522DeviceClass *mdc = MOS6522_GET_CLASS(ms);
+
+ mdc->parent_reset(dev);
+
+ ms->timers[0].frequency = VIA_TIMER_FREQ;
+ ms->timers[1].frequency = (SCALE_US * 6000) / 4700;
+
+ s->last_b = ms->b = TACK | TREQ;
+}
+
+static void mos6522_pmu_class_init(ObjectClass *oc, void *data)
+{
+ DeviceClass *dc = DEVICE_CLASS(oc);
+ MOS6522DeviceClass *mdc = MOS6522_CLASS(oc);
+
+ dc->reset = mos6522_pmu_reset;
+ mdc->portB_write = mos6522_pmu_portB_write;
+ mdc->portA_write = mos6522_pmu_portA_write;
+}
+
+static const TypeInfo mos6522_pmu_type_info = {
+ .name = TYPE_MOS6522_PMU,
+ .parent = TYPE_MOS6522,
+ .instance_size = sizeof(MOS6522PMUState),
+ .class_init = mos6522_pmu_class_init,
+};
+
+static void pmu_register_types(void)
+{
+ type_register_static(&pmu_type_info);
+ type_register_static(&mos6522_pmu_type_info);
+}
+
+type_init(pmu_register_types)