[MIPS] APM emu support
authorRodolfo Giometti <giometti@linux.it>
Mon, 5 Jun 2006 15:43:10 +0000 (17:43 +0200)
committerRalf Baechle <ralf@linux-mips.org>
Mon, 19 Jun 2006 16:39:22 +0000 (17:39 +0100)
Signed-off-by: Rodolfo Giometti <giometti@linux.it>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
arch/mips/Kconfig
arch/mips/kernel/Makefile
arch/mips/kernel/apm.c [new file with mode: 0644]
include/asm-mips/apm.h [new file with mode: 0644]

index 1c78661..ae33e0a 100644 (file)
@@ -1890,6 +1890,32 @@ config PM
        bool "Power Management support (EXPERIMENTAL)"
        depends on EXPERIMENTAL && SOC_AU1X00
 
+config APM
+        tristate "Advanced Power Management Emulation"
+       depends on PM
+        ---help---
+         APM is a BIOS specification for saving power using several different
+         techniques. This is mostly useful for battery powered systems with
+         APM compliant BIOSes. If you say Y here, the system time will be
+         reset after a RESUME operation, the /proc/apm device will provide
+         battery status information, and user-space programs will receive
+         notification of APM "events" (e.g. battery status change).
+
+         In order to use APM, you will need supporting software. For location
+         and more information, read <file:Documentation/pm.txt> and the
+         Battery Powered Linux mini-HOWTO, available from
+         <http://www.tldp.org/docs.html#howto>.
+
+         This driver does not spin down disk drives (see the hdparm(8)
+         manpage ("man 8 hdparm") for that), and it doesn't turn off
+         VESA-compliant "green" monitors.
+
+         Generally, if you don't have a battery in your machine, there isn't
+         much point in using this driver and you should say N. If you get
+         random kernel OOPSes or reboots that don't seem to be related to
+         anything, try disabling/enabling this option (or disabling/enabling
+         APM in your BIOS).
+
 endmenu
 
 source "net/Kconfig"
index 34e8a25..881c467 100644 (file)
@@ -13,6 +13,8 @@ binfmt_irix-objs      := irixelf.o irixinv.o irixioctl.o irixsig.o    \
 
 obj-$(CONFIG_MODULES)          += mips_ksyms.o module.o
 
+obj-$(CONFIG_APM)              += apm.o
+
 obj-$(CONFIG_CPU_R3000)                += r2300_fpu.o r2300_switch.o
 obj-$(CONFIG_CPU_TX39XX)       += r2300_fpu.o r2300_switch.o
 obj-$(CONFIG_CPU_TX49XX)       += r4k_fpu.o r4k_switch.o
diff --git a/arch/mips/kernel/apm.c b/arch/mips/kernel/apm.c
new file mode 100644 (file)
index 0000000..15f46b4
--- /dev/null
@@ -0,0 +1,605 @@
+/*
+ * bios-less APM driver for MIPS Linux
+ *  Jamey Hicks <jamey@crl.dec.com>
+ *  adapted from the APM BIOS driver for Linux by Stephen Rothwell (sfr@linuxcare.com)
+ *
+ * APM 1.2 Reference:
+ *   Intel Corporation, Microsoft Corporation. Advanced Power Management
+ *   (APM) BIOS Interface Specification, Revision 1.2, February 1996.
+ *
+ * [This document is available from Microsoft at:
+ *    http://www.microsoft.com/hwdev/busbios/amp_12.htm]
+ */
+#include <linux/config.h>
+#include <linux/module.h>
+#include <linux/poll.h>
+#include <linux/timer.h>
+#include <linux/slab.h>
+#include <linux/proc_fs.h>
+#include <linux/miscdevice.h>
+#include <linux/apm_bios.h>
+#include <linux/capability.h>
+#include <linux/sched.h>
+#include <linux/pm.h>
+#include <linux/device.h>
+#include <linux/kernel.h>
+#include <linux/list.h>
+#include <linux/init.h>
+#include <linux/completion.h>
+
+#include <asm/apm.h> /* apm_power_info */
+#include <asm/system.h>
+
+/*
+ * The apm_bios device is one of the misc char devices.
+ * This is its minor number.
+ */
+#define APM_MINOR_DEV  134
+
+/*
+ * See Documentation/Config.help for the configuration options.
+ *
+ * Various options can be changed at boot time as follows:
+ * (We allow underscores for compatibility with the modules code)
+ *     apm=on/off                      enable/disable APM
+ */
+
+/*
+ * Maximum number of events stored
+ */
+#define APM_MAX_EVENTS         16
+
+struct apm_queue {
+       unsigned int            event_head;
+       unsigned int            event_tail;
+       apm_event_t             events[APM_MAX_EVENTS];
+};
+
+/*
+ * The per-file APM data
+ */
+struct apm_user {
+       struct list_head        list;
+
+       unsigned int            suser: 1;
+       unsigned int            writer: 1;
+       unsigned int            reader: 1;
+
+       int                     suspend_result;
+       unsigned int            suspend_state;
+#define SUSPEND_NONE   0               /* no suspend pending */
+#define SUSPEND_PENDING        1               /* suspend pending read */
+#define SUSPEND_READ   2               /* suspend read, pending ack */
+#define SUSPEND_ACKED  3               /* suspend acked */
+#define SUSPEND_DONE   4               /* suspend completed */
+
+       struct apm_queue        queue;
+};
+
+/*
+ * Local variables
+ */
+static int suspends_pending;
+static int apm_disabled;
+static int mips_apm_active;
+
+static DECLARE_WAIT_QUEUE_HEAD(apm_waitqueue);
+static DECLARE_WAIT_QUEUE_HEAD(apm_suspend_waitqueue);
+
+/*
+ * This is a list of everyone who has opened /dev/apm_bios
+ */
+static DECLARE_RWSEM(user_list_lock);
+static LIST_HEAD(apm_user_list);
+
+/*
+ * kapmd info.  kapmd provides us a process context to handle
+ * "APM" events within - specifically necessary if we're going
+ * to be suspending the system.
+ */
+static DECLARE_WAIT_QUEUE_HEAD(kapmd_wait);
+static DECLARE_COMPLETION(kapmd_exit);
+static DEFINE_SPINLOCK(kapmd_queue_lock);
+static struct apm_queue kapmd_queue;
+
+
+static const char driver_version[] = "1.13";   /* no spaces */
+
+
+
+/*
+ * Compatibility cruft until the IPAQ people move over to the new
+ * interface.
+ */
+static void __apm_get_power_status(struct apm_power_info *info)
+{
+}
+
+/*
+ * This allows machines to provide their own "apm get power status" function.
+ */
+void (*apm_get_power_status)(struct apm_power_info *) = __apm_get_power_status;
+EXPORT_SYMBOL(apm_get_power_status);
+
+
+/*
+ * APM event queue management.
+ */
+static inline int queue_empty(struct apm_queue *q)
+{
+       return q->event_head == q->event_tail;
+}
+
+static inline apm_event_t queue_get_event(struct apm_queue *q)
+{
+       q->event_tail = (q->event_tail + 1) % APM_MAX_EVENTS;
+       return q->events[q->event_tail];
+}
+
+static void queue_add_event(struct apm_queue *q, apm_event_t event)
+{
+       q->event_head = (q->event_head + 1) % APM_MAX_EVENTS;
+       if (q->event_head == q->event_tail) {
+               static int notified;
+
+               if (notified++ == 0)
+                   printk(KERN_ERR "apm: an event queue overflowed\n");
+               q->event_tail = (q->event_tail + 1) % APM_MAX_EVENTS;
+       }
+       q->events[q->event_head] = event;
+}
+
+static void queue_event_one_user(struct apm_user *as, apm_event_t event)
+{
+       if (as->suser && as->writer) {
+               switch (event) {
+               case APM_SYS_SUSPEND:
+               case APM_USER_SUSPEND:
+                       /*
+                        * If this user already has a suspend pending,
+                        * don't queue another one.
+                        */
+                       if (as->suspend_state != SUSPEND_NONE)
+                               return;
+
+                       as->suspend_state = SUSPEND_PENDING;
+                       suspends_pending++;
+                       break;
+               }
+       }
+       queue_add_event(&as->queue, event);
+}
+
+static void queue_event(apm_event_t event, struct apm_user *sender)
+{
+       struct apm_user *as;
+
+       down_read(&user_list_lock);
+       list_for_each_entry(as, &apm_user_list, list) {
+               if (as != sender && as->reader)
+                       queue_event_one_user(as, event);
+       }
+       up_read(&user_list_lock);
+       wake_up_interruptible(&apm_waitqueue);
+}
+
+static void apm_suspend(void)
+{
+       struct apm_user *as;
+       int err = pm_suspend(PM_SUSPEND_MEM);
+
+       /*
+        * Anyone on the APM queues will think we're still suspended.
+        * Send a message so everyone knows we're now awake again.
+        */
+       queue_event(APM_NORMAL_RESUME, NULL);
+
+       /*
+        * Finally, wake up anyone who is sleeping on the suspend.
+        */
+       down_read(&user_list_lock);
+       list_for_each_entry(as, &apm_user_list, list) {
+               as->suspend_result = err;
+               as->suspend_state = SUSPEND_DONE;
+       }
+       up_read(&user_list_lock);
+
+       wake_up(&apm_suspend_waitqueue);
+}
+
+static ssize_t apm_read(struct file *fp, char __user *buf, size_t count, loff_t *ppos)
+{
+       struct apm_user *as = fp->private_data;
+       apm_event_t event;
+       int i = count, ret = 0;
+
+       if (count < sizeof(apm_event_t))
+               return -EINVAL;
+
+       if (queue_empty(&as->queue) && fp->f_flags & O_NONBLOCK)
+               return -EAGAIN;
+
+       wait_event_interruptible(apm_waitqueue, !queue_empty(&as->queue));
+
+       while ((i >= sizeof(event)) && !queue_empty(&as->queue)) {
+               event = queue_get_event(&as->queue);
+
+               ret = -EFAULT;
+               if (copy_to_user(buf, &event, sizeof(event)))
+                       break;
+
+               if (event == APM_SYS_SUSPEND || event == APM_USER_SUSPEND)
+                       as->suspend_state = SUSPEND_READ;
+
+               buf += sizeof(event);
+               i -= sizeof(event);
+       }
+
+       if (i < count)
+               ret = count - i;
+
+       return ret;
+}
+
+static unsigned int apm_poll(struct file *fp, poll_table * wait)
+{
+       struct apm_user *as = fp->private_data;
+
+       poll_wait(fp, &apm_waitqueue, wait);
+       return queue_empty(&as->queue) ? 0 : POLLIN | POLLRDNORM;
+}
+
+/*
+ * apm_ioctl - handle APM ioctl
+ *
+ * APM_IOC_SUSPEND
+ *   This IOCTL is overloaded, and performs two functions.  It is used to:
+ *     - initiate a suspend
+ *     - acknowledge a suspend read from /dev/apm_bios.
+ *   Only when everyone who has opened /dev/apm_bios with write permission
+ *   has acknowledge does the actual suspend happen.
+ */
+static int
+apm_ioctl(struct inode * inode, struct file *filp, u_int cmd, u_long arg)
+{
+       struct apm_user *as = filp->private_data;
+       unsigned long flags;
+       int err = -EINVAL;
+
+       if (!as->suser || !as->writer)
+               return -EPERM;
+
+       switch (cmd) {
+       case APM_IOC_SUSPEND:
+               as->suspend_result = -EINTR;
+
+               if (as->suspend_state == SUSPEND_READ) {
+                       /*
+                        * If we read a suspend command from /dev/apm_bios,
+                        * then the corresponding APM_IOC_SUSPEND ioctl is
+                        * interpreted as an acknowledge.
+                        */
+                       as->suspend_state = SUSPEND_ACKED;
+                       suspends_pending--;
+               } else {
+                       /*
+                        * Otherwise it is a request to suspend the system.
+                        * Queue an event for all readers, and expect an
+                        * acknowledge from all writers who haven't already
+                        * acknowledged.
+                        */
+                       queue_event(APM_USER_SUSPEND, as);
+               }
+
+               /*
+                * If there are no further acknowledges required, suspend
+                * the system.
+                */
+               if (suspends_pending == 0)
+                       apm_suspend();
+
+               /*
+                * Wait for the suspend/resume to complete.  If there are
+                * pending acknowledges, we wait here for them.
+                *
+                * Note that we need to ensure that the PM subsystem does
+                * not kick us out of the wait when it suspends the threads.
+                */
+               flags = current->flags;
+               current->flags |= PF_NOFREEZE;
+
+               /*
+                * Note: do not allow a thread which is acking the suspend
+                * to escape until the resume is complete.
+                */
+               if (as->suspend_state == SUSPEND_ACKED)
+                       wait_event(apm_suspend_waitqueue,
+                                        as->suspend_state == SUSPEND_DONE);
+               else
+                       wait_event_interruptible(apm_suspend_waitqueue,
+                                        as->suspend_state == SUSPEND_DONE);
+
+               current->flags = flags;
+               err = as->suspend_result;
+               as->suspend_state = SUSPEND_NONE;
+               break;
+       }
+
+       return err;
+}
+
+static int apm_release(struct inode * inode, struct file * filp)
+{
+       struct apm_user *as = filp->private_data;
+       filp->private_data = NULL;
+
+       down_write(&user_list_lock);
+       list_del(&as->list);
+       up_write(&user_list_lock);
+
+       /*
+        * We are now unhooked from the chain.  As far as new
+        * events are concerned, we no longer exist.  However, we
+        * need to balance suspends_pending, which means the
+        * possibility of sleeping.
+        */
+       if (as->suspend_state != SUSPEND_NONE) {
+               suspends_pending -= 1;
+               if (suspends_pending == 0)
+                       apm_suspend();
+       }
+
+       kfree(as);
+       return 0;
+}
+
+static int apm_open(struct inode * inode, struct file * filp)
+{
+       struct apm_user *as;
+
+       as = (struct apm_user *)kzalloc(sizeof(*as), GFP_KERNEL);
+       if (as) {
+               /*
+                * XXX - this is a tiny bit broken, when we consider BSD
+                * process accounting. If the device is opened by root, we
+                * instantly flag that we used superuser privs. Who knows,
+                * we might close the device immediately without doing a
+                * privileged operation -- cevans
+                */
+               as->suser = capable(CAP_SYS_ADMIN);
+               as->writer = (filp->f_mode & FMODE_WRITE) == FMODE_WRITE;
+               as->reader = (filp->f_mode & FMODE_READ) == FMODE_READ;
+
+               down_write(&user_list_lock);
+               list_add(&as->list, &apm_user_list);
+               up_write(&user_list_lock);
+
+               filp->private_data = as;
+       }
+
+       return as ? 0 : -ENOMEM;
+}
+
+static struct file_operations apm_bios_fops = {
+       .owner          = THIS_MODULE,
+       .read           = apm_read,
+       .poll           = apm_poll,
+       .ioctl          = apm_ioctl,
+       .open           = apm_open,
+       .release        = apm_release,
+};
+
+static struct miscdevice apm_device = {
+       .minor          = APM_MINOR_DEV,
+       .name           = "apm_bios",
+       .fops           = &apm_bios_fops
+};
+
+
+#ifdef CONFIG_PROC_FS
+/*
+ * Arguments, with symbols from linux/apm_bios.h.
+ *
+ *   0) Linux driver version (this will change if format changes)
+ *   1) APM BIOS Version.  Usually 1.0, 1.1 or 1.2.
+ *   2) APM flags from APM Installation Check (0x00):
+ *     bit 0: APM_16_BIT_SUPPORT
+ *     bit 1: APM_32_BIT_SUPPORT
+ *     bit 2: APM_IDLE_SLOWS_CLOCK
+ *     bit 3: APM_BIOS_DISABLED
+ *     bit 4: APM_BIOS_DISENGAGED
+ *   3) AC line status
+ *     0x00: Off-line
+ *     0x01: On-line
+ *     0x02: On backup power (BIOS >= 1.1 only)
+ *     0xff: Unknown
+ *   4) Battery status
+ *     0x00: High
+ *     0x01: Low
+ *     0x02: Critical
+ *     0x03: Charging
+ *     0x04: Selected battery not present (BIOS >= 1.2 only)
+ *     0xff: Unknown
+ *   5) Battery flag
+ *     bit 0: High
+ *     bit 1: Low
+ *     bit 2: Critical
+ *     bit 3: Charging
+ *     bit 7: No system battery
+ *     0xff: Unknown
+ *   6) Remaining battery life (percentage of charge):
+ *     0-100: valid
+ *     -1: Unknown
+ *   7) Remaining battery life (time units):
+ *     Number of remaining minutes or seconds
+ *     -1: Unknown
+ *   8) min = minutes; sec = seconds
+ */
+static int apm_get_info(char *buf, char **start, off_t fpos, int length)
+{
+       struct apm_power_info info;
+       char *units;
+       int ret;
+
+       info.ac_line_status = 0xff;
+       info.battery_status = 0xff;
+       info.battery_flag   = 0xff;
+       info.battery_life   = -1;
+       info.time           = -1;
+       info.units          = -1;
+
+       if (apm_get_power_status)
+               apm_get_power_status(&info);
+
+       switch (info.units) {
+       default:        units = "?";    break;
+       case 0:         units = "min";  break;
+       case 1:         units = "sec";  break;
+       }
+
+       ret = sprintf(buf, "%s 1.2 0x%02x 0x%02x 0x%02x 0x%02x %d%% %d %s\n",
+                    driver_version, APM_32_BIT_SUPPORT,
+                    info.ac_line_status, info.battery_status,
+                    info.battery_flag, info.battery_life,
+                    info.time, units);
+
+       return ret;
+}
+#endif
+
+static int kapmd(void *arg)
+{
+       daemonize("kapmd");
+       current->flags |= PF_NOFREEZE;
+
+       do {
+               apm_event_t event;
+
+               wait_event_interruptible(kapmd_wait,
+                               !queue_empty(&kapmd_queue) || !mips_apm_active);
+
+               if (!mips_apm_active)
+                       break;
+
+               spin_lock_irq(&kapmd_queue_lock);
+               event = 0;
+               if (!queue_empty(&kapmd_queue))
+                       event = queue_get_event(&kapmd_queue);
+               spin_unlock_irq(&kapmd_queue_lock);
+
+               switch (event) {
+               case 0:
+                       break;
+
+               case APM_LOW_BATTERY:
+               case APM_POWER_STATUS_CHANGE:
+                       queue_event(event, NULL);
+                       break;
+
+               case APM_USER_SUSPEND:
+               case APM_SYS_SUSPEND:
+                       queue_event(event, NULL);
+                       if (suspends_pending == 0)
+                               apm_suspend();
+                       break;
+
+               case APM_CRITICAL_SUSPEND:
+                       apm_suspend();
+                       break;
+               }
+       } while (1);
+
+       complete_and_exit(&kapmd_exit, 0);
+}
+
+static int __init apm_init(void)
+{
+       int ret;
+
+       if (apm_disabled) {
+               printk(KERN_NOTICE "apm: disabled on user request.\n");
+               return -ENODEV;
+       }
+
+       mips_apm_active = 1;
+
+       ret = kernel_thread(kapmd, NULL, CLONE_KERNEL);
+       if (ret < 0) {
+               mips_apm_active = 0;
+               return ret;
+       }
+
+#ifdef CONFIG_PROC_FS
+       create_proc_info_entry("apm", 0, NULL, apm_get_info);
+#endif
+
+       ret = misc_register(&apm_device);
+       if (ret != 0) {
+               remove_proc_entry("apm", NULL);
+
+               mips_apm_active = 0;
+               wake_up(&kapmd_wait);
+               wait_for_completion(&kapmd_exit);
+       }
+
+       return ret;
+}
+
+static void __exit apm_exit(void)
+{
+       misc_deregister(&apm_device);
+       remove_proc_entry("apm", NULL);
+
+       mips_apm_active = 0;
+       wake_up(&kapmd_wait);
+       wait_for_completion(&kapmd_exit);
+}
+
+module_init(apm_init);
+module_exit(apm_exit);
+
+MODULE_AUTHOR("Stephen Rothwell");
+MODULE_DESCRIPTION("Advanced Power Management");
+MODULE_LICENSE("GPL");
+
+#ifndef MODULE
+static int __init apm_setup(char *str)
+{
+       while ((str != NULL) && (*str != '\0')) {
+               if (strncmp(str, "off", 3) == 0)
+                       apm_disabled = 1;
+               if (strncmp(str, "on", 2) == 0)
+                       apm_disabled = 0;
+               str = strchr(str, ',');
+               if (str != NULL)
+                       str += strspn(str, ", \t");
+       }
+       return 1;
+}
+
+__setup("apm=", apm_setup);
+#endif
+
+/**
+ * apm_queue_event - queue an APM event for kapmd
+ * @event: APM event
+ *
+ * Queue an APM event for kapmd to process and ultimately take the
+ * appropriate action.  Only a subset of events are handled:
+ *   %APM_LOW_BATTERY
+ *   %APM_POWER_STATUS_CHANGE
+ *   %APM_USER_SUSPEND
+ *   %APM_SYS_SUSPEND
+ *   %APM_CRITICAL_SUSPEND
+ */
+void apm_queue_event(apm_event_t event)
+{
+       unsigned long flags;
+
+       spin_lock_irqsave(&kapmd_queue_lock, flags);
+       queue_add_event(&kapmd_queue, event);
+       spin_unlock_irqrestore(&kapmd_queue_lock, flags);
+
+       wake_up_interruptible(&kapmd_wait);
+}
+EXPORT_SYMBOL(apm_queue_event);
diff --git a/include/asm-mips/apm.h b/include/asm-mips/apm.h
new file mode 100644 (file)
index 0000000..e8c6920
--- /dev/null
@@ -0,0 +1,65 @@
+/* -*- linux-c -*-
+ *
+ * (C) 2003 zecke@handhelds.org
+ *
+ * GPL version 2
+ *
+ * based on arch/arm/kernel/apm.c
+ * factor out the information needed by architectures to provide
+ * apm status
+ *
+ *
+ */
+#ifndef MIPS_ASM_SA1100_APM_H
+#define MIPS_ASM_SA1100_APM_H
+
+#include <linux/config.h>
+#include <linux/apm_bios.h>
+
+/*
+ * This structure gets filled in by the machine specific 'get_power_status'
+ * implementation.  Any fields which are not set default to a safe value.
+ */
+struct apm_power_info {
+       unsigned char   ac_line_status;
+#define APM_AC_OFFLINE                 0
+#define APM_AC_ONLINE                  1
+#define APM_AC_BACKUP                  2
+#define APM_AC_UNKNOWN                 0xff
+
+       unsigned char   battery_status;
+#define APM_BATTERY_STATUS_HIGH                0
+#define APM_BATTERY_STATUS_LOW         1
+#define APM_BATTERY_STATUS_CRITICAL    2
+#define APM_BATTERY_STATUS_CHARGING    3
+#define APM_BATTERY_STATUS_NOT_PRESENT 4
+#define APM_BATTERY_STATUS_UNKNOWN     0xff
+
+       unsigned char   battery_flag;
+#define APM_BATTERY_FLAG_HIGH          (1 << 0)
+#define APM_BATTERY_FLAG_LOW           (1 << 1)
+#define APM_BATTERY_FLAG_CRITICAL      (1 << 2)
+#define APM_BATTERY_FLAG_CHARGING      (1 << 3)
+#define APM_BATTERY_FLAG_NOT_PRESENT   (1 << 7)
+#define APM_BATTERY_FLAG_UNKNOWN       0xff
+
+       int             battery_life;
+       int             time;
+       int             units;
+#define APM_UNITS_MINS                 0
+#define APM_UNITS_SECS                 1
+#define APM_UNITS_UNKNOWN              -1
+
+};
+
+/*
+ * This allows machines to provide their own "apm get power status" function.
+ */
+extern void (*apm_get_power_status)(struct apm_power_info *);
+
+/*
+ * Queue an event (APM_SYS_SUSPEND or APM_CRITICAL_SUSPEND)
+ */
+void apm_queue_event(apm_event_t event);
+
+#endif