kernel-fxtec-pro1x/arch/arm/mach-omap2/mux.c
Tero Kristo 13a3fe52f7 ARM: OMAP2+: mux: add support for PAD wakeup interrupts
OMAP mux now parses active wakeup events from pad registers and calls
corresponding hwmod ISRs once a wakeup is detected. This is
accomplished by registering an interrupt handler for PRCM IO event,
which is raised every time the HW detects wakeups.

[paul@pwsan.com: This patch is a merge of Govindraj R's "ARM: OMAP2+:
hwmod: Add API to check IO PAD wakeup status" patch, Tero Kristo's
"ARM: OMAP2+: mux: add support for PAD wakeup interrupts" patch, and
part of Tero's "ARM: OMAP: mux: add support for selecting mpu_irq for
each wakeup pad" patch.]

Signed-off-by: Tero Kristo <t-kristo@ti.com>
Cc: Govindraj.R <govindraj.raja@ti.com>
Tested-by: Kevin Hilman <khilman@ti.com>
Reviewed-by: Kevin Hilman <khilman@ti.com>
Acked-by: Tony Lindgren <tony@atomide.com>
[paul@pwsan.com: reduced indentation level; renamed omap_hwmod function;
 improved function documentation; modified to iterate only through dynamic
 pads; modified to skip pads where idle mode doesn't enable wakeups; split
 patches]
Signed-off-by: Paul Walmsley <paul@pwsan.com>
2011-12-16 14:36:59 -07:00

1159 lines
26 KiB
C

/*
* linux/arch/arm/mach-omap2/mux.c
*
* OMAP2, OMAP3 and OMAP4 pin multiplexing configurations
*
* Copyright (C) 2004 - 2010 Texas Instruments Inc.
* Copyright (C) 2003 - 2008 Nokia Corporation
*
* Written by Tony Lindgren
*
* 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, write to the Free Software
* Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
*
*/
#include <linux/kernel.h>
#include <linux/init.h>
#include <linux/io.h>
#include <linux/list.h>
#include <linux/slab.h>
#include <linux/ctype.h>
#include <linux/debugfs.h>
#include <linux/seq_file.h>
#include <linux/uaccess.h>
#include <linux/irq.h>
#include <linux/interrupt.h>
#include <asm/system.h>
#include <plat/omap_hwmod.h>
#include "control.h"
#include "mux.h"
#include "prm.h"
#define OMAP_MUX_BASE_OFFSET 0x30 /* Offset from CTRL_BASE */
#define OMAP_MUX_BASE_SZ 0x5ca
struct omap_mux_entry {
struct omap_mux mux;
struct list_head node;
};
static LIST_HEAD(mux_partitions);
static DEFINE_MUTEX(muxmode_mutex);
struct omap_mux_partition *omap_mux_get(const char *name)
{
struct omap_mux_partition *partition;
list_for_each_entry(partition, &mux_partitions, node) {
if (!strcmp(name, partition->name))
return partition;
}
return NULL;
}
u16 omap_mux_read(struct omap_mux_partition *partition, u16 reg)
{
if (partition->flags & OMAP_MUX_REG_8BIT)
return __raw_readb(partition->base + reg);
else
return __raw_readw(partition->base + reg);
}
void omap_mux_write(struct omap_mux_partition *partition, u16 val,
u16 reg)
{
if (partition->flags & OMAP_MUX_REG_8BIT)
__raw_writeb(val, partition->base + reg);
else
__raw_writew(val, partition->base + reg);
}
void omap_mux_write_array(struct omap_mux_partition *partition,
struct omap_board_mux *board_mux)
{
if (!board_mux)
return;
while (board_mux->reg_offset != OMAP_MUX_TERMINATOR) {
omap_mux_write(partition, board_mux->value,
board_mux->reg_offset);
board_mux++;
}
}
#ifdef CONFIG_OMAP_MUX
static char *omap_mux_options;
static int __init _omap_mux_init_gpio(struct omap_mux_partition *partition,
int gpio, int val)
{
struct omap_mux_entry *e;
struct omap_mux *gpio_mux = NULL;
u16 old_mode;
u16 mux_mode;
int found = 0;
struct list_head *muxmodes = &partition->muxmodes;
if (!gpio)
return -EINVAL;
list_for_each_entry(e, muxmodes, node) {
struct omap_mux *m = &e->mux;
if (gpio == m->gpio) {
gpio_mux = m;
found++;
}
}
if (found == 0) {
pr_err("%s: Could not set gpio%i\n", __func__, gpio);
return -ENODEV;
}
if (found > 1) {
pr_info("%s: Multiple gpio paths (%d) for gpio%i\n", __func__,
found, gpio);
return -EINVAL;
}
old_mode = omap_mux_read(partition, gpio_mux->reg_offset);
mux_mode = val & ~(OMAP_MUX_NR_MODES - 1);
if (partition->flags & OMAP_MUX_GPIO_IN_MODE3)
mux_mode |= OMAP_MUX_MODE3;
else
mux_mode |= OMAP_MUX_MODE4;
pr_debug("%s: Setting signal %s.gpio%i 0x%04x -> 0x%04x\n", __func__,
gpio_mux->muxnames[0], gpio, old_mode, mux_mode);
omap_mux_write(partition, mux_mode, gpio_mux->reg_offset);
return 0;
}
int __init omap_mux_init_gpio(int gpio, int val)
{
struct omap_mux_partition *partition;
int ret;
list_for_each_entry(partition, &mux_partitions, node) {
ret = _omap_mux_init_gpio(partition, gpio, val);
if (!ret)
return ret;
}
return -ENODEV;
}
static int __init _omap_mux_get_by_name(struct omap_mux_partition *partition,
const char *muxname,
struct omap_mux **found_mux)
{
struct omap_mux *mux = NULL;
struct omap_mux_entry *e;
const char *mode_name;
int found = 0, found_mode = 0, mode0_len = 0;
struct list_head *muxmodes = &partition->muxmodes;
mode_name = strchr(muxname, '.');
if (mode_name) {
mode0_len = strlen(muxname) - strlen(mode_name);
mode_name++;
} else {
mode_name = muxname;
}
list_for_each_entry(e, muxmodes, node) {
char *m0_entry;
int i;
mux = &e->mux;
m0_entry = mux->muxnames[0];
/* First check for full name in mode0.muxmode format */
if (mode0_len && strncmp(muxname, m0_entry, mode0_len))
continue;
/* Then check for muxmode only */
for (i = 0; i < OMAP_MUX_NR_MODES; i++) {
char *mode_cur = mux->muxnames[i];
if (!mode_cur)
continue;
if (!strcmp(mode_name, mode_cur)) {
*found_mux = mux;
found++;
found_mode = i;
}
}
}
if (found == 1) {
return found_mode;
}
if (found > 1) {
pr_err("%s: Multiple signal paths (%i) for %s\n", __func__,
found, muxname);
return -EINVAL;
}
pr_err("%s: Could not find signal %s\n", __func__, muxname);
return -ENODEV;
}
static int __init
omap_mux_get_by_name(const char *muxname,
struct omap_mux_partition **found_partition,
struct omap_mux **found_mux)
{
struct omap_mux_partition *partition;
list_for_each_entry(partition, &mux_partitions, node) {
struct omap_mux *mux = NULL;
int mux_mode = _omap_mux_get_by_name(partition, muxname, &mux);
if (mux_mode < 0)
continue;
*found_partition = partition;
*found_mux = mux;
return mux_mode;
}
return -ENODEV;
}
int __init omap_mux_init_signal(const char *muxname, int val)
{
struct omap_mux_partition *partition = NULL;
struct omap_mux *mux = NULL;
u16 old_mode;
int mux_mode;
mux_mode = omap_mux_get_by_name(muxname, &partition, &mux);
if (mux_mode < 0)
return mux_mode;
old_mode = omap_mux_read(partition, mux->reg_offset);
mux_mode |= val;
pr_debug("%s: Setting signal %s 0x%04x -> 0x%04x\n",
__func__, muxname, old_mode, mux_mode);
omap_mux_write(partition, mux_mode, mux->reg_offset);
return 0;
}
struct omap_hwmod_mux_info * __init
omap_hwmod_mux_init(struct omap_device_pad *bpads, int nr_pads)
{
struct omap_hwmod_mux_info *hmux;
int i, nr_pads_dynamic = 0;
if (!bpads || nr_pads < 1)
return NULL;
hmux = kzalloc(sizeof(struct omap_hwmod_mux_info), GFP_KERNEL);
if (!hmux)
goto err1;
hmux->nr_pads = nr_pads;
hmux->pads = kzalloc(sizeof(struct omap_device_pad) *
nr_pads, GFP_KERNEL);
if (!hmux->pads)
goto err2;
for (i = 0; i < hmux->nr_pads; i++) {
struct omap_mux_partition *partition;
struct omap_device_pad *bpad = &bpads[i], *pad = &hmux->pads[i];
struct omap_mux *mux;
int mux_mode;
mux_mode = omap_mux_get_by_name(bpad->name, &partition, &mux);
if (mux_mode < 0)
goto err3;
if (!pad->partition)
pad->partition = partition;
if (!pad->mux)
pad->mux = mux;
pad->name = kzalloc(strlen(bpad->name) + 1, GFP_KERNEL);
if (!pad->name) {
int j;
for (j = i - 1; j >= 0; j--)
kfree(hmux->pads[j].name);
goto err3;
}
strcpy(pad->name, bpad->name);
pad->flags = bpad->flags;
pad->enable = bpad->enable;
pad->idle = bpad->idle;
pad->off = bpad->off;
if (pad->flags &
(OMAP_DEVICE_PAD_REMUX | OMAP_DEVICE_PAD_WAKEUP))
nr_pads_dynamic++;
pr_debug("%s: Initialized %s\n", __func__, pad->name);
}
if (!nr_pads_dynamic)
return hmux;
/*
* Add pads that need dynamic muxing into a separate list
*/
hmux->nr_pads_dynamic = nr_pads_dynamic;
hmux->pads_dynamic = kzalloc(sizeof(struct omap_device_pad *) *
nr_pads_dynamic, GFP_KERNEL);
if (!hmux->pads_dynamic) {
pr_err("%s: Could not allocate dynamic pads\n", __func__);
return hmux;
}
nr_pads_dynamic = 0;
for (i = 0; i < hmux->nr_pads; i++) {
struct omap_device_pad *pad = &hmux->pads[i];
if (pad->flags &
(OMAP_DEVICE_PAD_REMUX | OMAP_DEVICE_PAD_WAKEUP)) {
pr_debug("%s: pad %s tagged dynamic\n",
__func__, pad->name);
hmux->pads_dynamic[nr_pads_dynamic] = pad;
nr_pads_dynamic++;
}
}
return hmux;
err3:
kfree(hmux->pads);
err2:
kfree(hmux);
err1:
pr_err("%s: Could not allocate device mux entry\n", __func__);
return NULL;
}
/**
* omap_hwmod_mux_scan_wakeups - omap hwmod scan wakeup pads
* @hmux: Pads for a hwmod
* @mpu_irqs: MPU irq array for a hwmod
*
* Scans the wakeup status of pads for a single hwmod. If an irq
* array is defined for this mux, the parser will call the registered
* ISRs for corresponding pads, otherwise the parser will stop at the
* first wakeup active pad and return. Returns true if there is a
* pending and non-served wakeup event for the mux, otherwise false.
*/
static bool omap_hwmod_mux_scan_wakeups(struct omap_hwmod_mux_info *hmux,
struct omap_hwmod_irq_info *mpu_irqs)
{
int i, irq;
unsigned int val;
u32 handled_irqs = 0;
for (i = 0; i < hmux->nr_pads_dynamic; i++) {
struct omap_device_pad *pad = hmux->pads_dynamic[i];
if (!(pad->flags & OMAP_DEVICE_PAD_WAKEUP) ||
!(pad->idle & OMAP_WAKEUP_EN))
continue;
val = omap_mux_read(pad->partition, pad->mux->reg_offset);
if (!(val & OMAP_WAKEUP_EVENT))
continue;
if (!hmux->irqs)
return true;
irq = hmux->irqs[i];
/* make sure we only handle each irq once */
if (handled_irqs & 1 << irq)
continue;
handled_irqs |= 1 << irq;
generic_handle_irq(mpu_irqs[irq].irq);
}
return false;
}
/**
* _omap_hwmod_mux_handle_irq - Process wakeup events for a single hwmod
*
* Checks a single hwmod for every wakeup capable pad to see if there is an
* active wakeup event. If this is the case, call the corresponding ISR.
*/
static int _omap_hwmod_mux_handle_irq(struct omap_hwmod *oh, void *data)
{
if (!oh->mux || !oh->mux->enabled)
return 0;
if (omap_hwmod_mux_scan_wakeups(oh->mux, oh->mpu_irqs))
generic_handle_irq(oh->mpu_irqs[0].irq);
return 0;
}
/**
* omap_hwmod_mux_handle_irq - Process pad wakeup irqs.
*
* Calls a function for each registered omap_hwmod to check
* pad wakeup statuses.
*/
static irqreturn_t omap_hwmod_mux_handle_irq(int irq, void *unused)
{
omap_hwmod_for_each(_omap_hwmod_mux_handle_irq, NULL);
return IRQ_HANDLED;
}
/* Assumes the calling function takes care of locking */
void omap_hwmod_mux(struct omap_hwmod_mux_info *hmux, u8 state)
{
int i;
/* Runtime idling of dynamic pads */
if (state == _HWMOD_STATE_IDLE && hmux->enabled) {
for (i = 0; i < hmux->nr_pads_dynamic; i++) {
struct omap_device_pad *pad = hmux->pads_dynamic[i];
int val = -EINVAL;
val = pad->idle;
omap_mux_write(pad->partition, val,
pad->mux->reg_offset);
}
return;
}
/* Runtime enabling of dynamic pads */
if ((state == _HWMOD_STATE_ENABLED) && hmux->pads_dynamic
&& hmux->enabled) {
for (i = 0; i < hmux->nr_pads_dynamic; i++) {
struct omap_device_pad *pad = hmux->pads_dynamic[i];
int val = -EINVAL;
val = pad->enable;
omap_mux_write(pad->partition, val,
pad->mux->reg_offset);
}
return;
}
/* Enabling or disabling of all pads */
for (i = 0; i < hmux->nr_pads; i++) {
struct omap_device_pad *pad = &hmux->pads[i];
int flags, val = -EINVAL;
flags = pad->flags;
switch (state) {
case _HWMOD_STATE_ENABLED:
val = pad->enable;
pr_debug("%s: Enabling %s %x\n", __func__,
pad->name, val);
break;
case _HWMOD_STATE_DISABLED:
/* Use safe mode unless OMAP_DEVICE_PAD_REMUX */
if (flags & OMAP_DEVICE_PAD_REMUX)
val = pad->off;
else
val = OMAP_MUX_MODE7;
pr_debug("%s: Disabling %s %x\n", __func__,
pad->name, val);
break;
default:
/* Nothing to be done */
break;
};
if (val >= 0) {
omap_mux_write(pad->partition, val,
pad->mux->reg_offset);
pad->flags = flags;
}
}
if (state == _HWMOD_STATE_ENABLED)
hmux->enabled = true;
else
hmux->enabled = false;
}
#ifdef CONFIG_DEBUG_FS
#define OMAP_MUX_MAX_NR_FLAGS 10
#define OMAP_MUX_TEST_FLAG(val, mask) \
if (((val) & (mask)) == (mask)) { \
i++; \
flags[i] = #mask; \
}
/* REVISIT: Add checking for non-optimal mux settings */
static inline void omap_mux_decode(struct seq_file *s, u16 val)
{
char *flags[OMAP_MUX_MAX_NR_FLAGS];
char mode[sizeof("OMAP_MUX_MODE") + 1];
int i = -1;
sprintf(mode, "OMAP_MUX_MODE%d", val & 0x7);
i++;
flags[i] = mode;
OMAP_MUX_TEST_FLAG(val, OMAP_PIN_OFF_WAKEUPENABLE);
if (val & OMAP_OFF_EN) {
if (!(val & OMAP_OFFOUT_EN)) {
if (!(val & OMAP_OFF_PULL_UP)) {
OMAP_MUX_TEST_FLAG(val,
OMAP_PIN_OFF_INPUT_PULLDOWN);
} else {
OMAP_MUX_TEST_FLAG(val,
OMAP_PIN_OFF_INPUT_PULLUP);
}
} else {
if (!(val & OMAP_OFFOUT_VAL)) {
OMAP_MUX_TEST_FLAG(val,
OMAP_PIN_OFF_OUTPUT_LOW);
} else {
OMAP_MUX_TEST_FLAG(val,
OMAP_PIN_OFF_OUTPUT_HIGH);
}
}
}
if (val & OMAP_INPUT_EN) {
if (val & OMAP_PULL_ENA) {
if (!(val & OMAP_PULL_UP)) {
OMAP_MUX_TEST_FLAG(val,
OMAP_PIN_INPUT_PULLDOWN);
} else {
OMAP_MUX_TEST_FLAG(val, OMAP_PIN_INPUT_PULLUP);
}
} else {
OMAP_MUX_TEST_FLAG(val, OMAP_PIN_INPUT);
}
} else {
i++;
flags[i] = "OMAP_PIN_OUTPUT";
}
do {
seq_printf(s, "%s", flags[i]);
if (i > 0)
seq_printf(s, " | ");
} while (i-- > 0);
}
#define OMAP_MUX_DEFNAME_LEN 32
static int omap_mux_dbg_board_show(struct seq_file *s, void *unused)
{
struct omap_mux_partition *partition = s->private;
struct omap_mux_entry *e;
u8 omap_gen = omap_rev() >> 28;
list_for_each_entry(e, &partition->muxmodes, node) {
struct omap_mux *m = &e->mux;
char m0_def[OMAP_MUX_DEFNAME_LEN];
char *m0_name = m->muxnames[0];
u16 val;
int i, mode;
if (!m0_name)
continue;
/* REVISIT: Needs to be updated if mode0 names get longer */
for (i = 0; i < OMAP_MUX_DEFNAME_LEN; i++) {
if (m0_name[i] == '\0') {
m0_def[i] = m0_name[i];
break;
}
m0_def[i] = toupper(m0_name[i]);
}
val = omap_mux_read(partition, m->reg_offset);
mode = val & OMAP_MUX_MODE7;
if (mode != 0)
seq_printf(s, "/* %s */\n", m->muxnames[mode]);
/*
* XXX: Might be revisited to support differences across
* same OMAP generation.
*/
seq_printf(s, "OMAP%d_MUX(%s, ", omap_gen, m0_def);
omap_mux_decode(s, val);
seq_printf(s, "),\n");
}
return 0;
}
static int omap_mux_dbg_board_open(struct inode *inode, struct file *file)
{
return single_open(file, omap_mux_dbg_board_show, inode->i_private);
}
static const struct file_operations omap_mux_dbg_board_fops = {
.open = omap_mux_dbg_board_open,
.read = seq_read,
.llseek = seq_lseek,
.release = single_release,
};
static struct omap_mux_partition *omap_mux_get_partition(struct omap_mux *mux)
{
struct omap_mux_partition *partition;
list_for_each_entry(partition, &mux_partitions, node) {
struct list_head *muxmodes = &partition->muxmodes;
struct omap_mux_entry *e;
list_for_each_entry(e, muxmodes, node) {
struct omap_mux *m = &e->mux;
if (m == mux)
return partition;
}
}
return NULL;
}
static int omap_mux_dbg_signal_show(struct seq_file *s, void *unused)
{
struct omap_mux *m = s->private;
struct omap_mux_partition *partition;
const char *none = "NA";
u16 val;
int mode;
partition = omap_mux_get_partition(m);
if (!partition)
return 0;
val = omap_mux_read(partition, m->reg_offset);
mode = val & OMAP_MUX_MODE7;
seq_printf(s, "name: %s.%s (0x%08x/0x%03x = 0x%04x), b %s, t %s\n",
m->muxnames[0], m->muxnames[mode],
partition->phys + m->reg_offset, m->reg_offset, val,
m->balls[0] ? m->balls[0] : none,
m->balls[1] ? m->balls[1] : none);
seq_printf(s, "mode: ");
omap_mux_decode(s, val);
seq_printf(s, "\n");
seq_printf(s, "signals: %s | %s | %s | %s | %s | %s | %s | %s\n",
m->muxnames[0] ? m->muxnames[0] : none,
m->muxnames[1] ? m->muxnames[1] : none,
m->muxnames[2] ? m->muxnames[2] : none,
m->muxnames[3] ? m->muxnames[3] : none,
m->muxnames[4] ? m->muxnames[4] : none,
m->muxnames[5] ? m->muxnames[5] : none,
m->muxnames[6] ? m->muxnames[6] : none,
m->muxnames[7] ? m->muxnames[7] : none);
return 0;
}
#define OMAP_MUX_MAX_ARG_CHAR 7
static ssize_t omap_mux_dbg_signal_write(struct file *file,
const char __user *user_buf,
size_t count, loff_t *ppos)
{
char buf[OMAP_MUX_MAX_ARG_CHAR];
struct seq_file *seqf;
struct omap_mux *m;
unsigned long val;
int buf_size, ret;
struct omap_mux_partition *partition;
if (count > OMAP_MUX_MAX_ARG_CHAR)
return -EINVAL;
memset(buf, 0, sizeof(buf));
buf_size = min(count, sizeof(buf) - 1);
if (copy_from_user(buf, user_buf, buf_size))
return -EFAULT;
ret = strict_strtoul(buf, 0x10, &val);
if (ret < 0)
return ret;
if (val > 0xffff)
return -EINVAL;
seqf = file->private_data;
m = seqf->private;
partition = omap_mux_get_partition(m);
if (!partition)
return -ENODEV;
omap_mux_write(partition, (u16)val, m->reg_offset);
*ppos += count;
return count;
}
static int omap_mux_dbg_signal_open(struct inode *inode, struct file *file)
{
return single_open(file, omap_mux_dbg_signal_show, inode->i_private);
}
static const struct file_operations omap_mux_dbg_signal_fops = {
.open = omap_mux_dbg_signal_open,
.read = seq_read,
.write = omap_mux_dbg_signal_write,
.llseek = seq_lseek,
.release = single_release,
};
static struct dentry *mux_dbg_dir;
static void __init omap_mux_dbg_create_entry(
struct omap_mux_partition *partition,
struct dentry *mux_dbg_dir)
{
struct omap_mux_entry *e;
list_for_each_entry(e, &partition->muxmodes, node) {
struct omap_mux *m = &e->mux;
(void)debugfs_create_file(m->muxnames[0], S_IWUSR, mux_dbg_dir,
m, &omap_mux_dbg_signal_fops);
}
}
static void __init omap_mux_dbg_init(void)
{
struct omap_mux_partition *partition;
static struct dentry *mux_dbg_board_dir;
mux_dbg_dir = debugfs_create_dir("omap_mux", NULL);
if (!mux_dbg_dir)
return;
mux_dbg_board_dir = debugfs_create_dir("board", mux_dbg_dir);
if (!mux_dbg_board_dir)
return;
list_for_each_entry(partition, &mux_partitions, node) {
omap_mux_dbg_create_entry(partition, mux_dbg_dir);
(void)debugfs_create_file(partition->name, S_IRUGO,
mux_dbg_board_dir, partition,
&omap_mux_dbg_board_fops);
}
}
#else
static inline void omap_mux_dbg_init(void)
{
}
#endif /* CONFIG_DEBUG_FS */
static void __init omap_mux_free_names(struct omap_mux *m)
{
int i;
for (i = 0; i < OMAP_MUX_NR_MODES; i++)
kfree(m->muxnames[i]);
#ifdef CONFIG_DEBUG_FS
for (i = 0; i < OMAP_MUX_NR_SIDES; i++)
kfree(m->balls[i]);
#endif
}
/* Free all data except for GPIO pins unless CONFIG_DEBUG_FS is set */
static int __init omap_mux_late_init(void)
{
struct omap_mux_partition *partition;
int ret;
list_for_each_entry(partition, &mux_partitions, node) {
struct omap_mux_entry *e, *tmp;
list_for_each_entry_safe(e, tmp, &partition->muxmodes, node) {
struct omap_mux *m = &e->mux;
u16 mode = omap_mux_read(partition, m->reg_offset);
if (OMAP_MODE_GPIO(mode))
continue;
#ifndef CONFIG_DEBUG_FS
mutex_lock(&muxmode_mutex);
list_del(&e->node);
mutex_unlock(&muxmode_mutex);
omap_mux_free_names(m);
kfree(m);
#endif
}
}
ret = request_irq(omap_prcm_event_to_irq("io"),
omap_hwmod_mux_handle_irq, IRQF_SHARED | IRQF_NO_SUSPEND,
"hwmod_io", omap_mux_late_init);
if (ret)
pr_warning("mux: Failed to setup hwmod io irq %d\n", ret);
omap_mux_dbg_init();
return 0;
}
late_initcall(omap_mux_late_init);
static void __init omap_mux_package_fixup(struct omap_mux *p,
struct omap_mux *superset)
{
while (p->reg_offset != OMAP_MUX_TERMINATOR) {
struct omap_mux *s = superset;
int found = 0;
while (s->reg_offset != OMAP_MUX_TERMINATOR) {
if (s->reg_offset == p->reg_offset) {
*s = *p;
found++;
break;
}
s++;
}
if (!found)
pr_err("%s: Unknown entry offset 0x%x\n", __func__,
p->reg_offset);
p++;
}
}
#ifdef CONFIG_DEBUG_FS
static void __init omap_mux_package_init_balls(struct omap_ball *b,
struct omap_mux *superset)
{
while (b->reg_offset != OMAP_MUX_TERMINATOR) {
struct omap_mux *s = superset;
int found = 0;
while (s->reg_offset != OMAP_MUX_TERMINATOR) {
if (s->reg_offset == b->reg_offset) {
s->balls[0] = b->balls[0];
s->balls[1] = b->balls[1];
found++;
break;
}
s++;
}
if (!found)
pr_err("%s: Unknown ball offset 0x%x\n", __func__,
b->reg_offset);
b++;
}
}
#else /* CONFIG_DEBUG_FS */
static inline void omap_mux_package_init_balls(struct omap_ball *b,
struct omap_mux *superset)
{
}
#endif /* CONFIG_DEBUG_FS */
static int __init omap_mux_setup(char *options)
{
if (!options)
return 0;
omap_mux_options = options;
return 1;
}
__setup("omap_mux=", omap_mux_setup);
/*
* Note that the omap_mux=some.signal1=0x1234,some.signal2=0x1234
* cmdline options only override the bootloader values.
* During development, please enable CONFIG_DEBUG_FS, and use the
* signal specific entries under debugfs.
*/
static void __init omap_mux_set_cmdline_signals(void)
{
char *options, *next_opt, *token;
if (!omap_mux_options)
return;
options = kstrdup(omap_mux_options, GFP_KERNEL);
if (!options)
return;
next_opt = options;
while ((token = strsep(&next_opt, ",")) != NULL) {
char *keyval, *name;
unsigned long val;
keyval = token;
name = strsep(&keyval, "=");
if (name) {
int res;
res = strict_strtoul(keyval, 0x10, &val);
if (res < 0)
continue;
omap_mux_init_signal(name, (u16)val);
}
}
kfree(options);
}
static int __init omap_mux_copy_names(struct omap_mux *src,
struct omap_mux *dst)
{
int i;
for (i = 0; i < OMAP_MUX_NR_MODES; i++) {
if (src->muxnames[i]) {
dst->muxnames[i] = kstrdup(src->muxnames[i],
GFP_KERNEL);
if (!dst->muxnames[i])
goto free;
}
}
#ifdef CONFIG_DEBUG_FS
for (i = 0; i < OMAP_MUX_NR_SIDES; i++) {
if (src->balls[i]) {
dst->balls[i] = kstrdup(src->balls[i], GFP_KERNEL);
if (!dst->balls[i])
goto free;
}
}
#endif
return 0;
free:
omap_mux_free_names(dst);
return -ENOMEM;
}
#endif /* CONFIG_OMAP_MUX */
static struct omap_mux *omap_mux_get_by_gpio(
struct omap_mux_partition *partition,
int gpio)
{
struct omap_mux_entry *e;
struct omap_mux *ret = NULL;
list_for_each_entry(e, &partition->muxmodes, node) {
struct omap_mux *m = &e->mux;
if (m->gpio == gpio) {
ret = m;
break;
}
}
return ret;
}
/* Needed for dynamic muxing of GPIO pins for off-idle */
u16 omap_mux_get_gpio(int gpio)
{
struct omap_mux_partition *partition;
struct omap_mux *m = NULL;
list_for_each_entry(partition, &mux_partitions, node) {
m = omap_mux_get_by_gpio(partition, gpio);
if (m)
return omap_mux_read(partition, m->reg_offset);
}
if (!m || m->reg_offset == OMAP_MUX_TERMINATOR)
pr_err("%s: Could not get gpio%i\n", __func__, gpio);
return OMAP_MUX_TERMINATOR;
}
/* Needed for dynamic muxing of GPIO pins for off-idle */
void omap_mux_set_gpio(u16 val, int gpio)
{
struct omap_mux_partition *partition;
struct omap_mux *m = NULL;
list_for_each_entry(partition, &mux_partitions, node) {
m = omap_mux_get_by_gpio(partition, gpio);
if (m) {
omap_mux_write(partition, val, m->reg_offset);
return;
}
}
if (!m || m->reg_offset == OMAP_MUX_TERMINATOR)
pr_err("%s: Could not set gpio%i\n", __func__, gpio);
}
static struct omap_mux * __init omap_mux_list_add(
struct omap_mux_partition *partition,
struct omap_mux *src)
{
struct omap_mux_entry *entry;
struct omap_mux *m;
entry = kzalloc(sizeof(struct omap_mux_entry), GFP_KERNEL);
if (!entry)
return NULL;
m = &entry->mux;
entry->mux = *src;
#ifdef CONFIG_OMAP_MUX
if (omap_mux_copy_names(src, m)) {
kfree(entry);
return NULL;
}
#endif
mutex_lock(&muxmode_mutex);
list_add_tail(&entry->node, &partition->muxmodes);
mutex_unlock(&muxmode_mutex);
return m;
}
/*
* Note if CONFIG_OMAP_MUX is not selected, we will only initialize
* the GPIO to mux offset mapping that is needed for dynamic muxing
* of GPIO pins for off-idle.
*/
static void __init omap_mux_init_list(struct omap_mux_partition *partition,
struct omap_mux *superset)
{
while (superset->reg_offset != OMAP_MUX_TERMINATOR) {
struct omap_mux *entry;
#ifdef CONFIG_OMAP_MUX
if (!superset->muxnames || !superset->muxnames[0]) {
superset++;
continue;
}
#else
/* Skip pins that are not muxed as GPIO by bootloader */
if (!OMAP_MODE_GPIO(omap_mux_read(partition,
superset->reg_offset))) {
superset++;
continue;
}
#endif
entry = omap_mux_list_add(partition, superset);
if (!entry) {
pr_err("%s: Could not add entry\n", __func__);
return;
}
superset++;
}
}
#ifdef CONFIG_OMAP_MUX
static void omap_mux_init_package(struct omap_mux *superset,
struct omap_mux *package_subset,
struct omap_ball *package_balls)
{
if (package_subset)
omap_mux_package_fixup(package_subset, superset);
if (package_balls)
omap_mux_package_init_balls(package_balls, superset);
}
static void omap_mux_init_signals(struct omap_mux_partition *partition,
struct omap_board_mux *board_mux)
{
omap_mux_set_cmdline_signals();
omap_mux_write_array(partition, board_mux);
}
#else
static void omap_mux_init_package(struct omap_mux *superset,
struct omap_mux *package_subset,
struct omap_ball *package_balls)
{
}
static void omap_mux_init_signals(struct omap_mux_partition *partition,
struct omap_board_mux *board_mux)
{
}
#endif
static u32 mux_partitions_cnt;
int __init omap_mux_init(const char *name, u32 flags,
u32 mux_pbase, u32 mux_size,
struct omap_mux *superset,
struct omap_mux *package_subset,
struct omap_board_mux *board_mux,
struct omap_ball *package_balls)
{
struct omap_mux_partition *partition;
partition = kzalloc(sizeof(struct omap_mux_partition), GFP_KERNEL);
if (!partition)
return -ENOMEM;
partition->name = name;
partition->flags = flags;
partition->size = mux_size;
partition->phys = mux_pbase;
partition->base = ioremap(mux_pbase, mux_size);
if (!partition->base) {
pr_err("%s: Could not ioremap mux partition at 0x%08x\n",
__func__, partition->phys);
kfree(partition);
return -ENODEV;
}
INIT_LIST_HEAD(&partition->muxmodes);
list_add_tail(&partition->node, &mux_partitions);
mux_partitions_cnt++;
pr_info("%s: Add partition: #%d: %s, flags: %x\n", __func__,
mux_partitions_cnt, partition->name, partition->flags);
omap_mux_init_package(superset, package_subset, package_balls);
omap_mux_init_list(partition, superset);
omap_mux_init_signals(partition, board_mux);
return 0;
}