You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
 
 
 
kernel_samsung_sm7125/drivers/clk/msm/clock.c

1405 lines
33 KiB

/* Copyright (C) 2007 Google, Inc.
* Copyright (c) 2007-2017, The Linux Foundation. All rights reserved.
*
* This software is licensed under the terms of the GNU General Public
* License version 2, as published by the Free Software Foundation, and
* may be copied, distributed, and modified under those terms.
*
* 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.
*/
#include <linux/kernel.h>
#include <linux/err.h>
#include <linux/spinlock.h>
#include <linux/string.h>
#include <linux/module.h>
#include <linux/clk.h>
#include <linux/clkdev.h>
#include <linux/list.h>
#include <linux/regulator/consumer.h>
#include <linux/mutex.h>
#include <linux/of.h>
#include <linux/clk/msm-clk-provider.h>
#include <linux/of_platform.h>
#include <linux/pm_opp.h>
#include <trace/events/power.h>
#include "clock.h"
struct handoff_clk {
struct list_head list;
struct clk *clk;
};
static LIST_HEAD(handoff_list);
struct handoff_vdd {
struct list_head list;
struct clk_vdd_class *vdd_class;
};
static LIST_HEAD(handoff_vdd_list);
static DEFINE_MUTEX(msm_clock_init_lock);
LIST_HEAD(orphan_clk_list);
static LIST_HEAD(clk_notifier_list);
/* Find the voltage level required for a given rate. */
int find_vdd_level(struct clk *clk, unsigned long rate)
{
int level;
for (level = 0; level < clk->num_fmax; level++)
if (rate <= clk->fmax[level])
break;
if (level == clk->num_fmax) {
pr_err("Rate %lu for %s is greater than highest Fmax\n", rate,
clk->dbg_name);
return -EINVAL;
}
return level;
}
/* Update voltage level given the current votes. */
static int update_vdd(struct clk_vdd_class *vdd_class)
{
int level, rc = 0, i, ignore;
struct regulator **r = vdd_class->regulator;
int *uv = vdd_class->vdd_uv;
int *ua = vdd_class->vdd_ua;
int n_reg = vdd_class->num_regulators;
int cur_lvl = vdd_class->cur_level;
int max_lvl = vdd_class->num_levels - 1;
int cur_base = cur_lvl * n_reg;
int new_base;
/* aggregate votes */
for (level = max_lvl; level > 0; level--)
if (vdd_class->level_votes[level])
break;
if (level == cur_lvl)
return 0;
max_lvl = max_lvl * n_reg;
new_base = level * n_reg;
for (i = 0; i < vdd_class->num_regulators; i++) {
rc = regulator_set_voltage(r[i], uv[new_base + i],
vdd_class->use_max_uV ? INT_MAX : uv[max_lvl + i]);
if (rc)
goto set_voltage_fail;
if (ua) {
rc = regulator_set_load(r[i], ua[new_base + i]);
rc = rc > 0 ? 0 : rc;
if (rc)
goto set_mode_fail;
}
if (cur_lvl == 0 || cur_lvl == vdd_class->num_levels)
rc = regulator_enable(r[i]);
else if (level == 0)
rc = regulator_disable(r[i]);
if (rc)
goto enable_disable_fail;
}
if (vdd_class->set_vdd && !vdd_class->num_regulators)
rc = vdd_class->set_vdd(vdd_class, level);
if (!rc)
vdd_class->cur_level = level;
return rc;
enable_disable_fail:
/*
* set_optimum_mode could use voltage to derive mode. Restore
* previous voltage setting for r[i] first.
*/
if (ua) {
regulator_set_voltage(r[i], uv[cur_base + i],
vdd_class->use_max_uV ? INT_MAX : uv[max_lvl + i]);
regulator_set_load(r[i], ua[cur_base + i]);
}
set_mode_fail:
regulator_set_voltage(r[i], uv[cur_base + i],
vdd_class->use_max_uV ? INT_MAX : uv[max_lvl + i]);
set_voltage_fail:
for (i--; i >= 0; i--) {
regulator_set_voltage(r[i], uv[cur_base + i],
vdd_class->use_max_uV ? INT_MAX : uv[max_lvl + i]);
if (ua)
regulator_set_load(r[i], ua[cur_base + i]);
if (cur_lvl == 0 || cur_lvl == vdd_class->num_levels)
regulator_disable(r[i]);
else if (level == 0)
ignore = regulator_enable(r[i]);
}
return rc;
}
/* Vote for a voltage level. */
int vote_vdd_level(struct clk_vdd_class *vdd_class, int level)
{
int rc;
if (level >= vdd_class->num_levels)
return -EINVAL;
mutex_lock(&vdd_class->lock);
vdd_class->level_votes[level]++;
rc = update_vdd(vdd_class);
if (rc)
vdd_class->level_votes[level]--;
mutex_unlock(&vdd_class->lock);
return rc;
}
/* Remove vote for a voltage level. */
int unvote_vdd_level(struct clk_vdd_class *vdd_class, int level)
{
int rc = 0;
if (level >= vdd_class->num_levels)
return -EINVAL;
mutex_lock(&vdd_class->lock);
if (WARN(!vdd_class->level_votes[level],
"Reference counts are incorrect for %s level %d\n",
vdd_class->class_name, level))
goto out;
vdd_class->level_votes[level]--;
rc = update_vdd(vdd_class);
if (rc)
vdd_class->level_votes[level]++;
out:
mutex_unlock(&vdd_class->lock);
return rc;
}
/* Vote for a voltage level corresponding to a clock's rate. */
static int vote_rate_vdd(struct clk *clk, unsigned long rate)
{
int level;
if (!clk->vdd_class)
return 0;
level = find_vdd_level(clk, rate);
if (level < 0)
return level;
return vote_vdd_level(clk->vdd_class, level);
}
/* Remove vote for a voltage level corresponding to a clock's rate. */
static void unvote_rate_vdd(struct clk *clk, unsigned long rate)
{
int level;
if (!clk->vdd_class)
return;
level = find_vdd_level(clk, rate);
if (level < 0)
return;
unvote_vdd_level(clk->vdd_class, level);
}
/* Check if the rate is within the voltage limits of the clock. */
bool is_rate_valid(struct clk *clk, unsigned long rate)
{
int level;
if (!clk->vdd_class)
return true;
level = find_vdd_level(clk, rate);
return level >= 0;
}
/**
* __clk_pre_reparent() - Set up the new parent before switching to it and
* prevent the enable state of the child clock from changing.
* @c: The child clock that's going to switch parents
* @new: The new parent that the child clock is going to switch to
* @flags: Pointer to scratch space to save spinlock flags
*
* Cannot be called from atomic context.
*
* Use this API to set up the @new parent clock to be able to support the
* current prepare and enable state of the child clock @c. Once the parent is
* set up, the child clock can safely switch to it.
*
* The caller shall grab the prepare_lock of clock @c before calling this API
* and only release it after calling __clk_post_reparent() for clock @c (or
* if this API fails). This is necessary to prevent the prepare state of the
* child clock @c from changing while the reparenting is in progress. Since
* this API takes care of grabbing the enable lock of @c, only atomic
* operation are allowed between calls to __clk_pre_reparent and
* __clk_post_reparent()
*
* The scratch space pointed to by @flags should not be altered before
* calling __clk_post_reparent() for clock @c.
*
* See also: __clk_post_reparent()
*/
int __clk_pre_reparent(struct clk *c, struct clk *new, unsigned long *flags)
{
int rc;
if (c->prepare_count) {
rc = clk_prepare(new);
if (rc)
return rc;
}
spin_lock_irqsave(&c->lock, *flags);
if (c->count) {
rc = clk_enable(new);
if (rc) {
spin_unlock_irqrestore(&c->lock, *flags);
clk_unprepare(new);
return rc;
}
}
return 0;
}
/**
* __clk_post_reparent() - Release requirements on old parent after switching
* away from it and allow changes to the child clock's enable state.
* @c: The child clock that switched parents
* @old: The old parent that the child clock switched away from or the new
* parent of a failed reparent attempt.
* @flags: Pointer to scratch space where spinlock flags were saved
*
* Cannot be called from atomic context.
*
* This API works in tandem with __clk_pre_reparent. Use this API to
* - Remove prepare and enable requirements from the @old parent after
* switching away from it
* - Or, undo the effects of __clk_pre_reparent() after a failed attempt to
* change parents
*
* The caller shall release the prepare_lock of @c that was grabbed before
* calling __clk_pre_reparent() only after this API is called (or if
* __clk_pre_reparent() fails). This is necessary to prevent the prepare
* state of the child clock @c from changing while the reparenting is in
* progress. Since this API releases the enable lock of @c, the limit to
* atomic operations set by __clk_pre_reparent() is no longer present.
*
* The scratch space pointed to by @flags shall not be altered since the call
* to __clk_pre_reparent() for clock @c.
*
* See also: __clk_pre_reparent()
*/
void __clk_post_reparent(struct clk *c, struct clk *old, unsigned long *flags)
{
if (c->count)
clk_disable(old);
spin_unlock_irqrestore(&c->lock, *flags);
if (c->prepare_count)
clk_unprepare(old);
}
int clk_prepare(struct clk *clk)
{
int ret = 0;
struct clk *parent;
if (!clk)
return 0;
if (IS_ERR(clk))
return -EINVAL;
mutex_lock(&clk->prepare_lock);
if (clk->prepare_count == 0) {
parent = clk->parent;
ret = clk_prepare(parent);
if (ret)
goto out;
ret = clk_prepare(clk->depends);
if (ret)
goto err_prepare_depends;
ret = vote_rate_vdd(clk, clk->rate);
if (ret)
goto err_vote_vdd;
if (clk->ops->prepare)
ret = clk->ops->prepare(clk);
if (ret)
goto err_prepare_clock;
}
clk->prepare_count++;
out:
mutex_unlock(&clk->prepare_lock);
return ret;
err_prepare_clock:
unvote_rate_vdd(clk, clk->rate);
err_vote_vdd:
clk_unprepare(clk->depends);
err_prepare_depends:
clk_unprepare(parent);
goto out;
}
EXPORT_SYMBOL(clk_prepare);
/*
* Standard clock functions defined in include/linux/clk.h
*/
int clk_enable(struct clk *clk)
{
int ret = 0;
unsigned long flags;
struct clk *parent;
const char *name;
if (!clk)
return 0;
if (IS_ERR(clk))
return -EINVAL;
name = clk->dbg_name;
spin_lock_irqsave(&clk->lock, flags);
WARN(!clk->prepare_count,
"%s: Don't call enable on unprepared clocks\n", name);
if (clk->count == 0) {
parent = clk->parent;
ret = clk_enable(parent);
if (ret)
goto err_enable_parent;
ret = clk_enable(clk->depends);
if (ret)
goto err_enable_depends;
trace_clock_enable(name, 1, smp_processor_id());
if (clk->ops->enable)
ret = clk->ops->enable(clk);
if (ret)
goto err_enable_clock;
}
clk->count++;
spin_unlock_irqrestore(&clk->lock, flags);
return 0;
err_enable_clock:
clk_disable(clk->depends);
err_enable_depends:
clk_disable(parent);
err_enable_parent:
spin_unlock_irqrestore(&clk->lock, flags);
return ret;
}
EXPORT_SYMBOL(clk_enable);
void clk_disable(struct clk *clk)
{
const char *name;
unsigned long flags;
if (IS_ERR_OR_NULL(clk))
return;
name = clk->dbg_name;
spin_lock_irqsave(&clk->lock, flags);
WARN(!clk->prepare_count,
"%s: Never called prepare or calling disable after unprepare\n",
name);
if (WARN(clk->count == 0, "%s is unbalanced", name))
goto out;
if (clk->count == 1) {
struct clk *parent = clk->parent;
trace_clock_disable(name, 0, smp_processor_id());
if (clk->ops->disable)
clk->ops->disable(clk);
clk_disable(clk->depends);
clk_disable(parent);
}
clk->count--;
out:
spin_unlock_irqrestore(&clk->lock, flags);
}
EXPORT_SYMBOL(clk_disable);
void clk_unprepare(struct clk *clk)
{
const char *name;
if (IS_ERR_OR_NULL(clk))
return;
name = clk->dbg_name;
mutex_lock(&clk->prepare_lock);
if (WARN(!clk->prepare_count, "%s is unbalanced (prepare)", name))
goto out;
if (clk->prepare_count == 1) {
struct clk *parent = clk->parent;
WARN(clk->count,
"%s: Don't call unprepare when the clock is enabled\n",
name);
if (clk->ops->unprepare)
clk->ops->unprepare(clk);
unvote_rate_vdd(clk, clk->rate);
clk_unprepare(clk->depends);
clk_unprepare(parent);
}
clk->prepare_count--;
out:
mutex_unlock(&clk->prepare_lock);
}
EXPORT_SYMBOL(clk_unprepare);
int clk_reset(struct clk *clk, enum clk_reset_action action)
{
if (IS_ERR_OR_NULL(clk))
return -EINVAL;
if (!clk->ops->reset)
return -EINVAL;
return clk->ops->reset(clk, action);
}
EXPORT_SYMBOL(clk_reset);
/**
* __clk_notify - call clk notifier chain
* @clk: struct clk * that is changing rate
* @msg: clk notifier type (see include/linux/clk.h)
* @old_rate: old clk rate
* @new_rate: new clk rate
*
* Triggers a notifier call chain on the clk rate-change notification
* for 'clk'. Passes a pointer to the struct clk and the previous
* and current rates to the notifier callback. Intended to be called by
* internal clock code only. Returns NOTIFY_DONE from the last driver
* called if all went well, or NOTIFY_STOP or NOTIFY_BAD immediately if
* a driver returns that.
*/
static int __clk_notify(struct clk *clk, unsigned long msg,
unsigned long old_rate, unsigned long new_rate)
{
struct msm_clk_notifier *cn;
struct msm_clk_notifier_data cnd;
int ret = NOTIFY_DONE;
cnd.clk = clk;
cnd.old_rate = old_rate;
cnd.new_rate = new_rate;
list_for_each_entry(cn, &clk_notifier_list, node) {
if (cn->clk == clk) {
ret = srcu_notifier_call_chain(&cn->notifier_head, msg,
&cnd);
break;
}
}
return ret;
}
/*
* clk rate change notifiers
*
* Note - The following notifier functionality is a verbatim copy
* of the implementation in the common clock framework, copied here
* until MSM switches to the common clock framework.
*/
/**
* msm_clk_notif_register - add a clk rate change notifier
* @clk: struct clk * to watch
* @nb: struct notifier_block * with callback info
*
* Request notification when clk's rate changes. This uses an SRCU
* notifier because we want it to block and notifier unregistrations are
* uncommon. The callbacks associated with the notifier must not
* re-enter into the clk framework by calling any top-level clk APIs;
* this will cause a nested prepare_lock mutex.
*
* Pre-change notifier callbacks will be passed the current, pre-change
* rate of the clk via struct msm_clk_notifier_data.old_rate. The new,
* post-change rate of the clk is passed via struct
* msm_clk_notifier_data.new_rate.
*
* Post-change notifiers will pass the now-current, post-change rate of
* the clk in both struct msm_clk_notifier_data.old_rate and struct
* msm_clk_notifier_data.new_rate.
*
* Abort-change notifiers are effectively the opposite of pre-change
* notifiers: the original pre-change clk rate is passed in via struct
* msm_clk_notifier_data.new_rate and the failed post-change rate is passed
* in via struct msm_clk_notifier_data.old_rate.
*
* msm_clk_notif_register() must be called from non-atomic context.
* Returns -EINVAL if called with null arguments, -ENOMEM upon
* allocation failure; otherwise, passes along the return value of
* srcu_notifier_chain_register().
*/
int msm_clk_notif_register(struct clk *clk, struct notifier_block *nb)
{
struct msm_clk_notifier *cn;
int ret = -ENOMEM;
if (!clk || !nb)
return -EINVAL;
mutex_lock(&clk->prepare_lock);
/* search the list of notifiers for this clk */
list_for_each_entry(cn, &clk_notifier_list, node)
if (cn->clk == clk)
break;
/* if clk wasn't in the notifier list, allocate new clk_notifier */
if (cn->clk != clk) {
cn = kzalloc(sizeof(struct msm_clk_notifier), GFP_KERNEL);
if (!cn)
goto out;
cn->clk = clk;
srcu_init_notifier_head(&cn->notifier_head);
list_add(&cn->node, &clk_notifier_list);
}
ret = srcu_notifier_chain_register(&cn->notifier_head, nb);
clk->notifier_count++;
out:
mutex_unlock(&clk->prepare_lock);
return ret;
}
/**
* msm_clk_notif_unregister - remove a clk rate change notifier
* @clk: struct clk *
* @nb: struct notifier_block * with callback info
*
* Request no further notification for changes to 'clk' and frees memory
* allocated in msm_clk_notifier_register.
*
* Returns -EINVAL if called with null arguments; otherwise, passes
* along the return value of srcu_notifier_chain_unregister().
*/
int msm_clk_notif_unregister(struct clk *clk, struct notifier_block *nb)
{
struct msm_clk_notifier *cn = NULL;
int ret = -EINVAL;
if (!clk || !nb)
return -EINVAL;
mutex_lock(&clk->prepare_lock);
list_for_each_entry(cn, &clk_notifier_list, node)
if (cn->clk == clk)
break;
if (cn->clk == clk) {
ret = srcu_notifier_chain_unregister(&cn->notifier_head, nb);
clk->notifier_count--;
/* XXX the notifier code should handle this better */
if (!cn->notifier_head.head) {
srcu_cleanup_notifier_head(&cn->notifier_head);
list_del(&cn->node);
kfree(cn);
}
} else {
ret = -ENOENT;
}
mutex_unlock(&clk->prepare_lock);
return ret;
}
unsigned long clk_get_rate(struct clk *clk)
{
if (IS_ERR_OR_NULL(clk))
return 0;
if (!clk->ops->get_rate)
return clk->rate;
return clk->ops->get_rate(clk);
}
EXPORT_SYMBOL(clk_get_rate);
int clk_set_rate(struct clk *clk, unsigned long rate)
{
unsigned long start_rate;
int rc = 0;
const char *name;
if (IS_ERR_OR_NULL(clk))
return -EINVAL;
name = clk->dbg_name;
if (!is_rate_valid(clk, rate))
return -EINVAL;
mutex_lock(&clk->prepare_lock);
/* Return early if the rate isn't going to change */
if (clk->rate == rate && !(clk->flags & CLKFLAG_NO_RATE_CACHE))
goto out;
if (!clk->ops->set_rate) {
rc = -EINVAL;
goto out;
}
trace_clock_set_rate(name, rate, raw_smp_processor_id());
start_rate = clk->rate;
if (clk->notifier_count)
__clk_notify(clk, PRE_RATE_CHANGE, clk->rate, rate);
if (clk->ops->pre_set_rate) {
rc = clk->ops->pre_set_rate(clk, rate);
if (rc)
goto abort_set_rate;
}
/* Enforce vdd requirements for target frequency. */
if (clk->prepare_count) {
rc = vote_rate_vdd(clk, rate);
if (rc)
goto err_vote_vdd;
}
rc = clk->ops->set_rate(clk, rate);
if (rc)
goto err_set_rate;
clk->rate = rate;
/* Release vdd requirements for starting frequency. */
if (clk->prepare_count)
unvote_rate_vdd(clk, start_rate);
if (clk->ops->post_set_rate)
clk->ops->post_set_rate(clk, start_rate);
if (clk->notifier_count)
__clk_notify(clk, POST_RATE_CHANGE, start_rate, clk->rate);
trace_clock_set_rate_complete(name, clk->rate, raw_smp_processor_id());
out:
mutex_unlock(&clk->prepare_lock);
return rc;
abort_set_rate:
__clk_notify(clk, ABORT_RATE_CHANGE, clk->rate, rate);
err_set_rate:
if (clk->prepare_count)
unvote_rate_vdd(clk, rate);
err_vote_vdd:
/* clk->rate is still the old rate. So, pass the new rate instead. */
if (clk->ops->post_set_rate)
clk->ops->post_set_rate(clk, rate);
goto out;
}
EXPORT_SYMBOL(clk_set_rate);
long clk_round_rate(struct clk *clk, unsigned long rate)
{
long rrate;
unsigned long fmax = 0, i;
if (IS_ERR_OR_NULL(clk))
return -EINVAL;
for (i = 0; i < clk->num_fmax; i++)
fmax = max(fmax, clk->fmax[i]);
if (!fmax)
fmax = ULONG_MAX;
rate = min(rate, fmax);
if (clk->ops->round_rate)
rrate = clk->ops->round_rate(clk, rate);
else if (clk->rate)
rrate = clk->rate;
else
return -EINVAL;
if (rrate > fmax)
return -EINVAL;
return rrate;
}
EXPORT_SYMBOL(clk_round_rate);
int clk_set_max_rate(struct clk *clk, unsigned long rate)
{
if (IS_ERR_OR_NULL(clk))
return -EINVAL;
if (!clk->ops->set_max_rate)
return -EINVAL;
return clk->ops->set_max_rate(clk, rate);
}
EXPORT_SYMBOL(clk_set_max_rate);
int parent_to_src_sel(struct clk_src *parents, int num_parents, struct clk *p)
{
int i;
for (i = 0; i < num_parents; i++) {
if (parents[i].src == p)
return parents[i].sel;
}
return -EINVAL;
}
EXPORT_SYMBOL(parent_to_src_sel);
int clk_get_parent_sel(struct clk *c, struct clk *parent)
{
return parent_to_src_sel(c->parents, c->num_parents, parent);
}
EXPORT_SYMBOL(clk_get_parent_sel);
int clk_set_parent(struct clk *clk, struct clk *parent)
{
int rc = 0;
if (IS_ERR_OR_NULL(clk))
return -EINVAL;
if (!clk->ops->set_parent && clk->parent == parent)
return 0;
if (!clk->ops->set_parent)
return -EINVAL;
mutex_lock(&clk->prepare_lock);
if (clk->parent == parent && !(clk->flags & CLKFLAG_NO_RATE_CACHE))
goto out;
rc = clk->ops->set_parent(clk, parent);
out:
mutex_unlock(&clk->prepare_lock);
return rc;
}
EXPORT_SYMBOL(clk_set_parent);
struct clk *clk_get_parent(struct clk *clk)
{
if (IS_ERR_OR_NULL(clk))
return NULL;
return clk->parent;
}
EXPORT_SYMBOL(clk_get_parent);
int clk_set_flags(struct clk *clk, unsigned long flags)
{
if (IS_ERR_OR_NULL(clk))
return -EINVAL;
if (!clk->ops->set_flags)
return -EINVAL;
return clk->ops->set_flags(clk, flags);
}
EXPORT_SYMBOL(clk_set_flags);
int clk_set_duty_cycle(struct clk *clk, unsigned int num, unsigned int den)
{
if (IS_ERR_OR_NULL(clk))
return -EINVAL;
if (num > den) {
pr_err("Numerator cannot be > denominator\n");
return -EINVAL;
}
if (!den) {
pr_err("Denominator can not be Zero\n");
return -EINVAL;
}
if (!clk->ops->set_duty_cycle)
return -EINVAL;
return clk->ops->set_duty_cycle(clk, num, den);
}
EXPORT_SYMBOL(clk_set_duty_cycle);
static LIST_HEAD(initdata_list);
static void init_sibling_lists(struct clk_lookup *clock_tbl, size_t num_clocks)
{
struct clk *clk, *parent;
unsigned long n;
for (n = 0; n < num_clocks; n++) {
clk = clock_tbl[n].clk;
parent = clk->parent;
if (parent && list_empty(&clk->siblings))
list_add(&clk->siblings, &parent->children);
}
}
static void vdd_class_init(struct clk_vdd_class *vdd)
{
struct handoff_vdd *v;
if (!vdd)
return;
if (vdd->skip_handoff)
return;
list_for_each_entry(v, &handoff_vdd_list, list) {
if (v->vdd_class == vdd)
return;
}
pr_debug("voting for vdd_class %s\n", vdd->class_name);
if (vote_vdd_level(vdd, vdd->num_levels - 1))
pr_err("failed to vote for %s\n", vdd->class_name);
v = kmalloc(sizeof(*v), GFP_KERNEL);
if (!v)
return;
v->vdd_class = vdd;
list_add_tail(&v->list, &handoff_vdd_list);
}
static int __handoff_clk(struct clk *clk)
{
enum handoff state = HANDOFF_DISABLED_CLK;
struct handoff_clk *h = NULL;
int rc, i;
if (clk == NULL || clk->flags & CLKFLAG_INIT_DONE ||
clk->flags & CLKFLAG_SKIP_HANDOFF)
return 0;
if (clk->flags & CLKFLAG_INIT_ERR)
return -ENXIO;
if (clk->flags & CLKFLAG_EPROBE_DEFER)
return -EPROBE_DEFER;
/* Handoff any 'depends' clock first. */
rc = __handoff_clk(clk->depends);
if (rc)
goto err;
/*
* Handoff functions for the parent must be called before the
* children can be handed off. Without handing off the parents and
* knowing their rate and state (on/off), it's impossible to figure
* out the rate and state of the children.
*/
if (clk->ops->get_parent)
clk->parent = clk->ops->get_parent(clk);
if (IS_ERR(clk->parent)) {
rc = PTR_ERR(clk->parent);
goto err;
}
rc = __handoff_clk(clk->parent);
if (rc)
goto err;
for (i = 0; i < clk->num_parents; i++) {
rc = __handoff_clk(clk->parents[i].src);
if (rc)
goto err;
}
if (clk->ops->handoff)
state = clk->ops->handoff(clk);
if (state == HANDOFF_ENABLED_CLK) {
h = kmalloc(sizeof(*h), GFP_KERNEL);
if (!h) {
rc = -ENOMEM;
goto err;
}
rc = clk_prepare_enable(clk->parent);
if (rc)
goto err;
rc = clk_prepare_enable(clk->depends);
if (rc)
goto err_depends;
rc = vote_rate_vdd(clk, clk->rate);
WARN(rc, "%s unable to vote for voltage!\n", clk->dbg_name);
clk->count = 1;
clk->prepare_count = 1;
h->clk = clk;
list_add_tail(&h->list, &handoff_list);
pr_debug("Handed off %s rate=%lu\n", clk->dbg_name, clk->rate);
}
if (clk->init_rate && clk_set_rate(clk, clk->init_rate))
pr_err("failed to set an init rate of %lu on %s\n",
clk->init_rate, clk->dbg_name);
if (clk->always_on && clk_prepare_enable(clk))
pr_err("failed to enable always-on clock %s\n",
clk->dbg_name);
clk->flags |= CLKFLAG_INIT_DONE;
/* if the clk is on orphan list, remove it */
list_del_init(&clk->list);
clock_debug_register(clk);
return 0;
err_depends:
clk_disable_unprepare(clk->parent);
err:
kfree(h);
if (rc == -EPROBE_DEFER) {
clk->flags |= CLKFLAG_EPROBE_DEFER;
if (list_empty(&clk->list))
list_add_tail(&clk->list, &orphan_clk_list);
} else {
pr_err("%s handoff failed (%d)\n", clk->dbg_name, rc);
clk->flags |= CLKFLAG_INIT_ERR;
}
return rc;
}
/**
* msm_clock_register() - Register additional clock tables
* @table: Table of clocks
* @size: Size of @table
*
* Upon return, clock APIs may be used to control clocks registered using this
* function.
*/
int msm_clock_register(struct clk_lookup *table, size_t size)
{
int n = 0, rc;
struct clk *c, *safe;
bool found_more_clks;
mutex_lock(&msm_clock_init_lock);
init_sibling_lists(table, size);
/*
* Enable regulators and temporarily set them up at maximum voltage.
* Once all the clocks have made their respective vote, remove this
* temporary vote. The removing of the temporary vote is done at
* late_init, by which time we assume all the clocks would have been
* handed off.
*/
for (n = 0; n < size; n++)
vdd_class_init(table[n].clk->vdd_class);
/*
* Detect and preserve initial clock state until clock_late_init() or
* a driver explicitly changes it, whichever is first.
*/
for (n = 0; n < size; n++)
__handoff_clk(table[n].clk);
/* maintain backwards compatibility */
if (table[0].con_id || table[0].dev_id)
clkdev_add_table(table, size);
do {
found_more_clks = false;
/* clear cached __handoff_clk return values */
list_for_each_entry_safe(c, safe, &orphan_clk_list, list)
c->flags &= ~CLKFLAG_EPROBE_DEFER;
list_for_each_entry_safe(c, safe, &orphan_clk_list, list) {
rc = __handoff_clk(c);
if (!rc)
found_more_clks = true;
}
} while (found_more_clks);
mutex_unlock(&msm_clock_init_lock);
return 0;
}
EXPORT_SYMBOL(msm_clock_register);
struct of_msm_provider_data {
struct clk_lookup *table;
size_t size;
};
static struct clk *of_clk_src_get(struct of_phandle_args *clkspec,
void *data)
{
struct of_msm_provider_data *ofdata = data;
int n;
for (n = 0; n < ofdata->size; n++) {
if (clkspec->args[0] == ofdata->table[n].of_idx)
return ofdata->table[n].clk;
}
return ERR_PTR(-ENOENT);
}
#define MAX_LEN_OPP_HANDLE 50
#define LEN_OPP_HANDLE 16
#define LEN_OPP_VCORNER_HANDLE 22
static struct device **derive_device_list(struct clk *clk,
struct device_node *np,
char *clk_handle_name, int len)
{
int j, count, cpu;
struct platform_device *pdev;
struct device_node *dev_node;
struct device **device_list;
count = len/sizeof(u32);
device_list = kmalloc_array(count, sizeof(struct device *),
GFP_KERNEL);
if (!device_list)
return ERR_PTR(-ENOMEM);
for (j = 0; j < count; j++) {
device_list[j] = NULL;
dev_node = of_parse_phandle(np, clk_handle_name, j);
if (!dev_node) {
pr_err("Unable to get device_node pointer for %s opp-handle (%s)\n",
clk->dbg_name, clk_handle_name);
goto err_parse_phandle;
}
for_each_possible_cpu(cpu) {
if (of_get_cpu_node(cpu, NULL) == dev_node)
device_list[j] = get_cpu_device(cpu);
}
if (device_list[j])
continue;
pdev = of_find_device_by_node(dev_node);
if (!pdev) {
pr_err("Unable to find platform_device node for %s opp-handle\n",
clk->dbg_name);
goto err_parse_phandle;
}
device_list[j] = &pdev->dev;
}
return device_list;
err_parse_phandle:
kfree(device_list);
return ERR_PTR(-EINVAL);
}
static int get_voltage(struct clk *clk, unsigned long rate,
int store_vcorner, int n)
{
struct clk_vdd_class *vdd;
int uv, level, corner;
/*
* Use the first regulator in the vdd class
* for the OPP table.
*/
vdd = clk->vdd_class;
if (vdd->num_regulators > 1) {
corner = vdd->vdd_uv[vdd->num_regulators * n];
} else {
level = find_vdd_level(clk, rate);
if (level < 0) {
pr_err("Could not find vdd level\n");
return -EINVAL;
}
corner = vdd->vdd_uv[level];
}
if (!corner) {
pr_err("%s: Unable to find vdd level for rate %lu\n",
clk->dbg_name, rate);
return -EINVAL;
}
if (store_vcorner) {
uv = corner;
return uv;
}
uv = regulator_list_corner_voltage(vdd->regulator[0], corner);
if (uv < 0) {
pr_err("%s: no uv for corner %d - err: %d\n",
clk->dbg_name, corner, uv);
return uv;
}
return uv;
}
static int add_and_print_opp(struct clk *clk, struct device **device_list,
int count, unsigned long rate, int uv, int n)
{
int j, ret = 0;
for (j = 0; j < count; j++) {
ret = dev_pm_opp_add(device_list[j], rate, uv);
if (ret) {
pr_err("%s: couldn't add OPP for %lu - err: %d\n",
clk->dbg_name, rate, ret);
return ret;
}
if (n == 1 || n == clk->num_fmax - 1 ||
rate == clk_round_rate(clk, INT_MAX))
pr_info("%s: set OPP pair(%lu Hz: %u uV) on %s\n",
clk->dbg_name, rate, uv,
dev_name(device_list[j]));
}
return ret;
}
static void populate_clock_opp_table(struct device_node *np,
struct clk_lookup *table, size_t size)
{
struct device **device_list;
struct clk *clk;
char clk_handle_name[MAX_LEN_OPP_HANDLE];
char clk_store_volt_corner[MAX_LEN_OPP_HANDLE];
size_t i;
int n, len, count, uv = 0;
unsigned long rate, ret = 0;
bool store_vcorner;
/* Iterate across all clocks in the clock controller */
for (i = 0; i < size; i++) {
n = 1;
rate = 0;
store_vcorner = false;
clk = table[i].clk;
if (!clk || !clk->num_fmax || clk->opp_table_populated)
continue;
if (strlen(clk->dbg_name) + LEN_OPP_HANDLE
< MAX_LEN_OPP_HANDLE) {
ret = snprintf(clk_handle_name,
ARRAY_SIZE(clk_handle_name),
"qcom,%s-opp-handle", clk->dbg_name);
if (ret < strlen(clk->dbg_name) + LEN_OPP_HANDLE) {
pr_err("Failed to hold clk_handle_name\n");
continue;
}
} else {
pr_err("clk name (%s) too large to fit in clk_handle_name\n",
clk->dbg_name);
continue;
}
if (strlen(clk->dbg_name) + LEN_OPP_VCORNER_HANDLE
< MAX_LEN_OPP_HANDLE) {
ret = snprintf(clk_store_volt_corner,
ARRAY_SIZE(clk_store_volt_corner),
"qcom,%s-opp-store-vcorner", clk->dbg_name);
if (ret < strlen(clk->dbg_name) +
LEN_OPP_VCORNER_HANDLE) {
pr_err("Failed to hold clk_store_volt_corner\n");
continue;
}
} else {
pr_err("clk name (%s) too large to fit in clk_store_volt_corner\n",
clk->dbg_name);
continue;
}
if (!of_find_property(np, clk_handle_name, &len)) {
pr_debug("Unable to find %s\n", clk_handle_name);
if (!of_find_property(np, clk_store_volt_corner,
&len)) {
pr_debug("Unable to find %s\n",
clk_store_volt_corner);
continue;
} else {
store_vcorner = true;
device_list = derive_device_list(clk, np,
clk_store_volt_corner, len);
}
} else
device_list = derive_device_list(clk, np,
clk_handle_name, len);
if (IS_ERR_OR_NULL(device_list)) {
pr_err("Failed to fill device_list\n");
continue;
}
count = len/sizeof(u32);
while (1) {
/*
* Calling clk_round_rate will not work for all clocks
* (eg. mux_div). Use their fmax values instead to get
* list of all available frequencies.
*/
if (clk->ops->list_rate) {
ret = clk_round_rate(clk, rate + 1);
if (ret < 0) {
pr_err("clk_round_rate failed for %s\n",
clk->dbg_name);
goto err_round_rate;
}
/*
* If clk_round_rate give the same value on
* consecutive iterations, exit loop since
* we're at the maximum clock frequency.
*/
if (rate == ret)
break;
rate = ret;
} else {
if (n < clk->num_fmax)
rate = clk->fmax[n];
else
break;
}
uv = get_voltage(clk, rate, store_vcorner, n);
if (uv < 0)
goto err_round_rate;
ret = add_and_print_opp(clk, device_list, count,
rate, uv, n);
if (ret)
goto err_round_rate;
n++;
}
err_round_rate:
/* If OPP table population was successful, set the flag */
if (uv >= 0 && ret >= 0)
clk->opp_table_populated = true;
kfree(device_list);
}
}
/**
* of_msm_clock_register() - Register clock tables with clkdev and with the
* clock DT framework
* @table: Table of clocks
* @size: Size of @table
* @np: Device pointer corresponding to the clock-provider device
*
* Upon return, clock APIs may be used to control clocks registered using this
* function.
*/
int of_msm_clock_register(struct device_node *np, struct clk_lookup *table,
size_t size)
{
int ret = 0;
struct of_msm_provider_data *data;
data = kzalloc(sizeof(*data), GFP_KERNEL);
if (!data)
return -ENOMEM;
data->table = table;
data->size = size;
ret = of_clk_add_provider(np, of_clk_src_get, data);
if (ret) {
kfree(data);
return -ENOMEM;
}
populate_clock_opp_table(np, table, size);
return msm_clock_register(table, size);
}
EXPORT_SYMBOL(of_msm_clock_register);
/**
* msm_clock_init() - Register and initialize a clock driver
* @data: Driver-specific clock initialization data
*
* Upon return from this call, clock APIs may be used to control
* clocks registered with this API.
*/
int __init msm_clock_init(struct clock_init_data *data)
{
if (!data)
return -EINVAL;
if (data->pre_init)
data->pre_init();
mutex_lock(&msm_clock_init_lock);
if (data->late_init)
list_add(&data->list, &initdata_list);
mutex_unlock(&msm_clock_init_lock);
msm_clock_register(data->table, data->size);
if (data->post_init)
data->post_init();
return 0;
}
static int __init clock_late_init(void)
{
struct handoff_clk *h, *h_temp;
struct handoff_vdd *v, *v_temp;
struct clock_init_data *initdata, *initdata_temp;
int ret = 0;
pr_info("%s: Removing enables held for handed-off clocks\n", __func__);
mutex_lock(&msm_clock_init_lock);
list_for_each_entry_safe(initdata, initdata_temp,
&initdata_list, list) {
ret = initdata->late_init();
if (ret)
pr_err("%s: %pS failed late_init.\n", __func__,
initdata);
}
list_for_each_entry_safe(h, h_temp, &handoff_list, list) {
clk_disable_unprepare(h->clk);
list_del(&h->list);
kfree(h);
}
list_for_each_entry_safe(v, v_temp, &handoff_vdd_list, list) {
unvote_vdd_level(v->vdd_class, v->vdd_class->num_levels - 1);
list_del(&v->list);
kfree(v);
}
mutex_unlock(&msm_clock_init_lock);
return ret;
}
/* clock_late_init should run only after all deferred probing
* (excluding DLKM probes) has completed.
*/
late_initcall_sync(clock_late_init);