linux/arch/arm/mach-omap2/clockdomain.c

1044 lines
28 KiB
C
Raw Normal View History

/*
* OMAP2/3/4 clockdomain framework functions
*
* Copyright (C) 2008-2010 Texas Instruments, Inc.
* Copyright (C) 2008-2010 Nokia Corporation
*
* Written by Paul Walmsley and Jouni Högander
* Added OMAP4 specific support by Abhijit Pagare <abhijitpagare@ti.com>
*
* This program is free software; you can redistribute it and/or modify
* it under the terms of the GNU General Public License version 2 as
* published by the Free Software Foundation.
*/
#undef DEBUG
#include <linux/kernel.h>
#include <linux/device.h>
#include <linux/list.h>
#include <linux/errno.h>
#include <linux/delay.h>
#include <linux/clk.h>
#include <linux/limits.h>
#include <linux/err.h>
#include <linux/io.h>
#include <linux/bitops.h>
#include "prm2xxx_3xxx.h"
#include "prm-regbits-24xx.h"
#include "cm2xxx_3xxx.h"
#include "cm-regbits-24xx.h"
#include "cminst44xx.h"
#include "prcm44xx.h"
#include <plat/clock.h>
#include "powerdomain.h"
#include "clockdomain.h"
#include <plat/prcm.h>
/* clkdm_list contains all registered struct clockdomains */
static LIST_HEAD(clkdm_list);
/* array of clockdomain deps to be added/removed when clkdm in hwsup mode */
static struct clkdm_autodep *autodeps;
/* Private functions */
static struct clockdomain *_clkdm_lookup(const char *name)
{
struct clockdomain *clkdm, *temp_clkdm;
if (!name)
return NULL;
clkdm = NULL;
list_for_each_entry(temp_clkdm, &clkdm_list, node) {
if (!strcmp(name, temp_clkdm->name)) {
clkdm = temp_clkdm;
break;
}
}
return clkdm;
}
/**
* _clkdm_register - register a clockdomain
* @clkdm: struct clockdomain * to register
*
* Adds a clockdomain to the internal clockdomain list.
* Returns -EINVAL if given a null pointer, -EEXIST if a clockdomain is
* already registered by the provided name, or 0 upon success.
*/
static int _clkdm_register(struct clockdomain *clkdm)
{
struct powerdomain *pwrdm;
if (!clkdm || !clkdm->name)
return -EINVAL;
if (!omap_chip_is(clkdm->omap_chip))
return -EINVAL;
pwrdm = pwrdm_lookup(clkdm->pwrdm.name);
if (!pwrdm) {
pr_err("clockdomain: %s: powerdomain %s does not exist\n",
clkdm->name, clkdm->pwrdm.name);
return -EINVAL;
}
clkdm->pwrdm.ptr = pwrdm;
/* Verify that the clockdomain is not already registered */
if (_clkdm_lookup(clkdm->name))
return -EEXIST;
list_add(&clkdm->node, &clkdm_list);
pwrdm_add_clkdm(pwrdm, clkdm);
pr_debug("clockdomain: registered %s\n", clkdm->name);
return 0;
}
/* _clkdm_deps_lookup - look up the specified clockdomain in a clkdm list */
static struct clkdm_dep *_clkdm_deps_lookup(struct clockdomain *clkdm,
struct clkdm_dep *deps)
{
struct clkdm_dep *cd;
if (!clkdm || !deps || !omap_chip_is(clkdm->omap_chip))
return ERR_PTR(-EINVAL);
for (cd = deps; cd->clkdm_name; cd++) {
if (!omap_chip_is(cd->omap_chip))
continue;
if (!cd->clkdm && cd->clkdm_name)
cd->clkdm = _clkdm_lookup(cd->clkdm_name);
if (cd->clkdm == clkdm)
break;
}
if (!cd->clkdm_name)
return ERR_PTR(-ENOENT);
return cd;
}
/*
* _autodep_lookup - resolve autodep clkdm names to clkdm pointers; store
* @autodep: struct clkdm_autodep * to resolve
*
* Resolve autodep clockdomain names to clockdomain pointers via
* clkdm_lookup() and store the pointers in the autodep structure. An
* "autodep" is a clockdomain sleep/wakeup dependency that is
* automatically added and removed whenever clocks in the associated
* clockdomain are enabled or disabled (respectively) when the
* clockdomain is in hardware-supervised mode. Meant to be called
* once at clockdomain layer initialization, since these should remain
* fixed for a particular architecture. No return value.
*
* XXX autodeps are deprecated and should be removed at the earliest
* opportunity
*/
static void _autodep_lookup(struct clkdm_autodep *autodep)
{
struct clockdomain *clkdm;
if (!autodep)
return;
if (!omap_chip_is(autodep->omap_chip))
return;
clkdm = clkdm_lookup(autodep->clkdm.name);
if (!clkdm) {
pr_err("clockdomain: autodeps: clockdomain %s does not exist\n",
autodep->clkdm.name);
clkdm = ERR_PTR(-ENOENT);
}
autodep->clkdm.ptr = clkdm;
}
/*
* _clkdm_add_autodeps - add auto sleepdeps/wkdeps to clkdm upon clock enable
* @clkdm: struct clockdomain *
*
* Add the "autodep" sleep & wakeup dependencies to clockdomain 'clkdm'
* in hardware-supervised mode. Meant to be called from clock framework
* when a clock inside clockdomain 'clkdm' is enabled. No return value.
*
* XXX autodeps are deprecated and should be removed at the earliest
* opportunity
*/
static void _clkdm_add_autodeps(struct clockdomain *clkdm)
{
struct clkdm_autodep *autodep;
if (!autodeps)
return;
for (autodep = autodeps; autodep->clkdm.ptr; autodep++) {
if (IS_ERR(autodep->clkdm.ptr))
continue;
if (!omap_chip_is(autodep->omap_chip))
continue;
pr_debug("clockdomain: adding %s sleepdep/wkdep for "
"clkdm %s\n", autodep->clkdm.ptr->name,
clkdm->name);
clkdm_add_sleepdep(clkdm, autodep->clkdm.ptr);
clkdm_add_wkdep(clkdm, autodep->clkdm.ptr);
}
}
/*
* _clkdm_add_autodeps - remove auto sleepdeps/wkdeps from clkdm
* @clkdm: struct clockdomain *
*
* Remove the "autodep" sleep & wakeup dependencies from clockdomain 'clkdm'
* in hardware-supervised mode. Meant to be called from clock framework
* when a clock inside clockdomain 'clkdm' is disabled. No return value.
*
* XXX autodeps are deprecated and should be removed at the earliest
* opportunity
*/
static void _clkdm_del_autodeps(struct clockdomain *clkdm)
{
struct clkdm_autodep *autodep;
if (!autodeps)
return;
for (autodep = autodeps; autodep->clkdm.ptr; autodep++) {
if (IS_ERR(autodep->clkdm.ptr))
continue;
if (!omap_chip_is(autodep->omap_chip))
continue;
pr_debug("clockdomain: removing %s sleepdep/wkdep for "
"clkdm %s\n", autodep->clkdm.ptr->name,
clkdm->name);
clkdm_del_sleepdep(clkdm, autodep->clkdm.ptr);
clkdm_del_wkdep(clkdm, autodep->clkdm.ptr);
}
}
/**
* _enable_hwsup - place a clockdomain into hardware-supervised idle
* @clkdm: struct clockdomain *
*
* Place the clockdomain into hardware-supervised idle mode. No return
* value.
*
* XXX Should this return an error if the clockdomain does not support
* hardware-supervised idle mode?
*/
static void _enable_hwsup(struct clockdomain *clkdm)
{
if (cpu_is_omap24xx())
omap2xxx_cm_clkdm_enable_hwsup(clkdm->pwrdm.ptr->prcm_offs,
clkdm->clktrctrl_mask);
else if (cpu_is_omap34xx())
omap3xxx_cm_clkdm_enable_hwsup(clkdm->pwrdm.ptr->prcm_offs,
clkdm->clktrctrl_mask);
else if (cpu_is_omap44xx())
return omap4_cminst_clkdm_enable_hwsup(clkdm->prcm_partition,
clkdm->cm_inst,
clkdm->clkdm_offs);
else
BUG();
}
/**
* _disable_hwsup - place a clockdomain into software-supervised idle
* @clkdm: struct clockdomain *
*
* Place the clockdomain @clkdm into software-supervised idle mode.
* No return value.
*
* XXX Should this return an error if the clockdomain does not support
* software-supervised idle mode?
*/
static void _disable_hwsup(struct clockdomain *clkdm)
{
if (cpu_is_omap24xx())
omap2xxx_cm_clkdm_disable_hwsup(clkdm->pwrdm.ptr->prcm_offs,
clkdm->clktrctrl_mask);
else if (cpu_is_omap34xx())
omap3xxx_cm_clkdm_disable_hwsup(clkdm->pwrdm.ptr->prcm_offs,
clkdm->clktrctrl_mask);
else if (cpu_is_omap44xx())
return omap4_cminst_clkdm_disable_hwsup(clkdm->prcm_partition,
clkdm->cm_inst,
clkdm->clkdm_offs);
else
BUG();
}
/* Public functions */
/**
* clkdm_init - set up the clockdomain layer
* @clkdms: optional pointer to an array of clockdomains to register
* @init_autodeps: optional pointer to an array of autodeps to register
*
* Set up internal state. If a pointer to an array of clockdomains
* @clkdms was supplied, loop through the list of clockdomains,
* register all that are available on the current platform. Similarly,
* if a pointer to an array of clockdomain autodependencies
* @init_autodeps was provided, register those. No return value.
*/
void clkdm_init(struct clockdomain **clkdms,
struct clkdm_autodep *init_autodeps)
{
struct clockdomain **c = NULL;
OMAP clockdomains: add usecounting for wakeup and sleep dependencies Add usecounting for wakeup and sleep dependencies. In the current situation, if several functions add dependencies on the same clockdomains, when the first dependency removal function is called, the dependency will be incorrectly removed from the hardware. Add clkdm_clear_all_wkdeps() and clkdm_clear_all_sleepdeps(), which provide a fast and usecounting-consistent way to clear all hardware clockdomain dependencies, since accesses to these registers can be quite slow. pm{2,3}4xx.c has been updated to use these new functions. The original version of this patch did not touch these files, which previously wrote directly to the wkdep registers, and thus confused the usecounting code. This problem was found by Kevin Hilman <khilman@deeprootsystems.com>. N.B.: This patch introduces one significant functional difference over the previous pm34xx.c code: sleepdeps are now cleared during clockdomain initialization, whereas previously they were left untouched. This has been tested by Kevin and confirmed to work. The original version of this patch also did not take into consideration that some clockdomains do not have sleep or wakeup dependency sources, which caused NULL pointer dereferences. This problem was debugged and fixed by Kevin Hilman <khilman@deeprootsystems.com>. Signed-off-by: Paul Walmsley <paul@pwsan.com> Signed-off-by: Kevin Hilman <khilman@deeprootsystems.com> Cc: Jouni Högander <jouni.hogander@nokia.com>
2010-01-27 03:13:01 +00:00
struct clockdomain *clkdm;
struct clkdm_autodep *autodep = NULL;
if (clkdms)
for (c = clkdms; *c; c++)
_clkdm_register(*c);
autodeps = init_autodeps;
if (autodeps)
for (autodep = autodeps; autodep->clkdm.ptr; autodep++)
_autodep_lookup(autodep);
OMAP clockdomains: add usecounting for wakeup and sleep dependencies Add usecounting for wakeup and sleep dependencies. In the current situation, if several functions add dependencies on the same clockdomains, when the first dependency removal function is called, the dependency will be incorrectly removed from the hardware. Add clkdm_clear_all_wkdeps() and clkdm_clear_all_sleepdeps(), which provide a fast and usecounting-consistent way to clear all hardware clockdomain dependencies, since accesses to these registers can be quite slow. pm{2,3}4xx.c has been updated to use these new functions. The original version of this patch did not touch these files, which previously wrote directly to the wkdep registers, and thus confused the usecounting code. This problem was found by Kevin Hilman <khilman@deeprootsystems.com>. N.B.: This patch introduces one significant functional difference over the previous pm34xx.c code: sleepdeps are now cleared during clockdomain initialization, whereas previously they were left untouched. This has been tested by Kevin and confirmed to work. The original version of this patch also did not take into consideration that some clockdomains do not have sleep or wakeup dependency sources, which caused NULL pointer dereferences. This problem was debugged and fixed by Kevin Hilman <khilman@deeprootsystems.com>. Signed-off-by: Paul Walmsley <paul@pwsan.com> Signed-off-by: Kevin Hilman <khilman@deeprootsystems.com> Cc: Jouni Högander <jouni.hogander@nokia.com>
2010-01-27 03:13:01 +00:00
/*
* Put all clockdomains into software-supervised mode; PM code
* should later enable hardware-supervised mode as appropriate
OMAP clockdomains: add usecounting for wakeup and sleep dependencies Add usecounting for wakeup and sleep dependencies. In the current situation, if several functions add dependencies on the same clockdomains, when the first dependency removal function is called, the dependency will be incorrectly removed from the hardware. Add clkdm_clear_all_wkdeps() and clkdm_clear_all_sleepdeps(), which provide a fast and usecounting-consistent way to clear all hardware clockdomain dependencies, since accesses to these registers can be quite slow. pm{2,3}4xx.c has been updated to use these new functions. The original version of this patch did not touch these files, which previously wrote directly to the wkdep registers, and thus confused the usecounting code. This problem was found by Kevin Hilman <khilman@deeprootsystems.com>. N.B.: This patch introduces one significant functional difference over the previous pm34xx.c code: sleepdeps are now cleared during clockdomain initialization, whereas previously they were left untouched. This has been tested by Kevin and confirmed to work. The original version of this patch also did not take into consideration that some clockdomains do not have sleep or wakeup dependency sources, which caused NULL pointer dereferences. This problem was debugged and fixed by Kevin Hilman <khilman@deeprootsystems.com>. Signed-off-by: Paul Walmsley <paul@pwsan.com> Signed-off-by: Kevin Hilman <khilman@deeprootsystems.com> Cc: Jouni Högander <jouni.hogander@nokia.com>
2010-01-27 03:13:01 +00:00
*/
list_for_each_entry(clkdm, &clkdm_list, node) {
if (clkdm->flags & CLKDM_CAN_FORCE_WAKEUP)
omap2_clkdm_wakeup(clkdm);
else if (clkdm->flags & CLKDM_CAN_DISABLE_AUTO)
omap2_clkdm_deny_idle(clkdm);
clkdm_clear_all_wkdeps(clkdm);
clkdm_clear_all_sleepdeps(clkdm);
OMAP clockdomains: add usecounting for wakeup and sleep dependencies Add usecounting for wakeup and sleep dependencies. In the current situation, if several functions add dependencies on the same clockdomains, when the first dependency removal function is called, the dependency will be incorrectly removed from the hardware. Add clkdm_clear_all_wkdeps() and clkdm_clear_all_sleepdeps(), which provide a fast and usecounting-consistent way to clear all hardware clockdomain dependencies, since accesses to these registers can be quite slow. pm{2,3}4xx.c has been updated to use these new functions. The original version of this patch did not touch these files, which previously wrote directly to the wkdep registers, and thus confused the usecounting code. This problem was found by Kevin Hilman <khilman@deeprootsystems.com>. N.B.: This patch introduces one significant functional difference over the previous pm34xx.c code: sleepdeps are now cleared during clockdomain initialization, whereas previously they were left untouched. This has been tested by Kevin and confirmed to work. The original version of this patch also did not take into consideration that some clockdomains do not have sleep or wakeup dependency sources, which caused NULL pointer dereferences. This problem was debugged and fixed by Kevin Hilman <khilman@deeprootsystems.com>. Signed-off-by: Paul Walmsley <paul@pwsan.com> Signed-off-by: Kevin Hilman <khilman@deeprootsystems.com> Cc: Jouni Högander <jouni.hogander@nokia.com>
2010-01-27 03:13:01 +00:00
}
}
/**
* clkdm_lookup - look up a clockdomain by name, return a pointer
* @name: name of clockdomain
*
* Find a registered clockdomain by its name @name. Returns a pointer
* to the struct clockdomain if found, or NULL otherwise.
*/
struct clockdomain *clkdm_lookup(const char *name)
{
struct clockdomain *clkdm, *temp_clkdm;
if (!name)
return NULL;
clkdm = NULL;
list_for_each_entry(temp_clkdm, &clkdm_list, node) {
if (!strcmp(name, temp_clkdm->name)) {
clkdm = temp_clkdm;
break;
}
}
return clkdm;
}
/**
* clkdm_for_each - call function on each registered clockdomain
* @fn: callback function *
*
* Call the supplied function @fn for each registered clockdomain.
* The callback function @fn can return anything but 0 to bail
* out early from the iterator. The callback function is called with
* the clkdm_mutex held, so no clockdomain structure manipulation
* functions should be called from the callback, although hardware
* clockdomain control functions are fine. Returns the last return
* value of the callback function, which should be 0 for success or
* anything else to indicate failure; or -EINVAL if the function pointer
* is null.
*/
int clkdm_for_each(int (*fn)(struct clockdomain *clkdm, void *user),
void *user)
{
struct clockdomain *clkdm;
int ret = 0;
if (!fn)
return -EINVAL;
list_for_each_entry(clkdm, &clkdm_list, node) {
ret = (*fn)(clkdm, user);
if (ret)
break;
}
return ret;
}
/**
* clkdm_get_pwrdm - return a ptr to the pwrdm that this clkdm resides in
* @clkdm: struct clockdomain *
*
* Return a pointer to the struct powerdomain that the specified clockdomain
* @clkdm exists in, or returns NULL if @clkdm is NULL.
*/
struct powerdomain *clkdm_get_pwrdm(struct clockdomain *clkdm)
{
if (!clkdm)
return NULL;
return clkdm->pwrdm.ptr;
}
/* Hardware clockdomain control */
/**
* clkdm_add_wkdep - add a wakeup dependency from clkdm2 to clkdm1
* @clkdm1: wake this struct clockdomain * up (dependent)
* @clkdm2: when this struct clockdomain * wakes up (source)
*
* When the clockdomain represented by @clkdm2 wakes up, wake up
* @clkdm1. Implemented in hardware on the OMAP, this feature is
* designed to reduce wakeup latency of the dependent clockdomain @clkdm1.
* Returns -EINVAL if presented with invalid clockdomain pointers,
* -ENOENT if @clkdm2 cannot wake up clkdm1 in hardware, or 0 upon
* success.
*/
int clkdm_add_wkdep(struct clockdomain *clkdm1, struct clockdomain *clkdm2)
{
struct clkdm_dep *cd;
if (!cpu_is_omap24xx() && !cpu_is_omap34xx()) {
pr_err("clockdomain: %s/%s: %s: not yet implemented\n",
clkdm1->name, clkdm2->name, __func__);
return -EINVAL;
}
if (!clkdm1 || !clkdm2)
return -EINVAL;
cd = _clkdm_deps_lookup(clkdm2, clkdm1->wkdep_srcs);
if (IS_ERR(cd)) {
pr_debug("clockdomain: hardware cannot set/clear wake up of "
"%s when %s wakes up\n", clkdm1->name, clkdm2->name);
return PTR_ERR(cd);
}
OMAP clockdomains: add usecounting for wakeup and sleep dependencies Add usecounting for wakeup and sleep dependencies. In the current situation, if several functions add dependencies on the same clockdomains, when the first dependency removal function is called, the dependency will be incorrectly removed from the hardware. Add clkdm_clear_all_wkdeps() and clkdm_clear_all_sleepdeps(), which provide a fast and usecounting-consistent way to clear all hardware clockdomain dependencies, since accesses to these registers can be quite slow. pm{2,3}4xx.c has been updated to use these new functions. The original version of this patch did not touch these files, which previously wrote directly to the wkdep registers, and thus confused the usecounting code. This problem was found by Kevin Hilman <khilman@deeprootsystems.com>. N.B.: This patch introduces one significant functional difference over the previous pm34xx.c code: sleepdeps are now cleared during clockdomain initialization, whereas previously they were left untouched. This has been tested by Kevin and confirmed to work. The original version of this patch also did not take into consideration that some clockdomains do not have sleep or wakeup dependency sources, which caused NULL pointer dereferences. This problem was debugged and fixed by Kevin Hilman <khilman@deeprootsystems.com>. Signed-off-by: Paul Walmsley <paul@pwsan.com> Signed-off-by: Kevin Hilman <khilman@deeprootsystems.com> Cc: Jouni Högander <jouni.hogander@nokia.com>
2010-01-27 03:13:01 +00:00
if (atomic_inc_return(&cd->wkdep_usecount) == 1) {
pr_debug("clockdomain: hardware will wake up %s when %s wakes "
"up\n", clkdm1->name, clkdm2->name);
omap2_prm_set_mod_reg_bits((1 << clkdm2->dep_bit),
OMAP clockdomains: add usecounting for wakeup and sleep dependencies Add usecounting for wakeup and sleep dependencies. In the current situation, if several functions add dependencies on the same clockdomains, when the first dependency removal function is called, the dependency will be incorrectly removed from the hardware. Add clkdm_clear_all_wkdeps() and clkdm_clear_all_sleepdeps(), which provide a fast and usecounting-consistent way to clear all hardware clockdomain dependencies, since accesses to these registers can be quite slow. pm{2,3}4xx.c has been updated to use these new functions. The original version of this patch did not touch these files, which previously wrote directly to the wkdep registers, and thus confused the usecounting code. This problem was found by Kevin Hilman <khilman@deeprootsystems.com>. N.B.: This patch introduces one significant functional difference over the previous pm34xx.c code: sleepdeps are now cleared during clockdomain initialization, whereas previously they were left untouched. This has been tested by Kevin and confirmed to work. The original version of this patch also did not take into consideration that some clockdomains do not have sleep or wakeup dependency sources, which caused NULL pointer dereferences. This problem was debugged and fixed by Kevin Hilman <khilman@deeprootsystems.com>. Signed-off-by: Paul Walmsley <paul@pwsan.com> Signed-off-by: Kevin Hilman <khilman@deeprootsystems.com> Cc: Jouni Högander <jouni.hogander@nokia.com>
2010-01-27 03:13:01 +00:00
clkdm1->pwrdm.ptr->prcm_offs, PM_WKDEP);
}
return 0;
}
/**
* clkdm_del_wkdep - remove a wakeup dependency from clkdm2 to clkdm1
* @clkdm1: wake this struct clockdomain * up (dependent)
* @clkdm2: when this struct clockdomain * wakes up (source)
*
* Remove a wakeup dependency causing @clkdm1 to wake up when @clkdm2
* wakes up. Returns -EINVAL if presented with invalid clockdomain
* pointers, -ENOENT if @clkdm2 cannot wake up clkdm1 in hardware, or
* 0 upon success.
*/
int clkdm_del_wkdep(struct clockdomain *clkdm1, struct clockdomain *clkdm2)
{
struct clkdm_dep *cd;
if (!cpu_is_omap24xx() && !cpu_is_omap34xx()) {
pr_err("clockdomain: %s/%s: %s: not yet implemented\n",
clkdm1->name, clkdm2->name, __func__);
return -EINVAL;
}
if (!clkdm1 || !clkdm2)
return -EINVAL;
cd = _clkdm_deps_lookup(clkdm2, clkdm1->wkdep_srcs);
if (IS_ERR(cd)) {
pr_debug("clockdomain: hardware cannot set/clear wake up of "
"%s when %s wakes up\n", clkdm1->name, clkdm2->name);
return PTR_ERR(cd);
}
OMAP clockdomains: add usecounting for wakeup and sleep dependencies Add usecounting for wakeup and sleep dependencies. In the current situation, if several functions add dependencies on the same clockdomains, when the first dependency removal function is called, the dependency will be incorrectly removed from the hardware. Add clkdm_clear_all_wkdeps() and clkdm_clear_all_sleepdeps(), which provide a fast and usecounting-consistent way to clear all hardware clockdomain dependencies, since accesses to these registers can be quite slow. pm{2,3}4xx.c has been updated to use these new functions. The original version of this patch did not touch these files, which previously wrote directly to the wkdep registers, and thus confused the usecounting code. This problem was found by Kevin Hilman <khilman@deeprootsystems.com>. N.B.: This patch introduces one significant functional difference over the previous pm34xx.c code: sleepdeps are now cleared during clockdomain initialization, whereas previously they were left untouched. This has been tested by Kevin and confirmed to work. The original version of this patch also did not take into consideration that some clockdomains do not have sleep or wakeup dependency sources, which caused NULL pointer dereferences. This problem was debugged and fixed by Kevin Hilman <khilman@deeprootsystems.com>. Signed-off-by: Paul Walmsley <paul@pwsan.com> Signed-off-by: Kevin Hilman <khilman@deeprootsystems.com> Cc: Jouni Högander <jouni.hogander@nokia.com>
2010-01-27 03:13:01 +00:00
if (atomic_dec_return(&cd->wkdep_usecount) == 0) {
pr_debug("clockdomain: hardware will no longer wake up %s "
"after %s wakes up\n", clkdm1->name, clkdm2->name);
omap2_prm_clear_mod_reg_bits((1 << clkdm2->dep_bit),
OMAP clockdomains: add usecounting for wakeup and sleep dependencies Add usecounting for wakeup and sleep dependencies. In the current situation, if several functions add dependencies on the same clockdomains, when the first dependency removal function is called, the dependency will be incorrectly removed from the hardware. Add clkdm_clear_all_wkdeps() and clkdm_clear_all_sleepdeps(), which provide a fast and usecounting-consistent way to clear all hardware clockdomain dependencies, since accesses to these registers can be quite slow. pm{2,3}4xx.c has been updated to use these new functions. The original version of this patch did not touch these files, which previously wrote directly to the wkdep registers, and thus confused the usecounting code. This problem was found by Kevin Hilman <khilman@deeprootsystems.com>. N.B.: This patch introduces one significant functional difference over the previous pm34xx.c code: sleepdeps are now cleared during clockdomain initialization, whereas previously they were left untouched. This has been tested by Kevin and confirmed to work. The original version of this patch also did not take into consideration that some clockdomains do not have sleep or wakeup dependency sources, which caused NULL pointer dereferences. This problem was debugged and fixed by Kevin Hilman <khilman@deeprootsystems.com>. Signed-off-by: Paul Walmsley <paul@pwsan.com> Signed-off-by: Kevin Hilman <khilman@deeprootsystems.com> Cc: Jouni Högander <jouni.hogander@nokia.com>
2010-01-27 03:13:01 +00:00
clkdm1->pwrdm.ptr->prcm_offs, PM_WKDEP);
}
return 0;
}
/**
* clkdm_read_wkdep - read wakeup dependency state from clkdm2 to clkdm1
* @clkdm1: wake this struct clockdomain * up (dependent)
* @clkdm2: when this struct clockdomain * wakes up (source)
*
* Return 1 if a hardware wakeup dependency exists wherein @clkdm1 will be
* awoken when @clkdm2 wakes up; 0 if dependency is not set; -EINVAL
* if either clockdomain pointer is invalid; or -ENOENT if the hardware
* is incapable.
*
* REVISIT: Currently this function only represents software-controllable
* wakeup dependencies. Wakeup dependencies fixed in hardware are not
* yet handled here.
*/
int clkdm_read_wkdep(struct clockdomain *clkdm1, struct clockdomain *clkdm2)
{
struct clkdm_dep *cd;
if (!clkdm1 || !clkdm2)
return -EINVAL;
if (!cpu_is_omap24xx() && !cpu_is_omap34xx()) {
pr_err("clockdomain: %s/%s: %s: not yet implemented\n",
clkdm1->name, clkdm2->name, __func__);
return -EINVAL;
}
cd = _clkdm_deps_lookup(clkdm2, clkdm1->wkdep_srcs);
if (IS_ERR(cd)) {
pr_debug("clockdomain: hardware cannot set/clear wake up of "
"%s when %s wakes up\n", clkdm1->name, clkdm2->name);
return PTR_ERR(cd);
}
OMAP clockdomains: add usecounting for wakeup and sleep dependencies Add usecounting for wakeup and sleep dependencies. In the current situation, if several functions add dependencies on the same clockdomains, when the first dependency removal function is called, the dependency will be incorrectly removed from the hardware. Add clkdm_clear_all_wkdeps() and clkdm_clear_all_sleepdeps(), which provide a fast and usecounting-consistent way to clear all hardware clockdomain dependencies, since accesses to these registers can be quite slow. pm{2,3}4xx.c has been updated to use these new functions. The original version of this patch did not touch these files, which previously wrote directly to the wkdep registers, and thus confused the usecounting code. This problem was found by Kevin Hilman <khilman@deeprootsystems.com>. N.B.: This patch introduces one significant functional difference over the previous pm34xx.c code: sleepdeps are now cleared during clockdomain initialization, whereas previously they were left untouched. This has been tested by Kevin and confirmed to work. The original version of this patch also did not take into consideration that some clockdomains do not have sleep or wakeup dependency sources, which caused NULL pointer dereferences. This problem was debugged and fixed by Kevin Hilman <khilman@deeprootsystems.com>. Signed-off-by: Paul Walmsley <paul@pwsan.com> Signed-off-by: Kevin Hilman <khilman@deeprootsystems.com> Cc: Jouni Högander <jouni.hogander@nokia.com>
2010-01-27 03:13:01 +00:00
/* XXX It's faster to return the atomic wkdep_usecount */
return omap2_prm_read_mod_bits_shift(clkdm1->pwrdm.ptr->prcm_offs, PM_WKDEP,
(1 << clkdm2->dep_bit));
}
OMAP clockdomains: add usecounting for wakeup and sleep dependencies Add usecounting for wakeup and sleep dependencies. In the current situation, if several functions add dependencies on the same clockdomains, when the first dependency removal function is called, the dependency will be incorrectly removed from the hardware. Add clkdm_clear_all_wkdeps() and clkdm_clear_all_sleepdeps(), which provide a fast and usecounting-consistent way to clear all hardware clockdomain dependencies, since accesses to these registers can be quite slow. pm{2,3}4xx.c has been updated to use these new functions. The original version of this patch did not touch these files, which previously wrote directly to the wkdep registers, and thus confused the usecounting code. This problem was found by Kevin Hilman <khilman@deeprootsystems.com>. N.B.: This patch introduces one significant functional difference over the previous pm34xx.c code: sleepdeps are now cleared during clockdomain initialization, whereas previously they were left untouched. This has been tested by Kevin and confirmed to work. The original version of this patch also did not take into consideration that some clockdomains do not have sleep or wakeup dependency sources, which caused NULL pointer dereferences. This problem was debugged and fixed by Kevin Hilman <khilman@deeprootsystems.com>. Signed-off-by: Paul Walmsley <paul@pwsan.com> Signed-off-by: Kevin Hilman <khilman@deeprootsystems.com> Cc: Jouni Högander <jouni.hogander@nokia.com>
2010-01-27 03:13:01 +00:00
/**
* clkdm_clear_all_wkdeps - remove all wakeup dependencies from target clkdm
* @clkdm: struct clockdomain * to remove all wakeup dependencies from
*
* Remove all inter-clockdomain wakeup dependencies that could cause
* @clkdm to wake. Intended to be used during boot to initialize the
* PRCM to a known state, after all clockdomains are put into swsup idle
* and woken up. Returns -EINVAL if @clkdm pointer is invalid, or
* 0 upon success.
*/
int clkdm_clear_all_wkdeps(struct clockdomain *clkdm)
{
struct clkdm_dep *cd;
u32 mask = 0;
if (!cpu_is_omap24xx() && !cpu_is_omap34xx()) {
pr_err("clockdomain: %s: %s: not yet implemented\n",
clkdm->name, __func__);
return -EINVAL;
}
OMAP clockdomains: add usecounting for wakeup and sleep dependencies Add usecounting for wakeup and sleep dependencies. In the current situation, if several functions add dependencies on the same clockdomains, when the first dependency removal function is called, the dependency will be incorrectly removed from the hardware. Add clkdm_clear_all_wkdeps() and clkdm_clear_all_sleepdeps(), which provide a fast and usecounting-consistent way to clear all hardware clockdomain dependencies, since accesses to these registers can be quite slow. pm{2,3}4xx.c has been updated to use these new functions. The original version of this patch did not touch these files, which previously wrote directly to the wkdep registers, and thus confused the usecounting code. This problem was found by Kevin Hilman <khilman@deeprootsystems.com>. N.B.: This patch introduces one significant functional difference over the previous pm34xx.c code: sleepdeps are now cleared during clockdomain initialization, whereas previously they were left untouched. This has been tested by Kevin and confirmed to work. The original version of this patch also did not take into consideration that some clockdomains do not have sleep or wakeup dependency sources, which caused NULL pointer dereferences. This problem was debugged and fixed by Kevin Hilman <khilman@deeprootsystems.com>. Signed-off-by: Paul Walmsley <paul@pwsan.com> Signed-off-by: Kevin Hilman <khilman@deeprootsystems.com> Cc: Jouni Högander <jouni.hogander@nokia.com>
2010-01-27 03:13:01 +00:00
if (!clkdm)
return -EINVAL;
for (cd = clkdm->wkdep_srcs; cd && cd->clkdm_name; cd++) {
if (!omap_chip_is(cd->omap_chip))
continue;
if (!cd->clkdm && cd->clkdm_name)
cd->clkdm = _clkdm_lookup(cd->clkdm_name);
OMAP clockdomains: add usecounting for wakeup and sleep dependencies Add usecounting for wakeup and sleep dependencies. In the current situation, if several functions add dependencies on the same clockdomains, when the first dependency removal function is called, the dependency will be incorrectly removed from the hardware. Add clkdm_clear_all_wkdeps() and clkdm_clear_all_sleepdeps(), which provide a fast and usecounting-consistent way to clear all hardware clockdomain dependencies, since accesses to these registers can be quite slow. pm{2,3}4xx.c has been updated to use these new functions. The original version of this patch did not touch these files, which previously wrote directly to the wkdep registers, and thus confused the usecounting code. This problem was found by Kevin Hilman <khilman@deeprootsystems.com>. N.B.: This patch introduces one significant functional difference over the previous pm34xx.c code: sleepdeps are now cleared during clockdomain initialization, whereas previously they were left untouched. This has been tested by Kevin and confirmed to work. The original version of this patch also did not take into consideration that some clockdomains do not have sleep or wakeup dependency sources, which caused NULL pointer dereferences. This problem was debugged and fixed by Kevin Hilman <khilman@deeprootsystems.com>. Signed-off-by: Paul Walmsley <paul@pwsan.com> Signed-off-by: Kevin Hilman <khilman@deeprootsystems.com> Cc: Jouni Högander <jouni.hogander@nokia.com>
2010-01-27 03:13:01 +00:00
/* PRM accesses are slow, so minimize them */
mask |= 1 << cd->clkdm->dep_bit;
atomic_set(&cd->wkdep_usecount, 0);
}
omap2_prm_clear_mod_reg_bits(mask, clkdm->pwrdm.ptr->prcm_offs, PM_WKDEP);
OMAP clockdomains: add usecounting for wakeup and sleep dependencies Add usecounting for wakeup and sleep dependencies. In the current situation, if several functions add dependencies on the same clockdomains, when the first dependency removal function is called, the dependency will be incorrectly removed from the hardware. Add clkdm_clear_all_wkdeps() and clkdm_clear_all_sleepdeps(), which provide a fast and usecounting-consistent way to clear all hardware clockdomain dependencies, since accesses to these registers can be quite slow. pm{2,3}4xx.c has been updated to use these new functions. The original version of this patch did not touch these files, which previously wrote directly to the wkdep registers, and thus confused the usecounting code. This problem was found by Kevin Hilman <khilman@deeprootsystems.com>. N.B.: This patch introduces one significant functional difference over the previous pm34xx.c code: sleepdeps are now cleared during clockdomain initialization, whereas previously they were left untouched. This has been tested by Kevin and confirmed to work. The original version of this patch also did not take into consideration that some clockdomains do not have sleep or wakeup dependency sources, which caused NULL pointer dereferences. This problem was debugged and fixed by Kevin Hilman <khilman@deeprootsystems.com>. Signed-off-by: Paul Walmsley <paul@pwsan.com> Signed-off-by: Kevin Hilman <khilman@deeprootsystems.com> Cc: Jouni Högander <jouni.hogander@nokia.com>
2010-01-27 03:13:01 +00:00
return 0;
}
/**
* clkdm_add_sleepdep - add a sleep dependency from clkdm2 to clkdm1
* @clkdm1: prevent this struct clockdomain * from sleeping (dependent)
* @clkdm2: when this struct clockdomain * is active (source)
*
* Prevent @clkdm1 from automatically going inactive (and then to
* retention or off) if @clkdm2 is active. Returns -EINVAL if
* presented with invalid clockdomain pointers or called on a machine
* that does not support software-configurable hardware sleep
* dependencies, -ENOENT if the specified dependency cannot be set in
* hardware, or 0 upon success.
*/
int clkdm_add_sleepdep(struct clockdomain *clkdm1, struct clockdomain *clkdm2)
{
struct clkdm_dep *cd;
if (!cpu_is_omap34xx())
return -EINVAL;
if (!clkdm1 || !clkdm2)
return -EINVAL;
cd = _clkdm_deps_lookup(clkdm2, clkdm1->sleepdep_srcs);
if (IS_ERR(cd)) {
pr_debug("clockdomain: hardware cannot set/clear sleep "
"dependency affecting %s from %s\n", clkdm1->name,
clkdm2->name);
return PTR_ERR(cd);
}
OMAP clockdomains: add usecounting for wakeup and sleep dependencies Add usecounting for wakeup and sleep dependencies. In the current situation, if several functions add dependencies on the same clockdomains, when the first dependency removal function is called, the dependency will be incorrectly removed from the hardware. Add clkdm_clear_all_wkdeps() and clkdm_clear_all_sleepdeps(), which provide a fast and usecounting-consistent way to clear all hardware clockdomain dependencies, since accesses to these registers can be quite slow. pm{2,3}4xx.c has been updated to use these new functions. The original version of this patch did not touch these files, which previously wrote directly to the wkdep registers, and thus confused the usecounting code. This problem was found by Kevin Hilman <khilman@deeprootsystems.com>. N.B.: This patch introduces one significant functional difference over the previous pm34xx.c code: sleepdeps are now cleared during clockdomain initialization, whereas previously they were left untouched. This has been tested by Kevin and confirmed to work. The original version of this patch also did not take into consideration that some clockdomains do not have sleep or wakeup dependency sources, which caused NULL pointer dereferences. This problem was debugged and fixed by Kevin Hilman <khilman@deeprootsystems.com>. Signed-off-by: Paul Walmsley <paul@pwsan.com> Signed-off-by: Kevin Hilman <khilman@deeprootsystems.com> Cc: Jouni Högander <jouni.hogander@nokia.com>
2010-01-27 03:13:01 +00:00
if (atomic_inc_return(&cd->sleepdep_usecount) == 1) {
pr_debug("clockdomain: will prevent %s from sleeping if %s "
"is active\n", clkdm1->name, clkdm2->name);
omap2_cm_set_mod_reg_bits((1 << clkdm2->dep_bit),
OMAP clockdomains: add usecounting for wakeup and sleep dependencies Add usecounting for wakeup and sleep dependencies. In the current situation, if several functions add dependencies on the same clockdomains, when the first dependency removal function is called, the dependency will be incorrectly removed from the hardware. Add clkdm_clear_all_wkdeps() and clkdm_clear_all_sleepdeps(), which provide a fast and usecounting-consistent way to clear all hardware clockdomain dependencies, since accesses to these registers can be quite slow. pm{2,3}4xx.c has been updated to use these new functions. The original version of this patch did not touch these files, which previously wrote directly to the wkdep registers, and thus confused the usecounting code. This problem was found by Kevin Hilman <khilman@deeprootsystems.com>. N.B.: This patch introduces one significant functional difference over the previous pm34xx.c code: sleepdeps are now cleared during clockdomain initialization, whereas previously they were left untouched. This has been tested by Kevin and confirmed to work. The original version of this patch also did not take into consideration that some clockdomains do not have sleep or wakeup dependency sources, which caused NULL pointer dereferences. This problem was debugged and fixed by Kevin Hilman <khilman@deeprootsystems.com>. Signed-off-by: Paul Walmsley <paul@pwsan.com> Signed-off-by: Kevin Hilman <khilman@deeprootsystems.com> Cc: Jouni Högander <jouni.hogander@nokia.com>
2010-01-27 03:13:01 +00:00
clkdm1->pwrdm.ptr->prcm_offs,
OMAP3430_CM_SLEEPDEP);
}
return 0;
}
/**
* clkdm_del_sleepdep - remove a sleep dependency from clkdm2 to clkdm1
* @clkdm1: prevent this struct clockdomain * from sleeping (dependent)
* @clkdm2: when this struct clockdomain * is active (source)
*
* Allow @clkdm1 to automatically go inactive (and then to retention or
* off), independent of the activity state of @clkdm2. Returns -EINVAL
* if presented with invalid clockdomain pointers or called on a machine
* that does not support software-configurable hardware sleep dependencies,
* -ENOENT if the specified dependency cannot be cleared in hardware, or
* 0 upon success.
*/
int clkdm_del_sleepdep(struct clockdomain *clkdm1, struct clockdomain *clkdm2)
{
struct clkdm_dep *cd;
if (!cpu_is_omap34xx())
return -EINVAL;
if (!clkdm1 || !clkdm2)
return -EINVAL;
cd = _clkdm_deps_lookup(clkdm2, clkdm1->sleepdep_srcs);
if (IS_ERR(cd)) {
pr_debug("clockdomain: hardware cannot set/clear sleep "
"dependency affecting %s from %s\n", clkdm1->name,
clkdm2->name);
return PTR_ERR(cd);
}
OMAP clockdomains: add usecounting for wakeup and sleep dependencies Add usecounting for wakeup and sleep dependencies. In the current situation, if several functions add dependencies on the same clockdomains, when the first dependency removal function is called, the dependency will be incorrectly removed from the hardware. Add clkdm_clear_all_wkdeps() and clkdm_clear_all_sleepdeps(), which provide a fast and usecounting-consistent way to clear all hardware clockdomain dependencies, since accesses to these registers can be quite slow. pm{2,3}4xx.c has been updated to use these new functions. The original version of this patch did not touch these files, which previously wrote directly to the wkdep registers, and thus confused the usecounting code. This problem was found by Kevin Hilman <khilman@deeprootsystems.com>. N.B.: This patch introduces one significant functional difference over the previous pm34xx.c code: sleepdeps are now cleared during clockdomain initialization, whereas previously they were left untouched. This has been tested by Kevin and confirmed to work. The original version of this patch also did not take into consideration that some clockdomains do not have sleep or wakeup dependency sources, which caused NULL pointer dereferences. This problem was debugged and fixed by Kevin Hilman <khilman@deeprootsystems.com>. Signed-off-by: Paul Walmsley <paul@pwsan.com> Signed-off-by: Kevin Hilman <khilman@deeprootsystems.com> Cc: Jouni Högander <jouni.hogander@nokia.com>
2010-01-27 03:13:01 +00:00
if (atomic_dec_return(&cd->sleepdep_usecount) == 0) {
pr_debug("clockdomain: will no longer prevent %s from "
"sleeping if %s is active\n", clkdm1->name,
clkdm2->name);
omap2_cm_clear_mod_reg_bits((1 << clkdm2->dep_bit),
OMAP clockdomains: add usecounting for wakeup and sleep dependencies Add usecounting for wakeup and sleep dependencies. In the current situation, if several functions add dependencies on the same clockdomains, when the first dependency removal function is called, the dependency will be incorrectly removed from the hardware. Add clkdm_clear_all_wkdeps() and clkdm_clear_all_sleepdeps(), which provide a fast and usecounting-consistent way to clear all hardware clockdomain dependencies, since accesses to these registers can be quite slow. pm{2,3}4xx.c has been updated to use these new functions. The original version of this patch did not touch these files, which previously wrote directly to the wkdep registers, and thus confused the usecounting code. This problem was found by Kevin Hilman <khilman@deeprootsystems.com>. N.B.: This patch introduces one significant functional difference over the previous pm34xx.c code: sleepdeps are now cleared during clockdomain initialization, whereas previously they were left untouched. This has been tested by Kevin and confirmed to work. The original version of this patch also did not take into consideration that some clockdomains do not have sleep or wakeup dependency sources, which caused NULL pointer dereferences. This problem was debugged and fixed by Kevin Hilman <khilman@deeprootsystems.com>. Signed-off-by: Paul Walmsley <paul@pwsan.com> Signed-off-by: Kevin Hilman <khilman@deeprootsystems.com> Cc: Jouni Högander <jouni.hogander@nokia.com>
2010-01-27 03:13:01 +00:00
clkdm1->pwrdm.ptr->prcm_offs,
OMAP3430_CM_SLEEPDEP);
}
return 0;
}
/**
* clkdm_read_sleepdep - read sleep dependency state from clkdm2 to clkdm1
* @clkdm1: prevent this struct clockdomain * from sleeping (dependent)
* @clkdm2: when this struct clockdomain * is active (source)
*
* Return 1 if a hardware sleep dependency exists wherein @clkdm1 will
* not be allowed to automatically go inactive if @clkdm2 is active;
* 0 if @clkdm1's automatic power state inactivity transition is independent
* of @clkdm2's; -EINVAL if either clockdomain pointer is invalid or called
* on a machine that does not support software-configurable hardware sleep
* dependencies; or -ENOENT if the hardware is incapable.
*
* REVISIT: Currently this function only represents software-controllable
* sleep dependencies. Sleep dependencies fixed in hardware are not
* yet handled here.
*/
int clkdm_read_sleepdep(struct clockdomain *clkdm1, struct clockdomain *clkdm2)
{
struct clkdm_dep *cd;
if (!cpu_is_omap34xx())
return -EINVAL;
if (!clkdm1 || !clkdm2)
return -EINVAL;
cd = _clkdm_deps_lookup(clkdm2, clkdm1->sleepdep_srcs);
if (IS_ERR(cd)) {
pr_debug("clockdomain: hardware cannot set/clear sleep "
"dependency affecting %s from %s\n", clkdm1->name,
clkdm2->name);
return PTR_ERR(cd);
}
OMAP clockdomains: add usecounting for wakeup and sleep dependencies Add usecounting for wakeup and sleep dependencies. In the current situation, if several functions add dependencies on the same clockdomains, when the first dependency removal function is called, the dependency will be incorrectly removed from the hardware. Add clkdm_clear_all_wkdeps() and clkdm_clear_all_sleepdeps(), which provide a fast and usecounting-consistent way to clear all hardware clockdomain dependencies, since accesses to these registers can be quite slow. pm{2,3}4xx.c has been updated to use these new functions. The original version of this patch did not touch these files, which previously wrote directly to the wkdep registers, and thus confused the usecounting code. This problem was found by Kevin Hilman <khilman@deeprootsystems.com>. N.B.: This patch introduces one significant functional difference over the previous pm34xx.c code: sleepdeps are now cleared during clockdomain initialization, whereas previously they were left untouched. This has been tested by Kevin and confirmed to work. The original version of this patch also did not take into consideration that some clockdomains do not have sleep or wakeup dependency sources, which caused NULL pointer dereferences. This problem was debugged and fixed by Kevin Hilman <khilman@deeprootsystems.com>. Signed-off-by: Paul Walmsley <paul@pwsan.com> Signed-off-by: Kevin Hilman <khilman@deeprootsystems.com> Cc: Jouni Högander <jouni.hogander@nokia.com>
2010-01-27 03:13:01 +00:00
/* XXX It's faster to return the atomic sleepdep_usecount */
return omap2_prm_read_mod_bits_shift(clkdm1->pwrdm.ptr->prcm_offs,
OMAP3430_CM_SLEEPDEP,
(1 << clkdm2->dep_bit));
}
OMAP clockdomains: add usecounting for wakeup and sleep dependencies Add usecounting for wakeup and sleep dependencies. In the current situation, if several functions add dependencies on the same clockdomains, when the first dependency removal function is called, the dependency will be incorrectly removed from the hardware. Add clkdm_clear_all_wkdeps() and clkdm_clear_all_sleepdeps(), which provide a fast and usecounting-consistent way to clear all hardware clockdomain dependencies, since accesses to these registers can be quite slow. pm{2,3}4xx.c has been updated to use these new functions. The original version of this patch did not touch these files, which previously wrote directly to the wkdep registers, and thus confused the usecounting code. This problem was found by Kevin Hilman <khilman@deeprootsystems.com>. N.B.: This patch introduces one significant functional difference over the previous pm34xx.c code: sleepdeps are now cleared during clockdomain initialization, whereas previously they were left untouched. This has been tested by Kevin and confirmed to work. The original version of this patch also did not take into consideration that some clockdomains do not have sleep or wakeup dependency sources, which caused NULL pointer dereferences. This problem was debugged and fixed by Kevin Hilman <khilman@deeprootsystems.com>. Signed-off-by: Paul Walmsley <paul@pwsan.com> Signed-off-by: Kevin Hilman <khilman@deeprootsystems.com> Cc: Jouni Högander <jouni.hogander@nokia.com>
2010-01-27 03:13:01 +00:00
/**
* clkdm_clear_all_sleepdeps - remove all sleep dependencies from target clkdm
* @clkdm: struct clockdomain * to remove all sleep dependencies from
*
* Remove all inter-clockdomain sleep dependencies that could prevent
* @clkdm from idling. Intended to be used during boot to initialize the
* PRCM to a known state, after all clockdomains are put into swsup idle
* and woken up. Returns -EINVAL if @clkdm pointer is invalid, or
* 0 upon success.
*/
int clkdm_clear_all_sleepdeps(struct clockdomain *clkdm)
{
struct clkdm_dep *cd;
u32 mask = 0;
if (!cpu_is_omap34xx())
return -EINVAL;
if (!clkdm)
return -EINVAL;
for (cd = clkdm->sleepdep_srcs; cd && cd->clkdm_name; cd++) {
if (!omap_chip_is(cd->omap_chip))
continue;
if (!cd->clkdm && cd->clkdm_name)
cd->clkdm = _clkdm_lookup(cd->clkdm_name);
OMAP clockdomains: add usecounting for wakeup and sleep dependencies Add usecounting for wakeup and sleep dependencies. In the current situation, if several functions add dependencies on the same clockdomains, when the first dependency removal function is called, the dependency will be incorrectly removed from the hardware. Add clkdm_clear_all_wkdeps() and clkdm_clear_all_sleepdeps(), which provide a fast and usecounting-consistent way to clear all hardware clockdomain dependencies, since accesses to these registers can be quite slow. pm{2,3}4xx.c has been updated to use these new functions. The original version of this patch did not touch these files, which previously wrote directly to the wkdep registers, and thus confused the usecounting code. This problem was found by Kevin Hilman <khilman@deeprootsystems.com>. N.B.: This patch introduces one significant functional difference over the previous pm34xx.c code: sleepdeps are now cleared during clockdomain initialization, whereas previously they were left untouched. This has been tested by Kevin and confirmed to work. The original version of this patch also did not take into consideration that some clockdomains do not have sleep or wakeup dependency sources, which caused NULL pointer dereferences. This problem was debugged and fixed by Kevin Hilman <khilman@deeprootsystems.com>. Signed-off-by: Paul Walmsley <paul@pwsan.com> Signed-off-by: Kevin Hilman <khilman@deeprootsystems.com> Cc: Jouni Högander <jouni.hogander@nokia.com>
2010-01-27 03:13:01 +00:00
/* PRM accesses are slow, so minimize them */
mask |= 1 << cd->clkdm->dep_bit;
atomic_set(&cd->sleepdep_usecount, 0);
}
omap2_prm_clear_mod_reg_bits(mask, clkdm->pwrdm.ptr->prcm_offs,
OMAP clockdomains: add usecounting for wakeup and sleep dependencies Add usecounting for wakeup and sleep dependencies. In the current situation, if several functions add dependencies on the same clockdomains, when the first dependency removal function is called, the dependency will be incorrectly removed from the hardware. Add clkdm_clear_all_wkdeps() and clkdm_clear_all_sleepdeps(), which provide a fast and usecounting-consistent way to clear all hardware clockdomain dependencies, since accesses to these registers can be quite slow. pm{2,3}4xx.c has been updated to use these new functions. The original version of this patch did not touch these files, which previously wrote directly to the wkdep registers, and thus confused the usecounting code. This problem was found by Kevin Hilman <khilman@deeprootsystems.com>. N.B.: This patch introduces one significant functional difference over the previous pm34xx.c code: sleepdeps are now cleared during clockdomain initialization, whereas previously they were left untouched. This has been tested by Kevin and confirmed to work. The original version of this patch also did not take into consideration that some clockdomains do not have sleep or wakeup dependency sources, which caused NULL pointer dereferences. This problem was debugged and fixed by Kevin Hilman <khilman@deeprootsystems.com>. Signed-off-by: Paul Walmsley <paul@pwsan.com> Signed-off-by: Kevin Hilman <khilman@deeprootsystems.com> Cc: Jouni Högander <jouni.hogander@nokia.com>
2010-01-27 03:13:01 +00:00
OMAP3430_CM_SLEEPDEP);
return 0;
}
/**
* omap2_clkdm_sleep - force clockdomain sleep transition
* @clkdm: struct clockdomain *
*
* Instruct the CM to force a sleep transition on the specified
* clockdomain @clkdm. Returns -EINVAL if @clkdm is NULL or if
* clockdomain does not support software-initiated sleep; 0 upon
* success.
*/
int omap2_clkdm_sleep(struct clockdomain *clkdm)
{
if (!clkdm)
return -EINVAL;
if (!(clkdm->flags & CLKDM_CAN_FORCE_SLEEP)) {
pr_debug("clockdomain: %s does not support forcing "
"sleep via software\n", clkdm->name);
return -EINVAL;
}
pr_debug("clockdomain: forcing sleep on %s\n", clkdm->name);
if (cpu_is_omap24xx()) {
omap2_cm_set_mod_reg_bits(OMAP24XX_FORCESTATE_MASK,
clkdm->pwrdm.ptr->prcm_offs, OMAP2_PM_PWSTCTRL);
} else if (cpu_is_omap34xx()) {
omap3xxx_cm_clkdm_force_sleep(clkdm->pwrdm.ptr->prcm_offs,
clkdm->clktrctrl_mask);
} else if (cpu_is_omap44xx()) {
omap4_cminst_clkdm_force_sleep(clkdm->prcm_partition,
clkdm->cm_inst,
clkdm->clkdm_offs);
} else {
BUG();
};
return 0;
}
/**
* omap2_clkdm_wakeup - force clockdomain wakeup transition
* @clkdm: struct clockdomain *
*
* Instruct the CM to force a wakeup transition on the specified
* clockdomain @clkdm. Returns -EINVAL if @clkdm is NULL or if the
* clockdomain does not support software-controlled wakeup; 0 upon
* success.
*/
int omap2_clkdm_wakeup(struct clockdomain *clkdm)
{
if (!clkdm)
return -EINVAL;
if (!(clkdm->flags & CLKDM_CAN_FORCE_WAKEUP)) {
pr_debug("clockdomain: %s does not support forcing "
"wakeup via software\n", clkdm->name);
return -EINVAL;
}
pr_debug("clockdomain: forcing wakeup on %s\n", clkdm->name);
if (cpu_is_omap24xx()) {
omap2_cm_clear_mod_reg_bits(OMAP24XX_FORCESTATE_MASK,
clkdm->pwrdm.ptr->prcm_offs, OMAP2_PM_PWSTCTRL);
} else if (cpu_is_omap34xx()) {
omap3xxx_cm_clkdm_force_wakeup(clkdm->pwrdm.ptr->prcm_offs,
clkdm->clktrctrl_mask);
} else if (cpu_is_omap44xx()) {
omap4_cminst_clkdm_force_wakeup(clkdm->prcm_partition,
clkdm->cm_inst,
clkdm->clkdm_offs);
} else {
BUG();
};
return 0;
}
/**
* omap2_clkdm_allow_idle - enable hwsup idle transitions for clkdm
* @clkdm: struct clockdomain *
*
* Allow the hardware to automatically switch the clockdomain @clkdm into
* active or idle states, as needed by downstream clocks. If the
* clockdomain has any downstream clocks enabled in the clock
* framework, wkdep/sleepdep autodependencies are added; this is so
* device drivers can read and write to the device. No return value.
*/
void omap2_clkdm_allow_idle(struct clockdomain *clkdm)
{
if (!clkdm)
return;
if (!(clkdm->flags & CLKDM_CAN_ENABLE_AUTO)) {
pr_debug("clock: automatic idle transitions cannot be enabled "
"on clockdomain %s\n", clkdm->name);
return;
}
pr_debug("clockdomain: enabling automatic idle transitions for %s\n",
clkdm->name);
/*
* XXX This should be removed once TI adds wakeup/sleep
* dependency code and data for OMAP4.
*/
if (cpu_is_omap44xx()) {
pr_err("clockdomain: %s: OMAP4 wakeup/sleep dependency support: not yet implemented\n", clkdm->name);
} else {
if (atomic_read(&clkdm->usecount) > 0)
_clkdm_add_autodeps(clkdm);
}
_enable_hwsup(clkdm);
pwrdm_clkdm_state_switch(clkdm);
}
/**
* omap2_clkdm_deny_idle - disable hwsup idle transitions for clkdm
* @clkdm: struct clockdomain *
*
* Prevent the hardware from automatically switching the clockdomain
* @clkdm into inactive or idle states. If the clockdomain has
* downstream clocks enabled in the clock framework, wkdep/sleepdep
* autodependencies are removed. No return value.
*/
void omap2_clkdm_deny_idle(struct clockdomain *clkdm)
{
if (!clkdm)
return;
if (!(clkdm->flags & CLKDM_CAN_DISABLE_AUTO)) {
pr_debug("clockdomain: automatic idle transitions cannot be "
"disabled on %s\n", clkdm->name);
return;
}
pr_debug("clockdomain: disabling automatic idle transitions for %s\n",
clkdm->name);
_disable_hwsup(clkdm);
/*
* XXX This should be removed once TI adds wakeup/sleep
* dependency code and data for OMAP4.
*/
if (cpu_is_omap44xx()) {
pr_err("clockdomain: %s: OMAP4 wakeup/sleep dependency support: not yet implemented\n", clkdm->name);
} else {
if (atomic_read(&clkdm->usecount) > 0)
_clkdm_del_autodeps(clkdm);
}
}
/* Clockdomain-to-clock framework interface code */
/**
* omap2_clkdm_clk_enable - add an enabled downstream clock to this clkdm
* @clkdm: struct clockdomain *
* @clk: struct clk * of the enabled downstream clock
*
* Increment the usecount of the clockdomain @clkdm and ensure that it
* is awake before @clk is enabled. Intended to be called by
* clk_enable() code. If the clockdomain is in software-supervised
* idle mode, force the clockdomain to wake. If the clockdomain is in
* hardware-supervised idle mode, add clkdm-pwrdm autodependencies, to
* ensure that devices in the clockdomain can be read from/written to
* by on-chip processors. Returns -EINVAL if passed null pointers;
* returns 0 upon success or if the clockdomain is in hwsup idle mode.
*/
int omap2_clkdm_clk_enable(struct clockdomain *clkdm, struct clk *clk)
{
bool hwsup = false;
/*
* XXX Rewrite this code to maintain a list of enabled
* downstream clocks for debugging purposes?
*/
OMAP2+ clock: revise omap2_clk_{disable,enable}() Simplify the code in the omap2_clk_disable() and omap2_clk_enable() functions, reducing levels of indentation. This makes the code easier to read. Add some additional debugging pr_debug()s here also to help others understand what is going on. Revise the omap2_clk_disable() logic so that it now attempts to disable the clock's clockdomain before recursing up the clock tree. Simultaneously, ensure that omap2_clk_enable() is called on parent clocks first, before enabling the clockdomain. This ensures that a parent clock's clockdomain is enabled before the child clock's clockdomain. These sequences should be the inverse of each other. Revise the omap2_clk_enable() logic so that it now cleans up after itself upon encountering an error. Previously, an error enabling a parent clock could have resulted in inconsistent usecounts on the enclosing clockdomain. Remove the trivial _omap2_clk_disable() and _omap2_clk_enable() static functions, and replace it with the clkops calls that they were executing. For all this to work, the clockdomain omap2_clkdm_clk_enable() and omap2_clkdm_clk_disable() code must not return an error on clockdomains without CLKSTCTRL registers; so modify those functions to simply return 0 in that case. While here, add some basic kerneldoc documentation on both functions, and get rid of some old non-CodingStyle-compliant comments that have existed since the dawn of time (at least, the OMAP clock framework's time). Signed-off-by: Paul Walmsley <paul@pwsan.com> Cc: Richard Woodruff <r-woodruff2@ti.com> Cc: Rajendra Nayak <rnayak@ti.com>
2010-02-23 05:09:38 +00:00
if (!clkdm || !clk)
return -EINVAL;
if (atomic_inc_return(&clkdm->usecount) > 1)
return 0;
/* Clockdomain now has one enabled downstream clock */
pr_debug("clockdomain: clkdm %s: clk %s now enabled\n", clkdm->name,
clk->name);
if (cpu_is_omap24xx() || cpu_is_omap34xx()) {
OMAP2+ clock: revise omap2_clk_{disable,enable}() Simplify the code in the omap2_clk_disable() and omap2_clk_enable() functions, reducing levels of indentation. This makes the code easier to read. Add some additional debugging pr_debug()s here also to help others understand what is going on. Revise the omap2_clk_disable() logic so that it now attempts to disable the clock's clockdomain before recursing up the clock tree. Simultaneously, ensure that omap2_clk_enable() is called on parent clocks first, before enabling the clockdomain. This ensures that a parent clock's clockdomain is enabled before the child clock's clockdomain. These sequences should be the inverse of each other. Revise the omap2_clk_enable() logic so that it now cleans up after itself upon encountering an error. Previously, an error enabling a parent clock could have resulted in inconsistent usecounts on the enclosing clockdomain. Remove the trivial _omap2_clk_disable() and _omap2_clk_enable() static functions, and replace it with the clkops calls that they were executing. For all this to work, the clockdomain omap2_clkdm_clk_enable() and omap2_clkdm_clk_disable() code must not return an error on clockdomains without CLKSTCTRL registers; so modify those functions to simply return 0 in that case. While here, add some basic kerneldoc documentation on both functions, and get rid of some old non-CodingStyle-compliant comments that have existed since the dawn of time (at least, the OMAP clock framework's time). Signed-off-by: Paul Walmsley <paul@pwsan.com> Cc: Richard Woodruff <r-woodruff2@ti.com> Cc: Rajendra Nayak <rnayak@ti.com>
2010-02-23 05:09:38 +00:00
if (!clkdm->clktrctrl_mask)
return 0;
hwsup = omap2_cm_is_clkdm_in_hwsup(clkdm->pwrdm.ptr->prcm_offs,
clkdm->clktrctrl_mask);
} else if (cpu_is_omap44xx()) {
hwsup = omap4_cminst_is_clkdm_in_hwsup(clkdm->prcm_partition,
clkdm->cm_inst,
clkdm->clkdm_offs);
}
if (hwsup) {
/* Disable HW transitions when we are changing deps */
_disable_hwsup(clkdm);
_clkdm_add_autodeps(clkdm);
_enable_hwsup(clkdm);
} else {
omap2_clkdm_wakeup(clkdm);
}
pwrdm_wait_transition(clkdm->pwrdm.ptr);
pwrdm_clkdm_state_switch(clkdm);
return 0;
}
/**
* omap2_clkdm_clk_disable - remove an enabled downstream clock from this clkdm
* @clkdm: struct clockdomain *
* @clk: struct clk * of the disabled downstream clock
*
* Decrement the usecount of this clockdomain @clkdm when @clk is
* disabled. Intended to be called by clk_disable() code. If the
* clockdomain usecount goes to 0, put the clockdomain to sleep
* (software-supervised mode) or remove the clkdm autodependencies
* (hardware-supervised mode). Returns -EINVAL if passed null
* pointers; -ERANGE if the @clkdm usecount underflows and debugging
* is enabled; or returns 0 upon success or if the clockdomain is in
* hwsup idle mode.
*/
int omap2_clkdm_clk_disable(struct clockdomain *clkdm, struct clk *clk)
{
bool hwsup = false;
/*
* XXX Rewrite this code to maintain a list of enabled
* downstream clocks for debugging purposes?
*/
OMAP2+ clock: revise omap2_clk_{disable,enable}() Simplify the code in the omap2_clk_disable() and omap2_clk_enable() functions, reducing levels of indentation. This makes the code easier to read. Add some additional debugging pr_debug()s here also to help others understand what is going on. Revise the omap2_clk_disable() logic so that it now attempts to disable the clock's clockdomain before recursing up the clock tree. Simultaneously, ensure that omap2_clk_enable() is called on parent clocks first, before enabling the clockdomain. This ensures that a parent clock's clockdomain is enabled before the child clock's clockdomain. These sequences should be the inverse of each other. Revise the omap2_clk_enable() logic so that it now cleans up after itself upon encountering an error. Previously, an error enabling a parent clock could have resulted in inconsistent usecounts on the enclosing clockdomain. Remove the trivial _omap2_clk_disable() and _omap2_clk_enable() static functions, and replace it with the clkops calls that they were executing. For all this to work, the clockdomain omap2_clkdm_clk_enable() and omap2_clkdm_clk_disable() code must not return an error on clockdomains without CLKSTCTRL registers; so modify those functions to simply return 0 in that case. While here, add some basic kerneldoc documentation on both functions, and get rid of some old non-CodingStyle-compliant comments that have existed since the dawn of time (at least, the OMAP clock framework's time). Signed-off-by: Paul Walmsley <paul@pwsan.com> Cc: Richard Woodruff <r-woodruff2@ti.com> Cc: Rajendra Nayak <rnayak@ti.com>
2010-02-23 05:09:38 +00:00
if (!clkdm || !clk)
return -EINVAL;
#ifdef DEBUG
if (atomic_read(&clkdm->usecount) == 0) {
WARN_ON(1); /* underflow */
return -ERANGE;
}
#endif
if (atomic_dec_return(&clkdm->usecount) > 0)
return 0;
/* All downstream clocks of this clockdomain are now disabled */
pr_debug("clockdomain: clkdm %s: clk %s now disabled\n", clkdm->name,
clk->name);
if (cpu_is_omap24xx() || cpu_is_omap34xx()) {
if (!clkdm->clktrctrl_mask)
return 0;
hwsup = omap2_cm_is_clkdm_in_hwsup(clkdm->pwrdm.ptr->prcm_offs,
clkdm->clktrctrl_mask);
OMAP2+ clock: revise omap2_clk_{disable,enable}() Simplify the code in the omap2_clk_disable() and omap2_clk_enable() functions, reducing levels of indentation. This makes the code easier to read. Add some additional debugging pr_debug()s here also to help others understand what is going on. Revise the omap2_clk_disable() logic so that it now attempts to disable the clock's clockdomain before recursing up the clock tree. Simultaneously, ensure that omap2_clk_enable() is called on parent clocks first, before enabling the clockdomain. This ensures that a parent clock's clockdomain is enabled before the child clock's clockdomain. These sequences should be the inverse of each other. Revise the omap2_clk_enable() logic so that it now cleans up after itself upon encountering an error. Previously, an error enabling a parent clock could have resulted in inconsistent usecounts on the enclosing clockdomain. Remove the trivial _omap2_clk_disable() and _omap2_clk_enable() static functions, and replace it with the clkops calls that they were executing. For all this to work, the clockdomain omap2_clkdm_clk_enable() and omap2_clkdm_clk_disable() code must not return an error on clockdomains without CLKSTCTRL registers; so modify those functions to simply return 0 in that case. While here, add some basic kerneldoc documentation on both functions, and get rid of some old non-CodingStyle-compliant comments that have existed since the dawn of time (at least, the OMAP clock framework's time). Signed-off-by: Paul Walmsley <paul@pwsan.com> Cc: Richard Woodruff <r-woodruff2@ti.com> Cc: Rajendra Nayak <rnayak@ti.com>
2010-02-23 05:09:38 +00:00
} else if (cpu_is_omap44xx()) {
hwsup = omap4_cminst_is_clkdm_in_hwsup(clkdm->prcm_partition,
clkdm->cm_inst,
clkdm->clkdm_offs);
}
if (hwsup) {
/* Disable HW transitions when we are changing deps */
_disable_hwsup(clkdm);
_clkdm_del_autodeps(clkdm);
_enable_hwsup(clkdm);
} else {
omap2_clkdm_sleep(clkdm);
}
pwrdm_clkdm_state_switch(clkdm);
return 0;
}