537 lines
15 KiB
C
537 lines
15 KiB
C
/*
|
|
* OMAP2+ common Power & Reset Management (PRM) IP block functions
|
|
*
|
|
* Copyright (C) 2011 Texas Instruments, Inc.
|
|
* Tero Kristo <t-kristo@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.
|
|
*
|
|
*
|
|
* For historical purposes, the API used to configure the PRM
|
|
* interrupt handler refers to it as the "PRCM interrupt." The
|
|
* underlying registers are located in the PRM on OMAP3/4.
|
|
*
|
|
* XXX This code should eventually be moved to a PRM driver.
|
|
*/
|
|
|
|
#include <linux/kernel.h>
|
|
#include <linux/module.h>
|
|
#include <linux/init.h>
|
|
#include <linux/io.h>
|
|
#include <linux/irq.h>
|
|
#include <linux/interrupt.h>
|
|
#include <linux/slab.h>
|
|
#include <linux/of.h>
|
|
#include <linux/of_address.h>
|
|
#include <linux/clk-provider.h>
|
|
#include <linux/clk/ti.h>
|
|
|
|
#include "soc.h"
|
|
#include "prm2xxx_3xxx.h"
|
|
#include "prm2xxx.h"
|
|
#include "prm3xxx.h"
|
|
#include "prm44xx.h"
|
|
#include "common.h"
|
|
#include "clock.h"
|
|
|
|
/*
|
|
* OMAP_PRCM_MAX_NR_PENDING_REG: maximum number of PRM_IRQ*_MPU regs
|
|
* XXX this is technically not needed, since
|
|
* omap_prcm_register_chain_handler() could allocate this based on the
|
|
* actual amount of memory needed for the SoC
|
|
*/
|
|
#define OMAP_PRCM_MAX_NR_PENDING_REG 2
|
|
|
|
/*
|
|
* prcm_irq_chips: an array of all of the "generic IRQ chips" in use
|
|
* by the PRCM interrupt handler code. There will be one 'chip' per
|
|
* PRM_{IRQSTATUS,IRQENABLE}_MPU register pair. (So OMAP3 will have
|
|
* one "chip" and OMAP4 will have two.)
|
|
*/
|
|
static struct irq_chip_generic **prcm_irq_chips;
|
|
|
|
/*
|
|
* prcm_irq_setup: the PRCM IRQ parameters for the hardware the code
|
|
* is currently running on. Defined and passed by initialization code
|
|
* that calls omap_prcm_register_chain_handler().
|
|
*/
|
|
static struct omap_prcm_irq_setup *prcm_irq_setup;
|
|
|
|
/* prm_base: base virtual address of the PRM IP block */
|
|
void __iomem *prm_base;
|
|
|
|
u16 prm_features;
|
|
|
|
/*
|
|
* prm_ll_data: function pointers to SoC-specific implementations of
|
|
* common PRM functions
|
|
*/
|
|
static struct prm_ll_data null_prm_ll_data;
|
|
static struct prm_ll_data *prm_ll_data = &null_prm_ll_data;
|
|
|
|
/* Private functions */
|
|
|
|
/*
|
|
* Move priority events from events to priority_events array
|
|
*/
|
|
static void omap_prcm_events_filter_priority(unsigned long *events,
|
|
unsigned long *priority_events)
|
|
{
|
|
int i;
|
|
|
|
for (i = 0; i < prcm_irq_setup->nr_regs; i++) {
|
|
priority_events[i] =
|
|
events[i] & prcm_irq_setup->priority_mask[i];
|
|
events[i] ^= priority_events[i];
|
|
}
|
|
}
|
|
|
|
/*
|
|
* PRCM Interrupt Handler
|
|
*
|
|
* This is a common handler for the OMAP PRCM interrupts. Pending
|
|
* interrupts are detected by a call to prcm_pending_events and
|
|
* dispatched accordingly. Clearing of the wakeup events should be
|
|
* done by the SoC specific individual handlers.
|
|
*/
|
|
static void omap_prcm_irq_handler(unsigned int irq, struct irq_desc *desc)
|
|
{
|
|
unsigned long pending[OMAP_PRCM_MAX_NR_PENDING_REG];
|
|
unsigned long priority_pending[OMAP_PRCM_MAX_NR_PENDING_REG];
|
|
struct irq_chip *chip = irq_desc_get_chip(desc);
|
|
unsigned int virtirq;
|
|
int nr_irq = prcm_irq_setup->nr_regs * 32;
|
|
|
|
/*
|
|
* If we are suspended, mask all interrupts from PRCM level,
|
|
* this does not ack them, and they will be pending until we
|
|
* re-enable the interrupts, at which point the
|
|
* omap_prcm_irq_handler will be executed again. The
|
|
* _save_and_clear_irqen() function must ensure that the PRM
|
|
* write to disable all IRQs has reached the PRM before
|
|
* returning, or spurious PRCM interrupts may occur during
|
|
* suspend.
|
|
*/
|
|
if (prcm_irq_setup->suspended) {
|
|
prcm_irq_setup->save_and_clear_irqen(prcm_irq_setup->saved_mask);
|
|
prcm_irq_setup->suspend_save_flag = true;
|
|
}
|
|
|
|
/*
|
|
* Loop until all pending irqs are handled, since
|
|
* generic_handle_irq() can cause new irqs to come
|
|
*/
|
|
while (!prcm_irq_setup->suspended) {
|
|
prcm_irq_setup->read_pending_irqs(pending);
|
|
|
|
/* No bit set, then all IRQs are handled */
|
|
if (find_first_bit(pending, nr_irq) >= nr_irq)
|
|
break;
|
|
|
|
omap_prcm_events_filter_priority(pending, priority_pending);
|
|
|
|
/*
|
|
* Loop on all currently pending irqs so that new irqs
|
|
* cannot starve previously pending irqs
|
|
*/
|
|
|
|
/* Serve priority events first */
|
|
for_each_set_bit(virtirq, priority_pending, nr_irq)
|
|
generic_handle_irq(prcm_irq_setup->base_irq + virtirq);
|
|
|
|
/* Serve normal events next */
|
|
for_each_set_bit(virtirq, pending, nr_irq)
|
|
generic_handle_irq(prcm_irq_setup->base_irq + virtirq);
|
|
}
|
|
if (chip->irq_ack)
|
|
chip->irq_ack(&desc->irq_data);
|
|
if (chip->irq_eoi)
|
|
chip->irq_eoi(&desc->irq_data);
|
|
chip->irq_unmask(&desc->irq_data);
|
|
|
|
prcm_irq_setup->ocp_barrier(); /* avoid spurious IRQs */
|
|
}
|
|
|
|
/* Public functions */
|
|
|
|
/**
|
|
* omap_prcm_event_to_irq - given a PRCM event name, returns the
|
|
* corresponding IRQ on which the handler should be registered
|
|
* @name: name of the PRCM interrupt bit to look up - see struct omap_prcm_irq
|
|
*
|
|
* Returns the Linux internal IRQ ID corresponding to @name upon success,
|
|
* or -ENOENT upon failure.
|
|
*/
|
|
int omap_prcm_event_to_irq(const char *name)
|
|
{
|
|
int i;
|
|
|
|
if (!prcm_irq_setup || !name)
|
|
return -ENOENT;
|
|
|
|
for (i = 0; i < prcm_irq_setup->nr_irqs; i++)
|
|
if (!strcmp(prcm_irq_setup->irqs[i].name, name))
|
|
return prcm_irq_setup->base_irq +
|
|
prcm_irq_setup->irqs[i].offset;
|
|
|
|
return -ENOENT;
|
|
}
|
|
|
|
/**
|
|
* omap_prcm_irq_cleanup - reverses memory allocated and other steps
|
|
* done by omap_prcm_register_chain_handler()
|
|
*
|
|
* No return value.
|
|
*/
|
|
void omap_prcm_irq_cleanup(void)
|
|
{
|
|
int i;
|
|
|
|
if (!prcm_irq_setup) {
|
|
pr_err("PRCM: IRQ handler not initialized; cannot cleanup\n");
|
|
return;
|
|
}
|
|
|
|
if (prcm_irq_chips) {
|
|
for (i = 0; i < prcm_irq_setup->nr_regs; i++) {
|
|
if (prcm_irq_chips[i])
|
|
irq_remove_generic_chip(prcm_irq_chips[i],
|
|
0xffffffff, 0, 0);
|
|
prcm_irq_chips[i] = NULL;
|
|
}
|
|
kfree(prcm_irq_chips);
|
|
prcm_irq_chips = NULL;
|
|
}
|
|
|
|
kfree(prcm_irq_setup->saved_mask);
|
|
prcm_irq_setup->saved_mask = NULL;
|
|
|
|
kfree(prcm_irq_setup->priority_mask);
|
|
prcm_irq_setup->priority_mask = NULL;
|
|
|
|
irq_set_chained_handler(prcm_irq_setup->irq, NULL);
|
|
|
|
if (prcm_irq_setup->base_irq > 0)
|
|
irq_free_descs(prcm_irq_setup->base_irq,
|
|
prcm_irq_setup->nr_regs * 32);
|
|
prcm_irq_setup->base_irq = 0;
|
|
}
|
|
|
|
void omap_prcm_irq_prepare(void)
|
|
{
|
|
prcm_irq_setup->suspended = true;
|
|
}
|
|
|
|
void omap_prcm_irq_complete(void)
|
|
{
|
|
prcm_irq_setup->suspended = false;
|
|
|
|
/* If we have not saved the masks, do not attempt to restore */
|
|
if (!prcm_irq_setup->suspend_save_flag)
|
|
return;
|
|
|
|
prcm_irq_setup->suspend_save_flag = false;
|
|
|
|
/*
|
|
* Re-enable all masked PRCM irq sources, this causes the PRCM
|
|
* interrupt to fire immediately if the events were masked
|
|
* previously in the chain handler
|
|
*/
|
|
prcm_irq_setup->restore_irqen(prcm_irq_setup->saved_mask);
|
|
}
|
|
|
|
/**
|
|
* omap_prcm_register_chain_handler - initializes the prcm chained interrupt
|
|
* handler based on provided parameters
|
|
* @irq_setup: hardware data about the underlying PRM/PRCM
|
|
*
|
|
* Set up the PRCM chained interrupt handler on the PRCM IRQ. Sets up
|
|
* one generic IRQ chip per PRM interrupt status/enable register pair.
|
|
* Returns 0 upon success, -EINVAL if called twice or if invalid
|
|
* arguments are passed, or -ENOMEM on any other error.
|
|
*/
|
|
int omap_prcm_register_chain_handler(struct omap_prcm_irq_setup *irq_setup)
|
|
{
|
|
int nr_regs;
|
|
u32 mask[OMAP_PRCM_MAX_NR_PENDING_REG];
|
|
int offset, i;
|
|
struct irq_chip_generic *gc;
|
|
struct irq_chip_type *ct;
|
|
|
|
if (!irq_setup)
|
|
return -EINVAL;
|
|
|
|
nr_regs = irq_setup->nr_regs;
|
|
|
|
if (prcm_irq_setup) {
|
|
pr_err("PRCM: already initialized; won't reinitialize\n");
|
|
return -EINVAL;
|
|
}
|
|
|
|
if (nr_regs > OMAP_PRCM_MAX_NR_PENDING_REG) {
|
|
pr_err("PRCM: nr_regs too large\n");
|
|
return -EINVAL;
|
|
}
|
|
|
|
prcm_irq_setup = irq_setup;
|
|
|
|
prcm_irq_chips = kzalloc(sizeof(void *) * nr_regs, GFP_KERNEL);
|
|
prcm_irq_setup->saved_mask = kzalloc(sizeof(u32) * nr_regs, GFP_KERNEL);
|
|
prcm_irq_setup->priority_mask = kzalloc(sizeof(u32) * nr_regs,
|
|
GFP_KERNEL);
|
|
|
|
if (!prcm_irq_chips || !prcm_irq_setup->saved_mask ||
|
|
!prcm_irq_setup->priority_mask) {
|
|
pr_err("PRCM: kzalloc failed\n");
|
|
goto err;
|
|
}
|
|
|
|
memset(mask, 0, sizeof(mask));
|
|
|
|
for (i = 0; i < irq_setup->nr_irqs; i++) {
|
|
offset = irq_setup->irqs[i].offset;
|
|
mask[offset >> 5] |= 1 << (offset & 0x1f);
|
|
if (irq_setup->irqs[i].priority)
|
|
irq_setup->priority_mask[offset >> 5] |=
|
|
1 << (offset & 0x1f);
|
|
}
|
|
|
|
irq_set_chained_handler(irq_setup->irq, omap_prcm_irq_handler);
|
|
|
|
irq_setup->base_irq = irq_alloc_descs(-1, 0, irq_setup->nr_regs * 32,
|
|
0);
|
|
|
|
if (irq_setup->base_irq < 0) {
|
|
pr_err("PRCM: failed to allocate irq descs: %d\n",
|
|
irq_setup->base_irq);
|
|
goto err;
|
|
}
|
|
|
|
for (i = 0; i < irq_setup->nr_regs; i++) {
|
|
gc = irq_alloc_generic_chip("PRCM", 1,
|
|
irq_setup->base_irq + i * 32, prm_base,
|
|
handle_level_irq);
|
|
|
|
if (!gc) {
|
|
pr_err("PRCM: failed to allocate generic chip\n");
|
|
goto err;
|
|
}
|
|
ct = gc->chip_types;
|
|
ct->chip.irq_ack = irq_gc_ack_set_bit;
|
|
ct->chip.irq_mask = irq_gc_mask_clr_bit;
|
|
ct->chip.irq_unmask = irq_gc_mask_set_bit;
|
|
|
|
ct->regs.ack = irq_setup->ack + i * 4;
|
|
ct->regs.mask = irq_setup->mask + i * 4;
|
|
|
|
irq_setup_generic_chip(gc, mask[i], 0, IRQ_NOREQUEST, 0);
|
|
prcm_irq_chips[i] = gc;
|
|
}
|
|
|
|
if (of_have_populated_dt()) {
|
|
int irq = omap_prcm_event_to_irq("io");
|
|
omap_pcs_legacy_init(irq, irq_setup->reconfigure_io_chain);
|
|
}
|
|
|
|
return 0;
|
|
|
|
err:
|
|
omap_prcm_irq_cleanup();
|
|
return -ENOMEM;
|
|
}
|
|
|
|
/**
|
|
* omap2_set_globals_prm - set the PRM base address (for early use)
|
|
* @prm: PRM base virtual address
|
|
*
|
|
* XXX Will be replaced when the PRM/CM drivers are completed.
|
|
*/
|
|
void __init omap2_set_globals_prm(void __iomem *prm)
|
|
{
|
|
prm_base = prm;
|
|
}
|
|
|
|
/**
|
|
* prm_read_reset_sources - return the sources of the SoC's last reset
|
|
*
|
|
* Return a u32 bitmask representing the reset sources that caused the
|
|
* SoC to reset. The low-level per-SoC functions called by this
|
|
* function remap the SoC-specific reset source bits into an
|
|
* OMAP-common set of reset source bits, defined in
|
|
* arch/arm/mach-omap2/prm.h. Returns the standardized reset source
|
|
* u32 bitmask from the hardware upon success, or returns (1 <<
|
|
* OMAP_UNKNOWN_RST_SRC_ID_SHIFT) if no low-level read_reset_sources()
|
|
* function was registered.
|
|
*/
|
|
u32 prm_read_reset_sources(void)
|
|
{
|
|
u32 ret = 1 << OMAP_UNKNOWN_RST_SRC_ID_SHIFT;
|
|
|
|
if (prm_ll_data->read_reset_sources)
|
|
ret = prm_ll_data->read_reset_sources();
|
|
else
|
|
WARN_ONCE(1, "prm: %s: no mapping function defined for reset sources\n", __func__);
|
|
|
|
return ret;
|
|
}
|
|
|
|
/**
|
|
* prm_was_any_context_lost_old - was device context lost? (old API)
|
|
* @part: PRM partition ID (e.g., OMAP4430_PRM_PARTITION)
|
|
* @inst: PRM instance offset (e.g., OMAP4430_PRM_MPU_INST)
|
|
* @idx: CONTEXT register offset
|
|
*
|
|
* Return 1 if any bits were set in the *_CONTEXT_* register
|
|
* identified by (@part, @inst, @idx), which means that some context
|
|
* was lost for that module; otherwise, return 0. XXX Deprecated;
|
|
* callers need to use a less-SoC-dependent way to identify hardware
|
|
* IP blocks.
|
|
*/
|
|
bool prm_was_any_context_lost_old(u8 part, s16 inst, u16 idx)
|
|
{
|
|
bool ret = true;
|
|
|
|
if (prm_ll_data->was_any_context_lost_old)
|
|
ret = prm_ll_data->was_any_context_lost_old(part, inst, idx);
|
|
else
|
|
WARN_ONCE(1, "prm: %s: no mapping function defined\n",
|
|
__func__);
|
|
|
|
return ret;
|
|
}
|
|
|
|
/**
|
|
* prm_clear_context_lost_flags_old - clear context loss flags (old API)
|
|
* @part: PRM partition ID (e.g., OMAP4430_PRM_PARTITION)
|
|
* @inst: PRM instance offset (e.g., OMAP4430_PRM_MPU_INST)
|
|
* @idx: CONTEXT register offset
|
|
*
|
|
* Clear hardware context loss bits for the module identified by
|
|
* (@part, @inst, @idx). No return value. XXX Deprecated; callers
|
|
* need to use a less-SoC-dependent way to identify hardware IP
|
|
* blocks.
|
|
*/
|
|
void prm_clear_context_loss_flags_old(u8 part, s16 inst, u16 idx)
|
|
{
|
|
if (prm_ll_data->clear_context_loss_flags_old)
|
|
prm_ll_data->clear_context_loss_flags_old(part, inst, idx);
|
|
else
|
|
WARN_ONCE(1, "prm: %s: no mapping function defined\n",
|
|
__func__);
|
|
}
|
|
|
|
/**
|
|
* prm_register - register per-SoC low-level data with the PRM
|
|
* @pld: low-level per-SoC OMAP PRM data & function pointers to register
|
|
*
|
|
* Register per-SoC low-level OMAP PRM data and function pointers with
|
|
* the OMAP PRM common interface. The caller must keep the data
|
|
* pointed to by @pld valid until it calls prm_unregister() and
|
|
* it returns successfully. Returns 0 upon success, -EINVAL if @pld
|
|
* is NULL, or -EEXIST if prm_register() has already been called
|
|
* without an intervening prm_unregister().
|
|
*/
|
|
int prm_register(struct prm_ll_data *pld)
|
|
{
|
|
if (!pld)
|
|
return -EINVAL;
|
|
|
|
if (prm_ll_data != &null_prm_ll_data)
|
|
return -EEXIST;
|
|
|
|
prm_ll_data = pld;
|
|
|
|
return 0;
|
|
}
|
|
|
|
/**
|
|
* prm_unregister - unregister per-SoC low-level data & function pointers
|
|
* @pld: low-level per-SoC OMAP PRM data & function pointers to unregister
|
|
*
|
|
* Unregister per-SoC low-level OMAP PRM data and function pointers
|
|
* that were previously registered with prm_register(). The
|
|
* caller may not destroy any of the data pointed to by @pld until
|
|
* this function returns successfully. Returns 0 upon success, or
|
|
* -EINVAL if @pld is NULL or if @pld does not match the struct
|
|
* prm_ll_data * previously registered by prm_register().
|
|
*/
|
|
int prm_unregister(struct prm_ll_data *pld)
|
|
{
|
|
if (!pld || prm_ll_data != pld)
|
|
return -EINVAL;
|
|
|
|
prm_ll_data = &null_prm_ll_data;
|
|
|
|
return 0;
|
|
}
|
|
|
|
static const struct of_device_id omap_prcm_dt_match_table[] = {
|
|
{ .compatible = "ti,am3-prcm" },
|
|
{ .compatible = "ti,am3-scrm" },
|
|
{ .compatible = "ti,am4-prcm" },
|
|
{ .compatible = "ti,am4-scrm" },
|
|
{ .compatible = "ti,omap2-prcm" },
|
|
{ .compatible = "ti,omap2-scrm" },
|
|
{ .compatible = "ti,omap3-prm" },
|
|
{ .compatible = "ti,omap3-cm" },
|
|
{ .compatible = "ti,omap3-scrm" },
|
|
{ .compatible = "ti,omap4-cm1" },
|
|
{ .compatible = "ti,omap4-prm" },
|
|
{ .compatible = "ti,omap4-cm2" },
|
|
{ .compatible = "ti,omap4-scrm" },
|
|
{ .compatible = "ti,omap5-prm" },
|
|
{ .compatible = "ti,omap5-cm-core-aon" },
|
|
{ .compatible = "ti,omap5-scrm" },
|
|
{ .compatible = "ti,omap5-cm-core" },
|
|
{ .compatible = "ti,dra7-prm" },
|
|
{ .compatible = "ti,dra7-cm-core-aon" },
|
|
{ .compatible = "ti,dra7-cm-core" },
|
|
{ }
|
|
};
|
|
|
|
static struct clk_hw_omap memmap_dummy_ck = {
|
|
.flags = MEMMAP_ADDRESSING,
|
|
};
|
|
|
|
static u32 prm_clk_readl(void __iomem *reg)
|
|
{
|
|
return omap2_clk_readl(&memmap_dummy_ck, reg);
|
|
}
|
|
|
|
static void prm_clk_writel(u32 val, void __iomem *reg)
|
|
{
|
|
omap2_clk_writel(val, &memmap_dummy_ck, reg);
|
|
}
|
|
|
|
static struct ti_clk_ll_ops omap_clk_ll_ops = {
|
|
.clk_readl = prm_clk_readl,
|
|
.clk_writel = prm_clk_writel,
|
|
};
|
|
|
|
int __init of_prcm_init(void)
|
|
{
|
|
struct device_node *np;
|
|
void __iomem *mem;
|
|
int memmap_index = 0;
|
|
|
|
ti_clk_ll_ops = &omap_clk_ll_ops;
|
|
|
|
for_each_matching_node(np, omap_prcm_dt_match_table) {
|
|
mem = of_iomap(np, 0);
|
|
clk_memmaps[memmap_index] = mem;
|
|
ti_dt_clk_init_provider(np, memmap_index);
|
|
memmap_index++;
|
|
}
|
|
|
|
return 0;
|
|
}
|
|
|
|
static int __init prm_late_init(void)
|
|
{
|
|
if (prm_ll_data->late_init)
|
|
return prm_ll_data->late_init();
|
|
return 0;
|
|
}
|
|
subsys_initcall(prm_late_init);
|