arm-trusted-firmware/lib/psci/psci_common.c
Boyan Karatotev 83ec7e452c perf(amu): greatly simplify AMU context management
The current code is incredibly resilient to updates to the spec and
has worked quite well so far. However, recent implementations expose a
weakness in that this is rather slow. A large part of it is written in
assembly, making it opaque to the compiler for optimisations. The
future proofness requires reading registers that are effectively
`volatile`, making it even harder for the compiler, as well as adding
lots of implicit barriers, making it hard for the microarchitecutre to
optimise as well.

We can make a few assumptions, checked by a few well placed asserts, and
remove a lot of this burden. For a start, at the moment there are 4
group 0 counters with static assignments. Contexting them is a trivial
affair that doesn't need a loop. Similarly, there can only be up to 16
group 1 counters. Contexting them is a bit harder, but we can do with a
single branch with a falling through switch. If/when both of these
change, we have a pair of asserts and the feature detection mechanism to
guard us against pretending that we support something we don't.

We can drop contexting of the offset registers. They are fully
accessible by EL2 and as such are its responsibility to preserve on
powerdown.

Another small thing we can do, is pass the core_pos into the hook.
The caller already knows which core we're running on, we don't need to
call this non-trivial function again.

Finally, knowing this, we don't really need the auxiliary AMUs to be
described by the device tree. Linux doesn't care at the moment, and any
information we need for EL3 can be neatly placed in a simple array.

All of this, combined with lifting the actual saving out of assembly,
reduces the instructions to save the context from 180 to 40, including a
lot fewer branches. The code is also much shorter and easier to read.

Also propagate to aarch32 so that the two don't diverge too much.

Change-Id: Ib62e6e9ba5be7fb9fb8965c8eee148d5598a5361
Signed-off-by: Boyan Karatotev <boyan.karatotev@arm.com>
2025-02-25 08:50:46 +00:00

1360 lines
47 KiB
C

/*
* Copyright (c) 2013-2025, Arm Limited and Contributors. All rights reserved.
*
* SPDX-License-Identifier: BSD-3-Clause
*/
#include <assert.h>
#include <string.h>
#include <arch.h>
#include <arch_features.h>
#include <arch_helpers.h>
#include <common/bl_common.h>
#include <common/debug.h>
#include <context.h>
#include <drivers/delay_timer.h>
#include <lib/el3_runtime/context_mgmt.h>
#include <lib/extensions/spe.h>
#include <lib/pmf/pmf.h>
#include <lib/runtime_instr.h>
#include <lib/utils.h>
#include <plat/common/platform.h>
#include "psci_private.h"
/*
* SPD power management operations, expected to be supplied by the registered
* SPD on successful SP initialization
*/
const spd_pm_ops_t *psci_spd_pm;
/*
* PSCI requested local power state map. This array is used to store the local
* power states requested by a CPU for power levels from level 1 to
* PLAT_MAX_PWR_LVL. It does not store the requested local power state for power
* level 0 (PSCI_CPU_PWR_LVL) as the requested and the target power state for a
* CPU are the same.
*
* During state coordination, the platform is passed an array containing the
* local states requested for a particular non cpu power domain by each cpu
* within the domain.
*
* TODO: Dense packing of the requested states will cause cache thrashing
* when multiple power domains write to it. If we allocate the requested
* states at each power level in a cache-line aligned per-domain memory,
* the cache thrashing can be avoided.
*/
static plat_local_state_t
psci_req_local_pwr_states[PLAT_MAX_PWR_LVL][PLATFORM_CORE_COUNT];
unsigned int psci_plat_core_count;
/*******************************************************************************
* Arrays that hold the platform's power domain tree information for state
* management of power domains.
* Each node in the array 'psci_non_cpu_pd_nodes' corresponds to a power domain
* which is an ancestor of a CPU power domain.
* Each node in the array 'psci_cpu_pd_nodes' corresponds to a cpu power domain
******************************************************************************/
non_cpu_pd_node_t psci_non_cpu_pd_nodes[PSCI_NUM_NON_CPU_PWR_DOMAINS]
#if USE_COHERENT_MEM
__section(".tzfw_coherent_mem")
#endif
;
/* Lock for PSCI state coordination */
DEFINE_PSCI_LOCK(psci_locks[PSCI_NUM_NON_CPU_PWR_DOMAINS]);
cpu_pd_node_t psci_cpu_pd_nodes[PLATFORM_CORE_COUNT];
/*******************************************************************************
* Pointer to functions exported by the platform to complete power mgmt. ops
******************************************************************************/
const plat_psci_ops_t *psci_plat_pm_ops;
/******************************************************************************
* Check that the maximum power level supported by the platform makes sense
*****************************************************************************/
CASSERT((PLAT_MAX_PWR_LVL <= PSCI_MAX_PWR_LVL) &&
(PLAT_MAX_PWR_LVL >= PSCI_CPU_PWR_LVL),
assert_platform_max_pwrlvl_check);
#if PSCI_OS_INIT_MODE
/*******************************************************************************
* The power state coordination mode used in CPU_SUSPEND.
* Defaults to platform-coordinated mode.
******************************************************************************/
suspend_mode_t psci_suspend_mode = PLAT_COORD;
#endif
/*
* The plat_local_state used by the platform is one of these types: RUN,
* RETENTION and OFF. The platform can define further sub-states for each type
* apart from RUN. This categorization is done to verify the sanity of the
* psci_power_state passed by the platform and to print debug information. The
* categorization is done on the basis of the following conditions:
*
* 1. If (plat_local_state == 0) then the category is STATE_TYPE_RUN.
*
* 2. If (0 < plat_local_state <= PLAT_MAX_RET_STATE), then the category is
* STATE_TYPE_RETN.
*
* 3. If (plat_local_state > PLAT_MAX_RET_STATE), then the category is
* STATE_TYPE_OFF.
*/
typedef enum plat_local_state_type {
STATE_TYPE_RUN = 0,
STATE_TYPE_RETN,
STATE_TYPE_OFF
} plat_local_state_type_t;
/* Function used to categorize plat_local_state. */
static plat_local_state_type_t find_local_state_type(plat_local_state_t state)
{
if (state != 0U) {
if (state > PLAT_MAX_RET_STATE) {
return STATE_TYPE_OFF;
} else {
return STATE_TYPE_RETN;
}
} else {
return STATE_TYPE_RUN;
}
}
/******************************************************************************
* Check that the maximum retention level supported by the platform is less
* than the maximum off level.
*****************************************************************************/
CASSERT(PLAT_MAX_RET_STATE < PLAT_MAX_OFF_STATE,
assert_platform_max_off_and_retn_state_check);
/******************************************************************************
* This function ensures that the power state parameter in a CPU_SUSPEND request
* is valid. If so, it returns the requested states for each power level.
*****************************************************************************/
int psci_validate_power_state(unsigned int power_state,
psci_power_state_t *state_info)
{
/* Check SBZ bits in power state are zero */
if (psci_check_power_state(power_state) != 0U)
return PSCI_E_INVALID_PARAMS;
assert(psci_plat_pm_ops->validate_power_state != NULL);
/* Validate the power_state using platform pm_ops */
return psci_plat_pm_ops->validate_power_state(power_state, state_info);
}
/******************************************************************************
* This function retrieves the `psci_power_state_t` for system suspend from
* the platform.
*****************************************************************************/
void psci_query_sys_suspend_pwrstate(psci_power_state_t *state_info)
{
/*
* Assert that the required pm_ops hook is implemented to ensure that
* the capability detected during psci_setup() is valid.
*/
assert(psci_plat_pm_ops->get_sys_suspend_power_state != NULL);
/*
* Query the platform for the power_state required for system suspend
*/
psci_plat_pm_ops->get_sys_suspend_power_state(state_info);
}
#if PSCI_OS_INIT_MODE
/*******************************************************************************
* This function verifies that all the other cores at the 'end_pwrlvl' have been
* idled and the current CPU is the last running CPU at the 'end_pwrlvl'.
* Returns 1 (true) if the current CPU is the last ON CPU or 0 (false)
* otherwise.
******************************************************************************/
static bool psci_is_last_cpu_to_idle_at_pwrlvl(unsigned int my_idx, unsigned int end_pwrlvl)
{
unsigned int lvl;
unsigned int parent_idx = 0;
unsigned int cpu_start_idx, ncpus, cpu_idx;
plat_local_state_t local_state;
if (end_pwrlvl == PSCI_CPU_PWR_LVL) {
return true;
}
parent_idx = psci_cpu_pd_nodes[my_idx].parent_node;
for (lvl = PSCI_CPU_PWR_LVL + U(1); lvl < end_pwrlvl; lvl++) {
parent_idx = psci_non_cpu_pd_nodes[parent_idx].parent_node;
}
cpu_start_idx = psci_non_cpu_pd_nodes[parent_idx].cpu_start_idx;
ncpus = psci_non_cpu_pd_nodes[parent_idx].ncpus;
for (cpu_idx = cpu_start_idx; cpu_idx < cpu_start_idx + ncpus;
cpu_idx++) {
local_state = psci_get_cpu_local_state_by_idx(cpu_idx);
if (cpu_idx == my_idx) {
assert(is_local_state_run(local_state) != 0);
continue;
}
if (is_local_state_run(local_state) != 0) {
return false;
}
}
return true;
}
#endif
/*******************************************************************************
* This function verifies that all the other cores in the system have been
* turned OFF and the current CPU is the last running CPU in the system.
* Returns true, if the current CPU is the last ON CPU or false otherwise.
******************************************************************************/
bool psci_is_last_on_cpu(unsigned int my_idx)
{
for (unsigned int cpu_idx = 0; cpu_idx < psci_plat_core_count; cpu_idx++) {
if (cpu_idx == my_idx) {
assert(psci_get_aff_info_state() == AFF_STATE_ON);
continue;
}
if (psci_get_aff_info_state_by_idx(cpu_idx) != AFF_STATE_OFF) {
VERBOSE("core=%u other than current core=%u %s\n",
cpu_idx, my_idx, "running in the system");
return false;
}
}
return true;
}
/*******************************************************************************
* This function verifies that all cores in the system have been turned ON.
* Returns true, if all CPUs are ON or false otherwise.
******************************************************************************/
static bool psci_are_all_cpus_on(void)
{
unsigned int cpu_idx;
for (cpu_idx = 0; cpu_idx < psci_plat_core_count; cpu_idx++) {
if (psci_get_aff_info_state_by_idx(cpu_idx) == AFF_STATE_OFF) {
return false;
}
}
return true;
}
/*******************************************************************************
* Routine to return the maximum power level to traverse to after a cpu has
* been physically powered up. It is expected to be called immediately after
* reset from assembler code.
******************************************************************************/
static unsigned int get_power_on_target_pwrlvl(void)
{
unsigned int pwrlvl;
/*
* Assume that this cpu was suspended and retrieve its target power
* level. If it wasn't, the cpu is off so this will be PLAT_MAX_PWR_LVL.
*/
pwrlvl = psci_get_suspend_pwrlvl();
assert(pwrlvl < PSCI_INVALID_PWR_LVL);
return pwrlvl;
}
/******************************************************************************
* Helper function to update the requested local power state array. This array
* does not store the requested state for the CPU power level. Hence an
* assertion is added to prevent us from accessing the CPU power level.
*****************************************************************************/
static void psci_set_req_local_pwr_state(unsigned int pwrlvl,
unsigned int cpu_idx,
plat_local_state_t req_pwr_state)
{
assert(pwrlvl > PSCI_CPU_PWR_LVL);
if ((pwrlvl > PSCI_CPU_PWR_LVL) && (pwrlvl <= PLAT_MAX_PWR_LVL) &&
(cpu_idx < psci_plat_core_count)) {
psci_req_local_pwr_states[pwrlvl - 1U][cpu_idx] = req_pwr_state;
}
}
/******************************************************************************
* This function initializes the psci_req_local_pwr_states.
*****************************************************************************/
void __init psci_init_req_local_pwr_states(void)
{
/* Initialize the requested state of all non CPU power domains as OFF */
unsigned int pwrlvl;
unsigned int core;
for (pwrlvl = 0U; pwrlvl < PLAT_MAX_PWR_LVL; pwrlvl++) {
for (core = 0; core < psci_plat_core_count; core++) {
psci_req_local_pwr_states[pwrlvl][core] =
PLAT_MAX_OFF_STATE;
}
}
}
/******************************************************************************
* Helper function to return a reference to an array containing the local power
* states requested by each cpu for a power domain at 'pwrlvl'. The size of the
* array will be the number of cpu power domains of which this power domain is
* an ancestor. These requested states will be used to determine a suitable
* target state for this power domain during psci state coordination. An
* assertion is added to prevent us from accessing the CPU power level.
*****************************************************************************/
static plat_local_state_t *psci_get_req_local_pwr_states(unsigned int pwrlvl,
unsigned int cpu_idx)
{
assert(pwrlvl > PSCI_CPU_PWR_LVL);
if ((pwrlvl > PSCI_CPU_PWR_LVL) && (pwrlvl <= PLAT_MAX_PWR_LVL) &&
(cpu_idx < psci_plat_core_count)) {
return &psci_req_local_pwr_states[pwrlvl - 1U][cpu_idx];
} else
return NULL;
}
#if PSCI_OS_INIT_MODE
/******************************************************************************
* Helper function to save a copy of the psci_req_local_pwr_states (prev) for a
* CPU (cpu_idx), and update psci_req_local_pwr_states with the new requested
* local power states (state_info).
*****************************************************************************/
void psci_update_req_local_pwr_states(unsigned int end_pwrlvl,
unsigned int cpu_idx,
psci_power_state_t *state_info,
plat_local_state_t *prev)
{
unsigned int lvl;
#ifdef PLAT_MAX_CPU_SUSPEND_PWR_LVL
unsigned int max_pwrlvl = PLAT_MAX_CPU_SUSPEND_PWR_LVL;
#else
unsigned int max_pwrlvl = PLAT_MAX_PWR_LVL;
#endif
plat_local_state_t req_state;
for (lvl = PSCI_CPU_PWR_LVL + 1U; lvl <= max_pwrlvl; lvl++) {
/* Save the previous requested local power state */
prev[lvl - 1U] = *psci_get_req_local_pwr_states(lvl, cpu_idx);
/* Update the new requested local power state */
if (lvl <= end_pwrlvl) {
req_state = state_info->pwr_domain_state[lvl];
} else {
req_state = state_info->pwr_domain_state[end_pwrlvl];
}
psci_set_req_local_pwr_state(lvl, cpu_idx, req_state);
}
}
/******************************************************************************
* Helper function to restore the previously saved requested local power states
* (prev) for a CPU (cpu_idx) to psci_req_local_pwr_states.
*****************************************************************************/
void psci_restore_req_local_pwr_states(unsigned int cpu_idx,
plat_local_state_t *prev)
{
unsigned int lvl;
#ifdef PLAT_MAX_CPU_SUSPEND_PWR_LVL
unsigned int max_pwrlvl = PLAT_MAX_CPU_SUSPEND_PWR_LVL;
#else
unsigned int max_pwrlvl = PLAT_MAX_PWR_LVL;
#endif
for (lvl = PSCI_CPU_PWR_LVL + 1U; lvl <= max_pwrlvl; lvl++) {
/* Restore the previous requested local power state */
psci_set_req_local_pwr_state(lvl, cpu_idx, prev[lvl - 1U]);
}
}
#endif
/*
* psci_non_cpu_pd_nodes can be placed either in normal memory or coherent
* memory.
*
* With !USE_COHERENT_MEM, psci_non_cpu_pd_nodes is placed in normal memory,
* it's accessed by both cached and non-cached participants. To serve the common
* minimum, perform a cache flush before read and after write so that non-cached
* participants operate on latest data in main memory.
*
* When USE_COHERENT_MEM is used, psci_non_cpu_pd_nodes is placed in coherent
* memory. With HW_ASSISTED_COHERENCY, all PSCI participants are cache-coherent.
* In both cases, no cache operations are required.
*/
/*
* Retrieve local state of non-CPU power domain node from a non-cached CPU,
* after any required cache maintenance operation.
*/
static plat_local_state_t get_non_cpu_pd_node_local_state(
unsigned int parent_idx)
{
#if !(USE_COHERENT_MEM || HW_ASSISTED_COHERENCY || WARMBOOT_ENABLE_DCACHE_EARLY)
flush_dcache_range(
(uintptr_t) &psci_non_cpu_pd_nodes[parent_idx],
sizeof(psci_non_cpu_pd_nodes[parent_idx]));
#endif
return psci_non_cpu_pd_nodes[parent_idx].local_state;
}
/*
* Update local state of non-CPU power domain node from a cached CPU; perform
* any required cache maintenance operation afterwards.
*/
static void set_non_cpu_pd_node_local_state(unsigned int parent_idx,
plat_local_state_t state)
{
psci_non_cpu_pd_nodes[parent_idx].local_state = state;
#if !(USE_COHERENT_MEM || HW_ASSISTED_COHERENCY || WARMBOOT_ENABLE_DCACHE_EARLY)
flush_dcache_range(
(uintptr_t) &psci_non_cpu_pd_nodes[parent_idx],
sizeof(psci_non_cpu_pd_nodes[parent_idx]));
#endif
}
/******************************************************************************
* Helper function to return the current local power state of each power domain
* from the current cpu power domain to its ancestor at the 'end_pwrlvl'. This
* function will be called after a cpu is powered on to find the local state
* each power domain has emerged from.
*****************************************************************************/
void psci_get_target_local_pwr_states(unsigned int cpu_idx, unsigned int end_pwrlvl,
psci_power_state_t *target_state)
{
unsigned int parent_idx, lvl;
plat_local_state_t *pd_state = target_state->pwr_domain_state;
pd_state[PSCI_CPU_PWR_LVL] = psci_get_cpu_local_state();
parent_idx = psci_cpu_pd_nodes[cpu_idx].parent_node;
/* Copy the local power state from node to state_info */
for (lvl = PSCI_CPU_PWR_LVL + 1U; lvl <= end_pwrlvl; lvl++) {
pd_state[lvl] = get_non_cpu_pd_node_local_state(parent_idx);
parent_idx = psci_non_cpu_pd_nodes[parent_idx].parent_node;
}
/* Set the the higher levels to RUN */
for (; lvl <= PLAT_MAX_PWR_LVL; lvl++)
target_state->pwr_domain_state[lvl] = PSCI_LOCAL_STATE_RUN;
}
/******************************************************************************
* Helper function to set the target local power state that each power domain
* from the current cpu power domain to its ancestor at the 'end_pwrlvl' will
* enter. This function will be called after coordination of requested power
* states has been done for each power level.
*****************************************************************************/
void psci_set_target_local_pwr_states(unsigned int cpu_idx, unsigned int end_pwrlvl,
const psci_power_state_t *target_state)
{
unsigned int parent_idx, lvl;
const plat_local_state_t *pd_state = target_state->pwr_domain_state;
psci_set_cpu_local_state(pd_state[PSCI_CPU_PWR_LVL]);
/*
* Need to flush as local_state might be accessed with Data Cache
* disabled during power on
*/
psci_flush_cpu_data(psci_svc_cpu_data.local_state);
parent_idx = psci_cpu_pd_nodes[cpu_idx].parent_node;
/* Copy the local_state from state_info */
for (lvl = 1U; lvl <= end_pwrlvl; lvl++) {
set_non_cpu_pd_node_local_state(parent_idx, pd_state[lvl]);
parent_idx = psci_non_cpu_pd_nodes[parent_idx].parent_node;
}
}
/*******************************************************************************
* PSCI helper function to get the parent nodes corresponding to a cpu_index.
******************************************************************************/
void psci_get_parent_pwr_domain_nodes(unsigned int cpu_idx,
unsigned int end_lvl,
unsigned int *node_index)
{
unsigned int parent_node = psci_cpu_pd_nodes[cpu_idx].parent_node;
unsigned int i;
unsigned int *node = node_index;
for (i = PSCI_CPU_PWR_LVL + 1U; i <= end_lvl; i++) {
*node = parent_node;
node++;
parent_node = psci_non_cpu_pd_nodes[parent_node].parent_node;
}
}
/******************************************************************************
* This function is invoked post CPU power up and initialization. It sets the
* affinity info state, target power state and requested power state for the
* current CPU and all its ancestor power domains to RUN.
*****************************************************************************/
void psci_set_pwr_domains_to_run(unsigned int cpu_idx, unsigned int end_pwrlvl)
{
unsigned int parent_idx, lvl;
parent_idx = psci_cpu_pd_nodes[cpu_idx].parent_node;
/* Reset the local_state to RUN for the non cpu power domains. */
for (lvl = PSCI_CPU_PWR_LVL + 1U; lvl <= end_pwrlvl; lvl++) {
set_non_cpu_pd_node_local_state(parent_idx,
PSCI_LOCAL_STATE_RUN);
psci_set_req_local_pwr_state(lvl,
cpu_idx,
PSCI_LOCAL_STATE_RUN);
parent_idx = psci_non_cpu_pd_nodes[parent_idx].parent_node;
}
/* Set the affinity info state to ON */
psci_set_aff_info_state(AFF_STATE_ON);
psci_set_cpu_local_state(PSCI_LOCAL_STATE_RUN);
psci_flush_cpu_data(psci_svc_cpu_data);
}
/******************************************************************************
* This function is used in platform-coordinated mode.
*
* This function is passed the local power states requested for each power
* domain (state_info) between the current CPU domain and its ancestors until
* the target power level (end_pwrlvl). It updates the array of requested power
* states with this information.
*
* Then, for each level (apart from the CPU level) until the 'end_pwrlvl', it
* retrieves the states requested by all the cpus of which the power domain at
* that level is an ancestor. It passes this information to the platform to
* coordinate and return the target power state. If the target state for a level
* is RUN then subsequent levels are not considered. At the CPU level, state
* coordination is not required. Hence, the requested and the target states are
* the same.
*
* The 'state_info' is updated with the target state for each level between the
* CPU and the 'end_pwrlvl' and returned to the caller.
*
* This function will only be invoked with data cache enabled and while
* powering down a core.
*****************************************************************************/
void psci_do_state_coordination(unsigned int cpu_idx, unsigned int end_pwrlvl,
psci_power_state_t *state_info)
{
unsigned int lvl, parent_idx;
unsigned int start_idx;
unsigned int ncpus;
plat_local_state_t target_state, *req_states;
assert(end_pwrlvl <= PLAT_MAX_PWR_LVL);
parent_idx = psci_cpu_pd_nodes[cpu_idx].parent_node;
/* For level 0, the requested state will be equivalent
to target state */
for (lvl = PSCI_CPU_PWR_LVL + 1U; lvl <= end_pwrlvl; lvl++) {
/* First update the requested power state */
psci_set_req_local_pwr_state(lvl, cpu_idx,
state_info->pwr_domain_state[lvl]);
/* Get the requested power states for this power level */
start_idx = psci_non_cpu_pd_nodes[parent_idx].cpu_start_idx;
req_states = psci_get_req_local_pwr_states(lvl, start_idx);
/*
* Let the platform coordinate amongst the requested states at
* this power level and return the target local power state.
*/
ncpus = psci_non_cpu_pd_nodes[parent_idx].ncpus;
target_state = plat_get_target_pwr_state(lvl,
req_states,
ncpus);
state_info->pwr_domain_state[lvl] = target_state;
/* Break early if the negotiated target power state is RUN */
if (is_local_state_run(state_info->pwr_domain_state[lvl]) != 0)
break;
parent_idx = psci_non_cpu_pd_nodes[parent_idx].parent_node;
}
/*
* This is for cases when we break out of the above loop early because
* the target power state is RUN at a power level < end_pwlvl.
* We update the requested power state from state_info and then
* set the target state as RUN.
*/
for (lvl = lvl + 1U; lvl <= end_pwrlvl; lvl++) {
psci_set_req_local_pwr_state(lvl, cpu_idx,
state_info->pwr_domain_state[lvl]);
state_info->pwr_domain_state[lvl] = PSCI_LOCAL_STATE_RUN;
}
}
#if PSCI_OS_INIT_MODE
/******************************************************************************
* This function is used in OS-initiated mode.
*
* This function is passed the local power states requested for each power
* domain (state_info) between the current CPU domain and its ancestors until
* the target power level (end_pwrlvl), and ensures the requested power states
* are valid. It updates the array of requested power states with this
* information.
*
* Then, for each level (apart from the CPU level) until the 'end_pwrlvl', it
* retrieves the states requested by all the cpus of which the power domain at
* that level is an ancestor. It passes this information to the platform to
* coordinate and return the target power state. If the requested state does
* not match the target state, the request is denied.
*
* The 'state_info' is not modified.
*
* This function will only be invoked with data cache enabled and while
* powering down a core.
*****************************************************************************/
int psci_validate_state_coordination(unsigned int cpu_idx, unsigned int end_pwrlvl,
psci_power_state_t *state_info)
{
int rc = PSCI_E_SUCCESS;
unsigned int lvl, parent_idx;
unsigned int start_idx;
unsigned int ncpus;
plat_local_state_t target_state, *req_states;
plat_local_state_t prev[PLAT_MAX_PWR_LVL];
assert(end_pwrlvl <= PLAT_MAX_PWR_LVL);
parent_idx = psci_cpu_pd_nodes[cpu_idx].parent_node;
/*
* Save a copy of the previous requested local power states and update
* the new requested local power states.
*/
psci_update_req_local_pwr_states(end_pwrlvl, cpu_idx, state_info, prev);
for (lvl = PSCI_CPU_PWR_LVL + 1U; lvl <= end_pwrlvl; lvl++) {
/* Get the requested power states for this power level */
start_idx = psci_non_cpu_pd_nodes[parent_idx].cpu_start_idx;
req_states = psci_get_req_local_pwr_states(lvl, start_idx);
/*
* Let the platform coordinate amongst the requested states at
* this power level and return the target local power state.
*/
ncpus = psci_non_cpu_pd_nodes[parent_idx].ncpus;
target_state = plat_get_target_pwr_state(lvl,
req_states,
ncpus);
/*
* Verify that the requested power state matches the target
* local power state.
*/
if (state_info->pwr_domain_state[lvl] != target_state) {
if (target_state == PSCI_LOCAL_STATE_RUN) {
rc = PSCI_E_DENIED;
} else {
rc = PSCI_E_INVALID_PARAMS;
}
goto exit;
}
parent_idx = psci_non_cpu_pd_nodes[parent_idx].parent_node;
}
/*
* Verify that the current core is the last running core at the
* specified power level.
*/
lvl = state_info->last_at_pwrlvl;
if (!psci_is_last_cpu_to_idle_at_pwrlvl(cpu_idx, lvl)) {
rc = PSCI_E_DENIED;
}
exit:
if (rc != PSCI_E_SUCCESS) {
/* Restore the previous requested local power states. */
psci_restore_req_local_pwr_states(cpu_idx, prev);
return rc;
}
return rc;
}
#endif
/******************************************************************************
* This function validates a suspend request by making sure that if a standby
* state is requested then no power level is turned off and the highest power
* level is placed in a standby/retention state.
*
* It also ensures that the state level X will enter is not shallower than the
* state level X + 1 will enter.
*
* This validation will be enabled only for DEBUG builds as the platform is
* expected to perform these validations as well.
*****************************************************************************/
int psci_validate_suspend_req(const psci_power_state_t *state_info,
unsigned int is_power_down_state)
{
unsigned int max_off_lvl, target_lvl, max_retn_lvl;
plat_local_state_t state;
plat_local_state_type_t req_state_type, deepest_state_type;
int i;
/* Find the target suspend power level */
target_lvl = psci_find_target_suspend_lvl(state_info);
if (target_lvl == PSCI_INVALID_PWR_LVL)
return PSCI_E_INVALID_PARAMS;
/* All power domain levels are in a RUN state to begin with */
deepest_state_type = STATE_TYPE_RUN;
for (i = (int) target_lvl; i >= (int) PSCI_CPU_PWR_LVL; i--) {
state = state_info->pwr_domain_state[i];
req_state_type = find_local_state_type(state);
/*
* While traversing from the highest power level to the lowest,
* the state requested for lower levels has to be the same or
* deeper i.e. equal to or greater than the state at the higher
* levels. If this condition is true, then the requested state
* becomes the deepest state encountered so far.
*/
if (req_state_type < deepest_state_type)
return PSCI_E_INVALID_PARAMS;
deepest_state_type = req_state_type;
}
/* Find the highest off power level */
max_off_lvl = psci_find_max_off_lvl(state_info);
/* The target_lvl is either equal to the max_off_lvl or max_retn_lvl */
max_retn_lvl = PSCI_INVALID_PWR_LVL;
if (target_lvl != max_off_lvl)
max_retn_lvl = target_lvl;
/*
* If this is not a request for a power down state then max off level
* has to be invalid and max retention level has to be a valid power
* level.
*/
if ((is_power_down_state == 0U) &&
((max_off_lvl != PSCI_INVALID_PWR_LVL) ||
(max_retn_lvl == PSCI_INVALID_PWR_LVL)))
return PSCI_E_INVALID_PARAMS;
return PSCI_E_SUCCESS;
}
/******************************************************************************
* This function finds the highest power level which will be powered down
* amongst all the power levels specified in the 'state_info' structure
*****************************************************************************/
unsigned int psci_find_max_off_lvl(const psci_power_state_t *state_info)
{
int i;
for (i = (int) PLAT_MAX_PWR_LVL; i >= (int) PSCI_CPU_PWR_LVL; i--) {
if (is_local_state_off(state_info->pwr_domain_state[i]) != 0)
return (unsigned int) i;
}
return PSCI_INVALID_PWR_LVL;
}
/******************************************************************************
* This functions finds the level of the highest power domain which will be
* placed in a low power state during a suspend operation.
*****************************************************************************/
unsigned int psci_find_target_suspend_lvl(const psci_power_state_t *state_info)
{
int i;
for (i = (int) PLAT_MAX_PWR_LVL; i >= (int) PSCI_CPU_PWR_LVL; i--) {
if (is_local_state_run(state_info->pwr_domain_state[i]) == 0)
return (unsigned int) i;
}
return PSCI_INVALID_PWR_LVL;
}
/*******************************************************************************
* This function is passed the highest level in the topology tree that the
* operation should be applied to and a list of node indexes. It picks up locks
* from the node index list in order of increasing power domain level in the
* range specified.
******************************************************************************/
void psci_acquire_pwr_domain_locks(unsigned int end_pwrlvl,
const unsigned int *parent_nodes)
{
unsigned int parent_idx;
unsigned int level;
/* No locking required for level 0. Hence start locking from level 1 */
for (level = PSCI_CPU_PWR_LVL + 1U; level <= end_pwrlvl; level++) {
parent_idx = parent_nodes[level - 1U];
psci_lock_get(&psci_non_cpu_pd_nodes[parent_idx]);
}
}
/*******************************************************************************
* This function is passed the highest level in the topology tree that the
* operation should be applied to and a list of node indexes. It releases the
* locks in order of decreasing power domain level in the range specified.
******************************************************************************/
void psci_release_pwr_domain_locks(unsigned int end_pwrlvl,
const unsigned int *parent_nodes)
{
unsigned int parent_idx;
unsigned int level;
/* Unlock top down. No unlocking required for level 0. */
for (level = end_pwrlvl; level >= (PSCI_CPU_PWR_LVL + 1U); level--) {
parent_idx = parent_nodes[level - 1U];
psci_lock_release(&psci_non_cpu_pd_nodes[parent_idx]);
}
}
/*******************************************************************************
* This function determines the full entrypoint information for the requested
* PSCI entrypoint on power on/resume and returns it.
******************************************************************************/
#ifdef __aarch64__
static int psci_get_ns_ep_info(entry_point_info_t *ep,
uintptr_t entrypoint,
u_register_t context_id)
{
u_register_t ep_attr, sctlr;
unsigned int daif, ee, mode;
u_register_t ns_scr_el3 = read_scr_el3();
u_register_t ns_sctlr_el1 = read_sctlr_el1();
sctlr = ((ns_scr_el3 & SCR_HCE_BIT) != 0U) ?
read_sctlr_el2() : ns_sctlr_el1;
ee = 0;
ep_attr = NON_SECURE | EP_ST_DISABLE;
if ((sctlr & SCTLR_EE_BIT) != 0U) {
ep_attr |= EP_EE_BIG;
ee = 1;
}
SET_PARAM_HEAD(ep, PARAM_EP, VERSION_1, ep_attr);
ep->pc = entrypoint;
zeromem(&ep->args, sizeof(ep->args));
ep->args.arg0 = context_id;
/*
* Figure out whether the cpu enters the non-secure address space
* in aarch32 or aarch64
*/
if ((ns_scr_el3 & SCR_RW_BIT) != 0U) {
/*
* Check whether a Thumb entry point has been provided for an
* aarch64 EL
*/
if ((entrypoint & 0x1UL) != 0UL)
return PSCI_E_INVALID_ADDRESS;
mode = ((ns_scr_el3 & SCR_HCE_BIT) != 0U) ? MODE_EL2 : MODE_EL1;
ep->spsr = SPSR_64((uint64_t)mode, MODE_SP_ELX,
DISABLE_ALL_EXCEPTIONS);
} else {
mode = ((ns_scr_el3 & SCR_HCE_BIT) != 0U) ?
MODE32_hyp : MODE32_svc;
/*
* TODO: Choose async. exception bits if HYP mode is not
* implemented according to the values of SCR.{AW, FW} bits
*/
daif = DAIF_ABT_BIT | DAIF_IRQ_BIT | DAIF_FIQ_BIT;
ep->spsr = SPSR_MODE32((uint64_t)mode, entrypoint & 0x1, ee,
daif);
}
return PSCI_E_SUCCESS;
}
#else /* !__aarch64__ */
static int psci_get_ns_ep_info(entry_point_info_t *ep,
uintptr_t entrypoint,
u_register_t context_id)
{
u_register_t ep_attr;
unsigned int aif, ee, mode;
u_register_t scr = read_scr();
u_register_t ns_sctlr, sctlr;
/* Switch to non secure state */
write_scr(scr | SCR_NS_BIT);
isb();
ns_sctlr = read_sctlr();
sctlr = scr & SCR_HCE_BIT ? read_hsctlr() : ns_sctlr;
/* Return to original state */
write_scr(scr);
isb();
ee = 0;
ep_attr = NON_SECURE | EP_ST_DISABLE;
if (sctlr & SCTLR_EE_BIT) {
ep_attr |= EP_EE_BIG;
ee = 1;
}
SET_PARAM_HEAD(ep, PARAM_EP, VERSION_1, ep_attr);
ep->pc = entrypoint;
zeromem(&ep->args, sizeof(ep->args));
ep->args.arg0 = context_id;
mode = scr & SCR_HCE_BIT ? MODE32_hyp : MODE32_svc;
/*
* TODO: Choose async. exception bits if HYP mode is not
* implemented according to the values of SCR.{AW, FW} bits
*/
aif = SPSR_ABT_BIT | SPSR_IRQ_BIT | SPSR_FIQ_BIT;
ep->spsr = SPSR_MODE32(mode, entrypoint & 0x1, ee, aif);
return PSCI_E_SUCCESS;
}
#endif /* __aarch64__ */
/*******************************************************************************
* This function validates the entrypoint with the platform layer if the
* appropriate pm_ops hook is exported by the platform and returns the
* 'entry_point_info'.
******************************************************************************/
int psci_validate_entry_point(entry_point_info_t *ep,
uintptr_t entrypoint,
u_register_t context_id)
{
int rc;
/* Validate the entrypoint using platform psci_ops */
if (psci_plat_pm_ops->validate_ns_entrypoint != NULL) {
rc = psci_plat_pm_ops->validate_ns_entrypoint(entrypoint);
if (rc != PSCI_E_SUCCESS)
return PSCI_E_INVALID_ADDRESS;
}
/*
* Verify and derive the re-entry information for
* the non-secure world from the non-secure state from
* where this call originated.
*/
rc = psci_get_ns_ep_info(ep, entrypoint, context_id);
return rc;
}
/*******************************************************************************
* Generic handler which is called when a cpu is physically powered on. It
* traverses the node information and finds the highest power level powered
* off and performs generic, architectural, platform setup and state management
* to power on that power level and power levels below it.
* e.g. For a cpu that's been powered on, it will call the platform specific
* code to enable the gic cpu interface and for a cluster it will enable
* coherency at the interconnect level in addition to gic cpu interface.
******************************************************************************/
void psci_warmboot_entrypoint(void)
{
unsigned int end_pwrlvl;
unsigned int cpu_idx = plat_my_core_pos();
unsigned int parent_nodes[PLAT_MAX_PWR_LVL] = {0};
psci_power_state_t state_info = { {PSCI_LOCAL_STATE_RUN} };
/* Init registers that never change for the lifetime of TF-A */
cm_manage_extensions_el3(cpu_idx);
/*
* Verify that we have been explicitly turned ON or resumed from
* suspend.
*/
if (psci_get_aff_info_state() == AFF_STATE_OFF) {
ERROR("Unexpected affinity info state.\n");
panic();
}
/*
* Get the maximum power domain level to traverse to after this cpu
* has been physically powered up.
*/
end_pwrlvl = get_power_on_target_pwrlvl();
/* Get the parent nodes */
psci_get_parent_pwr_domain_nodes(cpu_idx, end_pwrlvl, parent_nodes);
/*
* This function acquires the lock corresponding to each power level so
* that by the time all locks are taken, the system topology is snapshot
* and state management can be done safely.
*/
psci_acquire_pwr_domain_locks(end_pwrlvl, parent_nodes);
psci_get_target_local_pwr_states(cpu_idx, end_pwrlvl, &state_info);
#if ENABLE_PSCI_STAT
plat_psci_stat_accounting_stop(&state_info);
#endif
/*
* This CPU could be resuming from suspend or it could have just been
* turned on. To distinguish between these 2 cases, we examine the
* affinity state of the CPU:
* - If the affinity state is ON_PENDING then it has just been
* turned on.
* - Else it is resuming from suspend.
*
* Depending on the type of warm reset identified, choose the right set
* of power management handler and perform the generic, architecture
* and platform specific handling.
*/
if (psci_get_aff_info_state() == AFF_STATE_ON_PENDING)
psci_cpu_on_finish(cpu_idx, &state_info);
else {
unsigned int max_off_lvl = psci_find_max_off_lvl(&state_info);
assert(max_off_lvl != PSCI_INVALID_PWR_LVL);
psci_cpu_suspend_to_powerdown_finish(cpu_idx, max_off_lvl, &state_info);
}
/*
* Generic management: Now we just need to retrieve the
* information that we had stashed away during the cpu_on
* call to set this cpu on its way.
*/
cm_prepare_el3_exit_ns();
/*
* Set the requested and target state of this CPU and all the higher
* power domains which are ancestors of this CPU to run.
*/
psci_set_pwr_domains_to_run(cpu_idx, end_pwrlvl);
#if ENABLE_PSCI_STAT
psci_stats_update_pwr_up(cpu_idx, end_pwrlvl, &state_info);
#endif
/*
* This loop releases the lock corresponding to each power level
* in the reverse order to which they were acquired.
*/
psci_release_pwr_domain_locks(end_pwrlvl, parent_nodes);
}
/*******************************************************************************
* This function initializes the set of hooks that PSCI invokes as part of power
* management operation. The power management hooks are expected to be provided
* by the SPD, after it finishes all its initialization
******************************************************************************/
void psci_register_spd_pm_hook(const spd_pm_ops_t *pm)
{
assert(pm != NULL);
psci_spd_pm = pm;
if (pm->svc_migrate != NULL)
psci_caps |= define_psci_cap(PSCI_MIG_AARCH64);
if (pm->svc_migrate_info != NULL)
psci_caps |= define_psci_cap(PSCI_MIG_INFO_UP_CPU_AARCH64)
| define_psci_cap(PSCI_MIG_INFO_TYPE);
}
/*******************************************************************************
* This function invokes the migrate info hook in the spd_pm_ops. It performs
* the necessary return value validation. If the Secure Payload is UP and
* migrate capable, it returns the mpidr of the CPU on which the Secure payload
* is resident through the mpidr parameter. Else the value of the parameter on
* return is undefined.
******************************************************************************/
int psci_spd_migrate_info(u_register_t *mpidr)
{
int rc;
if ((psci_spd_pm == NULL) || (psci_spd_pm->svc_migrate_info == NULL))
return PSCI_E_NOT_SUPPORTED;
rc = psci_spd_pm->svc_migrate_info(mpidr);
assert((rc == PSCI_TOS_UP_MIG_CAP) || (rc == PSCI_TOS_NOT_UP_MIG_CAP) ||
(rc == PSCI_TOS_NOT_PRESENT_MP) || (rc == PSCI_E_NOT_SUPPORTED));
return rc;
}
/*******************************************************************************
* This function prints the state of all power domains present in the
* system
******************************************************************************/
void psci_print_power_domain_map(void)
{
#if LOG_LEVEL >= LOG_LEVEL_INFO
unsigned int idx;
plat_local_state_t state;
plat_local_state_type_t state_type;
/* This array maps to the PSCI_STATE_X definitions in psci.h */
static const char * const psci_state_type_str[] = {
"ON",
"RETENTION",
"OFF",
};
INFO("PSCI Power Domain Map:\n");
for (idx = 0; idx < (PSCI_NUM_PWR_DOMAINS - psci_plat_core_count);
idx++) {
state_type = find_local_state_type(
psci_non_cpu_pd_nodes[idx].local_state);
INFO(" Domain Node : Level %u, parent_node %u,"
" State %s (0x%x)\n",
psci_non_cpu_pd_nodes[idx].level,
psci_non_cpu_pd_nodes[idx].parent_node,
psci_state_type_str[state_type],
psci_non_cpu_pd_nodes[idx].local_state);
}
for (idx = 0; idx < psci_plat_core_count; idx++) {
state = psci_get_cpu_local_state_by_idx(idx);
state_type = find_local_state_type(state);
INFO(" CPU Node : MPID 0x%llx, parent_node %u,"
" State %s (0x%x)\n",
(unsigned long long)psci_cpu_pd_nodes[idx].mpidr,
psci_cpu_pd_nodes[idx].parent_node,
psci_state_type_str[state_type],
psci_get_cpu_local_state_by_idx(idx));
}
#endif
}
/******************************************************************************
* Return whether any secondaries were powered up with CPU_ON call. A CPU that
* have ever been powered up would have set its MPDIR value to something other
* than PSCI_INVALID_MPIDR. Note that MPDIR isn't reset back to
* PSCI_INVALID_MPIDR when a CPU is powered down later, so the return value is
* meaningful only when called on the primary CPU during early boot.
*****************************************************************************/
int psci_secondaries_brought_up(void)
{
unsigned int idx, n_valid = 0U;
for (idx = 0U; idx < ARRAY_SIZE(psci_cpu_pd_nodes); idx++) {
if (psci_cpu_pd_nodes[idx].mpidr != PSCI_INVALID_MPIDR)
n_valid++;
}
assert(n_valid > 0U);
return (n_valid > 1U) ? 1 : 0;
}
/*******************************************************************************
* Initiate power down sequence, by calling power down operations registered for
* this CPU.
******************************************************************************/
void psci_pwrdown_cpu_start(unsigned int power_level)
{
#if ENABLE_RUNTIME_INSTRUMENTATION
/*
* Flush cache line so that even if CPU power down happens
* the timestamp update is reflected in memory.
*/
PMF_CAPTURE_TIMESTAMP(rt_instr_svc,
RT_INSTR_ENTER_CFLUSH,
PMF_CACHE_MAINT);
#endif
#if HW_ASSISTED_COHERENCY
/*
* With hardware-assisted coherency, the CPU drivers only initiate the
* power down sequence, without performing cache-maintenance operations
* in software. Data caches enabled both before and after this call.
*/
prepare_cpu_pwr_dwn(power_level);
#else
/*
* Without hardware-assisted coherency, the CPU drivers disable data
* caches, then perform cache-maintenance operations in software.
*
* This also calls prepare_cpu_pwr_dwn() to initiate power down
* sequence, but that function will return with data caches disabled.
* We must ensure that the stack memory is flushed out to memory before
* we start popping from it again.
*/
psci_do_pwrdown_cache_maintenance(power_level);
#endif
#if ENABLE_RUNTIME_INSTRUMENTATION
PMF_CAPTURE_TIMESTAMP(rt_instr_svc,
RT_INSTR_EXIT_CFLUSH,
PMF_NO_CACHE_MAINT);
#endif
}
/*******************************************************************************
* Finish a terminal power down sequence, ending with a wfi. In case of wakeup
* will retry the sleep and panic if it persists.
******************************************************************************/
void __dead2 psci_pwrdown_cpu_end_terminal(void)
{
#if ERRATA_SME_POWER_DOWN
/*
* force SME off to not get power down rejected. Getting here is
* terminal so we don't care if we lose context because of another
* wakeup
*/
if (is_feat_sme_supported()) {
write_svcr(0);
isb();
}
#endif /* ERRATA_SME_POWER_DOWN */
/*
* Execute a wfi which, in most cases, will allow the power controller
* to physically power down this cpu. Under some circumstances that may
* be denied. Hopefully this is transient, retrying a few times should
* power down.
*/
for (int i = 0; i < 32; i++)
psci_power_down_wfi();
/* Wake up wasn't transient. System is probably in a bad state. */
ERROR("Could not power off CPU.\n");
panic();
}
/*******************************************************************************
* Finish a non-terminal power down sequence, ending with a wfi. In case of
* wakeup will unwind any CPU specific actions and return.
******************************************************************************/
void psci_pwrdown_cpu_end_wakeup(unsigned int power_level)
{
/*
* Usually, will be terminal. In some circumstances the powerdown will
* be denied and we'll need to unwind
*/
psci_power_down_wfi();
/*
* Waking up does not require hardware-assisted coherency, but that is
* the case for every core that can wake up. Untangling the cache
* coherency code from powerdown is a non-trivial effort which isn't
* needed for our purposes.
*/
#if !FEAT_PABANDON
ERROR("Systems without FEAT_PABANDON shouldn't wake up.\n");
panic();
#else /* FEAT_PABANDON */
/*
* Begin unwinding. Everything can be shared with CPU_ON and co later,
* except the CPU specific bit. Cores that have hardware-assisted
* coherency don't have much to do so just calling the hook again is
* the simplest way to achieve this
*/
prepare_cpu_pwr_dwn(power_level);
#endif /* FEAT_PABANDON */
}
/*******************************************************************************
* This function invokes the callback 'stop_func()' with the 'mpidr' of each
* online PE. Caller can pass suitable method to stop a remote core.
*
* 'wait_ms' is the timeout value in milliseconds for the other cores to
* transition to power down state. Passing '0' makes it non-blocking.
*
* The function returns 'PSCI_E_DENIED' if some cores failed to stop within the
* given timeout.
******************************************************************************/
int psci_stop_other_cores(unsigned int this_cpu_idx, unsigned int wait_ms,
void (*stop_func)(u_register_t mpidr))
{
/* Invoke stop_func for each core */
for (unsigned int idx = 0U; idx < psci_plat_core_count; idx++) {
/* skip current CPU */
if (idx == this_cpu_idx) {
continue;
}
/* Check if the CPU is ON */
if (psci_get_aff_info_state_by_idx(idx) == AFF_STATE_ON) {
(*stop_func)(psci_cpu_pd_nodes[idx].mpidr);
}
}
/* Need to wait for other cores to shutdown */
if (wait_ms != 0U) {
while ((wait_ms-- != 0U) && (!psci_is_last_on_cpu(this_cpu_idx))) {
mdelay(1U);
}
if (!psci_is_last_on_cpu(this_cpu_idx)) {
WARN("Failed to stop all cores!\n");
psci_print_power_domain_map();
return PSCI_E_DENIED;
}
}
return PSCI_E_SUCCESS;
}
/*******************************************************************************
* This function verifies that all the other cores in the system have been
* turned OFF and the current CPU is the last running CPU in the system.
* Returns true if the current CPU is the last ON CPU or false otherwise.
*
* This API has following differences with psci_is_last_on_cpu
* 1. PSCI states are locked
******************************************************************************/
bool psci_is_last_on_cpu_safe(unsigned int this_core)
{
unsigned int parent_nodes[PLAT_MAX_PWR_LVL] = {0};
psci_get_parent_pwr_domain_nodes(this_core, PLAT_MAX_PWR_LVL, parent_nodes);
psci_acquire_pwr_domain_locks(PLAT_MAX_PWR_LVL, parent_nodes);
if (!psci_is_last_on_cpu(this_core)) {
psci_release_pwr_domain_locks(PLAT_MAX_PWR_LVL, parent_nodes);
return false;
}
psci_release_pwr_domain_locks(PLAT_MAX_PWR_LVL, parent_nodes);
return true;
}
/*******************************************************************************
* This function verifies that all cores in the system have been turned ON.
* Returns true, if all CPUs are ON or false otherwise.
*
* This API has following differences with psci_are_all_cpus_on
* 1. PSCI states are locked
******************************************************************************/
bool psci_are_all_cpus_on_safe(unsigned int this_core)
{
unsigned int parent_nodes[PLAT_MAX_PWR_LVL] = {0};
psci_get_parent_pwr_domain_nodes(this_core, PLAT_MAX_PWR_LVL, parent_nodes);
psci_acquire_pwr_domain_locks(PLAT_MAX_PWR_LVL, parent_nodes);
if (!psci_are_all_cpus_on()) {
psci_release_pwr_domain_locks(PLAT_MAX_PWR_LVL, parent_nodes);
return false;
}
psci_release_pwr_domain_locks(PLAT_MAX_PWR_LVL, parent_nodes);
return true;
}