Djam/arm-trusted-firmware
1
0
Fork 0
mirror of https://github.com/ARM-software/arm-trusted-firmware.git synced 2025-04-27 23:35:10 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

Merge pull request #702 from jeenu-arm/psci-node-hw-state

Browse source 
Support for PSCI NODE_HW_STATE
This commit is contained in:
danh-arm 2016-09-19 11:55:56 +01:00 • committed by GitHub
commit 7a1b279430
13 changed files with 253 additions and 6 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

2
docs/firmware-design.md
View file

@ -716,7 +716,7 @@ required support.
|`PSCI_FEATURES` | Yes | |
|`CPU_FREEZE` | No | |
|`CPU_DEFAULT_SUSPEND` | No | |
|`CPU_HW_STATE` | No | |
|`NODE_HW_STATE` | Yes* | |
|`SYSTEM_SUSPEND` | Yes* | |
|`PSCI_SET_SUSPEND_MODE`| No | |
|`PSCI_STAT_RESIDENCY` | Yes* | |

14
docs/porting-guide.md
View file

@ -1832,6 +1832,20 @@ This function can also be used in case the platform wants to support local
power state encoding for `power_state` parameter of PSCI_STAT_COUNT/RESIDENCY
APIs as described in Section 5.18 of [PSCI].
#### plat_psci_ops.get_node_hw_state()
This is an optional function. If implemented this function is intended to return
the power state of a node (identified by the first parameter, the `MPIDR`) in
the power domain topology (identified by the second parameter, `power_level`),
as retrieved from a power controller or equivalent component on the platform.
Upon successful completion, the implementation must map and return the final
status among `HW_ON`, `HW_OFF` or `HW_STANDBY`. Upon encountering failures, it
must return either `PSCI_E_INVALID_PARAMS` or `PSCI_E_NOT_SUPPORTED` as
appropriate.
Implementations are not expected to handle `power_levels` greater than
`PLAT_MAX_PWR_LVL`.
3.6 Interrupt Management framework (in BL31)
----------------------------------------------
BL31 implements an Interrupt Management Framework (IMF) to manage interrupts

16
include/lib/psci/psci.h
View file

@ -78,6 +78,8 @@
#define PSCI_SYSTEM_OFF 0x84000008
#define PSCI_SYSTEM_RESET 0x84000009
#define PSCI_FEATURES 0x8400000A
#define PSCI_NODE_HW_STATE_AARCH32 0x8400000d
#define PSCI_NODE_HW_STATE_AARCH64 0xc400000d
#define PSCI_SYSTEM_SUSPEND_AARCH32 0x8400000E
#define PSCI_SYSTEM_SUSPEND_AARCH64 0xc400000E
#define PSCI_STAT_RESIDENCY_AARCH32 0x84000010
@ -199,6 +201,17 @@ typedef enum {
AFF_STATE_ON_PENDING = 2
} aff_info_state_t;
/*
* These are the power states reported by PSCI_NODE_HW_STATE API for the
* specified CPU. The definitions of these states can be found in Section 5.15.3
* of PSCI specification (ARM DEN 0022C).
*/
typedef enum {
HW_ON = 0,
HW_OFF = 1,
HW_STANDBY = 2
} node_hw_state_t;
/*
* Macro to represent invalid affinity level within PSCI.
*/
@ -293,6 +306,7 @@ typedef struct plat_psci_ops {
int (*translate_power_state_by_mpidr)(u_register_t mpidr,
unsigned int power_state,
psci_power_state_t *output_state);
int (*get_node_hw_state)(u_register_t mpidr, unsigned int power_level);
} plat_psci_ops_t;
/*******************************************************************************
@ -330,6 +344,8 @@ int psci_affinity_info(u_register_t target_affinity,
int psci_migrate(u_register_t target_cpu);
int psci_migrate_info_type(void);
long psci_migrate_info_up_cpu(void);
int psci_node_hw_state(u_register_t target_cpu,
unsigned int power_level);
int psci_features(unsigned int psci_fid);
void __dead2 psci_power_down_wfi(void);
void psci_arch_setup(void);

10
include/plat/arm/css/common/css_def.h
View file

@ -148,5 +148,15 @@
/* Trusted mailbox base address common to all CSS */
#define PLAT_ARM_TRUSTED_MAILBOX_BASE ARM_TRUSTED_SRAM_BASE
/*
* Parsing of CPU and Cluster states, as returned by 'Get CSS Power State' SCP
* command
*/
#define CSS_CLUSTER_PWR_STATE_ON 0
#define CSS_CLUSTER_PWR_STATE_OFF 3
#define CSS_CPU_PWR_STATE_ON 1
#define CSS_CPU_PWR_STATE_OFF 0
#define CSS_CPU_PWR_STATE(state, n) (((state) >> (n)) & 1)
#endif /* __CSS_DEF_H__ */

1
include/plat/arm/css/common/css_pm.h
View file

@ -45,5 +45,6 @@ void __dead2 css_system_off(void);
void __dead2 css_system_reset(void);
void css_cpu_standby(plat_local_state_t cpu_state);
void css_get_sys_suspend_power_state(psci_power_state_t *req_state);
int css_node_hw_state(u_register_t mpidr, unsigned int power_level);
#endif /* __CSS_PM_H__ */

31
lib/psci/psci_main.c
View file

@ -295,6 +295,31 @@ long psci_migrate_info_up_cpu(void)
return resident_cpu_mpidr;
}
int psci_node_hw_state(u_register_t target_cpu,
unsigned int power_level)
{
int rc;
/* Validate target_cpu */
rc = psci_validate_mpidr(target_cpu);
if (rc != PSCI_E_SUCCESS)
return PSCI_E_INVALID_PARAMS;
/* Validate power_level against PLAT_MAX_PWR_LVL */
if (power_level > PLAT_MAX_PWR_LVL)
return PSCI_E_INVALID_PARAMS;
/*
* Dispatch this call to platform to query power controller, and pass on
* to the caller what it returns
*/
assert(psci_plat_pm_ops->get_node_hw_state);
rc = psci_plat_pm_ops->get_node_hw_state(target_cpu, power_level);
assert((rc >= HW_ON && rc <= HW_STANDBY) || rc == PSCI_E_NOT_SUPPORTED
|| rc == PSCI_E_INVALID_PARAMS);
return rc;
}
int psci_features(unsigned int psci_fid)
{
unsigned int local_caps = psci_caps;
@ -378,6 +403,9 @@ u_register_t psci_smc_handler(uint32_t smc_fid,
case PSCI_MIG_INFO_UP_CPU_AARCH32:
return psci_migrate_info_up_cpu();
case PSCI_NODE_HW_STATE_AARCH32:
return psci_node_hw_state(x1, x2);
case PSCI_SYSTEM_SUSPEND_AARCH32:
return psci_system_suspend(x1, x2);
@ -422,6 +450,9 @@ u_register_t psci_smc_handler(uint32_t smc_fid,
case PSCI_MIG_INFO_UP_CPU_AARCH64:
return psci_migrate_info_up_cpu();
case PSCI_NODE_HW_STATE_AARCH64:
return psci_node_hw_state(x1, x2);
case PSCI_SYSTEM_SUSPEND_AARCH64:
return psci_system_suspend(x1, x2);

1
lib/psci/psci_private.h
View file

@ -68,6 +68,7 @@
define_psci_cap(PSCI_AFFINITY_INFO_AARCH64) | \
define_psci_cap(PSCI_MIG_AARCH64) | \
define_psci_cap(PSCI_MIG_INFO_UP_CPU_AARCH64) | \
define_psci_cap(PSCI_NODE_HW_STATE_AARCH64) | \
define_psci_cap(PSCI_SYSTEM_SUSPEND_AARCH64) | \
define_psci_cap(PSCI_STAT_RESIDENCY_AARCH64) | \
define_psci_cap(PSCI_STAT_COUNT_AARCH64))

2
lib/psci/psci_setup.c
View file

@ -263,6 +263,8 @@ int psci_setup(uintptr_t mailbox_ep)
psci_caps |= define_psci_cap(PSCI_SYSTEM_OFF);
if (psci_plat_pm_ops->system_reset)
psci_caps |= define_psci_cap(PSCI_SYSTEM_RESET);
if (psci_plat_pm_ops->get_node_hw_state)
psci_caps |= define_psci_cap(PSCI_NODE_HW_STATE_AARCH64);
#if ENABLE_PSCI_STAT
psci_caps |= define_psci_cap(PSCI_STAT_RESIDENCY_AARCH64);

39
plat/arm/board/fvp/fvp_pm.c
View file

@ -287,6 +287,42 @@ static void __dead2 fvp_system_reset(void)
panic();
}
static int fvp_node_hw_state(u_register_t target_cpu,
unsigned int power_level)
{
unsigned int psysr;
int ret;
/*
* The format of 'power_level' is implementation-defined, but 0 must
* mean a CPU. We also allow 1 to denote the cluster
*/
if (power_level != ARM_PWR_LVL0 && power_level != ARM_PWR_LVL1)
return PSCI_E_INVALID_PARAMS;
/*
* Read the status of the given MPDIR from FVP power controller. The
* power controller only gives us on/off status, so map that to expected
* return values of the PSCI call
*/
psysr = fvp_pwrc_read_psysr(target_cpu);
if (psysr == PSYSR_INVALID)
return PSCI_E_INVALID_PARAMS;
switch (power_level) {
case ARM_PWR_LVL0:
ret = (psysr & PSYSR_AFF_L0) ? HW_ON : HW_OFF;
break;
case ARM_PWR_LVL1:
ret = (psysr & PSYSR_AFF_L1) ? HW_ON : HW_OFF;
break;
default:
assert(0);
}
return ret;
}
/*******************************************************************************
* Export the platform handlers via plat_arm_psci_pm_ops. The ARM Standard
* platform layer will take care of registering the handlers with PSCI.
@ -301,5 +337,6 @@ const plat_psci_ops_t plat_arm_psci_pm_ops = {
.system_off = fvp_system_off,
.system_reset = fvp_system_reset,
.validate_power_state = arm_validate_power_state,
.validate_ns_entrypoint = arm_validate_ns_entrypoint
.validate_ns_entrypoint = arm_validate_ns_entrypoint,
.get_node_hw_state = fvp_node_hw_state
};

5
plat/arm/board/juno/juno_pm.c
View file

@ -1,5 +1,5 @@
/*
* Copyright (c) 2015, ARM Limited and Contributors. All rights reserved.
* Copyright (c) 2015-2016, ARM Limited and Contributors. All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions are met:
@ -88,5 +88,6 @@ const plat_psci_ops_t plat_arm_psci_pm_ops = {
.validate_power_state = juno_validate_power_state,
.validate_ns_entrypoint = arm_validate_ns_entrypoint,
.get_sys_suspend_power_state = css_get_sys_suspend_power_state,
.translate_power_state_by_mpidr = juno_translate_power_state_by_mpidr
.translate_power_state_by_mpidr = juno_translate_power_state_by_mpidr,
.get_node_hw_state = css_node_hw_state
};

40
plat/arm/css/common/css_pm.c
View file

@ -298,6 +298,43 @@ void css_get_sys_suspend_power_state(psci_power_state_t *req_state)
req_state->pwr_domain_state[i] = ARM_LOCAL_STATE_OFF;
}
/*******************************************************************************
* Handler to query CPU/cluster power states from SCP
******************************************************************************/
int css_node_hw_state(u_register_t mpidr, unsigned int power_level)
{
int rc, element;
unsigned int cpu_state, cluster_state;
/*
* The format of 'power_level' is implementation-defined, but 0 must
* mean a CPU. We also allow 1 to denote the cluster
*/
if (power_level != ARM_PWR_LVL0 && power_level != ARM_PWR_LVL1)
return PSCI_E_INVALID_PARAMS;
/* Query SCP */
rc = scpi_get_css_power_state(mpidr, &cpu_state, &cluster_state);
if (rc != 0)
return PSCI_E_INVALID_PARAMS;
/* Map power states of CPU and cluster to expected PSCI return codes */
if (power_level == ARM_PWR_LVL0) {
/*
* The CPU state returned by SCP is an 8-bit bit mask
* corresponding to each CPU in the cluster
*/
element = mpidr & MPIDR_AFFLVL_MASK;
return CSS_CPU_PWR_STATE(cpu_state, element) ==
CSS_CPU_PWR_STATE_ON ? HW_ON : HW_OFF;
} else {
assert(cluster_state == CSS_CLUSTER_PWR_STATE_ON ||
cluster_state == CSS_CLUSTER_PWR_STATE_OFF);
return cluster_state == CSS_CLUSTER_PWR_STATE_ON ? HW_ON :
HW_OFF;
}
}
/*******************************************************************************
* Export the platform handlers via plat_arm_psci_pm_ops. The ARM Standard
* platform will take care of registering the handlers with PSCI.
@ -312,5 +349,6 @@ const plat_psci_ops_t plat_arm_psci_pm_ops = {
.system_off = css_system_off,
.system_reset = css_system_reset,
.validate_power_state = arm_validate_power_state,
.validate_ns_entrypoint = arm_validate_ns_entrypoint
.validate_ns_entrypoint = arm_validate_ns_entrypoint,
.get_node_hw_state = css_node_hw_state
};

69
plat/arm/css/common/css_scpi.c
View file

@ -41,11 +41,18 @@
#define SCPI_SHARED_MEM_AP_TO_SCP (PLAT_CSS_SCP_COM_SHARED_MEM_BASE \
+ 0x100)
/* Header and payload addresses for commands from AP to SCP */
#define SCPI_CMD_HEADER_AP_TO_SCP \
((scpi_cmd_t *) SCPI_SHARED_MEM_AP_TO_SCP)
#define SCPI_CMD_PAYLOAD_AP_TO_SCP \
((void *) (SCPI_SHARED_MEM_AP_TO_SCP + sizeof(scpi_cmd_t)))
/* Header and payload addresses for responses from SCP to AP */
#define SCPI_RES_HEADER_SCP_TO_AP \
((scpi_cmd_t *) SCPI_SHARED_MEM_SCP_TO_AP)
#define SCPI_RES_PAYLOAD_SCP_TO_AP \
((void *) (SCPI_SHARED_MEM_SCP_TO_AP + sizeof(scpi_cmd_t)))
/* ID of the MHU slot used for the SCPI protocol */
#define SCPI_MHU_SLOT_ID 0
@ -163,6 +170,68 @@ void scpi_set_css_power_state(unsigned mpidr, scpi_power_state_t cpu_state,
scpi_secure_message_end();
}
/*
* Query and obtain CSS power state from SCP.
*
* In response to the query, SCP returns power states of all CPUs in all
* clusters of the system. The returned response is then filtered based on the
* supplied MPIDR. Power states of requested cluster and CPUs within are updated
* via. supplied non-NULL pointer arguments.
*
* Returns 0 on success, or -1 on errors.
*/
int scpi_get_css_power_state(unsigned int mpidr, unsigned int *cpu_state_p,
unsigned int *cluster_state_p)
{
scpi_cmd_t *cmd;
scpi_cmd_t response;
int power_state, cpu, cluster, rc = -1;
/*
* Extract CPU and cluster membership of the given MPIDR. SCPI caters
* for only up to 0xf clusters, and 8 CPUs per cluster
*/
cpu = mpidr & MPIDR_AFFLVL_MASK;
cluster = (mpidr >> MPIDR_AFF1_SHIFT) & MPIDR_AFFLVL_MASK;
if (cpu >= 8 || cluster >= 0xf)
return -1;
scpi_secure_message_start();
/* Populate request headers */
cmd = memset(SCPI_CMD_HEADER_AP_TO_SCP, 0, sizeof(*cmd));
cmd->id = SCPI_CMD_GET_CSS_POWER_STATE;
/*
* Send message and wait for SCP's response
*/
scpi_secure_message_send(0);
scpi_secure_message_receive(&response);
if (response.status != SCP_OK)
goto exit;
/* Validate SCP response */
if (!CHECK_RESPONSE(response, cluster))
goto exit;
/* Extract power states for required cluster */
power_state = *(((uint16_t *) SCPI_RES_PAYLOAD_SCP_TO_AP) + cluster);
if (CLUSTER_ID(power_state) != cluster)
goto exit;
/* Update power state via. pointers */
if (cluster_state_p)
*cluster_state_p = CLUSTER_POWER_STATE(power_state);
if (cpu_state_p)
*cpu_state_p = CPU_POWER_STATE(power_state);
rc = 0;
exit:
scpi_secure_message_end();
return rc;
}
uint32_t scpi_sys_power_state(scpi_system_state_t system_state)
{
scpi_cmd_t *cmd;

29
plat/arm/css/common/css_scpi.h
View file

@ -1,5 +1,5 @@
/*
* Copyright (c) 2014-2015, ARM Limited and Contributors. All rights reserved.
* Copyright (c) 2014-2016, ARM Limited and Contributors. All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions are met:
@ -81,9 +81,34 @@ typedef uint32_t scpi_status_t;
typedef enum {
SCPI_CMD_SCP_READY = 0x01,
SCPI_CMD_SET_CSS_POWER_STATE = 0x03,
SCPI_CMD_GET_CSS_POWER_STATE = 0x04,
SCPI_CMD_SYS_POWER_STATE = 0x05
} scpi_command_t;
/*
* Macros to parse SCP response to GET_CSS_POWER_STATE command
*
* [3:0] : cluster ID
* [7:4] : cluster state: 0 = on; 3 = off; rest are reserved
* [15:8]: on/off state for individual CPUs in the cluster
*
* Payload is in little-endian
*/
#define CLUSTER_ID(_resp) ((_resp) & 0xf)
#define CLUSTER_POWER_STATE(_resp) (((_resp) >> 4) & 0xf)
/* Result is a bit mask of CPU on/off states in the cluster */
#define CPU_POWER_STATE(_resp) (((_resp) >> 8) & 0xff)
/*
* For GET_CSS_POWER_STATE, SCP returns the power states of every cluster. The
* size of response depends on the number of clusters in the system. The
* SCP-to-AP payload contains 2 bytes per cluster. Make sure the response is
* large enough to contain power states of a given cluster
*/
#define CHECK_RESPONSE(_resp, _clus) \
(_resp.size >= (((_clus) + 1) * 2))
typedef enum {
scpi_power_on = 0,
scpi_power_retention = 1,
@ -101,6 +126,8 @@ extern void scpi_set_css_power_state(unsigned mpidr,
scpi_power_state_t cpu_state,
scpi_power_state_t cluster_state,
scpi_power_state_t css_state);
int scpi_get_css_power_state(unsigned int mpidr, unsigned int *cpu_state_p,
unsigned int *cluster_state_p);
uint32_t scpi_sys_power_state(scpi_system_state_t system_state);