arm-trusted-firmware/services/std_svc/psci/psci_afflvl_on.c

486 lines
15 KiB
C
Raw Normal View History

2013-10-25 09:08:21 +01:00
/*
* Copyright (c) 2013-2014, ARM Limited and Contributors. All rights reserved.
2013-10-25 09:08:21 +01:00
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions are met:
*
* Redistributions of source code must retain the above copyright notice, this
* list of conditions and the following disclaimer.
*
* Redistributions in binary form must reproduce the above copyright notice,
* this list of conditions and the following disclaimer in the documentation
* and/or other materials provided with the distribution.
*
* Neither the name of ARM nor the names of its contributors may be used
* to endorse or promote products derived from this software without specific
* prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
* AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE
* LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
* CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
* SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
* INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
* CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
* ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
* POSSIBILITY OF SUCH DAMAGE.
*/
#include <stdio.h>
#include <string.h>
#include <assert.h>
#include <debug.h>
2013-10-25 09:08:21 +01:00
#include <arch_helpers.h>
#include <console.h>
#include <platform.h>
#include <psci.h>
#include <context_mgmt.h>
#include "psci_private.h"
2013-10-25 09:08:21 +01:00
typedef int (*afflvl_on_handler)(unsigned long,
aff_map_node *,
unsigned long,
unsigned long);
/*******************************************************************************
* This function checks whether a cpu which has been requested to be turned on
* is OFF to begin with.
******************************************************************************/
psci: fix affinity level upgrade issue The psci implementation does not track target affinity level requests specified during cpu_suspend calls correctly as per the following example. 1. cpu0.cluster0 calls cpu_suspend with the target affinity level as 0 2. Only the cpu0.cluster0 is powered down while cluster0 remains powered up 3. cpu1.cluster0 calls cpu_off to power itself down to highest possible affinity level 4. cluster0 will be powered off even though cpu0.cluster0 does not allow cluster shutdown This patch introduces reference counts at affinity levels > 0 to track the number of cpus which want an affinity instance at level X to remain powered up. This instance can be turned off only if its reference count is 0. Cpus still undergo the normal state transitions (ON, OFF, ON_PENDING, SUSPEND) but the higher levels can only be either ON or OFF depending upon their reference count. The above issue is thus fixed as follows: 1. cluster0's reference count is incremented by two when cpu0 and cpu1 are initially powered on. 2. cpu0.cluster0 calls cpu_suspend with the target affinity level as 0. This does not affect the cluster0 reference count. 3. Only the cpu0.cluster0 is powered down while cluster0 remains powered up as it has a non-zero reference count. 4. cpu1.cluster0 call cpu_off to power itself down to highest possible affinity level. This decrements the cluster0 reference count. 5. cluster0 is still not powered off since its reference count will at least be 1 due to the restriction placed by cpu0. Change-Id: I433dfe82b946f5f6985b1602c2de87800504f7a9
2013-12-05 16:33:10 +00:00
static int cpu_on_validate_state(aff_map_node *node)
2013-10-25 09:08:21 +01:00
{
unsigned int psci_state;
/* Get the raw psci state */
psci: fix affinity level upgrade issue The psci implementation does not track target affinity level requests specified during cpu_suspend calls correctly as per the following example. 1. cpu0.cluster0 calls cpu_suspend with the target affinity level as 0 2. Only the cpu0.cluster0 is powered down while cluster0 remains powered up 3. cpu1.cluster0 calls cpu_off to power itself down to highest possible affinity level 4. cluster0 will be powered off even though cpu0.cluster0 does not allow cluster shutdown This patch introduces reference counts at affinity levels > 0 to track the number of cpus which want an affinity instance at level X to remain powered up. This instance can be turned off only if its reference count is 0. Cpus still undergo the normal state transitions (ON, OFF, ON_PENDING, SUSPEND) but the higher levels can only be either ON or OFF depending upon their reference count. The above issue is thus fixed as follows: 1. cluster0's reference count is incremented by two when cpu0 and cpu1 are initially powered on. 2. cpu0.cluster0 calls cpu_suspend with the target affinity level as 0. This does not affect the cluster0 reference count. 3. Only the cpu0.cluster0 is powered down while cluster0 remains powered up as it has a non-zero reference count. 4. cpu1.cluster0 call cpu_off to power itself down to highest possible affinity level. This decrements the cluster0 reference count. 5. cluster0 is still not powered off since its reference count will at least be 1 due to the restriction placed by cpu0. Change-Id: I433dfe82b946f5f6985b1602c2de87800504f7a9
2013-12-05 16:33:10 +00:00
psci_state = psci_get_state(node);
2013-10-25 09:08:21 +01:00
if (psci_state == PSCI_STATE_ON || psci_state == PSCI_STATE_SUSPEND)
return PSCI_E_ALREADY_ON;
if (psci_state == PSCI_STATE_ON_PENDING)
return PSCI_E_ON_PENDING;
assert(psci_state == PSCI_STATE_OFF);
return PSCI_E_SUCCESS;
}
/*******************************************************************************
* Handler routine to turn a cpu on. It takes care of any generic, architectural
* or platform specific setup required.
* TODO: Split this code across separate handlers for each type of setup?
******************************************************************************/
static int psci_afflvl0_on(unsigned long target_cpu,
aff_map_node *cpu_node,
unsigned long ns_entrypoint,
unsigned long context_id)
{
unsigned int index, plat_state;
unsigned long psci_entrypoint;
int rc;
/* Sanity check to safeguard against data corruption */
assert(cpu_node->level == MPIDR_AFFLVL0);
/*
* Generic management: Ensure that the cpu is off to be
* turned on
*/
psci: fix affinity level upgrade issue The psci implementation does not track target affinity level requests specified during cpu_suspend calls correctly as per the following example. 1. cpu0.cluster0 calls cpu_suspend with the target affinity level as 0 2. Only the cpu0.cluster0 is powered down while cluster0 remains powered up 3. cpu1.cluster0 calls cpu_off to power itself down to highest possible affinity level 4. cluster0 will be powered off even though cpu0.cluster0 does not allow cluster shutdown This patch introduces reference counts at affinity levels > 0 to track the number of cpus which want an affinity instance at level X to remain powered up. This instance can be turned off only if its reference count is 0. Cpus still undergo the normal state transitions (ON, OFF, ON_PENDING, SUSPEND) but the higher levels can only be either ON or OFF depending upon their reference count. The above issue is thus fixed as follows: 1. cluster0's reference count is incremented by two when cpu0 and cpu1 are initially powered on. 2. cpu0.cluster0 calls cpu_suspend with the target affinity level as 0. This does not affect the cluster0 reference count. 3. Only the cpu0.cluster0 is powered down while cluster0 remains powered up as it has a non-zero reference count. 4. cpu1.cluster0 call cpu_off to power itself down to highest possible affinity level. This decrements the cluster0 reference count. 5. cluster0 is still not powered off since its reference count will at least be 1 due to the restriction placed by cpu0. Change-Id: I433dfe82b946f5f6985b1602c2de87800504f7a9
2013-12-05 16:33:10 +00:00
rc = cpu_on_validate_state(cpu_node);
2013-10-25 09:08:21 +01:00
if (rc != PSCI_E_SUCCESS)
return rc;
/*
* Call the cpu on handler registered by the Secure Payload Dispatcher
* to let it do any bookeeping. If the handler encounters an error, it's
* expected to assert within
*/
if (psci_spd_pm && psci_spd_pm->svc_on)
psci_spd_pm->svc_on(target_cpu);
2013-10-25 09:08:21 +01:00
/*
* Arch. management: Derive the re-entry information for
* the non-secure world from the non-secure state from
* where this call originated.
*/
index = cpu_node->data;
rc = psci_set_ns_entry_info(index, ns_entrypoint, context_id);
if (rc != PSCI_E_SUCCESS)
return rc;
/* Set the secure world (EL3) re-entry point after BL1 */
psci_entrypoint = (unsigned long) psci_aff_on_finish_entry;
psci: fix affinity level upgrade issue The psci implementation does not track target affinity level requests specified during cpu_suspend calls correctly as per the following example. 1. cpu0.cluster0 calls cpu_suspend with the target affinity level as 0 2. Only the cpu0.cluster0 is powered down while cluster0 remains powered up 3. cpu1.cluster0 calls cpu_off to power itself down to highest possible affinity level 4. cluster0 will be powered off even though cpu0.cluster0 does not allow cluster shutdown This patch introduces reference counts at affinity levels > 0 to track the number of cpus which want an affinity instance at level X to remain powered up. This instance can be turned off only if its reference count is 0. Cpus still undergo the normal state transitions (ON, OFF, ON_PENDING, SUSPEND) but the higher levels can only be either ON or OFF depending upon their reference count. The above issue is thus fixed as follows: 1. cluster0's reference count is incremented by two when cpu0 and cpu1 are initially powered on. 2. cpu0.cluster0 calls cpu_suspend with the target affinity level as 0. This does not affect the cluster0 reference count. 3. Only the cpu0.cluster0 is powered down while cluster0 remains powered up as it has a non-zero reference count. 4. cpu1.cluster0 call cpu_off to power itself down to highest possible affinity level. This decrements the cluster0 reference count. 5. cluster0 is still not powered off since its reference count will at least be 1 due to the restriction placed by cpu0. Change-Id: I433dfe82b946f5f6985b1602c2de87800504f7a9
2013-12-05 16:33:10 +00:00
/* State management: Set this cpu's state as ON PENDING */
psci_set_state(cpu_node, PSCI_STATE_ON_PENDING);
2013-10-25 09:08:21 +01:00
/*
* Plat. management: Give the platform the current state
* of the target cpu to allow it to perform the necessary
* steps to power on.
*/
if (psci_plat_pm_ops->affinst_on) {
/* Get the current physical state of this cpu */
psci: fix affinity level upgrade issue The psci implementation does not track target affinity level requests specified during cpu_suspend calls correctly as per the following example. 1. cpu0.cluster0 calls cpu_suspend with the target affinity level as 0 2. Only the cpu0.cluster0 is powered down while cluster0 remains powered up 3. cpu1.cluster0 calls cpu_off to power itself down to highest possible affinity level 4. cluster0 will be powered off even though cpu0.cluster0 does not allow cluster shutdown This patch introduces reference counts at affinity levels > 0 to track the number of cpus which want an affinity instance at level X to remain powered up. This instance can be turned off only if its reference count is 0. Cpus still undergo the normal state transitions (ON, OFF, ON_PENDING, SUSPEND) but the higher levels can only be either ON or OFF depending upon their reference count. The above issue is thus fixed as follows: 1. cluster0's reference count is incremented by two when cpu0 and cpu1 are initially powered on. 2. cpu0.cluster0 calls cpu_suspend with the target affinity level as 0. This does not affect the cluster0 reference count. 3. Only the cpu0.cluster0 is powered down while cluster0 remains powered up as it has a non-zero reference count. 4. cpu1.cluster0 call cpu_off to power itself down to highest possible affinity level. This decrements the cluster0 reference count. 5. cluster0 is still not powered off since its reference count will at least be 1 due to the restriction placed by cpu0. Change-Id: I433dfe82b946f5f6985b1602c2de87800504f7a9
2013-12-05 16:33:10 +00:00
plat_state = psci_get_phys_state(cpu_node);
2013-10-25 09:08:21 +01:00
rc = psci_plat_pm_ops->affinst_on(target_cpu,
psci_entrypoint,
ns_entrypoint,
cpu_node->level,
plat_state);
}
return rc;
}
/*******************************************************************************
* Handler routine to turn a cluster on. It takes care or any generic, arch.
* or platform specific setup required.
* TODO: Split this code across separate handlers for each type of setup?
******************************************************************************/
static int psci_afflvl1_on(unsigned long target_cpu,
aff_map_node *cluster_node,
unsigned long ns_entrypoint,
unsigned long context_id)
{
int rc = PSCI_E_SUCCESS;
unsigned int plat_state;
unsigned long psci_entrypoint;
assert(cluster_node->level == MPIDR_AFFLVL1);
/*
* There is no generic and arch. specific cluster
* management required
*/
psci: fix affinity level upgrade issue The psci implementation does not track target affinity level requests specified during cpu_suspend calls correctly as per the following example. 1. cpu0.cluster0 calls cpu_suspend with the target affinity level as 0 2. Only the cpu0.cluster0 is powered down while cluster0 remains powered up 3. cpu1.cluster0 calls cpu_off to power itself down to highest possible affinity level 4. cluster0 will be powered off even though cpu0.cluster0 does not allow cluster shutdown This patch introduces reference counts at affinity levels > 0 to track the number of cpus which want an affinity instance at level X to remain powered up. This instance can be turned off only if its reference count is 0. Cpus still undergo the normal state transitions (ON, OFF, ON_PENDING, SUSPEND) but the higher levels can only be either ON or OFF depending upon their reference count. The above issue is thus fixed as follows: 1. cluster0's reference count is incremented by two when cpu0 and cpu1 are initially powered on. 2. cpu0.cluster0 calls cpu_suspend with the target affinity level as 0. This does not affect the cluster0 reference count. 3. Only the cpu0.cluster0 is powered down while cluster0 remains powered up as it has a non-zero reference count. 4. cpu1.cluster0 call cpu_off to power itself down to highest possible affinity level. This decrements the cluster0 reference count. 5. cluster0 is still not powered off since its reference count will at least be 1 due to the restriction placed by cpu0. Change-Id: I433dfe82b946f5f6985b1602c2de87800504f7a9
2013-12-05 16:33:10 +00:00
/* State management: Is not required while turning a cluster on */
2013-10-25 09:08:21 +01:00
/*
* Plat. management: Give the platform the current state
* of the target cpu to allow it to perform the necessary
* steps to power on.
*/
if (psci_plat_pm_ops->affinst_on) {
psci: fix affinity level upgrade issue The psci implementation does not track target affinity level requests specified during cpu_suspend calls correctly as per the following example. 1. cpu0.cluster0 calls cpu_suspend with the target affinity level as 0 2. Only the cpu0.cluster0 is powered down while cluster0 remains powered up 3. cpu1.cluster0 calls cpu_off to power itself down to highest possible affinity level 4. cluster0 will be powered off even though cpu0.cluster0 does not allow cluster shutdown This patch introduces reference counts at affinity levels > 0 to track the number of cpus which want an affinity instance at level X to remain powered up. This instance can be turned off only if its reference count is 0. Cpus still undergo the normal state transitions (ON, OFF, ON_PENDING, SUSPEND) but the higher levels can only be either ON or OFF depending upon their reference count. The above issue is thus fixed as follows: 1. cluster0's reference count is incremented by two when cpu0 and cpu1 are initially powered on. 2. cpu0.cluster0 calls cpu_suspend with the target affinity level as 0. This does not affect the cluster0 reference count. 3. Only the cpu0.cluster0 is powered down while cluster0 remains powered up as it has a non-zero reference count. 4. cpu1.cluster0 call cpu_off to power itself down to highest possible affinity level. This decrements the cluster0 reference count. 5. cluster0 is still not powered off since its reference count will at least be 1 due to the restriction placed by cpu0. Change-Id: I433dfe82b946f5f6985b1602c2de87800504f7a9
2013-12-05 16:33:10 +00:00
plat_state = psci_get_phys_state(cluster_node);
2013-10-25 09:08:21 +01:00
psci_entrypoint = (unsigned long) psci_aff_on_finish_entry;
rc = psci_plat_pm_ops->affinst_on(target_cpu,
psci_entrypoint,
ns_entrypoint,
cluster_node->level,
plat_state);
}
return rc;
}
/*******************************************************************************
* Handler routine to turn a cluster of clusters on. It takes care or any
* generic, arch. or platform specific setup required.
* TODO: Split this code across separate handlers for each type of setup?
******************************************************************************/
static int psci_afflvl2_on(unsigned long target_cpu,
aff_map_node *system_node,
unsigned long ns_entrypoint,
unsigned long context_id)
{
int rc = PSCI_E_SUCCESS;
unsigned int plat_state;
unsigned long psci_entrypoint;
/* Cannot go beyond affinity level 2 in this psci imp. */
assert(system_node->level == MPIDR_AFFLVL2);
/*
* There is no generic and arch. specific system management
* required
*/
psci: fix affinity level upgrade issue The psci implementation does not track target affinity level requests specified during cpu_suspend calls correctly as per the following example. 1. cpu0.cluster0 calls cpu_suspend with the target affinity level as 0 2. Only the cpu0.cluster0 is powered down while cluster0 remains powered up 3. cpu1.cluster0 calls cpu_off to power itself down to highest possible affinity level 4. cluster0 will be powered off even though cpu0.cluster0 does not allow cluster shutdown This patch introduces reference counts at affinity levels > 0 to track the number of cpus which want an affinity instance at level X to remain powered up. This instance can be turned off only if its reference count is 0. Cpus still undergo the normal state transitions (ON, OFF, ON_PENDING, SUSPEND) but the higher levels can only be either ON or OFF depending upon their reference count. The above issue is thus fixed as follows: 1. cluster0's reference count is incremented by two when cpu0 and cpu1 are initially powered on. 2. cpu0.cluster0 calls cpu_suspend with the target affinity level as 0. This does not affect the cluster0 reference count. 3. Only the cpu0.cluster0 is powered down while cluster0 remains powered up as it has a non-zero reference count. 4. cpu1.cluster0 call cpu_off to power itself down to highest possible affinity level. This decrements the cluster0 reference count. 5. cluster0 is still not powered off since its reference count will at least be 1 due to the restriction placed by cpu0. Change-Id: I433dfe82b946f5f6985b1602c2de87800504f7a9
2013-12-05 16:33:10 +00:00
/* State management: Is not required while turning a system on */
2013-10-25 09:08:21 +01:00
/*
* Plat. management: Give the platform the current state
* of the target cpu to allow it to perform the necessary
* steps to power on.
*/
if (psci_plat_pm_ops->affinst_on) {
psci: fix affinity level upgrade issue The psci implementation does not track target affinity level requests specified during cpu_suspend calls correctly as per the following example. 1. cpu0.cluster0 calls cpu_suspend with the target affinity level as 0 2. Only the cpu0.cluster0 is powered down while cluster0 remains powered up 3. cpu1.cluster0 calls cpu_off to power itself down to highest possible affinity level 4. cluster0 will be powered off even though cpu0.cluster0 does not allow cluster shutdown This patch introduces reference counts at affinity levels > 0 to track the number of cpus which want an affinity instance at level X to remain powered up. This instance can be turned off only if its reference count is 0. Cpus still undergo the normal state transitions (ON, OFF, ON_PENDING, SUSPEND) but the higher levels can only be either ON or OFF depending upon their reference count. The above issue is thus fixed as follows: 1. cluster0's reference count is incremented by two when cpu0 and cpu1 are initially powered on. 2. cpu0.cluster0 calls cpu_suspend with the target affinity level as 0. This does not affect the cluster0 reference count. 3. Only the cpu0.cluster0 is powered down while cluster0 remains powered up as it has a non-zero reference count. 4. cpu1.cluster0 call cpu_off to power itself down to highest possible affinity level. This decrements the cluster0 reference count. 5. cluster0 is still not powered off since its reference count will at least be 1 due to the restriction placed by cpu0. Change-Id: I433dfe82b946f5f6985b1602c2de87800504f7a9
2013-12-05 16:33:10 +00:00
plat_state = psci_get_phys_state(system_node);
2013-10-25 09:08:21 +01:00
psci_entrypoint = (unsigned long) psci_aff_on_finish_entry;
rc = psci_plat_pm_ops->affinst_on(target_cpu,
psci_entrypoint,
ns_entrypoint,
system_node->level,
plat_state);
}
return rc;
}
/* Private data structure to make this handlers accessible through indexing */
static const afflvl_on_handler psci_afflvl_on_handlers[] = {
psci_afflvl0_on,
psci_afflvl1_on,
psci_afflvl2_on,
};
/*******************************************************************************
* This function takes an array of pointers to affinity instance nodes in the
* topology tree and calls the on handler for the corresponding affinity
* levels
******************************************************************************/
static int psci_call_on_handlers(mpidr_aff_map_nodes target_cpu_nodes,
int start_afflvl,
int end_afflvl,
unsigned long target_cpu,
unsigned long entrypoint,
unsigned long context_id)
{
int rc = PSCI_E_INVALID_PARAMS, level;
aff_map_node *node;
for (level = end_afflvl; level >= start_afflvl; level--) {
node = target_cpu_nodes[level];
if (node == NULL)
continue;
/*
* TODO: In case of an error should there be a way
* of undoing what we might have setup at higher
* affinity levels.
*/
rc = psci_afflvl_on_handlers[level](target_cpu,
node,
entrypoint,
context_id);
if (rc != PSCI_E_SUCCESS)
break;
}
return rc;
}
/*******************************************************************************
* Generic handler which is called to physically power on a cpu identified by
* its mpidr. It traverses through all the affinity levels performing generic,
* architectural, platform setup and state management e.g. for a cpu that is
* to be powered on, it will ensure that enough information is stashed for it
* to resume execution in the non-secure security state.
*
* The state of all the relevant affinity levels is changed after calling the
* affinity level specific handlers as their actions would depend upon the state
* the affinity level is currently in.
*
* The affinity level specific handlers are called in descending order i.e. from
* the highest to the lowest affinity level implemented by the platform because
* to turn on affinity level X it is neccesary to turn on affinity level X + 1
* first.
2013-10-25 09:08:21 +01:00
******************************************************************************/
int psci_afflvl_on(unsigned long target_cpu,
unsigned long entrypoint,
unsigned long context_id,
int start_afflvl,
int end_afflvl)
2013-10-25 09:08:21 +01:00
{
int rc = PSCI_E_SUCCESS;
mpidr_aff_map_nodes target_cpu_nodes;
2013-10-25 09:08:21 +01:00
unsigned long mpidr = read_mpidr() & MPIDR_AFFINITY_MASK;
/*
* Collect the pointers to the nodes in the topology tree for
* each affinity instance in the mpidr. If this function does
* not return successfully then either the mpidr or the affinity
* levels are incorrect.
2013-10-25 09:08:21 +01:00
*/
rc = psci_get_aff_map_nodes(target_cpu,
start_afflvl,
end_afflvl,
target_cpu_nodes);
if (rc != PSCI_E_SUCCESS)
return rc;
2013-10-25 09:08:21 +01:00
/*
* This function acquires the lock corresponding to each affinity
* level so that by the time all locks are taken, the system topology
* is snapshot and state management can be done safely.
2013-10-25 09:08:21 +01:00
*/
psci_acquire_afflvl_locks(mpidr,
start_afflvl,
end_afflvl,
target_cpu_nodes);
/* Perform generic, architecture and platform specific handling. */
rc = psci_call_on_handlers(target_cpu_nodes,
start_afflvl,
end_afflvl,
target_cpu,
entrypoint,
context_id);
2013-10-25 09:08:21 +01:00
/*
* This loop releases the lock corresponding to each affinity level
* in the reverse order to which they were acquired.
2013-10-25 09:08:21 +01:00
*/
psci_release_afflvl_locks(mpidr,
start_afflvl,
end_afflvl,
target_cpu_nodes);
2013-10-25 09:08:21 +01:00
return rc;
}
/*******************************************************************************
* The following functions finish an earlier affinity power on request. They
* are called by the common finisher routine in psci_common.c.
******************************************************************************/
static unsigned int psci_afflvl0_on_finish(unsigned long mpidr,
aff_map_node *cpu_node)
2013-10-25 09:08:21 +01:00
{
unsigned int index, plat_state, state, rc = PSCI_E_SUCCESS;
2013-10-25 09:08:21 +01:00
assert(cpu_node->level == MPIDR_AFFLVL0);
/* Ensure we have been explicitly woken up by another cpu */
psci: fix affinity level upgrade issue The psci implementation does not track target affinity level requests specified during cpu_suspend calls correctly as per the following example. 1. cpu0.cluster0 calls cpu_suspend with the target affinity level as 0 2. Only the cpu0.cluster0 is powered down while cluster0 remains powered up 3. cpu1.cluster0 calls cpu_off to power itself down to highest possible affinity level 4. cluster0 will be powered off even though cpu0.cluster0 does not allow cluster shutdown This patch introduces reference counts at affinity levels > 0 to track the number of cpus which want an affinity instance at level X to remain powered up. This instance can be turned off only if its reference count is 0. Cpus still undergo the normal state transitions (ON, OFF, ON_PENDING, SUSPEND) but the higher levels can only be either ON or OFF depending upon their reference count. The above issue is thus fixed as follows: 1. cluster0's reference count is incremented by two when cpu0 and cpu1 are initially powered on. 2. cpu0.cluster0 calls cpu_suspend with the target affinity level as 0. This does not affect the cluster0 reference count. 3. Only the cpu0.cluster0 is powered down while cluster0 remains powered up as it has a non-zero reference count. 4. cpu1.cluster0 call cpu_off to power itself down to highest possible affinity level. This decrements the cluster0 reference count. 5. cluster0 is still not powered off since its reference count will at least be 1 due to the restriction placed by cpu0. Change-Id: I433dfe82b946f5f6985b1602c2de87800504f7a9
2013-12-05 16:33:10 +00:00
state = psci_get_state(cpu_node);
assert(state == PSCI_STATE_ON_PENDING);
2013-10-25 09:08:21 +01:00
/*
* Plat. management: Perform the platform specific actions
* for this cpu e.g. enabling the gic or zeroing the mailbox
* register. The actual state of this cpu has already been
* changed.
*/
if (psci_plat_pm_ops->affinst_on_finish) {
/* Get the physical state of this cpu */
psci: fix affinity level upgrade issue The psci implementation does not track target affinity level requests specified during cpu_suspend calls correctly as per the following example. 1. cpu0.cluster0 calls cpu_suspend with the target affinity level as 0 2. Only the cpu0.cluster0 is powered down while cluster0 remains powered up 3. cpu1.cluster0 calls cpu_off to power itself down to highest possible affinity level 4. cluster0 will be powered off even though cpu0.cluster0 does not allow cluster shutdown This patch introduces reference counts at affinity levels > 0 to track the number of cpus which want an affinity instance at level X to remain powered up. This instance can be turned off only if its reference count is 0. Cpus still undergo the normal state transitions (ON, OFF, ON_PENDING, SUSPEND) but the higher levels can only be either ON or OFF depending upon their reference count. The above issue is thus fixed as follows: 1. cluster0's reference count is incremented by two when cpu0 and cpu1 are initially powered on. 2. cpu0.cluster0 calls cpu_suspend with the target affinity level as 0. This does not affect the cluster0 reference count. 3. Only the cpu0.cluster0 is powered down while cluster0 remains powered up as it has a non-zero reference count. 4. cpu1.cluster0 call cpu_off to power itself down to highest possible affinity level. This decrements the cluster0 reference count. 5. cluster0 is still not powered off since its reference count will at least be 1 due to the restriction placed by cpu0. Change-Id: I433dfe82b946f5f6985b1602c2de87800504f7a9
2013-12-05 16:33:10 +00:00
plat_state = get_phys_state(state);
2013-10-25 09:08:21 +01:00
rc = psci_plat_pm_ops->affinst_on_finish(mpidr,
cpu_node->level,
plat_state);
assert(rc == PSCI_E_SUCCESS);
}
/*
* Arch. management: Turn on mmu & restore architectural state
*/
enable_mmu();
/*
* All the platform specific actions for turning this cpu
* on have completed. Perform enough arch.initialization
* to run in the non-secure address space.
*/
bl31_arch_setup();
/*
* Use the more complex exception vectors to enable SPD
* initialisation. SP_EL3 should point to a 'cpu_context'
* structure which has an exception stack allocated. The
* calling cpu should have set the context already
*/
assert(cm_get_context(mpidr, NON_SECURE));
cm_set_next_eret_context(NON_SECURE);
write_vbar_el3((uint64_t) runtime_exceptions);
/*
* Call the cpu on finish handler registered by the Secure Payload
* Dispatcher to let it do any bookeeping. If the handler encounters an
* error, it's expected to assert within
*/
if (psci_spd_pm && psci_spd_pm->svc_on_finish)
psci_spd_pm->svc_on_finish(0);
2013-10-25 09:08:21 +01:00
/*
* 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. First get the index
2013-10-25 09:08:21 +01:00
* for restoring the re-entry info
*/
index = cpu_node->data;
psci_get_ns_entry_info(index);
2013-10-25 09:08:21 +01:00
psci: fix affinity level upgrade issue The psci implementation does not track target affinity level requests specified during cpu_suspend calls correctly as per the following example. 1. cpu0.cluster0 calls cpu_suspend with the target affinity level as 0 2. Only the cpu0.cluster0 is powered down while cluster0 remains powered up 3. cpu1.cluster0 calls cpu_off to power itself down to highest possible affinity level 4. cluster0 will be powered off even though cpu0.cluster0 does not allow cluster shutdown This patch introduces reference counts at affinity levels > 0 to track the number of cpus which want an affinity instance at level X to remain powered up. This instance can be turned off only if its reference count is 0. Cpus still undergo the normal state transitions (ON, OFF, ON_PENDING, SUSPEND) but the higher levels can only be either ON or OFF depending upon their reference count. The above issue is thus fixed as follows: 1. cluster0's reference count is incremented by two when cpu0 and cpu1 are initially powered on. 2. cpu0.cluster0 calls cpu_suspend with the target affinity level as 0. This does not affect the cluster0 reference count. 3. Only the cpu0.cluster0 is powered down while cluster0 remains powered up as it has a non-zero reference count. 4. cpu1.cluster0 call cpu_off to power itself down to highest possible affinity level. This decrements the cluster0 reference count. 5. cluster0 is still not powered off since its reference count will at least be 1 due to the restriction placed by cpu0. Change-Id: I433dfe82b946f5f6985b1602c2de87800504f7a9
2013-12-05 16:33:10 +00:00
/* State management: mark this cpu as on */
psci_set_state(cpu_node, PSCI_STATE_ON);
2013-10-25 09:08:21 +01:00
/* Clean caches before re-entering normal world */
dcsw_op_louis(DCCSW);
return rc;
}
static unsigned int psci_afflvl1_on_finish(unsigned long mpidr,
aff_map_node *cluster_node)
2013-10-25 09:08:21 +01:00
{
unsigned int plat_state, rc = PSCI_E_SUCCESS;
2013-10-25 09:08:21 +01:00
assert(cluster_node->level == MPIDR_AFFLVL1);
/*
* Plat. management: Perform the platform specific actions
* as per the old state of the cluster e.g. enabling
* coherency at the interconnect depends upon the state with
* which this cluster was powered up. If anything goes wrong
* then assert as there is no way to recover from this
* situation.
*/
if (psci_plat_pm_ops->affinst_on_finish) {
/* Get the physical state of this cluster */
psci: fix affinity level upgrade issue The psci implementation does not track target affinity level requests specified during cpu_suspend calls correctly as per the following example. 1. cpu0.cluster0 calls cpu_suspend with the target affinity level as 0 2. Only the cpu0.cluster0 is powered down while cluster0 remains powered up 3. cpu1.cluster0 calls cpu_off to power itself down to highest possible affinity level 4. cluster0 will be powered off even though cpu0.cluster0 does not allow cluster shutdown This patch introduces reference counts at affinity levels > 0 to track the number of cpus which want an affinity instance at level X to remain powered up. This instance can be turned off only if its reference count is 0. Cpus still undergo the normal state transitions (ON, OFF, ON_PENDING, SUSPEND) but the higher levels can only be either ON or OFF depending upon their reference count. The above issue is thus fixed as follows: 1. cluster0's reference count is incremented by two when cpu0 and cpu1 are initially powered on. 2. cpu0.cluster0 calls cpu_suspend with the target affinity level as 0. This does not affect the cluster0 reference count. 3. Only the cpu0.cluster0 is powered down while cluster0 remains powered up as it has a non-zero reference count. 4. cpu1.cluster0 call cpu_off to power itself down to highest possible affinity level. This decrements the cluster0 reference count. 5. cluster0 is still not powered off since its reference count will at least be 1 due to the restriction placed by cpu0. Change-Id: I433dfe82b946f5f6985b1602c2de87800504f7a9
2013-12-05 16:33:10 +00:00
plat_state = psci_get_phys_state(cluster_node);
2013-10-25 09:08:21 +01:00
rc = psci_plat_pm_ops->affinst_on_finish(mpidr,
cluster_node->level,
plat_state);
assert(rc == PSCI_E_SUCCESS);
}
psci: fix affinity level upgrade issue The psci implementation does not track target affinity level requests specified during cpu_suspend calls correctly as per the following example. 1. cpu0.cluster0 calls cpu_suspend with the target affinity level as 0 2. Only the cpu0.cluster0 is powered down while cluster0 remains powered up 3. cpu1.cluster0 calls cpu_off to power itself down to highest possible affinity level 4. cluster0 will be powered off even though cpu0.cluster0 does not allow cluster shutdown This patch introduces reference counts at affinity levels > 0 to track the number of cpus which want an affinity instance at level X to remain powered up. This instance can be turned off only if its reference count is 0. Cpus still undergo the normal state transitions (ON, OFF, ON_PENDING, SUSPEND) but the higher levels can only be either ON or OFF depending upon their reference count. The above issue is thus fixed as follows: 1. cluster0's reference count is incremented by two when cpu0 and cpu1 are initially powered on. 2. cpu0.cluster0 calls cpu_suspend with the target affinity level as 0. This does not affect the cluster0 reference count. 3. Only the cpu0.cluster0 is powered down while cluster0 remains powered up as it has a non-zero reference count. 4. cpu1.cluster0 call cpu_off to power itself down to highest possible affinity level. This decrements the cluster0 reference count. 5. cluster0 is still not powered off since its reference count will at least be 1 due to the restriction placed by cpu0. Change-Id: I433dfe82b946f5f6985b1602c2de87800504f7a9
2013-12-05 16:33:10 +00:00
/* State management: Increment the cluster reference count */
psci_set_state(cluster_node, PSCI_STATE_ON);
2013-10-25 09:08:21 +01:00
return rc;
}
static unsigned int psci_afflvl2_on_finish(unsigned long mpidr,
aff_map_node *system_node)
2013-10-25 09:08:21 +01:00
{
unsigned int plat_state, rc = PSCI_E_SUCCESS;
2013-10-25 09:08:21 +01:00
/* Cannot go beyond this affinity level */
assert(system_node->level == MPIDR_AFFLVL2);
/*
* Currently, there are no architectural actions to perform
* at the system level.
*/
/*
* Plat. management: Perform the platform specific actions
* as per the old state of the cluster e.g. enabling
* coherency at the interconnect depends upon the state with
* which this cluster was powered up. If anything goes wrong
* then assert as there is no way to recover from this
* situation.
*/
if (psci_plat_pm_ops->affinst_on_finish) {
/* Get the physical state of the system */
psci: fix affinity level upgrade issue The psci implementation does not track target affinity level requests specified during cpu_suspend calls correctly as per the following example. 1. cpu0.cluster0 calls cpu_suspend with the target affinity level as 0 2. Only the cpu0.cluster0 is powered down while cluster0 remains powered up 3. cpu1.cluster0 calls cpu_off to power itself down to highest possible affinity level 4. cluster0 will be powered off even though cpu0.cluster0 does not allow cluster shutdown This patch introduces reference counts at affinity levels > 0 to track the number of cpus which want an affinity instance at level X to remain powered up. This instance can be turned off only if its reference count is 0. Cpus still undergo the normal state transitions (ON, OFF, ON_PENDING, SUSPEND) but the higher levels can only be either ON or OFF depending upon their reference count. The above issue is thus fixed as follows: 1. cluster0's reference count is incremented by two when cpu0 and cpu1 are initially powered on. 2. cpu0.cluster0 calls cpu_suspend with the target affinity level as 0. This does not affect the cluster0 reference count. 3. Only the cpu0.cluster0 is powered down while cluster0 remains powered up as it has a non-zero reference count. 4. cpu1.cluster0 call cpu_off to power itself down to highest possible affinity level. This decrements the cluster0 reference count. 5. cluster0 is still not powered off since its reference count will at least be 1 due to the restriction placed by cpu0. Change-Id: I433dfe82b946f5f6985b1602c2de87800504f7a9
2013-12-05 16:33:10 +00:00
plat_state = psci_get_phys_state(system_node);
2013-10-25 09:08:21 +01:00
rc = psci_plat_pm_ops->affinst_on_finish(mpidr,
system_node->level,
plat_state);
assert(rc == PSCI_E_SUCCESS);
}
psci: fix affinity level upgrade issue The psci implementation does not track target affinity level requests specified during cpu_suspend calls correctly as per the following example. 1. cpu0.cluster0 calls cpu_suspend with the target affinity level as 0 2. Only the cpu0.cluster0 is powered down while cluster0 remains powered up 3. cpu1.cluster0 calls cpu_off to power itself down to highest possible affinity level 4. cluster0 will be powered off even though cpu0.cluster0 does not allow cluster shutdown This patch introduces reference counts at affinity levels > 0 to track the number of cpus which want an affinity instance at level X to remain powered up. This instance can be turned off only if its reference count is 0. Cpus still undergo the normal state transitions (ON, OFF, ON_PENDING, SUSPEND) but the higher levels can only be either ON or OFF depending upon their reference count. The above issue is thus fixed as follows: 1. cluster0's reference count is incremented by two when cpu0 and cpu1 are initially powered on. 2. cpu0.cluster0 calls cpu_suspend with the target affinity level as 0. This does not affect the cluster0 reference count. 3. Only the cpu0.cluster0 is powered down while cluster0 remains powered up as it has a non-zero reference count. 4. cpu1.cluster0 call cpu_off to power itself down to highest possible affinity level. This decrements the cluster0 reference count. 5. cluster0 is still not powered off since its reference count will at least be 1 due to the restriction placed by cpu0. Change-Id: I433dfe82b946f5f6985b1602c2de87800504f7a9
2013-12-05 16:33:10 +00:00
/* State management: Increment the system reference count */
psci_set_state(system_node, PSCI_STATE_ON);
2013-10-25 09:08:21 +01:00
return rc;
}
const afflvl_power_on_finisher psci_afflvl_on_finishers[] = {
psci_afflvl0_on_finish,
psci_afflvl1_on_finish,
psci_afflvl2_on_finish,
};