You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.

820 lines
25 KiB

/*
* Copyright (c) 2013-2022, ARM Limited and Contributors. All rights reserved.
*
* SPDX-License-Identifier: BSD-3-Clause
*/
/*******************************************************************************
* This is the Secure Payload Dispatcher (SPD). The dispatcher is meant to be a
* plug-in component to the Secure Monitor, registered as a runtime service. The
* SPD is expected to be a functional extension of the Secure Payload (SP) that
* executes in Secure EL1. The Secure Monitor will delegate all SMCs targeting
* the Trusted OS/Applications range to the dispatcher. The SPD will either
* handle the request locally or delegate it to the Secure Payload. It is also
* responsible for initialising and maintaining communication with the SP.
******************************************************************************/
#include <assert.h>
#include <errno.h>
#include <stddef.h>
#include <string.h>
#include <arch_helpers.h>
#include <bl31/bl31.h>
#include <bl31/ehf.h>
#include <bl32/tsp/tsp.h>
#include <common/bl_common.h>
#include <common/debug.h>
#include <common/runtime_svc.h>
#include <lib/el3_runtime/context_mgmt.h>
#include <plat/common/platform.h>
#include <tools_share/uuid.h>
#include "tspd_private.h"
/*******************************************************************************
* Address of the entrypoint vector table in the Secure Payload. It is
* initialised once on the primary core after a cold boot.
******************************************************************************/
tsp_vectors_t *tsp_vectors;
/*******************************************************************************
* Array to keep track of per-cpu Secure Payload state
******************************************************************************/
tsp_context_t tspd_sp_context[TSPD_CORE_COUNT];
/* TSP UID */
DEFINE_SVC_UUID2(tsp_uuid,
0xa056305b, 0x9132, 0x7b42, 0x98, 0x11,
0x71, 0x68, 0xca, 0x50, 0xf3, 0xfa);
int32_t tspd_init(void);
Unify interrupt return paths from TSP into the TSPD The TSP is expected to pass control back to EL3 if it gets preempted due to an interrupt while handling a Standard SMC in the following scenarios: 1. An FIQ preempts Standard SMC execution and that FIQ is not a TSP Secure timer interrupt or is preempted by a higher priority interrupt by the time the TSP acknowledges it. In this case, the TSP issues an SMC with the ID as `TSP_EL3_FIQ`. Currently this case is never expected to happen as only the TSP Secure Timer is expected to generate FIQ. 2. An IRQ preempts Standard SMC execution and in this case the TSP issues an SMC with the ID as `TSP_PREEMPTED`. In both the cases, the TSPD hands control back to the normal world and returns returns an error code to the normal world to indicate that the standard SMC it had issued has been preempted but not completed. This patch unifies the handling of these two cases in the TSPD and ensures that the TSP only uses TSP_PREEMPTED instead of separate SMC IDs. Also instead of 2 separate error codes, SMC_PREEMPTED and TSP_EL3_FIQ, only SMC_PREEMPTED is returned as error code back to the normal world. Background information: On a GICv3 system, when the secure world has affinity routing enabled, in 2. an FIQ will preempt TSP execution instead of an IRQ. The FIQ could be a result of a Group 0 or a Group 1 NS interrupt. In both case, the TSPD passes control back to the normal world upon receipt of the TSP_PREEMPTED SMC. A Group 0 interrupt will immediately preempt execution to EL3 where it will be handled. This allows for unified interrupt handling in TSP for both GICv3 and GICv2 systems. Change-Id: I9895344db74b188021e3f6a694701ad272fb40d4
9 years ago
/*
* This helper function handles Secure EL1 preemption. The preemption could be
* due Non Secure interrupts or EL3 interrupts. In both the cases we context
* switch to the normal world and in case of EL3 interrupts, it will again be
* routed to EL3 which will get handled at the exception vectors.
*/
uint64_t tspd_handle_sp_preemption(void *handle)
{
cpu_context_t *ns_cpu_context;
Unify interrupt return paths from TSP into the TSPD The TSP is expected to pass control back to EL3 if it gets preempted due to an interrupt while handling a Standard SMC in the following scenarios: 1. An FIQ preempts Standard SMC execution and that FIQ is not a TSP Secure timer interrupt or is preempted by a higher priority interrupt by the time the TSP acknowledges it. In this case, the TSP issues an SMC with the ID as `TSP_EL3_FIQ`. Currently this case is never expected to happen as only the TSP Secure Timer is expected to generate FIQ. 2. An IRQ preempts Standard SMC execution and in this case the TSP issues an SMC with the ID as `TSP_PREEMPTED`. In both the cases, the TSPD hands control back to the normal world and returns returns an error code to the normal world to indicate that the standard SMC it had issued has been preempted but not completed. This patch unifies the handling of these two cases in the TSPD and ensures that the TSP only uses TSP_PREEMPTED instead of separate SMC IDs. Also instead of 2 separate error codes, SMC_PREEMPTED and TSP_EL3_FIQ, only SMC_PREEMPTED is returned as error code back to the normal world. Background information: On a GICv3 system, when the secure world has affinity routing enabled, in 2. an FIQ will preempt TSP execution instead of an IRQ. The FIQ could be a result of a Group 0 or a Group 1 NS interrupt. In both case, the TSPD passes control back to the normal world upon receipt of the TSP_PREEMPTED SMC. A Group 0 interrupt will immediately preempt execution to EL3 where it will be handled. This allows for unified interrupt handling in TSP for both GICv3 and GICv2 systems. Change-Id: I9895344db74b188021e3f6a694701ad272fb40d4
9 years ago
assert(handle == cm_get_context(SECURE));
cm_el1_sysregs_context_save(SECURE);
/* Get a reference to the non-secure context */
ns_cpu_context = cm_get_context(NON_SECURE);
assert(ns_cpu_context);
/*
* To allow Secure EL1 interrupt handler to re-enter TSP while TSP
* is preempted, the secure system register context which will get
* overwritten must be additionally saved. This is currently done
* by the TSPD S-EL1 interrupt handler.
*/
/*
* Restore non-secure state.
*/
cm_el1_sysregs_context_restore(NON_SECURE);
cm_set_next_eret_context(NON_SECURE);
Unify interrupt return paths from TSP into the TSPD The TSP is expected to pass control back to EL3 if it gets preempted due to an interrupt while handling a Standard SMC in the following scenarios: 1. An FIQ preempts Standard SMC execution and that FIQ is not a TSP Secure timer interrupt or is preempted by a higher priority interrupt by the time the TSP acknowledges it. In this case, the TSP issues an SMC with the ID as `TSP_EL3_FIQ`. Currently this case is never expected to happen as only the TSP Secure Timer is expected to generate FIQ. 2. An IRQ preempts Standard SMC execution and in this case the TSP issues an SMC with the ID as `TSP_PREEMPTED`. In both the cases, the TSPD hands control back to the normal world and returns returns an error code to the normal world to indicate that the standard SMC it had issued has been preempted but not completed. This patch unifies the handling of these two cases in the TSPD and ensures that the TSP only uses TSP_PREEMPTED instead of separate SMC IDs. Also instead of 2 separate error codes, SMC_PREEMPTED and TSP_EL3_FIQ, only SMC_PREEMPTED is returned as error code back to the normal world. Background information: On a GICv3 system, when the secure world has affinity routing enabled, in 2. an FIQ will preempt TSP execution instead of an IRQ. The FIQ could be a result of a Group 0 or a Group 1 NS interrupt. In both case, the TSPD passes control back to the normal world upon receipt of the TSP_PREEMPTED SMC. A Group 0 interrupt will immediately preempt execution to EL3 where it will be handled. This allows for unified interrupt handling in TSP for both GICv3 and GICv2 systems. Change-Id: I9895344db74b188021e3f6a694701ad272fb40d4
9 years ago
/*
* The TSP was preempted during execution of a Yielding SMC Call.
* Return back to the normal world with SMC_PREEMPTED as error
* code in x0.
Unify interrupt return paths from TSP into the TSPD The TSP is expected to pass control back to EL3 if it gets preempted due to an interrupt while handling a Standard SMC in the following scenarios: 1. An FIQ preempts Standard SMC execution and that FIQ is not a TSP Secure timer interrupt or is preempted by a higher priority interrupt by the time the TSP acknowledges it. In this case, the TSP issues an SMC with the ID as `TSP_EL3_FIQ`. Currently this case is never expected to happen as only the TSP Secure Timer is expected to generate FIQ. 2. An IRQ preempts Standard SMC execution and in this case the TSP issues an SMC with the ID as `TSP_PREEMPTED`. In both the cases, the TSPD hands control back to the normal world and returns returns an error code to the normal world to indicate that the standard SMC it had issued has been preempted but not completed. This patch unifies the handling of these two cases in the TSPD and ensures that the TSP only uses TSP_PREEMPTED instead of separate SMC IDs. Also instead of 2 separate error codes, SMC_PREEMPTED and TSP_EL3_FIQ, only SMC_PREEMPTED is returned as error code back to the normal world. Background information: On a GICv3 system, when the secure world has affinity routing enabled, in 2. an FIQ will preempt TSP execution instead of an IRQ. The FIQ could be a result of a Group 0 or a Group 1 NS interrupt. In both case, the TSPD passes control back to the normal world upon receipt of the TSP_PREEMPTED SMC. A Group 0 interrupt will immediately preempt execution to EL3 where it will be handled. This allows for unified interrupt handling in TSP for both GICv3 and GICv2 systems. Change-Id: I9895344db74b188021e3f6a694701ad272fb40d4
9 years ago
*/
SMC_RET1(ns_cpu_context, SMC_PREEMPTED);
}
Unify interrupt return paths from TSP into the TSPD The TSP is expected to pass control back to EL3 if it gets preempted due to an interrupt while handling a Standard SMC in the following scenarios: 1. An FIQ preempts Standard SMC execution and that FIQ is not a TSP Secure timer interrupt or is preempted by a higher priority interrupt by the time the TSP acknowledges it. In this case, the TSP issues an SMC with the ID as `TSP_EL3_FIQ`. Currently this case is never expected to happen as only the TSP Secure Timer is expected to generate FIQ. 2. An IRQ preempts Standard SMC execution and in this case the TSP issues an SMC with the ID as `TSP_PREEMPTED`. In both the cases, the TSPD hands control back to the normal world and returns returns an error code to the normal world to indicate that the standard SMC it had issued has been preempted but not completed. This patch unifies the handling of these two cases in the TSPD and ensures that the TSP only uses TSP_PREEMPTED instead of separate SMC IDs. Also instead of 2 separate error codes, SMC_PREEMPTED and TSP_EL3_FIQ, only SMC_PREEMPTED is returned as error code back to the normal world. Background information: On a GICv3 system, when the secure world has affinity routing enabled, in 2. an FIQ will preempt TSP execution instead of an IRQ. The FIQ could be a result of a Group 0 or a Group 1 NS interrupt. In both case, the TSPD passes control back to the normal world upon receipt of the TSP_PREEMPTED SMC. A Group 0 interrupt will immediately preempt execution to EL3 where it will be handled. This allows for unified interrupt handling in TSP for both GICv3 and GICv2 systems. Change-Id: I9895344db74b188021e3f6a694701ad272fb40d4
9 years ago
/*******************************************************************************
* This function is the handler registered for S-EL1 interrupts by the TSPD. It
* validates the interrupt and upon success arranges entry into the TSP at
* 'tsp_sel1_intr_entry()' for handling the interrupt.
Bug fix in tspd interrupt handling when TSP_NS_INTR_ASYNC_PREEMPT is enabled Typically, interrupts for a specific security state get handled in the same security execption level if the execution is in the same security state. For example, if a non-secure interrupt gets fired when CPU is executing in NS-EL2 it gets handled in the non-secure world. However, interrupts belonging to the opposite security state typically demand a world(context) switch. This is inline with the security principle which states a secure interrupt has to be handled in the secure world. Hence, the TSPD in EL3 expects the context(handle) for a secure interrupt to be non-secure and vice versa. The function &#34;tspd_sel1_interrupt_handler&#34; is the handler registered for S-EL1 interrupts by the TSPD. Based on the above assumption, it provides an assertion to validate if the interrupt originated from non-secure world and upon success arranges entry into the TSP at &#39;tsp_sel1_intr_entry&#39; for handling the interrupt. However, a race condition between non-secure and secure interrupts can lead to a scenario where the above assumptions do not hold true and further leading to following assert fail. This patch fixes the bug which causes this assert fail: ASSERT: services/spd/tspd/tspd_main.c:105 BACKTRACE: START: assert 0: EL3: 0x400c128 1: EL3: 0x400faf8 2: EL3: 0x40099a4 3: EL3: 0x4010d54 BACKTRACE: END: assert Change-Id: I359d30fb5dbb1429a4a3c3fff37fdc64c07e9414 Signed-off-by: Madhukar Pappireddy &lt;madhukar.pappireddy@arm.com&gt;
4 years ago
* Typically, interrupts for a specific security state get handled in the same
* security execption level if the execution is in the same security state. For
* example, if a non-secure interrupt gets fired when CPU is executing in NS-EL2
* it gets handled in the non-secure world.
* However, interrupts belonging to the opposite security state typically demand
* a world(context) switch. This is inline with the security principle which
* states a secure interrupt has to be handled in the secure world.
* Hence, the TSPD in EL3 expects the context(handle) for a secure interrupt to
* be non-secure and vice versa.
* However, a race condition between non-secure and secure interrupts can lead to
* a scenario where the above assumptions do not hold true. This is demonstrated
* below through Note 1.
******************************************************************************/
static uint64_t tspd_sel1_interrupt_handler(uint32_t id,
uint32_t flags,
void *handle,
void *cookie)
{
uint32_t linear_id;
tsp_context_t *tsp_ctx;
Bug fix in tspd interrupt handling when TSP_NS_INTR_ASYNC_PREEMPT is enabled Typically, interrupts for a specific security state get handled in the same security execption level if the execution is in the same security state. For example, if a non-secure interrupt gets fired when CPU is executing in NS-EL2 it gets handled in the non-secure world. However, interrupts belonging to the opposite security state typically demand a world(context) switch. This is inline with the security principle which states a secure interrupt has to be handled in the secure world. Hence, the TSPD in EL3 expects the context(handle) for a secure interrupt to be non-secure and vice versa. The function &#34;tspd_sel1_interrupt_handler&#34; is the handler registered for S-EL1 interrupts by the TSPD. Based on the above assumption, it provides an assertion to validate if the interrupt originated from non-secure world and upon success arranges entry into the TSP at &#39;tsp_sel1_intr_entry&#39; for handling the interrupt. However, a race condition between non-secure and secure interrupts can lead to a scenario where the above assumptions do not hold true and further leading to following assert fail. This patch fixes the bug which causes this assert fail: ASSERT: services/spd/tspd/tspd_main.c:105 BACKTRACE: START: assert 0: EL3: 0x400c128 1: EL3: 0x400faf8 2: EL3: 0x40099a4 3: EL3: 0x4010d54 BACKTRACE: END: assert Change-Id: I359d30fb5dbb1429a4a3c3fff37fdc64c07e9414 Signed-off-by: Madhukar Pappireddy &lt;madhukar.pappireddy@arm.com&gt;
4 years ago
/* Get a reference to this cpu's TSP context */
linear_id = plat_my_core_pos();
tsp_ctx = &tspd_sp_context[linear_id];
#if TSP_NS_INTR_ASYNC_PREEMPT
/*
* Note 1:
* Under the current interrupt routing model, interrupts from other
* world are routed to EL3 when TSP_NS_INTR_ASYNC_PREEMPT is enabled.
* Consider the following scenario:
* 1/ A non-secure payload(like tftf) requests a secure service from
* TSP by invoking a yielding SMC call.
* 2/ Later, execution jumps to TSP in S-EL1 with the help of TSP
* Dispatcher in Secure Monitor(EL3).
* 3/ While CPU is executing TSP, a Non-secure interrupt gets fired.
* this demands a context switch to the non-secure world through
* secure monitor.
* 4/ Consequently, TSP in S-EL1 get asynchronously pre-empted and
* execution switches to secure monitor(EL3).
* 5/ EL3 tries to triage the (Non-secure) interrupt based on the
* highest pending interrupt.
* 6/ However, while the NS Interrupt was pending, secure timer gets
* fired which makes a S-EL1 interrupt to be pending.
* 7/ Hence, execution jumps to this companion handler of S-EL1
* interrupt (i.e., tspd_sel1_interrupt_handler) even though the TSP
* was pre-empted due to non-secure interrupt.
* 8/ The above sequence of events explain how TSP was pre-empted by
* S-EL1 interrupt indirectly in an asynchronous way.
* 9/ Hence, we track the TSP pre-emption by S-EL1 interrupt using a
* boolean variable per each core.
* 10/ This helps us to indicate that SMC call for TSP service was
* pre-empted when execution resumes in non-secure world.
*/
/* Check the security state when the exception was generated */
if (get_interrupt_src_ss(flags) == NON_SECURE) {
/* Sanity check the pointer to this cpu's context */
assert(handle == cm_get_context(NON_SECURE));
/* Save the non-secure context before entering the TSP */
cm_el1_sysregs_context_save(NON_SECURE);
tsp_ctx->preempted_by_sel1_intr = false;
} else {
/* Sanity check the pointer to this cpu's context */
assert(handle == cm_get_context(SECURE));
/* Save the secure context before entering the TSP for S-EL1
* interrupt handling
*/
cm_el1_sysregs_context_save(SECURE);
tsp_ctx->preempted_by_sel1_intr = true;
}
#else
/* Check the security state when the exception was generated */
assert(get_interrupt_src_ss(flags) == NON_SECURE);
/* Sanity check the pointer to this cpu's context */
assert(handle == cm_get_context(NON_SECURE));
/* Save the non-secure context before entering the TSP */
cm_el1_sysregs_context_save(NON_SECURE);
Bug fix in tspd interrupt handling when TSP_NS_INTR_ASYNC_PREEMPT is enabled Typically, interrupts for a specific security state get handled in the same security execption level if the execution is in the same security state. For example, if a non-secure interrupt gets fired when CPU is executing in NS-EL2 it gets handled in the non-secure world. However, interrupts belonging to the opposite security state typically demand a world(context) switch. This is inline with the security principle which states a secure interrupt has to be handled in the secure world. Hence, the TSPD in EL3 expects the context(handle) for a secure interrupt to be non-secure and vice versa. The function &#34;tspd_sel1_interrupt_handler&#34; is the handler registered for S-EL1 interrupts by the TSPD. Based on the above assumption, it provides an assertion to validate if the interrupt originated from non-secure world and upon success arranges entry into the TSP at &#39;tsp_sel1_intr_entry&#39; for handling the interrupt. However, a race condition between non-secure and secure interrupts can lead to a scenario where the above assumptions do not hold true and further leading to following assert fail. This patch fixes the bug which causes this assert fail: ASSERT: services/spd/tspd/tspd_main.c:105 BACKTRACE: START: assert 0: EL3: 0x400c128 1: EL3: 0x400faf8 2: EL3: 0x40099a4 3: EL3: 0x4010d54 BACKTRACE: END: assert Change-Id: I359d30fb5dbb1429a4a3c3fff37fdc64c07e9414 Signed-off-by: Madhukar Pappireddy &lt;madhukar.pappireddy@arm.com&gt;
4 years ago
#endif
assert(&tsp_ctx->cpu_ctx == cm_get_context(SECURE));
/*
* Determine if the TSP was previously preempted. Its last known
* context has to be preserved in this case.
* The TSP should return control to the TSPD after handling this
* S-EL1 interrupt. Preserve essential EL3 context to allow entry into
* the TSP at the S-EL1 interrupt entry point using the 'cpu_context'
* structure. There is no need to save the secure system register
* context since the TSP is supposed to preserve it during S-EL1
* interrupt handling.
*/
if (get_yield_smc_active_flag(tsp_ctx->state)) {
tsp_ctx->saved_spsr_el3 = (uint32_t)SMC_GET_EL3(&tsp_ctx->cpu_ctx,
CTX_SPSR_EL3);
tsp_ctx->saved_elr_el3 = SMC_GET_EL3(&tsp_ctx->cpu_ctx,
CTX_ELR_EL3);
#if TSP_NS_INTR_ASYNC_PREEMPT
memcpy(&tsp_ctx->sp_ctx, &tsp_ctx->cpu_ctx, TSPD_SP_CTX_SIZE);
#endif
}
cm_el1_sysregs_context_restore(SECURE);
cm_set_elr_spsr_el3(SECURE, (uint64_t) &tsp_vectors->sel1_intr_entry,
SPSR_64(MODE_EL1, MODE_SP_ELX, DISABLE_ALL_EXCEPTIONS));
cm_set_next_eret_context(SECURE);
/*
* Tell the TSP that it has to handle a S-EL1 interrupt synchronously.
* Also the instruction in normal world where the interrupt was
* generated is passed for debugging purposes. It is safe to retrieve
* this address from ELR_EL3 as the secure context will not take effect
* until el3_exit().
*/
SMC_RET2(&tsp_ctx->cpu_ctx, TSP_HANDLE_SEL1_INTR_AND_RETURN, read_elr_el3());
}
#if TSP_NS_INTR_ASYNC_PREEMPT
/*******************************************************************************
* This function is the handler registered for Non secure interrupts by the
* TSPD. It validates the interrupt and upon success arranges entry into the
* normal world for handling the interrupt.
******************************************************************************/
static uint64_t tspd_ns_interrupt_handler(uint32_t id,
uint32_t flags,
void *handle,
void *cookie)
{
/* Check the security state when the exception was generated */
assert(get_interrupt_src_ss(flags) == SECURE);
/*
* Disable the routing of NS interrupts from secure world to EL3 while
* interrupted on this core.
*/
disable_intr_rm_local(INTR_TYPE_NS, SECURE);
return tspd_handle_sp_preemption(handle);
}
#endif
/*******************************************************************************
* Secure Payload Dispatcher setup. The SPD finds out the SP entrypoint and type
* (aarch32/aarch64) if not already known and initialises the context for entry
* into the SP for its initialisation.
******************************************************************************/
static int32_t tspd_setup(void)
{
entry_point_info_t *tsp_ep_info;
uint32_t linear_id;
linear_id = plat_my_core_pos();
/*
* Get information about the Secure Payload (BL32) image. Its
* absence is a critical failure. TODO: Add support to
* conditionally include the SPD service
*/
tsp_ep_info = bl31_plat_get_next_image_ep_info(SECURE);
if (!tsp_ep_info) {
WARN("No TSP provided by BL2 boot loader, Booting device"
" without TSP initialization. SMC`s destined for TSP"
" will return SMC_UNK\n");
return 1;
}
/*
* If there's no valid entry point for SP, we return a non-zero value
* signalling failure initializing the service. We bail out without
* registering any handlers
*/
if (!tsp_ep_info->pc)
return 1;
/*
* We could inspect the SP image and determine its execution
* state i.e whether AArch32 or AArch64. Assuming it's AArch64
* for the time being.
*/
tspd_init_tsp_ep_state(tsp_ep_info,
TSP_AARCH64,
tsp_ep_info->pc,
&tspd_sp_context[linear_id]);
#if TSP_INIT_ASYNC
bl31_set_next_image_type(SECURE);
#else
/*
* All TSPD initialization done. Now register our init function with
* BL31 for deferred invocation
*/
bl31_register_bl32_init(&tspd_init);
#endif
return 0;
}
/*******************************************************************************
* This function passes control to the Secure Payload image (BL32) for the first
* time on the primary cpu after a cold boot. It assumes that a valid secure
* context has already been created by tspd_setup() which can be directly used.
* It also assumes that a valid non-secure context has been initialised by PSCI
* so it does not need to save and restore any non-secure state. This function
* performs a synchronous entry into the Secure payload. The SP passes control
* back to this routine through a SMC.
******************************************************************************/
int32_t tspd_init(void)
{
uint32_t linear_id = plat_my_core_pos();
tsp_context_t *tsp_ctx = &tspd_sp_context[linear_id];
entry_point_info_t *tsp_entry_point;
uint64_t rc;
/*
* Get information about the Secure Payload (BL32) image. Its
* absence is a critical failure.
*/
tsp_entry_point = bl31_plat_get_next_image_ep_info(SECURE);
assert(tsp_entry_point);
cm_init_my_context(tsp_entry_point);
/*
* Arrange for an entry into the test secure payload. It will be
* returned via TSP_ENTRY_DONE case
*/
rc = tspd_synchronous_sp_entry(tsp_ctx);
assert(rc != 0);
return rc;
}
/*******************************************************************************
* This function is responsible for handling all SMCs in the Trusted OS/App
* range from the non-secure state as defined in the SMC Calling Convention
* Document. It is also responsible for communicating with the Secure payload
* to delegate work and return results back to the non-secure state. Lastly it
* will also return any information that the secure payload needs to do the
* work assigned to it.
******************************************************************************/
static uintptr_t tspd_smc_handler(uint32_t smc_fid,
u_register_t x1,
u_register_t x2,
u_register_t x3,
u_register_t x4,
void *cookie,
void *handle,
u_register_t flags)
{
cpu_context_t *ns_cpu_context;
uint32_t linear_id = plat_my_core_pos(), ns;
tsp_context_t *tsp_ctx = &tspd_sp_context[linear_id];
uint64_t rc;
#if TSP_INIT_ASYNC
entry_point_info_t *next_image_info;
#endif
/* Determine which security state this SMC originated from */
ns = is_caller_non_secure(flags);
switch (smc_fid) {
/*
* This function ID is used by TSP to indicate that it was
* preempted by a normal world IRQ.
*
*/
case TSP_PREEMPTED:
if (ns)
SMC_RET1(handle, SMC_UNK);
return tspd_handle_sp_preemption(handle);
/*
* This function ID is used only by the TSP to indicate that it has
* finished handling a S-EL1 interrupt or was preempted by a higher
* priority pending EL3 interrupt. Execution should resume
* in the normal world.
*/
case TSP_HANDLED_S_EL1_INTR:
if (ns)
SMC_RET1(handle, SMC_UNK);
assert(handle == cm_get_context(SECURE));
/*
* Restore the relevant EL3 state which saved to service
* this SMC.
*/
if (get_yield_smc_active_flag(tsp_ctx->state)) {
SMC_SET_EL3(&tsp_ctx->cpu_ctx,
CTX_SPSR_EL3,
tsp_ctx->saved_spsr_el3);
SMC_SET_EL3(&tsp_ctx->cpu_ctx,
CTX_ELR_EL3,
tsp_ctx->saved_elr_el3);
#if TSP_NS_INTR_ASYNC_PREEMPT
/*
* Need to restore the previously interrupted
* secure context.
*/
memcpy(&tsp_ctx->cpu_ctx, &tsp_ctx->sp_ctx,
TSPD_SP_CTX_SIZE);
#endif
}
/* Get a reference to the non-secure context */
ns_cpu_context = cm_get_context(NON_SECURE);
assert(ns_cpu_context);
/*
* Restore non-secure state. There is no need to save the
* secure system register context since the TSP was supposed
* to preserve it during S-EL1 interrupt handling.
*/
cm_el1_sysregs_context_restore(NON_SECURE);
cm_set_next_eret_context(NON_SECURE);
Bug fix in tspd interrupt handling when TSP_NS_INTR_ASYNC_PREEMPT is enabled Typically, interrupts for a specific security state get handled in the same security execption level if the execution is in the same security state. For example, if a non-secure interrupt gets fired when CPU is executing in NS-EL2 it gets handled in the non-secure world. However, interrupts belonging to the opposite security state typically demand a world(context) switch. This is inline with the security principle which states a secure interrupt has to be handled in the secure world. Hence, the TSPD in EL3 expects the context(handle) for a secure interrupt to be non-secure and vice versa. The function &#34;tspd_sel1_interrupt_handler&#34; is the handler registered for S-EL1 interrupts by the TSPD. Based on the above assumption, it provides an assertion to validate if the interrupt originated from non-secure world and upon success arranges entry into the TSP at &#39;tsp_sel1_intr_entry&#39; for handling the interrupt. However, a race condition between non-secure and secure interrupts can lead to a scenario where the above assumptions do not hold true and further leading to following assert fail. This patch fixes the bug which causes this assert fail: ASSERT: services/spd/tspd/tspd_main.c:105 BACKTRACE: START: assert 0: EL3: 0x400c128 1: EL3: 0x400faf8 2: EL3: 0x40099a4 3: EL3: 0x4010d54 BACKTRACE: END: assert Change-Id: I359d30fb5dbb1429a4a3c3fff37fdc64c07e9414 Signed-off-by: Madhukar Pappireddy &lt;madhukar.pappireddy@arm.com&gt;
4 years ago
/* Refer to Note 1 in function tspd_sel1_interrupt_handler()*/
#if TSP_NS_INTR_ASYNC_PREEMPT
if (tsp_ctx->preempted_by_sel1_intr) {
/* Reset the flag */
tsp_ctx->preempted_by_sel1_intr = false;
SMC_RET1(ns_cpu_context, SMC_PREEMPTED);
} else {
SMC_RET0((uint64_t) ns_cpu_context);
}
#else
SMC_RET0((uint64_t) ns_cpu_context);
Bug fix in tspd interrupt handling when TSP_NS_INTR_ASYNC_PREEMPT is enabled Typically, interrupts for a specific security state get handled in the same security execption level if the execution is in the same security state. For example, if a non-secure interrupt gets fired when CPU is executing in NS-EL2 it gets handled in the non-secure world. However, interrupts belonging to the opposite security state typically demand a world(context) switch. This is inline with the security principle which states a secure interrupt has to be handled in the secure world. Hence, the TSPD in EL3 expects the context(handle) for a secure interrupt to be non-secure and vice versa. The function &#34;tspd_sel1_interrupt_handler&#34; is the handler registered for S-EL1 interrupts by the TSPD. Based on the above assumption, it provides an assertion to validate if the interrupt originated from non-secure world and upon success arranges entry into the TSP at &#39;tsp_sel1_intr_entry&#39; for handling the interrupt. However, a race condition between non-secure and secure interrupts can lead to a scenario where the above assumptions do not hold true and further leading to following assert fail. This patch fixes the bug which causes this assert fail: ASSERT: services/spd/tspd/tspd_main.c:105 BACKTRACE: START: assert 0: EL3: 0x400c128 1: EL3: 0x400faf8 2: EL3: 0x40099a4 3: EL3: 0x4010d54 BACKTRACE: END: assert Change-Id: I359d30fb5dbb1429a4a3c3fff37fdc64c07e9414 Signed-off-by: Madhukar Pappireddy &lt;madhukar.pappireddy@arm.com&gt;
4 years ago
#endif
/*
* This function ID is used only by the SP to indicate it has
* finished initialising itself after a cold boot
*/
case TSP_ENTRY_DONE:
if (ns)
SMC_RET1(handle, SMC_UNK);
/*
* Stash the SP entry points information. This is done
* only once on the primary cpu
*/
assert(tsp_vectors == NULL);
tsp_vectors = (tsp_vectors_t *) x1;
if (tsp_vectors) {
set_tsp_pstate(tsp_ctx->state, TSP_PSTATE_ON);
/*
* TSP has been successfully initialized. Register power
* management hooks with PSCI
*/
psci_register_spd_pm_hook(&tspd_pm);
/*
* Register an interrupt handler for S-EL1 interrupts
* when generated during code executing in the
* non-secure state.
*/
flags = 0;
set_interrupt_rm_flag(flags, NON_SECURE);
rc = register_interrupt_type_handler(INTR_TYPE_S_EL1,
tspd_sel1_interrupt_handler,
flags);
if (rc)
panic();
#if TSP_NS_INTR_ASYNC_PREEMPT
/*
* Register an interrupt handler for NS interrupts when
* generated during code executing in secure state are
* routed to EL3.
*/
flags = 0;
set_interrupt_rm_flag(flags, SECURE);
rc = register_interrupt_type_handler(INTR_TYPE_NS,
tspd_ns_interrupt_handler,
flags);
if (rc)
panic();
/*
Unify interrupt return paths from TSP into the TSPD The TSP is expected to pass control back to EL3 if it gets preempted due to an interrupt while handling a Standard SMC in the following scenarios: 1. An FIQ preempts Standard SMC execution and that FIQ is not a TSP Secure timer interrupt or is preempted by a higher priority interrupt by the time the TSP acknowledges it. In this case, the TSP issues an SMC with the ID as `TSP_EL3_FIQ`. Currently this case is never expected to happen as only the TSP Secure Timer is expected to generate FIQ. 2. An IRQ preempts Standard SMC execution and in this case the TSP issues an SMC with the ID as `TSP_PREEMPTED`. In both the cases, the TSPD hands control back to the normal world and returns returns an error code to the normal world to indicate that the standard SMC it had issued has been preempted but not completed. This patch unifies the handling of these two cases in the TSPD and ensures that the TSP only uses TSP_PREEMPTED instead of separate SMC IDs. Also instead of 2 separate error codes, SMC_PREEMPTED and TSP_EL3_FIQ, only SMC_PREEMPTED is returned as error code back to the normal world. Background information: On a GICv3 system, when the secure world has affinity routing enabled, in 2. an FIQ will preempt TSP execution instead of an IRQ. The FIQ could be a result of a Group 0 or a Group 1 NS interrupt. In both case, the TSPD passes control back to the normal world upon receipt of the TSP_PREEMPTED SMC. A Group 0 interrupt will immediately preempt execution to EL3 where it will be handled. This allows for unified interrupt handling in TSP for both GICv3 and GICv2 systems. Change-Id: I9895344db74b188021e3f6a694701ad272fb40d4
9 years ago
* Disable the NS interrupt locally.
*/
disable_intr_rm_local(INTR_TYPE_NS, SECURE);
#endif
}
#if TSP_INIT_ASYNC
/* Save the Secure EL1 system register context */
assert(cm_get_context(SECURE) == &tsp_ctx->cpu_ctx);
cm_el1_sysregs_context_save(SECURE);
/* Program EL3 registers to enable entry into the next EL */
next_image_info = bl31_plat_get_next_image_ep_info(NON_SECURE);
assert(next_image_info);
assert(NON_SECURE ==
GET_SECURITY_STATE(next_image_info->h.attr));
cm_init_my_context(next_image_info);
cm_prepare_el3_exit(NON_SECURE);
SMC_RET0(cm_get_context(NON_SECURE));
#else
/*
* SP reports completion. The SPD must have initiated
* the original request through a synchronous entry
* into the SP. Jump back to the original C runtime
* context.
*/
tspd_synchronous_sp_exit(tsp_ctx, x1);
break;
#endif
/*
* This function ID is used only by the SP to indicate it has finished
* aborting a preempted Yielding SMC Call.
*/
case TSP_ABORT_DONE:
/*
* These function IDs are used only by the SP to indicate it has
* finished:
* 1. turning itself on in response to an earlier psci
* cpu_on request
* 2. resuming itself after an earlier psci cpu_suspend
* request.
*/
case TSP_ON_DONE:
case TSP_RESUME_DONE:
/*
* These function IDs are used only by the SP to indicate it has
* finished:
* 1. suspending itself after an earlier psci cpu_suspend
* request.
* 2. turning itself off in response to an earlier psci
* cpu_off request.
*/
case TSP_OFF_DONE:
case TSP_SUSPEND_DONE:
case TSP_SYSTEM_OFF_DONE:
case TSP_SYSTEM_RESET_DONE:
if (ns)
SMC_RET1(handle, SMC_UNK);
/*
* SP reports completion. The SPD must have initiated the
* original request through a synchronous entry into the SP.
* Jump back to the original C runtime context, and pass x1 as
* return value to the caller
*/
tspd_synchronous_sp_exit(tsp_ctx, x1);
break;
/*
* Request from non-secure client to perform an
* arithmetic operation or response from secure
* payload to an earlier request.
*/
case TSP_FAST_FID(TSP_ADD):
case TSP_FAST_FID(TSP_SUB):
case TSP_FAST_FID(TSP_MUL):
case TSP_FAST_FID(TSP_DIV):
case TSP_YIELD_FID(TSP_ADD):
case TSP_YIELD_FID(TSP_SUB):
case TSP_YIELD_FID(TSP_MUL):
case TSP_YIELD_FID(TSP_DIV):
/*
* Request from non-secure client to perform a check
* of the DIT PSTATE bit.
*/
case TSP_YIELD_FID(TSP_CHECK_DIT):
if (ns) {
/*
* This is a fresh request from the non-secure client.
* The parameters are in x1 and x2. Figure out which
* registers need to be preserved, save the non-secure
* state and send the request to the secure payload.
*/
assert(handle == cm_get_context(NON_SECURE));
/* Check if we are already preempted */
if (get_yield_smc_active_flag(tsp_ctx->state))
SMC_RET1(handle, SMC_UNK);
cm_el1_sysregs_context_save(NON_SECURE);
/* Save x1 and x2 for use by TSP_GET_ARGS call below */
store_tsp_args(tsp_ctx, x1, x2);
/*
* We are done stashing the non-secure context. Ask the
* secure payload to do the work now.
*/
/*
* Verify if there is a valid context to use, copy the
* operation type and parameters to the secure context
* and jump to the fast smc entry point in the secure
* payload. Entry into S-EL1 will take place upon exit
* from this function.
*/
assert(&tsp_ctx->cpu_ctx == cm_get_context(SECURE));
/* Set appropriate entry for SMC.
* We expect the TSP to manage the PSTATE.I and PSTATE.F
* flags as appropriate.
*/
if (GET_SMC_TYPE(smc_fid) == SMC_TYPE_FAST) {
cm_set_elr_el3(SECURE, (uint64_t)
&tsp_vectors->fast_smc_entry);
} else {
set_yield_smc_active_flag(tsp_ctx->state);
cm_set_elr_el3(SECURE, (uint64_t)
&tsp_vectors->yield_smc_entry);
#if TSP_NS_INTR_ASYNC_PREEMPT
/*
* Enable the routing of NS interrupts to EL3
* during processing of a Yielding SMC Call on
* this core.
*/
enable_intr_rm_local(INTR_TYPE_NS, SECURE);
#endif
#if EL3_EXCEPTION_HANDLING
/*
* With EL3 exception handling, while an SMC is
* being processed, Non-secure interrupts can't
* preempt Secure execution. However, for
* yielding SMCs, we want preemption to happen;
* so explicitly allow NS preemption in this
* case, and supply the preemption return code
* for TSP.
*/
ehf_allow_ns_preemption(TSP_PREEMPTED);
#endif
}
cm_el1_sysregs_context_restore(SECURE);
cm_set_next_eret_context(SECURE);
SMC_RET3(&tsp_ctx->cpu_ctx, smc_fid, x1, x2);
} else {
/*
* This is the result from the secure client of an
* earlier request. The results are in x1-x3. Copy it
* into the non-secure context, save the secure state
* and return to the non-secure state.
*/
assert(handle == cm_get_context(SECURE));
cm_el1_sysregs_context_save(SECURE);
/* Get a reference to the non-secure context */
ns_cpu_context = cm_get_context(NON_SECURE);
assert(ns_cpu_context);
/* Restore non-secure state */
cm_el1_sysregs_context_restore(NON_SECURE);
cm_set_next_eret_context(NON_SECURE);
if (GET_SMC_TYPE(smc_fid) == SMC_TYPE_YIELD) {
clr_yield_smc_active_flag(tsp_ctx->state);
#if TSP_NS_INTR_ASYNC_PREEMPT
/*
* Disable the routing of NS interrupts to EL3
* after processing of a Yielding SMC Call on
* this core is finished.
*/
disable_intr_rm_local(INTR_TYPE_NS, SECURE);
#endif
}
SMC_RET3(ns_cpu_context, x1, x2, x3);
}
assert(0); /* Unreachable */
/*
* Request from the non-secure world to abort a preempted Yielding SMC
* Call.
*/
case TSP_FID_ABORT:
/* ABORT should only be invoked by normal world */
if (!ns) {
assert(0);
break;
}
assert(handle == cm_get_context(NON_SECURE));
cm_el1_sysregs_context_save(NON_SECURE);
/* Abort the preempted SMC request */
if (!tspd_abort_preempted_smc(tsp_ctx)) {
/*
* If there was no preempted SMC to abort, return
* SMC_UNK.
*
* Restoring the NON_SECURE context is not necessary as
* the synchronous entry did not take place if the
* return code of tspd_abort_preempted_smc is zero.
*/
cm_set_next_eret_context(NON_SECURE);
break;
}
cm_el1_sysregs_context_restore(NON_SECURE);
cm_set_next_eret_context(NON_SECURE);
SMC_RET1(handle, SMC_OK);
/*
* Request from non secure world to resume the preempted
* Yielding SMC Call.
*/
case TSP_FID_RESUME:
/* RESUME should be invoked only by normal world */
if (!ns) {
assert(0);
break;
}
/*
* This is a resume request from the non-secure client.
* save the non-secure state and send the request to
* the secure payload.
*/
assert(handle == cm_get_context(NON_SECURE));
/* Check if we are already preempted before resume */
if (!get_yield_smc_active_flag(tsp_ctx->state))
SMC_RET1(handle, SMC_UNK);
cm_el1_sysregs_context_save(NON_SECURE);
/*
* We are done stashing the non-secure context. Ask the
* secure payload to do the work now.
*/
#if TSP_NS_INTR_ASYNC_PREEMPT
/*
* Enable the routing of NS interrupts to EL3 during resumption
* of a Yielding SMC Call on this core.
*/
enable_intr_rm_local(INTR_TYPE_NS, SECURE);
#endif
#if EL3_EXCEPTION_HANDLING
/*
* Allow the resumed yielding SMC processing to be preempted by
* Non-secure interrupts. Also, supply the preemption return
* code for TSP.
*/
ehf_allow_ns_preemption(TSP_PREEMPTED);
#endif
/* We just need to return to the preempted point in
* TSP and the execution will resume as normal.
*/
cm_el1_sysregs_context_restore(SECURE);
cm_set_next_eret_context(SECURE);
SMC_RET0(&tsp_ctx->cpu_ctx);
/*
* This is a request from the secure payload for more arguments
* for an ongoing arithmetic operation requested by the
* non-secure world. Simply return the arguments from the non-
* secure client in the original call.
*/
case TSP_GET_ARGS:
if (ns)
SMC_RET1(handle, SMC_UNK);
get_tsp_args(tsp_ctx, x1, x2);
SMC_RET2(handle, x1, x2);
case TOS_CALL_COUNT:
/*
* Return the number of service function IDs implemented to
* provide service to non-secure
*/
SMC_RET1(handle, TSP_NUM_FID);
case TOS_UID:
/* Return TSP UID to the caller */
SMC_UUID_RET(handle, tsp_uuid);
case TOS_CALL_VERSION:
/* Return the version of current implementation */
SMC_RET2(handle, TSP_VERSION_MAJOR, TSP_VERSION_MINOR);
default:
break;
}
SMC_RET1(handle, SMC_UNK);
}
/* Define a SPD runtime service descriptor for fast SMC calls */
DECLARE_RT_SVC(
tspd_fast,
OEN_TOS_START,
OEN_TOS_END,
SMC_TYPE_FAST,
tspd_setup,
tspd_smc_handler
);
/* Define a SPD runtime service descriptor for Yielding SMC Calls */
DECLARE_RT_SVC(
tspd_std,
OEN_TOS_START,
OEN_TOS_END,
SMC_TYPE_YIELD,
NULL,
tspd_smc_handler
);