fix: skip CPU cycle allocation for SP vCPU to reach message loop
With the support added in Hafnium SPMC for secondary CPU cold boot,
secondary execution contexts of SPs dont need a round of CPU cycles
through ffa_run to reach the message loop.
Signed-off-by: Madhukar Pappireddy <madhukar.pappireddy@arm.com>
Change-Id: Ib02f51abb31d30329d43f0937ba30d721504bc53
diff --git a/spm/cactus/cactus_main.c b/spm/cactus/cactus_main.c
index b608d7b..c15ee4e 100644
--- a/spm/cactus/cactus_main.c
+++ b/spm/cactus/cactus_main.c
@@ -61,8 +61,9 @@
ffa_ret = ffa_msg_wait();
for (;;) {
- VERBOSE("Woke up with func id: %s\n",
- ffa_func_name(ffa_func_id(ffa_ret)));
+ VERBOSE("Woke up with func:%s id: %x\n",
+ ffa_func_name(ffa_func_id(ffa_ret)),
+ ffa_func_id(ffa_ret));
if (ffa_func_id(ffa_ret) == FFA_ERROR) {
ERROR("Error: %s\n",
@@ -74,7 +75,7 @@
ffa_func_id(ffa_ret) != FFA_MSG_SEND_DIRECT_REQ_SMC64 &&
ffa_func_id(ffa_ret) != FFA_INTERRUPT &&
ffa_func_id(ffa_ret) != FFA_RUN) {
- ERROR("%s(%u) unknown func id %s\n", __func__, vm_id,
+ ERROR("%s(%x) unknown func id %s\n", __func__, vm_id,
ffa_func_name(ffa_func_id(ffa_ret)));
break;
}
@@ -87,10 +88,15 @@
* informational as we're running with virtual
* interrupts unmasked and the interrupt is processed
* by the interrupt handler.
- *
- * Received FFA_RUN in waiting state, the endpoint
- * simply returns by FFA_MSG_WAIT.
*/
+ if (ffa_func_id(ffa_ret) == FFA_RUN) {
+ /*
+ * Received FFA_RUN in waiting state, the
+ * endpoint simply returns by FFA_MSG_WAIT.
+ */
+ VERBOSE("Nothing to do. Exit to NWd\n");
+ }
+
ffa_ret = ffa_msg_wait();
continue;
}