andrius
/
arm-trusted-firmware
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

meson: gxl: Fix CPU hotplug 

The CPU[1-3] are reset to initial/cold boot state (with their reset
address set to 0x0). In this state the cpus are waiting for another
one to set the reset address to bl31_warm_entrypoint and wake them up.

The CPU0 needs a bit of a workaround as changing the reset address
either through PSCI mailbox or the mmio mapped RVBAR (at 0xda834650)
does not seem to have any effect. Thus the workaround consists in
emulating the other CPUs' behavior with a WFE loop and manually jumping
to bl31_warm_entrypoint when woken back up by another one.

Change-Id: I11265620b5fd0619285e3993253a3f9a3ff6a7a4
Signed-off-by: Remi Pommarel <repk@triplefau.lt>
This commit is contained in:
Remi Pommarel 2019-07-30 18:04:38 +02:00
parent 43d4a2910d
commit b4694a8677
1 changed files with 12 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

14
plat/meson/gxl/gxl_pm.c
View File

@ -162,7 +162,8 @@ static void gxbb_pwr_domain_off(const psci_power_state_t *target_state)
static void __dead2 gxbb_pwr_domain_pwr_down_wfi(const psci_power_state_t
*target_state)
{
unsigned int core = plat_gxbb_calc_core_pos(read_mpidr_el1());
u_register_t mpidr = read_mpidr_el1();
unsigned int core = plat_gxbb_calc_core_pos(mpidr);
/* CPU0 can't be turned OFF, emulate it with a WFE loop */
if (core == GXBB_PRIMARY_CPU) {
@ -173,10 +174,19 @@ static void __dead2 gxbb_pwr_domain_pwr_down_wfi(const psci_power_state_t
VERBOSE("BL31: CPU0 resumed.\n");
write_rmr_el3(RMR_EL3_RR_BIT | RMR_EL3_AA64_BIT);
/*
* Because setting CPU0's warm reset entrypoint through PSCI
* mailbox and/or mmio mapped RVBAR (0xda834650) does not seem
* to work, jump to it manually.
* In order to avoid an assert, mmu has to be disabled.
*/
disable_mmu_el3();
((void(*)(void))gxbb_sec_entrypoint)();
}
dsbsy();
gxl_pm_set_reset_addr(mpidr, 0);
gxl_pm_reset(mpidr);
for (;;)
wfi();