hv: pm: enhencement platform S5 entering operation

Now, we have assumption that SOS control whether the platform
should enter S5 or not. So when SOS tries enter S5, we just
forward the S5 request to native port which make sure platform
S5 is totally aligned with SOS S5.

With higher serverity guest introduced,this assumption is not
true any more. We need to extend the platform S5 process to
handle higher severity guest:
  - For DM launched RTVM, we need to make sure these guests
    is off before put the whole platfrom to S5.

  - For pre-launched VM, there are two cases:
    * if os running in it support S5, we wait for guests off.
    * if os running in it doesn't support S5, we expect it
      will invoke one hypercall to notify HV to shutdown it.
      NOTE: this case is not supported yet. Will add it in the
      future.

Tracked-On: #3564
Signed-off-by: Yin Fengwei <fengwei.yin@intel.com>
Reviewed-by: Li, Fei1 <fei1.li@intel.com>
This commit is contained in:
Yin Fengwei
2019-09-06 14:14:06 +08:00
committed by ACRN System Integration
parent ce9375874c
commit f039d75998
3 changed files with 62 additions and 9 deletions

View File

@@ -31,6 +31,7 @@ struct acpi_reset_reg {
struct pm_s_state_data *get_host_sstate_data(void);
void host_enter_s3(struct pm_s_state_data *sstate_data, uint32_t pm1a_cnt_val, uint32_t pm1b_cnt_val);
void host_enter_s5(struct pm_s_state_data *sstate_data, uint32_t pm1a_cnt_val, uint32_t pm1b_cnt_val);
extern void asm_enter_s3(struct pm_s_state_data *sstate_data, uint32_t pm1a_cnt_val, uint32_t pm1b_cnt_val);
extern void restore_s3_context(void);
struct cpu_state_info *get_cpu_pm_state_info(void);