MSM平台RPM

转载https://blog.csdn.net/loongembedded/article/details/51024610

oftware Component Block Diagram

RPM（Resource Power Manager）是高通MSM平台另外加的一块芯片，虽然与AP芯片打包在一起，但其是一个独立的ARM Core。之所以加这个东西，就是要控制整个电源相关的shared resources，比如ldo，clock。负责与SMP,MPM交互进入睡眠或者唤醒整个系统。 
以下是高通文档中对各个功能模块的说明。 

• Kernel – DALSys-based lightweight kernel
• RPM handler – RPM handler abstracts the RPM message protocol away 
  from other software
• Drivers – Drivers for each of the resources supported by the RPM will 
  register with RPM handler to request notification when requests are 
  received for the resource which the driver controls
• NPA – A driver may use the Node Power Architecture (NPA) to represent 
  resources controlled by the driver
• Clock driver – RPM clock driver directly handles aggregating requests 
  from each of the masters for any of the systemwide clock resources 
  controlled by the RPM
• Bus arbitration driver – RPM bus arbiter driver takes bus arbiter settings 
  as requests from the different masters in the system and aggregates them 
  to represent the frequency-independent system settings
• PMIC – RPM PMIC driver directly aggregates requests from each of the 
  masters for any of the systemwide PMIC resources controlled by the RPM
• Watch Dog driver – Watch Dog driver is a fail-safe for incorrect or stuck 
  code
• MPM driver – Used to program the MPM hardware block during 
  systemwide sleep
• RPM message driver – RPM message protocol driver abstracts the RPM 
  message protocol away from other subsystem software

Messaging Masters

与RPM通过shared memory region交互进行dynamic and static resource/power management的可以有很多种。 
这个可以查看smd_type.h中的smd_channel_type。但目前看只有AP,MODEM,RIVA,TZ与RPM有交互，这个可以看message_ram_malloc()函数中的设置。 
其实也可以间接从rpm_config.c文件中的SystempData temp_config_data这个变量中看出来到底有几个部分是与RPM进行交互的。

RPM Initialization

在main.c文件中会逐个调用init_fcns[]变量中的函数进行初始化。当然也包括上面的资源的初始化。

1. const init_fcn init_fcns[] =

2. {

3. populate_image_header,

4. npa_init,

5.  

6. #if (!defined(MSM8909_STUBS) )

7. railway_init_v2,

8. #endif

9. PlatformInfo_Init, /* pm_init is using PlatformInfo APIs */

10. pm_init,//LDO等资源的注册

11. acc_init,

12.  

13. #if (!defined(MSM8909_STUBS) )

14. railway_init_early_proxy_votes,

15. #endif

16. // xpu_init, /* cookie set here also indicates to SBL that railway is ready */

17. (init_fcn)Clock_Init, //Clock资源的注册

18. __init_stack_chk_guard,

19. ddr_init,

20. smem_init,

21. init_smdl,

22. version_init, /* Needs to be after smem_init */

23. rpmserver_init,

24. rpm_server_init_done,

25. railway_init_proxies_and_pins,

26.  

27. #if (!defined(MSM8909_STUBS) )

28. rbcpr_init,

29. #endif

30. svs_init,

31. vmpm_init,

32. sleep_init,

33.  

34. #if (!defined(MSM8909_STUBS) )

35. QDSSInit,

36. #endif

37. exceptions_enable,

38. swevent_qdss_init,

39. icb_init,

40.  

41. #if (!defined(MSM8909_STUBS) )

42. debug_init,

43. system_db_init,

44. zqcal_task_init,

45. #endif

46. rpm_settling_timer_init,

47. gpio_toggle_init,

48. rpm_set_changer_common_init,

49. }

pm_init ： LDO等资源的注册，然后接收rpm_message。接收rpm_message的部分高通代码没有给，所以看不到，但从rpm log来看，是有接收处理并反馈的过程的。

1. 107.764902: rpm_message_received (master: "APSS") (message id: 723)

2. 107.764908: rpm_svs (mode: RPM_SVS_FAST) (reason: imminent processing)

3. 107.764924: rpm_svs (mode: RPM_SVS_FAST) (reason: imminent processing)

4. 107.764939: rpm_process_request (master: "APSS") (resource type: ldoa) (id: 14)

5. 107.764942: rpm_xlate_request (resource type: ldoa) (resource id: 14)

6. 107.764946: rpm_apply_request (resource type: ldoa) (resource id: 14)

7. 107.765033: rpm_send_message_response (master: "APSS")

• ldoa对应的resource type为RPM_LDO_A_REQ。这个在pm_rom_device_init()里的pm_rpm_ldo_register_resources(RPM_LDO_A_REQ, num_of_ldoa); 这里被注册，所以看进去最后xlate和apply最终都会被pm_rpm_ldo_translation()和pm_rpm_ldo_apply进行处理。pm_rpm_ldo_tranlation()读取kvp内容，pm_rpm_ldo_apply最终把request的内容设置上去。

1. kernel端在msm-pm8916-rpm-regulator.dtsi文件中定如下ldo内容

2. (init_fcn)Clock_Init：Clock资源的注册，这个过程和上面的也差不多 
3. rpmserver_init：启动接收message的进程

比如要设置LDO3的电压和电流，kvp的内容如下：

1. {

2. “req\0” : {

3. {“rsrc” : “ldo\0”}

4. {“id” : 3}

5. {“set” : 0}

6. {“data” : {

7. {“uv\0\0” : 1100000}

8. {“mA\0\0” : 130}

9. }

10. }

11. }

12. }

RPM log里边显示是APSS，这个在kernel里边也是有设置的，在msm-pm8916-rpm-regulator.dtsi里的

1. rpm-regulator-ldoa14 {

2. compatible = "qcom,rpm-smd-regulator-resource";

3. qcom,resource-name = "ldoa";

4. qcom,resource-id = <14>;

5. qcom,regulator-type = <0>;

6. qcom,hpm-min-load = <5000>;

7. status = "disabled";

8.  

9. regulator-l14 {

10. compatible = "qcom,rpm-smd-regulator";

11. regulator-name = "8916_l14";

12. qcom,set = <3>;

13. status = "disabled";

14. };

15. };

MSM的retulator相关的驱动会读取这个，如果有需要发请求的话，会根据这个dtsi的内容，组织kvp，然后通过rpm_request通道发过去。 
rpm_request通道在rpm里边打开的地方是在rpm_handler.cpp里边的Hander::init()函数

1. void Handler::init()

2. {

3. smdlPort_ = smdl_open("rpm_requests",

4. rpm->ees[ee_].edge,

5. SMDL_OPEN_FLAGS_MODE_PACKET,

6. rpm->ees[ee_].smd_fifo_sz,

7. (smdl_callback_t)rpm_smd_handler,

8. this);

9. }

• 1
• 2
• 3
• 4
• 5
• 6
• 7
• 8
• 9

在对应的kernel端也有打开，可以参考msm8916.dtsi文件里边的设置。

1. rpm_bus: qcom,rpm-smd {

2. compatible = "qcom,rpm-smd";

3. rpm-channel-name = "rpm_requests";

4. rpm-channel-type = <15>; /* SMD_APPS_RPM */

5. };

• 1
• 2
• 3
• 4
• 5

以下是注册资源的函数和资源的类型： 
所有的资源都通过rpm_register_resource()函数注册。这些资源包括clock, ldo等。具体可以看下面的类型定义。

1. typedef enum

2. {

3. RPM_TEST_REQ = 0x74736574, // 'test' in little endiant

4. RPM_CLOCK_0_REQ = 0x306b6c63, // 'clk0' in little endian; misc clocks [CXO, QDSS, dcvs.ena]

5. RPM_CLOCK_1_REQ = 0x316b6c63, // 'clk1' in little endian; bus clocks [pcnoc, snoc, sysmmnoc]

6. RPM_CLOCK_2_REQ = 0x326b6c63, // 'clk2' in little endian; memory clocks [bimc]

7. RPM_CLOCK_QPIC_REQ = 0x63697071, // 'qpic' in little endian; clock [qpic]

8. RPM_BUS_SLAVE_REQ = 0x766c7362, // 'bslv' in little endian

9. RPM_BUS_MASTER_REQ = 0x73616d62, // 'bmas' in little endian

10. RPM_BUS_SPDM_CLK_REQ = 0x63707362, // 'bspc' in little endian

11. RPM_BUS_MASTER_LATENCY_REQ = 0x74616c62, // 'blat' in little endian

12. RPM_SMPS_A_REQ = 0x61706D73, // 'smpa' in little endian

13. RPM_LDO_A_REQ = 0x616F646C, // 'ldoa' in little endian

14. RPM_NCP_A_REQ = 0x6170636E, // 'ncpa' in little endian

15. RPM_VS_A_REQ = 0x617376, // 'vsa' in little endian

16. RPM_CLK_BUFFER_A_REQ = 0x616B6C63, // 'clka' in little endian

17. RPM_BOOST_A_REQ = 0x61747362, // 'bsta' in little endian

18. RPM_SMPS_B_REQ = 0x62706D73, // 'smpb' in little endian

19. RPM_LDO_B_REQ = 0x626F646C, // 'ldob' in little endian

20. RPM_NCP_B_REQ = 0x6270636E, // 'ncpb' in little endian

21. RPM_VS_B_REQ = 0x627376, // 'vsb' in little endian

22. RPM_CLK_BUFFER_B_REQ = 0x626B6C63, // 'clk' in little endian

23. RPM_BOOST_B_REQ = 0x62747362, // 'bstb' in little endian

24. RPM_SWEVENT_REQ = 0x76657773, // 'swev' in little endian

25. RPM_OCMEM_POWER_REQ = 0x706d636f, // 憃cmp?in little endian

26. RPM_RBCPR_REQ = 0x727063, // 'cpr'in little endian

27. RPM_GPIO_TOGGLE_REQ = 0x6F697067, // 'gpio' in little endian:[gpio0,gpio1,gpio2]

28. } rpm_resource_type;

• 1
• 2
• 3
• 4
• 5
• 6
• 7
• 8
• 9
• 10
• 11
• 12
• 13
• 14
• 15
• 16
• 17
• 18
• 19
• 20
• 21
• 22
• 23
• 24
• 25
• 26
• 27
• 28

RPM Resouce Registration

1. rpm_register_resource()：

1. Gpio_toggle.c (x:\j3_ctc\nhlos\rpm_proc\core\power\rpm\common): rpm_register_resource(RPM_GPIO_TOGGLE_REQ, 3, sizeof(gpio_toggle_inrep), gpio_toggle_xlate, gpio_toggle_apply, 0);

2.  

3. Icb_rpm_spdm_req.c (x:\j3_ctc\nhlos\rpm_proc\core\buses\icb\src\common): rpm_register_resource( RPM_BUS_SPDM_CLK_REQ,

4.  

5. Ocmem_resource.c (x:\j3_ctc\nhlos\rpm_proc\core\power\ocmem\src): rpm_register_resource(RPM_OCMEM_POWER_REQ, 3, sizeof(rpm_ocmem_vote_int_rep), rpm_ocmem_xlate, rpm_ocmem_apply, 0);

6.  

7. Pm_rpm_boost_trans_apply.c (x:\j3_ctc\nhlos\rpm_proc\core\systemdrivers\pmic\npa\src\rpm): rpm_register_resource(resourceType, num_npa_resources + 1 , sizeof(pm_npa_boost_int_rep), pm_rpm_boost_translation, pm_rpm_boost_apply, (void *)boost_data);

8. Pm_rpm_clk_buffer_trans_apply.c (x:\j3_ctc\nhlos\rpm_proc\core\systemdrivers\pmic\npa\src\rpm): rpm_register_resource(resourceType, num_npa_resources + 1,

9.  

10. Pm_rpm_ldo_trans_apply.c (x:\j3_ctc\nhlos\rpm_proc\core\systemdrivers\pmic\npa\src\rpm): rpm_register_resource(resourceType, num_npa_resources + 1, sizeof(pm_npa_ldo_int_rep), pm_rpm_ldo_translation, pm_rpm_ldo_apply, (void *)ldo_data);

11.  

12. Pm_rpm_smps_trans_apply.c (x:\j3_ctc\nhlos\rpm_proc\core\systemdrivers\pmic\npa\src\rpm): rpm_register_resource(resourceType, num_npa_resources + 1, sizeof(pm_npa_smps_int_rep), pm_rpm_smps_translation, pm_rpm_smps_apply, (void *)smps_data);

13.  

14. Pm_rpm_vs_trans_apply.c (x:\j3_ctc\nhlos\rpm_proc\core\systemdrivers\pmic\npa\src\rpm): rpm_register_resource(resourceType, num_npa_resources + 1, sizeof(pm_npa_vs_int_rep), pm_rpm_vs_translation, pm_rpm_vs_apply, (void *)vs_data);

15.  

16. Rpmserver.cpp (x:\j3_ctc\nhlos\rpm_proc\core\power\rpm\server):void rpm_register_resource

17.  

18. Rpmserver.h (x:\j3_ctc\nhlos\rpm_proc\core\api\power):void rpm_register_resource

19.  

20. Rpm_npa.cpp (x:\j3_ctc\nhlos\rpm_proc\core\power\rpm\server): rpm_register_resource(resource, num_npa_resources, sizeof(npa_request_data_t), rpm_npa_xlate, rpm_npa_apply, adapter);

21.  

22. Rpm_npa.cpp (x:\j3_ctc\nhlos\rpm_proc\core\power\rpm\server): rpm_register_resource(resource, num_npa_resources, sizeof(npa_request_data_t), rpm_npa_xlate, rpm_npa_apply, adapter);

23.  

24. Rpm_test_resource.cpp (x:\j3_ctc\nhlos\rpm_proc\core\power\rpm\server): rpm_register_resource(RPM_TEST_REQ, 1, 4, rpm_test_xlate, rpm_test_apply, 0);

25.  

26. Swevent.c (x:\j3_ctc\nhlos\rpm_proc\core\power\rpm\common): rpm_register_resource(RPM_SWEVENT_REQ, 1, sizeof(swevent_inrep), rpm_swevent_xlate, rpm_trace_control, 0);

• 1
• 2
• 3
• 4
• 5
• 6
• 7
• 8
• 9
• 10
• 11
• 12
• 13
• 14
• 15
• 16
• 17
• 18
• 19
• 20
• 21
• 22
• 23
• 24
• 25
• 26

所有的资源都是通过rpm_register_resource函数注册。可以看到像RPM_BUS_SPDM_CLK_REQ，RPM_GPIO_TOGGLE_REQ都是通过这个接口直接注册的。

1. NPA相关也是通过rpm_register_resource注册，但是用了NPA的driver去实际注册资源和使用资源。 
   NPA相关的两个注册接口是：rpm_create_npa_adapter()，rpm_create_npa_settling_adapter()。 
   可以看到这两个就注册了三个CLOCK相关的资源，

1. ---- rpm_create_npa_adapter Matches (8 in 5 files) ----

2. ClockRPM.c : clk0_adapter = rpm_create_npa_adapter(RPM_CLOCK_0_REQ, 3); // Misc clocks: [CXO, QDSS, dcvs.ena]

3. ClockRPM.c : clk1_adapter = rpm_create_npa_adapter(RPM_CLOCK_1_REQ, 2); // Bus clocks: [pcnoc, snoc]

4. Rpmserver.h :rpm_npa_adapter rpm_create_npa_adapter(rpm_resource_type resource, unsigned num_npa_resources);

5. Rpm_npa.cpp :rpm_npa_adapter rpm_create_npa_adapter(rpm_resource_type resource, unsigned num_npa_resources)

6.  

7. ---- rpm_create_npa_settling_adapter Matches (5 in 5 files) ----

8. ClockRPM.c : clk2_adapter = rpm_create_npa_settling_adapter(RPM_CLOCK_2_REQ, 1); // Memory clocks: [bimc ]

9. Rpmserver.h :rpm_npa_adapter rpm_create_npa_settling_adapter(rpm_resource_type resource, unsigned num_npa_resources);

10. Rpm_npa.cpp :rpm_npa_adapter rpm_create_npa_settling_adapter(rpm_resource_type resource, unsigned num_npa_resources)

• 1
• 2
• 3
• 4
• 5
• 6
• 7
• 8
• 9
• 10

NPA client的创建函数是:npa_create_sync_client() 
NPA clien request的函数是:npa_issue_required_request() 
NPA client创建并request这个资源，必须要适用像下面这样的npa node。这个可以直接像下面这样定义。然后再使用。步骤如下：

1.定义npa node definition

1. static npa_resource_definition sleep_uber_resource[] =

2. {

3. {

4. "/sleep/uber", /* Name */

5. "on/off", /* Units */

6. 0x7, /* Max State */

7. &npa_or_plugin, /* Plugin */

8. NPA_RESOURCE_DEFAULT, /* Attributes */

9. NULL, /* User Data */

10. }

11. };

12.  

13. npa_node_definition sleep_uber_node =

14. {

15. "/node/sleep/uber", /* name */

16. sleep_uber_driver, /* driver_fcn */

17. NPA_NODE_DEFAULT, /* attributes */

18. NULL, /* data */

19. 0, NULL, /* dependency count, dependency list */

20. NPA_ARRAY(sleep_uber_resource)

21. };

• 1
• 2
• 3
• 4
• 5
• 6
• 7
• 8
• 9
• 10
• 11
• 12
• 13
• 14
• 15
• 16
• 17
• 18
• 19
• 20
• 21

2.npa_define_node(&sleep_uber_node, initial_state, NULL)，调用这个函数初始化这个NPA node 
3.创建Client

1. uber_node_handle = npa_create_sync_client("/sleep/uber",

2. "sleep",

3. NPA_CLIENT_REQUIRED);

• 1
• 2
• 3

4.npa_issue_required_request(uber_node_handle, request) ： request

RPM Resource handle

当子系统通过share memory发送请求给RPM。RPM负责处理这些请求并设置。RPM处理请求的流程如下： 
SMD IRQ ->smd_isr()->rpm_smd_handler() [在Handler::init函数中注册的，rpm_request channel的SMD处理函数]->Handler::queue()->schedule_me()->Handler::execute_until()->Handler::processMessage()->resource_ee_request()->发到每个资源注册xlate然后再调用apply等

下面以高通控制DDR频率(BIMC)的过程为例，看一下kernel这边怎么发送请求给RPM的。 
device tree设置如下：

1. cpubw: qcom,cpubw@0 {

2. reg = <0 4>;

3. compatible = "qcom,devbw";

4. governor = "cpufreq";

5. qcom,src-dst-ports = <1 512>;

6. qcom,active-only;

7. qcom,bw-tbl = //这个频率表对应RPM中BIMC频率表，不过删掉了50MHz和9.6MHz。

8. /* 73 9.60 MHz */

9. /* 381 50MHz */

10. < 762 /* 100 MHz */>,

11. < 1525 /* 200 MHz */>,

12. < 3051 /* 400 MHz */>,

13. < 4066 /* 533 MHz */>;

14. };

• 1
• 2
• 3
• 4
• 5
• 6
• 7
• 8
• 9
• 10
• 11
• 12
• 13
• 14

对应的kernel代码在devfeq_devbw.c文件。根据算法算出当前应该设定的ddr总线频率之后，最后通过以下顺序发消息给RPM。这里暂时不讨论按什么规则选择需要的频率的，只看按什么路径发频率给RPM的。 
set_bw()->msm_bus_scale_client_update_request()->update_request_adhoc()[msm_bus_arb_adhoc.c]->update_path()->msm_bus_commit_data()->flush_bw_data()->send_rpm_msg()->msm_rpm_send_message()

System Sleep Overview

RPM除了对Clock和LDO等资源的管理之外，还管理整个系统睡眠。 

可以看到睡眠并不是通过shared memory发送消息给RPM的，而是子系统通过设置对应的SPM，SPM触发RPM相应的中断来完成的。 
kernel这边设置spm的模式的设置如下：

1. enum {

2. MSM_SPM_MODE_DISABLED,

3. MSM_SPM_MODE_CLOCK_GATING,

4. MSM_SPM_MODE_RETENTION,

5. MSM_SPM_MODE_GDHS,

6. MSM_SPM_MODE_POWER_COLLAPSE,//设置成这个状态之后，对应的SPM应该就会触发RPM的对应的shutdown中断

7. MSM_SPM_MODE_NR

8. };

• 1
• 2
• 3
• 4
• 5
• 6
• 7
• 8

rpm中，与哪几个子系统传递接收message，然后和哪几个子系统的SPM进行交互，交互的中断号是多少的设置如下：

1. static SystemData temp_config_data =

2. {

3. .num_ees = 4, // 4 EE's, [apps, modem, pronto, tz]

4. .ees = (EEData[] ) {

5. [0] = {

6. .edge = SMD_APPS_RPM,

7. .smd_fifo_sz = 1024,

8. .ee_buflen = 256,

9. .priority = 4,

10. .wakeupInt = (1 << 5) | (1 << 7),

11. .spm = {

12. .numCores = 1,

13. .bringupInts = (unsigned[]) { 15 },

14. .bringupAcks = (unsigned[]) { 20 },

15. .shutdownInts = (unsigned[]) { 14 },//这个应该是SPM睡眠的中断号

16. .shutdownAcks = (unsigned[]) { 4 },

17. },

18. },

19. [1] = {

20. .edge = SMD_MODEM_RPM,

21. .smd_fifo_sz = 1024,

22. .ee_buflen = 1024,

23. .priority = 2,

24. .wakeupInt = (1 << 13) | (1 << 15),

25. .spm = {

26. .numCores = 1,

27. .bringupInts = (unsigned[]) { 25 },

28. .bringupAcks = (unsigned[]) { 22 },

29. .shutdownInts = (unsigned[]) { 24 },

30. .shutdownAcks = (unsigned[]) { 6 },

31. },

32. },

33. [2] = {

34. .edge = SMD_RIVA_RPM,

35. .smd_fifo_sz = 1024,

36. .ee_buflen = 256,

37. .priority = 1,

38. .wakeupInt = (1 << 17) | (1 << 19),

39. .spm = {

40. .numCores = 1,

41. .bringupInts = (unsigned[]) { 31 },

42. .bringupAcks = (unsigned[]) { 23 },

43. .shutdownInts = (unsigned[]) { 30 },

44. .shutdownAcks = (unsigned[]) { 7 },

45. },

46. },

47. [3] = {

48. .edge = SMD_RPM_TZ,

49. .smd_fifo_sz = 1024,

50. .ee_buflen = 256,

51. .priority = 5,

52. .wakeupInt = 0,

53. .spm = {

54. .numCores = 0,

55. .bringupInts = (unsigned[]) { 31 },

56. .bringupAcks = (unsigned[]) { 23 },

57. .shutdownInts = (unsigned[]) { 30 },

58. .shutdownAcks = (unsigned[]) { 7 },

59. },

60. },

61. },

62. .supported_classes = SUPPORTED_CLASSES,

63. .supported_resources = SUPPORTED_RESOURCES,

64. .classes = (ResourceClassData[SUPPORTED_CLASSES]) { 0 },

65. .resources = (ResourceData[SUPPORTED_RESOURCES]) { 0 },

66. .resource_seeds = (int16_t[SUPPORTED_RESOURCES]) { 0 },

67. }

• 1
• 2
• 3
• 4
• 5
• 6
• 7
• 8
• 9
• 10
• 11
• 12
• 13
• 14
• 15
• 16
• 17
• 18
• 19
• 20
• 21
• 22
• 23
• 24
• 25
• 26
• 27
• 28
• 29
• 30
• 31
• 32
• 33
• 34
• 35
• 36
• 37
• 38
• 39
• 40
• 41
• 42
• 43
• 44
• 45
• 46
• 47
• 48
• 49
• 50
• 51
• 52
• 53
• 54
• 55
• 56
• 57
• 58
• 59
• 60
• 61
• 62
• 63
• 64
• 65
• 66
• 67

但还不知道这些中断号到底怎么来的，，中断的个数都是1个，，这个怎么来的也还是不知~ 
这些内容在rpm_spm_init中会读取，然后设置中断。shutdownISR对应的中断处理函数为rpm_spm_shutdown_high_isr。 
这个中断函数会调用rpm_spm_state_machine()处理RPM状态机，进入或者阻止进入睡眠模式等。

1. void rpm_spm_state_machine(unsigned ee, rpm_spm_entry_reason reason)

2. {

3. INTLOCK();

4.  

5. bool changed_state = false;

6. EEData *ee_state = &(rpm->ees[ee]);

7. SetChanger *changer = ee_state->changer;

8.  

9. do

10. {

11. switch(ee_state->subsystem_status)

12. {

13. case SPM_AWAKE:

14. changed_state = FALSE;

15.  

16. if(0 == ee_state->num_active_cores)

17. {//等待所有的core都进入睡眠！！之后才能走到GOITONG TO SLEEP状态！！

18. SPM_CHANGE_STATE(SPM_GOING_TO_SLEEP);

19. }

20. else

21. {

22. // We're awake, so make sure we keep up with any incoming bringup reqs.

23. rpm_acknowledge_spm_handshakes(ee);

24. }

25. break;

26.  

27. case SPM_GOING_TO_SLEEP:

28. if(changed_state)

29. {

30. // check for scheduled wakeup

31. uint64_t deadline = 0;

32.  

33. if(! rpm_get_wakeup_deadline(ee, deadline))

34. {

35. deadline = 0;

36. }

37. changer->setWakeTime (deadline);

38. // enqueue immediate set transition to sleep

39. changer->enqueue(RPM_SLEEP_SET, 0);

40. }

41. changed_state = FALSE;

42.  

43. // When we've finished selecting the sleep set, we're officially asleep.

44. if((SPM_TRANSITION_COMPLETE == reason) && (RPM_SLEEP_SET == changer->currentSet()))

45. {

46. SPM_CHANGE_STATE(SPM_SLEEPING);

47. }

48. // However, we might get a wakeup request before we've made it all the way to sleep.

49. if(SPM_BRINGUP_REQ == reason)

50. {

51. // Set the preempt flag; this will force the set change to recycle if

52. // it's currently running. It will notice the processor has woken up

53. // and stop performing its work.

54. theSchedule().preempt();

55. }

56. break;

57.  

58. case SPM_SLEEPING:

59. if(changed_state)

60. {

61. // check for scheduled wakeup

62. uint64_t deadline = changer->getWakeTime ();

63.  

64. // enqueue scheduled wakeup request

65. changer->enqueue(RPM_ACTIVE_SET, deadline);

66. }

67. changed_state = FALSE;

68.  

69. if(ee_state->num_active_cores > 0)

70. {

71. SPM_CHANGE_STATE(SPM_WAKING_UP);

72. }

73. break;

74.  

75. case SPM_WAKING_UP:

76. if(changed_state)

77. {

78. // work our way back to the active set

79. if(RPM_SLEEP_SET == changer->currentSet() || changer->inTransition())

80. {

81. changer->enqueue(RPM_ACTIVE_SET, 0);

82. }

83. }

84. changed_state = FALSE;

85.  

86. // check for completion

87. if(RPM_ACTIVE_SET == changer->currentSet() && !changer->inTransition())

88. {

89. SPM_CHANGE_STATE(SPM_AWAKE);

90. }

91. break;

92. }

93. } while(changed_state);

94.  

95. INTFREE();

96. }

• 1
• 2
• 3
• 4
• 5
• 6
• 7
• 8
• 9
• 10
• 11
• 12
• 13
• 14
• 15
• 16
• 17
• 18
• 19
• 20
• 21
• 22
• 23
• 24
• 25
• 26
• 27
• 28
• 29
• 30
• 31
• 32
• 33
• 34
• 35
• 36
• 37
• 38
• 39
• 40
• 41
• 42
• 43
• 44
• 45
• 46
• 47
• 48
• 49
• 50
• 51
• 52
• 53
• 54
• 55
• 56
• 57
• 58
• 59
• 60
• 61
• 62
• 63
• 64
• 65
• 66
• 67
• 68
• 69
• 70
• 71
• 72
• 73
• 74
• 75
• 76
• 77
• 78
• 79
• 80
• 81
• 82
• 83
• 84
• 85
• 86
• 87
• 88
• 89
• 90
• 91
• 92
• 93
• 94
• 95
• 96

最终会跑到SetChanger::enqueue()里。

RPM与MPM的交互

RPM 状态读取

rpm的状态读取，在/d/rpm_stats里边可以读到AP这边设置的几个APSS,MPSS,PRONTO等几个对应的RPM状态。 
读的内容当然也是从rpm_request这个shared memory里边读的。相应的设置在msm8916-pm.dts文件里边有。

1. qcom,rpm-stats@29dba0 {

2. compatible = "qcom,rpm-stats";

3. reg = <0x29dba0 0x1000>;

4. reg-names = "phys_addr_base";

5. qcom,sleep-stats-version = <2>;

6. };

7. qcom,rpm-master-stats@60150 {

8. compatible = "qcom,rpm-master-stats";

9. reg = <0x60150 0x2030>;

10. qcom,masters = "APSS", "MPSS", "PRONTO";

11. qcom,master-stats-version = <2>;

12. qcom,master-offset = <4096>;

13. };

14. qcom,rpm-rbcpr-stats@0x29daa0 {

15. compatible = "qcom,rpmrbcpr-stats";

16. reg = <0x29daa0 0x1a0000>;

17. qcom,start-offset = <0x190010>;

18. };