coresight-cpu-debug.rst 8.4 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193
  1. ==========================
  2. Coresight CPU Debug Module
  3. ==========================
  4. :Author: Leo Yan <leo.yan@linaro.org>
  5. :Date: April 5th, 2017
  6. Introduction
  7. ------------
  8. Coresight CPU debug module is defined in ARMv8-a architecture reference manual
  9. (ARM DDI 0487A.k) Chapter 'Part H: External debug', the CPU can integrate
  10. debug module and it is mainly used for two modes: self-hosted debug and
  11. external debug. Usually the external debug mode is well known as the external
  12. debugger connects with SoC from JTAG port; on the other hand the program can
  13. explore debugging method which rely on self-hosted debug mode, this document
  14. is to focus on this part.
  15. The debug module provides sample-based profiling extension, which can be used
  16. to sample CPU program counter, secure state and exception level, etc; usually
  17. every CPU has one dedicated debug module to be connected. Based on self-hosted
  18. debug mechanism, Linux kernel can access these related registers from mmio
  19. region when the kernel panic happens. The callback notifier for kernel panic
  20. will dump related registers for every CPU; finally this is good for assistant
  21. analysis for panic.
  22. Implementation
  23. --------------
  24. - During driver registration, it uses EDDEVID and EDDEVID1 - two device ID
  25. registers to decide if sample-based profiling is implemented or not. On some
  26. platforms this hardware feature is fully or partially implemented; and if
  27. this feature is not supported then registration will fail.
  28. - At the time this documentation was written, the debug driver mainly relies on
  29. information gathered by the kernel panic callback notifier from three
  30. sampling registers: EDPCSR, EDVIDSR and EDCIDSR: from EDPCSR we can get
  31. program counter; EDVIDSR has information for secure state, exception level,
  32. bit width, etc; EDCIDSR is context ID value which contains the sampled value
  33. of CONTEXTIDR_EL1.
  34. - The driver supports a CPU running in either AArch64 or AArch32 mode. The
  35. registers naming convention is a bit different between them, AArch64 uses
  36. 'ED' for register prefix (ARM DDI 0487A.k, chapter H9.1) and AArch32 uses
  37. 'DBG' as prefix (ARM DDI 0487A.k, chapter G5.1). The driver is unified to
  38. use AArch64 naming convention.
  39. - ARMv8-a (ARM DDI 0487A.k) and ARMv7-a (ARM DDI 0406C.b) have different
  40. register bits definition. So the driver consolidates two difference:
  41. If PCSROffset=0b0000, on ARMv8-a the feature of EDPCSR is not implemented;
  42. but ARMv7-a defines "PCSR samples are offset by a value that depends on the
  43. instruction set state". For ARMv7-a, the driver checks furthermore if CPU
  44. runs with ARM or thumb instruction set and calibrate PCSR value, the
  45. detailed description for offset is in ARMv7-a ARM (ARM DDI 0406C.b) chapter
  46. C11.11.34 "DBGPCSR, Program Counter Sampling Register".
  47. If PCSROffset=0b0010, ARMv8-a defines "EDPCSR implemented, and samples have
  48. no offset applied and do not sample the instruction set state in AArch32
  49. state". So on ARMv8 if EDDEVID1.PCSROffset is 0b0010 and the CPU operates
  50. in AArch32 state, EDPCSR is not sampled; when the CPU operates in AArch64
  51. state EDPCSR is sampled and no offset are applied.
  52. Clock and power domain
  53. ----------------------
  54. Before accessing debug registers, we should ensure the clock and power domain
  55. have been enabled properly. In ARMv8-a ARM (ARM DDI 0487A.k) chapter 'H9.1
  56. Debug registers', the debug registers are spread into two domains: the debug
  57. domain and the CPU domain.
  58. ::
  59. +---------------+
  60. | |
  61. | |
  62. +----------+--+ |
  63. dbg_clock -->| |**| |<-- cpu_clock
  64. | Debug |**| CPU |
  65. dbg_power_domain -->| |**| |<-- cpu_power_domain
  66. +----------+--+ |
  67. | |
  68. | |
  69. +---------------+
  70. For debug domain, the user uses DT binding "clocks" and "power-domains" to
  71. specify the corresponding clock source and power supply for the debug logic.
  72. The driver calls the pm_runtime_{put|get} operations as needed to handle the
  73. debug power domain.
  74. For CPU domain, the different SoC designs have different power management
  75. schemes and finally this heavily impacts external debug module. So we can
  76. divide into below cases:
  77. - On systems with a sane power controller which can behave correctly with
  78. respect to CPU power domain, the CPU power domain can be controlled by
  79. register EDPRCR in driver. The driver firstly writes bit EDPRCR.COREPURQ
  80. to power up the CPU, and then writes bit EDPRCR.CORENPDRQ for emulation
  81. of CPU power down. As result, this can ensure the CPU power domain is
  82. powered on properly during the period when access debug related registers;
  83. - Some designs will power down an entire cluster if all CPUs on the cluster
  84. are powered down - including the parts of the debug registers that should
  85. remain powered in the debug power domain. The bits in EDPRCR are not
  86. respected in these cases, so these designs do not support debug over
  87. power down in the way that the CoreSight / Debug designers anticipated.
  88. This means that even checking EDPRSR has the potential to cause a bus hang
  89. if the target register is unpowered.
  90. In this case, accessing to the debug registers while they are not powered
  91. is a recipe for disaster; so we need preventing CPU low power states at boot
  92. time or when user enable module at the run time. Please see chapter
  93. "How to use the module" for detailed usage info for this.
  94. Device Tree Bindings
  95. --------------------
  96. See Documentation/devicetree/bindings/arm/arm,coresight-cpu-debug.yaml for
  97. details.
  98. How to use the module
  99. ---------------------
  100. If you want to enable debugging functionality at boot time, you can add
  101. "coresight_cpu_debug.enable=1" to the kernel command line parameter.
  102. The driver also can work as module, so can enable the debugging when insmod
  103. module::
  104. # insmod coresight_cpu_debug.ko debug=1
  105. When boot time or insmod module you have not enabled the debugging, the driver
  106. uses the debugfs file system to provide a knob to dynamically enable or disable
  107. debugging:
  108. To enable it, write a '1' into /sys/kernel/debug/coresight_cpu_debug/enable::
  109. # echo 1 > /sys/kernel/debug/coresight_cpu_debug/enable
  110. To disable it, write a '0' into /sys/kernel/debug/coresight_cpu_debug/enable::
  111. # echo 0 > /sys/kernel/debug/coresight_cpu_debug/enable
  112. As explained in chapter "Clock and power domain", if you are working on one
  113. platform which has idle states to power off debug logic and the power
  114. controller cannot work well for the request from EDPRCR, then you should
  115. firstly constraint CPU idle states before enable CPU debugging feature; so can
  116. ensure the accessing to debug logic.
  117. If you want to limit idle states at boot time, you can use "nohlt" or
  118. "cpuidle.off=1" in the kernel command line.
  119. At the runtime you can disable idle states with below methods:
  120. It is possible to disable CPU idle states by way of the PM QoS
  121. subsystem, more specifically by using the "/dev/cpu_dma_latency"
  122. interface (see Documentation/power/pm_qos_interface.rst for more
  123. details). As specified in the PM QoS documentation the requested
  124. parameter will stay in effect until the file descriptor is released.
  125. For example::
  126. # exec 3<> /dev/cpu_dma_latency; echo 0 >&3
  127. ...
  128. Do some work...
  129. ...
  130. # exec 3<>-
  131. The same can also be done from an application program.
  132. Disable specific CPU's specific idle state from cpuidle sysfs (see
  133. Documentation/admin-guide/pm/cpuidle.rst)::
  134. # echo 1 > /sys/devices/system/cpu/cpu$cpu/cpuidle/state$state/disable
  135. Output format
  136. -------------
  137. Here is an example of the debugging output format::
  138. ARM external debug module:
  139. coresight-cpu-debug 850000.debug: CPU[0]:
  140. coresight-cpu-debug 850000.debug: EDPRSR: 00000001 (Power:On DLK:Unlock)
  141. coresight-cpu-debug 850000.debug: EDPCSR: handle_IPI+0x174/0x1d8
  142. coresight-cpu-debug 850000.debug: EDCIDSR: 00000000
  143. coresight-cpu-debug 850000.debug: EDVIDSR: 90000000 (State:Non-secure Mode:EL1/0 Width:64bits VMID:0)
  144. coresight-cpu-debug 852000.debug: CPU[1]:
  145. coresight-cpu-debug 852000.debug: EDPRSR: 00000001 (Power:On DLK:Unlock)
  146. coresight-cpu-debug 852000.debug: EDPCSR: debug_notifier_call+0x23c/0x358
  147. coresight-cpu-debug 852000.debug: EDCIDSR: 00000000
  148. coresight-cpu-debug 852000.debug: EDVIDSR: 90000000 (State:Non-secure Mode:EL1/0 Width:64bits VMID:0)