xref: /rk3399_ARM-atf/docs/porting-guide.rst (revision 430f246e58d146949d399d72294f56403672bee0)
1Porting Guide
2=============
3
4Introduction
5------------
6
7Porting Trusted Firmware-A (TF-A) to a new platform involves making some
8mandatory and optional modifications for both the cold and warm boot paths.
9Modifications consist of:
10
11-  Implementing a platform-specific function or variable,
12-  Setting up the execution context in a certain way, or
13-  Defining certain constants (for example #defines).
14
15The platform-specific functions and variables are declared in
16``include/plat/common/platform.h``. The firmware provides a default
17implementation of variables and functions to fulfill the optional requirements
18in order to ease the porting effort. Each platform port can use them as is or
19provide their own implementation if the default implementation is inadequate.
20
21   .. note::
22
23      TF-A historically provided default implementations of platform interfaces
24      as *weak* functions. This practice is now discouraged and new platform
25      interfaces as they get introduced in the code base should be *strongly*
26      defined. We intend to convert existing weak functions over time. Until
27      then, you will find references to *weak* functions in this document.
28
29Please review the :ref:`Threat Model` documents as part of the porting
30effort. Some platform interfaces play a key role in mitigating against some of
31the threats. Failing to fulfill these expectations could undermine the security
32guarantees offered by TF-A. These platform responsibilities are highlighted in
33the threat assessment section, under the "`Mitigations implemented?`" box for
34each threat.
35
36Some modifications are common to all Boot Loader (BL) stages. Section 2
37discusses these in detail. The subsequent sections discuss the remaining
38modifications for each BL stage in detail.
39
40Please refer to the :ref:`Platform Ports Policy` for the policy regarding
41compatibility and deprecation of these porting interfaces.
42
43Only Arm development platforms (such as FVP and Juno) may use the
44functions/definitions in ``include/plat/arm/common/`` and the corresponding
45source files in ``plat/arm/common/``. This is done so that there are no
46dependencies between platforms maintained by different people/companies. If you
47want to use any of the functionality present in ``plat/arm`` files, please
48propose a patch that moves the code to ``plat/common`` so that it can be
49discussed.
50
51Common modifications
52--------------------
53
54This section covers the modifications that should be made by the platform for
55each BL stage to correctly port the firmware stack. They are categorized as
56either mandatory or optional.
57
58Common mandatory modifications
59------------------------------
60
61A platform port must enable the Memory Management Unit (MMU) as well as the
62instruction and data caches for each BL stage. Setting up the translation
63tables is the responsibility of the platform port because memory maps differ
64across platforms. A memory translation library (see ``lib/xlat_tables_v2/``) is
65provided to help in this setup.
66
67Note that although this library supports non-identity mappings, this is intended
68only for re-mapping peripheral physical addresses and allows platforms with high
69I/O addresses to reduce their virtual address space. All other addresses
70corresponding to code and data must currently use an identity mapping.
71
72Also, the only translation granule size supported in TF-A is 4KB, as various
73parts of the code assume that is the case. It is not possible to switch to
7416 KB or 64 KB granule sizes at the moment.
75
76In Arm standard platforms, each BL stage configures the MMU in the
77platform-specific architecture setup function, ``blX_plat_arch_setup()``, and uses
78an identity mapping for all addresses.
79
80If the build option ``USE_COHERENT_MEM`` is enabled, each platform can allocate a
81block of identity mapped secure memory with Device-nGnRE attributes aligned to
82page boundary (4K) for each BL stage. All sections which allocate coherent
83memory are grouped under ``.coherent_ram``. For ex: Bakery locks are placed in a
84section identified by name ``.bakery_lock`` inside ``.coherent_ram`` so that its
85possible for the firmware to place variables in it using the following C code
86directive:
87
88::
89
90    __section(".bakery_lock")
91
92Or alternatively the following assembler code directive:
93
94::
95
96    .section .bakery_lock
97
98The ``.coherent_ram`` section is a sum of all sections like ``.bakery_lock`` which are
99used to allocate any data structures that are accessed both when a CPU is
100executing with its MMU and caches enabled, and when it's running with its MMU
101and caches disabled. Examples are given below.
102
103The following variables, functions and constants must be defined by the platform
104for the firmware to work correctly.
105
106.. _platform_def_mandatory:
107
108File : platform_def.h [mandatory]
109~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
110
111Each platform must ensure that a header file of this name is in the system
112include path with the following constants defined. This will require updating
113the list of ``PLAT_INCLUDES`` in the ``platform.mk`` file.
114
115Platform ports may optionally use the file ``include/plat/common/common_def.h``,
116which provides typical values for some of the constants below. These values are
117likely to be suitable for all platform ports.
118
119-  **#define : PLATFORM_LINKER_FORMAT**
120
121   Defines the linker format used by the platform, for example
122   ``elf64-littleaarch64``.
123
124-  **#define : PLATFORM_LINKER_ARCH**
125
126   Defines the processor architecture for the linker by the platform, for
127   example ``aarch64``.
128
129-  **#define : PLATFORM_STACK_SIZE**
130
131   Defines the normal stack memory available to each CPU. This constant is used
132   by ``plat/common/aarch64/platform_mp_stack.S`` and
133   ``plat/common/aarch64/platform_up_stack.S``.
134
135-  **#define : CACHE_WRITEBACK_GRANULE**
136
137   Defines the size in bytes of the largest cache line across all the cache
138   levels in the platform.
139
140-  **#define : FIRMWARE_WELCOME_STR**
141
142   Defines the character string printed by BL1 upon entry into the ``bl1_main()``
143   function.
144
145-  **#define : PLATFORM_CORE_COUNT**
146
147   Defines the total number of CPUs implemented by the platform across all
148   clusters in the system.
149
150-  **#define : PLAT_NUM_PWR_DOMAINS**
151
152   Defines the total number of nodes in the power domain topology
153   tree at all the power domain levels used by the platform.
154   This macro is used by the PSCI implementation to allocate
155   data structures to represent power domain topology.
156
157-  **#define : PLAT_MAX_PWR_LVL**
158
159   Defines the maximum power domain level that the power management operations
160   should apply to. More often, but not always, the power domain level
161   corresponds to affinity level. This macro allows the PSCI implementation
162   to know the highest power domain level that it should consider for power
163   management operations in the system that the platform implements. For
164   example, the Base AEM FVP implements two clusters with a configurable
165   number of CPUs and it reports the maximum power domain level as 1.
166
167-  **#define : PLAT_MAX_OFF_STATE**
168
169   Defines the local power state corresponding to the deepest power down
170   possible at every power domain level in the platform. The local power
171   states for each level may be sparsely allocated between 0 and this value
172   with 0 being reserved for the RUN state. The PSCI implementation uses this
173   value to initialize the local power states of the power domain nodes and
174   to specify the requested power state for a PSCI_CPU_OFF call.
175
176-  **#define : PLAT_MAX_RET_STATE**
177
178   Defines the local power state corresponding to the deepest retention state
179   possible at every power domain level in the platform. This macro should be
180   a value less than PLAT_MAX_OFF_STATE and greater than 0. It is used by the
181   PSCI implementation to distinguish between retention and power down local
182   power states within PSCI_CPU_SUSPEND call.
183
184-  **#define : PLAT_MAX_PWR_LVL_STATES**
185
186   Defines the maximum number of local power states per power domain level
187   that the platform supports. The default value of this macro is 2 since
188   most platforms just support a maximum of two local power states at each
189   power domain level (power-down and retention). If the platform needs to
190   account for more local power states, then it must redefine this macro.
191
192   Currently, this macro is used by the Generic PSCI implementation to size
193   the array used for PSCI_STAT_COUNT/RESIDENCY accounting.
194
195-  **#define : BL1_RO_BASE**
196
197   Defines the base address in secure ROM where BL1 originally lives. Must be
198   aligned on a page-size boundary.
199
200-  **#define : BL1_RO_LIMIT**
201
202   Defines the maximum address in secure ROM that BL1's actual content (i.e.
203   excluding any data section allocated at runtime) can occupy.
204
205-  **#define : BL1_RW_BASE**
206
207   Defines the base address in secure RAM where BL1's read-write data will live
208   at runtime. Must be aligned on a page-size boundary.
209
210-  **#define : BL1_RW_LIMIT**
211
212   Defines the maximum address in secure RAM that BL1's read-write data can
213   occupy at runtime.
214
215-  **#define : BL2_BASE**
216
217   Defines the base address in secure RAM where BL1 loads the BL2 binary image.
218   Must be aligned on a page-size boundary. This constant is not applicable
219   when BL2_IN_XIP_MEM is set to '1'.
220
221-  **#define : BL2_LIMIT**
222
223   Defines the maximum address in secure RAM that the BL2 image can occupy.
224   This constant is not applicable when BL2_IN_XIP_MEM is set to '1'.
225
226-  **#define : BL2_RO_BASE**
227
228   Defines the base address in secure XIP memory where BL2 RO section originally
229   lives. Must be aligned on a page-size boundary. This constant is only needed
230   when BL2_IN_XIP_MEM is set to '1'.
231
232-  **#define : BL2_RO_LIMIT**
233
234   Defines the maximum address in secure XIP memory that BL2's actual content
235   (i.e. excluding any data section allocated at runtime) can occupy. This
236   constant is only needed when BL2_IN_XIP_MEM is set to '1'.
237
238-  **#define : BL2_RW_BASE**
239
240   Defines the base address in secure RAM where BL2's read-write data will live
241   at runtime. Must be aligned on a page-size boundary. This constant is only
242   needed when BL2_IN_XIP_MEM is set to '1'.
243
244-  **#define : BL2_RW_LIMIT**
245
246   Defines the maximum address in secure RAM that BL2's read-write data can
247   occupy at runtime. This constant is only needed when BL2_IN_XIP_MEM is set
248   to '1'.
249
250-  **#define : BL31_BASE**
251
252   Defines the base address in secure RAM where BL2 loads the BL31 binary
253   image. Must be aligned on a page-size boundary.
254
255-  **#define : BL31_LIMIT**
256
257   Defines the maximum address in secure RAM that the BL31 image can occupy.
258
259-  **#define : PLAT_RSE_COMMS_PAYLOAD_MAX_SIZE**
260
261   Defines the maximum message size between AP and RSE. Need to define if
262   platform supports RSE.
263
264For every image, the platform must define individual identifiers that will be
265used by BL1 or BL2 to load the corresponding image into memory from non-volatile
266storage. For the sake of performance, integer numbers will be used as
267identifiers. The platform will use those identifiers to return the relevant
268information about the image to be loaded (file handler, load address,
269authentication information, etc.). The following image identifiers are
270mandatory:
271
272-  **#define : BL2_IMAGE_ID**
273
274   BL2 image identifier, used by BL1 to load BL2.
275
276-  **#define : BL31_IMAGE_ID**
277
278   BL31 image identifier, used by BL2 to load BL31.
279
280-  **#define : BL33_IMAGE_ID**
281
282   BL33 image identifier, used by BL2 to load BL33.
283
284If Trusted Board Boot is enabled, the following certificate identifiers must
285also be defined:
286
287-  **#define : TRUSTED_BOOT_FW_CERT_ID**
288
289   BL2 content certificate identifier, used by BL1 to load the BL2 content
290   certificate.
291
292-  **#define : TRUSTED_KEY_CERT_ID**
293
294   Trusted key certificate identifier, used by BL2 to load the trusted key
295   certificate.
296
297-  **#define : SOC_FW_KEY_CERT_ID**
298
299   BL31 key certificate identifier, used by BL2 to load the BL31 key
300   certificate.
301
302-  **#define : SOC_FW_CONTENT_CERT_ID**
303
304   BL31 content certificate identifier, used by BL2 to load the BL31 content
305   certificate.
306
307-  **#define : NON_TRUSTED_FW_KEY_CERT_ID**
308
309   BL33 key certificate identifier, used by BL2 to load the BL33 key
310   certificate.
311
312-  **#define : NON_TRUSTED_FW_CONTENT_CERT_ID**
313
314   BL33 content certificate identifier, used by BL2 to load the BL33 content
315   certificate.
316
317-  **#define : FWU_CERT_ID**
318
319   Firmware Update (FWU) certificate identifier, used by NS_BL1U to load the
320   FWU content certificate.
321
322If the AP Firmware Updater Configuration image, BL2U is used, the following
323must also be defined:
324
325-  **#define : BL2U_BASE**
326
327   Defines the base address in secure memory where BL1 copies the BL2U binary
328   image. Must be aligned on a page-size boundary.
329
330-  **#define : BL2U_LIMIT**
331
332   Defines the maximum address in secure memory that the BL2U image can occupy.
333
334-  **#define : BL2U_IMAGE_ID**
335
336   BL2U image identifier, used by BL1 to fetch an image descriptor
337   corresponding to BL2U.
338
339If the SCP Firmware Update Configuration Image, SCP_BL2U is used, the following
340must also be defined:
341
342-  **#define : SCP_BL2U_IMAGE_ID**
343
344   SCP_BL2U image identifier, used by BL1 to fetch an image descriptor
345   corresponding to SCP_BL2U.
346
347   .. note::
348      TF-A does not provide source code for this image.
349
350If the Non-Secure Firmware Updater ROM, NS_BL1U is used, the following must
351also be defined:
352
353-  **#define : NS_BL1U_BASE**
354
355   Defines the base address in non-secure ROM where NS_BL1U executes.
356   Must be aligned on a page-size boundary.
357
358   .. note::
359      TF-A does not provide source code for this image.
360
361-  **#define : NS_BL1U_IMAGE_ID**
362
363   NS_BL1U image identifier, used by BL1 to fetch an image descriptor
364   corresponding to NS_BL1U.
365
366If the Non-Secure Firmware Updater, NS_BL2U is used, the following must also
367be defined:
368
369-  **#define : NS_BL2U_BASE**
370
371   Defines the base address in non-secure memory where NS_BL2U executes.
372   Must be aligned on a page-size boundary.
373
374   .. note::
375      TF-A does not provide source code for this image.
376
377-  **#define : NS_BL2U_IMAGE_ID**
378
379   NS_BL2U image identifier, used by BL1 to fetch an image descriptor
380   corresponding to NS_BL2U.
381
382For the the Firmware update capability of TRUSTED BOARD BOOT, the following
383macros may also be defined:
384
385-  **#define : PLAT_FWU_MAX_SIMULTANEOUS_IMAGES**
386
387   Total number of images that can be loaded simultaneously. If the platform
388   doesn't specify any value, it defaults to 10.
389
390If a SCP_BL2 image is supported by the platform, the following constants must
391also be defined:
392
393-  **#define : SCP_BL2_IMAGE_ID**
394
395   SCP_BL2 image identifier, used by BL2 to load SCP_BL2 into secure memory
396   from platform storage before being transferred to the SCP.
397
398-  **#define : SCP_FW_KEY_CERT_ID**
399
400   SCP_BL2 key certificate identifier, used by BL2 to load the SCP_BL2 key
401   certificate (mandatory when Trusted Board Boot is enabled).
402
403-  **#define : SCP_FW_CONTENT_CERT_ID**
404
405   SCP_BL2 content certificate identifier, used by BL2 to load the SCP_BL2
406   content certificate (mandatory when Trusted Board Boot is enabled).
407
408If a BL32 image is supported by the platform, the following constants must
409also be defined:
410
411-  **#define : BL32_IMAGE_ID**
412
413   BL32 image identifier, used by BL2 to load BL32.
414
415-  **#define : TRUSTED_OS_FW_KEY_CERT_ID**
416
417   BL32 key certificate identifier, used by BL2 to load the BL32 key
418   certificate (mandatory when Trusted Board Boot is enabled).
419
420-  **#define : TRUSTED_OS_FW_CONTENT_CERT_ID**
421
422   BL32 content certificate identifier, used by BL2 to load the BL32 content
423   certificate (mandatory when Trusted Board Boot is enabled).
424
425-  **#define : BL32_BASE**
426
427   Defines the base address in secure memory where BL2 loads the BL32 binary
428   image. Must be aligned on a page-size boundary.
429
430-  **#define : BL32_LIMIT**
431
432   Defines the maximum address that the BL32 image can occupy.
433
434If the Test Secure-EL1 Payload (TSP) instantiation of BL32 is supported by the
435platform, the following constants must also be defined:
436
437-  **#define : TSP_SEC_MEM_BASE**
438
439   Defines the base address of the secure memory used by the TSP image on the
440   platform. This must be at the same address or below ``BL32_BASE``.
441
442-  **#define : TSP_SEC_MEM_SIZE**
443
444   Defines the size of the secure memory used by the BL32 image on the
445   platform. ``TSP_SEC_MEM_BASE`` and ``TSP_SEC_MEM_SIZE`` must fully
446   accommodate the memory required by the BL32 image, defined by ``BL32_BASE``
447   and ``BL32_LIMIT``.
448
449-  **#define : TSP_IRQ_SEC_PHY_TIMER**
450
451   Defines the ID of the secure physical generic timer interrupt used by the
452   TSP's interrupt handling code.
453
454If the platform port uses the translation table library code, the following
455constants must also be defined:
456
457-  **#define : PLAT_XLAT_TABLES_DYNAMIC**
458
459   Optional flag that can be set per-image to enable the dynamic allocation of
460   regions even when the MMU is enabled. If not defined, only static
461   functionality will be available, if defined and set to 1 it will also
462   include the dynamic functionality.
463
464-  **#define : MAX_XLAT_TABLES**
465
466   Defines the maximum number of translation tables that are allocated by the
467   translation table library code. To minimize the amount of runtime memory
468   used, choose the smallest value needed to map the required virtual addresses
469   for each BL stage. If ``PLAT_XLAT_TABLES_DYNAMIC`` flag is enabled for a BL
470   image, ``MAX_XLAT_TABLES`` must be defined to accommodate the dynamic regions
471   as well.
472
473-  **#define : MAX_MMAP_REGIONS**
474
475   Defines the maximum number of regions that are allocated by the translation
476   table library code. A region consists of physical base address, virtual base
477   address, size and attributes (Device/Memory, RO/RW, Secure/Non-Secure), as
478   defined in the ``mmap_region_t`` structure. The platform defines the regions
479   that should be mapped. Then, the translation table library will create the
480   corresponding tables and descriptors at runtime. To minimize the amount of
481   runtime memory used, choose the smallest value needed to register the
482   required regions for each BL stage. If ``PLAT_XLAT_TABLES_DYNAMIC`` flag is
483   enabled for a BL image, ``MAX_MMAP_REGIONS`` must be defined to accommodate
484   the dynamic regions as well.
485
486-  **#define : PLAT_VIRT_ADDR_SPACE_SIZE**
487
488   Defines the total size of the virtual address space in bytes. For example,
489   for a 32 bit virtual address space, this value should be ``(1ULL << 32)``.
490
491-  **#define : PLAT_PHY_ADDR_SPACE_SIZE**
492
493   Defines the total size of the physical address space in bytes. For example,
494   for a 32 bit physical address space, this value should be ``(1ULL << 32)``.
495
496If the platform port uses the IO storage framework, the following constants
497must also be defined:
498
499-  **#define : MAX_IO_DEVICES**
500
501   Defines the maximum number of registered IO devices. Attempting to register
502   more devices than this value using ``io_register_device()`` will fail with
503   -ENOMEM.
504
505-  **#define : MAX_IO_HANDLES**
506
507   Defines the maximum number of open IO handles. Attempting to open more IO
508   entities than this value using ``io_open()`` will fail with -ENOMEM.
509
510-  **#define : MAX_IO_BLOCK_DEVICES**
511
512   Defines the maximum number of registered IO block devices. Attempting to
513   register more devices this value using ``io_dev_open()`` will fail
514   with -ENOMEM. MAX_IO_BLOCK_DEVICES should be less than MAX_IO_DEVICES.
515   With this macro, multiple block devices could be supported at the same
516   time.
517
518If the platform needs to allocate data within the per-cpu data framework in
519BL31, it should define the following macro. Currently this is only required if
520the platform decides not to use the coherent memory section by undefining the
521``USE_COHERENT_MEM`` build flag. In this case, the framework allocates the
522required memory within the the per-cpu data to minimize wastage.
523
524-  **#define : PLAT_PCPU_DATA_SIZE**
525
526   Defines the memory (in bytes) to be reserved within the per-cpu data
527   structure for use by the platform layer.
528
529The following constants are optional. They should be defined when the platform
530memory layout implies some image overlaying like in Arm standard platforms.
531
532-  **#define : BL31_PROGBITS_LIMIT**
533
534   Defines the maximum address in secure RAM that the BL31's progbits sections
535   can occupy.
536
537-  **#define : TSP_PROGBITS_LIMIT**
538
539   Defines the maximum address that the TSP's progbits sections can occupy.
540
541If the platform supports OS-initiated mode, i.e. the build option
542``PSCI_OS_INIT_MODE`` is enabled, and if the platform's maximum power domain
543level for PSCI_CPU_SUSPEND differs from ``PLAT_MAX_PWR_LVL``, the following
544constant must be defined.
545
546-  **#define : PLAT_MAX_CPU_SUSPEND_PWR_LVL**
547
548   Defines the maximum power domain level that PSCI_CPU_SUSPEND should apply to.
549
550If the platform port uses the PL061 GPIO driver, the following constant may
551optionally be defined:
552
553-  **PLAT_PL061_MAX_GPIOS**
554   Maximum number of GPIOs required by the platform. This allows control how
555   much memory is allocated for PL061 GPIO controllers. The default value is
556
557   #. $(eval $(call add_define,PLAT_PL061_MAX_GPIOS))
558
559If the platform port uses the partition driver, the following constant may
560optionally be defined:
561
562-  **PLAT_PARTITION_MAX_ENTRIES**
563   Maximum number of partition entries required by the platform. This allows
564   control how much memory is allocated for partition entries. The default
565   value is 128.
566   For example, define the build flag in ``platform.mk``:
567   PLAT_PARTITION_MAX_ENTRIES := 12
568   $(eval $(call add_define,PLAT_PARTITION_MAX_ENTRIES))
569
570-  **PLAT_PARTITION_BLOCK_SIZE**
571   The size of partition block. It could be either 512 bytes or 4096 bytes.
572   The default value is 512.
573   For example, define the build flag in ``platform.mk``:
574   PLAT_PARTITION_BLOCK_SIZE := 4096
575   $(eval $(call add_define,PLAT_PARTITION_BLOCK_SIZE))
576
577If the platform port uses the Arm® Ethos™-N NPU driver, the following
578configuration must be performed:
579
580- The NPU SiP service handler must be hooked up. This consists of both the
581  initial setup (``ethosn_smc_setup``) and the handler itself
582  (``ethosn_smc_handler``)
583
584If the platform port uses the Arm® Ethos™-N NPU driver with TZMP1 support
585enabled, the following constants and configuration must also be defined:
586
587- **ETHOSN_NPU_PROT_FW_NSAID**
588
589  Defines the Non-secure Access IDentity (NSAID) that the NPU shall use to
590  access the protected memory that contains the NPU's firmware.
591
592- **ETHOSN_NPU_PROT_DATA_RW_NSAID**
593
594  Defines the Non-secure Access IDentity (NSAID) that the NPU shall use for
595  read/write access to the protected memory that contains inference data.
596
597- **ETHOSN_NPU_PROT_DATA_RO_NSAID**
598
599  Defines the Non-secure Access IDentity (NSAID) that the NPU shall use for
600  read-only access to the protected memory that contains inference data.
601
602- **ETHOSN_NPU_NS_RW_DATA_NSAID**
603
604  Defines the Non-secure Access IDentity (NSAID) that the NPU shall use for
605  read/write access to the non-protected memory.
606
607- **ETHOSN_NPU_NS_RO_DATA_NSAID**
608
609  Defines the Non-secure Access IDentity (NSAID) that the NPU shall use for
610  read-only access to the non-protected memory.
611
612- **ETHOSN_NPU_FW_IMAGE_BASE** and **ETHOSN_NPU_FW_IMAGE_LIMIT**
613
614  Defines the physical address range that the NPU's firmware will be loaded
615  into and executed from.
616
617- Configure the platforms TrustZone Controller (TZC) with appropriate regions
618  of protected memory. At minimum this must include a region for the NPU's
619  firmware code and a region for protected inference data, and these must be
620  accessible using the NSAIDs defined above.
621
622- Include the NPU firmware and certificates in the FIP.
623
624- Provide FCONF entries to configure the image source for the NPU firmware
625  and certificates.
626
627- Add MMU mappings such that:
628
629 - BL2 can write the NPU firmware into the region defined by
630   ``ETHOSN_NPU_FW_IMAGE_BASE`` and ``ETHOSN_NPU_FW_IMAGE_LIMIT``
631 - BL31 (SiP service) can read the NPU firmware from the same region
632
633- Add the firmware image ID ``ETHOSN_NPU_FW_IMAGE_ID`` to the list of images
634  loaded by BL2.
635
636Please see the reference implementation code for the Juno platform as an example.
637
638
639The following constant is optional. It should be defined to override the default
640behaviour of the ``assert()`` function (for example, to save memory).
641
642-  **PLAT_LOG_LEVEL_ASSERT**
643   If ``PLAT_LOG_LEVEL_ASSERT`` is higher or equal than ``LOG_LEVEL_VERBOSE``,
644   ``assert()`` prints the name of the file, the line number and the asserted
645   expression. Else if it is higher than ``LOG_LEVEL_INFO``, it prints the file
646   name and the line number. Else if it is lower than ``LOG_LEVEL_INFO``, it
647   doesn't print anything to the console. If ``PLAT_LOG_LEVEL_ASSERT`` isn't
648   defined, it defaults to ``LOG_LEVEL``.
649
650If the platform port uses the DRTM feature, the following constants must be
651defined:
652
653-  **#define : PLAT_DRTM_EVENT_LOG_MAX_SIZE**
654
655   Maximum Event Log size used by the platform. Platform can decide the maximum
656   size of the Event Log buffer, depending upon the highest hash algorithm
657   chosen and the number of components selected to measure during the DRTM
658   execution flow.
659
660-  **#define : PLAT_DRTM_MMAP_ENTRIES**
661
662   Number of the MMAP entries used by the DRTM implementation to calculate the
663   size of address map region of the platform.
664
665File : plat_macros.S [mandatory]
666~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
667
668Each platform must ensure a file of this name is in the system include path with
669the following macro defined. In the Arm development platforms, this file is
670found in ``plat/arm/board/<plat_name>/include/plat_macros.S``.
671
672-  **Macro : plat_crash_print_regs**
673
674   This macro allows the crash reporting routine to print relevant platform
675   registers in case of an unhandled exception in BL31. This aids in debugging
676   and this macro can be defined to be empty in case register reporting is not
677   desired.
678
679   For instance, GIC or interconnect registers may be helpful for
680   troubleshooting.
681
682Handling Reset
683--------------
684
685BL1 by default implements the reset vector where execution starts from a cold
686or warm boot. BL31 can be optionally set as a reset vector using the
687``RESET_TO_BL31`` make variable.
688
689For each CPU, the reset vector code is responsible for the following tasks:
690
691#. Distinguishing between a cold boot and a warm boot.
692
693#. In the case of a cold boot and the CPU being a secondary CPU, ensuring that
694   the CPU is placed in a platform-specific state until the primary CPU
695   performs the necessary steps to remove it from this state.
696
697#. In the case of a warm boot, ensuring that the CPU jumps to a platform-
698   specific address in the BL31 image in the same processor mode as it was
699   when released from reset.
700
701The following functions need to be implemented by the platform port to enable
702reset vector code to perform the above tasks.
703
704Function : plat_get_my_entrypoint() [mandatory when PROGRAMMABLE_RESET_ADDRESS == 0]
705~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
706
707::
708
709    Argument : void
710    Return   : uintptr_t
711
712This function is called with the MMU and caches disabled
713(``SCTLR_EL3.M`` = 0 and ``SCTLR_EL3.C`` = 0). The function is responsible for
714distinguishing between a warm and cold reset for the current CPU using
715platform-specific means. If it's a warm reset, then it returns the warm
716reset entrypoint point provided to ``plat_setup_psci_ops()`` during
717BL31 initialization. If it's a cold reset then this function must return zero.
718
719This function does not follow the Procedure Call Standard used by the
720Application Binary Interface for the Arm 64-bit architecture. The caller should
721not assume that callee saved registers are preserved across a call to this
722function.
723
724This function fulfills requirement 1 and 3 listed above.
725
726Note that for platforms that support programming the reset address, it is
727expected that a CPU will start executing code directly at the right address,
728both on a cold and warm reset. In this case, there is no need to identify the
729type of reset nor to query the warm reset entrypoint. Therefore, implementing
730this function is not required on such platforms.
731
732Function : plat_secondary_cold_boot_setup() [mandatory when COLD_BOOT_SINGLE_CPU == 0]
733~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
734
735::
736
737    Argument : void
738
739This function is called with the MMU and data caches disabled. It is responsible
740for placing the executing secondary CPU in a platform-specific state until the
741primary CPU performs the necessary actions to bring it out of that state and
742allow entry into the OS. This function must not return.
743
744In the Arm FVP port, when using the normal boot flow, each secondary CPU powers
745itself off. The primary CPU is responsible for powering up the secondary CPUs
746when normal world software requires them. When booting an EL3 payload instead,
747they stay powered on and are put in a holding pen until their mailbox gets
748populated.
749
750This function fulfills requirement 2 above.
751
752Note that for platforms that can't release secondary CPUs out of reset, only the
753primary CPU will execute the cold boot code. Therefore, implementing this
754function is not required on such platforms.
755
756Generic hold pen helpers
757~~~~~~~~~~~~~~~~~~~~~~~~
758
759When ``COLD_BOOT_SINGLE_CPU`` is not set, TF-A provides a generic hold pen
760implementation that platforms can use for SMP secondary CPU bring-up. Each
761core gets a cache-line-aligned slot containing an entry field and two magic
762tags, avoiding common pitfalls with cache coherency during the boot handshake
763between primary and secondary CPUs.
764
765The C API consists of ``plat_hold_pen_init()``, to be called during platform
766setup (typically from ``plat_setup_psci_ops()``), ``plat_hold_pen_signal()``
767to be called from ``pwr_domain_on()``, and the assembly macro
768``plat_hold_pen_wait_and_jump`` to be used by waking cores, usually from
769``plat_secondary_cold_boot_setup()``. Detailed documentation for each is in
770``include/plat/common/plat_hold_pen.h`` and
771``include/plat/common/plat_hold_pen.S``. See
772``plat/arm/board/corstone1000/`` and ``plat/qemu/`` for usage examples.
773
774Platforms using this helper must add ``plat/common/plat_hold_pen.c`` to
775``BL31_SOURCES``.
776
777Function : plat_is_my_cpu_primary() [mandatory when COLD_BOOT_SINGLE_CPU == 0]
778~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
779
780::
781
782    Argument : void
783    Return   : unsigned int
784
785This function identifies whether the current CPU is the primary CPU or a
786secondary CPU. A return value of zero indicates that the CPU is not the
787primary CPU, while a non-zero return value indicates that the CPU is the
788primary CPU.
789
790Note that for platforms that can't release secondary CPUs out of reset, only the
791primary CPU will execute the cold boot code. Therefore, there is no need to
792distinguish between primary and secondary CPUs and implementing this function is
793not required.
794
795Function : platform_mem_init() [mandatory]
796~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
797
798::
799
800    Argument : void
801    Return   : void
802
803This function is called before any access to data is made by the firmware, in
804order to carry out any essential memory initialization.
805
806Function: plat_get_rotpk_info()
807~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
808
809::
810
811    Argument : void *, void **, unsigned int *, unsigned int *
812    Return   : int
813
814This function is mandatory when Trusted Board Boot is enabled. It returns a
815pointer to the ROTPK stored in the platform (or a hash of it) and its length.
816The ROTPK must be encoded in DER format according to the following ASN.1
817structure:
818
819::
820
821    AlgorithmIdentifier  ::=  SEQUENCE  {
822        algorithm         OBJECT IDENTIFIER,
823        parameters        ANY DEFINED BY algorithm OPTIONAL
824    }
825
826    SubjectPublicKeyInfo  ::=  SEQUENCE  {
827        algorithm         AlgorithmIdentifier,
828        subjectPublicKey  BIT STRING
829    }
830
831In case the function returns a hash of the key:
832
833::
834
835    DigestInfo ::= SEQUENCE {
836        digestAlgorithm   AlgorithmIdentifier,
837        digest            OCTET STRING
838    }
839
840The function returns 0 on success. Any other value is treated as error by the
841Trusted Board Boot. The function also reports extra information related
842to the ROTPK in the flags parameter:
843
844::
845
846    ROTPK_IS_HASH      : Indicates that the ROTPK returned by the platform is a
847                         hash.
848    ROTPK_NOT_DEPLOYED : This allows the platform to skip certificate ROTPK
849                         verification while the platform ROTPK is not deployed.
850                         When this flag is set, the function does not need to
851                         return a platform ROTPK, and the authentication
852                         framework uses the ROTPK in the certificate without
853                         verifying it against the platform value. This flag
854                         must not be used in a deployed production environment.
855
856Function: plat_get_nv_ctr()
857~~~~~~~~~~~~~~~~~~~~~~~~~~~
858
859::
860
861    Argument : void *, unsigned int *
862    Return   : int
863
864This function is mandatory when Trusted Board Boot is enabled. It returns the
865non-volatile counter value stored in the platform in the second argument. The
866cookie in the first argument may be used to select the counter in case the
867platform provides more than one (for example, on platforms that use the default
868TBBR CoT, the cookie will correspond to the OID values defined in
869TRUSTED_FW_NVCOUNTER_OID or NON_TRUSTED_FW_NVCOUNTER_OID).
870
871The function returns 0 on success. Any other value means the counter value could
872not be retrieved from the platform.
873
874Function: plat_set_nv_ctr()
875~~~~~~~~~~~~~~~~~~~~~~~~~~~
876
877::
878
879    Argument : void *, unsigned int
880    Return   : int
881
882This function is mandatory when Trusted Board Boot is enabled. It sets a new
883counter value in the platform. The cookie in the first argument may be used to
884select the counter (as explained in plat_get_nv_ctr()). The second argument is
885the updated counter value to be written to the NV counter.
886
887The function returns 0 on success. Any other value means the counter value could
888not be updated.
889
890Function: plat_set_nv_ctr2()
891~~~~~~~~~~~~~~~~~~~~~~~~~~~~
892
893::
894
895    Argument : void *, const auth_img_desc_t *, unsigned int
896    Return   : int
897
898This function is optional when Trusted Board Boot is enabled. If this
899interface is defined, then ``plat_set_nv_ctr()`` need not be defined. The
900first argument passed is a cookie and is typically used to
901differentiate between a Non Trusted NV Counter and a Trusted NV
902Counter. The second argument is a pointer to an authentication image
903descriptor and may be used to decide if the counter is allowed to be
904updated or not. The third argument is the updated counter value to
905be written to the NV counter.
906
907The function returns 0 on success. Any other value means the counter value
908either could not be updated or the authentication image descriptor indicates
909that it is not allowed to be updated.
910
911Dynamic Root of Trust for Measurement support (in BL31)
912-------------------------------------------------------
913
914The functions mentioned in this section are mandatory, when platform enables
915DRTM_SUPPORT build flag.
916
917Function : plat_get_addr_mmap()
918~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
919
920::
921
922    Argument : void
923    Return   : const mmap_region_t *
924
925This function is used to return the address of the platform *address-map* table,
926which describes the regions of normal memory, memory mapped I/O
927and non-volatile memory.
928
929Function : plat_has_non_host_platforms()
930~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
931
932::
933
934    Argument : void
935    Return   : bool
936
937This function returns *true* if the platform has any trusted devices capable of
938DMA, otherwise returns *false*.
939
940Function : plat_has_unmanaged_dma_peripherals()
941~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
942
943::
944
945    Argument : void
946    Return   : bool
947
948This function returns *true* if platform uses peripherals whose DMA is not
949managed by an SMMU, otherwise returns *false*.
950
951Note -
952If the platform has peripherals that are not managed by the SMMU, then the
953platform should investigate such peripherals to determine whether they can
954be trusted, and such peripherals should be moved under "Non-host platforms"
955if they can be trusted.
956
957Function : plat_get_total_num_smmus()
958~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
959
960::
961
962    Argument : void
963    Return   : unsigned int
964
965This function returns the total number of SMMUs in the platform.
966
967Function : plat_enumerate_smmus()
968~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
969::
970
971
972    Argument : void
973    Return   : const uintptr_t *, size_t
974
975This function returns an array of SMMU addresses and the actual number of SMMUs
976reported by the platform.
977
978Function : plat_drtm_get_dma_prot_features()
979~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
980
981::
982
983    Argument : void
984    Return   : const plat_drtm_dma_prot_features_t*
985
986This function returns the address of plat_drtm_dma_prot_features_t structure
987containing the maximum number of protected regions and bitmap with the types
988of DMA protection supported by the platform.
989For more details see section 3.3 Table 6 of `DRTM`_ specification.
990
991Function : plat_drtm_dma_prot_get_max_table_bytes()
992~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
993
994::
995
996    Argument : void
997    Return   : uint64_t
998
999This function returns the maximum size of DMA protected regions table in
1000bytes.
1001
1002Function : plat_drtm_get_tpm_features()
1003~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1004
1005::
1006
1007    Argument : void
1008    Return   : const plat_drtm_tpm_features_t*
1009
1010This function returns the address of *plat_drtm_tpm_features_t* structure
1011containing PCR usage schema, TPM-based hash, and firmware hash algorithm
1012supported by the platform.
1013
1014Function : plat_drtm_get_min_size_normal_world_dce()
1015~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1016
1017::
1018
1019    Argument : void
1020    Return   : uint64_t
1021
1022This function returns the size normal-world DCE of the platform.
1023
1024Function : plat_drtm_get_imp_def_dlme_region_size()
1025~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1026
1027::
1028
1029    Argument : void
1030    Return   : uint64_t
1031
1032This function returns the size of implementation defined DLME region
1033of the platform.
1034
1035Function : plat_drtm_get_tcb_hash_table_size()
1036~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1037
1038::
1039
1040    Argument : void
1041    Return   : uint64_t
1042
1043This function returns the size of TCB hash table of the platform.
1044
1045Function : plat_drtm_get_acpi_tables_region_size()
1046~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1047
1048::
1049
1050    Argument : void
1051    Return   : uint64_t
1052
1053This function returns the size of ACPI tables region of the platform.
1054
1055Function : plat_drtm_get_tcb_hash_features()
1056~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1057
1058::
1059
1060    Argument : void
1061    Return   : uint64_t
1062
1063This function returns the Maximum number of TCB hashes recorded by the
1064platform.
1065For more details see section 3.3 Table 6 of `DRTM`_ specification.
1066
1067Function : plat_drtm_get_dlme_img_auth_features()
1068~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1069
1070::
1071
1072    Argument : void
1073    Return   : uint64_t
1074
1075This function returns the DLME image authentication features.
1076For more details see section 3.3 Table 6 of `DRTM`_ specification.
1077
1078Function : plat_drtm_validate_ns_region()
1079~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1080
1081::
1082
1083    Argument : uintptr_t, uintptr_t
1084    Return   : int
1085
1086This function validates that given region is within the Non-Secure region
1087of DRAM. This function takes a region start address and size an input
1088arguments, and returns 0 on success and -1 on failure.
1089
1090Function : plat_set_drtm_error()
1091~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1092
1093::
1094
1095    Argument : uint64_t
1096    Return   : int
1097
1098This function writes a 64 bit error code received as input into
1099non-volatile storage and returns 0 on success and -1 on failure.
1100
1101Function : plat_get_drtm_error()
1102~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1103
1104::
1105
1106    Argument : uint64_t*
1107    Return   : int
1108
1109This function reads a 64 bit error code from the non-volatile storage
1110into the received address, and returns 0 on success and -1 on failure.
1111
1112Common mandatory function modifications
1113---------------------------------------
1114
1115The following functions are mandatory functions which need to be implemented
1116by the platform port.
1117
1118Function : plat_my_core_pos()
1119~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1120
1121::
1122
1123    Argument : void
1124    Return   : unsigned int
1125
1126This function returns the index of the calling CPU which is used as a
1127CPU-specific linear index into blocks of memory (for example while allocating
1128per-CPU stacks). This function will be invoked very early in the
1129initialization sequence which mandates that this function should be
1130implemented in assembly and should not rely on the availability of a C
1131runtime environment. This function can clobber x0 - x8 and must preserve
1132x9 - x29.
1133
1134This function plays a crucial role in the power domain topology framework in
1135PSCI and details of this can be found in
1136:ref:`PSCI Power Domain Tree Structure`.
1137
1138Function : plat_core_pos_by_mpidr()
1139~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1140
1141::
1142
1143    Argument : u_register_t
1144    Return   : int
1145
1146This function validates the ``MPIDR`` of a CPU and converts it to an index,
1147which can be used as a CPU-specific linear index into blocks of memory. In
1148case the ``MPIDR`` is invalid, this function returns -1. This function will only
1149be invoked by BL31 after the power domain topology is initialized and can
1150utilize the C runtime environment. For further details about how TF-A
1151represents the power domain topology and how this relates to the linear CPU
1152index, please refer :ref:`PSCI Power Domain Tree Structure`.
1153
1154Function : plat_get_mbedtls_heap() [when TRUSTED_BOARD_BOOT == 1]
1155~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1156
1157::
1158
1159    Arguments : void **heap_addr, size_t *heap_size
1160    Return    : int
1161
1162This function is invoked during Mbed TLS library initialisation to get a heap,
1163by means of a starting address and a size. This heap will then be used
1164internally by the Mbed TLS library. Hence, each BL stage that utilises Mbed TLS
1165must be able to provide a heap to it.
1166
1167A helper function can be found in `drivers/auth/mbedtls/mbedtls_common.c` in
1168which a heap is statically reserved during compile time inside every image
1169(i.e. every BL stage) that utilises Mbed TLS. In this default implementation,
1170the function simply returns the address and size of this "pre-allocated" heap.
1171For a platform to use this default implementation, only a call to the helper
1172from inside plat_get_mbedtls_heap() body is enough and nothing else is needed.
1173
1174However, by writting their own implementation, platforms have the potential to
1175optimise memory usage. For example, on some Arm platforms, the Mbed TLS heap is
1176shared between BL1 and BL2 stages and, thus, the necessary space is not reserved
1177twice.
1178
1179On success the function should return 0 and a negative error code otherwise.
1180
1181Function : plat_get_enc_key_info() [when FW_ENC_STATUS == 0 or 1]
1182~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1183
1184::
1185
1186    Arguments : enum fw_enc_status_t fw_enc_status, uint8_t *key,
1187                size_t *key_len, unsigned int *flags, const uint8_t *img_id,
1188                size_t img_id_len
1189    Return    : int
1190
1191This function provides a symmetric key (either SSK or BSSK depending on
1192fw_enc_status) which is invoked during runtime decryption of encrypted
1193firmware images. `plat/common/plat_bl_common.c` provides a dummy weak
1194implementation for testing purposes which must be overridden by the platform
1195trying to implement a real world firmware encryption use-case.
1196
1197It also allows the platform to pass symmetric key identifier rather than
1198actual symmetric key which is useful in cases where the crypto backend provides
1199secure storage for the symmetric key. So in this case ``ENC_KEY_IS_IDENTIFIER``
1200flag must be set in ``flags``.
1201
1202In addition to above a platform may also choose to provide an image specific
1203symmetric key/identifier using img_id.
1204
1205On success the function should return 0 and a negative error code otherwise.
1206
1207Note that this API depends on ``DECRYPTION_SUPPORT`` build flag.
1208
1209Function : plat_fwu_set_images_source() [when PSA_FWU_SUPPORT == 1]
1210~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1211
1212::
1213
1214    Argument : const struct fwu_metadata *metadata
1215    Return   : void
1216
1217This function is mandatory when PSA_FWU_SUPPORT is enabled.
1218It provides a means to retrieve image specification (offset in
1219non-volatile storage and length) of active/updated images using the passed
1220FWU metadata, and update I/O policies of active/updated images using retrieved
1221image specification information.
1222Further I/O layer operations such as I/O open, I/O read, etc. on these
1223images rely on this function call.
1224
1225In Arm platforms, this function is used to set an I/O policy of the FIP image,
1226container of all active/updated secure and non-secure images.
1227
1228Function : plat_fwu_set_metadata_image_source() [when PSA_FWU_SUPPORT == 1]
1229~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1230
1231::
1232
1233    Argument : unsigned int image_id, uintptr_t *dev_handle,
1234               uintptr_t *image_spec
1235    Return   : int
1236
1237This function is mandatory when PSA_FWU_SUPPORT is enabled. It is
1238responsible for setting up the platform I/O policy of the requested metadata
1239image (either FWU_METADATA_IMAGE_ID or BKUP_FWU_METADATA_IMAGE_ID) that will
1240be used to load this image from the platform's non-volatile storage.
1241
1242FWU metadata can not be always stored as a raw image in non-volatile storage
1243to define its image specification (offset in non-volatile storage and length)
1244statically in I/O policy.
1245For example, the FWU metadata image is stored as a partition inside the GUID
1246partition table image. Its specification is defined in the partition table
1247that needs to be parsed dynamically.
1248This function provides a means to retrieve such dynamic information to set
1249the I/O policy of the FWU metadata image.
1250Further I/O layer operations such as I/O open, I/O read, etc. on FWU metadata
1251image relies on this function call.
1252
1253It returns '0' on success, otherwise a negative error value on error.
1254Alongside, returns device handle and image specification from the I/O policy
1255of the requested FWU metadata image.
1256
1257Function : plat_fwu_get_boot_idx() [when PSA_FWU_SUPPORT == 1]
1258~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1259
1260::
1261
1262    Argument : void
1263    Return   : uint32_t
1264
1265This function is mandatory when PSA_FWU_SUPPORT is enabled. It provides the
1266means to retrieve the boot index value from the platform. The boot index is the
1267bank from which the platform has booted the firmware images.
1268
1269By default, the platform will read the metadata structure and try to boot from
1270the active bank. If the platform fails to boot from the active bank due to
1271reasons like an Authentication failure, or on crossing a set number of watchdog
1272resets while booting from the active bank, the platform can then switch to boot
1273from a different bank. This function then returns the bank that the platform
1274should boot its images from.
1275
1276Common optional modifications
1277-----------------------------
1278
1279The following are helper functions implemented by the firmware that perform
1280common platform-specific tasks. A platform may choose to override these
1281definitions.
1282
1283Function : plat_set_my_stack()
1284~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1285
1286::
1287
1288    Argument : void
1289    Return   : void
1290
1291This function sets the current stack pointer to the normal memory stack that
1292has been allocated for the current CPU. For BL images that only require a
1293stack for the primary CPU, the UP version of the function is used. The size
1294of the stack allocated to each CPU is specified by the platform defined
1295constant ``PLATFORM_STACK_SIZE``.
1296
1297Common implementations of this function for the UP and MP BL images are
1298provided in ``plat/common/aarch64/platform_up_stack.S`` and
1299``plat/common/aarch64/platform_mp_stack.S``
1300
1301Function : plat_get_my_stack()
1302~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1303
1304::
1305
1306    Argument : void
1307    Return   : uintptr_t
1308
1309This function returns the base address of the normal memory stack that
1310has been allocated for the current CPU. For BL images that only require a
1311stack for the primary CPU, the UP version of the function is used. The size
1312of the stack allocated to each CPU is specified by the platform defined
1313constant ``PLATFORM_STACK_SIZE``.
1314
1315Common implementations of this function for the UP and MP BL images are
1316provided in ``plat/common/aarch64/platform_up_stack.S`` and
1317``plat/common/aarch64/platform_mp_stack.S``
1318
1319Function : plat_report_exception()
1320~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1321
1322::
1323
1324    Argument : unsigned int
1325    Return   : void
1326
1327A platform may need to report various information about its status when an
1328exception is taken, for example the current exception level, the CPU security
1329state (secure/non-secure), the exception type, and so on. This function is
1330called in the following circumstances:
1331
1332-  In BL1, whenever an exception is taken.
1333-  In BL2, whenever an exception is taken.
1334
1335The default implementation doesn't do anything, to avoid making assumptions
1336about the way the platform displays its status information.
1337
1338For AArch64, this function receives the exception type as its argument.
1339Possible values for exceptions types are listed in the
1340``include/common/bl_common.h`` header file. Note that these constants are not
1341related to any architectural exception code; they are just a TF-A convention.
1342
1343For AArch32, this function receives the exception mode as its argument.
1344Possible values for exception modes are listed in the
1345``include/lib/aarch32/arch.h`` header file.
1346
1347Function : plat_reset_handler()
1348~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1349
1350::
1351
1352    Argument : void
1353    Return   : void
1354
1355A platform may need to do additional initialization after reset. This function
1356allows the platform to do the platform specific initializations. Platform
1357specific errata workarounds could also be implemented here. The API should
1358preserve the values of callee saved registers x19 to x29.
1359
1360The default implementation doesn't do anything. If a platform needs to override
1361the default implementation, refer to the :ref:`Firmware Design` for general
1362guidelines.
1363
1364Function : plat_disable_acp()
1365~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1366
1367::
1368
1369    Argument : void
1370    Return   : void
1371
1372This API allows a platform to disable the Accelerator Coherency Port (if
1373present) during a cluster power down sequence. The default weak implementation
1374doesn't do anything. Since this API is called during the power down sequence,
1375it has restrictions for stack usage and it can use the registers x0 - x17 as
1376scratch registers. It should preserve the value in x18 register as it is used
1377by the caller to store the return address.
1378
1379Function : plat_error_handler()
1380~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1381
1382::
1383
1384    Argument : int
1385    Return   : void
1386
1387This API is called when the generic code encounters an error situation from
1388which it cannot continue. It allows the platform to perform error reporting or
1389recovery actions (for example, reset the system). This function must not return.
1390
1391The parameter indicates the type of error using standard codes from ``errno.h``.
1392Possible errors reported by the generic code are:
1393
1394-  ``-EAUTH``: a certificate or image could not be authenticated (when Trusted
1395   Board Boot is enabled)
1396-  ``-ENOENT``: the requested image or certificate could not be found or an IO
1397   error was detected
1398-  ``-ENOMEM``: resources exhausted. TF-A does not use dynamic memory, so this
1399   error is usually an indication of an incorrect array size
1400
1401The default implementation simply spins.
1402
1403Function : plat_panic_handler()
1404~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1405
1406::
1407
1408    Argument : void
1409    Return   : void
1410
1411This API is called when the generic code encounters an unexpected error
1412situation from which it cannot recover. This function must not return,
1413and must be implemented in assembly because it may be called before the C
1414environment is initialized.
1415
1416.. note::
1417   The address from where it was called is stored in x30 (Link Register).
1418   The default implementation simply spins.
1419
1420Function : plat_get_bl_image_load_info()
1421~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1422
1423::
1424
1425    Argument : void
1426    Return   : bl_load_info_t *
1427
1428This function returns pointer to the list of images that the platform has
1429populated to load. This function is invoked in BL2 to load the
1430BL3xx images.
1431
1432Function : plat_get_next_bl_params()
1433~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1434
1435::
1436
1437    Argument : void
1438    Return   : bl_params_t *
1439
1440This function returns a pointer to the shared memory that the platform has
1441kept aside to pass TF-A related information that next BL image needs. This
1442function is invoked in BL2 to pass this information to the next BL
1443image.
1444
1445Function : plat_get_stack_protector_canary()
1446~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1447
1448::
1449
1450    Argument : void
1451    Return   : u_register_t
1452
1453This function returns a random value that is used to initialize the canary used
1454when the stack protector is enabled with ENABLE_STACK_PROTECTOR. A predictable
1455value will weaken the protection as the attacker could easily write the right
1456value as part of the attack most of the time. Therefore, it should return a
1457true random number.
1458
1459.. warning::
1460   For the protection to be effective, the global data need to be placed at
1461   a lower address than the stack bases. Failure to do so would allow an
1462   attacker to overwrite the canary as part of the stack buffer overflow attack.
1463
1464Function : plat_flush_next_bl_params()
1465~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1466
1467::
1468
1469    Argument : void
1470    Return   : void
1471
1472This function flushes to main memory all the image params that are passed to
1473next image. This function is invoked in BL2 to flush this information
1474to the next BL image.
1475
1476Function : plat_log_get_prefix()
1477~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1478
1479::
1480
1481    Argument : unsigned int
1482    Return   : const char *
1483
1484This function defines the prefix string corresponding to the `log_level` to be
1485prepended to all the log output from TF-A. The `log_level` (argument) will
1486correspond to one of the standard log levels defined in debug.h. The platform
1487can override the common implementation to define a different prefix string for
1488the log output. The implementation should be robust to future changes that
1489increase the number of log levels.
1490
1491Function : plat_get_soc_version()
1492~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1493
1494::
1495
1496    Argument : void
1497    Return   : int32_t
1498
1499This function returns soc version which mainly consist of below fields
1500
1501::
1502
1503    soc_version[30:24] = JEP-106 continuation code for the SiP
1504    soc_version[23:16] = JEP-106 identification code with parity bit for the SiP
1505    soc_version[15:0]  = Implementation defined SoC ID
1506
1507Function : plat_get_soc_revision()
1508~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1509
1510::
1511
1512    Argument : void
1513    Return   : int32_t
1514
1515This function returns soc revision in below format
1516
1517::
1518
1519    soc_revision[0:30] = SOC revision of specific SOC
1520
1521Function : plat_get_soc_name()
1522~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1523
1524::
1525
1526    Argument : char **
1527    Return   : int32_t
1528
1529The plat_get_soc_name() function allows a platform to expose the SoC name to
1530the firmware. It takes a pointer to a character pointer as an argument, which
1531must be set to point to a static, null-terminated SoC name string. The string
1532must be encoded in UTF-8 and should use only printable ASCII characters for
1533compatibility. It must not exceed 136 bytes, including the null terminator. On
1534success, the function returns SMC_ARCH_CALL_SUCCESS. If the platform does not
1535support SoC name retrieval, it returns SMC_ARCH_CALL_NOT_SUPPORTED. This API
1536allows platforms to support SoC name queries via SMCCC_ARCH_SOC_ID.
1537
1538Function : plat_is_smccc_feature_available()
1539~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1540
1541::
1542
1543    Argument : u_register_t
1544    Return   : int32_t
1545
1546This function returns SMC_ARCH_CALL_SUCCESS if the platform supports
1547the SMCCC function specified in the argument; otherwise returns
1548SMC_ARCH_CALL_NOT_SUPPORTED.
1549
1550Function : plat_can_cmo()
1551~~~~~~~~~~~~~~~~~~~~~~~~~
1552
1553::
1554
1555    Argument : void
1556    Return   : uint64_t
1557
1558When CONDITIONAL_CMO flag is enabled:
1559
1560- This function indicates whether cache management operations should be
1561  performed. It returns 0 if CMOs should be skipped and non-zero
1562  otherwise.
1563- The function must not clobber x1, x2 and x3. It's also not safe to rely on
1564  stack. Otherwise obey AAPCS.
1565
1566Struct: plat_try_images_ops [optional]
1567~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1568
1569This optional structure holds platform hooks for alternative images load.
1570It has to be defined in platform code and registered by calling
1571plat_setup_try_img_ops() function, passing it the address of the
1572plat_try_images_ops struct.
1573
1574Function : plat_setup_try_img_ops [optional]
1575............................................
1576
1577::
1578
1579    Argument : const struct plat_try_images_ops *
1580    Return   : void
1581
1582This optional function is called to register platform try images ops, given
1583as argument.
1584
1585Function : plat_try_images_ops.next_instance [optional]
1586.......................................................
1587
1588::
1589
1590    Argument : unsigned int image_id
1591    Return   : int
1592
1593This optional function tries to load images from alternative places.
1594In case PSA FWU is not used, it can be any instance or media. If PSA FWU is
1595used, it is mandatory that the backup image is on the same media.
1596This is required for MTD devices like NAND.
1597The argument is the ID of the image for which we are looking for an alternative
1598place. It returns 0 in case of success and a negative errno value otherwise.
1599
1600Function : plat_setup_log_gpt_corrupted [optional]
1601..................................................
1602
1603::
1604
1605    Argument : const struct plat_log_gpt_corrupted *
1606    Return   : void
1607
1608This optional function is called to register platform log GPT corrupted functions,
1609given as argument.
1610
1611Function : plat_setup_log_gpt_corrupted.plat_set_gpt_corruption [optional]
1612..........................................................................
1613
1614::
1615
1616    Argument : uintptr_t gpt_corrupted_info_ptr, uint8_t flags
1617    Return   : void
1618
1619This optional function will log error information if the GPT is corrupted, by
1620setting the address passed by gpt_corrupted_info_ptr to flags value, currently
1621bit[0] is used for logging primary GPT corruption, bit[7:1] are reserved.
1622The data type passed by reference is uint8_t.
1623
1624Function : plat_setup_log_gpt_corrupted.plat_log_gpt_corruption [optional]
1625..........................................................................
1626
1627::
1628
1629    Argument : uintptr_t log_address, uint8_t gpt_corrupted_info
1630    Return   : void
1631
1632This optional function will log if the primary GPT is corrupted, by writing
1633the value of gpt_corrupted_info to the address passed by log_address.
1634
1635Modifications specific to a Boot Loader stage
1636---------------------------------------------
1637
1638Boot Loader Stage 1 (BL1)
1639-------------------------
1640
1641BL1 implements the reset vector where execution starts from after a cold or
1642warm boot. For each CPU, BL1 is responsible for the following tasks:
1643
1644#. Handling the reset as described in section 2.2
1645
1646#. In the case of a cold boot and the CPU being the primary CPU, ensuring that
1647   only this CPU executes the remaining BL1 code, including loading and passing
1648   control to the BL2 stage.
1649
1650#. Identifying and starting the Firmware Update process (if required).
1651
1652#. Loading the BL2 image from non-volatile storage into secure memory at the
1653   address specified by the platform defined constant ``BL2_BASE``.
1654
1655#. Populating a ``meminfo`` structure with the following information in memory,
1656   accessible by BL2 immediately upon entry.
1657
1658   ::
1659
1660       meminfo.total_base = Base address of secure RAM visible to BL2
1661       meminfo.total_size = Size of secure RAM visible to BL2
1662
1663   By default, BL1 places this ``meminfo`` structure at the end of secure
1664   memory visible to BL2.
1665
1666   It is possible for the platform to decide where it wants to place the
1667   ``meminfo`` structure for BL2 or restrict the amount of memory visible to
1668   BL2 by overriding the weak default implementation of
1669   ``bl1_plat_handle_post_image_load`` API.
1670
1671The following functions need to be implemented by the platform port to enable
1672BL1 to perform the above tasks.
1673
1674Function : bl1_early_platform_setup() [mandatory]
1675~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1676
1677::
1678
1679    Argument : void
1680    Return   : void
1681
1682This function executes with the MMU and data caches disabled. It is only called
1683by the primary CPU.
1684
1685On Arm standard platforms, this function:
1686
1687-  Enables a secure instance of SP805 to act as the Trusted Watchdog.
1688
1689-  Initializes a UART (PL011 console), which enables access to the ``printf``
1690   family of functions in BL1.
1691
1692-  Enables issuing of snoop and DVM (Distributed Virtual Memory) requests to
1693   the CCI slave interface corresponding to the cluster that includes the
1694   primary CPU.
1695
1696Function : bl1_plat_arch_setup() [mandatory]
1697~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1698
1699::
1700
1701    Argument : void
1702    Return   : void
1703
1704This function performs any platform-specific and architectural setup that the
1705platform requires. Platform-specific setup might include configuration of
1706memory controllers and the interconnect.
1707
1708In Arm standard platforms, this function enables the MMU.
1709
1710This function helps fulfill requirement 2 above.
1711
1712Function : bl1_platform_setup() [mandatory]
1713~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1714
1715::
1716
1717    Argument : void
1718    Return   : void
1719
1720This function executes with the MMU and data caches enabled. It is responsible
1721for performing any remaining platform-specific setup that can occur after the
1722MMU and data cache have been enabled.
1723
1724In Arm standard platforms, this function initializes the storage abstraction
1725layer used to load the next bootloader image.
1726
1727This function helps fulfill requirement 4 above.
1728
1729Function : bl1_plat_sec_mem_layout() [mandatory]
1730~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1731
1732::
1733
1734    Argument : void
1735    Return   : meminfo *
1736
1737This function should only be called on the cold boot path. It executes with the
1738MMU and data caches enabled. The pointer returned by this function must point to
1739a ``meminfo`` structure containing the extents and availability of secure RAM for
1740the BL1 stage.
1741
1742::
1743
1744    meminfo.total_base = Base address of secure RAM visible to BL1
1745    meminfo.total_size = Size of secure RAM visible to BL1
1746
1747This information is used by BL1 to load the BL2 image in secure RAM. BL1 also
1748populates a similar structure to tell BL2 the extents of memory available for
1749its own use.
1750
1751This function helps fulfill requirements 4 and 5 above.
1752
1753Function : bl1_plat_prepare_exit() [optional]
1754~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1755
1756::
1757
1758    Argument : entry_point_info_t *
1759    Return   : void
1760
1761This function is called prior to exiting BL1 in response to the
1762``BL1_SMC_RUN_IMAGE`` SMC request raised by BL2. It should be used to perform
1763platform specific clean up or bookkeeping operations before transferring
1764control to the next image. It receives the address of the ``entry_point_info_t``
1765structure passed from BL2. This function runs with MMU disabled.
1766
1767Function : bl1_plat_set_ep_info() [optional]
1768~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1769
1770::
1771
1772    Argument : unsigned int image_id, entry_point_info_t *ep_info
1773    Return   : void
1774
1775This function allows platforms to override ``ep_info`` for the given ``image_id``.
1776
1777The default implementation just returns.
1778
1779Function : bl1_plat_get_next_image_id() [optional]
1780~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1781
1782::
1783
1784    Argument : void
1785    Return   : unsigned int
1786
1787This and the following function must be overridden to enable the FWU feature.
1788
1789BL1 calls this function after platform setup to identify the next image to be
1790loaded and executed. If the platform returns ``BL2_IMAGE_ID`` then BL1 proceeds
1791with the normal boot sequence, which loads and executes BL2. If the platform
1792returns a different image id, BL1 assumes that Firmware Update is required.
1793
1794The default implementation always returns ``BL2_IMAGE_ID``. The Arm development
1795platforms override this function to detect if firmware update is required, and
1796if so, return the first image in the firmware update process.
1797
1798Function : bl1_plat_get_image_desc() [optional]
1799~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1800
1801::
1802
1803    Argument : unsigned int image_id
1804    Return   : image_desc_t *
1805
1806BL1 calls this function to get the image descriptor information ``image_desc_t``
1807for the provided ``image_id`` from the platform.
1808
1809The default implementation always returns a common BL2 image descriptor. Arm
1810standard platforms return an image descriptor corresponding to BL2 or one of
1811the firmware update images defined in the Trusted Board Boot Requirements
1812specification.
1813
1814Function : bl1_plat_handle_pre_image_load() [optional]
1815~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1816
1817::
1818
1819    Argument : unsigned int image_id
1820    Return   : int
1821
1822This function can be used by the platforms to update/use image information
1823corresponding to ``image_id``. This function is invoked in BL1, both in cold
1824boot and FWU code path, before loading the image.
1825
1826Function : bl1_plat_calc_bl2_layout() [optional]
1827~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1828
1829::
1830
1831    Argument : const meminfo_t *bl1_mem_layout, meminfo_t *bl2_mem_layout
1832    Return   : void
1833
1834This utility function calculates the memory layout of BL2, representing it in a
1835`meminfo_t` structure. The default implementation derives this layout from the
1836positioning of BL1’s RW data at the top of the memory layout.
1837
1838Function : bl1_plat_handle_post_image_load() [optional]
1839~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1840
1841::
1842
1843    Argument : unsigned int image_id
1844    Return   : int
1845
1846This function can be used by the platforms to update/use image information
1847corresponding to ``image_id``. This function is invoked in BL1, both in cold
1848boot and FWU code path, after loading and authenticating the image.
1849
1850The default weak implementation of this function calculates the amount of
1851Trusted SRAM that can be used by BL2 and allocates a ``meminfo_t``
1852structure at the beginning of this free memory and populates it. The address
1853of ``meminfo_t`` structure is updated in ``arg1`` of the entrypoint
1854information to BL2.
1855
1856Function : bl1_plat_fwu_done() [optional]
1857~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1858
1859::
1860
1861    Argument : unsigned int image_id, uintptr_t image_src,
1862               unsigned int image_size
1863    Return   : void
1864
1865BL1 calls this function when the FWU process is complete. It must not return.
1866The platform may override this function to take platform specific action, for
1867example to initiate the normal boot flow.
1868
1869The default implementation spins forever.
1870
1871Function : bl1_plat_mem_check() [mandatory]
1872~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1873
1874::
1875
1876    Argument : uintptr_t mem_base, unsigned int mem_size,
1877               unsigned int flags
1878    Return   : int
1879
1880BL1 calls this function while handling FWU related SMCs, more specifically when
1881copying or authenticating an image. Its responsibility is to ensure that the
1882region of memory identified by ``mem_base`` and ``mem_size`` is mapped in BL1, and
1883that this memory corresponds to either a secure or non-secure memory region as
1884indicated by the security state of the ``flags`` argument.
1885
1886This function can safely assume that the value resulting from the addition of
1887``mem_base`` and ``mem_size`` fits into a ``uintptr_t`` type variable and does not
1888overflow.
1889
1890This function must return 0 on success, a non-null error code otherwise.
1891
1892The default implementation of this function asserts therefore platforms must
1893override it when using the FWU feature.
1894
1895Function : bl1_plat_is_shared_nv_ctr() [optional]
1896~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1897
1898::
1899
1900    Argument : void
1901    Return   : bool
1902
1903This function inquiry the platform if the non-volatile counter is shared
1904across all secure images (BL2, BL31, BL32, etc.). It is used only in BL1
1905and when `PSA_FWU_SUPPORT` is enabled.
1906
1907Boot Loader Stage 2 (BL2)
1908-------------------------
1909
1910The BL2 stage is executed only by the primary CPU, which is determined in BL1
1911using the ``platform_is_primary_cpu()`` function. BL1 passed control to BL2 at
1912``BL2_BASE``. BL2 executes in Secure EL1 and and invokes
1913``plat_get_bl_image_load_info()`` to retrieve the list of images to load from
1914non-volatile storage to secure/non-secure RAM. After all the images are loaded
1915then BL2 invokes ``plat_get_next_bl_params()`` to get the list of executable
1916images to be passed to the next BL image.
1917
1918The following functions must be implemented by the platform port to enable BL2
1919to perform the above tasks.
1920
1921Function : bl2_early_platform_setup2() [mandatory]
1922~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1923
1924::
1925
1926    Argument : u_register_t, u_register_t, u_register_t, u_register_t
1927    Return   : void
1928
1929This function executes with the MMU and data caches disabled. It is only called
1930by the primary CPU. The 4 arguments are passed by BL1 to BL2 and these arguments
1931are platform specific.
1932
1933On Arm standard platforms when RESET_TO_BL2 is not set, the arguments received
1934are :
1935
1936    arg0 - Points to load address of FW_CONFIG
1937
1938    arg1 - ``meminfo`` structure populated by BL1. The platform copies
1939    the contents of ``meminfo`` as it may be subsequently overwritten by BL2.
1940
1941On Arm standard platforms, this function also:
1942
1943-  Initializes a UART (PL011 console), which enables access to the ``printf``
1944   family of functions in BL2.
1945
1946-  Initializes the storage abstraction layer used to load further bootloader
1947   images. It is necessary to do this early on platforms with a SCP_BL2 image,
1948   since the later ``bl2_platform_setup`` must be done after SCP_BL2 is loaded.
1949
1950Function : bl2_plat_arch_setup() [mandatory]
1951~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1952
1953::
1954
1955    Argument : void
1956    Return   : void
1957
1958This function executes with the MMU and data caches disabled. It is only called
1959by the primary CPU.
1960
1961The purpose of this function is to perform any architectural initialization
1962that varies across platforms.
1963
1964On Arm standard platforms, this function enables the MMU.
1965
1966Function : bl2_platform_setup() [mandatory]
1967~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1968
1969::
1970
1971    Argument : void
1972    Return   : void
1973
1974This function may execute with the MMU and data caches enabled if the platform
1975port does the necessary initialization in ``bl2_plat_arch_setup()``. It is only
1976called by the primary CPU.
1977
1978The purpose of this function is to perform any platform initialization
1979specific to BL2.
1980
1981In Arm standard platforms, this function performs security setup, including
1982configuration of the TrustZone controller to allow non-secure masters access
1983to most of DRAM. Part of DRAM is reserved for secure world use.
1984
1985Function : bl2_plat_handle_pre_image_load() [optional]
1986~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1987
1988::
1989
1990    Argument : unsigned int
1991    Return   : int
1992
1993This function can be used by the platforms to update/use image information
1994for given ``image_id``. This function is currently invoked in BL2 before
1995loading each image.
1996
1997Function : bl2_plat_handle_post_image_load() [optional]
1998~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1999
2000::
2001
2002    Argument : unsigned int
2003    Return   : int
2004
2005This function can be used by the platforms to update/use image information
2006for given ``image_id``. This function is currently invoked in BL2 after
2007loading each image.
2008
2009Function : bl2_plat_preload_setup [optional]
2010~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2011
2012::
2013
2014    Argument : void
2015    Return   : void
2016
2017This optional function performs any BL2 platform initialization
2018required before image loading, that is not done later in
2019bl2_platform_setup().
2020
2021Boot Loader Stage 2 (BL2) at EL3
2022--------------------------------
2023
2024When the platform has a non-TF-A Boot ROM it is desirable to jump
2025directly to BL2 instead of TF-A BL1. In this case BL2 is expected to
2026execute at EL3 instead of executing at EL1. Also on RME enabled systems BL2 runs
2027at EL3. Refer to the :ref:`Firmware Design` document for more information.
2028
2029The existing mandatory functions of BL2 must handle BL2 running at EL3.
2030
2031FWU Boot Loader Stage 2 (BL2U)
2032------------------------------
2033
2034The AP Firmware Updater Configuration, BL2U, is an optional part of the FWU
2035process and is executed only by the primary CPU. BL1 passes control to BL2U at
2036``BL2U_BASE``. BL2U executes in Secure-EL1 and is responsible for:
2037
2038#. (Optional) Transferring the optional SCP_BL2U binary image from AP secure
2039   memory to SCP RAM. BL2U uses the SCP_BL2U ``image_info`` passed by BL1.
2040   ``SCP_BL2U_BASE`` defines the address in AP secure memory where SCP_BL2U
2041   should be copied from. Subsequent handling of the SCP_BL2U image is
2042   implemented by the platform specific ``bl2u_plat_handle_scp_bl2u()`` function.
2043   If ``SCP_BL2U_BASE`` is not defined then this step is not performed.
2044
2045#. Any platform specific setup required to perform the FWU process. For
2046   example, Arm standard platforms initialize the TZC controller so that the
2047   normal world can access DDR memory.
2048
2049The following functions must be implemented by the platform port to enable
2050BL2U to perform the tasks mentioned above.
2051
2052Function : bl2u_early_platform_setup() [mandatory]
2053~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2054
2055::
2056
2057    Argument : meminfo *mem_info, void *plat_info
2058    Return   : void
2059
2060This function executes with the MMU and data caches disabled. It is only
2061called by the primary CPU. The arguments to this function is the address
2062of the ``meminfo`` structure and platform specific info provided by BL1.
2063
2064The platform may copy the contents of the ``mem_info`` and ``plat_info`` into
2065private storage as the original memory may be subsequently overwritten by BL2U.
2066
2067On Arm CSS platforms ``plat_info`` is interpreted as an ``image_info_t`` structure,
2068to extract SCP_BL2U image information, which is then copied into a private
2069variable.
2070
2071Function : bl2u_plat_arch_setup() [mandatory]
2072~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2073
2074::
2075
2076    Argument : void
2077    Return   : void
2078
2079This function executes with the MMU and data caches disabled. It is only
2080called by the primary CPU.
2081
2082The purpose of this function is to perform any architectural initialization
2083that varies across platforms, for example enabling the MMU (since the memory
2084map differs across platforms).
2085
2086Function : bl2u_platform_setup() [mandatory]
2087~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2088
2089::
2090
2091    Argument : void
2092    Return   : void
2093
2094This function may execute with the MMU and data caches enabled if the platform
2095port does the necessary initialization in ``bl2u_plat_arch_setup()``. It is only
2096called by the primary CPU.
2097
2098The purpose of this function is to perform any platform initialization
2099specific to BL2U.
2100
2101In Arm standard platforms, this function performs security setup, including
2102configuration of the TrustZone controller to allow non-secure masters access
2103to most of DRAM. Part of DRAM is reserved for secure world use.
2104
2105Function : bl2u_plat_handle_scp_bl2u() [optional]
2106~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2107
2108::
2109
2110    Argument : void
2111    Return   : int
2112
2113This function is used to perform any platform-specific actions required to
2114handle the SCP firmware. Typically it transfers the image into SCP memory using
2115a platform-specific protocol and waits until SCP executes it and signals to the
2116Application Processor (AP) for BL2U execution to continue.
2117
2118This function returns 0 on success, a negative error code otherwise.
2119This function is included if SCP_BL2U_BASE is defined.
2120
2121Boot Loader Stage 3-1 (BL31)
2122----------------------------
2123
2124During cold boot, the BL31 stage is executed only by the primary CPU. This is
2125determined in BL1 using the ``platform_is_primary_cpu()`` function. BL1 passes
2126control to BL31 at ``BL31_BASE``. During warm boot, BL31 is executed by all
2127CPUs. BL31 executes at EL3 and is responsible for:
2128
2129#. Re-initializing all architectural and platform state. Although BL1 performs
2130   some of this initialization, BL31 remains resident in EL3 and must ensure
2131   that EL3 architectural and platform state is completely initialized. It
2132   should make no assumptions about the system state when it receives control.
2133
2134#. Passing control to a normal world BL image, pre-loaded at a platform-
2135   specific address by BL2. On ARM platforms, BL31 uses the ``bl_params`` list
2136   populated by BL2 in memory to do this.
2137
2138#. Providing runtime firmware services. Currently, BL31 only implements a
2139   subset of the Power State Coordination Interface (PSCI) API as a runtime
2140   service. See :ref:`psci_in_bl31` below for details of porting the PSCI
2141   implementation.
2142
2143#. Optionally passing control to the BL32 image, pre-loaded at a platform-
2144   specific address by BL2. BL31 exports a set of APIs that allow runtime
2145   services to specify the security state in which the next image should be
2146   executed and run the corresponding image. On ARM platforms, BL31 uses the
2147   ``bl_params`` list populated by BL2 in memory to do this.
2148
2149If BL31 is a reset vector, It also needs to handle the reset as specified in
2150section 2.2 before the tasks described above.
2151
2152The following functions must be implemented by the platform port to enable BL31
2153to perform the above tasks.
2154
2155Function : bl31_early_platform_setup2() [mandatory]
2156~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2157
2158::
2159
2160    Argument : u_register_t, u_register_t, u_register_t, u_register_t
2161    Return   : void
2162
2163This function executes with the MMU and data caches disabled. It is only called
2164by the primary CPU. BL2 can pass 4 arguments to BL31 and these arguments are
2165platform specific.
2166
2167In Arm standard platforms, the arguments received are :
2168
2169    arg0 - The pointer to the head of `bl_params_t` list
2170    which is list of executable images following BL31,
2171
2172    arg1 - Points to load address of SOC_FW_CONFIG if present
2173           except in case of Arm FVP and Juno platform.
2174
2175           In case of Arm FVP and Juno platform, points to load address
2176           of FW_CONFIG.
2177
2178    arg2 - Points to load address of HW_CONFIG if present
2179
2180    arg3 - A special value to verify platform parameters from BL2 to BL31. Not
2181    used in release builds.
2182
2183The function runs through the `bl_param_t` list and extracts the entry point
2184information for BL32 and BL33. It also performs the following:
2185
2186-  Initialize a UART (PL011 console), which enables access to the ``printf``
2187   family of functions in BL31.
2188
2189-  Enable issuing of snoop and DVM (Distributed Virtual Memory) requests to the
2190   CCI slave interface corresponding to the cluster that includes the primary
2191   CPU.
2192
2193Function : bl31_plat_arch_setup() [mandatory]
2194~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2195
2196::
2197
2198    Argument : void
2199    Return   : void
2200
2201This function executes with the MMU and data caches disabled. It is only called
2202by the primary CPU.
2203
2204The purpose of this function is to perform any architectural initialization
2205that varies across platforms.
2206
2207On Arm standard platforms, this function enables the MMU.
2208
2209Function : bl31_platform_setup() [mandatory]
2210~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2211
2212::
2213
2214    Argument : void
2215    Return   : void
2216
2217This function may execute with the MMU and data caches enabled if the platform
2218port does the necessary initialization in ``bl31_plat_arch_setup()``. It is only
2219called by the primary CPU.
2220
2221The purpose of this function is to complete platform initialization so that both
2222BL31 runtime services and normal world software can function correctly.
2223
2224On Arm standard platforms, this function does the following:
2225
2226-  Initialize the generic interrupt controller.
2227
2228   Depending on the GIC driver selected by the platform, the appropriate GICv2
2229   or GICv3 initialization will be done, which mainly consists of:
2230
2231   -  Enable secure interrupts in the GIC CPU interface.
2232   -  Disable the legacy interrupt bypass mechanism.
2233   -  Configure the priority mask register to allow interrupts of all priorities
2234      to be signaled to the CPU interface.
2235   -  Mark SGIs 8-15 and the other secure interrupts on the platform as secure.
2236   -  Target all secure SPIs to CPU0.
2237   -  Enable these secure interrupts in the GIC distributor.
2238   -  Configure all other interrupts as non-secure.
2239   -  Enable signaling of secure interrupts in the GIC distributor.
2240
2241-  Enable system-level implementation of the generic timer counter through the
2242   memory mapped interface.
2243
2244-  Grant access to the system counter timer module
2245
2246-  Initialize the power controller device.
2247
2248   In particular, initialise the locks that prevent concurrent accesses to the
2249   power controller device.
2250
2251Function : bl31_plat_runtime_setup() [optional]
2252~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2253
2254::
2255
2256    Argument : void
2257    Return   : void
2258
2259The purpose of this function is to allow the platform to perform any BL31 runtime
2260setup just prior to BL31 exit during cold boot. The default weak implementation
2261of this function is empty. Any platform that needs to perform additional runtime
2262setup, before BL31 exits, will need to override this function.
2263
2264Function : bl31_plat_get_next_image_ep_info() [mandatory]
2265~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2266
2267::
2268
2269    Argument : uint32_t
2270    Return   : entry_point_info *
2271
2272This function may execute with the MMU and data caches enabled if the platform
2273port does the necessary initializations in ``bl31_plat_arch_setup()``.
2274
2275This function is called by ``bl31_main()`` to retrieve information provided by
2276BL2 for the next image in the security state specified by the argument. BL31
2277uses this information to pass control to that image in the specified security
2278state. This function must return a pointer to the ``entry_point_info`` structure
2279(that was copied during ``bl31_early_platform_setup()``) if the image exists. It
2280should return NULL otherwise.
2281
2282Function : plat_rmmd_get_cca_attest_token() [mandatory when ENABLE_RMM == 1]
2283~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2284
2285::
2286
2287    Argument : uintptr_t, size_t *, uintptr_t, size_t, size_t *
2288    Return   : int
2289
2290This function returns the Platform attestation token. If the full token does
2291not fit in the buffer, the function will return a hunk of the token and
2292indicate how many bytes were copied and how many are pending. Multiple calls
2293to this function may be needed to retrieve the entire token.
2294
2295The parameters of the function are:
2296
2297    arg0 - A pointer to the buffer where the Platform token should be copied by
2298           this function. If the platform token does not completely fit in the
2299           buffer, the function may return a piece of the token only.
2300
2301    arg1 - Contains the size (in bytes) of the buffer passed in arg0. In
2302           addition, this parameter is used by the function to return the size
2303           of the platform token length hunk copied to the buffer.
2304
2305    arg2 - A pointer to the buffer where the challenge object is stored.
2306
2307    arg3 - The length of the challenge object in bytes. Possible values are 32,
2308           48 and 64. This argument must be zero for subsequent calls to
2309           retrieve the remaining hunks of the token.
2310
2311    arg4 - Returns the remaining length of the token (in bytes) that is yet to
2312           be returned in further calls.
2313
2314The function returns 0 on success, -EINVAL on failure and -EAGAIN if the
2315resource associated with the platform token retrieval is busy.
2316
2317Function : plat_rmmd_get_cca_realm_attest_key() [mandatory when ENABLE_RMM == 1]
2318~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2319
2320::
2321
2322    Argument : uintptr_t, size_t *, unsigned int
2323    Return   : int
2324
2325This function returns the delegated realm attestation key which will be used to
2326sign Realm attestation token. The API currently only supports P-384 ECC curve
2327key.
2328
2329The parameters of the function are:
2330
2331    arg0 - A pointer to the buffer where the attestation key should be copied
2332           by this function. The buffer must be big enough to hold the
2333           attestation key.
2334
2335    arg1 - Contains the size (in bytes) of the buffer passed in arg0. The
2336           function returns the attestation key length in this parameter.
2337
2338    arg2 - The type of the elliptic curve to which the requested attestation key
2339           belongs.
2340
2341The function returns 0 on success, -EINVAL on failure.
2342
2343Function : plat_rmmd_get_el3_rmm_shared_mem() [when ENABLE_RMM == 1]
2344~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2345
2346::
2347
2348   Argument : uintptr_t *
2349   Return   : size_t
2350
2351This function returns the size of the shared area between EL3 and RMM (or 0 on
2352failure). A pointer to the shared area (or a NULL pointer on failure) is stored
2353in the pointer passed as argument.
2354
2355Function : plat_rmmd_load_manifest() [when ENABLE_RMM == 1]
2356~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2357
2358::
2359
2360    Arguments : rmm_manifest_t *manifest
2361    Return    : int
2362
2363When ENABLE_RMM is enabled, this function populates a boot manifest for the
2364RMM image and stores it in the area specified by manifest.
2365
2366When ENABLE_RMM is disabled, this function is not used.
2367
2368Function : plat_rmmd_mecid_key_update() [when ENABLE_RMM == 1]
2369~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2370
2371::
2372
2373    Argument : uint16_t, unsigned int
2374    Return   : int
2375
2376This function is invoked by BL31's RMMD when there is a request from the RMM
2377monitor to update the tweak for the encryption key associated to a MECID.
2378
2379The first parameter (``uint16_t mecid``) contains the MECID for which the
2380encryption key is to be updated. The second argument specifies the reason
2381for key update. Possible values are: 0 - Realm creation, 1 - Realm destruction.
2382
2383Return value is 0 upon success and -EFAULT otherwise.
2384
2385This function needs to be implemented by a platform if it enables RME.
2386
2387Function : plat_rmmd_reserve_memory() [when ENABLE_RMM == 1]
2388~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2389
2390::
2391
2392    Arguments : size_t size, unsigned long alignment
2393    Return    : uintptr_t
2394
2395Reserve memory to be used by the RMM. This could be memory simply taken from a pool of reserved
2396memory, for instance from a carveout dedicated to RMM.
2397
2398Return value is the physical address of a memory region of at least ``size`` bytes, which needs
2399to be aligned to ``alignment`` bytes.
2400
2401This function needs to be implemented if a platform enables RME and the RMM requires the memory
2402reservation feature.
2403
2404Function : plat_rmmd_el3_token_sign_push_req() [mandatory when RMMD_ENABLE_EL3_TOKEN_SIGN == 1]
2405~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2406
2407::
2408
2409    Arguments : const struct el3_token_sign_request *req
2410    Return    : int
2411
2412Queue realm attestation token signing request from the RMM in EL3. The interface between
2413the RMM and EL3 is modeled as a queue but the underlying implementation may be different,
2414so long as the semantics of queuing and the error codes are used as defined below.
2415
2416See :ref:`el3_token_sign_request_struct` for definition of the request structure.
2417
2418Optional interface from the RMM-EL3 interface v0.4 onwards.
2419
2420The parameters of the functions are:
2421      arg0: Pointer to the token sign request to be pushed to EL3.
2422      The structure must be located in the RMM-EL3 shared
2423      memory buffer and must be locked before use.
2424
2425Return codes:
2426        - E_RMM_OK	On Success.
2427        - E_RMM_INVAL   If the arguments are invalid.
2428        - E_RMM_AGAIN   Indicates that the request was not queued since the
2429	  queue in EL3 is full. This may also be returned for any reason
2430	  or situation in the system, that prevents accepting the request
2431	  from the RMM.
2432        - E_RMM_UNK     If the SMC is not implemented or if interface
2433	  version is < 0.4.
2434
2435Function : plat_rmmd_el3_token_sign_pull_resp() [mandatory when RMMD_ENABLE_EL3_TOKEN_SIGN == 1]
2436~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2437
2438::
2439
2440    Arguments : struct el3_token_sign_response *resp
2441    Return    : int
2442
2443Populate the attestation signing response in the ``resp`` parameter. The interface between
2444the RMM and EL3 is modeled as a queue for responses but the underlying implementation may
2445be different, so long as the semantics of queuing and the error codes are used as defined
2446below.
2447
2448See :ref:`el3_token_sign_response_struct` for definition of the response structure.
2449
2450Optional interface from the RMM-EL3 interface v0.4 onwards.
2451
2452The parameters of the functions are:
2453          resp: Pointer to the token sign response to get from EL3.
2454	  The structure must be located in the RMM-EL3 shared
2455	  memory buffer and must be locked before use.
2456
2457Return:
2458        - E_RMM_OK      On Success.
2459        - E_RMM_INVAL   If the arguments are invalid.
2460        - E_RMM_AGAIN   Indicates that a response is not ready yet.
2461        - E_RMM_UNK     If the SMC is not implemented or if interface
2462	  version is < 0.4.
2463
2464Function : plat_rmmd_el3_token_sign_get_rak_pub() [mandatory when RMMD_ENABLE_EL3_TOKEN_SIGN == 1]
2465~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2466
2467::
2468
2469    Argument : uintptr_t, size_t *, unsigned int
2470    Return   : int
2471
2472This function returns the public portion of the realm attestation key which will be used to
2473sign Realm attestation token. Typically, with delegated attestation, the private key is
2474returned, however, there may be platforms where the private key bits are better protected
2475in a platform specific manner such that the private key is not exposed. In such cases,
2476the RMM will only cache the public key and forward any requests such as signing, that
2477uses the private key to EL3. The API currently only supports P-384 ECC curve key.
2478
2479This is an optional interface from the RMM-EL3 interface v0.4 onwards.
2480
2481The parameters of the function are:
2482
2483    arg0 - A pointer to the buffer where the public key should be copied
2484    by this function. The buffer must be big enough to hold the
2485    attestation key.
2486
2487    arg1 - Contains the size (in bytes) of the buffer passed in arg0. The
2488    function returns the attestation key length in this parameter.
2489
2490    arg2 - The type of the elliptic curve to which the requested attestation key
2491    belongs.
2492
2493The function returns E_RMM_OK on success, RMM_E_INVAL if arguments are invalid and
2494E_RMM_UNK if the SMC is not implemented or if interface version is < 0.4.
2495
2496Function : plat_rmmd_el3_ide_key_program() [mandatory when RMMD_ENABLE_IDE_KEY_PROG == 1]
2497~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2498
2499::
2500
2501    Argument : uint64_t, uint64_t, uint64_t, struct rp_ide_key_info_t *, uint64_t, uint64_t
2502    Return   : int
2503
2504This function sets the key/IV info for an IDE stream at the Root port. The key is 256 bits
2505and IV is 96 bits. The caller calls this SMC to program this key to the Rx and Tx ports
2506and for each substream corresponding to a single keyset. The platform should validate
2507the arguments `Ecam address` and `Rootport ID` before acting on it. The arguments `request ID`
2508and `cookie` are to be ignored for blocking mode and are pass-through to the response for
2509non-blocking mode.
2510
2511The platform needs to ensure proper exclusives are in place when accessed from multiple CPUs.
2512Depending on the expected latency for IDE-KM interface, the platform should choose blocking
2513or non-blocking semantics. More details about IDE Setup flow can be found
2514in this `RFC <https://github.com/TF-RMM/tf-rmm/wiki/RFC:-EL3-RMM-IDE-KM-Interface>`_.
2515
2516The parameters of the function are:
2517
2518    arg0 - The ecam address, to access and configure PCI devices in a system.
2519
2520    arg1 - The rootport ID used to identify the PCIe rootport of a connected device.
2521
2522    arg2 - The IDE stream info associated with a physical device, this parameter packs the
2523    the keyset, direction, substream and stream ID info.
2524
2525    arg3 - Structure with key and IV info.
2526
2527    arg4 - The request ID, is used in non-blocking mode only and can be ignored in blocking mode.
2528
2529    arg5 - The cookie variable, is used in non-blocking mode only and can be ignored in blocking
2530    mode.
2531
2532The function returns E_RMM_OK on success, E_RMM_INVAL if arguments are invalid, E_RMM_FAULT
2533if the key programming is unsuccesful, E_RMM_UNK for an unknown error, E_RMM_AGAIN returned
2534only for non-blocking mode if the IDE-KM interface is busy or the request queue is full.
2535E_RMM_INPROGRESS returned if the request is queued successfully and used only in non-blocking
2536mode.
2537
2538Function : plat_rmmd_el3_ide_key_set_go() [mandatory when RMMD_ENABLE_IDE_KEY_PROG == 1]
2539~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2540
2541::
2542
2543    Argument : uint64_t, uint64_t, uint64_t, uint64_t, uint64_t
2544    Return   : int
2545
2546This function activates the IDE stream at the Root Port once all the keys have been
2547programmed. The platform should validate the arguments `Ecam address` and `Rootport ID`
2548before acting on it. The arguments `request ID` and `cookie` are to be ignored for blocking
2549mode and are pass-through to the response for non-blocking mode.
2550
2551The platform needs to ensure proper exclusives are in place when accessed from multiple CPUs.
2552Depending on the expected latency for IDE-KM interface, the platform should choose blocking
2553or non-blocking semantics. More details about IDE Setup flow can be found
2554in this `RFC <https://github.com/TF-RMM/tf-rmm/wiki/RFC:-EL3-RMM-IDE-KM-Interface>`_.
2555
2556The parameters of the function are:
2557
2558    arg0 - The ecam address, to access and configure PCI devices in a system.
2559
2560    arg1 - The rootport ID used to identify the PCIe rootport of a connected device.
2561
2562    arg2 - The IDE stream info associated with a physical device, this parameter packs the
2563    the keyset, direction, substream and stream ID info.
2564
2565    arg3 - The request ID, is used in non-blocking mode only and can be ignored in blocking mode.
2566
2567    arg4 - The cookie variable, is used in non-blocking mode only and can be ignored in blocking
2568    mode.
2569
2570The function returns E_RMM_OK on success, E_RMM_INVAL if arguments are invalid, E_RMM_FAULT
2571if the key programming is unsuccesful, E_RMM_UNK for an unknown error, E_RMM_AGAIN returned
2572only for non-blocking mode if the IDE-KM interface is busy or the request queue is full.
2573E_RMM_INPROGRESS returned if the request is queued successfully and used only in non-blocking
2574mode.
2575
2576Function : plat_rmmd_el3_ide_key_set_stop() [mandatory when RMMD_ENABLE_IDE_KEY_PROG == 1]
2577~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2578
2579::
2580
2581    Argument : uint64_t, uint64_t, uint64_t, uint64_t, uint64_t
2582    Return   : int
2583
2584This function stops the IDE stream and is used to tear down the IDE stream at Root Port.
2585The platform should validate the arguments `Ecam address` and `Rootport ID` before acting
2586on it. The arguments `request ID` and `cookie` are to be ignored for blocking
2587mode and are pass-through to the response for non-blocking mode.
2588
2589The platform needs to ensure proper exclusives are in place when accessed from multiple CPUs.
2590Depending on the expected latency for IDE-KM interface, the platform should choose blocking
2591or non-blocking semantics. More details about IDE Setup flow can be found
2592in this `RFC <https://github.com/TF-RMM/tf-rmm/wiki/RFC:-EL3-RMM-IDE-KM-Interface>`_.
2593
2594The parameters of the function are:
2595
2596    arg0 - The ecam address, to access and configure PCI devices in a system.
2597
2598    arg1 - The rootport ID used to identify the PCIe rootport of a connected device.
2599
2600    arg2 - The IDE stream info associated with a physical device, this parameter packs the
2601    the keyset, direction, substream and stream ID info.
2602
2603    arg3 - The request ID, is used in non-blocking mode only and can be ignored in blocking mode.
2604
2605    arg4 - The cookie variable, is used in non-blocking mode only and can be ignored in blocking
2606    mode.
2607
2608The function returns E_RMM_OK on success, E_RMM_INVAL if arguments are invalid, E_RMM_FAULT
2609if the key programming is unsuccesful, E_RMM_UNK for an unknown error, E_RMM_AGAIN returned
2610only for non-blocking mode if the IDE-KM interface is busy or the request queue is full.
2611E_RMM_INPROGRESS returned if the request is queued successfully and used only in non-blocking
2612mode.
2613
2614Function : plat_rmmd_el3_ide_km_pull_response() [mandatory when RMMD_ENABLE_IDE_KEY_PROG == 1]
2615~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2616
2617::
2618
2619    Argument : uint64_t, uint64_t, uint64_t *, uint64_t *, uint64_t *
2620    Return   : int
2621
2622This function retrieves a reponse for any of the prior non-blocking IDE-KM requests. The
2623caller has to identify the request and populate the accurate response. For blocking calls,
2624this function always returns E_RMM_UNK.
2625
2626The platform needs to ensure proper exclusives are in place when accessed from multiple CPUs.
2627Depending on the expected latency for IDE-KM interface, the platform should choose blocking
2628or non-blocking semantics. More details about IDE Setup flow can be found
2629in this `RFC <https://github.com/TF-RMM/tf-rmm/wiki/RFC:-EL3-RMM-IDE-KM-Interface>`_.
2630
2631The parameters of the function are:
2632
2633    arg0 - The ecam address, to access and configure PCI devices in a system.
2634
2635    arg1 - The rootport ID used to identify the PCIe rootport of a connected device.
2636
2637    arg2 - Retrieved response corresponding to the previous IDE_KM request.
2638
2639    arg3 - returns the passthrough request ID of the retrieved response.
2640
2641    arg4 - returns the passthrough cookie of the retrieved response.
2642
2643The function returns E_RMM_OK if response is retrieved successfully, E_RMM_INVAL if arguments
2644to this function are invalid, E_RMM_UNK if response retrieval failed for an unknown error or
2645IDE-KM interface is having blocking semantics, E_RMM_AGAIN if the response queue is empty.
2646
2647The `arg2` return parameter can return the following values:
2648E_RMM_OK - The previous request was successful.
2649E_RMM_FAULT - The previous request was not successful.
2650E_RMM_INVAL - Arguments to previous request were incorrect.
2651E_RMM_UNK - Previous request returned Unknown error.
2652
2653Function : bl31_plat_enable_mmu [optional]
2654~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2655
2656::
2657
2658    Argument : uint32_t
2659    Return   : void
2660
2661This function enables the MMU. The boot code calls this function with MMU and
2662caches disabled. This function should program necessary registers to enable
2663translation, and upon return, the MMU on the calling PE must be enabled.
2664
2665The function must honor flags passed in the first argument. These flags are
2666defined by the translation library, and can be found in the file
2667``include/lib/xlat_tables/xlat_mmu_helpers.h``.
2668
2669On DynamIQ systems, this function must not use stack while enabling MMU, which
2670is how the function in xlat table library version 2 is implemented.
2671
2672Function : plat_init_apkey [optional]
2673~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2674
2675::
2676
2677    Argument : void
2678    Return   : uint128_t
2679
2680This function returns the 128-bit value which can be used to program ARMv8.3
2681pointer authentication keys.
2682
2683The value should be obtained from a reliable source of randomness. It will be
2684called each time a core powers up and it is the platform's responsibility to
2685decide when to regenerate the keys if generating them is an expensive operation.
2686
2687This function is only needed if ARMv8.3 pointer authentication is used in the
2688Trusted Firmware by building with ``BRANCH_PROTECTION`` option set to 1, 2 or 3.
2689
2690Function : plat_get_syscnt_freq2() [mandatory]
2691~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2692
2693::
2694
2695    Argument : void
2696    Return   : unsigned int
2697
2698This function is used by the architecture setup code to retrieve the counter
2699frequency for the CPU's generic timer. This value will be programmed into the
2700``CNTFRQ_EL0`` register. In Arm standard platforms, it returns the base frequency
2701of the system counter, which is retrieved from the first entry in the frequency
2702modes table.
2703
2704#define : PLAT_PERCPU_BAKERY_LOCK_SIZE [optional]
2705~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2706
2707When ``USE_COHERENT_MEM = 0``, this constant defines the total memory (in
2708bytes) aligned to the cache line boundary that should be allocated per-cpu to
2709accommodate all the bakery locks.
2710
2711If this constant is not defined when ``USE_COHERENT_MEM = 0``, the linker
2712calculates the size of the ``.bakery_lock`` input section, aligns it to the
2713nearest ``CACHE_WRITEBACK_GRANULE``, multiplies it with ``PLATFORM_CORE_COUNT``
2714and stores the result in a linker symbol. This constant prevents a platform
2715from relying on the linker and provide a more efficient mechanism for
2716accessing per-cpu bakery lock information.
2717
2718If this constant is defined and its value is not equal to the value
2719calculated by the linker then a link time assertion is raised. A compile time
2720assertion is raised if the value of the constant is not aligned to the cache
2721line boundary.
2722
2723.. _porting_guide_sdei_requirements:
2724
2725SDEI porting requirements
2726~~~~~~~~~~~~~~~~~~~~~~~~~
2727
2728The |SDEI| dispatcher requires the platform to provide the following macros
2729and functions, of which some are optional, and some others mandatory.
2730
2731Macros
2732......
2733
2734Macro: PLAT_SDEI_NORMAL_PRI [mandatory]
2735^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
2736
2737This macro must be defined to the EL3 exception priority level associated with
2738Normal |SDEI| events on the platform. This must have a higher value
2739(therefore of lower priority) than ``PLAT_SDEI_CRITICAL_PRI``.
2740
2741Macro: PLAT_SDEI_CRITICAL_PRI [mandatory]
2742^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
2743
2744This macro must be defined to the EL3 exception priority level associated with
2745Critical |SDEI| events on the platform. This must have a lower value
2746(therefore of higher priority) than ``PLAT_SDEI_NORMAL_PRI``.
2747
2748**Note**: |SDEI| exception priorities must be the lowest among Secure
2749priorities. Among the |SDEI| exceptions, Critical |SDEI| priority must
2750be higher than Normal |SDEI| priority.
2751
2752Functions
2753.........
2754
2755Function: int plat_sdei_validate_entry_point() [optional]
2756^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
2757
2758::
2759
2760  Argument: uintptr_t ep, unsigned int client_mode
2761  Return: int
2762
2763This function validates the entry point address of the event handler provided by
2764the client for both event registration and *Complete and Resume* |SDEI| calls.
2765The function ensures that the address is valid in the client translation regime.
2766
2767The second argument is the exception level that the client is executing in. It
2768can be Non-Secure EL1 or Non-Secure EL2.
2769
2770The function must return ``0`` for successful validation, or ``-1`` upon failure.
2771
2772The default implementation always returns ``0``. On Arm platforms, this function
2773translates the entry point address within the client translation regime and
2774further ensures that the resulting physical address is located in Non-secure
2775DRAM.
2776
2777Function: void plat_sdei_handle_masked_trigger(uint64_t mpidr, unsigned int intr) [optional]
2778^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
2779
2780::
2781
2782  Argument: uint64_t
2783  Argument: unsigned int
2784  Return: void
2785
2786|SDEI| specification requires that a PE comes out of reset with the events
2787masked. The client therefore is expected to call ``PE_UNMASK`` to unmask
2788|SDEI| events on the PE. No |SDEI| events can be dispatched until such
2789time.
2790
2791Should a PE receive an interrupt that was bound to an |SDEI| event while the
2792events are masked on the PE, the dispatcher implementation invokes the function
2793``plat_sdei_handle_masked_trigger``. The MPIDR of the PE that received the
2794interrupt and the interrupt ID are passed as parameters.
2795
2796The default implementation only prints out a warning message.
2797
2798.. _porting_guide_trng_requirements:
2799
2800TRNG porting requirements
2801~~~~~~~~~~~~~~~~~~~~~~~~~
2802
2803The |TRNG| backend requires the platform to provide the following values
2804and mandatory functions.
2805
2806Values
2807......
2808
2809value: uuid_t plat_trng_uuid [mandatory]
2810^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
2811
2812This value must be defined to the UUID of the TRNG backend that is specific to
2813the hardware after ``plat_entropy_setup`` function is called. This value must
2814conform to the SMCCC calling convention; The most significant 32 bits of the
2815UUID must not equal ``0xffffffff`` or the signed integer ``-1`` as this value in
2816w0 indicates failure to get a TRNG source.
2817
2818Functions
2819.........
2820
2821Function: void plat_entropy_setup(void) [mandatory]
2822^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
2823
2824::
2825
2826  Argument: none
2827  Return: none
2828
2829This function is expected to do platform-specific initialization of any TRNG
2830hardware. This may include generating a UUID from a hardware-specific seed.
2831
2832Function: bool plat_get_entropy(uint64_t \*out) [mandatory]
2833^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
2834
2835::
2836
2837  Argument: uint64_t *
2838  Return: bool
2839  Out : when the return value is true, the entropy has been written into the
2840  storage pointed to
2841
2842This function writes entropy into storage provided by the caller. If no entropy
2843is available, it must return false and the storage must not be written.
2844
2845.. _psci_in_bl31:
2846
2847Power State Coordination Interface (PSCI)
2848--------------------------------------------
2849
2850The TF-A implementation of the PSCI API is based around the concept of a
2851*power domain*. A *power domain* is a CPU or a logical group of CPUs which
2852share some state on which power management operations can be performed as
2853specified by `PSCI`_. Each CPU in the system is assigned a cpu index which is
2854a unique number between ``0`` and ``PLATFORM_CORE_COUNT - 1``. The
2855*power domains* are arranged in a hierarchical tree structure and each
2856*power domain* can be identified in a system by the cpu index of any CPU that
2857is part of that domain and a *power domain level*. A processing element (for
2858example, a CPU) is at level 0. If the *power domain* node above a CPU is a
2859logical grouping of CPUs that share some state, then level 1 is that group of
2860CPUs (for example, a cluster), and level 2 is a group of clusters (for
2861example, the system). More details on the power domain topology and its
2862organization can be found in :ref:`PSCI Power Domain Tree Structure`.
2863
2864BL31's platform initialization code exports a pointer to the platform-specific
2865power management operations required for the PSCI implementation to function
2866correctly. This information is populated in the ``plat_psci_ops`` structure. The
2867PSCI implementation calls members of the ``plat_psci_ops`` structure for performing
2868power management operations on the power domains. For example, the target
2869CPU is specified by its ``MPIDR`` in a PSCI ``CPU_ON`` call. The ``pwr_domain_on()``
2870handler (if present) is called for the CPU power domain.
2871
2872The ``power_state`` parameter of a PSCI ``CPU_SUSPEND`` call can be used to
2873describe composite power states specific to a platform. The PSCI implementation
2874defines a generic representation of the power-state parameter, which is an
2875array of local power states where each index corresponds to a power domain
2876level. Each entry contains the local power state the power domain at that power
2877level could enter. It depends on the ``validate_power_state()`` handler to
2878convert the power-state parameter (possibly encoding a composite power state)
2879passed in a PSCI ``CPU_SUSPEND`` call to this representation.
2880
2881This 32-bit ``power_state`` parameter has two encoding formats (as described in the
2882ARM `PSCI`_ spec), controlled by the ``PSCI_EXTENDED_STATE_ID`` build option.
2883
2884The state ID field in this is entirely platform-defined.
2885Platforms must document their state ID conventions for OS developers. The
2886``power_state`` value is passed from the consumers (Like OSs) through generic PSCI
2887layers to the respective platform handler.
2888
2889The following functions form part of platform port of PSCI functionality.
2890
2891Function : plat_psci_stat_accounting_start() [optional]
2892~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2893
2894::
2895
2896    Argument : const psci_power_state_t *
2897    Return   : void
2898
2899This is an optional hook that platforms can implement for residency statistics
2900accounting before entering a low power state. The ``pwr_domain_state`` field of
2901``state_info`` (first argument) can be inspected if stat accounting is done
2902differently at CPU level versus higher levels. As an example, if the element at
2903index 0 (CPU power level) in the ``pwr_domain_state`` array indicates a power down
2904state, special hardware logic may be programmed in order to keep track of the
2905residency statistics. For higher levels (array indices > 0), the residency
2906statistics could be tracked in software using PMF. If ``ENABLE_PMF`` is set, the
2907default implementation will use PMF to capture timestamps.
2908
2909Function : plat_psci_stat_accounting_stop() [optional]
2910~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2911
2912::
2913
2914    Argument : const psci_power_state_t *
2915    Return   : void
2916
2917This is an optional hook that platforms can implement for residency statistics
2918accounting after exiting from a low power state. The ``pwr_domain_state`` field
2919of ``state_info`` (first argument) can be inspected if stat accounting is done
2920differently at CPU level versus higher levels. As an example, if the element at
2921index 0 (CPU power level) in the ``pwr_domain_state`` array indicates a power down
2922state, special hardware logic may be programmed in order to keep track of the
2923residency statistics. For higher levels (array indices > 0), the residency
2924statistics could be tracked in software using PMF. If ``ENABLE_PMF`` is set, the
2925default implementation will use PMF to capture timestamps.
2926
2927Function : plat_psci_stat_get_residency() [optional]
2928~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2929
2930::
2931
2932    Argument : unsigned int, const psci_power_state_t *, unsigned int
2933    Return   : u_register_t
2934
2935This is an optional interface that is is invoked after resuming from a low power
2936state and provides the time spent resident in that low power state by the power
2937domain at a particular power domain level. When a CPU wakes up from suspend,
2938all its parent power domain levels are also woken up. The generic PSCI code
2939invokes this function for each parent power domain that is resumed and it
2940identified by the ``lvl`` (first argument) parameter. The ``state_info`` (second
2941argument) describes the low power state that the power domain has resumed from.
2942The current CPU is the first CPU in the power domain to resume from the low
2943power state and the ``last_cpu_idx`` (third parameter) is the index of the last
2944CPU in the power domain to suspend and may be needed to calculate the residency
2945for that power domain.
2946
2947Function : plat_get_target_pwr_state() [optional]
2948~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2949
2950::
2951
2952    Argument : unsigned int, const plat_local_state_t *, unsigned int
2953    Return   : plat_local_state_t
2954
2955The PSCI generic code uses this function to let the platform participate in
2956state coordination during a power management operation. The function is passed
2957a pointer to an array of platform specific local power state ``states`` (second
2958argument) which contains the requested power state for each CPU at a particular
2959power domain level ``lvl`` (first argument) within the power domain. The function
2960is expected to traverse this array of upto ``ncpus`` (third argument) and return
2961a coordinated target power state by the comparing all the requested power
2962states. The target power state should not be deeper than any of the requested
2963power states.
2964
2965A weak definition of this API is provided by default wherein it assumes
2966that the platform assigns a local state value in order of increasing depth
2967of the power state i.e. for two power states X & Y, if X < Y
2968then X represents a shallower power state than Y. As a result, the
2969coordinated target local power state for a power domain will be the minimum
2970of the requested local power state values.
2971
2972Function : plat_get_power_domain_tree_desc() [mandatory]
2973~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2974
2975::
2976
2977    Argument : void
2978    Return   : const unsigned char *
2979
2980This function returns a pointer to the byte array containing the power domain
2981topology tree description. The format and method to construct this array are
2982described in :ref:`PSCI Power Domain Tree Structure`. The BL31 PSCI
2983initialization code requires this array to be described by the platform, either
2984statically or dynamically, to initialize the power domain topology tree. In case
2985the array is populated dynamically, then plat_core_pos_by_mpidr() and
2986plat_my_core_pos() should also be implemented suitably so that the topology tree
2987description matches the CPU indices returned by these APIs. These APIs together
2988form the platform interface for the PSCI topology framework.
2989
2990Function : plat_setup_psci_ops() [mandatory]
2991~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2992
2993::
2994
2995    Argument : uintptr_t, const plat_psci_ops **
2996    Return   : int
2997
2998This function may execute with the MMU and data caches enabled if the platform
2999port does the necessary initializations in ``bl31_plat_arch_setup()``. It is only
3000called by the primary CPU.
3001
3002This function is called by PSCI initialization code. Its purpose is to let
3003the platform layer know about the warm boot entrypoint through the
3004``sec_entrypoint`` (first argument) and to export handler routines for
3005platform-specific psci power management actions by populating the passed
3006pointer with a pointer to BL31's private ``plat_psci_ops`` structure.
3007
3008A description of each member of this structure is given below. Please refer to
3009the Arm FVP specific implementation of these handlers in
3010``plat/arm/board/fvp/fvp_pm.c`` as an example. For each PSCI function that the
3011platform wants to support, the associated operation or operations in this
3012structure must be provided and implemented (Refer section 4 of
3013:ref:`Firmware Design` for the PSCI API supported in TF-A). To disable a PSCI
3014function in a platform port, the operation should be removed from this
3015structure instead of providing an empty implementation.
3016
3017plat_psci_ops.cpu_standby()
3018...........................
3019
3020Perform the platform-specific actions to enter the standby state for a cpu
3021indicated by the passed argument. This provides a fast path for CPU standby
3022wherein overheads of PSCI state management and lock acquisition is avoided.
3023For this handler to be invoked by the PSCI ``CPU_SUSPEND`` API implementation,
3024the suspend state type specified in the ``power-state`` parameter should be
3025STANDBY and the target power domain level specified should be the CPU. The
3026handler should put the CPU into a low power retention state (usually by
3027issuing a wfi instruction) and ensure that it can be woken up from that
3028state by a normal interrupt. The generic code expects the handler to succeed.
3029
3030plat_psci_ops.pwr_domain_on()
3031.............................
3032
3033Perform the platform specific actions to power on a CPU, specified
3034by the ``MPIDR`` (first argument). The generic code expects the platform to
3035return PSCI_E_SUCCESS on success or PSCI_E_INTERN_FAIL for any failure.
3036
3037plat_psci_ops.pwr_domain_off_early() [optional]
3038...............................................
3039
3040This optional function performs the platform specific actions to check if
3041powering off the calling CPU and its higher parent power domain levels as
3042indicated by the ``target_state`` (first argument) is possible or allowed.
3043
3044The ``target_state`` encodes the platform coordinated target local power states
3045for the CPU power domain and its parent power domain levels.
3046
3047For this handler, the local power state for the CPU power domain will be a
3048power down state where as it could be either power down, retention or run state
3049for the higher power domain levels depending on the result of state
3050coordination. The generic code expects PSCI_E_DENIED return code if the
3051platform thinks that CPU_OFF should not proceed on the calling CPU.
3052
3053plat_psci_ops.pwr_domain_off()
3054..............................
3055
3056Perform the platform specific actions to prepare to power off the calling CPU
3057and its higher parent power domain levels as indicated by the ``target_state``
3058(first argument). It is called by the PSCI ``CPU_OFF`` API implementation.
3059
3060The ``target_state`` encodes the platform coordinated target local power states
3061for the CPU power domain and its parent power domain levels. The handler
3062needs to perform power management operation corresponding to the local state
3063at each power level.
3064
3065For this handler, the local power state for the CPU power domain will be a
3066power down state where as it could be either power down, retention or run state
3067for the higher power domain levels depending on the result of state
3068coordination. The generic code expects the handler to succeed.
3069
3070plat_psci_ops.pwr_domain_validate_suspend() [optional]
3071......................................................
3072
3073This is an optional function that is only compiled into the build if the build
3074option ``PSCI_OS_INIT_MODE`` is enabled.
3075
3076If implemented, this function allows the platform to perform platform specific
3077validations based on hardware states. The generic code expects this function to
3078return PSCI_E_SUCCESS on success, or either PSCI_E_DENIED or
3079PSCI_E_INVALID_PARAMS as appropriate for any invalid requests.
3080
3081plat_psci_ops.pwr_domain_suspend_pwrdown_early() [optional]
3082...........................................................
3083
3084This optional function may be used as a performance optimization to replace
3085or complement pwr_domain_suspend() on some platforms. Its calling semantics
3086are identical to pwr_domain_suspend(), except the PSCI implementation only
3087calls this function when suspending to a power down state, and it guarantees
3088that data caches are enabled.
3089
3090When HW_ASSISTED_COHERENCY = 0, the PSCI implementation disables data caches
3091before calling pwr_domain_suspend(). If the target_state corresponds to a
3092power down state and it is safe to perform some or all of the platform
3093specific actions in that function with data caches enabled, it may be more
3094efficient to move those actions to this function. When HW_ASSISTED_COHERENCY
3095= 1, data caches remain enabled throughout, and so there is no advantage to
3096moving platform specific actions to this function.
3097
3098plat_psci_ops.pwr_domain_suspend()
3099..................................
3100
3101Perform the platform specific actions to prepare to suspend the calling
3102CPU and its higher parent power domain levels as indicated by the
3103``target_state`` (first argument). It is called by the PSCI ``CPU_SUSPEND``
3104API implementation.
3105
3106The ``target_state`` has a similar meaning as described in
3107the ``pwr_domain_off()`` operation. It encodes the platform coordinated
3108target local power states for the CPU power domain and its parent
3109power domain levels. The handler needs to perform power management operation
3110corresponding to the local state at each power level. The generic code
3111expects the handler to succeed.
3112
3113The difference between turning a power domain off versus suspending it is that
3114in the former case, the power domain is expected to re-initialize its state
3115when it is next powered on (see ``pwr_domain_on_finish()``). In the latter
3116case, the power domain is expected to save enough state so that it can resume
3117execution by restoring this state when its powered on (see
3118``pwr_domain_suspend_finish()``).
3119
3120When suspending a core, the platform can also choose to power off the GICv3
3121Redistributor and ITS through an implementation-defined sequence. To achieve
3122this safely, the ITS context must be saved first. The architectural part is
3123implemented by the ``gicv3_its_save_disable()`` helper, but most of the needed
3124sequence is implementation defined and it is therefore the responsibility of
3125the platform code to implement the necessary sequence. Then the GIC
3126Redistributor context can be saved using the ``gicv3_rdistif_save()`` helper.
3127Powering off the Redistributor requires the implementation to support it and it
3128is the responsibility of the platform code to execute the right implementation
3129defined sequence.
3130
3131When a system suspend is requested, the platform can also make use of the
3132``gicv3_distif_save()`` helper to save the context of the GIC Distributor after
3133it has saved the context of the Redistributors and ITS of all the cores in the
3134system. The context of the Distributor can be large and may require it to be
3135allocated in a special area if it cannot fit in the platform's global static
3136data, for example in DRAM. The Distributor can then be powered down using an
3137implementation-defined sequence.
3138
3139plat_psci_ops.pwr_domain_pwr_down()
3140.......................................
3141
3142This is an optional function and, if implemented, is expected to perform
3143platform specific actions before the CPU is powered down. Since this function is
3144invoked outside the PSCI locks, the actions performed in this hook must be local
3145to the CPU or the platform must ensure that races between multiple CPUs cannot
3146occur.
3147
3148The ``target_state`` has a similar meaning as described in the ``pwr_domain_off()``
3149operation and it encodes the platform coordinated target local power states for
3150the CPU power domain and its parent power domain levels.
3151
3152It is preferred that this function returns. The caller will invoke
3153``wfi()`` to powerdown the CPU, mitigate any powerdown errata,
3154and handle any wakeups that may arise. Previously, this function did not return
3155and instead called ``wfi`` (in an infinite loop) directly. This is still
3156possible on platforms where this is guaranteed to be terminal, however, it is
3157strongly discouraged going forward.
3158
3159Previously this function was called ``pwr_domain_pwr_down_wfi()`` and invoked
3160``psci_power_down_wfi()`` (now removed).
3161
3162plat_psci_ops.pwr_domain_on_finish()
3163....................................
3164
3165This function is called by the PSCI implementation after the calling CPU is
3166powered on and released from reset in response to an earlier PSCI ``CPU_ON`` call.
3167It performs the platform-specific setup required to initialize enough state for
3168this CPU to enter the normal world and also provide secure runtime firmware
3169services.
3170
3171The ``target_state`` (first argument) is the prior state of the power domains
3172immediately before the CPU was turned on. It indicates which power domains
3173above the CPU might require initialization due to having previously been in
3174low power states. The generic code expects the handler to succeed.
3175
3176plat_psci_ops.pwr_domain_on_finish_late() [optional]
3177...........................................................
3178
3179This optional function is called by the PSCI implementation after the calling
3180CPU is fully powered on with respective data caches enabled. The calling CPU and
3181the associated cluster are guaranteed to be participating in coherency. This
3182function gives the flexibility to perform any platform-specific actions safely,
3183such as initialization or modification of shared data structures, without the
3184overhead of explicit cache maintainace operations.
3185
3186The ``target_state`` has a similar meaning as described in the ``pwr_domain_on_finish()``
3187operation. The generic code expects the handler to succeed.
3188
3189plat_psci_ops.pwr_domain_suspend_finish()
3190.........................................
3191
3192This function is called by the PSCI implementation after the calling CPU is
3193powered on and released from reset in response to an asynchronous wakeup
3194event, for example a timer interrupt that was programmed by the CPU during the
3195``CPU_SUSPEND`` call or ``SYSTEM_SUSPEND`` call. It performs the platform-specific
3196setup required to restore the saved state for this CPU to resume execution
3197in the normal world and also provide secure runtime firmware services.
3198
3199The ``target_state`` (first argument) has a similar meaning as described in
3200the ``pwr_domain_on_finish()`` operation. The generic code expects the platform
3201to succeed.
3202
3203If the Distributor, Redistributors or ITS have been powered off as part of a
3204suspend, their context must be restored in this function in the reverse order
3205to how they were saved during suspend sequence.
3206
3207plat_psci_ops.system_off()
3208..........................
3209
3210This function is called by PSCI implementation in response to a ``SYSTEM_OFF``
3211call. It performs the platform-specific system poweroff sequence after
3212notifying the Secure Payload Dispatcher. The caller will call ``wfi`` if this
3213function returns, similar to `plat_psci_ops.pwr_domain_pwr_down()`_.
3214
3215plat_psci_ops.system_reset()
3216............................
3217
3218This function is called by PSCI implementation in response to a ``SYSTEM_RESET``
3219call. It performs the platform-specific system reset sequence after
3220notifying the Secure Payload Dispatcher. The caller will call ``wfi`` if this
3221function returns, similar to `plat_psci_ops.pwr_domain_pwr_down()`_.
3222
3223plat_psci_ops.validate_power_state()
3224....................................
3225
3226This function is called by the PSCI implementation during the ``CPU_SUSPEND``
3227call to validate the ``power_state`` parameter of the PSCI API and if valid,
3228populate it in ``req_state`` (second argument) array as power domain level
3229specific local states. If the ``power_state`` is invalid, the platform must
3230return PSCI_E_INVALID_PARAMS as error, which is propagated back to the
3231normal world PSCI client.
3232
3233plat_psci_ops.validate_ns_entrypoint()
3234......................................
3235
3236This function is called by the PSCI implementation during the ``CPU_SUSPEND``,
3237``SYSTEM_SUSPEND`` and ``CPU_ON`` calls to validate the non-secure ``entry_point``
3238parameter passed by the normal world. If the ``entry_point`` is invalid,
3239the platform must return PSCI_E_INVALID_ADDRESS as error, which is
3240propagated back to the normal world PSCI client.
3241
3242plat_psci_ops.get_sys_suspend_power_state()
3243...........................................
3244
3245This function is called by the PSCI implementation during the ``SYSTEM_SUSPEND``
3246call to get the ``req_state`` parameter from platform which encodes the power
3247domain level specific local states to suspend to system affinity level. The
3248``req_state`` will be utilized to do the PSCI state coordination and
3249``pwr_domain_suspend()`` will be invoked with the coordinated target state to
3250enter system suspend.
3251
3252plat_psci_ops.get_pwr_lvl_state_idx()
3253.....................................
3254
3255This is an optional function and, if implemented, is invoked by the PSCI
3256implementation to convert the ``local_state`` (first argument) at a specified
3257``pwr_lvl`` (second argument) to an index between 0 and
3258``PLAT_MAX_PWR_LVL_STATES`` - 1. This function is only needed if the platform
3259supports more than two local power states at each power domain level, that is
3260``PLAT_MAX_PWR_LVL_STATES`` is greater than 2, and needs to account for these
3261local power states.
3262
3263plat_psci_ops.translate_power_state_by_mpidr()
3264..............................................
3265
3266This is an optional function and, if implemented, verifies the ``power_state``
3267(second argument) parameter of the PSCI API corresponding to a target power
3268domain. The target power domain is identified by using both ``MPIDR`` (first
3269argument) and the power domain level encoded in ``power_state``. The power domain
3270level specific local states are to be extracted from ``power_state`` and be
3271populated in the ``output_state`` (third argument) array. The functionality
3272is similar to the ``validate_power_state`` function described above and is
3273envisaged to be used in case the validity of ``power_state`` depend on the
3274targeted power domain. If the ``power_state`` is invalid for the targeted power
3275domain, the platform must return PSCI_E_INVALID_PARAMS as error. If this
3276function is not implemented, then the generic implementation relies on
3277``validate_power_state`` function to translate the ``power_state``.
3278
3279This function can also be used in case the platform wants to support local
3280power state encoding for ``power_state`` parameter of PSCI_STAT_COUNT/RESIDENCY
3281APIs as described in Section 5.18 of `PSCI`_.
3282
3283plat_psci_ops.get_node_hw_state()
3284.................................
3285
3286This is an optional function. If implemented this function is intended to return
3287the power state of a node (identified by the first parameter, the ``MPIDR``) in
3288the power domain topology (identified by the second parameter, ``power_level``),
3289as retrieved from a power controller or equivalent component on the platform.
3290Upon successful completion, the implementation must map and return the final
3291status among ``HW_ON``, ``HW_OFF`` or ``HW_STANDBY``. Upon encountering failures, it
3292must return either ``PSCI_E_INVALID_PARAMS`` or ``PSCI_E_NOT_SUPPORTED`` as
3293appropriate.
3294
3295Implementations are not expected to handle ``power_levels`` greater than
3296``PLAT_MAX_PWR_LVL``.
3297
3298plat_psci_ops.system_reset2()
3299.............................
3300
3301This is an optional function. If implemented this function is
3302called during the ``SYSTEM_RESET2`` call to perform a reset
3303based on the first parameter ``reset_type`` as specified in
3304`PSCI`_. The parameter ``cookie`` can be used to pass additional
3305reset information. If the ``reset_type`` is not supported, the
3306function must return ``PSCI_E_NOT_SUPPORTED``. For architectural
3307resets, all failures must return ``PSCI_E_INVALID_PARAMETERS``
3308and vendor reset can return other PSCI error codes as defined
3309in `PSCI`_. If this function returns success, the caller will call
3310``wfi`` similar to `plat_psci_ops.pwr_domain_pwr_down()`_.
3311
3312plat_psci_ops.write_mem_protect()
3313.................................
3314
3315This is an optional function. If implemented it enables or disables the
3316``MEM_PROTECT`` functionality based on the value of ``val``.
3317A non-zero value enables ``MEM_PROTECT`` and a value of zero
3318disables it. Upon encountering failures it must return a negative value
3319and on success it must return 0.
3320
3321plat_psci_ops.read_mem_protect()
3322................................
3323
3324This is an optional function. If implemented it returns the current
3325state of ``MEM_PROTECT`` via the ``val`` parameter.  Upon encountering
3326failures it must return a negative value and on success it must
3327return 0.
3328
3329plat_psci_ops.mem_protect_chk()
3330...............................
3331
3332This is an optional function. If implemented it checks if a memory
3333region defined by a base address ``base`` and with a size of ``length``
3334bytes is protected by ``MEM_PROTECT``.  If the region is protected
3335then it must return 0, otherwise it must return a negative number.
3336
3337.. _porting_guide_imf_in_bl31:
3338
3339Secure payload power management callback
3340~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
3341
3342During PSCI power management operations, the EL3 Runtime Software may
3343need to perform some bookkeeping, and PSCI library provides
3344``spd_pm_ops_t`` callbacks for this purpose. These hooks must be
3345populated and registered by using ``psci_register_spd_pm_hook()`` PSCI
3346library interface.
3347
3348Typical bookkeeping during PSCI power management calls include save/restore
3349of the EL3 Runtime Software context. Also if the EL3 Runtime Software makes
3350use of secure interrupts, then these interrupts must also be managed
3351appropriately during CPU power down/power up. Any secure interrupt targeted
3352to the current CPU must be disabled or re-targeted to other running CPU prior
3353to power down of the current CPU. During power up, these interrupt can be
3354enabled/re-targeted back to the current CPU.
3355
3356.. code:: c
3357
3358        typedef struct spd_pm_ops {
3359                void (*svc_on)(u_register_t target_cpu);
3360                int32_t (*svc_off)(u_register_t __unused);
3361                void (*svc_suspend)(u_register_t max_off_pwrlvl);
3362                void (*svc_on_finish)(u_register_t __unused);
3363                void (*svc_suspend_finish)(u_register_t max_off_pwrlvl);
3364                int32_t (*svc_migrate)(u_register_t from_cpu, u_register_t to_cpu);
3365                int32_t (*svc_migrate_info)(u_register_t *resident_cpu);
3366                void (*svc_system_off)(void);
3367                void (*svc_system_reset)(void);
3368        } spd_pm_ops_t;
3369
3370A brief description of each callback is given below:
3371
3372-  svc_on, svc_off, svc_on_finish
3373
3374   The ``svc_on``, ``svc_off`` callbacks are called during PSCI_CPU_ON,
3375   PSCI_CPU_OFF APIs respectively. The ``svc_on_finish`` is called when the
3376   target CPU of PSCI_CPU_ON API powers up and executes the
3377   ``psci_warmboot_entrypoint()`` PSCI library interface.
3378
3379-  svc_suspend, svc_suspend_finish
3380
3381   The ``svc_suspend`` callback is called during power down bu either
3382   PSCI_SUSPEND or PSCI_SYSTEM_SUSPEND APIs. The ``svc_suspend_finish`` is
3383   called when the CPU wakes up from suspend and executes the
3384   ``psci_warmboot_entrypoint()`` PSCI library interface. The ``max_off_pwrlvl``
3385   (first parameter) denotes the highest power domain level being powered down
3386   to or woken up from suspend.
3387
3388-  svc_system_off, svc_system_reset
3389
3390   These callbacks are called during PSCI_SYSTEM_OFF and PSCI_SYSTEM_RESET
3391   PSCI APIs respectively.
3392
3393-  svc_migrate_info
3394
3395   This callback is called in response to PSCI_MIGRATE_INFO_TYPE or
3396   PSCI_MIGRATE_INFO_UP_CPU APIs. The return value of this callback must
3397   correspond to the return value of PSCI_MIGRATE_INFO_TYPE API as described
3398   in `PSCI`_. If the secure payload is a Uniprocessor (UP)
3399   implementation, then it must update the mpidr of the CPU it is resident in
3400   via ``resident_cpu`` (first argument). The updates to ``resident_cpu`` is
3401   ignored if the secure payload is a multiprocessor (MP) implementation.
3402
3403-  svc_migrate
3404
3405   This callback is only relevant if the secure payload in EL3 Runtime
3406   Software is a Uniprocessor (UP) implementation and supports migration from
3407   the current CPU ``from_cpu`` (first argument) to another CPU ``to_cpu``
3408   (second argument). This callback is called in response to PSCI_MIGRATE
3409   API. This callback is never called if the secure payload is a
3410   Multiprocessor (MP) implementation.
3411
3412Interrupt Management framework (in BL31)
3413----------------------------------------
3414
3415BL31 implements an Interrupt Management Framework (IMF) to manage interrupts
3416generated in either security state and targeted to EL1 or EL2 in the non-secure
3417state or EL3/S-EL1 in the secure state. The design of this framework is
3418described in the :ref:`Interrupt Management Framework`
3419
3420A platform should export the following APIs to support the IMF. The following
3421text briefly describes each API and its implementation in Arm standard
3422platforms. The API implementation depends upon the type of interrupt controller
3423present in the platform. Arm standard platform layer supports both
3424`Arm Generic Interrupt Controller version 2.0 (GICv2)`_
3425and `3.0 (GICv3)`_. Juno builds the Arm platform layer to use GICv2 and the
3426FVP can be configured to use either GICv2 or GICv3 depending on the build flag
3427``FVP_USE_GIC_DRIVER`` (See :ref:`build_options_arm_fvp_platform` for more
3428details).
3429
3430See also: :ref:`Interrupt Controller Abstraction APIs<Platform Interrupt Controller API>`.
3431
3432Function : plat_interrupt_type_to_line() [mandatory]
3433~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
3434
3435::
3436
3437    Argument : uint32_t, uint32_t
3438    Return   : uint32_t
3439
3440The Arm processor signals an interrupt exception either through the IRQ or FIQ
3441interrupt line. The specific line that is signaled depends on how the interrupt
3442controller (IC) reports different interrupt types from an execution context in
3443either security state. The IMF uses this API to determine which interrupt line
3444the platform IC uses to signal each type of interrupt supported by the framework
3445from a given security state. This API must be invoked at EL3.
3446
3447The first parameter will be one of the ``INTR_TYPE_*`` values (see
3448:ref:`Interrupt Management Framework`) indicating the target type of the
3449interrupt, the second parameter is the security state of the originating
3450execution context. The return result is the bit position in the ``SCR_EL3``
3451register of the respective interrupt trap: IRQ=1, FIQ=2.
3452
3453In the case of Arm standard platforms using GICv2, S-EL1 interrupts are
3454configured as FIQs and Non-secure interrupts as IRQs from either security
3455state.
3456
3457In the case of Arm standard platforms using GICv3, the interrupt line to be
3458configured depends on the security state of the execution context when the
3459interrupt is signalled and are as follows:
3460
3461-  The S-EL1 interrupts are signaled as IRQ in S-EL0/1 context and as FIQ in
3462   NS-EL0/1/2 context.
3463-  The Non secure interrupts are signaled as FIQ in S-EL0/1 context and as IRQ
3464   in the NS-EL0/1/2 context.
3465-  The EL3 interrupts are signaled as FIQ in both S-EL0/1 and NS-EL0/1/2
3466   context.
3467
3468Function : plat_ic_get_pending_interrupt_type() [mandatory]
3469~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
3470
3471::
3472
3473    Argument : void
3474    Return   : uint32_t
3475
3476This API returns the type of the highest priority pending interrupt at the
3477platform IC. The IMF uses the interrupt type to retrieve the corresponding
3478handler function. ``INTR_TYPE_INVAL`` is returned when there is no interrupt
3479pending. The valid interrupt types that can be returned are ``INTR_TYPE_EL3``,
3480``INTR_TYPE_S_EL1`` and ``INTR_TYPE_NS``. This API must be invoked at EL3.
3481
3482In the case of Arm standard platforms using GICv2, the *Highest Priority
3483Pending Interrupt Register* (``GICC_HPPIR``) is read to determine the id of
3484the pending interrupt. The type of interrupt depends upon the id value as
3485follows.
3486
3487#. id < 1022 is reported as a S-EL1 interrupt
3488#. id = 1022 is reported as a Non-secure interrupt.
3489#. id = 1023 is reported as an invalid interrupt type.
3490
3491In the case of Arm standard platforms using GICv3, the system register
3492``ICC_HPPIR0_EL1``, *Highest Priority Pending group 0 Interrupt Register*,
3493is read to determine the id of the pending interrupt. The type of interrupt
3494depends upon the id value as follows.
3495
3496#. id = ``PENDING_G1S_INTID`` (1020) is reported as a S-EL1 interrupt
3497#. id = ``PENDING_G1NS_INTID`` (1021) is reported as a Non-secure interrupt.
3498#. id = ``GIC_SPURIOUS_INTERRUPT`` (1023) is reported as an invalid interrupt type.
3499#. All other interrupt id's are reported as EL3 interrupt.
3500
3501Function : plat_ic_get_pending_interrupt_id() [mandatory]
3502~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
3503
3504::
3505
3506    Argument : void
3507    Return   : uint32_t
3508
3509This API returns the id of the highest priority pending interrupt at the
3510platform IC. ``INTR_ID_UNAVAILABLE`` is returned when there is no interrupt
3511pending.
3512
3513In the case of Arm standard platforms using GICv2, the *Highest Priority
3514Pending Interrupt Register* (``GICC_HPPIR``) is read to determine the id of the
3515pending interrupt. The id that is returned by API depends upon the value of
3516the id read from the interrupt controller as follows.
3517
3518#. id < 1022. id is returned as is.
3519#. id = 1022. The *Aliased Highest Priority Pending Interrupt Register*
3520   (``GICC_AHPPIR``) is read to determine the id of the non-secure interrupt.
3521   This id is returned by the API.
3522#. id = 1023. ``INTR_ID_UNAVAILABLE`` is returned.
3523
3524In the case of Arm standard platforms using GICv3, if the API is invoked from
3525EL3, the system register ``ICC_HPPIR0_EL1``, *Highest Priority Pending Interrupt
3526group 0 Register*, is read to determine the id of the pending interrupt. The id
3527that is returned by API depends upon the value of the id read from the
3528interrupt controller as follows.
3529
3530#. id < ``PENDING_G1S_INTID`` (1020). id is returned as is.
3531#. id = ``PENDING_G1S_INTID`` (1020) or ``PENDING_G1NS_INTID`` (1021). The system
3532   register ``ICC_HPPIR1_EL1``, *Highest Priority Pending Interrupt group 1
3533   Register* is read to determine the id of the group 1 interrupt. This id
3534   is returned by the API as long as it is a valid interrupt id
3535#. If the id is any of the special interrupt identifiers,
3536   ``INTR_ID_UNAVAILABLE`` is returned.
3537
3538When the API invoked from S-EL1 for GICv3 systems, the id read from system
3539register ``ICC_HPPIR1_EL1``, *Highest Priority Pending group 1 Interrupt
3540Register*, is returned if is not equal to GIC_SPURIOUS_INTERRUPT (1023) else
3541``INTR_ID_UNAVAILABLE`` is returned.
3542
3543Function : plat_ic_acknowledge_interrupt() [mandatory]
3544~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
3545
3546::
3547
3548    Argument : void
3549    Return   : uint32_t
3550
3551This API is used by the CPU to indicate to the platform IC that processing of
3552the highest pending interrupt has begun. It should return the raw, unmodified
3553value obtained from the interrupt controller when acknowledging an interrupt.
3554The actual interrupt number shall be extracted from this raw value using the API
3555`plat_ic_get_interrupt_id()<plat_ic_get_interrupt_id>`.
3556
3557This function in Arm standard platforms using GICv2, reads the *Interrupt
3558Acknowledge Register* (``GICC_IAR``). This changes the state of the highest
3559priority pending interrupt from pending to active in the interrupt controller.
3560It returns the value read from the ``GICC_IAR``, unmodified.
3561
3562In the case of Arm standard platforms using GICv3, if the API is invoked
3563from EL3, the function reads the system register ``ICC_IAR0_EL1``, *Interrupt
3564Acknowledge Register group 0*. If the API is invoked from S-EL1, the function
3565reads the system register ``ICC_IAR1_EL1``, *Interrupt Acknowledge Register
3566group 1*. The read changes the state of the highest pending interrupt from
3567pending to active in the interrupt controller. The value read is returned
3568unmodified.
3569
3570The TSP uses this API to start processing of the secure physical timer
3571interrupt.
3572
3573Function : plat_ic_end_of_interrupt() [mandatory]
3574~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
3575
3576::
3577
3578    Argument : uint32_t
3579    Return   : void
3580
3581This API is used by the CPU to indicate to the platform IC that processing of
3582the interrupt corresponding to the id (passed as the parameter) has
3583finished. The id should be the same as the id returned by the
3584``plat_ic_acknowledge_interrupt()`` API.
3585
3586Arm standard platforms write the id to the *End of Interrupt Register*
3587(``GICC_EOIR``) in case of GICv2, and to ``ICC_EOIR0_EL1`` or ``ICC_EOIR1_EL1``
3588system register in case of GICv3 depending on where the API is invoked from,
3589EL3 or S-EL1. This deactivates the corresponding interrupt in the interrupt
3590controller.
3591
3592The TSP uses this API to finish processing of the secure physical timer
3593interrupt.
3594
3595Function : plat_ic_get_interrupt_type() [mandatory]
3596~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
3597
3598::
3599
3600    Argument : uint32_t
3601    Return   : uint32_t
3602
3603This API returns the type of the interrupt id passed as the parameter.
3604``INTR_TYPE_INVAL`` is returned if the id is invalid. If the id is valid, a valid
3605interrupt type (one of ``INTR_TYPE_EL3``, ``INTR_TYPE_S_EL1`` and ``INTR_TYPE_NS``) is
3606returned depending upon how the interrupt has been configured by the platform
3607IC. This API must be invoked at EL3.
3608
3609Arm standard platforms using GICv2 configures S-EL1 interrupts as Group0 interrupts
3610and Non-secure interrupts as Group1 interrupts. It reads the group value
3611corresponding to the interrupt id from the relevant *Interrupt Group Register*
3612(``GICD_IGROUPRn``). It uses the group value to determine the type of interrupt.
3613
3614In the case of Arm standard platforms using GICv3, both the *Interrupt Group
3615Register* (``GICD_IGROUPRn``) and *Interrupt Group Modifier Register*
3616(``GICD_IGRPMODRn``) is read to figure out whether the interrupt is configured
3617as Group 0 secure interrupt, Group 1 secure interrupt or Group 1 NS interrupt.
3618
3619Registering a console
3620---------------------
3621
3622Platforms will need to implement the TF-A console framework to register and use
3623a console for visual data output in TF-A. These can be used for data output during
3624the different stages of the firmware boot process and also for debugging purposes.
3625
3626The console framework can be used to output data on to a console using a number of
3627TF-A supported UARTs. Multiple consoles can be registered at the same time with
3628different output scopes (BOOT, RUNTIME, CRASH) so that data can be displayed on
3629their respective consoles without unnecessary cluttering of a single console.
3630
3631Information for registering a console can be found in the :ref:`Console Framework` section
3632of the :ref:`System Design` documentation.
3633
3634Common helper functions
3635-----------------------
3636Function : elx_panic()
3637~~~~~~~~~~~~~~~~~~~~~~
3638
3639::
3640
3641    Argument : void
3642    Return   : void
3643
3644This API is called from assembly files when reporting a critical failure
3645that has occured in lower EL and is been trapped in EL3. This call
3646**must not** return.
3647
3648Function : el3_panic()
3649~~~~~~~~~~~~~~~~~~~~~~
3650
3651::
3652
3653    Argument : void
3654    Return   : void
3655
3656This API is called from assembly files when encountering a critical failure that
3657cannot be recovered from. This function assumes that it is invoked from a C
3658runtime environment i.e. valid stack exists. This call **must not** return.
3659
3660Function : panic()
3661~~~~~~~~~~~~~~~~~~
3662
3663::
3664
3665    Argument : void
3666    Return   : void
3667
3668This API called from C files when encountering a critical failure that cannot
3669be recovered from. This function in turn prints backtrace (if enabled) and calls
3670el3_panic(). This call **must not** return.
3671
3672Crash Reporting mechanism (in BL31)
3673-----------------------------------
3674
3675BL31 implements a crash reporting mechanism which prints the various registers
3676of the CPU to enable quick crash analysis and debugging. This mechanism relies
3677on the platform implementing ``plat_crash_console_init``,
3678``plat_crash_console_putc`` and ``plat_crash_console_flush``.
3679
3680The file ``plat/common/aarch64/crash_console_helpers.S`` contains sample
3681implementation of all of them. Platforms may include this file to their
3682makefiles in order to benefit from them. By default, they will cause the crash
3683output to be routed over the normal console infrastructure and get printed on
3684consoles configured to output in crash state. ``console_set_scope()`` can be
3685used to control whether a console is used for crash output.
3686
3687.. note::
3688   Platforms are responsible for making sure that they only mark consoles for
3689   use in the crash scope that are able to support this, i.e. that are written
3690   in assembly and conform with the register clobber rules for putc()
3691   (x0-x2, x16-x17) and flush() (x0-x3, x16-x17) crash callbacks.
3692
3693In some cases (such as debugging very early crashes that happen before the
3694normal boot console can be set up), platforms may want to control crash output
3695more explicitly. These platforms may instead provide custom implementations for
3696these. They are executed outside of a C environment and without a stack. Many
3697console drivers provide functions named ``console_xxx_core_init/putc/flush``
3698that are designed to be used by these functions. See Arm platforms (like juno)
3699for an example of this.
3700
3701Function : plat_crash_console_init [mandatory]
3702~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
3703
3704::
3705
3706    Argument : void
3707    Return   : int
3708
3709This API is used by the crash reporting mechanism to initialize the crash
3710console. It must only use the general purpose registers x0 through x7 to do the
3711initialization and returns 1 on success.
3712
3713Function : plat_crash_console_putc [mandatory]
3714~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
3715
3716::
3717
3718    Argument : int
3719    Return   : int
3720
3721This API is used by the crash reporting mechanism to print a character on the
3722designated crash console. It must only use general purpose registers x1 and
3723x2 to do its work. The parameter and the return value are in general purpose
3724register x0.
3725
3726Function : plat_crash_console_flush [mandatory]
3727~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
3728
3729::
3730
3731    Argument : void
3732    Return   : void
3733
3734This API is used by the crash reporting mechanism to force write of all buffered
3735data on the designated crash console. It should only use general purpose
3736registers x0 through x5 to do its work.
3737
3738Function : plat_setup_early_console [optional]
3739~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
3740
3741::
3742
3743    Argument : void
3744    Return   : void
3745
3746This API is used to setup the early console, it is required only if the flag
3747``EARLY_CONSOLE`` is enabled.
3748
3749.. _External Abort handling and RAS Support:
3750
3751External Abort handling and RAS Support
3752---------------------------------------
3753
3754If any cores on the platform support powerdown abandon (check the "Core powerup
3755and powerdown sequence" in their TRMs), then
3756these functions should be able to handle being called with power domains off and
3757after the powerdown ``wfi``. In other words it may run after a call to
3758``pwr_domain_suspend()`` and before a call to ``pwr_domain_suspend_finish()``
3759(and their power off counterparts).
3760
3761Should this not be desirable, or if there is no powerdown abandon support, then
3762RAS errors should be masked by writing any relevant error records in any
3763powerdown hooks to prevent deadlocks due to a RAS error after the point of no
3764return. See the core's TRM for further information.
3765
3766Function : plat_ea_handler
3767~~~~~~~~~~~~~~~~~~~~~~~~~~
3768
3769::
3770
3771    Argument : int
3772    Argument : uint64_t
3773    Argument : void *
3774    Argument : void *
3775    Argument : uint64_t
3776    Return   : void
3777
3778This function is invoked by the runtime exception handling framework for the
3779platform to handle an External Abort received at EL3. The intention of the
3780function is to attempt to resolve the cause of External Abort and return;
3781if that's not possible then an orderly shutdown of the system is initiated.
3782
3783The first parameter (``int ea_reason``) indicates the reason for External Abort.
3784Its value is one of ``ERROR_EA_*`` constants defined in ``ea_handle.h``.
3785
3786The second parameter (``uint64_t syndrome``) is the respective syndrome
3787presented to EL3 after having received the External Abort. Depending on the
3788nature of the abort (as can be inferred from the ``ea_reason`` parameter), this
3789can be the content of either ``ESR_EL3`` or ``DISR_EL1``.
3790
3791The third parameter (``void *cookie``) is unused for now. The fourth parameter
3792(``void *handle``) is a pointer to the preempted context. The fifth parameter
3793(``uint64_t flags``) indicates the preempted security state. These parameters
3794are received from the top-level exception handler.
3795
3796This function must be implemented if a platform expects Firmware First handling
3797of External Aborts.
3798
3799Function : plat_handle_uncontainable_ea
3800~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
3801
3802::
3803
3804    Argument : int
3805    Argument : uint64_t
3806    Return   : void
3807
3808This function is invoked by the RAS framework when an External Abort of
3809Uncontainable type is received at EL3. Due to the critical nature of
3810Uncontainable errors, the intention of this function is to initiate orderly
3811shutdown of the system, and is not expected to return.
3812
3813This function must be implemented in assembly.
3814
3815The first and second parameters are the same as that of ``plat_ea_handler``.
3816
3817The default implementation of this function calls
3818``report_unhandled_exception``.
3819
3820Function : plat_handle_double_fault
3821~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
3822
3823::
3824
3825    Argument : int
3826    Argument : uint64_t
3827    Return   : void
3828
3829This function is invoked by the RAS framework when another External Abort is
3830received at EL3 while one is already being handled. I.e., a call to
3831``plat_ea_handler`` is outstanding. Due to its critical nature, the intention of
3832this function is to initiate orderly shutdown of the system, and is not expected
3833recover or return.
3834
3835This function must be implemented in assembly.
3836
3837The first and second parameters are the same as that of ``plat_ea_handler``.
3838
3839The default implementation of this function calls
3840``report_unhandled_exception``.
3841
3842Function : plat_handle_el3_ea
3843~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
3844
3845::
3846
3847    Return   : void
3848
3849This function is invoked when an External Abort is received while executing in
3850EL3. Due to its critical nature, the intention of this function is to initiate
3851orderly shutdown of the system, and is not expected recover or return.
3852
3853This function must be implemented in assembly.
3854
3855The default implementation of this function calls
3856``report_unhandled_exception``.
3857
3858Function : plat_handle_rng_trap
3859~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
3860
3861::
3862
3863    Argument : uint64_t
3864    Argument : cpu_context_t *
3865    Return   : int
3866
3867This function is invoked by BL31's exception handler when there is a synchronous
3868system register trap caused by access to the RNDR or RNDRRS registers. It allows
3869platforms implementing ``FEAT_RNG_TRAP`` and enabling ``ENABLE_FEAT_RNG_TRAP`` to
3870emulate those system registers by returing back some entropy to the lower EL.
3871
3872The first parameter (``uint64_t esr_el3``) contains the content of the ESR_EL3
3873syndrome register, which encodes the instruction that was trapped. The interesting
3874information in there is the target register (``get_sysreg_iss_rt()``).
3875
3876The second parameter (``cpu_context_t *ctx``) represents the CPU state in the
3877lower exception level, at the time when the execution of the ``mrs`` instruction
3878was trapped. Its content can be changed, to put the entropy into the target
3879register.
3880
3881The return value indicates how to proceed:
3882
3883-  When returning ``TRAP_RET_UNHANDLED`` (-1), the machine will panic.
3884-  When returning ``TRAP_RET_REPEAT`` (0), the exception handler will return
3885   to the same instruction, so its execution will be repeated.
3886-  When returning ``TRAP_RET_CONTINUE`` (1), the exception handler will return
3887   to the next instruction.
3888
3889This function needs to be implemented by a platform if it enables FEAT_RNG_TRAP.
3890
3891Function : plat_handle_impdef_trap
3892~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
3893
3894::
3895
3896    Argument : uint64_t
3897    Argument : cpu_context_t *
3898    Return   : int
3899
3900This function is invoked by BL31's exception handler when there is a synchronous
3901system register trap caused by access to the implementation defined registers.
3902It allows platforms enabling ``IMPDEF_SYSREG_TRAP`` to emulate those system
3903registers choosing to program bits of their choice. If using in combination with
3904``ARCH_FEATURE_AVAILABILITY``, the macros
3905{SCR,MDCR,CPTR}_PLAT_{BITS,IGNORED,FLIPPED} should be defined to report correct
3906results.
3907
3908The first parameter (``uint64_t esr_el3``) contains the content of the ESR_EL3
3909syndrome register, which encodes the instruction that was trapped.
3910
3911The second parameter (``cpu_context_t *ctx``) represents the CPU state in the
3912lower exception level, at the time when the execution of the ``mrs`` instruction
3913was trapped.
3914
3915The return value indicates how to proceed:
3916
3917-  When returning ``TRAP_RET_UNHANDLED`` (-1), the machine will panic.
3918-  When returning ``TRAP_RET_REPEAT`` (0), the exception handler will return
3919   to the same instruction, so its execution will be repeated.
3920-  When returning ``TRAP_RET_CONTINUE`` (1), the exception handler will return
3921   to the next instruction.
3922
3923This function needs to be implemented by a platform if it enables
3924IMPDEF_SYSREG_TRAP.
3925
3926Build flags
3927-----------
3928
3929There are some build flags which can be defined by the platform to control
3930inclusion or exclusion of certain BL stages from the FIP image. These flags
3931need to be defined in the platform makefile which will get included by the
3932build system.
3933
3934-  **NEED_BL33**
3935   By default, this flag is defined ``yes`` by the build system and ``BL33``
3936   build option should be supplied as a build option. The platform has the
3937   option of excluding the BL33 image in the ``fip`` image by defining this flag
3938   to ``no``. If any of the options ``EL3_PAYLOAD_BASE`` or ``PRELOADED_BL33_BASE``
3939   are used, this flag will be set to ``no`` automatically.
3940
3941-  **ARM_ARCH_MAJOR and ARM_ARCH_MINOR**
3942   By default, ARM_ARCH_MAJOR.ARM_ARCH_MINOR is set to 8.0 in ``defaults.mk``,
3943   if the platform makefile/build defines or uses the correct ARM_ARCH_MAJOR and
3944   ARM_ARCH_MINOR then mandatory Architectural features available for that Arch
3945   version will be enabled by default and any optional Arch feature supported by
3946   the Architecture and available in TF-A can be enabled from platform specific
3947   makefile. Look up to ``arch_features.mk`` for details pertaining to mandatory
3948   and optional Arch specific features.
3949
3950Platform include paths
3951----------------------
3952
3953Platforms are allowed to add more include paths to be passed to the compiler.
3954The ``PLAT_INCLUDES`` variable is used for this purpose. This is needed in
3955particular for the file ``platform_def.h``.
3956
3957Example:
3958
3959.. code:: c
3960
3961  PLAT_INCLUDES  += -Iinclude/plat/myplat/include
3962
3963C Library
3964---------
3965
3966To avoid subtle toolchain behavioral dependencies, the header files provided
3967by the compiler are not used. The software is built with the ``-nostdinc`` flag
3968to ensure no headers are included from the toolchain inadvertently. Instead the
3969required headers are included in the TF-A source tree. The library only
3970contains those C library definitions required by the local implementation. If
3971more functionality is required, the needed library functions will need to be
3972added to the local implementation.
3973
3974Some C headers have been obtained from `FreeBSD`_ and `SCC`_, while others have
3975been written specifically for TF-A. Some implementation files have been obtained
3976from `FreeBSD`_, others have been written specifically for TF-A as well. The
3977files can be found in ``include/lib/libc`` and ``lib/libc``.
3978
3979SCC can be found in http://www.simple-cc.org/. A copy of the `FreeBSD`_ sources
3980can be obtained from http://github.com/freebsd/freebsd.
3981
3982Storage abstraction layer
3983-------------------------
3984
3985In order to improve platform independence and portability a storage abstraction
3986layer is used to load data from non-volatile platform storage. Currently
3987storage access is only required by BL1 and BL2 phases and performed inside the
3988``load_image()`` function in ``bl_common.c``.
3989
3990.. uml:: resources/diagrams/plantuml/io_framework_usage_overview.puml
3991
3992It is mandatory to implement at least one storage driver. For the Arm
3993development platforms the Firmware Image Package (FIP) driver is provided as
3994the default means to load data from storage (see :ref:`firmware_design_fip`).
3995The storage layer is described in the header file
3996``include/drivers/io/io_storage.h``. The implementation of the common library is
3997in ``drivers/io/io_storage.c`` and the driver files are located in
3998``drivers/io/``.
3999
4000.. uml:: resources/diagrams/plantuml/io_arm_class_diagram.puml
4001
4002Each IO driver must provide ``io_dev_*`` structures, as described in
4003``drivers/io/io_driver.h``. These are returned via a mandatory registration
4004function that is called on platform initialization. The semi-hosting driver
4005implementation in ``io_semihosting.c`` can be used as an example.
4006
4007Each platform should register devices and their drivers via the storage
4008abstraction layer. These drivers then need to be initialized by bootloader
4009phases as required in their respective ``blx_platform_setup()`` functions.
4010
4011.. uml:: resources/diagrams/plantuml/io_dev_registration.puml
4012
4013The storage abstraction layer provides mechanisms (``io_dev_init()``) to
4014initialize storage devices before IO operations are called.
4015
4016.. uml:: resources/diagrams/plantuml/io_dev_init_and_check.puml
4017
4018The basic operations supported by the layer
4019include ``open()``, ``close()``, ``read()``, ``write()``, ``size()`` and ``seek()``.
4020Drivers do not have to implement all operations, but each platform must
4021provide at least one driver for a device capable of supporting generic
4022operations such as loading a bootloader image.
4023
4024The current implementation only allows for known images to be loaded by the
4025firmware. These images are specified by using their identifiers, as defined in
4026``include/plat/common/common_def.h`` (or a separate header file included from
4027there). The platform layer (``plat_get_image_source()``) then returns a reference
4028to a device and a driver-specific ``spec`` which will be understood by the driver
4029to allow access to the image data.
4030
4031The layer is designed in such a way that is it possible to chain drivers with
4032other drivers. For example, file-system drivers may be implemented on top of
4033physical block devices, both represented by IO devices with corresponding
4034drivers. In such a case, the file-system "binding" with the block device may
4035be deferred until the file-system device is initialised.
4036
4037The abstraction currently depends on structures being statically allocated
4038by the drivers and callers, as the system does not yet provide a means of
4039dynamically allocating memory. This may also have the affect of limiting the
4040amount of open resources per driver.
4041
4042Measured Boot Platform Interface
4043--------------------------------
4044
4045Enabling the MEASURED_BOOT flag adds extra platform requirements. Please refer
4046to :ref:`Measured Boot Design` for more details.
4047
4048Live Firmware Activation Interface
4049----------------------------------
4050
4051Function : plat_lfa_get_components()
4052~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
4053
4054::
4055
4056    Argument : plat_lfa_component_info_t **
4057    Return   : int
4058
4059This platform API provides the list of LFA components available for activation.
4060It populates a pointer to an array of ``plat_lfa_component_info_t`` structures,
4061which contain information about each component (like UUID, ID, etc.). It returns
40620 on success, or a standard error code on failure.
4063
4064Function : is_plat_lfa_activation_pending()
4065~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
4066
4067::
4068
4069    Argument : uint32_t
4070    Return   : bool
4071
4072This platform API checks if the specified LFA component, identified
4073by its ``lfa_component_id``, is available for activation. It returns
4074true if available, otherwise false.
4075
4076Function : plat_lfa_cancel()
4077~~~~~~~~~~~~~~~~~~~~~~~~~~~~
4078
4079::
4080
4081    Argument : uint32_t
4082    Return   : int
4083
4084This platform API allows the platform to cancel an ongoing update or activation
4085process for the specified ``lfa_component_id``. It returns 0 on success or
4086a standard error code on failure.
4087
4088Function : plat_lfa_load_auth_image()
4089~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
4090
4091::
4092
4093    Argument : uint32_t
4094    Return   : int
4095
4096The platform uses this API to load, authenticate and measure the component
4097specified by ``lfa_component_id``. It should return 0 on success or appropriate
4098error codes for load/authentication failures.
4099
4100NOTE: The error code -EAGAIN is not treated as an error, it means the operation
4101is incomplete and the function should be called again by the caller.
4102
4103Function : plat_lfa_notify_activate()
4104~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
4105
4106::
4107
4108    Argument : uint32_t
4109    Return   : int
4110
4111This API is invoked by the platform to notify its security engine to initiate
4112the required steps for component activation. The function takes the component
4113identifier ``lfa_component_id`` as an argument. It should return 0 on success
4114or appropriate negative error codes on failures.
4115
4116--------------
4117
4118*Copyright (c) 2013-2025, Arm Limited and Contributors. All rights reserved.*
4119
4120.. _PSCI: https://developer.arm.com/documentation/den0022/latest/
4121.. _Arm Generic Interrupt Controller version 2.0 (GICv2): https://developer.arm.com/documentation/ihi0048/b/
4122.. _3.0 (GICv3): https://developer.arm.com/documentation/ihi0069
4123.. _FreeBSD: https://www.freebsd.org
4124.. _SCC: http://www.simple-cc.org/
4125.. _DRTM: https://developer.arm.com/documentation/den0113
4126