1Firmware Design 2=============== 3 4Trusted Firmware-A (TF-A) implements a subset of the Trusted Board Boot 5Requirements (TBBR) Platform Design Document (PDD) [1]_ for Arm reference 6platforms. The TBB sequence starts when the platform is powered on and runs up 7to the stage where it hands-off control to firmware running in the normal 8world in DRAM. This is the cold boot path. 9 10TF-A also implements the Power State Coordination Interface PDD [2]_ as a 11runtime service. PSCI is the interface from normal world software to firmware 12implementing power management use-cases (for example, secondary CPU boot, 13hotplug and idle). Normal world software can access TF-A runtime services via 14the Arm SMC (Secure Monitor Call) instruction. The SMC instruction must be 15used as mandated by the SMC Calling Convention [3]_. 16 17TF-A implements a framework for configuring and managing interrupts generated 18in either security state. The details of the interrupt management framework 19and its design can be found in TF-A Interrupt Management Design guide [4]_. 20 21TF-A also implements a library for setting up and managing the translation 22tables. The details of this library can be found in `Xlat_tables design`_. 23 24TF-A can be built to support either AArch64 or AArch32 execution state. 25 26Cold boot 27--------- 28 29The cold boot path starts when the platform is physically turned on. If 30``COLD_BOOT_SINGLE_CPU=0``, one of the CPUs released from reset is chosen as the 31primary CPU, and the remaining CPUs are considered secondary CPUs. The primary 32CPU is chosen through platform-specific means. The cold boot path is mainly 33executed by the primary CPU, other than essential CPU initialization executed by 34all CPUs. The secondary CPUs are kept in a safe platform-specific state until 35the primary CPU has performed enough initialization to boot them. 36 37Refer to the `Reset Design`_ for more information on the effect of the 38``COLD_BOOT_SINGLE_CPU`` platform build option. 39 40The cold boot path in this implementation of TF-A depends on the execution 41state. For AArch64, it is divided into five steps (in order of execution): 42 43- Boot Loader stage 1 (BL1) *AP Trusted ROM* 44- Boot Loader stage 2 (BL2) *Trusted Boot Firmware* 45- Boot Loader stage 3-1 (BL31) *EL3 Runtime Software* 46- Boot Loader stage 3-2 (BL32) *Secure-EL1 Payload* (optional) 47- Boot Loader stage 3-3 (BL33) *Non-trusted Firmware* 48 49For AArch32, it is divided into four steps (in order of execution): 50 51- Boot Loader stage 1 (BL1) *AP Trusted ROM* 52- Boot Loader stage 2 (BL2) *Trusted Boot Firmware* 53- Boot Loader stage 3-2 (BL32) *EL3 Runtime Software* 54- Boot Loader stage 3-3 (BL33) *Non-trusted Firmware* 55 56Arm development platforms (Fixed Virtual Platforms (FVPs) and Juno) implement a 57combination of the following types of memory regions. Each bootloader stage uses 58one or more of these memory regions. 59 60- Regions accessible from both non-secure and secure states. For example, 61 non-trusted SRAM, ROM and DRAM. 62- Regions accessible from only the secure state. For example, trusted SRAM and 63 ROM. The FVPs also implement the trusted DRAM which is statically 64 configured. Additionally, the Base FVPs and Juno development platform 65 configure the TrustZone Controller (TZC) to create a region in the DRAM 66 which is accessible only from the secure state. 67 68The sections below provide the following details: 69 70- dynamic configuration of Boot Loader stages 71- initialization and execution of the first three stages during cold boot 72- specification of the EL3 Runtime Software (BL31 for AArch64 and BL32 for 73 AArch32) entrypoint requirements for use by alternative Trusted Boot 74 Firmware in place of the provided BL1 and BL2 75 76Dynamic Configuration during cold boot 77~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 78 79Each of the Boot Loader stages may be dynamically configured if required by the 80platform. The Boot Loader stage may optionally specify a firmware 81configuration file and/or hardware configuration file as listed below: 82 83- HW_CONFIG - The hardware configuration file. Can be shared by all Boot Loader 84 stages and also by the Normal World Rich OS. 85- TB_FW_CONFIG - Trusted Boot Firmware configuration file. Shared between BL1 86 and BL2. 87- SOC_FW_CONFIG - SoC Firmware configuration file. Used by BL31. 88- TOS_FW_CONFIG - Trusted OS Firmware configuration file. Used by Trusted OS 89 (BL32). 90- NT_FW_CONFIG - Non Trusted Firmware configuration file. Used by Non-trusted 91 firmware (BL33). 92 93The Arm development platforms use the Flattened Device Tree format for the 94dynamic configuration files. 95 96Each Boot Loader stage can pass up to 4 arguments via registers to the next 97stage. BL2 passes the list of the next images to execute to the *EL3 Runtime 98Software* (BL31 for AArch64 and BL32 for AArch32) via `arg0`. All the other 99arguments are platform defined. The Arm development platforms use the following 100convention: 101 102- BL1 passes the address of a meminfo_t structure to BL2 via ``arg1``. This 103 structure contains the memory layout available to BL2. 104- When dynamic configuration files are present, the firmware configuration for 105 the next Boot Loader stage is populated in the first available argument and 106 the generic hardware configuration is passed the next available argument. 107 For example, 108 109 - If TB_FW_CONFIG is loaded by BL1, then its address is passed in ``arg0`` 110 to BL2. 111 - If HW_CONFIG is loaded by BL1, then its address is passed in ``arg2`` to 112 BL2. Note, ``arg1`` is already used for meminfo_t. 113 - If SOC_FW_CONFIG is loaded by BL2, then its address is passed in ``arg1`` 114 to BL31. Note, ``arg0`` is used to pass the list of executable images. 115 - Similarly, if HW_CONFIG is loaded by BL1 or BL2, then its address is 116 passed in ``arg2`` to BL31. 117 - For other BL3x images, if the firmware configuration file is loaded by 118 BL2, then its address is passed in ``arg0`` and if HW_CONFIG is loaded 119 then its address is passed in ``arg1``. 120 121BL1 122~~~ 123 124This stage begins execution from the platform's reset vector at EL3. The reset 125address is platform dependent but it is usually located in a Trusted ROM area. 126The BL1 data section is copied to trusted SRAM at runtime. 127 128On the Arm development platforms, BL1 code starts execution from the reset 129vector defined by the constant ``BL1_RO_BASE``. The BL1 data section is copied 130to the top of trusted SRAM as defined by the constant ``BL1_RW_BASE``. 131 132The functionality implemented by this stage is as follows. 133 134Determination of boot path 135^^^^^^^^^^^^^^^^^^^^^^^^^^ 136 137Whenever a CPU is released from reset, BL1 needs to distinguish between a warm 138boot and a cold boot. This is done using platform-specific mechanisms (see the 139``plat_get_my_entrypoint()`` function in the `Porting Guide`_). In the case of a 140warm boot, a CPU is expected to continue execution from a separate 141entrypoint. In the case of a cold boot, the secondary CPUs are placed in a safe 142platform-specific state (see the ``plat_secondary_cold_boot_setup()`` function in 143the `Porting Guide`_) while the primary CPU executes the remaining cold boot path 144as described in the following sections. 145 146This step only applies when ``PROGRAMMABLE_RESET_ADDRESS=0``. Refer to the 147`Reset Design`_ for more information on the effect of the 148``PROGRAMMABLE_RESET_ADDRESS`` platform build option. 149 150Architectural initialization 151^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 152 153BL1 performs minimal architectural initialization as follows. 154 155- Exception vectors 156 157 BL1 sets up simple exception vectors for both synchronous and asynchronous 158 exceptions. The default behavior upon receiving an exception is to populate 159 a status code in the general purpose register ``X0/R0`` and call the 160 ``plat_report_exception()`` function (see the `Porting Guide`_). The status 161 code is one of: 162 163 For AArch64: 164 165 :: 166 167 0x0 : Synchronous exception from Current EL with SP_EL0 168 0x1 : IRQ exception from Current EL with SP_EL0 169 0x2 : FIQ exception from Current EL with SP_EL0 170 0x3 : System Error exception from Current EL with SP_EL0 171 0x4 : Synchronous exception from Current EL with SP_ELx 172 0x5 : IRQ exception from Current EL with SP_ELx 173 0x6 : FIQ exception from Current EL with SP_ELx 174 0x7 : System Error exception from Current EL with SP_ELx 175 0x8 : Synchronous exception from Lower EL using aarch64 176 0x9 : IRQ exception from Lower EL using aarch64 177 0xa : FIQ exception from Lower EL using aarch64 178 0xb : System Error exception from Lower EL using aarch64 179 0xc : Synchronous exception from Lower EL using aarch32 180 0xd : IRQ exception from Lower EL using aarch32 181 0xe : FIQ exception from Lower EL using aarch32 182 0xf : System Error exception from Lower EL using aarch32 183 184 For AArch32: 185 186 :: 187 188 0x10 : User mode 189 0x11 : FIQ mode 190 0x12 : IRQ mode 191 0x13 : SVC mode 192 0x16 : Monitor mode 193 0x17 : Abort mode 194 0x1a : Hypervisor mode 195 0x1b : Undefined mode 196 0x1f : System mode 197 198 The ``plat_report_exception()`` implementation on the Arm FVP port programs 199 the Versatile Express System LED register in the following format to 200 indicate the occurrence of an unexpected exception: 201 202 :: 203 204 SYS_LED[0] - Security state (Secure=0/Non-Secure=1) 205 SYS_LED[2:1] - Exception Level (EL3=0x3, EL2=0x2, EL1=0x1, EL0=0x0) 206 For AArch32 it is always 0x0 207 SYS_LED[7:3] - Exception Class (Sync/Async & origin). This is the value 208 of the status code 209 210 A write to the LED register reflects in the System LEDs (S6LED0..7) in the 211 CLCD window of the FVP. 212 213 BL1 does not expect to receive any exceptions other than the SMC exception. 214 For the latter, BL1 installs a simple stub. The stub expects to receive a 215 limited set of SMC types (determined by their function IDs in the general 216 purpose register ``X0/R0``): 217 218 - ``BL1_SMC_RUN_IMAGE``: This SMC is raised by BL2 to make BL1 pass control 219 to EL3 Runtime Software. 220 - All SMCs listed in section "BL1 SMC Interface" in the `Firmware Update`_ 221 Design Guide are supported for AArch64 only. These SMCs are currently 222 not supported when BL1 is built for AArch32. 223 224 Any other SMC leads to an assertion failure. 225 226- CPU initialization 227 228 BL1 calls the ``reset_handler()`` function which in turn calls the CPU 229 specific reset handler function (see the section: "CPU specific operations 230 framework"). 231 232- Control register setup (for AArch64) 233 234 - ``SCTLR_EL3``. Instruction cache is enabled by setting the ``SCTLR_EL3.I`` 235 bit. Alignment and stack alignment checking is enabled by setting the 236 ``SCTLR_EL3.A`` and ``SCTLR_EL3.SA`` bits. Exception endianness is set to 237 little-endian by clearing the ``SCTLR_EL3.EE`` bit. 238 239 - ``SCR_EL3``. The register width of the next lower exception level is set 240 to AArch64 by setting the ``SCR.RW`` bit. The ``SCR.EA`` bit is set to trap 241 both External Aborts and SError Interrupts in EL3. The ``SCR.SIF`` bit is 242 also set to disable instruction fetches from Non-secure memory when in 243 secure state. 244 245 - ``CPTR_EL3``. Accesses to the ``CPACR_EL1`` register from EL1 or EL2, or the 246 ``CPTR_EL2`` register from EL2 are configured to not trap to EL3 by 247 clearing the ``CPTR_EL3.TCPAC`` bit. Access to the trace functionality is 248 configured not to trap to EL3 by clearing the ``CPTR_EL3.TTA`` bit. 249 Instructions that access the registers associated with Floating Point 250 and Advanced SIMD execution are configured to not trap to EL3 by 251 clearing the ``CPTR_EL3.TFP`` bit. 252 253 - ``DAIF``. The SError interrupt is enabled by clearing the SError interrupt 254 mask bit. 255 256 - ``MDCR_EL3``. The trap controls, ``MDCR_EL3.TDOSA``, ``MDCR_EL3.TDA`` and 257 ``MDCR_EL3.TPM``, are set so that accesses to the registers they control 258 do not trap to EL3. AArch64 Secure self-hosted debug is disabled by 259 setting the ``MDCR_EL3.SDD`` bit. Also ``MDCR_EL3.SPD32`` is set to 260 disable AArch32 Secure self-hosted privileged debug from S-EL1. 261 262- Control register setup (for AArch32) 263 264 - ``SCTLR``. Instruction cache is enabled by setting the ``SCTLR.I`` bit. 265 Alignment checking is enabled by setting the ``SCTLR.A`` bit. 266 Exception endianness is set to little-endian by clearing the 267 ``SCTLR.EE`` bit. 268 269 - ``SCR``. The ``SCR.SIF`` bit is set to disable instruction fetches from 270 Non-secure memory when in secure state. 271 272 - ``CPACR``. Allow execution of Advanced SIMD instructions at PL0 and PL1, 273 by clearing the ``CPACR.ASEDIS`` bit. Access to the trace functionality 274 is configured not to trap to undefined mode by clearing the 275 ``CPACR.TRCDIS`` bit. 276 277 - ``NSACR``. Enable non-secure access to Advanced SIMD functionality and 278 system register access to implemented trace registers. 279 280 - ``FPEXC``. Enable access to the Advanced SIMD and floating-point 281 functionality from all Exception levels. 282 283 - ``CPSR.A``. The Asynchronous data abort interrupt is enabled by clearing 284 the Asynchronous data abort interrupt mask bit. 285 286 - ``SDCR``. The ``SDCR.SPD`` field is set to disable AArch32 Secure 287 self-hosted privileged debug. 288 289Platform initialization 290^^^^^^^^^^^^^^^^^^^^^^^ 291 292On Arm platforms, BL1 performs the following platform initializations: 293 294- Enable the Trusted Watchdog. 295- Initialize the console. 296- Configure the Interconnect to enable hardware coherency. 297- Enable the MMU and map the memory it needs to access. 298- Configure any required platform storage to load the next bootloader image 299 (BL2). 300- If the BL1 dynamic configuration file, ``TB_FW_CONFIG``, is available, then 301 load it to the platform defined address and make it available to BL2 via 302 ``arg0``. 303- Configure the system timer and program the `CNTFRQ_EL0` for use by NS-BL1U 304 and NS-BL2U firmware update images. 305 306Firmware Update detection and execution 307^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 308 309After performing platform setup, BL1 common code calls 310``bl1_plat_get_next_image_id()`` to determine if `Firmware Update`_ is required or 311to proceed with the normal boot process. If the platform code returns 312``BL2_IMAGE_ID`` then the normal boot sequence is executed as described in the 313next section, else BL1 assumes that `Firmware Update`_ is required and execution 314passes to the first image in the `Firmware Update`_ process. In either case, BL1 315retrieves a descriptor of the next image by calling ``bl1_plat_get_image_desc()``. 316The image descriptor contains an ``entry_point_info_t`` structure, which BL1 317uses to initialize the execution state of the next image. 318 319BL2 image load and execution 320^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 321 322In the normal boot flow, BL1 execution continues as follows: 323 324#. BL1 prints the following string from the primary CPU to indicate successful 325 execution of the BL1 stage: 326 327 :: 328 329 "Booting Trusted Firmware" 330 331#. BL1 loads a BL2 raw binary image from platform storage, at a 332 platform-specific base address. Prior to the load, BL1 invokes 333 ``bl1_plat_handle_pre_image_load()`` which allows the platform to update or 334 use the image information. If the BL2 image file is not present or if 335 there is not enough free trusted SRAM the following error message is 336 printed: 337 338 :: 339 340 "Failed to load BL2 firmware." 341 342#. BL1 invokes ``bl1_plat_handle_post_image_load()`` which again is intended 343 for platforms to take further action after image load. This function must 344 populate the necessary arguments for BL2, which may also include the memory 345 layout. Further description of the memory layout can be found later 346 in this document. 347 348#. BL1 passes control to the BL2 image at Secure EL1 (for AArch64) or at 349 Secure SVC mode (for AArch32), starting from its load address. 350 351BL2 352~~~ 353 354BL1 loads and passes control to BL2 at Secure-EL1 (for AArch64) or at Secure 355SVC mode (for AArch32) . BL2 is linked against and loaded at a platform-specific 356base address (more information can be found later in this document). 357The functionality implemented by BL2 is as follows. 358 359Architectural initialization 360^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 361 362For AArch64, BL2 performs the minimal architectural initialization required 363for subsequent stages of TF-A and normal world software. EL1 and EL0 are given 364access to Floating Point and Advanced SIMD registers by clearing the 365``CPACR.FPEN`` bits. 366 367For AArch32, the minimal architectural initialization required for subsequent 368stages of TF-A and normal world software is taken care of in BL1 as both BL1 369and BL2 execute at PL1. 370 371Platform initialization 372^^^^^^^^^^^^^^^^^^^^^^^ 373 374On Arm platforms, BL2 performs the following platform initializations: 375 376- Initialize the console. 377- Configure any required platform storage to allow loading further bootloader 378 images. 379- Enable the MMU and map the memory it needs to access. 380- Perform platform security setup to allow access to controlled components. 381- Reserve some memory for passing information to the next bootloader image 382 EL3 Runtime Software and populate it. 383- Define the extents of memory available for loading each subsequent 384 bootloader image. 385- If BL1 has passed TB_FW_CONFIG dynamic configuration file in ``arg0``, 386 then parse it. 387 388Image loading in BL2 389^^^^^^^^^^^^^^^^^^^^ 390 391BL2 generic code loads the images based on the list of loadable images 392provided by the platform. BL2 passes the list of executable images 393provided by the platform to the next handover BL image. 394 395The list of loadable images provided by the platform may also contain 396dynamic configuration files. The files are loaded and can be parsed as 397needed in the ``bl2_plat_handle_post_image_load()`` function. These 398configuration files can be passed to next Boot Loader stages as arguments 399by updating the corresponding entrypoint information in this function. 400 401SCP_BL2 (System Control Processor Firmware) image load 402^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 403 404Some systems have a separate System Control Processor (SCP) for power, clock, 405reset and system control. BL2 loads the optional SCP_BL2 image from platform 406storage into a platform-specific region of secure memory. The subsequent 407handling of SCP_BL2 is platform specific. For example, on the Juno Arm 408development platform port the image is transferred into SCP's internal memory 409using the Boot Over MHU (BOM) protocol after being loaded in the trusted SRAM 410memory. The SCP executes SCP_BL2 and signals to the Application Processor (AP) 411for BL2 execution to continue. 412 413EL3 Runtime Software image load 414^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 415 416BL2 loads the EL3 Runtime Software image from platform storage into a platform- 417specific address in trusted SRAM. If there is not enough memory to load the 418image or image is missing it leads to an assertion failure. 419 420AArch64 BL32 (Secure-EL1 Payload) image load 421^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 422 423BL2 loads the optional BL32 image from platform storage into a platform- 424specific region of secure memory. The image executes in the secure world. BL2 425relies on BL31 to pass control to the BL32 image, if present. Hence, BL2 426populates a platform-specific area of memory with the entrypoint/load-address 427of the BL32 image. The value of the Saved Processor Status Register (``SPSR``) 428for entry into BL32 is not determined by BL2, it is initialized by the 429Secure-EL1 Payload Dispatcher (see later) within BL31, which is responsible for 430managing interaction with BL32. This information is passed to BL31. 431 432BL33 (Non-trusted Firmware) image load 433^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 434 435BL2 loads the BL33 image (e.g. UEFI or other test or boot software) from 436platform storage into non-secure memory as defined by the platform. 437 438BL2 relies on EL3 Runtime Software to pass control to BL33 once secure state 439initialization is complete. Hence, BL2 populates a platform-specific area of 440memory with the entrypoint and Saved Program Status Register (``SPSR``) of the 441normal world software image. The entrypoint is the load address of the BL33 442image. The ``SPSR`` is determined as specified in Section 5.13 of the 443`PSCI PDD`_. This information is passed to the EL3 Runtime Software. 444 445AArch64 BL31 (EL3 Runtime Software) execution 446^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 447 448BL2 execution continues as follows: 449 450#. BL2 passes control back to BL1 by raising an SMC, providing BL1 with the 451 BL31 entrypoint. The exception is handled by the SMC exception handler 452 installed by BL1. 453 454#. BL1 turns off the MMU and flushes the caches. It clears the 455 ``SCTLR_EL3.M/I/C`` bits, flushes the data cache to the point of coherency 456 and invalidates the TLBs. 457 458#. BL1 passes control to BL31 at the specified entrypoint at EL3. 459 460Running BL2 at EL3 execution level 461~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 462 463Some platforms have a non-TF-A Boot ROM that expects the next boot stage 464to execute at EL3. On these platforms, TF-A BL1 is a waste of memory 465as its only purpose is to ensure TF-A BL2 is entered at S-EL1. To avoid 466this waste, a special mode enables BL2 to execute at EL3, which allows 467a non-TF-A Boot ROM to load and jump directly to BL2. This mode is selected 468when the build flag BL2_AT_EL3 is enabled. The main differences in this 469mode are: 470 471#. BL2 includes the reset code and the mailbox mechanism to differentiate 472 cold boot and warm boot. It runs at EL3 doing the arch 473 initialization required for EL3. 474 475#. BL2 does not receive the meminfo information from BL1 anymore. This 476 information can be passed by the Boot ROM or be internal to the 477 BL2 image. 478 479#. Since BL2 executes at EL3, BL2 jumps directly to the next image, 480 instead of invoking the RUN_IMAGE SMC call. 481 482 483We assume 3 different types of BootROM support on the platform: 484 485#. The Boot ROM always jumps to the same address, for both cold 486 and warm boot. In this case, we will need to keep a resident part 487 of BL2 whose memory cannot be reclaimed by any other image. The 488 linker script defines the symbols __TEXT_RESIDENT_START__ and 489 __TEXT_RESIDENT_END__ that allows the platform to configure 490 correctly the memory map. 491#. The platform has some mechanism to indicate the jump address to the 492 Boot ROM. Platform code can then program the jump address with 493 psci_warmboot_entrypoint during cold boot. 494#. The platform has some mechanism to program the reset address using 495 the PROGRAMMABLE_RESET_ADDRESS feature. Platform code can then 496 program the reset address with psci_warmboot_entrypoint during 497 cold boot, bypassing the boot ROM for warm boot. 498 499In the last 2 cases, no part of BL2 needs to remain resident at 500runtime. In the first 2 cases, we expect the Boot ROM to be able to 501differentiate between warm and cold boot, to avoid loading BL2 again 502during warm boot. 503 504This functionality can be tested with FVP loading the image directly 505in memory and changing the address where the system jumps at reset. 506For example: 507 508 -C cluster0.cpu0.RVBAR=0x4022000 509 --data cluster0.cpu0=bl2.bin@0x4022000 510 511With this configuration, FVP is like a platform of the first case, 512where the Boot ROM jumps always to the same address. For simplification, 513BL32 is loaded in DRAM in this case, to avoid other images reclaiming 514BL2 memory. 515 516 517AArch64 BL31 518~~~~~~~~~~~~ 519 520The image for this stage is loaded by BL2 and BL1 passes control to BL31 at 521EL3. BL31 executes solely in trusted SRAM. BL31 is linked against and 522loaded at a platform-specific base address (more information can be found later 523in this document). The functionality implemented by BL31 is as follows. 524 525Architectural initialization 526^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 527 528Currently, BL31 performs a similar architectural initialization to BL1 as 529far as system register settings are concerned. Since BL1 code resides in ROM, 530architectural initialization in BL31 allows override of any previous 531initialization done by BL1. 532 533BL31 initializes the per-CPU data framework, which provides a cache of 534frequently accessed per-CPU data optimised for fast, concurrent manipulation 535on different CPUs. This buffer includes pointers to per-CPU contexts, crash 536buffer, CPU reset and power down operations, PSCI data, platform data and so on. 537 538It then replaces the exception vectors populated by BL1 with its own. BL31 539exception vectors implement more elaborate support for handling SMCs since this 540is the only mechanism to access the runtime services implemented by BL31 (PSCI 541for example). BL31 checks each SMC for validity as specified by the 542`SMC calling convention PDD`_ before passing control to the required SMC 543handler routine. 544 545BL31 programs the ``CNTFRQ_EL0`` register with the clock frequency of the system 546counter, which is provided by the platform. 547 548Platform initialization 549^^^^^^^^^^^^^^^^^^^^^^^ 550 551BL31 performs detailed platform initialization, which enables normal world 552software to function correctly. 553 554On Arm platforms, this consists of the following: 555 556- Initialize the console. 557- Configure the Interconnect to enable hardware coherency. 558- Enable the MMU and map the memory it needs to access. 559- Initialize the generic interrupt controller. 560- Initialize the power controller device. 561- Detect the system topology. 562 563Runtime services initialization 564^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 565 566BL31 is responsible for initializing the runtime services. One of them is PSCI. 567 568As part of the PSCI initializations, BL31 detects the system topology. It also 569initializes the data structures that implement the state machine used to track 570the state of power domain nodes. The state can be one of ``OFF``, ``RUN`` or 571``RETENTION``. All secondary CPUs are initially in the ``OFF`` state. The cluster 572that the primary CPU belongs to is ``ON``; any other cluster is ``OFF``. It also 573initializes the locks that protect them. BL31 accesses the state of a CPU or 574cluster immediately after reset and before the data cache is enabled in the 575warm boot path. It is not currently possible to use 'exclusive' based spinlocks, 576therefore BL31 uses locks based on Lamport's Bakery algorithm instead. 577 578The runtime service framework and its initialization is described in more 579detail in the "EL3 runtime services framework" section below. 580 581Details about the status of the PSCI implementation are provided in the 582"Power State Coordination Interface" section below. 583 584AArch64 BL32 (Secure-EL1 Payload) image initialization 585^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 586 587If a BL32 image is present then there must be a matching Secure-EL1 Payload 588Dispatcher (SPD) service (see later for details). During initialization 589that service must register a function to carry out initialization of BL32 590once the runtime services are fully initialized. BL31 invokes such a 591registered function to initialize BL32 before running BL33. This initialization 592is not necessary for AArch32 SPs. 593 594Details on BL32 initialization and the SPD's role are described in the 595"Secure-EL1 Payloads and Dispatchers" section below. 596 597BL33 (Non-trusted Firmware) execution 598^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 599 600EL3 Runtime Software initializes the EL2 or EL1 processor context for normal- 601world cold boot, ensuring that no secure state information finds its way into 602the non-secure execution state. EL3 Runtime Software uses the entrypoint 603information provided by BL2 to jump to the Non-trusted firmware image (BL33) 604at the highest available Exception Level (EL2 if available, otherwise EL1). 605 606Using alternative Trusted Boot Firmware in place of BL1 & BL2 (AArch64 only) 607~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 608 609Some platforms have existing implementations of Trusted Boot Firmware that 610would like to use TF-A BL31 for the EL3 Runtime Software. To enable this 611firmware architecture it is important to provide a fully documented and stable 612interface between the Trusted Boot Firmware and BL31. 613 614Future changes to the BL31 interface will be done in a backwards compatible 615way, and this enables these firmware components to be independently enhanced/ 616updated to develop and exploit new functionality. 617 618Required CPU state when calling ``bl31_entrypoint()`` during cold boot 619^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 620 621This function must only be called by the primary CPU. 622 623On entry to this function the calling primary CPU must be executing in AArch64 624EL3, little-endian data access, and all interrupt sources masked: 625 626:: 627 628 PSTATE.EL = 3 629 PSTATE.RW = 1 630 PSTATE.DAIF = 0xf 631 SCTLR_EL3.EE = 0 632 633X0 and X1 can be used to pass information from the Trusted Boot Firmware to the 634platform code in BL31: 635 636:: 637 638 X0 : Reserved for common TF-A information 639 X1 : Platform specific information 640 641BL31 zero-init sections (e.g. ``.bss``) should not contain valid data on entry, 642these will be zero filled prior to invoking platform setup code. 643 644Use of the X0 and X1 parameters 645''''''''''''''''''''''''''''''' 646 647The parameters are platform specific and passed from ``bl31_entrypoint()`` to 648``bl31_early_platform_setup()``. The value of these parameters is never directly 649used by the common BL31 code. 650 651The convention is that ``X0`` conveys information regarding the BL31, BL32 and 652BL33 images from the Trusted Boot firmware and ``X1`` can be used for other 653platform specific purpose. This convention allows platforms which use TF-A's 654BL1 and BL2 images to transfer additional platform specific information from 655Secure Boot without conflicting with future evolution of TF-A using ``X0`` to 656pass a ``bl31_params`` structure. 657 658BL31 common and SPD initialization code depends on image and entrypoint 659information about BL33 and BL32, which is provided via BL31 platform APIs. 660This information is required until the start of execution of BL33. This 661information can be provided in a platform defined manner, e.g. compiled into 662the platform code in BL31, or provided in a platform defined memory location 663by the Trusted Boot firmware, or passed from the Trusted Boot Firmware via the 664Cold boot Initialization parameters. This data may need to be cleaned out of 665the CPU caches if it is provided by an earlier boot stage and then accessed by 666BL31 platform code before the caches are enabled. 667 668TF-A's BL2 implementation passes a ``bl31_params`` structure in 669``X0`` and the Arm development platforms interpret this in the BL31 platform 670code. 671 672MMU, Data caches & Coherency 673'''''''''''''''''''''''''''' 674 675BL31 does not depend on the enabled state of the MMU, data caches or 676interconnect coherency on entry to ``bl31_entrypoint()``. If these are disabled 677on entry, these should be enabled during ``bl31_plat_arch_setup()``. 678 679Data structures used in the BL31 cold boot interface 680'''''''''''''''''''''''''''''''''''''''''''''''''''' 681 682These structures are designed to support compatibility and independent 683evolution of the structures and the firmware images. For example, a version of 684BL31 that can interpret the BL3x image information from different versions of 685BL2, a platform that uses an extended entry_point_info structure to convey 686additional register information to BL31, or a ELF image loader that can convey 687more details about the firmware images. 688 689To support these scenarios the structures are versioned and sized, which enables 690BL31 to detect which information is present and respond appropriately. The 691``param_header`` is defined to capture this information: 692 693.. code:: c 694 695 typedef struct param_header { 696 uint8_t type; /* type of the structure */ 697 uint8_t version; /* version of this structure */ 698 uint16_t size; /* size of this structure in bytes */ 699 uint32_t attr; /* attributes: unused bits SBZ */ 700 } param_header_t; 701 702The structures using this format are ``entry_point_info``, ``image_info`` and 703``bl31_params``. The code that allocates and populates these structures must set 704the header fields appropriately, and the ``SET_PARAM_HEAD()`` a macro is defined 705to simplify this action. 706 707Required CPU state for BL31 Warm boot initialization 708^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 709 710When requesting a CPU power-on, or suspending a running CPU, TF-A provides 711the platform power management code with a Warm boot initialization 712entry-point, to be invoked by the CPU immediately after the reset handler. 713On entry to the Warm boot initialization function the calling CPU must be in 714AArch64 EL3, little-endian data access and all interrupt sources masked: 715 716:: 717 718 PSTATE.EL = 3 719 PSTATE.RW = 1 720 PSTATE.DAIF = 0xf 721 SCTLR_EL3.EE = 0 722 723The PSCI implementation will initialize the processor state and ensure that the 724platform power management code is then invoked as required to initialize all 725necessary system, cluster and CPU resources. 726 727AArch32 EL3 Runtime Software entrypoint interface 728~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 729 730To enable this firmware architecture it is important to provide a fully 731documented and stable interface between the Trusted Boot Firmware and the 732AArch32 EL3 Runtime Software. 733 734Future changes to the entrypoint interface will be done in a backwards 735compatible way, and this enables these firmware components to be independently 736enhanced/updated to develop and exploit new functionality. 737 738Required CPU state when entering during cold boot 739^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 740 741This function must only be called by the primary CPU. 742 743On entry to this function the calling primary CPU must be executing in AArch32 744EL3, little-endian data access, and all interrupt sources masked: 745 746:: 747 748 PSTATE.AIF = 0x7 749 SCTLR.EE = 0 750 751R0 and R1 are used to pass information from the Trusted Boot Firmware to the 752platform code in AArch32 EL3 Runtime Software: 753 754:: 755 756 R0 : Reserved for common TF-A information 757 R1 : Platform specific information 758 759Use of the R0 and R1 parameters 760''''''''''''''''''''''''''''''' 761 762The parameters are platform specific and the convention is that ``R0`` conveys 763information regarding the BL3x images from the Trusted Boot firmware and ``R1`` 764can be used for other platform specific purpose. This convention allows 765platforms which use TF-A's BL1 and BL2 images to transfer additional platform 766specific information from Secure Boot without conflicting with future 767evolution of TF-A using ``R0`` to pass a ``bl_params`` structure. 768 769The AArch32 EL3 Runtime Software is responsible for entry into BL33. This 770information can be obtained in a platform defined manner, e.g. compiled into 771the AArch32 EL3 Runtime Software, or provided in a platform defined memory 772location by the Trusted Boot firmware, or passed from the Trusted Boot Firmware 773via the Cold boot Initialization parameters. This data may need to be cleaned 774out of the CPU caches if it is provided by an earlier boot stage and then 775accessed by AArch32 EL3 Runtime Software before the caches are enabled. 776 777When using AArch32 EL3 Runtime Software, the Arm development platforms pass a 778``bl_params`` structure in ``R0`` from BL2 to be interpreted by AArch32 EL3 Runtime 779Software platform code. 780 781MMU, Data caches & Coherency 782'''''''''''''''''''''''''''' 783 784AArch32 EL3 Runtime Software must not depend on the enabled state of the MMU, 785data caches or interconnect coherency in its entrypoint. They must be explicitly 786enabled if required. 787 788Data structures used in cold boot interface 789''''''''''''''''''''''''''''''''''''''''''' 790 791The AArch32 EL3 Runtime Software cold boot interface uses ``bl_params`` instead 792of ``bl31_params``. The ``bl_params`` structure is based on the convention 793described in AArch64 BL31 cold boot interface section. 794 795Required CPU state for warm boot initialization 796^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 797 798When requesting a CPU power-on, or suspending a running CPU, AArch32 EL3 799Runtime Software must ensure execution of a warm boot initialization entrypoint. 800If TF-A BL1 is used and the PROGRAMMABLE_RESET_ADDRESS build flag is false, 801then AArch32 EL3 Runtime Software must ensure that BL1 branches to the warm 802boot entrypoint by arranging for the BL1 platform function, 803plat_get_my_entrypoint(), to return a non-zero value. 804 805In this case, the warm boot entrypoint must be in AArch32 EL3, little-endian 806data access and all interrupt sources masked: 807 808:: 809 810 PSTATE.AIF = 0x7 811 SCTLR.EE = 0 812 813The warm boot entrypoint may be implemented by using TF-A 814``psci_warmboot_entrypoint()`` function. In that case, the platform must fulfil 815the pre-requisites mentioned in the `PSCI Library integration guide`_. 816 817EL3 runtime services framework 818------------------------------ 819 820Software executing in the non-secure state and in the secure state at exception 821levels lower than EL3 will request runtime services using the Secure Monitor 822Call (SMC) instruction. These requests will follow the convention described in 823the SMC Calling Convention PDD (`SMCCC`_). The `SMCCC`_ assigns function 824identifiers to each SMC request and describes how arguments are passed and 825returned. 826 827The EL3 runtime services framework enables the development of services by 828different providers that can be easily integrated into final product firmware. 829The following sections describe the framework which facilitates the 830registration, initialization and use of runtime services in EL3 Runtime 831Software (BL31). 832 833The design of the runtime services depends heavily on the concepts and 834definitions described in the `SMCCC`_, in particular SMC Function IDs, Owning 835Entity Numbers (OEN), Fast and Yielding calls, and the SMC32 and SMC64 calling 836conventions. Please refer to that document for more detailed explanation of 837these terms. 838 839The following runtime services are expected to be implemented first. They have 840not all been instantiated in the current implementation. 841 842#. Standard service calls 843 844 This service is for management of the entire system. The Power State 845 Coordination Interface (`PSCI`_) is the first set of standard service calls 846 defined by Arm (see PSCI section later). 847 848#. Secure-EL1 Payload Dispatcher service 849 850 If a system runs a Trusted OS or other Secure-EL1 Payload (SP) then 851 it also requires a *Secure Monitor* at EL3 to switch the EL1 processor 852 context between the normal world (EL1/EL2) and trusted world (Secure-EL1). 853 The Secure Monitor will make these world switches in response to SMCs. The 854 `SMCCC`_ provides for such SMCs with the Trusted OS Call and Trusted 855 Application Call OEN ranges. 856 857 The interface between the EL3 Runtime Software and the Secure-EL1 Payload is 858 not defined by the `SMCCC`_ or any other standard. As a result, each 859 Secure-EL1 Payload requires a specific Secure Monitor that runs as a runtime 860 service - within TF-A this service is referred to as the Secure-EL1 Payload 861 Dispatcher (SPD). 862 863 TF-A provides a Test Secure-EL1 Payload (TSP) and its associated Dispatcher 864 (TSPD). Details of SPD design and TSP/TSPD operation are described in the 865 "Secure-EL1 Payloads and Dispatchers" section below. 866 867#. CPU implementation service 868 869 This service will provide an interface to CPU implementation specific 870 services for a given platform e.g. access to processor errata workarounds. 871 This service is currently unimplemented. 872 873Additional services for Arm Architecture, SiP and OEM calls can be implemented. 874Each implemented service handles a range of SMC function identifiers as 875described in the `SMCCC`_. 876 877Registration 878~~~~~~~~~~~~ 879 880A runtime service is registered using the ``DECLARE_RT_SVC()`` macro, specifying 881the name of the service, the range of OENs covered, the type of service and 882initialization and call handler functions. This macro instantiates a ``const struct rt_svc_desc`` for the service with these details (see ``runtime_svc.h``). 883This structure is allocated in a special ELF section ``rt_svc_descs``, enabling 884the framework to find all service descriptors included into BL31. 885 886The specific service for a SMC Function is selected based on the OEN and call 887type of the Function ID, and the framework uses that information in the service 888descriptor to identify the handler for the SMC Call. 889 890The service descriptors do not include information to identify the precise set 891of SMC function identifiers supported by this service implementation, the 892security state from which such calls are valid nor the capability to support 89364-bit and/or 32-bit callers (using SMC32 or SMC64). Responding appropriately 894to these aspects of a SMC call is the responsibility of the service 895implementation, the framework is focused on integration of services from 896different providers and minimizing the time taken by the framework before the 897service handler is invoked. 898 899Details of the parameters, requirements and behavior of the initialization and 900call handling functions are provided in the following sections. 901 902Initialization 903~~~~~~~~~~~~~~ 904 905``runtime_svc_init()`` in ``runtime_svc.c`` initializes the runtime services 906framework running on the primary CPU during cold boot as part of the BL31 907initialization. This happens prior to initializing a Trusted OS and running 908Normal world boot firmware that might in turn use these services. 909Initialization involves validating each of the declared runtime service 910descriptors, calling the service initialization function and populating the 911index used for runtime lookup of the service. 912 913The BL31 linker script collects all of the declared service descriptors into a 914single array and defines symbols that allow the framework to locate and traverse 915the array, and determine its size. 916 917The framework does basic validation of each descriptor to halt firmware 918initialization if service declaration errors are detected. The framework does 919not check descriptors for the following error conditions, and may behave in an 920unpredictable manner under such scenarios: 921 922#. Overlapping OEN ranges 923#. Multiple descriptors for the same range of OENs and ``call_type`` 924#. Incorrect range of owning entity numbers for a given ``call_type`` 925 926Once validated, the service ``init()`` callback is invoked. This function carries 927out any essential EL3 initialization before servicing requests. The ``init()`` 928function is only invoked on the primary CPU during cold boot. If the service 929uses per-CPU data this must either be initialized for all CPUs during this call, 930or be done lazily when a CPU first issues an SMC call to that service. If 931``init()`` returns anything other than ``0``, this is treated as an initialization 932error and the service is ignored: this does not cause the firmware to halt. 933 934The OEN and call type fields present in the SMC Function ID cover a total of 935128 distinct services, but in practice a single descriptor can cover a range of 936OENs, e.g. SMCs to call a Trusted OS function. To optimize the lookup of a 937service handler, the framework uses an array of 128 indices that map every 938distinct OEN/call-type combination either to one of the declared services or to 939indicate the service is not handled. This ``rt_svc_descs_indices[]`` array is 940populated for all of the OENs covered by a service after the service ``init()`` 941function has reported success. So a service that fails to initialize will never 942have it's ``handle()`` function invoked. 943 944The following figure shows how the ``rt_svc_descs_indices[]`` index maps the SMC 945Function ID call type and OEN onto a specific service handler in the 946``rt_svc_descs[]`` array. 947 948|Image 1| 949 950Handling an SMC 951~~~~~~~~~~~~~~~ 952 953When the EL3 runtime services framework receives a Secure Monitor Call, the SMC 954Function ID is passed in W0 from the lower exception level (as per the 955`SMCCC`_). If the calling register width is AArch32, it is invalid to invoke an 956SMC Function which indicates the SMC64 calling convention: such calls are 957ignored and return the Unknown SMC Function Identifier result code ``0xFFFFFFFF`` 958in R0/X0. 959 960Bit[31] (fast/yielding call) and bits[29:24] (owning entity number) of the SMC 961Function ID are combined to index into the ``rt_svc_descs_indices[]`` array. The 962resulting value might indicate a service that has no handler, in this case the 963framework will also report an Unknown SMC Function ID. Otherwise, the value is 964used as a further index into the ``rt_svc_descs[]`` array to locate the required 965service and handler. 966 967The service's ``handle()`` callback is provided with five of the SMC parameters 968directly, the others are saved into memory for retrieval (if needed) by the 969handler. The handler is also provided with an opaque ``handle`` for use with the 970supporting library for parameter retrieval, setting return values and context 971manipulation; and with ``flags`` indicating the security state of the caller. The 972framework finally sets up the execution stack for the handler, and invokes the 973services ``handle()`` function. 974 975On return from the handler the result registers are populated in X0-X3 before 976restoring the stack and CPU state and returning from the original SMC. 977 978Exception Handling Framework 979---------------------------- 980 981Please refer to the `Exception Handling Framework`_ document. 982 983Power State Coordination Interface 984---------------------------------- 985 986TODO: Provide design walkthrough of PSCI implementation. 987 988The PSCI v1.1 specification categorizes APIs as optional and mandatory. All the 989mandatory APIs in PSCI v1.1, PSCI v1.0 and in PSCI v0.2 draft specification 990`Power State Coordination Interface PDD`_ are implemented. The table lists 991the PSCI v1.1 APIs and their support in generic code. 992 993An API implementation might have a dependency on platform code e.g. CPU_SUSPEND 994requires the platform to export a part of the implementation. Hence the level 995of support of the mandatory APIs depends upon the support exported by the 996platform port as well. The Juno and FVP (all variants) platforms export all the 997required support. 998 999+-----------------------------+-------------+-------------------------------+ 1000| PSCI v1.1 API | Supported | Comments | 1001+=============================+=============+===============================+ 1002| ``PSCI_VERSION`` | Yes | The version returned is 1.1 | 1003+-----------------------------+-------------+-------------------------------+ 1004| ``CPU_SUSPEND`` | Yes\* | | 1005+-----------------------------+-------------+-------------------------------+ 1006| ``CPU_OFF`` | Yes\* | | 1007+-----------------------------+-------------+-------------------------------+ 1008| ``CPU_ON`` | Yes\* | | 1009+-----------------------------+-------------+-------------------------------+ 1010| ``AFFINITY_INFO`` | Yes | | 1011+-----------------------------+-------------+-------------------------------+ 1012| ``MIGRATE`` | Yes\*\* | | 1013+-----------------------------+-------------+-------------------------------+ 1014| ``MIGRATE_INFO_TYPE`` | Yes\*\* | | 1015+-----------------------------+-------------+-------------------------------+ 1016| ``MIGRATE_INFO_CPU`` | Yes\*\* | | 1017+-----------------------------+-------------+-------------------------------+ 1018| ``SYSTEM_OFF`` | Yes\* | | 1019+-----------------------------+-------------+-------------------------------+ 1020| ``SYSTEM_RESET`` | Yes\* | | 1021+-----------------------------+-------------+-------------------------------+ 1022| ``PSCI_FEATURES`` | Yes | | 1023+-----------------------------+-------------+-------------------------------+ 1024| ``CPU_FREEZE`` | No | | 1025+-----------------------------+-------------+-------------------------------+ 1026| ``CPU_DEFAULT_SUSPEND`` | No | | 1027+-----------------------------+-------------+-------------------------------+ 1028| ``NODE_HW_STATE`` | Yes\* | | 1029+-----------------------------+-------------+-------------------------------+ 1030| ``SYSTEM_SUSPEND`` | Yes\* | | 1031+-----------------------------+-------------+-------------------------------+ 1032| ``PSCI_SET_SUSPEND_MODE`` | No | | 1033+-----------------------------+-------------+-------------------------------+ 1034| ``PSCI_STAT_RESIDENCY`` | Yes\* | | 1035+-----------------------------+-------------+-------------------------------+ 1036| ``PSCI_STAT_COUNT`` | Yes\* | | 1037+-----------------------------+-------------+-------------------------------+ 1038| ``SYSTEM_RESET2`` | Yes\* | | 1039+-----------------------------+-------------+-------------------------------+ 1040| ``MEM_PROTECT`` | Yes\* | | 1041+-----------------------------+-------------+-------------------------------+ 1042| ``MEM_PROTECT_CHECK_RANGE`` | Yes\* | | 1043+-----------------------------+-------------+-------------------------------+ 1044 1045\*Note : These PSCI APIs require platform power management hooks to be 1046registered with the generic PSCI code to be supported. 1047 1048\*\*Note : These PSCI APIs require appropriate Secure Payload Dispatcher 1049hooks to be registered with the generic PSCI code to be supported. 1050 1051The PSCI implementation in TF-A is a library which can be integrated with 1052AArch64 or AArch32 EL3 Runtime Software for Armv8-A systems. A guide to 1053integrating PSCI library with AArch32 EL3 Runtime Software can be found 1054`here`_. 1055 1056Secure-EL1 Payloads and Dispatchers 1057----------------------------------- 1058 1059On a production system that includes a Trusted OS running in Secure-EL1/EL0, 1060the Trusted OS is coupled with a companion runtime service in the BL31 1061firmware. This service is responsible for the initialisation of the Trusted 1062OS and all communications with it. The Trusted OS is the BL32 stage of the 1063boot flow in TF-A. The firmware will attempt to locate, load and execute a 1064BL32 image. 1065 1066TF-A uses a more general term for the BL32 software that runs at Secure-EL1 - 1067the *Secure-EL1 Payload* - as it is not always a Trusted OS. 1068 1069TF-A provides a Test Secure-EL1 Payload (TSP) and a Test Secure-EL1 Payload 1070Dispatcher (TSPD) service as an example of how a Trusted OS is supported on a 1071production system using the Runtime Services Framework. On such a system, the 1072Test BL32 image and service are replaced by the Trusted OS and its dispatcher 1073service. The TF-A build system expects that the dispatcher will define the 1074build flag ``NEED_BL32`` to enable it to include the BL32 in the build either 1075as a binary or to compile from source depending on whether the ``BL32`` build 1076option is specified or not. 1077 1078The TSP runs in Secure-EL1. It is designed to demonstrate synchronous 1079communication with the normal-world software running in EL1/EL2. Communication 1080is initiated by the normal-world software 1081 1082- either directly through a Fast SMC (as defined in the `SMCCC`_) 1083 1084- or indirectly through a `PSCI`_ SMC. The `PSCI`_ implementation in turn 1085 informs the TSPD about the requested power management operation. This allows 1086 the TSP to prepare for or respond to the power state change 1087 1088The TSPD service is responsible for. 1089 1090- Initializing the TSP 1091 1092- Routing requests and responses between the secure and the non-secure 1093 states during the two types of communications just described 1094 1095Initializing a BL32 Image 1096~~~~~~~~~~~~~~~~~~~~~~~~~ 1097 1098The Secure-EL1 Payload Dispatcher (SPD) service is responsible for initializing 1099the BL32 image. It needs access to the information passed by BL2 to BL31 to do 1100so. This is provided by: 1101 1102.. code:: c 1103 1104 entry_point_info_t *bl31_plat_get_next_image_ep_info(uint32_t); 1105 1106which returns a reference to the ``entry_point_info`` structure corresponding to 1107the image which will be run in the specified security state. The SPD uses this 1108API to get entry point information for the SECURE image, BL32. 1109 1110In the absence of a BL32 image, BL31 passes control to the normal world 1111bootloader image (BL33). When the BL32 image is present, it is typical 1112that the SPD wants control to be passed to BL32 first and then later to BL33. 1113 1114To do this the SPD has to register a BL32 initialization function during 1115initialization of the SPD service. The BL32 initialization function has this 1116prototype: 1117 1118.. code:: c 1119 1120 int32_t init(void); 1121 1122and is registered using the ``bl31_register_bl32_init()`` function. 1123 1124TF-A supports two approaches for the SPD to pass control to BL32 before 1125returning through EL3 and running the non-trusted firmware (BL33): 1126 1127#. In the BL32 setup function, use ``bl31_set_next_image_type()`` to 1128 request that the exit from ``bl31_main()`` is to the BL32 entrypoint in 1129 Secure-EL1. BL31 will exit to BL32 using the asynchronous method by 1130 calling ``bl31_prepare_next_image_entry()`` and ``el3_exit()``. 1131 1132 When the BL32 has completed initialization at Secure-EL1, it returns to 1133 BL31 by issuing an SMC, using a Function ID allocated to the SPD. On 1134 receipt of this SMC, the SPD service handler should switch the CPU context 1135 from trusted to normal world and use the ``bl31_set_next_image_type()`` and 1136 ``bl31_prepare_next_image_entry()`` functions to set up the initial return to 1137 the normal world firmware BL33. On return from the handler the framework 1138 will exit to EL2 and run BL33. 1139 1140#. The BL32 setup function registers an initialization function using 1141 ``bl31_register_bl32_init()`` which provides a SPD-defined mechanism to 1142 invoke a 'world-switch synchronous call' to Secure-EL1 to run the BL32 1143 entrypoint. 1144 NOTE: The Test SPD service included with TF-A provides one implementation 1145 of such a mechanism. 1146 1147 On completion BL32 returns control to BL31 via a SMC, and on receipt the 1148 SPD service handler invokes the synchronous call return mechanism to return 1149 to the BL32 initialization function. On return from this function, 1150 ``bl31_main()`` will set up the return to the normal world firmware BL33 and 1151 continue the boot process in the normal world. 1152 1153Crash Reporting in BL31 1154----------------------- 1155 1156BL31 implements a scheme for reporting the processor state when an unhandled 1157exception is encountered. The reporting mechanism attempts to preserve all the 1158register contents and report it via a dedicated UART (PL011 console). BL31 1159reports the general purpose, EL3, Secure EL1 and some EL2 state registers. 1160 1161A dedicated per-CPU crash stack is maintained by BL31 and this is retrieved via 1162the per-CPU pointer cache. The implementation attempts to minimise the memory 1163required for this feature. The file ``crash_reporting.S`` contains the 1164implementation for crash reporting. 1165 1166The sample crash output is shown below. 1167 1168:: 1169 1170 x0 :0x000000004F00007C 1171 x1 :0x0000000007FFFFFF 1172 x2 :0x0000000004014D50 1173 x3 :0x0000000000000000 1174 x4 :0x0000000088007998 1175 x5 :0x00000000001343AC 1176 x6 :0x0000000000000016 1177 x7 :0x00000000000B8A38 1178 x8 :0x00000000001343AC 1179 x9 :0x00000000000101A8 1180 x10 :0x0000000000000002 1181 x11 :0x000000000000011C 1182 x12 :0x00000000FEFDC644 1183 x13 :0x00000000FED93FFC 1184 x14 :0x0000000000247950 1185 x15 :0x00000000000007A2 1186 x16 :0x00000000000007A4 1187 x17 :0x0000000000247950 1188 x18 :0x0000000000000000 1189 x19 :0x00000000FFFFFFFF 1190 x20 :0x0000000004014D50 1191 x21 :0x000000000400A38C 1192 x22 :0x0000000000247950 1193 x23 :0x0000000000000010 1194 x24 :0x0000000000000024 1195 x25 :0x00000000FEFDC868 1196 x26 :0x00000000FEFDC86A 1197 x27 :0x00000000019EDEDC 1198 x28 :0x000000000A7CFDAA 1199 x29 :0x0000000004010780 1200 x30 :0x000000000400F004 1201 scr_el3 :0x0000000000000D3D 1202 sctlr_el3 :0x0000000000C8181F 1203 cptr_el3 :0x0000000000000000 1204 tcr_el3 :0x0000000080803520 1205 daif :0x00000000000003C0 1206 mair_el3 :0x00000000000004FF 1207 spsr_el3 :0x00000000800003CC 1208 elr_el3 :0x000000000400C0CC 1209 ttbr0_el3 :0x00000000040172A0 1210 esr_el3 :0x0000000096000210 1211 sp_el3 :0x0000000004014D50 1212 far_el3 :0x000000004F00007C 1213 spsr_el1 :0x0000000000000000 1214 elr_el1 :0x0000000000000000 1215 spsr_abt :0x0000000000000000 1216 spsr_und :0x0000000000000000 1217 spsr_irq :0x0000000000000000 1218 spsr_fiq :0x0000000000000000 1219 sctlr_el1 :0x0000000030C81807 1220 actlr_el1 :0x0000000000000000 1221 cpacr_el1 :0x0000000000300000 1222 csselr_el1 :0x0000000000000002 1223 sp_el1 :0x0000000004028800 1224 esr_el1 :0x0000000000000000 1225 ttbr0_el1 :0x000000000402C200 1226 ttbr1_el1 :0x0000000000000000 1227 mair_el1 :0x00000000000004FF 1228 amair_el1 :0x0000000000000000 1229 tcr_el1 :0x0000000000003520 1230 tpidr_el1 :0x0000000000000000 1231 tpidr_el0 :0x0000000000000000 1232 tpidrro_el0 :0x0000000000000000 1233 dacr32_el2 :0x0000000000000000 1234 ifsr32_el2 :0x0000000000000000 1235 par_el1 :0x0000000000000000 1236 far_el1 :0x0000000000000000 1237 afsr0_el1 :0x0000000000000000 1238 afsr1_el1 :0x0000000000000000 1239 contextidr_el1 :0x0000000000000000 1240 vbar_el1 :0x0000000004027000 1241 cntp_ctl_el0 :0x0000000000000000 1242 cntp_cval_el0 :0x0000000000000000 1243 cntv_ctl_el0 :0x0000000000000000 1244 cntv_cval_el0 :0x0000000000000000 1245 cntkctl_el1 :0x0000000000000000 1246 sp_el0 :0x0000000004010780 1247 1248Guidelines for Reset Handlers 1249----------------------------- 1250 1251TF-A implements a framework that allows CPU and platform ports to perform 1252actions very early after a CPU is released from reset in both the cold and warm 1253boot paths. This is done by calling the ``reset_handler()`` function in both 1254the BL1 and BL31 images. It in turn calls the platform and CPU specific reset 1255handling functions. 1256 1257Details for implementing a CPU specific reset handler can be found in 1258Section 8. Details for implementing a platform specific reset handler can be 1259found in the `Porting Guide`_ (see the ``plat_reset_handler()`` function). 1260 1261When adding functionality to a reset handler, keep in mind that if a different 1262reset handling behavior is required between the first and the subsequent 1263invocations of the reset handling code, this should be detected at runtime. 1264In other words, the reset handler should be able to detect whether an action has 1265already been performed and act as appropriate. Possible courses of actions are, 1266e.g. skip the action the second time, or undo/redo it. 1267 1268Configuring secure interrupts 1269----------------------------- 1270 1271The GIC driver is responsible for performing initial configuration of secure 1272interrupts on the platform. To this end, the platform is expected to provide the 1273GIC driver (either GICv2 or GICv3, as selected by the platform) with the 1274interrupt configuration during the driver initialisation. 1275 1276Secure interrupt configuration are specified in an array of secure interrupt 1277properties. In this scheme, in both GICv2 and GICv3 driver data structures, the 1278``interrupt_props`` member points to an array of interrupt properties. Each 1279element of the array specifies the interrupt number and its attributes 1280(priority, group, configuration). Each element of the array shall be populated 1281by the macro ``INTR_PROP_DESC()``. The macro takes the following arguments: 1282 1283- 10-bit interrupt number, 1284 1285- 8-bit interrupt priority, 1286 1287- Interrupt type (one of ``INTR_TYPE_EL3``, ``INTR_TYPE_S_EL1``, 1288 ``INTR_TYPE_NS``), 1289 1290- Interrupt configuration (either ``GIC_INTR_CFG_LEVEL`` or 1291 ``GIC_INTR_CFG_EDGE``). 1292 1293CPU specific operations framework 1294--------------------------------- 1295 1296Certain aspects of the Armv8-A architecture are implementation defined, 1297that is, certain behaviours are not architecturally defined, but must be 1298defined and documented by individual processor implementations. TF-A 1299implements a framework which categorises the common implementation defined 1300behaviours and allows a processor to export its implementation of that 1301behaviour. The categories are: 1302 1303#. Processor specific reset sequence. 1304 1305#. Processor specific power down sequences. 1306 1307#. Processor specific register dumping as a part of crash reporting. 1308 1309#. Errata status reporting. 1310 1311Each of the above categories fulfils a different requirement. 1312 1313#. allows any processor specific initialization before the caches and MMU 1314 are turned on, like implementation of errata workarounds, entry into 1315 the intra-cluster coherency domain etc. 1316 1317#. allows each processor to implement the power down sequence mandated in 1318 its Technical Reference Manual (TRM). 1319 1320#. allows a processor to provide additional information to the developer 1321 in the event of a crash, for example Cortex-A53 has registers which 1322 can expose the data cache contents. 1323 1324#. allows a processor to define a function that inspects and reports the status 1325 of all errata workarounds on that processor. 1326 1327Please note that only 2. is mandated by the TRM. 1328 1329The CPU specific operations framework scales to accommodate a large number of 1330different CPUs during power down and reset handling. The platform can specify 1331any CPU optimization it wants to enable for each CPU. It can also specify 1332the CPU errata workarounds to be applied for each CPU type during reset 1333handling by defining CPU errata compile time macros. Details on these macros 1334can be found in the `cpu-specific-build-macros.rst`_ file. 1335 1336The CPU specific operations framework depends on the ``cpu_ops`` structure which 1337needs to be exported for each type of CPU in the platform. It is defined in 1338``include/lib/cpus/aarch64/cpu_macros.S`` and has the following fields : ``midr``, 1339``reset_func()``, ``cpu_pwr_down_ops`` (array of power down functions) and 1340``cpu_reg_dump()``. 1341 1342The CPU specific files in ``lib/cpus`` export a ``cpu_ops`` data structure with 1343suitable handlers for that CPU. For example, ``lib/cpus/aarch64/cortex_a53.S`` 1344exports the ``cpu_ops`` for Cortex-A53 CPU. According to the platform 1345configuration, these CPU specific files must be included in the build by 1346the platform makefile. The generic CPU specific operations framework code exists 1347in ``lib/cpus/aarch64/cpu_helpers.S``. 1348 1349CPU specific Reset Handling 1350~~~~~~~~~~~~~~~~~~~~~~~~~~~ 1351 1352After a reset, the state of the CPU when it calls generic reset handler is: 1353MMU turned off, both instruction and data caches turned off and not part 1354of any coherency domain. 1355 1356The BL entrypoint code first invokes the ``plat_reset_handler()`` to allow 1357the platform to perform any system initialization required and any system 1358errata workarounds that needs to be applied. The ``get_cpu_ops_ptr()`` reads 1359the current CPU midr, finds the matching ``cpu_ops`` entry in the ``cpu_ops`` 1360array and returns it. Note that only the part number and implementer fields 1361in midr are used to find the matching ``cpu_ops`` entry. The ``reset_func()`` in 1362the returned ``cpu_ops`` is then invoked which executes the required reset 1363handling for that CPU and also any errata workarounds enabled by the platform. 1364This function must preserve the values of general purpose registers x20 to x29. 1365 1366Refer to Section "Guidelines for Reset Handlers" for general guidelines 1367regarding placement of code in a reset handler. 1368 1369CPU specific power down sequence 1370~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 1371 1372During the BL31 initialization sequence, the pointer to the matching ``cpu_ops`` 1373entry is stored in per-CPU data by ``init_cpu_ops()`` so that it can be quickly 1374retrieved during power down sequences. 1375 1376Various CPU drivers register handlers to perform power down at certain power 1377levels for that specific CPU. The PSCI service, upon receiving a power down 1378request, determines the highest power level at which to execute power down 1379sequence for a particular CPU. It uses the ``prepare_cpu_pwr_dwn()`` function to 1380pick the right power down handler for the requested level. The function 1381retrieves ``cpu_ops`` pointer member of per-CPU data, and from that, further 1382retrieves ``cpu_pwr_down_ops`` array, and indexes into the required level. If the 1383requested power level is higher than what a CPU driver supports, the handler 1384registered for highest level is invoked. 1385 1386At runtime the platform hooks for power down are invoked by the PSCI service to 1387perform platform specific operations during a power down sequence, for example 1388turning off CCI coherency during a cluster power down. 1389 1390CPU specific register reporting during crash 1391~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 1392 1393If the crash reporting is enabled in BL31, when a crash occurs, the crash 1394reporting framework calls ``do_cpu_reg_dump`` which retrieves the matching 1395``cpu_ops`` using ``get_cpu_ops_ptr()`` function. The ``cpu_reg_dump()`` in 1396``cpu_ops`` is invoked, which then returns the CPU specific register values to 1397be reported and a pointer to the ASCII list of register names in a format 1398expected by the crash reporting framework. 1399 1400CPU errata status reporting 1401~~~~~~~~~~~~~~~~~~~~~~~~~~~ 1402 1403Errata workarounds for CPUs supported in TF-A are applied during both cold and 1404warm boots, shortly after reset. Individual Errata workarounds are enabled as 1405build options. Some errata workarounds have potential run-time implications; 1406therefore some are enabled by default, others not. Platform ports shall 1407override build options to enable or disable errata as appropriate. The CPU 1408drivers take care of applying errata workarounds that are enabled and applicable 1409to a given CPU. Refer to the section titled *CPU Errata Workarounds* in `CPUBM`_ 1410for more information. 1411 1412Functions in CPU drivers that apply errata workaround must follow the 1413conventions listed below. 1414 1415The errata workaround must be authored as two separate functions: 1416 1417- One that checks for errata. This function must determine whether that errata 1418 applies to the current CPU. Typically this involves matching the current 1419 CPUs revision and variant against a value that's known to be affected by the 1420 errata. If the function determines that the errata applies to this CPU, it 1421 must return ``ERRATA_APPLIES``; otherwise, it must return 1422 ``ERRATA_NOT_APPLIES``. The utility functions ``cpu_get_rev_var`` and 1423 ``cpu_rev_var_ls`` functions may come in handy for this purpose. 1424 1425For an errata identified as ``E``, the check function must be named 1426``check_errata_E``. 1427 1428This function will be invoked at different times, both from assembly and from 1429C run time. Therefore it must follow AAPCS, and must not use stack. 1430 1431- Another one that applies the errata workaround. This function would call the 1432 check function described above, and applies errata workaround if required. 1433 1434CPU drivers that apply errata workaround can optionally implement an assembly 1435function that report the status of errata workarounds pertaining to that CPU. 1436For a driver that registers the CPU, for example, ``cpux`` via ``declare_cpu_ops`` 1437macro, the errata reporting function, if it exists, must be named 1438``cpux_errata_report``. This function will always be called with MMU enabled; it 1439must follow AAPCS and may use stack. 1440 1441In a debug build of TF-A, on a CPU that comes out of reset, both BL1 and the 1442runtime firmware (BL31 in AArch64, and BL32 in AArch32) will invoke errata 1443status reporting function, if one exists, for that type of CPU. 1444 1445To report the status of each errata workaround, the function shall use the 1446assembler macro ``report_errata``, passing it: 1447 1448- The build option that enables the errata; 1449 1450- The name of the CPU: this must be the same identifier that CPU driver 1451 registered itself with, using ``declare_cpu_ops``; 1452 1453- And the errata identifier: the identifier must match what's used in the 1454 errata's check function described above. 1455 1456The errata status reporting function will be called once per CPU type/errata 1457combination during the software's active life time. 1458 1459It's expected that whenever an errata workaround is submitted to TF-A, the 1460errata reporting function is appropriately extended to report its status as 1461well. 1462 1463Reporting the status of errata workaround is for informational purpose only; it 1464has no functional significance. 1465 1466Memory layout of BL images 1467-------------------------- 1468 1469Each bootloader image can be divided in 2 parts: 1470 1471- the static contents of the image. These are data actually stored in the 1472 binary on the disk. In the ELF terminology, they are called ``PROGBITS`` 1473 sections; 1474 1475- the run-time contents of the image. These are data that don't occupy any 1476 space in the binary on the disk. The ELF binary just contains some 1477 metadata indicating where these data will be stored at run-time and the 1478 corresponding sections need to be allocated and initialized at run-time. 1479 In the ELF terminology, they are called ``NOBITS`` sections. 1480 1481All PROGBITS sections are grouped together at the beginning of the image, 1482followed by all NOBITS sections. This is true for all TF-A images and it is 1483governed by the linker scripts. This ensures that the raw binary images are 1484as small as possible. If a NOBITS section was inserted in between PROGBITS 1485sections then the resulting binary file would contain zero bytes in place of 1486this NOBITS section, making the image unnecessarily bigger. Smaller images 1487allow faster loading from the FIP to the main memory. 1488 1489Linker scripts and symbols 1490~~~~~~~~~~~~~~~~~~~~~~~~~~ 1491 1492Each bootloader stage image layout is described by its own linker script. The 1493linker scripts export some symbols into the program symbol table. Their values 1494correspond to particular addresses. TF-A code can refer to these symbols to 1495figure out the image memory layout. 1496 1497Linker symbols follow the following naming convention in TF-A. 1498 1499- ``__<SECTION>_START__`` 1500 1501 Start address of a given section named ``<SECTION>``. 1502 1503- ``__<SECTION>_END__`` 1504 1505 End address of a given section named ``<SECTION>``. If there is an alignment 1506 constraint on the section's end address then ``__<SECTION>_END__`` corresponds 1507 to the end address of the section's actual contents, rounded up to the right 1508 boundary. Refer to the value of ``__<SECTION>_UNALIGNED_END__`` to know the 1509 actual end address of the section's contents. 1510 1511- ``__<SECTION>_UNALIGNED_END__`` 1512 1513 End address of a given section named ``<SECTION>`` without any padding or 1514 rounding up due to some alignment constraint. 1515 1516- ``__<SECTION>_SIZE__`` 1517 1518 Size (in bytes) of a given section named ``<SECTION>``. If there is an 1519 alignment constraint on the section's end address then ``__<SECTION>_SIZE__`` 1520 corresponds to the size of the section's actual contents, rounded up to the 1521 right boundary. In other words, ``__<SECTION>_SIZE__ = __<SECTION>_END__ - _<SECTION>_START__``. Refer to the value of ``__<SECTION>_UNALIGNED_SIZE__`` 1522 to know the actual size of the section's contents. 1523 1524- ``__<SECTION>_UNALIGNED_SIZE__`` 1525 1526 Size (in bytes) of a given section named ``<SECTION>`` without any padding or 1527 rounding up due to some alignment constraint. In other words, 1528 ``__<SECTION>_UNALIGNED_SIZE__ = __<SECTION>_UNALIGNED_END__ - __<SECTION>_START__``. 1529 1530Some of the linker symbols are mandatory as TF-A code relies on them to be 1531defined. They are listed in the following subsections. Some of them must be 1532provided for each bootloader stage and some are specific to a given bootloader 1533stage. 1534 1535The linker scripts define some extra, optional symbols. They are not actually 1536used by any code but they help in understanding the bootloader images' memory 1537layout as they are easy to spot in the link map files. 1538 1539Common linker symbols 1540^^^^^^^^^^^^^^^^^^^^^ 1541 1542All BL images share the following requirements: 1543 1544- The BSS section must be zero-initialised before executing any C code. 1545- The coherent memory section (if enabled) must be zero-initialised as well. 1546- The MMU setup code needs to know the extents of the coherent and read-only 1547 memory regions to set the right memory attributes. When 1548 ``SEPARATE_CODE_AND_RODATA=1``, it needs to know more specifically how the 1549 read-only memory region is divided between code and data. 1550 1551The following linker symbols are defined for this purpose: 1552 1553- ``__BSS_START__`` 1554- ``__BSS_SIZE__`` 1555- ``__COHERENT_RAM_START__`` Must be aligned on a page-size boundary. 1556- ``__COHERENT_RAM_END__`` Must be aligned on a page-size boundary. 1557- ``__COHERENT_RAM_UNALIGNED_SIZE__`` 1558- ``__RO_START__`` 1559- ``__RO_END__`` 1560- ``__TEXT_START__`` 1561- ``__TEXT_END__`` 1562- ``__RODATA_START__`` 1563- ``__RODATA_END__`` 1564 1565BL1's linker symbols 1566^^^^^^^^^^^^^^^^^^^^ 1567 1568BL1 being the ROM image, it has additional requirements. BL1 resides in ROM and 1569it is entirely executed in place but it needs some read-write memory for its 1570mutable data. Its ``.data`` section (i.e. its allocated read-write data) must be 1571relocated from ROM to RAM before executing any C code. 1572 1573The following additional linker symbols are defined for BL1: 1574 1575- ``__BL1_ROM_END__`` End address of BL1's ROM contents, covering its code 1576 and ``.data`` section in ROM. 1577- ``__DATA_ROM_START__`` Start address of the ``.data`` section in ROM. Must be 1578 aligned on a 16-byte boundary. 1579- ``__DATA_RAM_START__`` Address in RAM where the ``.data`` section should be 1580 copied over. Must be aligned on a 16-byte boundary. 1581- ``__DATA_SIZE__`` Size of the ``.data`` section (in ROM or RAM). 1582- ``__BL1_RAM_START__`` Start address of BL1 read-write data. 1583- ``__BL1_RAM_END__`` End address of BL1 read-write data. 1584 1585How to choose the right base addresses for each bootloader stage image 1586~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 1587 1588There is currently no support for dynamic image loading in TF-A. This means 1589that all bootloader images need to be linked against their ultimate runtime 1590locations and the base addresses of each image must be chosen carefully such 1591that images don't overlap each other in an undesired way. As the code grows, 1592the base addresses might need adjustments to cope with the new memory layout. 1593 1594The memory layout is completely specific to the platform and so there is no 1595general recipe for choosing the right base addresses for each bootloader image. 1596However, there are tools to aid in understanding the memory layout. These are 1597the link map files: ``build/<platform>/<build-type>/bl<x>/bl<x>.map``, with ``<x>`` 1598being the stage bootloader. They provide a detailed view of the memory usage of 1599each image. Among other useful information, they provide the end address of 1600each image. 1601 1602- ``bl1.map`` link map file provides ``__BL1_RAM_END__`` address. 1603- ``bl2.map`` link map file provides ``__BL2_END__`` address. 1604- ``bl31.map`` link map file provides ``__BL31_END__`` address. 1605- ``bl32.map`` link map file provides ``__BL32_END__`` address. 1606 1607For each bootloader image, the platform code must provide its start address 1608as well as a limit address that it must not overstep. The latter is used in the 1609linker scripts to check that the image doesn't grow past that address. If that 1610happens, the linker will issue a message similar to the following: 1611 1612:: 1613 1614 aarch64-none-elf-ld: BLx has exceeded its limit. 1615 1616Additionally, if the platform memory layout implies some image overlaying like 1617on FVP, BL31 and TSP need to know the limit address that their PROGBITS 1618sections must not overstep. The platform code must provide those. 1619 1620TF-A does not provide any mechanism to verify at boot time that the memory 1621to load a new image is free to prevent overwriting a previously loaded image. 1622The platform must specify the memory available in the system for all the 1623relevant BL images to be loaded. 1624 1625For example, in the case of BL1 loading BL2, ``bl1_plat_sec_mem_layout()`` will 1626return the region defined by the platform where BL1 intends to load BL2. The 1627``load_image()`` function performs bounds check for the image size based on the 1628base and maximum image size provided by the platforms. Platforms must take 1629this behaviour into account when defining the base/size for each of the images. 1630 1631Memory layout on Arm development platforms 1632^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ 1633 1634The following list describes the memory layout on the Arm development platforms: 1635 1636- A 4KB page of shared memory is used for communication between Trusted 1637 Firmware and the platform's power controller. This is located at the base of 1638 Trusted SRAM. The amount of Trusted SRAM available to load the bootloader 1639 images is reduced by the size of the shared memory. 1640 1641 The shared memory is used to store the CPUs' entrypoint mailbox. On Juno, 1642 this is also used for the MHU payload when passing messages to and from the 1643 SCP. 1644 1645- Another 4 KB page is reserved for passing memory layout between BL1 and BL2 1646 and also the dynamic firmware configurations. 1647 1648- On FVP, BL1 is originally sitting in the Trusted ROM at address ``0x0``. On 1649 Juno, BL1 resides in flash memory at address ``0x0BEC0000``. BL1 read-write 1650 data are relocated to the top of Trusted SRAM at runtime. 1651 1652- BL2 is loaded below BL1 RW 1653 1654- EL3 Runtime Software, BL31 for AArch64 and BL32 for AArch32 (e.g. SP_MIN), 1655 is loaded at the top of the Trusted SRAM, such that its NOBITS sections will 1656 overwrite BL1 R/W data and BL2. This implies that BL1 global variables 1657 remain valid only until execution reaches the EL3 Runtime Software entry 1658 point during a cold boot. 1659 1660- On Juno, SCP_BL2 is loaded temporarily into the EL3 Runtime Software memory 1661 region and transfered to the SCP before being overwritten by EL3 Runtime 1662 Software. 1663 1664- BL32 (for AArch64) can be loaded in one of the following locations: 1665 1666 - Trusted SRAM 1667 - Trusted DRAM (FVP only) 1668 - Secure region of DRAM (top 16MB of DRAM configured by the TrustZone 1669 controller) 1670 1671 When BL32 (for AArch64) is loaded into Trusted SRAM, it is loaded below 1672 BL31. 1673 1674The location of the BL32 image will result in different memory maps. This is 1675illustrated for both FVP and Juno in the following diagrams, using the TSP as 1676an example. 1677 1678Note: Loading the BL32 image in TZC secured DRAM doesn't change the memory 1679layout of the other images in Trusted SRAM. 1680 1681CONFIG section in memory layouts shown below contains: 1682 1683:: 1684 1685 +--------------------+ 1686 |bl2_mem_params_descs| 1687 |--------------------| 1688 | fw_configs | 1689 +--------------------+ 1690 1691``bl2_mem_params_descs`` contains parameters passed from BL2 to next the 1692BL image during boot. 1693 1694``fw_configs`` includes soc_fw_config, tos_fw_config and tb_fw_config. 1695 1696**FVP with TSP in Trusted SRAM with firmware configs :** 1697(These diagrams only cover the AArch64 case) 1698 1699:: 1700 1701 DRAM 1702 0xffffffff +----------+ 1703 : : 1704 |----------| 1705 |HW_CONFIG | 1706 0x83000000 |----------| (non-secure) 1707 | | 1708 0x80000000 +----------+ 1709 1710 Trusted SRAM 1711 0x04040000 +----------+ loaded by BL2 +----------------+ 1712 | BL1 (rw) | <<<<<<<<<<<<< | | 1713 |----------| <<<<<<<<<<<<< | BL31 NOBITS | 1714 | BL2 | <<<<<<<<<<<<< | | 1715 |----------| <<<<<<<<<<<<< |----------------| 1716 | | <<<<<<<<<<<<< | BL31 PROGBITS | 1717 | | <<<<<<<<<<<<< |----------------| 1718 | | <<<<<<<<<<<<< | BL32 | 1719 0x04002000 +----------+ +----------------+ 1720 | CONFIG | 1721 0x04001000 +----------+ 1722 | Shared | 1723 0x04000000 +----------+ 1724 1725 Trusted ROM 1726 0x04000000 +----------+ 1727 | BL1 (ro) | 1728 0x00000000 +----------+ 1729 1730**FVP with TSP in Trusted DRAM with firmware configs (default option):** 1731 1732:: 1733 1734 DRAM 1735 0xffffffff +--------------+ 1736 : : 1737 |--------------| 1738 | HW_CONFIG | 1739 0x83000000 |--------------| (non-secure) 1740 | | 1741 0x80000000 +--------------+ 1742 1743 Trusted DRAM 1744 0x08000000 +--------------+ 1745 | BL32 | 1746 0x06000000 +--------------+ 1747 1748 Trusted SRAM 1749 0x04040000 +--------------+ loaded by BL2 +----------------+ 1750 | BL1 (rw) | <<<<<<<<<<<<< | | 1751 |--------------| <<<<<<<<<<<<< | BL31 NOBITS | 1752 | BL2 | <<<<<<<<<<<<< | | 1753 |--------------| <<<<<<<<<<<<< |----------------| 1754 | | <<<<<<<<<<<<< | BL31 PROGBITS | 1755 | | +----------------+ 1756 +--------------+ 1757 | CONFIG | 1758 0x04001000 +--------------+ 1759 | Shared | 1760 0x04000000 +--------------+ 1761 1762 Trusted ROM 1763 0x04000000 +--------------+ 1764 | BL1 (ro) | 1765 0x00000000 +--------------+ 1766 1767**FVP with TSP in TZC-Secured DRAM with firmware configs :** 1768 1769:: 1770 1771 DRAM 1772 0xffffffff +----------+ 1773 | BL32 | (secure) 1774 0xff000000 +----------+ 1775 | | 1776 |----------| 1777 |HW_CONFIG | 1778 0x83000000 |----------| (non-secure) 1779 | | 1780 0x80000000 +----------+ 1781 1782 Trusted SRAM 1783 0x04040000 +----------+ loaded by BL2 +----------------+ 1784 | BL1 (rw) | <<<<<<<<<<<<< | | 1785 |----------| <<<<<<<<<<<<< | BL31 NOBITS | 1786 | BL2 | <<<<<<<<<<<<< | | 1787 |----------| <<<<<<<<<<<<< |----------------| 1788 | | <<<<<<<<<<<<< | BL31 PROGBITS | 1789 | | +----------------+ 1790 0x04002000 +----------+ 1791 | CONFIG | 1792 0x04001000 +----------+ 1793 | Shared | 1794 0x04000000 +----------+ 1795 1796 Trusted ROM 1797 0x04000000 +----------+ 1798 | BL1 (ro) | 1799 0x00000000 +----------+ 1800 1801**Juno with BL32 in Trusted SRAM :** 1802 1803:: 1804 1805 Flash0 1806 0x0C000000 +----------+ 1807 : : 1808 0x0BED0000 |----------| 1809 | BL1 (ro) | 1810 0x0BEC0000 |----------| 1811 : : 1812 0x08000000 +----------+ BL31 is loaded 1813 after SCP_BL2 has 1814 Trusted SRAM been sent to SCP 1815 0x04040000 +----------+ loaded by BL2 +----------------+ 1816 | BL1 (rw) | <<<<<<<<<<<<< | | 1817 |----------| <<<<<<<<<<<<< | BL31 NOBITS | 1818 | BL2 | <<<<<<<<<<<<< | | 1819 |----------| <<<<<<<<<<<<< |----------------| 1820 | SCP_BL2 | <<<<<<<<<<<<< | BL31 PROGBITS | 1821 |----------| <<<<<<<<<<<<< |----------------| 1822 | | <<<<<<<<<<<<< | BL32 | 1823 | | +----------------+ 1824 | | 1825 0x04001000 +----------+ 1826 | MHU | 1827 0x04000000 +----------+ 1828 1829**Juno with BL32 in TZC-secured DRAM :** 1830 1831:: 1832 1833 DRAM 1834 0xFFE00000 +----------+ 1835 | BL32 | (secure) 1836 0xFF000000 |----------| 1837 | | 1838 : : (non-secure) 1839 | | 1840 0x80000000 +----------+ 1841 1842 Flash0 1843 0x0C000000 +----------+ 1844 : : 1845 0x0BED0000 |----------| 1846 | BL1 (ro) | 1847 0x0BEC0000 |----------| 1848 : : 1849 0x08000000 +----------+ BL31 is loaded 1850 after SCP_BL2 has 1851 Trusted SRAM been sent to SCP 1852 0x04040000 +----------+ loaded by BL2 +----------------+ 1853 | BL1 (rw) | <<<<<<<<<<<<< | | 1854 |----------| <<<<<<<<<<<<< | BL31 NOBITS | 1855 | BL2 | <<<<<<<<<<<<< | | 1856 |----------| <<<<<<<<<<<<< |----------------| 1857 | SCP_BL2 | <<<<<<<<<<<<< | BL31 PROGBITS | 1858 |----------| +----------------+ 1859 0x04001000 +----------+ 1860 | MHU | 1861 0x04000000 +----------+ 1862 1863Library at ROM 1864--------------- 1865 1866Please refer to the `ROMLIB Design`_ document. 1867 1868Firmware Image Package (FIP) 1869---------------------------- 1870 1871Using a Firmware Image Package (FIP) allows for packing bootloader images (and 1872potentially other payloads) into a single archive that can be loaded by TF-A 1873from non-volatile platform storage. A driver to load images from a FIP has 1874been added to the storage layer and allows a package to be read from supported 1875platform storage. A tool to create Firmware Image Packages is also provided 1876and described below. 1877 1878Firmware Image Package layout 1879~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 1880 1881The FIP layout consists of a table of contents (ToC) followed by payload data. 1882The ToC itself has a header followed by one or more table entries. The ToC is 1883terminated by an end marker entry, and since the size of the ToC is 0 bytes, 1884the offset equals the total size of the FIP file. All ToC entries describe some 1885payload data that has been appended to the end of the binary package. With the 1886information provided in the ToC entry the corresponding payload data can be 1887retrieved. 1888 1889:: 1890 1891 ------------------ 1892 | ToC Header | 1893 |----------------| 1894 | ToC Entry 0 | 1895 |----------------| 1896 | ToC Entry 1 | 1897 |----------------| 1898 | ToC End Marker | 1899 |----------------| 1900 | | 1901 | Data 0 | 1902 | | 1903 |----------------| 1904 | | 1905 | Data 1 | 1906 | | 1907 ------------------ 1908 1909The ToC header and entry formats are described in the header file 1910``include/tools_share/firmware_image_package.h``. This file is used by both the 1911tool and TF-A. 1912 1913The ToC header has the following fields: 1914 1915:: 1916 1917 `name`: The name of the ToC. This is currently used to validate the header. 1918 `serial_number`: A non-zero number provided by the creation tool 1919 `flags`: Flags associated with this data. 1920 Bits 0-31: Reserved 1921 Bits 32-47: Platform defined 1922 Bits 48-63: Reserved 1923 1924A ToC entry has the following fields: 1925 1926:: 1927 1928 `uuid`: All files are referred to by a pre-defined Universally Unique 1929 IDentifier [UUID] . The UUIDs are defined in 1930 `include/tools_share/firmware_image_package.h`. The platform translates 1931 the requested image name into the corresponding UUID when accessing the 1932 package. 1933 `offset_address`: The offset address at which the corresponding payload data 1934 can be found. The offset is calculated from the ToC base address. 1935 `size`: The size of the corresponding payload data in bytes. 1936 `flags`: Flags associated with this entry. None are yet defined. 1937 1938Firmware Image Package creation tool 1939~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 1940 1941The FIP creation tool can be used to pack specified images into a binary 1942package that can be loaded by TF-A from platform storage. The tool currently 1943only supports packing bootloader images. Additional image definitions can be 1944added to the tool as required. 1945 1946The tool can be found in ``tools/fiptool``. 1947 1948Loading from a Firmware Image Package (FIP) 1949~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 1950 1951The Firmware Image Package (FIP) driver can load images from a binary package on 1952non-volatile platform storage. For the Arm development platforms, this is 1953currently NOR FLASH. 1954 1955Bootloader images are loaded according to the platform policy as specified by 1956the function ``plat_get_image_source()``. For the Arm development platforms, this 1957means the platform will attempt to load images from a Firmware Image Package 1958located at the start of NOR FLASH0. 1959 1960The Arm development platforms' policy is to only allow loading of a known set of 1961images. The platform policy can be modified to allow additional images. 1962 1963Use of coherent memory in TF-A 1964------------------------------ 1965 1966There might be loss of coherency when physical memory with mismatched 1967shareability, cacheability and memory attributes is accessed by multiple CPUs 1968(refer to section B2.9 of `Arm ARM`_ for more details). This possibility occurs 1969in TF-A during power up/down sequences when coherency, MMU and caches are 1970turned on/off incrementally. 1971 1972TF-A defines coherent memory as a region of memory with Device nGnRE attributes 1973in the translation tables. The translation granule size in TF-A is 4KB. This 1974is the smallest possible size of the coherent memory region. 1975 1976By default, all data structures which are susceptible to accesses with 1977mismatched attributes from various CPUs are allocated in a coherent memory 1978region (refer to section 2.1 of `Porting Guide`_). The coherent memory region 1979accesses are Outer Shareable, non-cacheable and they can be accessed 1980with the Device nGnRE attributes when the MMU is turned on. Hence, at the 1981expense of at least an extra page of memory, TF-A is able to work around 1982coherency issues due to mismatched memory attributes. 1983 1984The alternative to the above approach is to allocate the susceptible data 1985structures in Normal WriteBack WriteAllocate Inner shareable memory. This 1986approach requires the data structures to be designed so that it is possible to 1987work around the issue of mismatched memory attributes by performing software 1988cache maintenance on them. 1989 1990Disabling the use of coherent memory in TF-A 1991~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 1992 1993It might be desirable to avoid the cost of allocating coherent memory on 1994platforms which are memory constrained. TF-A enables inclusion of coherent 1995memory in firmware images through the build flag ``USE_COHERENT_MEM``. 1996This flag is enabled by default. It can be disabled to choose the second 1997approach described above. 1998 1999The below sections analyze the data structures allocated in the coherent memory 2000region and the changes required to allocate them in normal memory. 2001 2002Coherent memory usage in PSCI implementation 2003~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 2004 2005The ``psci_non_cpu_pd_nodes`` data structure stores the platform's power domain 2006tree information for state management of power domains. By default, this data 2007structure is allocated in the coherent memory region in TF-A because it can be 2008accessed by multiple CPUs, either with caches enabled or disabled. 2009 2010.. code:: c 2011 2012 typedef struct non_cpu_pwr_domain_node { 2013 /* 2014 * Index of the first CPU power domain node level 0 which has this node 2015 * as its parent. 2016 */ 2017 unsigned int cpu_start_idx; 2018 2019 /* 2020 * Number of CPU power domains which are siblings of the domain indexed 2021 * by 'cpu_start_idx' i.e. all the domains in the range 'cpu_start_idx 2022 * -> cpu_start_idx + ncpus' have this node as their parent. 2023 */ 2024 unsigned int ncpus; 2025 2026 /* 2027 * Index of the parent power domain node. 2028 */ 2029 unsigned int parent_node; 2030 2031 plat_local_state_t local_state; 2032 2033 unsigned char level; 2034 2035 /* For indexing the psci_lock array*/ 2036 unsigned char lock_index; 2037 } non_cpu_pd_node_t; 2038 2039In order to move this data structure to normal memory, the use of each of its 2040fields must be analyzed. Fields like ``cpu_start_idx``, ``ncpus``, ``parent_node`` 2041``level`` and ``lock_index`` are only written once during cold boot. Hence removing 2042them from coherent memory involves only doing a clean and invalidate of the 2043cache lines after these fields are written. 2044 2045The field ``local_state`` can be concurrently accessed by multiple CPUs in 2046different cache states. A Lamport's Bakery lock ``psci_locks`` is used to ensure 2047mutual exclusion to this field and a clean and invalidate is needed after it 2048is written. 2049 2050Bakery lock data 2051~~~~~~~~~~~~~~~~ 2052 2053The bakery lock data structure ``bakery_lock_t`` is allocated in coherent memory 2054and is accessed by multiple CPUs with mismatched attributes. ``bakery_lock_t`` is 2055defined as follows: 2056 2057.. code:: c 2058 2059 typedef struct bakery_lock { 2060 /* 2061 * The lock_data is a bit-field of 2 members: 2062 * Bit[0] : choosing. This field is set when the CPU is 2063 * choosing its bakery number. 2064 * Bits[1 - 15] : number. This is the bakery number allocated. 2065 */ 2066 volatile uint16_t lock_data[BAKERY_LOCK_MAX_CPUS]; 2067 } bakery_lock_t; 2068 2069It is a characteristic of Lamport's Bakery algorithm that the volatile per-CPU 2070fields can be read by all CPUs but only written to by the owning CPU. 2071 2072Depending upon the data cache line size, the per-CPU fields of the 2073``bakery_lock_t`` structure for multiple CPUs may exist on a single cache line. 2074These per-CPU fields can be read and written during lock contention by multiple 2075CPUs with mismatched memory attributes. Since these fields are a part of the 2076lock implementation, they do not have access to any other locking primitive to 2077safeguard against the resulting coherency issues. As a result, simple software 2078cache maintenance is not enough to allocate them in coherent memory. Consider 2079the following example. 2080 2081CPU0 updates its per-CPU field with data cache enabled. This write updates a 2082local cache line which contains a copy of the fields for other CPUs as well. Now 2083CPU1 updates its per-CPU field of the ``bakery_lock_t`` structure with data cache 2084disabled. CPU1 then issues a DCIVAC operation to invalidate any stale copies of 2085its field in any other cache line in the system. This operation will invalidate 2086the update made by CPU0 as well. 2087 2088To use bakery locks when ``USE_COHERENT_MEM`` is disabled, the lock data structure 2089has been redesigned. The changes utilise the characteristic of Lamport's Bakery 2090algorithm mentioned earlier. The bakery_lock structure only allocates the memory 2091for a single CPU. The macro ``DEFINE_BAKERY_LOCK`` allocates all the bakery locks 2092needed for a CPU into a section ``bakery_lock``. The linker allocates the memory 2093for other cores by using the total size allocated for the bakery_lock section 2094and multiplying it with (PLATFORM_CORE_COUNT - 1). This enables software to 2095perform software cache maintenance on the lock data structure without running 2096into coherency issues associated with mismatched attributes. 2097 2098The bakery lock data structure ``bakery_info_t`` is defined for use when 2099``USE_COHERENT_MEM`` is disabled as follows: 2100 2101.. code:: c 2102 2103 typedef struct bakery_info { 2104 /* 2105 * The lock_data is a bit-field of 2 members: 2106 * Bit[0] : choosing. This field is set when the CPU is 2107 * choosing its bakery number. 2108 * Bits[1 - 15] : number. This is the bakery number allocated. 2109 */ 2110 volatile uint16_t lock_data; 2111 } bakery_info_t; 2112 2113The ``bakery_info_t`` represents a single per-CPU field of one lock and 2114the combination of corresponding ``bakery_info_t`` structures for all CPUs in the 2115system represents the complete bakery lock. The view in memory for a system 2116with n bakery locks are: 2117 2118:: 2119 2120 bakery_lock section start 2121 |----------------| 2122 | `bakery_info_t`| <-- Lock_0 per-CPU field 2123 | Lock_0 | for CPU0 2124 |----------------| 2125 | `bakery_info_t`| <-- Lock_1 per-CPU field 2126 | Lock_1 | for CPU0 2127 |----------------| 2128 | .... | 2129 |----------------| 2130 | `bakery_info_t`| <-- Lock_N per-CPU field 2131 | Lock_N | for CPU0 2132 ------------------ 2133 | XXXXX | 2134 | Padding to | 2135 | next Cache WB | <--- Calculate PERCPU_BAKERY_LOCK_SIZE, allocate 2136 | Granule | continuous memory for remaining CPUs. 2137 ------------------ 2138 | `bakery_info_t`| <-- Lock_0 per-CPU field 2139 | Lock_0 | for CPU1 2140 |----------------| 2141 | `bakery_info_t`| <-- Lock_1 per-CPU field 2142 | Lock_1 | for CPU1 2143 |----------------| 2144 | .... | 2145 |----------------| 2146 | `bakery_info_t`| <-- Lock_N per-CPU field 2147 | Lock_N | for CPU1 2148 ------------------ 2149 | XXXXX | 2150 | Padding to | 2151 | next Cache WB | 2152 | Granule | 2153 ------------------ 2154 2155Consider a system of 2 CPUs with 'N' bakery locks as shown above. For an 2156operation on Lock_N, the corresponding ``bakery_info_t`` in both CPU0 and CPU1 2157``bakery_lock`` section need to be fetched and appropriate cache operations need 2158to be performed for each access. 2159 2160On Arm Platforms, bakery locks are used in psci (``psci_locks``) and power controller 2161driver (``arm_lock``). 2162 2163Non Functional Impact of removing coherent memory 2164~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 2165 2166Removal of the coherent memory region leads to the additional software overhead 2167of performing cache maintenance for the affected data structures. However, since 2168the memory where the data structures are allocated is cacheable, the overhead is 2169mostly mitigated by an increase in performance. 2170 2171There is however a performance impact for bakery locks, due to: 2172 2173- Additional cache maintenance operations, and 2174- Multiple cache line reads for each lock operation, since the bakery locks 2175 for each CPU are distributed across different cache lines. 2176 2177The implementation has been optimized to minimize this additional overhead. 2178Measurements indicate that when bakery locks are allocated in Normal memory, the 2179minimum latency of acquiring a lock is on an average 3-4 micro seconds whereas 2180in Device memory the same is 2 micro seconds. The measurements were done on the 2181Juno Arm development platform. 2182 2183As mentioned earlier, almost a page of memory can be saved by disabling 2184``USE_COHERENT_MEM``. Each platform needs to consider these trade-offs to decide 2185whether coherent memory should be used. If a platform disables 2186``USE_COHERENT_MEM`` and needs to use bakery locks in the porting layer, it can 2187optionally define macro ``PLAT_PERCPU_BAKERY_LOCK_SIZE`` (see the 2188`Porting Guide`_). Refer to the reference platform code for examples. 2189 2190Isolating code and read-only data on separate memory pages 2191---------------------------------------------------------- 2192 2193In the Armv8-A VMSA, translation table entries include fields that define the 2194properties of the target memory region, such as its access permissions. The 2195smallest unit of memory that can be addressed by a translation table entry is 2196a memory page. Therefore, if software needs to set different permissions on two 2197memory regions then it needs to map them using different memory pages. 2198 2199The default memory layout for each BL image is as follows: 2200 2201:: 2202 2203 | ... | 2204 +-------------------+ 2205 | Read-write data | 2206 +-------------------+ Page boundary 2207 | <Padding> | 2208 +-------------------+ 2209 | Exception vectors | 2210 +-------------------+ 2 KB boundary 2211 | <Padding> | 2212 +-------------------+ 2213 | Read-only data | 2214 +-------------------+ 2215 | Code | 2216 +-------------------+ BLx_BASE 2217 2218Note: The 2KB alignment for the exception vectors is an architectural 2219requirement. 2220 2221The read-write data start on a new memory page so that they can be mapped with 2222read-write permissions, whereas the code and read-only data below are configured 2223as read-only. 2224 2225However, the read-only data are not aligned on a page boundary. They are 2226contiguous to the code. Therefore, the end of the code section and the beginning 2227of the read-only data one might share a memory page. This forces both to be 2228mapped with the same memory attributes. As the code needs to be executable, this 2229means that the read-only data stored on the same memory page as the code are 2230executable as well. This could potentially be exploited as part of a security 2231attack. 2232 2233TF provides the build flag ``SEPARATE_CODE_AND_RODATA`` to isolate the code and 2234read-only data on separate memory pages. This in turn allows independent control 2235of the access permissions for the code and read-only data. In this case, 2236platform code gets a finer-grained view of the image layout and can 2237appropriately map the code region as executable and the read-only data as 2238execute-never. 2239 2240This has an impact on memory footprint, as padding bytes need to be introduced 2241between the code and read-only data to ensure the segregation of the two. To 2242limit the memory cost, this flag also changes the memory layout such that the 2243code and exception vectors are now contiguous, like so: 2244 2245:: 2246 2247 | ... | 2248 +-------------------+ 2249 | Read-write data | 2250 +-------------------+ Page boundary 2251 | <Padding> | 2252 +-------------------+ 2253 | Read-only data | 2254 +-------------------+ Page boundary 2255 | <Padding> | 2256 +-------------------+ 2257 | Exception vectors | 2258 +-------------------+ 2 KB boundary 2259 | <Padding> | 2260 +-------------------+ 2261 | Code | 2262 +-------------------+ BLx_BASE 2263 2264With this more condensed memory layout, the separation of read-only data will 2265add zero or one page to the memory footprint of each BL image. Each platform 2266should consider the trade-off between memory footprint and security. 2267 2268This build flag is disabled by default, minimising memory footprint. On Arm 2269platforms, it is enabled. 2270 2271Publish and Subscribe Framework 2272------------------------------- 2273 2274The Publish and Subscribe Framework allows EL3 components to define and publish 2275events, to which other EL3 components can subscribe. 2276 2277The following macros are provided by the framework: 2278 2279- ``REGISTER_PUBSUB_EVENT(event)``: Defines an event, and takes one argument, 2280 the event name, which must be a valid C identifier. All calls to 2281 ``REGISTER_PUBSUB_EVENT`` macro must be placed in the file 2282 ``pubsub_events.h``. 2283 2284- ``PUBLISH_EVENT_ARG(event, arg)``: Publishes a defined event, by iterating 2285 subscribed handlers and calling them in turn. The handlers will be passed the 2286 parameter ``arg``. The expected use-case is to broadcast an event. 2287 2288- ``PUBLISH_EVENT(event)``: Like ``PUBLISH_EVENT_ARG``, except that the value 2289 ``NULL`` is passed to subscribed handlers. 2290 2291- ``SUBSCRIBE_TO_EVENT(event, handler)``: Registers the ``handler`` to 2292 subscribe to ``event``. The handler will be executed whenever the ``event`` 2293 is published. 2294 2295- ``for_each_subscriber(event, subscriber)``: Iterates through all handlers 2296 subscribed for ``event``. ``subscriber`` must be a local variable of type 2297 ``pubsub_cb_t *``, and will point to each subscribed handler in turn during 2298 iteration. This macro can be used for those patterns that none of the 2299 ``PUBLISH_EVENT_*()`` macros cover. 2300 2301Publishing an event that wasn't defined using ``REGISTER_PUBSUB_EVENT`` will 2302result in build error. Subscribing to an undefined event however won't. 2303 2304Subscribed handlers must be of type ``pubsub_cb_t``, with following function 2305signature: 2306 2307:: 2308 2309 typedef void* (*pubsub_cb_t)(const void *arg); 2310 2311There may be arbitrary number of handlers registered to the same event. The 2312order in which subscribed handlers are notified when that event is published is 2313not defined. Subscribed handlers may be executed in any order; handlers should 2314not assume any relative ordering amongst them. 2315 2316Publishing an event on a PE will result in subscribed handlers executing on that 2317PE only; it won't cause handlers to execute on a different PE. 2318 2319Note that publishing an event on a PE blocks until all the subscribed handlers 2320finish executing on the PE. 2321 2322TF-A generic code publishes and subscribes to some events within. Platform 2323ports are discouraged from subscribing to them. These events may be withdrawn, 2324renamed, or have their semantics altered in the future. Platforms may however 2325register, publish, and subscribe to platform-specific events. 2326 2327Publish and Subscribe Example 2328~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 2329 2330A publisher that wants to publish event ``foo`` would: 2331 2332- Define the event ``foo`` in the ``pubsub_events.h``. 2333 2334 :: 2335 2336 REGISTER_PUBSUB_EVENT(foo); 2337 2338- Depending on the nature of event, use one of ``PUBLISH_EVENT_*()`` macros to 2339 publish the event at the appropriate path and time of execution. 2340 2341A subscriber that wants to subscribe to event ``foo`` published above would 2342implement: 2343 2344.. code:: c 2345 2346 void *foo_handler(const void *arg) 2347 { 2348 void *result; 2349 2350 /* Do handling ... */ 2351 2352 return result; 2353 } 2354 2355 SUBSCRIBE_TO_EVENT(foo, foo_handler); 2356 2357 2358Reclaiming the BL31 initialization code 2359~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 2360 2361A significant amount of the code used for the initialization of BL31 is never 2362needed again after boot time. In order to reduce the runtime memory 2363footprint, the memory used for this code can be reclaimed after initialization 2364has finished and be used for runtime data. 2365 2366The build option ``RECLAIM_INIT_CODE`` can be set to mark this boot time code 2367with a ``.text.init.*`` attribute which can be filtered and placed suitably 2368within the BL image for later reclamation by the platform. The platform can 2369specify the filter and the memory region for this init section in BL31 via the 2370plat.ld.S linker script. For example, on the FVP, this section is placed 2371overlapping the secondary CPU stacks so that after the cold boot is done, this 2372memory can be reclaimed for the stacks. The init memory section is initially 2373mapped with ``RO``, ``EXECUTE`` attributes. After BL31 initialization has 2374completed, the FVP changes the attributes of this section to ``RW``, 2375``EXECUTE_NEVER`` allowing it to be used for runtime data. The memory attributes 2376are changed within the ``bl31_plat_runtime_setup`` platform hook. The init 2377section section can be reclaimed for any data which is accessed after cold 2378boot initialization and it is upto the platform to make the decision. 2379 2380Performance Measurement Framework 2381--------------------------------- 2382 2383The Performance Measurement Framework (PMF) facilitates collection of 2384timestamps by registered services and provides interfaces to retrieve them 2385from within TF-A. A platform can choose to expose appropriate SMCs to 2386retrieve these collected timestamps. 2387 2388By default, the global physical counter is used for the timestamp 2389value and is read via ``CNTPCT_EL0``. The framework allows to retrieve 2390timestamps captured by other CPUs. 2391 2392Timestamp identifier format 2393~~~~~~~~~~~~~~~~~~~~~~~~~~~ 2394 2395A PMF timestamp is uniquely identified across the system via the 2396timestamp ID or ``tid``. The ``tid`` is composed as follows: 2397 2398:: 2399 2400 Bits 0-7: The local timestamp identifier. 2401 Bits 8-9: Reserved. 2402 Bits 10-15: The service identifier. 2403 Bits 16-31: Reserved. 2404 2405#. The service identifier. Each PMF service is identified by a 2406 service name and a service identifier. Both the service name and 2407 identifier are unique within the system as a whole. 2408 2409#. The local timestamp identifier. This identifier is unique within a given 2410 service. 2411 2412Registering a PMF service 2413~~~~~~~~~~~~~~~~~~~~~~~~~ 2414 2415To register a PMF service, the ``PMF_REGISTER_SERVICE()`` macro from ``pmf.h`` 2416is used. The arguments required are the service name, the service ID, 2417the total number of local timestamps to be captured and a set of flags. 2418 2419The ``flags`` field can be specified as a bitwise-OR of the following values: 2420 2421:: 2422 2423 PMF_STORE_ENABLE: The timestamp is stored in memory for later retrieval. 2424 PMF_DUMP_ENABLE: The timestamp is dumped on the serial console. 2425 2426The ``PMF_REGISTER_SERVICE()`` reserves memory to store captured 2427timestamps in a PMF specific linker section at build time. 2428Additionally, it defines necessary functions to capture and 2429retrieve a particular timestamp for the given service at runtime. 2430 2431The macro ``PMF_REGISTER_SERVICE()`` only enables capturing PMF timestamps 2432from within TF-A. In order to retrieve timestamps from outside of TF-A, the 2433``PMF_REGISTER_SERVICE_SMC()`` macro must be used instead. This macro 2434accepts the same set of arguments as the ``PMF_REGISTER_SERVICE()`` 2435macro but additionally supports retrieving timestamps using SMCs. 2436 2437Capturing a timestamp 2438~~~~~~~~~~~~~~~~~~~~~ 2439 2440PMF timestamps are stored in a per-service timestamp region. On a 2441system with multiple CPUs, each timestamp is captured and stored 2442in a per-CPU cache line aligned memory region. 2443 2444Having registered the service, the ``PMF_CAPTURE_TIMESTAMP()`` macro can be 2445used to capture a timestamp at the location where it is used. The macro 2446takes the service name, a local timestamp identifier and a flag as arguments. 2447 2448The ``flags`` field argument can be zero, or ``PMF_CACHE_MAINT`` which 2449instructs PMF to do cache maintenance following the capture. Cache 2450maintenance is required if any of the service's timestamps are captured 2451with data cache disabled. 2452 2453To capture a timestamp in assembly code, the caller should use 2454``pmf_calc_timestamp_addr`` macro (defined in ``pmf_asm_macros.S``) to 2455calculate the address of where the timestamp would be stored. The 2456caller should then read ``CNTPCT_EL0`` register to obtain the timestamp 2457and store it at the determined address for later retrieval. 2458 2459Retrieving a timestamp 2460~~~~~~~~~~~~~~~~~~~~~~ 2461 2462From within TF-A, timestamps for individual CPUs can be retrieved using either 2463``PMF_GET_TIMESTAMP_BY_MPIDR()`` or ``PMF_GET_TIMESTAMP_BY_INDEX()`` macros. 2464These macros accept the CPU's MPIDR value, or its ordinal position 2465respectively. 2466 2467From outside TF-A, timestamps for individual CPUs can be retrieved by calling 2468into ``pmf_smc_handler()``. 2469 2470.. code:: c 2471 2472 Interface : pmf_smc_handler() 2473 Argument : unsigned int smc_fid, u_register_t x1, 2474 u_register_t x2, u_register_t x3, 2475 u_register_t x4, void *cookie, 2476 void *handle, u_register_t flags 2477 Return : uintptr_t 2478 2479 smc_fid: Holds the SMC identifier which is either `PMF_SMC_GET_TIMESTAMP_32` 2480 when the caller of the SMC is running in AArch32 mode 2481 or `PMF_SMC_GET_TIMESTAMP_64` when the caller is running in AArch64 mode. 2482 x1: Timestamp identifier. 2483 x2: The `mpidr` of the CPU for which the timestamp has to be retrieved. 2484 This can be the `mpidr` of a different core to the one initiating 2485 the SMC. In that case, service specific cache maintenance may be 2486 required to ensure the updated copy of the timestamp is returned. 2487 x3: A flags value that is either 0 or `PMF_CACHE_MAINT`. If 2488 `PMF_CACHE_MAINT` is passed, then the PMF code will perform a 2489 cache invalidate before reading the timestamp. This ensures 2490 an updated copy is returned. 2491 2492The remaining arguments, ``x4``, ``cookie``, ``handle`` and ``flags`` are unused 2493in this implementation. 2494 2495PMF code structure 2496~~~~~~~~~~~~~~~~~~ 2497 2498#. ``pmf_main.c`` consists of core functions that implement service registration, 2499 initialization, storing, dumping and retrieving timestamps. 2500 2501#. ``pmf_smc.c`` contains the SMC handling for registered PMF services. 2502 2503#. ``pmf.h`` contains the public interface to Performance Measurement Framework. 2504 2505#. ``pmf_asm_macros.S`` consists of macros to facilitate capturing timestamps in 2506 assembly code. 2507 2508#. ``pmf_helpers.h`` is an internal header used by ``pmf.h``. 2509 2510Armv8-A Architecture Extensions 2511------------------------------- 2512 2513TF-A makes use of Armv8-A Architecture Extensions where applicable. This 2514section lists the usage of Architecture Extensions, and build flags 2515controlling them. 2516 2517In general, and unless individually mentioned, the build options 2518``ARM_ARCH_MAJOR`` and ``ARM_ARCH_MINOR`` select the Architecture Extension to 2519target when building TF-A. Subsequent Arm Architecture Extensions are backward 2520compatible with previous versions. 2521 2522The build system only requires that ``ARM_ARCH_MAJOR`` and ``ARM_ARCH_MINOR`` have a 2523valid numeric value. These build options only control whether or not 2524Architecture Extension-specific code is included in the build. Otherwise, TF-A 2525targets the base Armv8.0-A architecture; i.e. as if ``ARM_ARCH_MAJOR`` == 8 2526and ``ARM_ARCH_MINOR`` == 0, which are also their respective default values. 2527 2528See also the *Summary of build options* in `User Guide`_. 2529 2530For details on the Architecture Extension and available features, please refer 2531to the respective Architecture Extension Supplement. 2532 2533Armv8.1-A 2534~~~~~~~~~ 2535 2536This Architecture Extension is targeted when ``ARM_ARCH_MAJOR`` >= 8, or when 2537``ARM_ARCH_MAJOR`` == 8 and ``ARM_ARCH_MINOR`` >= 1. 2538 2539- The Compare and Swap instruction is used to implement spinlocks. Otherwise, 2540 the load-/store-exclusive instruction pair is used. 2541 2542Armv8.2-A 2543~~~~~~~~~ 2544 2545- The presence of ARMv8.2-TTCNP is detected at runtime. When it is present, the 2546 Common not Private (TTBRn_ELx.CnP) bit is enabled to indicate that multiple 2547 Processing Elements in the same Inner Shareable domain use the same 2548 translation table entries for a given stage of translation for a particular 2549 translation regime. 2550 2551Armv8.3-A 2552~~~~~~~~~ 2553 2554- Pointer authentication features of Armv8.3-A are unconditionally enabled in 2555 the Non-secure world so that lower ELs are allowed to use them without 2556 causing a trap to EL3. 2557 2558 In order to enable the Secure world to use it, ``CTX_INCLUDE_PAUTH_REGS`` 2559 must be set to 1. This will add all pointer authentication system registers 2560 to the context that is saved when doing a world switch. 2561 2562 The TF-A itself has support for pointer authentication at runtime 2563 that can be enabled by setting both options ``ENABLE_PAUTH`` and 2564 ``CTX_INCLUDE_PAUTH_REGS`` to 1. This enables pointer authentication in BL1, 2565 BL2, BL31, and the TSP if it is used. 2566 2567 These options are experimental features. 2568 2569 Note that Pointer Authentication is enabled for Non-secure world irrespective 2570 of the value of these build flags if the CPU supports it. 2571 2572 If ``ARM_ARCH_MAJOR == 8`` and ``ARM_ARCH_MINOR >= 3`` the code footprint of 2573 enabling PAuth is lower because the compiler will use the optimized 2574 PAuth instructions rather than the backwards-compatible ones. 2575 2576Armv7-A 2577~~~~~~~ 2578 2579This Architecture Extension is targeted when ``ARM_ARCH_MAJOR`` == 7. 2580 2581There are several Armv7-A extensions available. Obviously the TrustZone 2582extension is mandatory to support the TF-A bootloader and runtime services. 2583 2584Platform implementing an Armv7-A system can to define from its target 2585Cortex-A architecture through ``ARM_CORTEX_A<X> = yes`` in their 2586``platform.mk`` script. For example ``ARM_CORTEX_A15=yes`` for a 2587Cortex-A15 target. 2588 2589Platform can also set ``ARM_WITH_NEON=yes`` to enable neon support. 2590Note that using neon at runtime has constraints on non secure wolrd context. 2591TF-A does not yet provide VFP context management. 2592 2593Directive ``ARM_CORTEX_A<x>`` and ``ARM_WITH_NEON`` are used to set 2594the toolchain target architecture directive. 2595 2596Platform may choose to not define straight the toolchain target architecture 2597directive by defining ``MARCH32_DIRECTIVE``. 2598I.e: 2599 2600:: 2601 2602 MARCH32_DIRECTIVE := -mach=armv7-a 2603 2604Code Structure 2605-------------- 2606 2607TF-A code is logically divided between the three boot loader stages mentioned 2608in the previous sections. The code is also divided into the following 2609categories (present as directories in the source code): 2610 2611- **Platform specific.** Choice of architecture specific code depends upon 2612 the platform. 2613- **Common code.** This is platform and architecture agnostic code. 2614- **Library code.** This code comprises of functionality commonly used by all 2615 other code. The PSCI implementation and other EL3 runtime frameworks reside 2616 as Library components. 2617- **Stage specific.** Code specific to a boot stage. 2618- **Drivers.** 2619- **Services.** EL3 runtime services (eg: SPD). Specific SPD services 2620 reside in the ``services/spd`` directory (e.g. ``services/spd/tspd``). 2621 2622Each boot loader stage uses code from one or more of the above mentioned 2623categories. Based upon the above, the code layout looks like this: 2624 2625:: 2626 2627 Directory Used by BL1? Used by BL2? Used by BL31? 2628 bl1 Yes No No 2629 bl2 No Yes No 2630 bl31 No No Yes 2631 plat Yes Yes Yes 2632 drivers Yes No Yes 2633 common Yes Yes Yes 2634 lib Yes Yes Yes 2635 services No No Yes 2636 2637The build system provides a non configurable build option IMAGE_BLx for each 2638boot loader stage (where x = BL stage). e.g. for BL1 , IMAGE_BL1 will be 2639defined by the build system. This enables TF-A to compile certain code only 2640for specific boot loader stages 2641 2642All assembler files have the ``.S`` extension. The linker source files for each 2643boot stage have the extension ``.ld.S``. These are processed by GCC to create the 2644linker scripts which have the extension ``.ld``. 2645 2646FDTs provide a description of the hardware platform and are used by the Linux 2647kernel at boot time. These can be found in the ``fdts`` directory. 2648 2649References 2650---------- 2651 2652.. [#] `Trusted Board Boot Requirements CLIENT (TBBR-CLIENT) Armv8-A (ARM DEN0006D)`_ 2653.. [#] `Power State Coordination Interface PDD`_ 2654.. [#] `SMC Calling Convention PDD`_ 2655.. [#] `TF-A Interrupt Management Design guide`_. 2656 2657-------------- 2658 2659*Copyright (c) 2013-2019, Arm Limited and Contributors. All rights reserved.* 2660 2661.. _Reset Design: ./reset-design.rst 2662.. _Porting Guide: ../getting_started/porting-guide.rst 2663.. _Firmware Update: ./firmware-update.rst 2664.. _PSCI PDD: http://infocenter.arm.com/help/topic/com.arm.doc.den0022d/Power_State_Coordination_Interface_PDD_v1_1_DEN0022D.pdf 2665.. _SMC calling convention PDD: http://infocenter.arm.com/help/topic/com.arm.doc.den0028b/ARM_DEN0028B_SMC_Calling_Convention.pdf 2666.. _PSCI Library integration guide: ../getting_started/psci-lib-integration-guide.rst 2667.. _SMCCC: http://infocenter.arm.com/help/topic/com.arm.doc.den0028b/ARM_DEN0028B_SMC_Calling_Convention.pdf 2668.. _PSCI: http://infocenter.arm.com/help/topic/com.arm.doc.den0022d/Power_State_Coordination_Interface_PDD_v1_1_DEN0022D.pdf 2669.. _Power State Coordination Interface PDD: http://infocenter.arm.com/help/topic/com.arm.doc.den0022d/Power_State_Coordination_Interface_PDD_v1_1_DEN0022D.pdf 2670.. _here: ../getting_started/psci-lib-integration-guide.rst 2671.. _cpu-specific-build-macros.rst: ./cpu-specific-build-macros.rst 2672.. _CPUBM: ./cpu-specific-build-macros.rst 2673.. _Arm ARM: http://infocenter.arm.com/help/index.jsp?topic=/com.arm.doc.ddi0487a.e/index.html 2674.. _User Guide: ../getting_started/user-guide.rst 2675.. _SMC Calling Convention PDD: http://infocenter.arm.com/help/topic/com.arm.doc.den0028b/ARM_DEN0028B_SMC_Calling_Convention.pdf 2676.. _TF-A Interrupt Management Design guide: ./interrupt-framework-design.rst 2677.. _Xlat_tables design: xlat-tables-lib-v2-design.rst 2678.. _Exception Handling Framework: exception-handling.rst 2679.. _ROMLIB Design: romlib-design.rst 2680.. _Trusted Board Boot Requirements CLIENT (TBBR-CLIENT) Armv8-A (ARM DEN0006D): https://developer.arm.com/docs/den0006/latest/trusted-board-boot-requirements-client-tbbr-client-armv8-a 2681 2682.. |Image 1| image:: diagrams/rt-svc-descs-layout.png?raw=true 2683
