]>
Commit | Line | Data |
---|---|---|
16a30b34 BB |
1 | .. SPDX-License-Identifier: GPL-2.0+ OR BSD-3-Clause |
2 | .. sectionauthor:: Bryan Brattlof <[email protected]> | |
3 | ||
4 | K3 Generation | |
5 | ============= | |
6 | ||
7 | Summary | |
8 | ------- | |
9 | ||
10 | Texas Instrument's K3 family of SoCs utilize a heterogeneous multicore | |
11 | and highly integrated device architecture targeted to maximize | |
12 | performance and power efficiency for a wide range of industrial, | |
13 | automotive and other broad market segments. | |
14 | ||
15 | Typically the processing cores and the peripherals for these devices are | |
16 | partitioned into three functional domains to provide ultra-low power | |
17 | modes as well as accommodating application and industrial safety systems | |
18 | on the same SoC. These functional domains are typically called the: | |
19 | ||
20 | * Wakeup (WKUP) domain | |
21 | * Micro-controller (MCU) domain | |
22 | * Main domain | |
23 | ||
24 | For a more detailed view of what peripherals are attached to each | |
25 | domain, consult the device specific documentation. | |
26 | ||
27 | K3 Based SoCs | |
28 | ------------- | |
29 | ||
30 | .. toctree:: | |
31 | :maxdepth: 1 | |
32 | ||
16a30b34 | 33 | am62x_sk |
7d1a1065 | 34 | ../toradex/verdin-am62 |
4bf49bad | 35 | am64x_evm |
1ee652ab | 36 | am65x_evm |
5c86c57f NM |
37 | j7200_evm |
38 | j721e_evm | |
16a30b34 BB |
39 | |
40 | Boot Flow Overview | |
41 | ------------------ | |
42 | ||
43 | For all K3 SoCs the first core started will be inside the Security | |
44 | Management Subsystem (SMS) which will secure the device and start a core | |
45 | in the wakeup domain to run the ROM code. ROM will then initialize the | |
46 | boot media needed to load the binaries packaged inside `tiboot3.bin`, | |
47 | including a 32bit U-Boot SPL, (called the wakup SPL) that ROM will jump | |
48 | to after it has finished loading everything into internal SRAM. | |
49 | ||
6e8fa061 | 50 | .. image:: img/boot_flow_01.svg |
7f62928a | 51 | :alt: Boot flow up to wakeup domain SPL |
16a30b34 BB |
52 | |
53 | The wakeup SPL, running on a wakeup domain core, will initialize DDR and | |
54 | any peripherals needed load the larger binaries inside the `tispl.bin` | |
55 | into DDR. Once loaded the wakeup SPL will start one of the 'big' | |
56 | application cores inside the main domain to initialize the main domain, | |
1ee652ab NMF |
57 | starting with Trusted Firmware-A (TF-A), before moving on to start |
58 | OP-TEE and the main domain's U-Boot SPL. | |
16a30b34 | 59 | |
6e8fa061 | 60 | .. image:: img/boot_flow_02.svg |
7f62928a | 61 | :alt: Boot flow up to main domain SPL |
16a30b34 BB |
62 | |
63 | The main domain's SPL, running on a 64bit application core, has | |
64 | virtually unlimited space (billions of bytes now that DDR is working) to | |
65 | initialize even more peripherals needed to load in the `u-boot.img` | |
66 | which loads more firmware into the micro-controller & wakeup domains and | |
67 | finally prepare the main domain to run Linux. | |
68 | ||
6e8fa061 | 69 | .. image:: img/boot_flow_03.svg |
7f62928a | 70 | :alt: Complete boot flow up to Linux |
16a30b34 BB |
71 | |
72 | This is the typical boot flow for all K3 based SoCs, however this flow | |
73 | offers quite a lot in the terms of flexibility, especially on High | |
74 | Security (HS) SoCs. | |
75 | ||
76 | Boot Flow Variations | |
77 | ^^^^^^^^^^^^^^^^^^^^ | |
78 | ||
79 | All K3 SoCs will generally use the above boot flow with two main | |
80 | differences depending on the capabilities of the boot ROM and the number | |
81 | of cores inside the device. These differences split the bootflow into | |
82 | essentially 4 unique but very similar flows: | |
83 | ||
84 | * Split binary with a combined firmware: (eg: AM65) | |
85 | * Combined binary with a combined firmware: (eg: AM64) | |
86 | * Split binary with a split firmware: (eg: J721E) | |
87 | * Combined binary with a split firmware: (eg: AM62) | |
88 | ||
89 | For devices that utilize the split binary approach, ROM is not capable | |
90 | of loading the firmware into the SoC requiring the wakeup domain's | |
91 | U-Boot SPL to load the firmware. | |
92 | ||
93 | Devices with a split firmware will have two firmwares loaded into the | |
94 | device at different times during the bootup process. TI's Foundational | |
95 | Security (TIFS), needed to operate the Security Management Subsystem, | |
96 | will either be loaded by ROM or the WKUP U-Boot SPL, then once the | |
97 | wakeup U-Boot SPL has completed, the second Device Management (DM) | |
98 | firmware can be loaded on the now free core in the wakeup domain. | |
99 | ||
100 | For more information on the bootup process of your SoC, consult the | |
101 | device specific boot flow documentation. | |
102 | ||
103 | Software Sources | |
104 | ---------------- | |
105 | ||
106 | All scripts and code needed to build the `tiboot3.bin`, `tispl.bin` and | |
107 | `u-boot.img` for all K3 SoCs can be located at the following places | |
108 | online | |
109 | ||
cce3e7a2 NM |
110 | .. k3_rst_include_start_boot_sources |
111 | ||
16a30b34 BB |
112 | * **Das U-Boot** |
113 | ||
114 | | **source:** https://source.denx.de/u-boot/u-boot.git | |
115 | | **branch:** master | |
116 | ||
1ee652ab | 117 | * **Trusted Firmware-A (TF-A)** |
16a30b34 | 118 | |
1ee652ab | 119 | | **source:** https://git.trustedfirmware.org/TF-A/trusted-firmware-a.git/ |
16a30b34 BB |
120 | | **branch:** master |
121 | ||
1ee652ab | 122 | * **Open Portable Trusted Execution Environment (OP-TEE)** |
16a30b34 BB |
123 | |
124 | | **source:** https://github.com/OP-TEE/optee_os.git | |
125 | | **branch:** master | |
126 | ||
4e4f344e | 127 | * **TI Firmware (TIFS, DM, SYSFW)** |
16a30b34 BB |
128 | |
129 | | **source:** https://git.ti.com/git/processor-firmware/ti-linux-firmware.git | |
130 | | **branch:** ti-linux-firmware | |
131 | ||
4e4f344e NM |
132 | .. note:: |
133 | ||
134 | The TI Firmware required for functionality of the system can be | |
135 | one of the following combination (see platform specific boot diagram for | |
136 | further information as to which component runs on which processor): | |
137 | ||
138 | * **TIFS** - TI Foundational Security Firmware - Consists of purely firmware | |
139 | meant to run on the security enclave. | |
140 | * **DM** - Device Management firmware also called TI System Control Interface | |
141 | server (TISCI Server) - This component purely plays the role of managing | |
142 | device resources such as power, clock, interrupts, dma etc. This firmware | |
143 | runs on a dedicated or multi-use microcontroller outside the security | |
144 | enclave. | |
145 | ||
146 | OR | |
147 | ||
148 | * **SYSFW** - System firmware - consists of both TIFS and DM both running on | |
149 | the security enclave. | |
150 | ||
cce3e7a2 NM |
151 | .. k3_rst_include_end_boot_sources |
152 | ||
16a30b34 BB |
153 | Build Procedure |
154 | --------------- | |
155 | ||
156 | Depending on the specifics of your device, you will need three or more | |
157 | binaries to boot your SoC. | |
158 | ||
159 | * `tiboot3.bin` (bootloader for the wakeup domain) | |
160 | * `tispl.bin` (bootloader for the main domain) | |
161 | * `u-boot.img` | |
162 | ||
163 | During the bootup process, both the 32bit wakeup domain and the 64bit | |
164 | main domains will be involved. This means everything inside the | |
165 | `tiboot3.bin` running in the wakeup domain will need to be compiled for | |
166 | 32bit cores and most binaries in the `tispl.bin` will need to be | |
167 | compiled for 64bit main domain CPU cores. | |
168 | ||
169 | All of that to say you will need both a 32bit and 64bit cross compiler | |
170 | (assuming you're using an x86 desktop) | |
171 | ||
c727b81d NM |
172 | .. k3_rst_include_start_common_env_vars_desc |
173 | .. list-table:: Generic environment variables | |
174 | :widths: 25 25 50 | |
175 | :header-rows: 1 | |
176 | ||
177 | * - S/w Component | |
178 | - Env Variable | |
179 | - Description | |
180 | * - All Software | |
181 | - CC32 | |
182 | - Cross compiler for ARMv7 (ARM 32bit), typically arm-linux-gnueabihf- | |
183 | * - All Software | |
184 | - CC64 | |
185 | - Cross compiler for ARMv8 (ARM 64bit), typically aarch64-linux-gnu- | |
186 | * - All Software | |
187 | - LNX_FW_PATH | |
188 | - Path to TI Linux firmware repository | |
189 | * - All Software | |
190 | - TFA_PATH | |
191 | - Path to source of Trusted Firmware-A | |
192 | * - All Software | |
193 | - OPTEE_PATH | |
194 | - Path to source of OP-TEE | |
195 | .. k3_rst_include_end_common_env_vars_desc | |
196 | ||
197 | .. k3_rst_include_start_common_env_vars_defn | |
febc7f10 | 198 | .. prompt:: bash |
16a30b34 | 199 | |
febc7f10 NM |
200 | export CC32=arm-linux-gnueabihf- |
201 | export CC64=aarch64-linux-gnu- | |
202 | export LNX_FW_PATH=path/to/ti-linux-firmware | |
203 | export TFA_PATH=path/to/trusted-firmware-a | |
204 | export OPTEE_PATH=path/to/optee_os | |
c727b81d NM |
205 | .. k3_rst_include_end_common_env_vars_defn |
206 | ||
207 | We will also need some common environment variables set up for the various | |
208 | other build sources. we shall use the following, in the build descriptions below: | |
209 | ||
210 | .. k3_rst_include_start_board_env_vars_desc | |
211 | .. list-table:: Board specific environment variables | |
212 | :widths: 25 25 50 | |
213 | :header-rows: 1 | |
214 | ||
215 | * - S/w Component | |
216 | - Env Variable | |
217 | - Description | |
218 | * - U-Boot | |
219 | - UBOOT_CFG_CORTEXR | |
220 | - Defconfig for Cortex-R (Boot processor). | |
221 | * - U-Boot | |
222 | - UBOOT_CFG_CORTEXA | |
223 | - Defconfig for Cortex-A (MPU processor). | |
224 | * - Trusted Firmware-A | |
225 | - TFA_BOARD | |
226 | - Platform name used for building TF-A for Cortex-A Processor. | |
227 | * - Trusted Firmware-A | |
228 | - TFA_EXTRA_ARGS | |
229 | - Any extra arguments used for building TF-A. | |
230 | * - OP-TEE | |
231 | - OPTEE_PLATFORM | |
232 | - Platform name used for building OP-TEE for Cortex-A Processor. | |
233 | * - OP-TEE | |
234 | - OPTEE_EXTRA_ARGS | |
235 | - Any extra arguments used for building OP-TEE. | |
236 | .. k3_rst_include_end_board_env_vars_desc | |
16a30b34 BB |
237 | |
238 | Building tiboot3.bin | |
239 | ^^^^^^^^^^^^^^^^^^^^^ | |
240 | ||
241 | 1. To generate the U-Boot SPL for the wakeup domain, use the following | |
242 | commands, substituting :code:`{SOC}` for the name of your device (eg: | |
1ee652ab NMF |
243 | am62x) to package the various firmware and the wakeup UBoot SPL into |
244 | the final `tiboot3.bin` binary. (or the `sysfw.itb` if your device | |
245 | uses the split binary flow) | |
16a30b34 | 246 | |
c727b81d | 247 | .. k3_rst_include_start_build_steps_spl_r5 |
febc7f10 | 248 | .. prompt:: bash |
16a30b34 | 249 | |
febc7f10 NM |
250 | # inside u-boot source |
251 | make $UBOOT_CFG_CORTEXR | |
252 | make CROSS_COMPILE=$CC32 BINMAN_INDIRS=$LNX_FW_PATH | |
c727b81d | 253 | .. k3_rst_include_end_build_steps_spl_r5 |
16a30b34 BB |
254 | |
255 | At this point you should have all the needed binaries to boot the wakeup | |
256 | domain of your K3 SoC. | |
257 | ||
258 | **Combined Binary Boot Flow** (eg: am62x, am64x, ... ) | |
259 | ||
1ee652ab | 260 | `tiboot3-{SOC}-{gp/hs-fs/hs}.bin` |
16a30b34 BB |
261 | |
262 | **Split Binary Boot Flow** (eg: j721e, am65x) | |
263 | ||
1ee652ab NMF |
264 | | `tiboot3-{SOC}-{gp/hs-fs/hs}.bin` |
265 | | `sysfw-{SOC}-{gp/hs-fs/hs}-evm.itb` | |
16a30b34 BB |
266 | |
267 | .. note :: | |
268 | ||
269 | It's important to rename the generated `tiboot3.bin` and `sysfw.itb` | |
270 | to match exactly `tiboot3.bin` and `sysfw.itb` as ROM and the wakeup | |
271 | UBoot SPL will only look for and load the files with these names. | |
272 | ||
273 | Building tispl.bin | |
274 | ^^^^^^^^^^^^^^^^^^^ | |
275 | ||
276 | The `tispl.bin` is a standard fitImage combining the firmware need for | |
277 | the main domain to function properly as well as Device Management (DM) | |
278 | firmware if your device using a split firmware. | |
279 | ||
1ee652ab | 280 | 2. We will first need TF-A, as it's the first thing to run on the 'big' |
16a30b34 BB |
281 | application cores on the main domain. |
282 | ||
c727b81d | 283 | .. k3_rst_include_start_build_steps_tfa |
febc7f10 | 284 | .. prompt:: bash |
16a30b34 | 285 | |
febc7f10 NM |
286 | # inside trusted-firmware-a source |
287 | make CROSS_COMPILE=$CC64 ARCH=aarch64 PLAT=k3 SPD=opteed $TFA_EXTRA_ARGS \ | |
288 | TARGET_BOARD=$TFA_BOARD | |
c727b81d | 289 | .. k3_rst_include_end_build_steps_tfa |
16a30b34 | 290 | |
1ee652ab | 291 | Typically all `j7*` devices will use `TARGET_BOARD=generic` or `TARGET_BOARD |
c727b81d | 292 | =j784s4` (if it is a J784S4 device), while typical Sitara (`am6*`) devices |
1ee652ab | 293 | use the `lite` option. |
16a30b34 | 294 | |
1ee652ab | 295 | 3. The Open Portable Trusted Execution Environment (OP-TEE) is designed |
16a30b34 BB |
296 | to run as a companion to a non-secure Linux kernel for Cortex-A cores |
297 | using the TrustZone technology built into the core. | |
298 | ||
c727b81d | 299 | .. k3_rst_include_start_build_steps_optee |
febc7f10 | 300 | .. prompt:: bash |
16a30b34 | 301 | |
febc7f10 NM |
302 | # inside optee_os source |
303 | make CROSS_COMPILE=$CC32 CROSS_COMPILE64=$CC64 CFG_ARM64_core=y $OPTEE_EXTRA_ARGS \ | |
304 | PLATFORM=$OPTEE_PLATFORM | |
c727b81d | 305 | .. k3_rst_include_end_build_steps_optee |
16a30b34 | 306 | |
1ee652ab | 307 | 4. Finally, after TF-A has initialized the main domain and OP-TEE has |
16a30b34 BB |
308 | finished, we can jump back into U-Boot again, this time running on a |
309 | 64bit core in the main domain. | |
310 | ||
c727b81d | 311 | .. k3_rst_include_start_build_steps_uboot |
febc7f10 | 312 | .. prompt:: bash |
16a30b34 | 313 | |
febc7f10 NM |
314 | # inside u-boot source |
315 | make $UBOOT_CFG_CORTEXA | |
316 | make CROSS_COMPILE=$CC64 BINMAN_INDIRS=$LNX_FW_PATH \ | |
c727b81d NM |
317 | BL31=$TFA_PATH/build/k3/$TFA_BOARD/release/bl31.bin \ |
318 | TEE=$OPTEE_PATH/out/arm-plat-k3/core/tee-raw.bin | |
319 | .. k3_rst_include_end_build_steps_uboot | |
16a30b34 BB |
320 | |
321 | At this point you should have every binary needed initialize both the | |
322 | wakeup and main domain and to boot to the U-Boot prompt | |
323 | ||
324 | **Main Domain Bootloader** | |
325 | ||
1ee652ab NMF |
326 | | `tispl.bin` for HS devices or `tispl.bin_unsigned` for GP devices |
327 | | `u-boot.img` for HS devices or `u-boot.img_unsigned` for GP devices | |
a5e8678e MC |
328 | |
329 | Fit Signature Signing | |
330 | --------------------- | |
331 | ||
332 | K3 Platforms have fit signature signing enabled by default on their primary | |
333 | platforms. Here we'll take an example for creating fit image for J721e platform | |
334 | and the same can be extended to other platforms | |
335 | ||
336 | 1. Describing FIT source | |
337 | ||
338 | .. code-block:: bash | |
339 | ||
340 | /dts-v1/; | |
341 | ||
342 | / { | |
343 | description = "Kernel fitImage for j721e-hs-evm"; | |
344 | #address-cells = <1>; | |
345 | ||
346 | images { | |
347 | kernel-1 { | |
348 | description = "Linux kernel"; | |
349 | data = /incbin/("Image"); | |
350 | type = "kernel"; | |
351 | arch = "arm64"; | |
352 | os = "linux"; | |
353 | compression = "none"; | |
354 | load = <0x80080000>; | |
355 | entry = <0x80080000>; | |
356 | hash-1 { | |
357 | algo = "sha512"; | |
358 | }; | |
359 | ||
360 | }; | |
361 | fdt-ti_k3-j721e-common-proc-board.dtb { | |
362 | description = "Flattened Device Tree blob"; | |
363 | data = /incbin/("k3-j721e-common-proc-board.dtb"); | |
364 | type = "flat_dt"; | |
365 | arch = "arm64"; | |
366 | compression = "none"; | |
367 | load = <0x83000000>; | |
368 | hash-1 { | |
369 | algo = "sha512"; | |
370 | }; | |
371 | ||
372 | }; | |
373 | }; | |
374 | ||
375 | configurations { | |
376 | default = "conf-ti_k3-j721e-common-proc-board.dtb"; | |
377 | conf-ti_k3-j721e-common-proc-board.dtb { | |
378 | description = "Linux kernel, FDT blob"; | |
379 | fdt = "fdt-ti_k3-j721e-common-proc-board.dtb"; | |
380 | kernel = "kernel-1"; | |
381 | signature-1 { | |
382 | algo = "sha512,rsa4096"; | |
383 | key-name-hint = "custMpk"; | |
384 | sign-images = "kernel", "fdt"; | |
385 | }; | |
386 | }; | |
387 | }; | |
388 | }; | |
389 | ||
390 | You would require to change the '/incbin/' lines to point to the respective | |
391 | files in your local machine and the key-name-hint also needs to be changed | |
392 | if you are using some other key other than the TI dummy key that we are | |
393 | using for this example. | |
394 | ||
395 | 2. Compile U-boot for the respective board | |
396 | ||
c727b81d NM |
397 | .. include:: k3.rst |
398 | :start-after: .. k3_rst_include_start_build_steps_uboot | |
399 | :end-before: .. k3_rst_include_end_build_steps_uboot | |
a5e8678e | 400 | |
c727b81d | 401 | .. note:: |
a5e8678e MC |
402 | |
403 | The changes only affect a72 binaries so the example just builds that | |
404 | ||
405 | 3. Sign the fit image and embed the dtb in uboot | |
406 | ||
407 | Now once the build is done, you'll have a dtb for your board that you'll | |
408 | be passing to mkimage for signing the fitImage and embedding the key in | |
409 | the u-boot dtb. | |
410 | ||
febc7f10 | 411 | .. prompt:: bash |
a5e8678e MC |
412 | |
413 | mkimage -r -f fitImage.its -k $UBOOT_PATH/board/ti/keys -K | |
414 | $UBOOT_PATH/build/a72/dts/dt.dtb | |
415 | ||
416 | For signing a secondary platform, pass the -K parameter to that DTB | |
417 | ||
febc7f10 | 418 | .. prompt:: bash |
a5e8678e MC |
419 | |
420 | mkimage -f fitImage.its -k $UBOOT_PATH/board/ti/keys -K | |
421 | $UBOOT_PATH/build/a72/arch/arm/dts/k3-j721e-sk.dtb | |
422 | ||
423 | .. note:: | |
424 | ||
425 | If changing `CONFIG_DEFAULT_DEVICE_TREE` to the secondary platform, | |
426 | binman changes would also be required so that correct dtb gets packaged. | |
427 | ||
428 | .. code-block:: bash | |
429 | ||
430 | diff --git a/arch/arm/dts/k3-j721e-binman.dtsi b/arch/arm/dts/k3-j721e-binman.dtsi | |
431 | index 673be646b1e3..752fa805fe8d 100644 | |
432 | --- a/arch/arm/dts/k3-j721e-binman.dtsi | |
433 | +++ b/arch/arm/dts/k3-j721e-binman.dtsi | |
434 | @@ -299,8 +299,8 @@ | |
435 | #define SPL_J721E_SK_DTB "spl/dts/k3-j721e-sk.dtb" | |
436 | ||
437 | #define UBOOT_NODTB "u-boot-nodtb.bin" | |
438 | -#define J721E_EVM_DTB "u-boot.dtb" | |
439 | -#define J721E_SK_DTB "arch/arm/dts/k3-j721e-sk.dtb" | |
440 | +#define J721E_EVM_DTB "arch/arm/dts/k3-j721e-common-proc-board.dtb" | |
441 | +#define J721E_SK_DTB "u-boot.dtb" | |
442 | ||
443 | 5. Rebuilt u-boot | |
444 | ||
445 | This is required so that the modified dtb gets updated in u-boot.img | |
446 | ||
c727b81d NM |
447 | .. include:: k3.rst |
448 | :start-after: .. k3_rst_include_start_build_steps_uboot | |
449 | :end-before: .. k3_rst_include_end_build_steps_uboot | |
a5e8678e MC |
450 | |
451 | 6. (Optional) Enabled FIT_SIGNATURE_ENFORCED | |
452 | ||
453 | By default u-boot will boot up the fit image without any authentication as | |
454 | such if the public key is not embedded properly, to check if the public key | |
455 | nodes are proper you can enable FIT_SIGNATURE_ENFORCED that would not rely | |
456 | on the dtb for anything else then the signature node for checking the fit | |
457 | image, rest other things will be enforced such as the property of | |
458 | required-keys. This is not an extensive check so do manual checks also | |
459 | ||
460 | This is by default enabled for devices with TI_SECURE_DEVICE enabled. | |
461 | ||
462 | .. note:: | |
463 | ||
464 | The devices now also have distroboot enabled so if the fit image doesn't | |
465 | work then the fallback to normal distroboot will be there on hs devices, | |
466 | this will need to be explicitly disabled by changing the boot_targets. | |
467 | ||
468 | Saving environment | |
469 | ------------------ | |
470 | ||
471 | SAVEENV is disabled by default and for the new flow uses Uenv.txt as the default | |
472 | way for saving the environments. This has been done as Uenv.txt is more granular | |
473 | then the saveenv command and can be used across various bootmodes too. | |
474 | ||
475 | **Writing to MMC/EMMC** | |
476 | ||
febc7f10 NM |
477 | .. prompt:: bash |
478 | :prompts: => | |
a5e8678e | 479 | |
febc7f10 NM |
480 | env export -t $loadaddr <list of variables> |
481 | fatwrite mmc ${mmcdev} ${loadaddr} ${bootenvfile} ${filesize} | |
a5e8678e MC |
482 | |
483 | **Reading from MMC/EMMC** | |
484 | ||
485 | By default run envboot will read it from the MMC/EMMC partition ( based on | |
486 | mmcdev) and set the environments. | |
487 | ||
488 | If manually needs to be done then the environment can be read from the | |
489 | filesystem and then imported | |
490 | ||
febc7f10 NM |
491 | .. prompt:: bash |
492 | :prompts: => | |
a5e8678e | 493 | |
febc7f10 NM |
494 | fatload mmc ${mmcdev} ${loadaddr} ${bootenvfile} |
495 | env import -t ${loadaddr} ${filesize} | |
effe5085 JK |
496 | |
497 | .. _k3_rst_refer_openocd: | |
498 | ||
499 | Common Debugging environment - OpenOCD | |
500 | -------------------------------------- | |
501 | ||
502 | This section will show you how to connect a board to `OpenOCD | |
503 | <https://openocd.org/>`_ and load the SPL symbols for debugging with | |
504 | a K3 generation device. To follow this guide, you must build custom | |
505 | u-boot binaries, start your board from a boot media such as an SD | |
506 | card, and use an OpenOCD environment. This section uses generic | |
507 | examples, though you can apply these instructions to any supported K3 | |
508 | generation device. | |
509 | ||
510 | The overall structure of this setup is in the following figure. | |
511 | ||
512 | .. image:: img/openocd-overview.svg | |
c6df5289 | 513 | :alt: Overview of OpenOCD setup. |
effe5085 JK |
514 | |
515 | .. note:: | |
516 | ||
517 | If you find these instructions useful, please consider `donating | |
518 | <https://openocd.org/pages/donations.html>`_ to OpenOCD. | |
519 | ||
520 | Step 1: Download and install OpenOCD | |
521 | ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ | |
522 | ||
523 | To get started, it is more convenient if the distribution you | |
524 | use supports OpenOCD by default. Follow the instructions in the | |
525 | `getting OpenOCD <https://openocd.org/pages/getting-openocd.html>`_ | |
526 | documentation to pick the installation steps appropriate to your | |
527 | environment. Some references to OpenOCD documentation: | |
528 | ||
529 | * `OpenOCD User Guide <https://openocd.org/doc/html/index.html>`_ | |
530 | * `OpenOCD Developer's Guide <https://openocd.org/doc/doxygen/html/index.html>`_ | |
531 | ||
532 | Refer to the release notes corresponding to the `OpenOCD version | |
533 | <https://github.com/openocd-org/openocd/releases>`_ to ensure | |
534 | ||
535 | * Processor support: In general, processor support shouldn't present | |
536 | any difficulties since OpenOCD provides solid support for both ARMv8 | |
537 | and ARMv7. | |
538 | * SoC support: When working with System-on-a-Chip (SoC), the support | |
539 | usually comes as a TCL config file. It is vital to ensure the correct | |
540 | version of OpenOCD or to use the TCL files from the latest release or | |
541 | the one mentioned. | |
542 | * Board or the JTAG adapter support: In most cases, board support is | |
543 | a relatively easy problem if the board has a JTAG pin header. All | |
544 | you need to do is ensure that the adapter you select is compatible | |
545 | with OpenOCD. Some boards come with an onboard JTAG adapter that | |
546 | requires a USB cable to be plugged into the board, in which case, it | |
547 | is vital to ensure that the JTAG adapter is supported. Fortunately, | |
548 | almost all TI K3 SK/EVMs come with TI's XDS110, which has out of the | |
549 | box support by OpenOCD. The board-specific documentation will | |
550 | cover the details and any adapter/dongle recommendations. | |
551 | ||
febc7f10 | 552 | .. prompt:: bash |
effe5085 JK |
553 | |
554 | openocd -v | |
555 | ||
556 | .. note:: | |
557 | ||
558 | OpenOCD version 0.12.0 is usually required to connect to most K3 | |
559 | devices. If your device is only supported by a newer version than the | |
560 | one provided by your distribution, you may need to build it from the source. | |
561 | ||
562 | Building OpenOCD from source | |
563 | """""""""""""""""""""""""""" | |
564 | ||
565 | The dependency package installation instructions below are for Debian | |
566 | systems, but equivalent instructions should exist for systems with | |
567 | other package managers. Please refer to the `OpenOCD Documentation | |
568 | <https://openocd.org/>`_ for more recent installation steps. | |
569 | ||
febc7f10 | 570 | .. prompt:: bash |
effe5085 | 571 | |
febc7f10 NM |
572 | # Check the packages to be installed: needs deb-src in sources.list |
573 | sudo apt build-dep openocd | |
574 | # The following list is NOT complete - please check the latest | |
575 | sudo apt-get install libtool pkg-config texinfo libusb-dev \ | |
effe5085 | 576 | libusb-1.0.0-dev libftdi-dev libhidapi-dev autoconf automake |
febc7f10 NM |
577 | git clone https://github.com/openocd-org/openocd.git openocd |
578 | cd openocd | |
579 | git submodule init | |
580 | git submodule update | |
581 | ./bootstrap | |
582 | ./configure --prefix=/usr/local/ | |
583 | make -j`nproc` | |
584 | sudo make install | |
effe5085 JK |
585 | |
586 | .. note:: | |
587 | ||
588 | The example above uses the GitHub mirror site. See | |
589 | `git repo information <https://openocd.org/doc/html/Developers.html#OpenOCD-Git-Repository>`_ | |
590 | information to pick the official git repo. | |
591 | If a specific version is desired, select the version using `git checkout tag`. | |
592 | ||
593 | Installing OpenOCD udev rules | |
594 | """"""""""""""""""""""""""""" | |
595 | ||
596 | The step is not necessary if the distribution supports the OpenOCD, but | |
597 | if building from a source, ensure that the udev rules are installed | |
598 | correctly to ensure a sane system. | |
599 | ||
febc7f10 | 600 | .. prompt:: bash |
effe5085 JK |
601 | |
602 | # Go to the OpenOCD source directory | |
febc7f10 NM |
603 | cd openocd |
604 | Copy the udev rules to the correct system location | |
605 | sudo cp ./contrib/60-openocd.rules \ | |
975103f1 | 606 | ./src/jtag/drivers/libjaylink/contrib/99-libjaylink.rules \ |
effe5085 JK |
607 | /etc/udev/rules.d/ |
608 | # Get Udev to load the new rules up | |
febc7f10 | 609 | sudo udevadm control --reload-rules |
effe5085 | 610 | # Use the new rules on existing connected devices |
febc7f10 | 611 | sudo udevadm trigger |
effe5085 JK |
612 | |
613 | Step 2: Setup GDB | |
614 | ^^^^^^^^^^^^^^^^^ | |
615 | ||
616 | Most systems come with gdb-multiarch package. | |
617 | ||
febc7f10 | 618 | .. prompt:: bash |
effe5085 JK |
619 | |
620 | # Install gdb-multiarch package | |
febc7f10 | 621 | sudo apt-get install gdb-multiarch |
effe5085 JK |
622 | |
623 | Though using GDB natively is normal, developers with interest in using IDE | |
624 | may find a few of these interesting: | |
625 | ||
626 | * `gdb-dashboard <https://github.com/cyrus-and/gdb-dashboard>`_ | |
627 | * `gef <https://github.com/hugsy/gef>`_ | |
628 | * `peda <https://github.com/longld/peda>`_ | |
629 | * `pwndbg <https://github.com/pwndbg/pwndbg>`_ | |
630 | * `voltron <https://github.com/snare/voltron>`_ | |
631 | * `ddd <https://www.gnu.org/software/ddd/>`_ | |
632 | * `vscode <https://www.justinmklam.com/posts/2017/10/vscode-debugger-setup/>`_ | |
633 | * `vim conque-gdb <https://github.com/vim-scripts/Conque-GDB>`_ | |
634 | * `emacs realgud <https://github.com/realgud/realgud/wiki/gdb-notes>`_ | |
635 | * `Lauterbach IDE <https://www2.lauterbach.com/pdf/backend_gdb.pdf>`_ | |
636 | ||
637 | .. warning:: | |
638 | LLDB support for OpenOCD is still a work in progress as of this writing. | |
639 | Using GDB is probably the safest option at this point in time. | |
640 | ||
641 | Step 3: Connect board to PC | |
642 | ^^^^^^^^^^^^^^^^^^^^^^^^^^^ | |
643 | There are few patterns of boards in the ecosystem | |
644 | ||
645 | .. k3_rst_include_start_openocd_connect_XDS110 | |
646 | ||
647 | **Integrated JTAG adapter/dongle**: The board has a micro-USB connector labelled | |
648 | XDS110 USB or JTAG. Connect a USB cable to the board to the mentioned port. | |
649 | ||
650 | .. note:: | |
651 | ||
652 | There are multiple USB ports on a typical board, So, ensure you have read | |
653 | the user guide for the board and confirmed the silk screen label to ensure | |
654 | connecting to the correct port. | |
655 | ||
656 | .. k3_rst_include_end_openocd_connect_XDS110 | |
657 | ||
658 | .. k3_rst_include_start_openocd_connect_cti20 | |
659 | ||
660 | **cTI20 connector**: The TI's `cTI20 | |
661 | <https://software-dl.ti.com/ccs/esd/documents/xdsdebugprobes/emu_JTAG_connectors.html#cti-20-pin-header-information>`_ connector | |
662 | is probably the most prevelant on TI platforms. Though many | |
663 | TI boards have an onboard XDS110, cTI20 connector is usually | |
664 | provided as an alternate scheme to connect alternatives such | |
665 | as `Lauterbach <https://www.lauterbach.com/>`_ or `XDS560 | |
666 | <https://www.ti.com/tool/TMDSEMU560V2STM-U>`_. | |
667 | ||
668 | To debug on these boards, the following combinations is suggested: | |
669 | ||
670 | * `TUMPA <https://www.diygadget.com/JTAG-cables-and-microcontroller-programmers/tiao-usb-multi-protocol-adapter-JTAG-spi-i2c-serial>`_ | |
671 | or `equivalent dongles supported by OpenOCD. <https://openocd.org/doc/html/Debug-Adapter-Hardware.html#Debug-Adapter-Hardware>`_ | |
672 | * Cable such as `Tag-connect ribbon cable <https://www.tag-connect.com/product/20-pin-cortex-ribbon-cable-4-length-with-50-mil-connectors>`_ | |
673 | * Adapter to convert cTI20 to ARM20 such as those from | |
674 | `Segger <https://www.segger.com/products/debug-probes/j-link/accessories/adapters/ti-cti-20-adapter/>`_ | |
675 | or `Lauterbach LA-3780 <https://www.lauterbach.com/ad3780.html>`_ | |
676 | Or optionally, if you have manufacturing capability then you could try | |
677 | `BeagleBone JTAG Adapter <https://github.com/mmorawiec/BeagleBone-Black-JTAG-Adapters>`_ | |
678 | ||
679 | .. warning:: | |
680 | XDS560 and Lauterbach are proprietary solutions and is not supported by | |
681 | OpenOCD. | |
682 | When purchasing an off the shelf adapter/dongle, you do want to be careful | |
683 | about the signalling though. Please | |
684 | `read for additional info <https://software-dl.ti.com/ccs/esd/xdsdebugprobes/emu_JTAG_connectors.html>`_. | |
685 | ||
686 | .. k3_rst_include_end_openocd_connect_cti20 | |
687 | ||
688 | .. k3_rst_include_start_openocd_connect_tag_connect | |
689 | ||
690 | **Tag-Connect**: `Tag-Connect <https://www.tag-connect.com/>`_ | |
691 | pads on the boards which require special cable. Please check the documentation | |
692 | to `identify <https://www.tag-connect.com/info/legs-or-no-legs>`_ if "legged" | |
693 | or "no-leg" version of the cable is appropriate for the board. | |
694 | ||
695 | To debug on these boards, you will need: | |
696 | ||
697 | * `TUMPA <https://www.diygadget.com/JTAG-cables-and-microcontroller-programmers/tiao-usb-multi-protocol-adapter-JTAG-spi-i2c-serial>`_ | |
698 | or `equivalent dongles supported by OpenOCD <https://openocd.org/doc/html/Debug-Adapter-Hardware.html#Debug-Adapter-Hardware>`_. | |
699 | * Tag-Connect cable appropriate to the board such as | |
700 | `TC2050-IDC-NL <https://www.tag-connect.com/product/TC2050-IDC-NL-10-pin-no-legs-cable-with-ribbon-connector>`_ | |
701 | * In case of no-leg, version, a | |
702 | `retaining clip <https://www.tag-connect.com/product/tc2050-clip-3pack-retaining-clip>`_ | |
703 | * Tag-Connect to ARM20 | |
704 | `adapter <https://www.tag-connect.com/product/tc2050-arm2010-arm-20-pin-to-tc2050-adapter>`_ | |
705 | ||
706 | .. note:: | |
707 | You can optionally use a 3d printed solution such as | |
708 | `Protective cap <https://www.thingiverse.com/thing:3025584>`_ or | |
709 | `clip <https://www.thingiverse.com/thing:3035278>`_ to replace | |
710 | the retaining clip. | |
711 | ||
712 | .. warning:: | |
713 | With the Tag-Connect to ARM20 adapter, Please solder the "Trst" signal for | |
714 | connection to work. | |
715 | ||
716 | .. k3_rst_include_end_openocd_connect_tag_connect | |
717 | ||
718 | Debugging with OpenOCD | |
719 | ^^^^^^^^^^^^^^^^^^^^^^ | |
720 | ||
721 | Debugging U-Boot is different from debugging regular user space | |
722 | applications. The bootloader initialization process involves many boot | |
723 | media and hardware configuration operations. For K3 devices, there | |
724 | are also interactions with security firmware. While reloading the | |
725 | "elf" file works through GDB, developers must be mindful of cascading | |
726 | initialization's potential consequences. | |
727 | ||
728 | Consider the following code change: | |
729 | ||
730 | .. code-block:: diff | |
731 | ||
732 | --- a/file.c 2023-07-29 10:55:29.647928811 -0500 | |
733 | +++ b/file.c 2023-07-29 10:55:46.091856816 -0500 | |
734 | @@ -1,3 +1,3 @@ | |
735 | val = readl(reg); | |
736 | -val |= 0x2; | |
737 | +val |= 0x1; | |
738 | writel(val, reg); | |
739 | ||
740 | Re-running the elf file with the above change will result in the | |
741 | register setting 0x3 instead of the intended 0x1. There are other | |
742 | hardware blocks which may not behave very well with a re-initialization | |
743 | without proper shutdown. | |
744 | ||
745 | To help narrow the debug down, it is usually simpler to use the | |
746 | standard boot media to get to the bootloader and debug only in the area | |
747 | of interest. | |
748 | ||
749 | In general, to debug u-boot spl/u-boot with OpenOCD there are three steps: | |
750 | ||
751 | * Modify the code adding a loop to allow the debugger to attach | |
752 | near the point of interest. Boot up normally to stop at the loop. | |
753 | * Connect with OpenOCD and step out of the loop. | |
754 | * Step through the code to find the root of issue. | |
755 | ||
756 | Typical debugging involves a few iterations of the above sequence. | |
757 | Though most bootloader developers like to use printf to debug, | |
758 | debug with JTAG tends to be most efficient since it is possible to | |
759 | investigate the code flow and inspect hardware registers without | |
760 | repeated iterations. | |
761 | ||
762 | Code modification | |
763 | """"""""""""""""" | |
764 | ||
765 | * **start.S**: Adding an infinite while loop at the very entry of | |
766 | U-Boot. For this, look for the corresponding start.S entry file. | |
767 | This is usually only required when debugging some core SoC or | |
768 | processor related function. For example: arch/arm/cpu/armv8/start.S or | |
769 | arch/arm/cpu/armv7/start.S | |
770 | ||
771 | .. code-block:: diff | |
772 | ||
773 | diff --git a/arch/arm/cpu/armv7/start.S b/arch/arm/cpu/armv7/start.S | |
774 | index 69e281b086..744929e825 100644 | |
775 | --- a/arch/arm/cpu/armv7/start.S | |
776 | +++ b/arch/arm/cpu/armv7/start.S | |
777 | @@ -37,6 +37,8 @@ | |
778 | #endif | |
779 | ||
780 | reset: | |
781 | +dead_loop: | |
782 | + b dead_loop | |
783 | /* Allow the board to save important registers */ | |
784 | b save_boot_params | |
785 | save_boot_params_ret: | |
786 | ||
787 | * **board_init_f**: Adding an infinite while loop at the board entry | |
788 | function. In many cases, it is important to debug the boot process if | |
789 | any changes are made for board-specific applications. Below is a step | |
790 | by step process for debugging the boot SPL or Armv8 SPL: | |
791 | ||
792 | To debug the boot process in either domain, we will first | |
793 | add a modification to the code we would like to debug. | |
794 | In this example, we will debug ``board_init_f`` inside | |
795 | ``arch/arm/mach-k3/{soc}_init.c``. Since some sections of U-Boot | |
796 | will be executed multiple times during the bootup process of K3 | |
975103f1 | 797 | devices, we will need to include either ``CONFIG_ARM64`` or |
effe5085 JK |
798 | ``CONFIG_CPU_V7R`` to catch the CPU at the desired place during the |
799 | bootup process (Main or Wakeup domains). For example, modify the | |
800 | file as follows (depending on need): | |
801 | ||
802 | .. code-block:: c | |
803 | ||
804 | void board_init_f(ulong dummy) | |
805 | { | |
806 | . | |
807 | . | |
808 | /* Code to run on the R5F (Wakeup/Boot Domain) */ | |
809 | if (IS_ENABLED(CONFIG_CPU_V7R)) { | |
810 | volatile int x = 1; | |
811 | while(x) {}; | |
812 | } | |
813 | ... | |
814 | /* Code to run on the ARMV8 (Main Domain) */ | |
975103f1 | 815 | if (IS_ENABLED(CONFIG_ARM64)) { |
effe5085 JK |
816 | volatile int x = 1; |
817 | while(x) {}; | |
818 | } | |
819 | . | |
820 | . | |
821 | } | |
822 | ||
823 | Connecting with OpenOCD for a debug session | |
824 | """"""""""""""""""""""""""""""""""""""""""" | |
825 | ||
826 | Startup OpenOCD to debug the platform as follows: | |
827 | ||
828 | * **Integrated JTAG interface**: If the evm has a debugger such as | |
829 | XDS110 inbuilt, there is typically an evm board support added and a | |
830 | cfg file will be available. | |
831 | ||
832 | .. k3_rst_include_start_openocd_cfg_XDS110 | |
833 | ||
febc7f10 | 834 | .. prompt:: bash |
effe5085 JK |
835 | |
836 | openocd -f board/{board_of_choice}.cfg | |
837 | ||
838 | .. k3_rst_include_end_openocd_cfg_XDS110 | |
839 | ||
840 | .. k3_rst_include_start_openocd_cfg_external_intro | |
841 | ||
842 | * **External JTAG adapter/interface**: In other cases, where an | |
843 | adapter/dongle is used, a simple cfg file can be created to integrate the | |
844 | SoC and adapter information. See `supported TI K3 SoCs | |
845 | <https://github.com/openocd-org/openocd/blob/master/tcl/target/ti_k3.cfg#L59>`_ | |
846 | to decide if the SoC is supported or not. | |
847 | ||
febc7f10 | 848 | .. prompt:: bash |
effe5085 JK |
849 | |
850 | openocd -f openocd_connect.cfg | |
851 | ||
852 | .. k3_rst_include_end_openocd_cfg_external_intro | |
853 | ||
854 | For example, with BeaglePlay (AM62X platform), the openocd_connect.cfg: | |
855 | ||
856 | .. code-block:: tcl | |
857 | ||
858 | # TUMPA example: | |
859 | # http://www.tiaowiki.com/w/TIAO_USB_Multi_Protocol_Adapter_User's_Manual | |
860 | source [find interface/ftdi/tumpa.cfg] | |
861 | ||
862 | transport select jtag | |
863 | ||
864 | # default JTAG configuration has only SRST and no TRST | |
865 | reset_config srst_only srst_push_pull | |
866 | ||
867 | # delay after SRST goes inactive | |
868 | adapter srst delay 20 | |
869 | ||
870 | if { ![info exists SOC] } { | |
871 | # Set the SoC of interest | |
872 | set SOC am625 | |
873 | } | |
874 | ||
875 | source [find target/ti_k3.cfg] | |
876 | ||
877 | ftdi tdo_sample_edge falling | |
878 | ||
879 | # Speeds for FT2232H are in multiples of 2, and 32MHz is tops | |
880 | # max speed we seem to achieve is ~20MHz.. so we pick 16MHz | |
881 | adapter speed 16000 | |
882 | ||
883 | Below is an example of the output of this command: | |
884 | ||
885 | .. code-block:: console | |
886 | ||
887 | Info : Listening on port 6666 for tcl connections | |
888 | Info : Listening on port 4444 for telnet connections | |
889 | Info : XDS110: connected | |
890 | Info : XDS110: vid/pid = 0451/bef3 | |
891 | Info : XDS110: firmware version = 3.0.0.20 | |
892 | Info : XDS110: hardware version = 0x002f | |
893 | Info : XDS110: connected to target via JTAG | |
894 | Info : XDS110: TCK set to 2500 kHz | |
895 | Info : clock speed 2500 kHz | |
896 | Info : JTAG tap: am625.cpu tap/device found: 0x0bb7e02f (mfg: 0x017 (Texas Instruments), part: 0xbb7e, ver: 0x0) | |
897 | Info : starting gdb server for am625.cpu.sysctrl on 3333 | |
898 | Info : Listening on port 3333 for gdb connections | |
899 | Info : starting gdb server for am625.cpu.a53.0 on 3334 | |
900 | Info : Listening on port 3334 for gdb connections | |
901 | Info : starting gdb server for am625.cpu.a53.1 on 3335 | |
902 | Info : Listening on port 3335 for gdb connections | |
903 | Info : starting gdb server for am625.cpu.a53.2 on 3336 | |
904 | Info : Listening on port 3336 for gdb connections | |
905 | Info : starting gdb server for am625.cpu.a53.3 on 3337 | |
906 | Info : Listening on port 3337 for gdb connections | |
907 | Info : starting gdb server for am625.cpu.main0_r5.0 on 3338 | |
908 | Info : Listening on port 3338 for gdb connections | |
909 | Info : starting gdb server for am625.cpu.gp_mcu on 3339 | |
910 | Info : Listening on port 3339 for gdb connections | |
911 | ||
912 | .. note:: | |
913 | Notice the default configuration is non-SMP configuration allowing | |
914 | for each of the core to be attached and debugged simultaneously. | |
915 | ARMv8 SPL/U-Boot starts up on cpu0 of a53/a72. | |
916 | ||
917 | .. k3_rst_include_start_openocd_cfg_external_gdb | |
918 | ||
919 | To debug using this server, use GDB directly or your preferred | |
920 | GDB-based IDE. To start up GDB in the terminal, run the following | |
921 | command. | |
922 | ||
febc7f10 | 923 | .. prompt:: bash |
effe5085 JK |
924 | |
925 | gdb-multiarch | |
926 | ||
927 | To connect to your desired core, run the following command within GDB: | |
928 | ||
929 | .. code-block:: bash | |
930 | ||
931 | target extended-remote localhost:{port for desired core} | |
932 | ||
933 | To load symbols: | |
934 | ||
935 | .. warning:: | |
936 | ||
937 | SPL and U-Boot does a re-location of address compared to where it | |
938 | is loaded originally. This step takes place after the DDR size is | |
939 | determined from dt parsing. So, debugging can be split into either | |
940 | "before re-location" or "after re-location". Please refer to the | |
941 | file ''doc/README.arm-relocation'' to see how to grab the relocation | |
942 | address. | |
943 | ||
944 | * Prior to relocation: | |
945 | ||
946 | .. code-block:: bash | |
947 | ||
948 | symbol-file {path to elf file} | |
949 | ||
950 | * After relocation: | |
951 | ||
952 | .. code-block:: bash | |
953 | ||
954 | # Drop old symbol file | |
955 | symbol-file | |
956 | # Pick up new relocaddr | |
957 | add-symbol-file {path to elf file} {relocaddr} | |
958 | ||
959 | .. k3_rst_include_end_openocd_cfg_external_gdb | |
960 | ||
961 | In the above example of AM625, | |
962 | ||
963 | .. code-block:: bash | |
964 | ||
965 | target extended-remote localhost:3338 <- R5F (Wakeup Domain) | |
966 | target extended-remote localhost:3334 <- A53 (Main Domain) | |
967 | ||
968 | The core can now be debugged directly within GDB using GDB commands or | |
969 | if using IDE, as appropriate to the IDE. | |
970 | ||
971 | Stepping through the code | |
972 | """"""""""""""""""""""""" | |
973 | ||
974 | `GDB TUI Commands | |
975 | <https://sourceware.org/gdb/onlinedocs/gdb/TUI-Commands.html>`_ can | |
976 | help set up the display more sensible for debug. Provide the name | |
977 | of the layout that can be used to debug. For example, use the GDB | |
978 | command ``layout src`` after loading the symbols to see the code and | |
979 | breakpoints. To exit the debug loop added above, add any breakpoints | |
980 | needed and run the following GDB commands to step out of the debug | |
981 | loop set in the ``board_init_f`` function. | |
982 | ||
983 | .. code-block:: bash | |
984 | ||
985 | set x = 0 | |
986 | continue | |
987 | ||
988 | The platform has now been successfully setup to debug with OpenOCD | |
989 | using GDB commands or a GDB-based IDE. See `OpenOCD documentation for | |
990 | GDB <https://openocd.org/doc/html/GDB-and-OpenOCD.html>`_ for further | |
991 | information. | |
992 | ||
993 | .. warning:: | |
994 | ||
995 | On the K3 family of devices, a watchdog timer within the DMSC is | |
996 | enabled by default by the ROM bootcode with a timeout of 3 minutes. | |
997 | The watchdog timer is serviced by System Firmware (SYSFW) or TI | |
998 | Foundational Security (TIFS) during normal operation. If debugging | |
999 | the SPL before the SYSFW is loaded, the watchdog timer will not get | |
1000 | serviced automatically and the debug session will reset after 3 | |
1001 | minutes. It is recommended to start debugging SPL code only after | |
1002 | the startup of SYSFW to avoid running into the watchdog timer reset. | |
1003 | ||
1004 | Miscellaneous notes with OpenOCD | |
1005 | ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ | |
1006 | ||
1007 | Currently, OpenOCD does not support tracing for K3 platforms. Tracing | |
1008 | function could be beneficial if the bug in code occurs deep within | |
1009 | nested function and can optionally save developers major trouble of | |
1010 | stepping through a large quantity of code. |