Last edited one week ago

How to debug OP-TEE

Applicable for STM32MP13x lines, STM32MP15x lines

1. Purpose[edit | edit source]

This article explains how to debug the OP-TEE secure world binaries.
This debug information is specifically linked to the CPU secure state (Arm® TrustZone®).

The OP-TEE secure world binaries include OP-TEE core (privilege firmware) and OP-TEE trusted applications and libraries (user space context):

  • There are two main ways to debug OP-TEE core: using embedded traces, or using JTAG/SWD to access the secure world.
    The focus here is on the solution integrated in OpenSTLinux: debug over GDB (ST-LINK or JTAG/SWD based).
  • The OP-TEE trusted applications and libraries provide debug support relying on embedded traces only.
    Debugging the secure userland binaries through JTAG/SWD resources is not recommended, and OP-TEE does not provide any means to embed a GDB server in the secure world.
Warning white.png Warning
This article focuses on OP-TEE debug.
Refer to STM32MP1 Platform trace and debug environment overview article for more generic information.

2. Debugging OP-TEE core[edit | edit source]

2.1. OP-TEE Version number[edit | edit source]

The starting point for debugging OP-TEE core is to identify the OP-TEE version embedded in the target. A version identifier is displayed on the console with the following format:

I/TC: OP-TEE version:<tag> #<buildcount> <date> <arch>

that is:

I/TC: OP-TEE version: openstlinux-19-01-11-10-g56ef3b0 #1 Wed Jan 30 09:12:56 UTC 2019 arm

2.2. Embedded assertions[edit | edit source]

OP-TEE core can embed debug assertions that panic the system when the tested condition is not met. Embedded assertions are implemented using the assert() function, that is, in the following code snippet, the system panics if argument1 is negative, while the function returns the decremented value of the input argument:

static int increment_argument(int argument)
{
   assert(argument > 0);
   return argument - 1;
}

Assertions are embedded (or not) depending on the configuration directive CFG_TEE_CORE_DEBUG={n|y} when the OP-TEE core is built.

2.3. Debug with traces[edit | edit source]

OP-TEE provides trace message support and a configurable trace level build directive CFG_TEE_CORE_LOG_LEVEL={0|1|2|3|4}. This directive defines the trace levels that are embedded inside the firmware and output through the OP-TEE console. A low value reduces the memory footprint of the firmware and increases its runtime performance.

Value Name Description with related macros
0 - All trace messages are disabled
1 Error trace level Only non-tagged and error trace messages are embedded: MSG(), MSG_RAW(), EMSG(), EMSG_RAW()
2 Info trace level + info trace messages: IMSG(), IMSG_RAW()
3 Debug trace level + debug trace messages: DMSG(), DMSG_RAW()
4 Flow trace level + flow trace messages: FMSG(), FMSG_RAW()
Warning white.png Warning
Concurrent enabling of all debug supports may not be possible due to the internal memory size constraint. If required, change some DMSG() into MSG() to force messages to output even at a low verbosity level.

Traces and errors are available on the console defined in the chosen node of the device tree by the stdout-path property:

chosen {
        stdout-path = "serial0:115200n8";
};

More information about OP-TEE build and update is available in the How to configure OP-TEE article.

2.4. Stack unwind support[edit | edit source]

The OP-TEE core can trace the execution backtrace when its panics. The execution backtrace allows analysis of the execution call sequence that led to the panic. The OP-TEE OS provides tools to analyse such backtraces based on the OP-TEE core ELF file generated at build time.

Backtrace unwind is embedded (or not) depending on the configuration directive CFG_UNWIND={y|n} Note that backtrace unwind increases the size of the OP-TEE core firmware. When the OP-TEE core executes from a small secure RAM, enabling stack unwind penalizes the OP-TEE core performance.

More information is available from the OP-TEE abort-dumps documentation[1].

2.5. Debug with GDB[edit | edit source]

The Debug OpenSTLinux BSP Components with GDB article describes how to set up the GDB / OpenOCD environment.
OP-TEE can be debugged through JTAG/SWD using an ST-LINK or the JTAG/SWD output, depending on the target board.

Note that when the OP-TEE core executes in a small secure memory with the support of its pager (case build directive CFG_WITH_PAGER=y), use of hardware breakpoints rather than software breakpoints is highly recommended. Since most of the OP-TEE core instructions are dynamically loaded into the small secure memory, a loaded software breakpoint is likely to be discarded when the OP-TEE pager wipes memory content to load other OP-TEE core pages.

When OP-TEE executes in the large main memory (case build directive CFG_WITH_PAGER is disabled), all OP-TEE core resources are resident. In this case, hardware breakpoints as well as software breakpoints can be used without any issues.

2.5.1. Debug boot sequence[edit | edit source]

Load symbols to the target offset:

(gdb) add-symbol-file <path_to_build_folder>/tee.elf <load_address>

OP-TEE load address is available from the generated tee-init_load_addr.txt file.
It can also be found in the generated tee.map file:

...
Linker script and memory map

                0x000000002ffc0000                . = 0x2ffc0000
                0x0000000000000001                ASSERT (0x1, text start should align to 32bytes)
                0x000000002ffc0000                __text_start = .
                0x000000002ffc0000                __flatmap_unpg_rx_start = ((__text_start / 0x1000) * 0x1000)

.text           0x000000002ffc0000     0xc538
 *(SORT_BY_ALIGNMENT(.text._start))
 .text._start   0x000000002ffc0000       0x98 out/stm32mp157c-ev1/core/arch/arm/kernel/entry_a32.o
                0x000000002ffc0000                _start  -> OP-TEE Load address
...

In this example, the OP-TEE load address is 0x2ffc0000.

All OP-TEE core symbols can be loaded:

(gdb) add-symbol-file <path_to_build_folder>/tee.elf 0x2ffc0000

Thanks to the Wrapper_for_FSBL_images, you will be able to debug the initial boot sequence. Once the board starts and waits into the FSBL debug wrapper, a hardware breakpoint can be set at the OP-TEE core entry point.

(gdb) hb _start
Warning white.png Warning
OP-TEE symbols may override the FSBL one, so you cannot load both TF-A and OP-TEE symbols at the same time.

2.5.2. Debugging during runtime execution[edit | edit source]

Once U-Boot or the Linux kernel is running, secure memory or regions cannot be accessed, but it is possible to break by setting a hardware breakpoint on OP-TEE service handler. GDB breaks once it has switched into the secure world and reached the break instruction. Once halted, GDB can access secure resources as peripheral interfaces or memories. For example, to break into U-Boot use the following GDB instructions:

(gdb) hb stm32_sip_service
(gdb) continue

On the first service call occurrence GDB breaks into the stm32_sip_service() entry in the OP-TEE core.

3. Debugging OP-TEE trusted applications[edit | edit source]

3.1. Embedded assertions[edit | edit source]

OP-TEE trusted applications can embed assertions as well as the OP-TEE core, as described above.
Assertions in trusted applications are embedded on the configuration directive CFG_TEE_CORE_DEBUG={y|n} when the OP-TEE OS package is built.

3.2. Debug with traces[edit | edit source]

OP-TEE trusted applications can embed trace messages using the same macros as the OP-TEE core (EMSG() and similar functions described above).
The build directive that sets the trace level for a trusted application is CFG_TEE_TA_LOG_LEVEL={0|1|2|3|4}.
The level values match those described for CFG_TEE_CORE_LOG_LEVEL in the section above.

3.3. Stack unwind support[edit | edit source]

When an OP-TEE trusted application panics, it may output backtrace messages on the OP-TEE console through the OP-TEE core.
These backtrace messages are generated by the OP-TEE OS when built with CFG_UNWIND=y. OP-TEE OS package provides a tool to analyse backtrace messages.

More information is available from the OP-TEE abort-dumps documentation[1].

4. References[edit | edit source]