Linker Script Generation

[中文]

Overview

There are several memory regions where code and data can be placed. Code and read-only data are placed by default in flash, writable data in RAM, etc. However, it is sometimes necessary to change these default placements.

For example, it may be necessary to place:

  • critical code in RAM for performance reasons.

  • executable code in IRAM so that it can be ran while cache is disabled.

  • code in RTC memory for use in a wake stub.

  • code in RTC memory for use by the ULP coprocessor.

With the linker script generation mechanism, it is possible to specify these placements at the component level within ESP-IDF. The component presents information on how it would like to place its symbols, objects or the entire archive. During build, the information presented by the components are collected, parsed and processed; and the placement rules generated is used to link the app.

Quick Start

This section presents a guide for quickly placing code/data to RAM and RTC memory - placements ESP-IDF provides out-of-the-box.

For this guide, suppose we have the following:

components
└── my_component
    ├── CMakeLists.txt
    ├── Kconfig
    ├── src/
    │   ├── my_src1.c
    │   ├── my_src2.c
    │   └── my_src3.c
    └── my_linker_fragment_file.lf

  • a component named my_component that is archived as library libmy_component.a during build

  • three source files archived under the library, my_src1.c, my_src2.c and my_src3.c which are compiled as my_src1.o, my_src2.o and my_src3.o, respectively

  • under my_src1.o, the function my_function1 is defined; under my_src2.o, the function my_function2 is defined

  • there is bool-type config PERFORMANCE_MODE (y/n) and int type config PERFORMANCE_LEVEL (with range 0-3) in my_component's Kconfig

Creating and Specifying a Linker Fragment File

Before anything else, a linker fragment file needs to be created. A linker fragment file is simply a text file with a .lf extension upon which the desired placements will be written. After creating the file, it is then necessary to present it to the build system. The instructions for the build systems supported by ESP-IDF are as follows:

In the component's CMakeLists.txt file, specify argument LDFRAGMENTS in the idf_component_register call. The value of LDFRAGMENTS can either be an absolute path or a relative path from the component directory to the created linker fragment file.

# file paths relative to CMakeLists.txt
idf_component_register(...
                       LDFRAGMENTS "path/to/linker_fragment_file.lf" "path/to/another_linker_fragment_file.lf"
                       ...
                       )

Specifying Placements

It is possible to specify placements at the following levels of granularity:

  • object file (.obj or .o files)

  • symbol (function/variable)

  • archive (.a files)

Placing Object Files

Suppose the entirety of my_src1.o is performance-critical, so it is desirable to place it in RAM. On the other hand, the entirety of my_src2.o contains symbols needed coming out of deep sleep, so it needs to be put under RTC memory.

In the linker fragment file, we can write:

[mapping:my_component]
archive: libmy_component.a
entries:
    my_src1 (noflash)     # places all my_src1 code/read-only data under IRAM/DRAM
    my_src2 (rtc)         # places all my_src2 code/ data and read-only data under RTC fast memory/RTC slow memory

What happens to my_src3.o? Since it is not specified, default placements are used for my_src3.o. More on default placements here.

Placing Symbols

Continuing our example, suppose that among functions defined under object1.o, only my_function1 is performance-critical; and under object2.o, only my_function2 needs to execute after the chip comes out of deep sleep. This could be accomplished by writing:

[mapping:my_component]
archive: libmy_component.a
entries:
    my_src1:my_function1 (noflash)
    my_src2:my_function2 (rtc)

The default placements are used for the rest of the functions in my_src1.o and my_src2.o and the entire object3.o. Something similar can be achieved for placing data by writing the variable name instead of the function name, like so:

my_src1:my_variable (noflash)

Warning

There are limitations in placing code/data at symbol granularity. In order to ensure proper placements, an alternative would be to group relevant code and data into source files, and use object-granularity placements.

Placing Entire Archive

In this example, suppose that the entire component archive needs to be placed in RAM. This can be written as:

[mapping:my_component]
archive: libmy_component.a
entries:
    * (noflash)

Similarly, this places the entire component in RTC memory:

[mapping:my_component]
archive: libmy_component.a
entries:
    * (rtc)

Configuration-Dependent Placements

Suppose that the entire component library should only have special placement when a certain condition is true; for example, when CONFIG_PERFORMANCE_MODE == y. This could be written as:

[mapping:my_component]
archive: libmy_component.a
entries:
    if PERFORMANCE_MODE = y:
        * (noflash)
    else:
        * (default)

For a more complex config-dependent placement, suppose the following requirements: when CONFIG_PERFORMANCE_LEVEL == 1, only object1.o is put in RAM; when CONFIG_PERFORMANCE_LEVEL == 2, object1.o and object2.o; and when CONFIG_PERFORMANCE_LEVEL == 3 all object files under the archive are to be put into RAM. When these three are false however, put entire library in RTC memory. This scenario is a bit contrived, but, it can be written as:

[mapping:my_component]
archive: libmy_component.a
entries:
    if PERFORMANCE_LEVEL = 1:
        my_src1 (noflash)
    elif PERFORMANCE_LEVEL = 2:
        my_src1 (noflash)
        my_src2 (noflash)
    elif PERFORMANCE_LEVEL = 3:
        my_src1 (noflash)
        my_src2 (noflash)
        my_src3 (noflash)
    else:
        * (rtc)

Nesting condition-checking is also possible. The following is equivalent to the snippet above:

[mapping:my_component]
archive: libmy_component.a
entries:
    if PERFORMANCE_LEVEL <= 3 && PERFORMANCE_LEVEL > 0:
        if PERFORMANCE_LEVEL >= 1:
            object1 (noflash)
            if PERFORMANCE_LEVEL >= 2:
                object2 (noflash)
                if PERFORMANCE_LEVEL >= 3:
                    object2 (noflash)
    else:
        * (rtc)

The 'default' Placements

Up until this point, the term 'default placements' has been mentioned as fallback placements when the placement rules rtc and noflash are not specified. It is important to note that the tokens noflash or rtc are not merely keywords, but are actually entities called fragments, specifically schemes.

In the same manner as rtc and noflash are schemes, there exists a default scheme which defines what the default placement rules should be. As the name suggests, it is where code and data are usually placed, i.e., code/constants is placed in flash, variables placed in RAM, etc. More on the default scheme here.

Note

For an example of an ESP-IDF component using the linker script generation mechanism, see freertos/CMakeLists.txt. freertos uses this to place its object files to the instruction RAM for performance reasons.

This marks the end of the quick start guide. The following text discusses the internals of the mechanism in a little bit more detail. The following sections should be helpful in creating custom placements or modifying default behavior.

Linker Script Generation Internals

Linking is the last step in the process of turning C/C++ source files into an executable. It is performed by the toolchain's linker, and accepts linker scripts which specify code/data placements, among other things. With the linker script generation mechanism, this process is no different, except that the linker script passed to the linker is dynamically generated from: (1) the collected linker fragment files and (2) linker script template.

Note

The tool that implements the linker script generation mechanism lives under tools/ldgen.

Linker Fragment Files

As mentioned in the quick start guide, fragment files are simple text files with the .lf extension containing the desired placements. This is a simplified description of what fragment files contain, however. What fragment files actually contain are 'fragments'. Fragments are entities which contain pieces of information which, when put together, form placement rules that tell where to place sections of object files in the output binary. There are three types of fragments: sections, scheme and mapping.

Grammar

The three fragment types share a common grammar:

[type:name]
key: value
key:
    value
    value
    value
    ...

  • type: Corresponds to the fragment type, can either be sections, scheme or mapping.

  • name: The name of the fragment, should be unique for the specified fragment type.

  • key, value: Contents of the fragment; each fragment type may support different keys and different grammars for the key values.

    • For sections and scheme, the only supported key is entries

    • For mappings, both archive and entries are supported.

Note

In cases where multiple fragments of the same type and name are encountered, an exception is thrown.

Note

The only valid characters for fragment names and keys are alphanumeric characters and underscore.

Condition Checking

Condition checking enable the linker script generation to be configuration-aware. Depending on whether expressions involving configuration values are true or not, a particular set of values for a key can be used. The evaluation uses eval_string from kconfiglib package and adheres to its required syntax and limitations. Supported operators are as follows:

  • comparison

    • LessThan <

    • LessThanOrEqualTo <=

    • MoreThan >

    • MoreThanOrEqualTo >=

    • Equal =

    • NotEqual !=

  • logical

    • Or ||

    • And &&

    • Negation !

  • grouping

    • Parenthesis ()

Condition checking behaves as you would expect an if...elseif/elif...else block in other languages. Condition-checking is possible for both key values and entire fragments. The two sample fragments below are equivalent:

# Value for keys is dependent on config
[type:name]
key_1:
    if CONDITION = y:
        value_1
    else:
        value_2
key_2:
    if CONDITION = y:
        value_a
    else:
        value_b

# Entire fragment definition is dependent on config
if CONDITION = y:
    [type:name]
    key_1:
        value_1
    key_2:
        value_a
else:
    [type:name]
    key_1:
        value_2
    key_2:
        value_b

Comments

Comment in linker fragment files begin with #. Like in other languages, comment are used to provide helpful descriptions and documentation and are ignored during processing.

Types

Sections

Sections fragments defines a list of object file sections that the GCC compiler emits. It may be a default section (e.g., .text, .data) or it may be user defined section through the __attribute__ keyword.

The use of an optional '+' indicates the inclusion of the section in the list, as well as sections that start with it. This is the preferred method over listing both explicitly.

[sections:name]
entries:
    .section+
    .section
    ...

Example:

# Non-preferred
[sections:text]
entries:
    .text
    .text.*
    .literal
    .literal.*

# Preferred, equivalent to the one above
[sections:text]
entries:
    .text+              # means .text and .text.*
    .literal+           # means .literal and .literal.*

Scheme

Scheme fragments define what target a sections fragment is assigned to.

[scheme:name]
entries:
    sections -> target
    sections -> target
    ...

Example:

[scheme:noflash]
entries:
    text -> iram0_text          # the entries under the sections fragment named text will go to iram0_text
    rodata -> dram0_data        # the entries under the sections fragment named rodata will go to dram0_data

The default scheme

There exists a special scheme with the name default. This scheme is special because catch-all placement rules are generated from its entries. This means that, if one of its entries is text -> flash_text, the placement rule will be generated for the target flash_text.

*(.literal .literal.* .text .text.*)

These catch-all rules then effectively serve as fallback rules for those whose mappings were not specified.

The default scheme is defined in esp_system/app.lf. The noflash and rtc scheme fragments which are built-in schemes referenced in the quick start guide are also defined in this file.

Mapping

Mapping fragments define what scheme fragment to use for mappable entities, i.e., object files, function names, variable names, archives.

[mapping:name]
archive: archive                # output archive file name, as built (i.e., libxxx.a)
entries:
    object:symbol (scheme)      # symbol granularity
    object (scheme)             # object granularity
    * (scheme)                  # archive granularity

There are three levels of placement granularity:

  • symbol: The object file name and symbol name are specified. The symbol name can be a function name or a variable name.

  • object: Only the object file name is specified.

  • archive: * is specified, which is a short-hand for all the object files under the archive.

To know what an entry means, let us expand a sample object-granularity placement:

object (scheme)

Then expanding the scheme fragment from its entries definitions, we have:

object (sections -> target,
        sections -> target,
        ...)

Expanding the sections fragment with its entries definition:

object (.section,      # given this object file
        .section,      # put its sections listed here at this
        ... -> target, # target

        .section,
        .section,      # same should be done for these sections
        ... -> target,

        ...)           # and so on

Example:

[mapping:map]
archive: libfreertos.a
entries:
    * (noflash)

Aside from the entity and scheme, flags can also be specified in an entry. The following flags are supported (note: <> = argument name, [] = optional):

  1. ALIGN(<alignment>[, pre, post])

    Align the placement by the amount specified in alignment. Generates

  1. SORT([<sort_by_first>, <sort_by_second>])

    Emits SORT_BY_NAME, SORT_BY_ALIGNMENT, SORT_BY_INIT_PRIORITY or SORT in the input section description.

    Possible values for sort_by_first and sort_by_second are: name, alignment, init_priority.

    If both sort_by_first and sort_by_second are not specified, the input sections are sorted by name. If both are specified, then the nested sorting follows the same rules discussed in https://sourceware.org/binutils/docs/ld/Input-Section-Wildcards.html.

  2. KEEP()

    Prevent the linker from discarding the placement by surrounding the input section description with KEEP command. See https://sourceware.org/binutils/docs/ld/Input-Section-Keep.html for more details.

4.SURROUND(<name>)

Generate symbols before and after the placement. The generated symbols follow the naming _<name>_start and _<name>_end. For example, if name == sym1,

When adding flags, the specific section -> target in the scheme needs to be specified. For multiple section -> target, use a comma as a separator. For example,

# Notes:
# A. semicolon after entity-scheme
# B. comma before section2 -> target2
# C. section1 -> target1 and section2 -> target2 should be defined in entries of scheme1
entity1 (scheme1);
    section1 -> target1 KEEP() ALIGN(4, pre, post),
    section2 -> target2 SURROUND(sym) ALIGN(4, post) SORT()

Putting it all together, the following mapping fragment, for example,

[mapping:name]
archive: lib1.a
entries:
    obj1 (noflash);
        rodata -> dram0_data KEEP() SORT() ALIGN(8) SURROUND(my_sym)

generates an output on the linker script:

. = ALIGN(8)
_my_sym_start = ABSOLUTE(.)
KEEP(lib1.a:obj1.*( SORT(.rodata) SORT(.rodata.*) ))
_my_sym_end = ABSOLUTE(.)

Note that ALIGN and SURROUND, as mentioned in the flag descriptions, are order sensitive. Therefore, if for the same mapping fragment these two are switched, the following is generated instead:

_my_sym_start = ABSOLUTE(.)
. = ALIGN(8)
KEEP(lib1.a:obj1.*( SORT(.rodata) SORT(.rodata.*) ))
_my_sym_end = ABSOLUTE(.)

On Symbol-Granularity Placements

Symbol granularity placements is possible due to compiler flags -ffunction-sections and -ffdata-sections. ESP-IDF compiles with these flags by default. If the user opts to remove these flags, then the symbol-granularity placements will not work. Furthermore, even with the presence of these flags, there are still other limitations to keep in mind due to the dependence on the compiler's emitted output sections.

For example, with -ffunction-sections, separate sections are emitted for each function; with section names predictably constructed i.e., .text.{func_name} and .literal.{func_name}. This is not the case for string literals within the function, as they go to pooled or generated section names.

With -fdata-sections, for global scope data the compiler predictably emits either .data.{var_name}, .rodata.{var_name} or .bss.{var_name}; and so Type I mapping entry works for these. However, this is not the case for static data declared in function scope, as the generated section name is a result of mangling the variable name with some other information.

Linker Script Template

The linker script template is the skeleton in which the generated placement rules are put into. It is an otherwise ordinary linker script, with a specific marker syntax that indicates where the generated placement rules are placed.

To reference the placement rules collected under a target token, the following syntax is used:

mapping[target]

Example:

The example below is an excerpt from a possible linker script template. It defines an output section .iram0.text, and inside is a marker referencing the target iram0_text.

.iram0.text :
{
    /* Code marked as running out of IRAM */
    _iram_text_start = ABSOLUTE(.);

    /* Marker referencing iram0_text */
    mapping[iram0_text]

    _iram_text_end = ABSOLUTE(.);
} > iram0_0_seg

Suppose the generator collected the fragment definitions below:

[sections:text]
    .text+
    .literal+

[sections:iram]
    .iram1+

[scheme:default]
entries:
    text -> flash_text
    iram -> iram0_text

[scheme:noflash]
entries:
    text -> iram0_text

[mapping:freertos]
archive: libfreertos.a
entries:
    * (noflash)

Then the corresponding excerpt from the generated linker script will be as follows:

.iram0.text :
{
    /* Code marked as running out of IRAM */
    _iram_text_start = ABSOLUTE(.);

    /* Placement rules generated from the processed fragments, placed where the marker was in the template */
    *(.iram1 .iram1.*)
    *libfreertos.a:(.literal .text .literal.* .text.*)

    _iram_text_end = ABSOLUTE(.);
} > iram0_0_seg

*libfreertos.a:(.literal .text .literal.* .text.*)

Rule generated from the entry * (noflash) of the freertos mapping fragment. All text sections of all object files under the archive libfreertos.a will be collected under the target iram0_text (as per the noflash scheme) and placed wherever in the template iram0_text is referenced by a marker.

*(.iram1 .iram1.*)

Rule generated from the default scheme entry iram -> iram0_text. Since the default scheme specifies an iram -> iram0_text entry, it too is placed wherever iram0_text is referenced by a marker. Since it is a rule generated from the default scheme, it comes first among all other rules collected under the same target name.

The linker script template currently used is esp_system/ld/esp32/sections.ld.in; the generated output script sections.ld is put under its build directory.

Provide feedback about this document