Chapter 3 Executable and Linking Format

The type of object files created by the SPARC assembler version for SunOS 5.x are now Executable and Linking Format (ELF) files. These relocatable ELF files hold code and data suitable for linking with other object files to create an executable or a shared object file, and are the assembler normal output. The assembler can also write information to standard output (for example, under the -S option) and to standard error (for example, under the -V option). The SPARC assembler creates a default output file when standard input or multiple files are used.

This chapter is organized into the following sections:

• 3.1 ELF Header

• 3.2 Sections

• 3.3 Locations

• 3.5 Relocation Tables

• 3.6 Symbol Tables

• 3.4 Addresses

• 3.7 String Tables

• 3.8 Assembler Directives

The ELF object file format consists of:

• Header

• Sections

• Locations

• Addresses

• Relocation tables

• Symbol tables

• String tables

For more information, see the System V Application Binary Interface: SPARC^TM Processor Supplement.

3.1 ELF Header

The ELF header is always located at the beginning of the ELF file. It describes the ELF file organization and contains the actual sizes of the object file control structures. The initial bytes of an ELF header specify how the file is to be interpreted.

The ELF header contains the following information:

ehsize

ELF header size in bytes.

entry

Virtual address at which the process is to start. A value of 0 indicates no associated entry point.

flag

Processor-specific flags associated with the file.

ident

Marks the file as an object file and provides machine-independent data to decode and interpret the file contents.

machine

Specifies the required architecture for an individual file. A value of 2 specifies SPARC.

phentsize

Size in bytes of entries in the program header table. All entries are the same size.

phnum

Number of entries in program header table. A value of 0 indicates the file has no program header table.

phoff

Program header table file offset in bytes. The value of 0 indicates no program header.

shentsize

Size in bytes of the section header. A section header is one entry in the section header table; all entries are the same size.

shnum

Number of entries in section header table. A value of 0 indicates the file has no section header table.

shoff

Section header table file offset in bytes. The value of 0 indicates no section header.

shstrndx

Section header table index of the entry associated with the section name string table. A value of SHN_UNDEF indicates the file does not have a section name string table.

type

Identifies the object file type. Table 3–1 describes the reserved object file types.

version

Identifies the object file version.

Table 3–1 shows reserved object file types:

3.2 Sections

A section is the smallest unit of an object that can be relocated. The following sections are commonly present in an ELF file:

• Section header

• Executable text

• Read-only data

• Read-write data

• Read-write uninitialized data (section header only)

Sections do not need to be specified in any particular order. The current section is the section to which code is generated.

These sections contain all other information in an object file and satisfy several conditions.

1. Every section must have one section header describing the section. However, a section header does not need to be followed by a section.

2. Each section occupies one contiguous sequence of bytes within a file. The section may be empty (that is, of zero-length).

3. A byte in a file can reside in only one section. Sections in a file cannot overlap.

4. An object file may have inactive space. The contents of the data in the inactive space are unspecified.

Sections can be added for multiple text or data segments, shared data, user-defined sections, or information in the object file for debugging.

Not all of the sections need to be present.

3.2.1 Section Header

The section header allows you to locate all of the file sections. An entry in a section header table contains information characterizing the data in a section.

The section header contains the following information:

addr

Address at which the first byte resides if the section appears in the memory image of a process; the default value is 0.

addralign

Aligns the address if a section has an address alignment constraint; for example, if a section contains a double-word, the entire section must be ensured double-word alignment. Only 0 and positive integral powers of 2 are currently allowed. A value of 0 or 1 indicates no address alignment constraints.

entsize

Size in bytes for entries in fixed-size tables such as the symbol table.

flags

One-bit descriptions of section attributes. Table 3–2 describes the section attribute flags.

info

Extra information. The interpretation of this information depends on the section type, as described in Table 3–3.

link

Section header table index link. The interpretation of this information depends on the section type, as described in Table 3–3.

name

Specifies the section name. An index into the section header string table section specifies the location of a null-terminated string.

offset

Specifies the byte offset from the beginning of the file to the first byte in the section.

---

Note –

If the section type is SHT_NOBITS, offset specifies the conceptual placement of the file.

---

size

Specifies the size of the section in bytes.

---

Note –

If the section type is SHT_NOBITS, size may be non-zero; however, the section still occupies no space in the file.

---

type

Categorizes the section contents and semantics. Table 3–3 describes the section types.

Some section header table indexes are reserved and the object file will not contain sections for these special indexes.

3.2.2 Predefined User Sections

A section that can be manipulated by the section control directives is known as a user section. You can use the section control directives to change the user section in which code or data is generated. Table 3–4 lists the predefined user sections that can be named in the section control directives.

3.2.2.1 Creating an .init Section in an Object File

The .init sections contain codes that are to be executed before the the main program is executed. To create an .init section in an object file, use the assembler pseudo-ops shown in Example 3–1.

Example 3–1 Creating an .init Section

.section ".init"

.align   4

<instructions>

At link time, the .init sections in a sequence of .o files are concatenated into an .init section in the linker output file. The code in the .init section are executed before the main program is executed.

Because the whole .init section is treated as a single function body, it is recommented that the only code added to these sections be in the following form:.

call routine_name

nop

The called routine should be located in another section. This will prevent conflicting register and stack usage within the .init sections.

3.2.2.2 Creating a .fini Section in an Object File

.fini sections contain codes that are to be executed after the the main program is executed. To create an .fini section in an object file, use the assembler pseudo-ops shown in Example 3–2.

Example 3–2 Creating an .fini Section

.section ".fini"

.align   4

<instructions>

At link time, the .fini sections in a sequence of .o files are concatenated into a .fini section in the linker output file. The codes in the .fini section are executed after the main program is executed.

Because the whole .fini section is treated as a single function body, it is recommended that the only code added to these section be in the following form:.

call routine_name

nop

The called routine should be located in another section. This will prevent conflicting register and stack usage within the .fini sections.

3.2.3 Predefined Non-User Sections

Table 3–5 lists sections that are predefined but cannot be named in the section control directives because they are not under user control.

3.3 Locations

A location is a specific position within a section. Each location is identified by a section and a byte offset from the beginning of the section. The current location is the location within the current section where code is generated.

A location counter tracks the current offset within each section where code or data is being generated. When a section control directive (for example, the .section pseudo-op) is processed, the location information from the location counter associated with the new section is assigned to and stored with the name and value of the current location.

The current location is updated at the end of processing each statement, but can be updated during processing of data-generating assembler directives (for example, the .word pseudo-op).

Each section has one location counter; if more than one section is present, only one location can be current at any time.

3.4 Addresses

Locations represent addresses in memory if a section is allocatable; that is, its contents are to be placed in memory at program runtime. Symbolic references to these locations must be changed to addresses by the SPARC link editor.

3.5 Relocation Tables

The assembler produces a companion relocation table for each relocatable section. The table contains a list of relocations (that is, adjustments to data in the section) to be performed by the link editor.

3.6 Symbol Tables

A symbol table contains information to locate and relocate symbolic definitions and references. The SPARC assembler creates a symbol table section for the object file. It makes an entry in the symbol table for each symbol that is defined or referenced in the input file and is needed during linking. The symbol table is then used by the SPARC link editor during relocation. The section header contains the symbol table index for the first non-local symbol.

A symbol table contains the following information:

name

Index into the object file symbol string table. A value of zero indicates the symbol table entry has no name; otherwise, the value represents the string table index that gives the symbol name.

value

Value of the associated symbol. This value is dependent on the context; for example, it may be an address, or it may be an absolute value.

size

Size of symbol. A value of 0 indicates that the symbol has either no size or an unknown size.

info

Specifies the symbol type and binding attributes. Table 3–6 and Table 3–7 describe these values.

other

Undefined meaning. Current value is 0.

shndx

Contains the section header table index to another relevant section, if specified. As a section moves during relocation, references to the symbol will continue to point to the same location because the value of the symbol will change as well.

Table 3–7 shows the symbol binding attributes.

3.7 String Tables

A string table is a section which contains null-terminated variable-length character sequences, or strings, in the object file; for example, symbol names and file names. The strings are referenced in the section header as indexes into the string table section.

• A string table index may refer to any byte in the section.

• Empty string table sections are permitted; however, the index referencing this section must contain zero.

A string may appear multiple times and may also be referenced multiple times. References to substrings may exist, and unreferenced strings are allowed.

3.8 Assembler Directives

Assembler directives, or pseudo-operations (pseudo-ops), are commands to the assembler that may or may not result in the generation of code. The different types of assembler directives are:

• Section Control Directives

• Symbol Attribute Directives

• Assignment Directives

• Data Generating Directives

• Optimizer Directives

See Appendix A, Pseudo-Operations for a complete description of the pseudo-ops supported by the SPARC assembler.

3.8.1 Section Control Directives

When a section is created, a section header is generated and entered in the ELF object file section header table. The section control pseudo-ops allow you to make entries in this table. Sections that can be manipulated with the section control directives are known as user sections. You can also use the section control directives to change the user section in which code or data is generated.

The symbol table, relocation table, and string table sections are created implicitly. The section control pseudo-ops cannot be used to manipulate these sections.

The section control directives also create a section symbol which is associated with the location at the beginning of each created section. The section symbol has an offset value of zero.

3.8.2 Symbol Attribute Directives

The symbol attribute pseudo-ops declare the symbol type and size and whether it is local or global.

3.8.3 Assignment Directive

The assignment directive associates the value and type of expression with the symbol and creates a symbol table entry for the symbol. This directive constitutes a definition of the symbol and, therefore, must be the only definition of the symbol.

3.8.4 Data Generating Directives

The data generating directives are used for allocating storage and loading values.