scnhdr(4) — File Formats
NAME
scnhdr − Section header for an object file
SYNOPSIS
#include <scnhdr.h>
DESCRIPTION
Every object file has a table of section headers to specify the layout of the data within the file. Each section within an object file has its own header. The C structure appears below:
struct scnhdr
{
char s_name[8]; /∗ section name ∗/
long s_paddr; /∗ physical address, aliased s_nlib ∗/
long s_vaddr; /∗ virtual address ∗/
long s_size; /∗ section size ∗/
long s_scnptr; /∗ file ptr to raw data for section ∗/
long s_relptr; /∗ file ptr to relocation ∗/
long s_lnnoptr; /∗ file ptr to gp table ∗/
unsigned short s_nreloc; /∗ number of relocation entries ∗/
unsigned short s_nlnno; /∗ number of gp table entries ∗/
int s_flags; /∗ flags ∗/
};
File pointers are byte offsets into the file; they can be used as the offset in a call to FSEEK [see ldfcn(4)]. If a section is initialized, the file contains the actual bytes. An uninitialized section is somewhat different. It has a size, symbols defined in it, and symbols that refer to it. But it can have no relocation entries or data. Consequently, an uninitialized section has no raw data in the object file, and the values for s_scnptr, s_relptr, and s_nreloc are zero.
The entries that refer to line numbers (s_lnnoptr, and s_nlnno) are not used. See the header file sym.h for the entries to get to the line number table. The entries that were for line numbers in the section header are used for gp tables.
The number of relocation entries for a section is found in the s_nreloc field of the section header. This field being a ‘C’ language short and can overflow with large objects. If this field overflows the section header s_flags field has the S_NRELOC_OVFL bit set. In this case the true number of relocation entries is found in the r_vaddr field of the first relocation entry for that section. That relocation entry has a type of R_ABS so it is ignored when the relocation takes place. This is a stopgap solution.
The gp table gives the section size corresponding to each applicable value of the compiler option -G num (always including 0), sorted by smallest size first. It is pointed to by the s_lnnoptr field in the section header and its number of entries (including the header) is in the s_nlnno field in the section header. This table only needs to exist for the .sdata and .sbss sections. If there is no “small” section then the gp table for it is attached to the corresponding “large” section so the information still gets to the link editor, ld(1). The C union for the gp table appears below.
union gp_table
{
struct {
int current_g_value; /∗ actual value ∗/
int unused;
} header;
struct {
int g_value; /∗ hypothetical value ∗/
int bytes; /∗ section size corresponding to hypothetical value ∗/
} entry;
};
Each gp table has one header structure that contains the actual value of the −G num option used to produce the object file. An entry must exist for every applicable value of the −G num option. The applicable values are all the sizes of the data items in that section.
For .lib sections the number of shared libraries is in the s_nlib field (an alias to s_paddr). The .lib section is made up of s_nlib descriptions of shared libraries. Each description of a shared library is a libscn structure followed by the path name to the shared library. The C structure appears below and is defined in scnhdr.h .
struct libscn
{
int size; /∗ size of this entry (including target name) ∗/
int offset; /∗ offset from start of entry to target name ∗/
long tsize; /∗ text size in bytes, padded to DW boundary ∗/
long dsize; /∗ initialized data size ∗/
long bsize; /∗ uninitialized data ∗/
long text_start; /∗ base of text used for this library ∗/
long data_start; /∗ base of data used for this library ∗/
long bss_start; /∗ base of bss used for this library ∗/
/∗ pathname of target shared library ∗/
};