LD(1) UNIX Programmer's Manual LD(1) NAME ld - link editor (2BSD) SYNOPSIS ld [ option ] ... file ... DESCRIPTION _L_d combines several object programs into one, resolves external references, and searches libraries. In the sim- plest case several object _f_i_l_e_s are given, and _l_d combines them, producing an object module which can be either exe- cuted or become the input for a further _l_d run. (In the latter case, the -r option must be given to preserve the relocation bits.) The output of _l_d is left on a.out. This file is made executable only if no errors occurred during the load. The argument routines are concatenated in the order speci- fied. The entry point of the output is the beginning of the first routine (unless the -e option is specified). If any argument is a library, it is searched exactly once at the point it is encountered in the argument list. Only those routines defining an unresolved external reference are loaded. If a routine from a library references another rou- tine in the library, and the library has not been processed by _r_a_n_l_i_b(1), the referenced routine must appear after the referencing routine in the library. Thus the order of pro- grams within libraries may be important. The first member of a library should be a file named `__.SYMDEF', which is understood to be a dictionary for the library as produced by _r_a_n_l_i_b(1); the dictionary is searched iteratively to satisfy as many references as possible. The symbols `_etext', `_edata' and `_end' (`etext', `edata' and `end' in C) are reserved, and if referred to, are set to the first location above the program, the first location above initialized data, and the first location above all data respectively. It is erroneous to define these symbols. _L_d understands several options. Except for -l, they should appear before the file names. -D Take the next argument as a decimal number and pad the data segment with zero bytes to the indicated length. -d Force definition of common storage even if the -r flag is present. -e The following argument is taken to be the name of the entry point of the loaded program; location 0 is the default. Printed 11/24/99 May 08, 1995 1 LD(1) UNIX Programmer's Manual LD(1) -L_d_i_r Add _d_i_r to the list of directories in which libraries are searched for. Directories specified with -L are searched before the standard directories. -l_x This option is an abbreviation for the library name `lib_x.a', where _x is a string. _L_d searches for libraries first in any directories specified with -L options, then in the standard directories `/lib', `/usr/lib', and `/usr/local/lib'. A library is searched when its name is encountered, so the placement of a -l is significant. -M produce a primitive load map, listing the names of the files which will be loaded. -n Arrange (by giving the output file a 0410 "magic number") that when the output file is executed, the text portion will be read-only and shared among all users executing the file. This involves moving the data areas up to the first possible 8K byte boundary following the end of the text. This option creates a `pure executable' format. -i When the output file is executed, the program text and data areas will live in separate address spaces. The only difference between this option and -n is that here the text and data segments are in separate address spaces and both start at location 0. This option creates a `separate executable' format. -z This option is a synonym for the -i option. On other systems (4.3BSD for example) the -z option causes a demand paged executable to be built. This option was added to 2.11BSD because some systems (those which use gcc) do not safely ignore (with a warning) the -i option. Adding the -z option to 2.11BSD allows makefiles to be copied freely between multiple plat- forms once again. -O This is a text replacement overlay file; only the text segment will be replaced by _e_x_e_c_v_e(2). Shared data must have the same size as the program overlaid, other- wise the _e_x_e_c_v_e(2) will fail. The entry point to the overlay may be defined with the -e option. This option allows the creation of a `replacement executable' for- mat. -o The _n_a_m_e argument after -o is used as the name of the _l_d output file, instead of a.out. -r Generate relocation bits in the output file so that it Printed 11/24/99 May 08, 1995 2 LD(1) UNIX Programmer's Manual LD(1) can be the subject of another _l_d run. This flag also prevents final definitions from being given to common symbols, and suppresses the `undefined symbol' diagnos- tics. (Note that this option cannot be used with over- lays (-Z) since they cannot be reloaded.) -s `Strip' the output, that is, remove the symbol table and relocation bits to save space (but impair the use- fulness of the debuggers). This information can also be removed by _s_t_r_i_p(1). -q ("quiet") Suppress the reporting of undefined symbols. Normally only used when building networked kernels - the large number of undefined symbols is normal (due to the three phase link proceedure) but can be distracting none the less. -t ("trace") Print the name of each file as it is pro- cessed. -u Take the following argument as a symbol and enter it as undefined in the symbol table. This is useful for loading wholly from a library, since initially the sym- bol table is empty and an unresolved reference is needed to force the loading of the first routine. -v ("verbose") Print the VM statistics. Printing out the number of pages swapped to and from the VM tmp file is now optional and only used when a problem is suspected (or if you are voyeuristic). -X Save local symbols except for those whose names begin with `L'. This option is used by _c_c(1) to discard internally-generated labels while retaining symbols local to routines. -x Do not preserve local (non-.globl) symbols in the out- put symbol table; only enter external symbols. This option saves some space in the output file. It also allows temporary labels to be discarded to prevent redefinition in sucessive ld's. Warning: _a_d_b uses these local symbols, especially when debugging overlaid programs, so some debugging information is necessarily lost if this option is used. -Z Indicate the creation of an automatic-overlay format. In addition a -i or -n must be present as overlays only work with shared text objects. Repeated instances of -Z bracket the modules that will be loaded into a given overlay. Modules before the first -Z or after the con- cluding -Y will be loaded into the non-overlaid text (base) area. Note that there may be a maximum of NOVL Printed 11/24/99 May 08, 1995 3 LD(1) UNIX Programmer's Manual LD(1) (currently 15) overlays. This option produces the `overlaid pure executable' and the `overlaid separate executable' formats. The loader creates a small entry interface in the base segment for each subroutine in an overlay. This interface ("thunk") arranges for the correct overlay to be present before the actual routine is entered. -Y Terminate text overlays. This allows any remaining modules or libraries to be loaded into the base area. Note that the -Y option used to be -_L, but had to be changed when the loader was brought up to date with the _4._3_B_S_D loader which uses -_L to indicate a directory to be searched for library references. AUTOMATIC TEXT OVERLAY EXAMPLE To set up an automatic text overlay object with the loader, use a command of the form: ld -n -X /lib/crt0.o base.o base2.o -Z ov1a.o ov1b.o ... -Z ov2a.o ov2b.o ... -Y base3.o ... -lovc Assembly source code must be compiled using the assembler overlay flags: "_a_s -_V prog.s" which causes the assembler to leave certain symbols unresolved so that _l_d may rearrange them. The various system compilers automatically use this option. When arranging modules into overlays, the following rules control the maximum sizes for an executable file. The magic numbers are due to the granularity of PDP-11 segmentation registers (there are 8 registers, each controlling an 8192- byte segment). The program is made up of four areas: base text, overlay text, data + bss, and stack sections. The size of the overlay section is controlled by the size of the largest of the overlays. Each section starts at an address that is a multiple of 8Kb, thus the size of each section is rounded up to a multiple of 8Kb. In the case of separate overlaid executable files, the text and overlays share one 64Kb byte address space; and the data + bss and stack share the other. Thus, the total of the base text size (rounded up to an 8Kb boundary) plus the max- imum overlay size (similarly rounded) must be less than or equal to 64Kb. Or, put another way, since there are only 8 segmentation registers available, the number of segmentation registers needed for an overlaid object must be less than or equal to 8. As an example, if the base text segment has 36800 bytes and the largest overlay takes 14144, the base Printed 11/24/99 May 08, 1995 4 LD(1) UNIX Programmer's Manual LD(1) will fit in 5 segments and the overlays in 2 segments; leav- ing one to spare. The data and bss together must fit in 7 segments (56K bytes), leaving one 8Kb segment for the stack. All of the limits can be checked by using _c_h_e_c_k_o_b_j(1). For pure overlaid programs, the rules are similar except that all four sections share one 64K-byte address space. The number of segments required by the text, overlay, data + bss and stack are calculated in the same way. The sum of the segments required, including one for the stack, must be less than or equal to 8. Example: a program has 8128 bytes of base text, the largest overlay is 16248 bytes, and the data and bss total 19500. The text requires 1 8Kb segment, the overlays 2, and data and bss use 4, leaving one for the stack. FILES /lib/lib*.a libraries /usr/lib/lib*.a more libraries /usr/local/lib/lib*.a still more libraries a.out output file SEE ALSO adb(1), ar(1), as(1), cc(1), checkobj(1), f77(1), file(1), ranlib(1), size(1), a.out(5) BUGS The text overlay scheme presented is unique to the PDP-11 and 2BSD. The -i, -P, -Z, -Y options are specific to 2BSD. The -q and -v options are new with 2.11BSD. Printed 11/24/99 May 08, 1995 5