Rework ifunc support to address a number of short comings:
- Move to a shared _rtld_call_ifunc for rel and rela architectures
- Architectures using rel format must patch IRELATIVE non-PLT
relocations like RELATIVE in additition to the later ifunc handling
- Consistently record the delta to the end of the relocation group for
non-PLT IRELATIVE relocations

Hidden ifunc is now supported on all ifunc platforms, even when using
-fno-plt. The combination of -fno-plt and relro is broken due to
incorrect GNU ld output though.

branches: 1.8.82;
Explicitly prefix ld(1) options in LDFLAGS with "-Wl," rather than
using the ${LDFLAGS:@F@-Wl,$F@} make substitution because that breaks
if LDFLAGS is already set.

Move common LDFLAGS into Makefile from arch/*/Makefile.inc

Whitespace.

Remove vestiges of VARPSZ.

Do self-relocation similar to sparc.

branches: 1.3.2;
Grumble. Put compiler options back into CPPFLAGS for now, since
they're needed to build .S files with the correct PIC options.

XXX Really needs cleanup in <bsd.*.mk> to do this properly.

Don't put compiler options into CPPFLAGS.

branches: 1.1.2;
Added support for the SPARC v9 ABI.

ld.elf_so(1): Nix trailing whitespace.

No functional change intended.

ld.elf_so: Cite reference for SPARC ELF relocations

PR lib/58455: Missing references for processor-specific ELF
relocation semantics

branches: 1.70.2;
Fix interactions of initial-exec TLS model and dlopen

(1) If an initial-exec relocation was used for a non-local symbol
(i.e. the definition of the symbol is in a different DSO), the
computation of the static TLS offset used the wrong DSO.
This would effectively mean the wrong address was computed
(PR toolchain/50277, PR pkg/57445).

Fix this by forcing the computation of the correct DSO (the one defining
the symbol).

This code uses __UNCONST to avoid the vast interface changes for this
special case.

(2) If symbols from a DSO loaded via dlopen are used with both
global-dynamic/local-dynamic and initial-exec relocations AND
a initial-exec relocation was resolved first in a thread, a split brain
situation could exist where the dynamic relocations would use one memory
block (separate allocation) and the initial-exec relocations the static
per-thread TLS space.

(3) If the initial-exec relocation in (2) is seen after any thread has
already used a GD/LD allocation, bail out. Since IE relocations are used
only in the GOT, this will prevent the dlopen. This is a bit more
aggressive than necessary, but a full blown reference counting doesn't
seem to be justified.

branches: 1.69.4; 1.69.12;
Rework ifunc support to address a number of short comings:
- Move to a shared _rtld_call_ifunc for rel and rela architectures
- Architectures using rel format must patch IRELATIVE non-PLT
relocations like RELATIVE in additition to the later ifunc handling
- Consistently record the delta to the end of the relocation group for
non-PLT IRELATIVE relocations

Hidden ifunc is now supported on all ifunc platforms, even when using
-fno-plt. The combination of -fno-plt and relro is broken due to
incorrect GNU ld output though.

Move the complex logic for dynamically writing branches from ld.elf_so
into a header for reuse in crt0.o for static ifunc support. Change the
existing logic for sparc64 to use the Bicc variant of ba,a as it allows
+-8MB displacement compared to the BPcc variant's +-1MB. Teach the sparc
variant the same trick for using ba,a and not sethi+jmp when possible.

branches: 1.67.2;
Since GNU ld doesn't honor the PLT entry reservation, also compensate
for JMP_IREL relocations at the start of the PLT.

init/fini for the main program is handled by crt0.o, so ifunc handling
is skipped right now as it iterates the same list. Don't repeat that
mistake and explicitly take care of it in the dynamic linker.

Add sparc/sparc64 support for irelative relocations.

Add IRELATIVE support for ARM, X86 and PowerPC.

Fix thinko in previous: even if the address is an unsigned value and we
have verified the range before, we still need to mask the bit pattern
to the target instruction field.

Simplify and fix the offset calculation when doing a %pc relative
branch from the PLT slot to the target.

Replace COMBREL with just-in-time check in _rtld_relocate_nonplt_objects.

The COMBREL logic predates thread-safety of the dynamic linker and
breaks the use of shared locks for the common symbol lookup case. It is
unlikely to provide any benefit for lazy binding or PLT lookups, so
provide equivalent functionality in the non-PLT relocation handling loop
by checking if the symbol used by the current relocation is the same as
the one used during the last lookup. No inter-object cachine is done as
it is also unlikely to be benefical.

Testing with Firefox startup on AMD64 shows a small performance gain by
the new method.

Fix indentation.

branches: 1.59.6;
Despite the 10 in the name, the OLO10 relocation targets a 13 bit field,
the offset (O) providing the additional data.

Fix a few bounds and instruction sequences generated in the PLT; exercised
by ASLR and verified to work with the aslr fixed random debug sysctls.

Add basic support for indirect functions. It allows providing a public
function symbol with an implementation choosen at run time.
Refactor calls to functions by address in ld.elf_so to create temporary
function descriptors on the stack, if the address is not leaked outside.

Limitations:
- no support for initialising static storage with function pointers
- no support for unnamed resolver functions

Inspired by FreeBSD's r228435 by kib@freebsd.org.

branches: 1.56.4;
The table of masks applied for each relocation type was wrong for some
relocs with 64bit targets, like GLOB_DAT.
Change some instruction patterns used for PLT slots, so -current
objdump -d shows what the comment says.
Fix encoding of %lox() in PLT slots "close" to -1.
Enhance debug output slightly.

The R_SPARC_L44 relocation type is documented to be truncating and
calculated as "(S + A) & 0xfff" - so while it (usually) sets a "imm13" field
in the opcode, only 12 bits of the calculated value can be used.

Missing , in table initializer

branches: 1.53.2;
Add special handling needed for OLO10 relocations.
First part of fixing PR port-sparc64/46724.

branches: 1.52.4; 1.52.6;
Add support for 64 bit TLS relocations.

Add basic locking to ld.elf_so.

Comment reformat.

Wrap long line.

Whitespace.

Reduce header pollution for mdreloc.c. Make Obj_Entry argument of
_rtld_relocate_nonplt_objects non-const in preparation for TLS support.

PR/39240: Satoshi Suetake: Don't fail when attempting to resolve weak symbols
when we are doing immediate binding, leave them alone and they will be dealt
with later during lazy binding. From skrll@

Use the new Elf_SOff (signed offset) type to fix a few signed/unsigned
problems.

Fix const issues (cast const pointers to "const uint8_t *" instead of "caddr_t")

NOTE: change based on ../i386/mdreloc.c 1.27
XXX: not compile tested

branches: 1.43.4; 1.43.6;
Refactor common code to _rtld_relocate_plt_object to i386 and arm so they
act like the other versions.
In _rtld_bind, if the result is 0, call _rtld_die.
Initialize _rtld_sym_zero.st_value to -_rtld_objself.maprelocbase. Now when
the symbol is resolved, st_value + maprelocbase will equal 0 and the above
check in _rtld_bind will fire and a call to NULL will be avoided.

branches: 1.42.2;
Remove clause 3 and 4 from TNF licenses

branches: 1.41.10;
Remove MD COMBRELOC support since it's now MI.
Enable COMBRELOC for all arches

Add COMBRELOC support

quell GCC 4.1 uninitialised variable warnings.

XXX: we should audit the tree for which old ones are no longer needed
after getting the older compilers out of the tree..

Remove leading __ from __(const|inline|signed|volatile) -- it is obsolete.

Add __RCSID.

KNF

Add LD_BIND_NOW support for sparc64.

ANSIfy and de-__P().

Fix alignment problem in debug output.
Happens only under RTLD_DEBUG_RELOC.

_rtld_bind(): Don't use "rela" until after we've (potentially) adjusted
it.

Remove the `self' args to _rtld_relocate_objects() and
_rtld_relocate_nonplt_objects().

Remove vestigal comments about _rtld_bind_start_0() -- it works now.

Make room for the argument save area in _rtld_start(). The PLT stub already
made room from _rtld_bind_start(), but add a comment explaining the instruction
it uses.

Add a comment explaining some of the magick.

Fix a bug that caused PLT entries to effectively not be patched.
Also, fix more bugs in the large-PLT case -- my test program works now.

Swap registers %o0 and %o1 in the binder stub to say one instruction.

Fix an obvious bug in the 64-bit PLT fixup: the SLLX was by 12 bits, when it
should be 32.

Fix what *appear* to be two bugs in the >32768 PLT entry stub:
* One division was wrong (/14 rather than /24).
* We need to subtract 1048576 (to make the offset relative to the beginning of
the upper section), not add it.
This path is still untested, and buggy.

_rtld_bind_start_0_stub() is not actually used.

Push a multiple from _rtld_bind_start_[01]() to _rtld_bind().

Push the _rtld_bind() interface into MD code -- it's just a trivial wrapper
anyway.

%ld -> %p

Nuke -DRTLD_RELOCATE_SELF and `dodebug' from orbit.

Prototype different functions.

_rtld_bind_start() is not used in MI code, so declare it in the MD code.

Do self-relocation similar to sparc.

Expand the comment about the R_TYPE(RELATIVE) special-case, as it's not quite
obvious from the code. Also, just do it all the time, as an optimization.

Make the reloc tables const.

Pass down an additional flag to _rtld_relocate_nonplt_objects() which
indicates whether we're relocating ld.elf_so itself. Use this in some places
rather than hackish tests on `dodebug'. (The Alpha and HPPA `dodebug' tests
were actually noops, because RTLD_RELOCATE_SELF is not set, and therefore
dodebug is always true.)

Re-poison a lot of consts, now that the mark shite is gone.

Split _rtld_relocate_plt_object() into two MD functions:
* _rtld_relocate_plt_lazy() fixes up all the relocs pointing to the PLT. (On
most platforms it just does a simple base-relative fixup; on SPARC it does
nothing.)
* _rtld_relocate_plt_object() does immediate binding for a PLT entry.
The basic gist is that this saves a bit of time on SPARC (where the iteration
through the pltrela table was gratuitous), and a little less time on all other
platforms. A whole lot of #ifdef'ed crap is moved out of reloc.c, too.

NOT tested on: hppa sh x86_64

Repair a slight biff I made earlier -- do not exit the loop in
_rtld_relocate_nonplt_objects() early.

Partially fix up some debug printf()s that don't need to use defobj.

A few things:
* Pass a symbol number to _rtld_find_symdef(), not a r_info.
* Don't try to do a symbol lookup when we find an unsupported relocation;
instead get the symbol name from the referencing object's strtab.
* Add preliminary support for `-z combreloc'-style startup optimization on
i386, `#ifdef COMBRELOC'.

Rename _rtld_relocate_nonplt_object() to _rtld_relocate_nonplt_objects(),
and push the outer loop into it. This actually shaves a couple % off startup
time at least on PCs.

Split _rtld_relocate_nonplt_object() into separate MD files.

Remove unused arguments of _rtld_find_symdef().

Make _rtld_setup_pltgot() a consistent interface on all platforms.

branches: 1.5.2;
Sync with <sys/exec_elf.h> rev. 1.45: ElfNN_RelA -> ElfNN_Rela.

Wrong register in machine code description.

branches: 1.3.2;
Catch up with type changes. Also:
* Apply DT_PLTRELSZ to (one of) pltrel or pltrela *after* we've finished
parsing the headers, so we know which one.
* Fix sparc64 bogons. (It works now!)

Fix some issues with SPARC V9 ELF binaries.

Added support for the SPARC v9 ABI.

ld.elf_so(1): Nix trailing whitespace.

No functional change intended.

branches: 1.17.74;
Remove clause 3 and 4 from TNF licenses

branches: 1.16.30;
Update copyright.

Adjust for the fact that the GOT is all 0s in new binutils -- do the trick of
disassembling a call to _DYNAMIC to determine its real address, and using the
first entry of the GOT as its base-relative address.

It's evil, but it works.

Remove vestigal comments about _rtld_bind_start_0() -- it works now.

Make room for the argument save area in _rtld_start(). The PLT stub already
made room from _rtld_bind_start(), but add a comment explaining the instruction
it uses.

Fix a bug that caused PLT entries to effectively not be patched.
Also, fix more bugs in the large-PLT case -- my test program works now.

Swap registers %o0 and %o1 in the binder stub to say one instruction.

Fix an obvious bug in the 64-bit PLT fixup: the SLLX was by 12 bits, when it
should be 32.

Fix what *appear* to be two bugs in the >32768 PLT entry stub:
* One division was wrong (/14 rather than /24).
* We need to subtract 1048576 (to make the offset relative to the beginning of
the upper section), not add it.
This path is still untested, and buggy.

Push a multiple from _rtld_bind_start_[01]() to _rtld_bind().

Update comments.

Pass the calculated relocation offset down to _rtld().

Do self-relocation similar to sparc.

branches: 1.6.2;
Add .register directives to appease newer gas.

Correct some comments.

Get rid of some debug code that should no longer be needed.

Sync with <sys/exec_elf.h> rev. 1.45: ElfNN_RelA -> ElfNN_Rela.

branches: 1.2.2;
Fix some issues with SPARC V9 ELF binaries.

Added support for the SPARC v9 ABI.