Get rid of static Atffiles and let bsd.test.mk generate them on the fly.

add regression tests for dlinfo()

Fix static linking for the tests: -lrump is also used by -lrumpuser,
so we also need -lrump after -lrumpuser. Fixes build for sun2.

branches: 1.3.8; 1.3.18;
Turn warns on for all tests and fix all the bugs.

branches: 1.2.4;
Get rid of static Atffiles and let bsd.test.mk generate them on the fly.

add regression tests for dlinfo()

add Makefile.inc everywhere so that we can set WARNS=4 by default. Amazing
how many bugs this found :-)

Get rid of static Atffiles and let bsd.test.mk generate them on the fly.

add regression tests for dlinfo()

tests/libexec/ld.elf_so/Makefile: Undo recent -Wl,-Map change.

This was almost certainly a mistake in a commit that was otherwise
entirely about omap4/pandaboard support:

https://mail-index.NetBSD.org/source-changes/2025/12/16/msg159556.html

The change added LDADD linker arguments of the form

-Wl,-Map foo.map

except that doesn't actually pass `-Map foo.map' through to the
underlying linker -- it only passes `-Map' to the linker, and then
treats `foo.map' as an extra input file for another object to link
(which doesn't work because the file doesn't exist). If we want to
keep this change (not sure why), then we would have to use

-Wl,-Map,foo.map

to actually pass `-Map foo.map' through to the underlying linker.

Should fix clang build failures:

aarch64--netbsd-clang: error: no such file or directory: 't_ifunc.map'
--- t_ifunc.link ---

*** Failed target: t_ifunc.link
*** In directory: /home/source/ab/HEAD-llvm/src/tests/libexec/ld.elf_so
*** Failed commands:
${_MKTARGET_LINK}
=> @# " link " ld.elf_so/t_ifunc.link
${_CCLINK.${:Ut_ifunc}} ${_LDFLAGS.${:Ut_ifunc}} ${_LDSTATIC.${:Ut_ifunc}} -o ${.TARGET} ${OBJS.${:Ut_ifunc}} ${_PROGLDOPTS} ${_LDADD.${:Ut_ifunc}}
=> /home/builds/ab/HEAD-llvm/evbarm-aarch64/20251217070550Z-tools/bin/aarch64--netbsd-clang --sysroot=/home/builds/ab/HEAD-llvm/evbarm-aarch64/20251217070550Z-dest -Wl,--fatal-warnings -Wl,--warn-shared-textrel -Wl,-z,relro -pie -o t_ifunc.link t_ifunc.o -Wl,-rpath-link,/home/builds/ab/HEAD-llvm/evbarm-aarch64/20251217070550Z-dest/lib -L=/lib -Wl,-z,relro -Wl,-Map t_ifunc.map -Wl,-rpath,/usr/tests/libexec/ld.elf_so -lutil -latf-c
${CTFMERGE} ${CTFMFLAGS} -o ${.TARGET} ${OBJS.${:Ut_ifunc}}
=> /home/builds/ab/HEAD-llvm/evbarm-aarch64/20251217070550Z-tools/bin/nbctfmerge -t -g -L VERSION -g -o t_ifunc.link t_ifunc.o
*** [t_ifunc.link] Error code 1

tests/lib/csu, tests/libexec/ld.elf_so: Apply CTFMERGE=: workaround.

The workaround for

PR toolchain/59364: ctf tools needs update

required an update after fixing

PR toolchain/57241: mips64el--netbsd-install core dumps randomly

by splitting the recipes for ${PROG} and ${PROG}.debug into an
intermediate ${PROG}.link to avoid overwriting ${PROG} inside the
recipe for ${PROG}.debug.

This is kludgey (writing the `.link' suffix into a Makefile isn't
great) but I see only two cases of it so this'll do for now.

Pandaboard work from Rui-Xiang Guo via port-arm with updates from me.

https://github.com/picohive/netbsd-pandaboard

dl_addr(3): Add some automatic tests.

XXX Should also test the libc stubs, maybe move this into
tests/lib/libc/dlfcn and test both static and (for MKPIC=yes)
dynamic, but this'll do for now.

PR lib/59567: dladdr(3) doesn't work properly especially when main
executable is loaded at high memory address

Add a testcase for PR/59652.

ld.elf_so(1): Test concurrent dlopen/dlclose.

PR lib/59751: dlclose is not MT-safe depending on the libraries
unloaded

branches: 1.28.2;
centralize pack-relative-relocs handling

ld.elf_so: Add a trivial test for R_*_RELATIVE relocations.

PR bin/59360: ld.elf_so(8): missing RELR support

tests/libexec/ld.elf_so/Makefile: Sort.

Preparation for:

PR bin/59360: ld.elf_so(8): missing RELR support

t_rtld_r_debug: Test this as a non-PIE executable too.

On MIPS, non-PIEs can use DT_MIPS_RLD_MAP or DT_MIPS_RLD_MAP_REL,
while PIEs can't use DT_MIPS_RLD_MAP because it relies on an absolute
virtual address baked into the file. So it's important to test both.

PR port-mips/59296: t_rtld_r_debug test is failing

tests/libexec/ld.elf_so/t_ifunc: Handle MKRELRO=no harder.

Don't even pass -Wl,-z,norelro if MKRELRO=no, because the toolchain's
ld(1) might not support relro _or_ norelro.

Should resolve, e.g.:

/home/builds/ab/HEAD-llvm/evbarm-earmv7hf/202407182350Z-tools/bin/armv7--netbsdelf-eabihf-ld: warning: -z norelro ignored
armv7--netbsdelf-eabihf-clang: error: linker command failed with exit code 1 (use -v to see invocation)
--- t_ifunc_norelro ---
*** Failed target: t_ifunc_norelro
*** In directory: /home/source/ab/HEAD-llvm/src/tests/libexec/ld.elf_so
*** Failed commands:
${_MKTARGET_LINK}
...

tests/libexec/ld.elf_so/t_ifunc: Handle MKRELRO=no.

Just build t_ifunc and t_ifunc_now normally. If they wind up being
built with the same options as t_ifunc_norelro and
t_ifunc_norelro_now, so be it; easier to do it this way than to put
more conditionals into set lists.

Should resolve, e.g.:

/home/builds/ab/HEAD-llvm/evbarm-earmv7hf/202407141900Z-tools/bin/armv7--netbsdelf-eabihf-ld: warning: -z relro ignored
armv7--netbsdelf-eabihf-clang: error: linker command failed with exit code 1 (use -v to see invocation)
--- t_ifunc ---
*** Failed target: t_ifunc
*** In directory: /home/source/ab/HEAD-llvm/src/tests/libexec/ld.elf_so
*** Failed commands:
${_MKTARGET_LINK}
=> @# " link " ld.elf_so/t_ifunc
${_CCLINK.${:Ut_ifunc}} ${_LDFLAGS.${:Ut_ifunc}} ${_LDSTATIC.${:Ut_ifunc}} -o ${.TARGET} ${OBJS.${:Ut_ifunc}} ${_PROGLDOPTS} ${_LDADD.${:Ut_ifunc}}
=> /home/builds/ab/HEAD-llvm/evbarm-earmv7hf/202407141900Z-tools/bin/armv7--netbsdelf-eabihf-clang --sysroot=/home/builds/ab/HEAD-llvm/evbarm-earmv7hf/202407141900Z-dest -Wl,--fatal-warnings -Wl,--warn-shared-textrel -pie -o t_ifunc t_ifunc.o -Wl,-rpath-link,/home/builds/ab/HEAD-llvm/evbarm-earmv7hf/202407141900Z-dest/lib -L=/lib -Wl,-z,relro -Wl,-rpath,/usr/tests/libexec/ld.elf_so -lutil -latf-c
${CTFMERGE} ${CTFMFLAGS} -o ${.TARGET} ${OBJS.${:Ut_ifunc}}
=> /home/builds/ab/HEAD-llvm/evbarm-earmv7hf/202407141900Z-tools/bin/nbctfmerge -t -g -L VERSION -g -o t_ifunc t_ifunc.o

t_ifunc: Test with all combinations of {relro, bindnow}.

PR lib/57792

branches: 1.21.2;
ld.elf_so: Add some known-answer tests for hash functions.

Make sure the testing mechanism detects the traditional overflow bug.

adapt to ${CC_WNO_MAYBE_UNINITIALIZED}

Use ${CC_WNO_MAYBE_UNINITIALIZED} instead of
the older style more complex expressions.

Remove workarounds if they were for a specific
version of gcc < 10.

ld.elf_so: Test dynamic-then-static abuse via ctor.

XXX pullup-10

ld.elf_so: Test another edge case of mismatched TLS models.

One library defines a symbol and _doesn't_ use it, so it has no
indication of whether the symbol is for static TLS or dynamic TLS,
and then two other libraries use it in different ways.

XXX pullup-10

ld.elf_so: Test TLS abuse of static def, dynamic use and vice versa.

XXX pullup-10

ld.elf_so: Test extern dynamic TLS too.

XXX pullup-10

ld.elf_so: Nix inadvertently committed private test program.

ld.elf_so: Fix extern TLS test to match PR toolchain/50277.

Now it's actually testing the problem.

ld.elf_so: New test for extern initial-exec TLS, PR toolchain/50277.

XXX pullup-10

branches: 1.12.4;
Drop explicit -msecure-plt, it's the default for both clang and gcc.

branches: 1.11.2;
Add new RTLD test file for r_debug

New tests:
- self
- dlopen

Both check whether the r_debug structure seems to be well-formed, without
and with a dlopen(3) call.

branches: 1.10.2;
clang ifunc's on powerpc require secure-plt.

branches: 1.9.6;
Implement __cxa_thread_atexit and __cxa_thread_atexit_impl. This
functions are used for destructors of thread_local objects.

If a pending destructor exists, prevent unloading of shared objects.
Introduce __dl_cxa_refcount interface for this purpose. When the last
reference is gone and the object has been dlclose'd before, the
unloading is finalized.

Ideally, __cxa_thread_atexit_impl wouldn't exist, but libstdc++ insists
on providing __cxa_thread_atexit as direct wrapper without further
patching.

branches: 1.8.8;
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.

avoid building stuff that depends upon pic objects if MKPIC == no.

only build the ld.elf_so tests if MKPIC != no.

PR/45015: ld.elf_so: support ELF symbol versioning
Applied latest patch.

Add initial regression test for ld.elf_so locking.

Add support for DF_1_BIND_NOW, DF_1_NODELETE and DF_1_NOOPEN marked
objects, and the RTLD_NODELETE and RTLD_NOLOAD flags to dlopen(3).

Mark libpthread as DF_1_NOOPEN and use it to test the functionality.

Somewhat taken from FreeBSD.

Fixes PR 42029.

OK from christos and joerg.

Convert a couple of old regress tests to ATF.

Remove the regress/usr.bin/rtld/dlopen/dlopen.c test as well as it's
covered by existing ATF test(s).

add regression tests for dlinfo()

Drop explicit -msecure-plt, it's the default for both clang and gcc.

branches: 1.2.6;
clang ifunc's on powerpc require secure-plt.

branches: 1.1.2; 1.1.44;
Turn warns on for all tests and fix all the bugs.

Use proper format string

Blame where blame is due.

Add support for DF_1_BIND_NOW, DF_1_NODELETE and DF_1_NOOPEN marked
objects, and the RTLD_NODELETE and RTLD_NOLOAD flags to dlopen(3).

Mark libpthread as DF_1_NOOPEN and use it to test the functionality.

Somewhat taken from FreeBSD.

Fixes PR 42029.

OK from christos and joerg.

PR/45015: ld.elf_so: support ELF symbol versioning
Applied latest patch.

Disable attempt to use ifunc on hppa. It doesn't work and new binutils
throws a build error.

Avoid casting fun by switching ifunc helper functions to return
long long. Dead beef is too useful to switch to a different constant.

branches: 1.1.14;
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.

Add initial regression test for ld.elf_so locking.

branches: 1.1.4;
ld.elf_so: Add a trivial test for R_*_RELATIVE relocations.

PR bin/59360: ld.elf_so(8): missing RELR support

branches: 1.2.2;
Avoid a race condition between dlclose and thread termination.

Implement __cxa_thread_atexit and __cxa_thread_atexit_impl. This
functions are used for destructors of thread_local objects.

If a pending destructor exists, prevent unloading of shared objects.
Introduce __dl_cxa_refcount interface for this purpose. When the last
reference is gone and the object has been dlclose'd before, the
unloading is finalized.

Ideally, __cxa_thread_atexit_impl wouldn't exist, but libstdc++ insists
on providing __cxa_thread_atexit as direct wrapper without further
patching.

branches: 1.1.2; 1.1.4;
tests/libexec/ld.elf_so: Fix helper library makefiles.

1. Consolidate logic into a single helper.mk to reduce duplication.
2. Set NO* variables, not MK* variables which are reserved for user.
3. Avoid eager X!= in favour of lazy ${X:sh}.
4. Mark _g.a set list entries obsolete. Never should've been built!

PR misc/57462

Blame where blame is due.

test helper program exit code, not test program exit code

Add support for DF_1_BIND_NOW, DF_1_NODELETE and DF_1_NOOPEN marked
objects, and the RTLD_NODELETE and RTLD_NOLOAD flags to dlopen(3).

Mark libpthread as DF_1_NOOPEN and use it to test the functionality.

Somewhat taken from FreeBSD.

Fixes PR 42029.

OK from christos and joerg.

PR/45015: ld.elf_so: support ELF symbol versioning
Applied latest patch.

branches: 1.2.2;
dladdr(3): Fix determination of object address intervals.

This appears to have been broken in rev. 1.62 back in 2017 with a
change to compute the largest distance from the first LOAD segment's
starting address to the maximum address of any LOAD segment -- we
started by finding the distance from _zero_ to the maximum address of
the first LOAD segment because we subtracted obj->vaddrbase which
wasn't initialized yet, so we would overshoot obj->mapsize and
obj->textsize by whatever the value of obj->vaddrbase turned out to
be.

XXX The commit message for 1.62 suggests we are dropping the
assumption that the LOAD segments appear in order. But it looks to
me like we still assume the _first_ load segment has the lowest
virtual address, because we keep it as obj->vaddrbase/obj->mapbase
and use it as a reference point for measuring the size of an object
by subtracting it from each load segment's maximum address. Is the
original assumption that LOAD segments appear in order valid after
all, or do we need to do extra work here to handle objects whose LOAD
segments are so much out of order that the first LOAD segment isn't
lowest in virtual address space? Unclear!

Might help with wine.

Based on a patch by Akihiro Sagawa, thanks!

PR lib/59567: dladdr(3) doesn't work properly especially when main
executable is loaded at high memory address

dl_addr(3): Add some automatic tests.

XXX Should also test the libc stubs, maybe move this into
tests/lib/libc/dlfcn and test both static and (for MKPIC=yes)
dynamic, but this'll do for now.

PR lib/59567: dladdr(3) doesn't work properly especially when main
executable is loaded at high memory address

branches: 1.1.2;
ld.elf_so(1): Test concurrent dlopen/dlclose.

PR lib/59751: dlclose is not MT-safe depending on the libraries
unloaded

return return atf_no_error() instead of 0 for consistency.

suggested by moritzbuhl in https://github.com/NetBSD/src/pull/11/

branches: 1.2.14;
Don't play with "../.." in includes for h_macros.h; deal with it centrally.
Minor fixes.

branches: 1.1.28;
Convert a couple of old regress tests to ATF.

Remove the regress/usr.bin/rtld/dlopen/dlopen.c test as well as it's
covered by existing ATF test(s).

Trailing whitespace

Don't play with "../.." in includes for h_macros.h; deal with it centrally.
Minor fixes.

branches: 1.1.28;
Convert a couple of old regress tests to ATF.

Remove the regress/usr.bin/rtld/dlopen/dlopen.c test as well as it's
covered by existing ATF test(s).

t_dlinfo, t_rtld_r_debug: Audit ATF_REQUIRE/CHECK, sprinkle messages.

Make sure to use ATF_REQUIRE when the subsequent code doesn't work if
the check fails.

rtld tests: Don't use RZ for dlinfo.

Use

ATF_REQUIRE_EQ_MSG(dlinfo(...), 0, "dlinfo: %s", dlerror())

instead, in order to accurately report the error on failure. RZ is
only for functions like pthread_create(3) that return zero on success
and errno(3) code on failure, but dlinfo returns -1 on failure and
sets dlerror() to report the nature of the error.

return return atf_no_error() instead of 0 for consistency.

suggested by moritzbuhl in https://github.com/NetBSD/src/pull/11/

branches: 1.5.14;
Don't play with "../.." in includes for h_macros.h; deal with it centrally.
Minor fixes.

branches: 1.4.26;
check that we can dlopen() the linkmap name, too

check that dlopen()d dso shows up in the linkmap

add Makefile.inc everywhere so that we can set WARNS=4 by default. Amazing
how many bugs this found :-)

add regression tests for dlinfo()

PR/45015: ld.elf_so: support ELF symbol versioning
Applied latest patch.

tests/libexec/ld.elf_so/t_hash: Add web.archive.org link.

The original link still works for now but let's make sure it doesn't
bitrot later.

No functional change intended: comment only.

branches: 1.1.2; 1.1.4; 1.1.6; 1.1.8;
ld.elf_so: Add some known-answer tests for hash functions.

Make sure the testing mechanism detects the traditional overflow bug.

sort; it is the same list as in h_ifunc_static.c; perhaps it should be
a HAVE_ something?

Support ifunc on aarch64. The tests pass at least.

Wrap long #if defined(...) || ... and sort. NFCI.

Whitespace

return return atf_no_error() instead of 0 for consistency.

suggested by moritzbuhl in https://github.com/NetBSD/src/pull/11/

branches: 1.8.2;
Avoid casting fun by switching ifunc helper functions to return
long long. Dead beef is too useful to switch to a different constant.

branches: 1.7.2;
Build less of the test code if there's no linker ifunc support.

r1.6 changed the code in such a way that it was hitting a BFD assert on
MIPS, causing the builds to fail.

we were hitting the assert at binutils/dist/bfd/elfxx-mips.c:10879

Skip all tests on architectures w/o ifunc linker/ld.elf_so support.

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.

Don't play with "../.." in includes for h_macros.h; deal with it centrally.
Minor fixes.

branches: 1.1.2;
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: 1.2.4;
ld.elf_so: Implement RELR relocations.

The SHT_RELR-type .relr.dyn section, pointed to by the DT_RELR
dynamic tag, holds compressed R_*_RELATIVE-type relocations,
substantially reducing the disk space occupied by many programs.

This commit adds support in ld.elf_so for relocating dynamically
linked executables with RELR relocations.

Reference:

Rahul Chaudhry, `Re: Proposal for a new section type SHT_RELR',
generic-abi mailing list, 2018-02-07.

https://groups.google.com/g/generic-abi/c/bX460iggiKg/m/Jnz1lgLJAgAJ
https://web.archive.org/web/20241213012330/https://groups.google.com/g/generic-abi/c/bX460iggiKg/m/Jnz1lgLJAgAJ

Proposed on tech-userlevel:
https://mail-index.NetBSD.org/tech-userlevel/2025/04/27/msg014727.html

PR bin/59360: ld.elf_so(8): missing RELR support

ld.elf_so: Add a trivial test for R_*_RELATIVE relocations.

PR bin/59360: ld.elf_so(8): missing RELR support

t_rtld_r_debug: Break once we've found DT_MIPS_RLD_MAP_REL.

No need to keep searching at that point (though maybe it wouldn't
hurt to verify that if we have both DT_MIPS_RLD_MAP and
DT_MIPS_RLD_MAP_REL, they agree).

PR port-mips/59296: t_rtld_r_debug test is failing

ld.elf_so: Teach this to handle MIPS PIE rtld debug data.

Adapt t_rtld_r_debug to handle the two MIPS cases too.

XXX t_rtld_r_debug should be tested both as PIE and non-PIE to
exercise both cases.

Context:

The value of a DT_DEBUG .dynamic entry is initialized at load-time,
by ld.elf_so, to a pointer to a data structure set up by ld.elf_so
describing the shared objects loaded by the executable, so debuggers
can find them from, e.g., core dumps. None of this is really
documented anywhere that I can find. Best reference is this post on
the gdb mailing list from a quarter century ago saying there's no
real documentation:

https://web.archive.org/web/20250414021320/https://sourceware.org/pipermail/gdb/2000-April/004509.html

However, on MIPS, the .dynamic section is mapped read-only, so
ld.elf_so can't properly modify it (I imagine technically it could
with some mprotect shenanigans but that's not how it's done on MIPS).
Instead, the linker reserves a location in read/write memory and uses
a DT_MIPS_RLD_MAP entry with a pointer to that location.

However, in position-independent executables, the .dynamic entry
can't have an absolute pointer to that location because it's not
known up front. Instead, the the linker uses a DT_MIPS_RLD_MAP_REL
entry with the relative offset to that location from the Elf_Dyn
entry itself.

I would add a reference for this but it's basically a matter of UTSL
plus some oblique mentions on the web and mailing list discussions:

https://web.archive.org/web/20250414024823/https://cygwin.com/legacy-ml/binutils/2016-04/msg00244.html
https://web.archive.org/web/20250403151803/https://maskray.me/blog/2023-09-04-toolchain-notes-on-mips
https://web.archive.org/web/20211024050833/https://reviews.llvm.org/D12794?id=34533
https://web.archive.org/web/20250407052145/https://wiki.debian.org/MIPSPort
https://web.archive.org/web/20250414024924/https://reviews.freebsd.org/D17867?id=50122

PR port-mips/59296: t_rtld_r_debug test is failing

t_rtld_r_debug: Fix xfail on mips.

There was an earlier assertion, oops.

Add __RCSID while here.

PR port-mips/59296: t_rtld_r_debug test is failing

t_rtld_r_debug: Mark this xfail on mips.

PR port-mips/59296: t_rtld_r_debug test is failing

t_rtld_r_debug: Print some diagnostics as we go.

t_rtld_r_debug: Use ATF_REQUIRE where we can't proceed on failure.

Not helpful for the test program to crash with SIGSEGV on null
pointer dereference -- just make it stop when we detect the null
pointer.

fix a/an grammar in obvious cases

branches: 1.5.2;
t_dlinfo, t_rtld_r_debug: Audit ATF_REQUIRE/CHECK, sprinkle messages.

Make sure to use ATF_REQUIRE when the subsequent code doesn't work if
the check fails.

rtld tests: Don't use RZ for dlinfo.

Use

ATF_REQUIRE_EQ_MSG(dlinfo(...), 0, "dlinfo: %s", dlerror())

instead, in order to accurately report the error on failure. RZ is
only for functions like pthread_create(3) that return zero on success
and errno(3) code on failure, but dlinfo returns -1 on failure and
sets dlerror() to report the nature of the error.

Whitespace

Fix build with clang.

Add new RTLD test file for r_debug

New tests:
- self
- dlopen

Both check whether the r_debug structure seems to be well-formed, without
and with a dlopen(3) call.

branches: 1.1.4;
Implement __cxa_thread_atexit and __cxa_thread_atexit_impl. This
functions are used for destructors of thread_local objects.

If a pending destructor exists, prevent unloading of shared objects.
Introduce __dl_cxa_refcount interface for this purpose. When the last
reference is gone and the object has been dlclose'd before, the
unloading is finalized.

Ideally, __cxa_thread_atexit_impl wouldn't exist, but libstdc++ insists
on providing __cxa_thread_atexit as direct wrapper without further
patching.

branches: 1.1.2;
Add a testcase for PR/59652.

ld.elf_so aarch64/rtld_start.S: Fix dynamic TLS fast path branch.

Bug found and patch prepared by pho@.

PR lib/58154

tests/libexec/ld.elf_so/t_tls_extern: Test PR lib/58154.

Alter prev to just explicitly use 0 instead of NULL, so there's no
need to ensure NULL is defined.

Compensate for change in ATF_REQUIRE_MSG

ATF_REQUIRE_MSG() was recently changed so that ATF_REQUIRE_MSG(x = y, ..)
would generate a warning message, as it probably intended to be x == y.

In ATF_REQUIRE_DL() the use is always ATF_REQUIRE_DL(var = dlopen(...), ...)
where the '=' is certainly not intended to be ==, but is instead checking
that the result of dlopen() is not NULL (while also saving it for later).

Alter the definition of ATF_REQUIRE_DL() to make that explicit.

(The s/\t/ / is just for 80 column police avoidance)

branches: 1.12.2; 1.12.4; 1.12.6;
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.

ld.elf_so: Test dynamic-then-static abuse via ctor.

XXX pullup-10

ld.elf_so: Test another edge case of mismatched TLS models.

One library defines a symbol and _doesn't_ use it, so it has no
indication of whether the symbol is for static TLS or dynamic TLS,
and then two other libraries use it in different ways.

XXX pullup-10

ld.elf_so: tls_extern dynamic_defabuse_eager must xfail differently.

If a symbol has already been resolved as dynamic TLS, any library
that tries to use it as static TLS cannot be dlopened.

XXX pullup-10

ld.elf_so: Separately test eager and lazy resolution of def tls ptr.

eager: before loading use library
lazy: after loading use library

XXX pullup-10

ld.elf_so: Shorter test names.

No functional non-cosmetic change intended.

XXX pullup-10

ld.elf_so: Test TLS abuse of static def, dynamic use and vice versa.

XXX pullup-10

ld.elf_so: Factor out logic in TLS tests to make writing more easier.

No functional change intended.

XXX pullup-10

ld.elf_so: Test extern dynamic TLS too.

XXX pullup-10

ld.elf_so: Test variations on PR toolchain/50277.

XXX pullup-10

ld.elf_so: Fix extern TLS test to match PR toolchain/50277.

Now it's actually testing the problem.

ld.elf_so: New test for extern initial-exec TLS, PR toolchain/50277.

XXX pullup-10

PR/45015: ld.elf_so: support ELF symbol versioning
Applied latest patch.

PR/45015: ld.elf_so: support ELF symbol versioning
Applied latest patch.

PR/45015: ld.elf_so: support ELF symbol versioning
Applied latest patch.

PR/45015: ld.elf_so: support ELF symbol versioning
Applied latest patch.

PR/45015: ld.elf_so: support ELF symbol versioning
Applied latest patch.

PR/45015: ld.elf_so: support ELF symbol versioning
Applied latest patch.

PR/45015: ld.elf_so: support ELF symbol versioning
Applied latest patch.

PR/45015: ld.elf_so: support ELF symbol versioning
Applied latest patch.

PR/45015: ld.elf_so: support ELF symbol versioning
Applied latest patch.

PR/45015: ld.elf_so: support ELF symbol versioning
Applied latest patch.

PR/45015: ld.elf_so: support ELF symbol versioning
Applied latest patch.

PR/45015: ld.elf_so: support ELF symbol versioning
Applied latest patch.

PR/45015: ld.elf_so: support ELF symbol versioning
Applied latest patch.

PR/45015: ld.elf_so: support ELF symbol versioning
Applied latest patch.

PR/45015: ld.elf_so: support ELF symbol versioning
Applied latest patch.

PR/45015: ld.elf_so: support ELF symbol versioning
Applied latest patch.

PR/45015: ld.elf_so: support ELF symbol versioning
Applied latest patch.

PR/45015: ld.elf_so: support ELF symbol versioning
Applied latest patch.

PR/45015: ld.elf_so: support ELF symbol versioning
Applied latest patch.

branches: 1.2.2; 1.2.4;
tests/libexec/ld.elf_so: Fix helper library makefiles.

1. Consolidate logic into a single helper.mk to reduce duplication.
2. Set NO* variables, not MK* variables which are reserved for user.
3. Avoid eager X!= in favour of lazy ${X:sh}.
4. Mark _g.a set list entries obsolete. Never should've been built!

PR misc/57462

ld.elf_so: Test TLS abuse of static def, dynamic use and vice versa.

XXX pullup-10

branches: 1.1.2; 1.1.4;
ld.elf_so: Test TLS abuse of static def, dynamic use and vice versa.

XXX pullup-10

branches: 1.2.2; 1.2.4;
tests/libexec/ld.elf_so: Fix helper library makefiles.

1. Consolidate logic into a single helper.mk to reduce duplication.
2. Set NO* variables, not MK* variables which are reserved for user.
3. Avoid eager X!= in favour of lazy ${X:sh}.
4. Mark _g.a set list entries obsolete. Never should've been built!

PR misc/57462

ld.elf_so: Test TLS abuse of static def, dynamic use and vice versa.

XXX pullup-10

branches: 1.1.2; 1.1.4;
ld.elf_so: Test TLS abuse of static def, dynamic use and vice versa.

XXX pullup-10

branches: 1.2.2; 1.2.4;
tests/libexec/ld.elf_so: Fix helper library makefiles.

1. Consolidate logic into a single helper.mk to reduce duplication.
2. Set NO* variables, not MK* variables which are reserved for user.
3. Avoid eager X!= in favour of lazy ${X:sh}.
4. Mark _g.a set list entries obsolete. Never should've been built!

PR misc/57462

ld.elf_so: Test extern dynamic TLS too.

XXX pullup-10

branches: 1.1.2; 1.1.4;
ld.elf_so: Test extern dynamic TLS too.

XXX pullup-10

branches: 1.2.2; 1.2.4;
tests/libexec/ld.elf_so: Fix helper library makefiles.

1. Consolidate logic into a single helper.mk to reduce duplication.
2. Set NO* variables, not MK* variables which are reserved for user.
3. Avoid eager X!= in favour of lazy ${X:sh}.
4. Mark _g.a set list entries obsolete. Never should've been built!

PR misc/57462

ld.elf_so: New test for extern initial-exec TLS, PR toolchain/50277.

XXX pullup-10

branches: 1.1.2; 1.1.4;
ld.elf_so: New test for extern initial-exec TLS, PR toolchain/50277.

XXX pullup-10

tests/libexec/ld.elf_so: Fix helper library makefiles.

1. Consolidate logic into a single helper.mk to reduce duplication.
2. Set NO* variables, not MK* variables which are reserved for user.
3. Avoid eager X!= in favour of lazy ${X:sh}.
4. Mark _g.a set list entries obsolete. Never should've been built!

PR misc/57462

branches: 1.1.46; 1.1.54;
Add initial regression test for ld.elf_so locking.

Introduce __HAVE_NO___THREAD for sun2 and vax to disable the TLS usage.
Require __HAVE_TLS_VARIANT_I or __HAVE_TLS_VARIANT_II as well as
__lwp_getprivate_fast / __lwp_gettcb_fast to exist for libpthread.
Define VAX as going to use TLS variant I, if it is ever implemented.

branches: 1.3.2;
Turn warns on for all tests and fix all the bugs.

branches: 1.2.2;
Don't fail on platforms without TLS

Add initial regression test for ld.elf_so locking.

tests/libexec/ld.elf_so: Fix helper library makefiles.

1. Consolidate logic into a single helper.mk to reduce duplication.
2. Set NO* variables, not MK* variables which are reserved for user.
3. Avoid eager X!= in favour of lazy ${X:sh}.
4. Mark _g.a set list entries obsolete. Never should've been built!

PR misc/57462

branches: 1.1.46; 1.1.54;
Add initial regression test for ld.elf_so locking.

Turn warns on for all tests and fix all the bugs.

branches: 1.1.2;
Add initial regression test for ld.elf_so locking.

tests/libexec/ld.elf_so: Fix helper library makefiles.

1. Consolidate logic into a single helper.mk to reduce duplication.
2. Set NO* variables, not MK* variables which are reserved for user.
3. Avoid eager X!= in favour of lazy ${X:sh}.
4. Mark _g.a set list entries obsolete. Never should've been built!

PR misc/57462

branches: 1.1.4; 1.1.10; 1.1.18;
Implement __cxa_thread_atexit and __cxa_thread_atexit_impl. This
functions are used for destructors of thread_local objects.

If a pending destructor exists, prevent unloading of shared objects.
Introduce __dl_cxa_refcount interface for this purpose. When the last
reference is gone and the object has been dlclose'd before, the
unloading is finalized.

Ideally, __cxa_thread_atexit_impl wouldn't exist, but libstdc++ insists
on providing __cxa_thread_atexit as direct wrapper without further
patching.

branches: 1.1.4;
Implement __cxa_thread_atexit and __cxa_thread_atexit_impl. This
functions are used for destructors of thread_local objects.

If a pending destructor exists, prevent unloading of shared objects.
Introduce __dl_cxa_refcount interface for this purpose. When the last
reference is gone and the object has been dlclose'd before, the
unloading is finalized.

Ideally, __cxa_thread_atexit_impl wouldn't exist, but libstdc++ insists
on providing __cxa_thread_atexit as direct wrapper without further
patching.

tests/libexec/ld.elf_so: Fix helper library makefiles.

1. Consolidate logic into a single helper.mk to reduce duplication.
2. Set NO* variables, not MK* variables which are reserved for user.
3. Avoid eager X!= in favour of lazy ${X:sh}.
4. Mark _g.a set list entries obsolete. Never should've been built!

PR misc/57462

branches: 1.1.18; 1.1.26;
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.

Good bye oabi! Tested building earmv7hf.

PR/56230: Don't try to use ifunc with arm and oabi

branches: 1.8.4;
Disable attempt to use ifunc on hppa. It doesn't work and new binutils
throws a build error.

Avoid casting fun by switching ifunc helper functions to return
long long. Dead beef is too useful to switch to a different constant.

branches: 1.6.2;
Add sparc/sparc64 support for irelative relocations.

Add IRELATIVE support for ARM, X86 and PowerPC.

Really revert to the original version.

Revert. There is no such thing as __HAVE_IFUNC and there should not be
one in first place.

effectively disable the test to fix the build and explain why.

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: 1.2.2; 1.2.4;
tests/libexec/ld.elf_so: Fix helper library makefiles.

1. Consolidate logic into a single helper.mk to reduce duplication.
2. Set NO* variables, not MK* variables which are reserved for user.
3. Avoid eager X!= in favour of lazy ${X:sh}.
4. Mark _g.a set list entries obsolete. Never should've been built!

PR misc/57462

ld.elf_so: Test dynamic-then-static abuse via ctor.

XXX pullup-10

branches: 1.1.2; 1.1.4;
ld.elf_so: Test dynamic-then-static abuse via ctor.

XXX pullup-10

branches: 1.2.2; 1.2.4;
tests/libexec/ld.elf_so: Fix helper library makefiles.

1. Consolidate logic into a single helper.mk to reduce duplication.
2. Set NO* variables, not MK* variables which are reserved for user.
3. Avoid eager X!= in favour of lazy ${X:sh}.
4. Mark _g.a set list entries obsolete. Never should've been built!

PR misc/57462

ld.elf_so: Test another edge case of mismatched TLS models.

One library defines a symbol and _doesn't_ use it, so it has no
indication of whether the symbol is for static TLS or dynamic TLS,
and then two other libraries use it in different ways.

XXX pullup-10

branches: 1.1.2; 1.1.4;
ld.elf_so: Test another edge case of mismatched TLS models.

One library defines a symbol and _doesn't_ use it, so it has no
indication of whether the symbol is for static TLS or dynamic TLS,
and then two other libraries use it in different ways.

XXX pullup-10

branches: 1.2.2; 1.2.4;
tests/libexec/ld.elf_so: Fix helper library makefiles.

1. Consolidate logic into a single helper.mk to reduce duplication.
2. Set NO* variables, not MK* variables which are reserved for user.
3. Avoid eager X!= in favour of lazy ${X:sh}.
4. Mark _g.a set list entries obsolete. Never should've been built!

PR misc/57462

ld.elf_so: Test another edge case of mismatched TLS models.

One library defines a symbol and _doesn't_ use it, so it has no
indication of whether the symbol is for static TLS or dynamic TLS,
and then two other libraries use it in different ways.

XXX pullup-10

branches: 1.1.2; 1.1.4;
ld.elf_so: Test another edge case of mismatched TLS models.

One library defines a symbol and _doesn't_ use it, so it has no
indication of whether the symbol is for static TLS or dynamic TLS,
and then two other libraries use it in different ways.

XXX pullup-10

branches: 1.2.2; 1.2.4;
tests/libexec/ld.elf_so: Fix helper library makefiles.

1. Consolidate logic into a single helper.mk to reduce duplication.
2. Set NO* variables, not MK* variables which are reserved for user.
3. Avoid eager X!= in favour of lazy ${X:sh}.
4. Mark _g.a set list entries obsolete. Never should've been built!

PR misc/57462

ld.elf_so: Test another edge case of mismatched TLS models.

One library defines a symbol and _doesn't_ use it, so it has no
indication of whether the symbol is for static TLS or dynamic TLS,
and then two other libraries use it in different ways.

XXX pullup-10

branches: 1.1.2; 1.1.4;
ld.elf_so: Test another edge case of mismatched TLS models.

One library defines a symbol and _doesn't_ use it, so it has no
indication of whether the symbol is for static TLS or dynamic TLS,
and then two other libraries use it in different ways.

XXX pullup-10

branches: 1.2.2; 1.2.4;
tests/libexec/ld.elf_so: Fix helper library makefiles.

1. Consolidate logic into a single helper.mk to reduce duplication.
2. Set NO* variables, not MK* variables which are reserved for user.
3. Avoid eager X!= in favour of lazy ${X:sh}.
4. Mark _g.a set list entries obsolete. Never should've been built!

PR misc/57462

ld.elf_so: Test another edge case of mismatched TLS models.

One library defines a symbol and _doesn't_ use it, so it has no
indication of whether the symbol is for static TLS or dynamic TLS,
and then two other libraries use it in different ways.

XXX pullup-10

branches: 1.1.2; 1.1.4;
ld.elf_so: Test another edge case of mismatched TLS models.

One library defines a symbol and _doesn't_ use it, so it has no
indication of whether the symbol is for static TLS or dynamic TLS,
and then two other libraries use it in different ways.

XXX pullup-10

tests/libexec/ld.elf_so: Fix helper library makefiles.

1. Consolidate logic into a single helper.mk to reduce duplication.
2. Set NO* variables, not MK* variables which are reserved for user.
3. Avoid eager X!= in favour of lazy ${X:sh}.
4. Mark _g.a set list entries obsolete. Never should've been built!

PR misc/57462

branches: 1.1.44; 1.1.52;
PR/45015: ld.elf_so: support ELF symbol versioning
Applied latest patch.

PR/45015: ld.elf_so: support ELF symbol versioning
Applied latest patch.

tests/libexec/ld.elf_so: Fix helper library makefiles.

1. Consolidate logic into a single helper.mk to reduce duplication.
2. Set NO* variables, not MK* variables which are reserved for user.
3. Avoid eager X!= in favour of lazy ${X:sh}.
4. Mark _g.a set list entries obsolete. Never should've been built!

PR misc/57462

branches: 1.3.28; 1.3.36;
ARM now works with integrated-as.

Disable integrated-assembler for clang on ARM for two files that are
victims of the @ comment mess on that platform. Add a reference to the
upstream PR.

branches: 1.1.2; 1.1.8;
PR/45015: ld.elf_so: support ELF symbol versioning
Applied latest patch.

PR/45015: ld.elf_so: support ELF symbol versioning
Applied latest patch.

PR/45015: ld.elf_so: support ELF symbol versioning
Applied latest patch.

tests/libexec/ld.elf_so: Fix helper library makefiles.

1. Consolidate logic into a single helper.mk to reduce duplication.
2. Set NO* variables, not MK* variables which are reserved for user.
3. Avoid eager X!= in favour of lazy ${X:sh}.
4. Mark _g.a set list entries obsolete. Never should've been built!

PR misc/57462

branches: 1.3.28; 1.3.36;
ARM now works with integrated-as.

Disable integrated-assembler for clang on ARM for two files that are
victims of the @ comment mess on that platform. Add a reference to the
upstream PR.

PR/45015: ld.elf_so: support ELF symbol versioning
Applied latest patch.

PR/45015: ld.elf_so: support ELF symbol versioning
Applied latest patch.

PR/45015: ld.elf_so: support ELF symbol versioning
Applied latest patch.

branches: 1.2.2; 1.2.4;
tests/libexec/ld.elf_so: Fix helper library makefiles.

1. Consolidate logic into a single helper.mk to reduce duplication.
2. Set NO* variables, not MK* variables which are reserved for user.
3. Avoid eager X!= in favour of lazy ${X:sh}.
4. Mark _g.a set list entries obsolete. Never should've been built!

PR misc/57462

ld.elf_so: Test extern dynamic TLS too.

XXX pullup-10

branches: 1.1.2; 1.1.4;
ld.elf_so: Test extern dynamic TLS too.

XXX pullup-10

branches: 1.3.2; 1.3.4;
tests/libexec/ld.elf_so: Fix helper library makefiles.

1. Consolidate logic into a single helper.mk to reduce duplication.
2. Set NO* variables, not MK* variables which are reserved for user.
3. Avoid eager X!= in favour of lazy ${X:sh}.
4. Mark _g.a set list entries obsolete. Never should've been built!

PR misc/57462

ld.elf_so: Fix extern TLS test to match PR toolchain/50277.

Now it's actually testing the problem.

ld.elf_so: New test for extern initial-exec TLS, PR toolchain/50277.

XXX pullup-10

branches: 1.2.2; 1.2.4;
ld.elf_so: Fix extern TLS test to match PR toolchain/50277.

Now it's actually testing the problem.

ld.elf_so: New test for extern initial-exec TLS, PR toolchain/50277.

XXX pullup-10