The right unwinding stop indicator should be CFI-undefined PC.
https://dwarfstd.org/doc/Dwarf3.pdf - page 118:
If a Return Address register is defined in the virtual unwind table,
and its rule is undefined (for example, by DW_CFA_undefined), then
there is no return address and no call address, and the virtual
unwind of stack activations is complete.
The hack localizing _start1 symbol removed.
Reviewed by: kib
Differential Revision: https://reviews.freebsd.org/D40624
Normally, modern unwinders uses Dwarf information to unwind stack,
however in case when the code is not annotated by Dwarf instructions,
unwinders fallbacks to a frame-pointer based algorithm.
That is allows libunwind to unwind stack from global constructors and
destructors. Also it makes gdb happy as it printed nonexistent frame
before.
Reviewed by: kib, imp
Differential Revision: https://reviews.freebsd.org/D40948
Why? Most trivial point, it shaves around 600 bytes from the dynamic
binaries on amd64. Less trivial, the removed code is no longer part of
the ABI, and we can ship updates to it with libc updates. Right now most
of the csu is linked into the binaries and require us to do somewhat
tricky ABI compat when it needs to change. For instance, the init_array
change would be much simpler and does not require note tagging if we
have init calling code in libc.
This could be improved more, by splitting dynamic and static
initialization. For instance, &_DYNAMIC tests can be removed then.
Such change, nonetheless, would require building libc three times.
I left this for later, after this change stabilizes, if ever.
Reviewed by: markj
Discussed with: jrtc27 (some objections, see the review), imp
Tested by: markj (aarch64)
Sponsored by: The FreeBSD Foundation
MFC after: 3 weeks
Differential revision: https://reviews.freebsd.org/D37220
The ELF notes compiled in C were placed in a section with the wrong type
(SHT_PROGBITS instead of SHT_NOTE). Previously, sed was used on the
generated assembly to rewrite the section type. Instead, write the notes
in assembly which permits setting the correct section type directly.
While here, move inline assembly entry points out of C and into assembly
for aarch64, arm, and riscv.
Reviewed by: kib
Tested on: amd64 (cirrus-ci), riscv64
Sponsored by: DARPA
Differential Revision: https://reviews.freebsd.org/D25211
These are needed for .ctors/.dtors and .jcr handling. The former needs
all the function pointers to be called in the correct order from the
.init/.fini section. The latter just needs to call a gcj specific function
if it exists with a pointer to the start of the .jcr section.
This is currently disabled until __dso_handle support is added.
Reviewed by: emaste
MFC after: 1 month
Sponsored by: DARPA, AFRL
Differential Revision: https://reviews.freebsd.org/D17587
This makes statically linked binaries with ifuncs operational.
Reported and tested by: mjg
Reviewed by: emaste, markj
Sponsored by: The FreeBSD Foundation
Approved by: re (rgrimes)
Differential revision: https://reviews.freebsd.org/D17363
On most architectures crt objects are compiled in a multiple-step process
so that sed can be run on the generated assembly. As the final step,
the C compiler generates an object file from the modified assembly output.
Currently this last step uses $CC with only $ACFLAGS. However, for other
uses in the tree, $ACFLAGS is meant to include assembly-specific compiler
flags that are in addition to $CFLAGS (see default .S.o rules
bsd.suffixes.mk). In particular, external toolchains may require
additional flags to select a non-default target which will be present
in CFLAGS but not ACFLAGS. To support this while still mitigating the
issue with CFLAGS described in r234502, include a modified CFLAGS that
excludes "-g" when assembling the modified assembly files.
Note that normally an assembler ($AS) is used to assemble .s flags to
object files (see bsd.suffixes.mk). However, llvm-based toolchains do
not currently have a stand-alone assembler.
Reviewed by: imp
Sponsored by: DARPA / AFRL
Differential Revision: https://reviews.freebsd.org/D10085
The requirement is for a GCC-compatible compiler and not necessarily
GCC itself. However, we currently expect any compiler used for building
the whole of FreeBSD to be GCC-compatible and many things will break if
not; there's no longer a need to have an explicit test for this in csu.
Sponsored by: The FreeBSD Foundation
FILES is not used when LIBRARIES_ONLY is set, which is used to build and
install the lib32 sysroot. All of the csu files do quality as "libraries"
for this case so just undefine LIBRARIES_ONLY.
This is still better than the previous realinstall handling as it does
not hook into META_MODE properly.
Sponsored by: EMC / Isilon Storage Division
This partially reverts r270170 for lib/csu/i386 while retaining the
change for using bsd.lib.mk.
These FILES groups could go into lib/csu/Makefile.inc but I've kept them
in the Makefiles for clarity.
Sponsored by: EMC / Isilon Storage Division
Off by default, build behaves normally.
WITH_META_MODE we get auto objdir creation, the ability to
start build from anywhere in the tree.
Still need to add real targets under targets/ to build packages.
Differential Revision: D2796
Reviewed by: brooks imp
This was a leftover from when we had both i386 a.out and ELF.
Reviewed by: kib, imp
Sponsored by: The FreeBSD Foundation
Differential Revision: https://reviews.freebsd.org/D2591
Also, make an internal _getprogname() that is used only inside
libc. For libc, getprogname(3) is a weak symbol in case a
function of the same name is defined in userland.
Move a.out libraries to /usr/lib/aout to make space for ELF libs.
Make rtld usr /usr/lib/aout as default library path.
Make ldconfig reject /usr/lib as an a.out library path.
Fix various Makefiles for LIBDIR!=/usr/lib breakage.
This will after a make world & reboot give a system that no
longer uses /usr/lib/*, infact one could remove all the old
libraries there, they are not used anymore.
We are getting close to an ELF make world, but I'll let this
all settle for a week or two...
into libc. This reduces the size of every dynamically linked
executable by 248 bytes, and it reduces the size of static executables
by a lesser amount. It also eliminates some global namespace
pollution.
With this change in place, the source for dlfcn.h should probably
be moved to "/usr/src/include". I'll save that for another day.
Compatibility note: Programs which use dlopen, if compiled on
systems with this change, will not run on systems with a libc from
prior to this change. Very few programs use dlopen, so I think
that is OK.
common. Add one do-nothing element to each set. This ensures that
the linker realizes that they are linker sets rather than simple
commons, and makes it possible to link c++rt0.o into every shared
library regardless of whether it is a C++ library or not. Without
this change, the constructors and destructors in the main program
could be executed multiple times.
This change is going to make it possible to get rid of the
CPLUSPLUSLIB makefile variable once and for all. It is a piece of
the solution to PR gnu/3505 (gcc -shared). Finally, it fixes a
heretofore unreported bug: If CPLUSPLUSLIB was set in a makefile
for a C++ shared library that had no static constructors or
destructors in it, then the main program's constructors and
destructors would be executed multiple times.
This will make a number of things easier in the future, as well as (finally!)
avoiding the Id-smashing problem which has plagued developers for so long.
Boy, I'm glad we're not using sup anymore. This update would have been
insane otherwise.
Add descriptions of RTLD_LAZY and RTLD_NOW.
Correct the synopsis to agree with the actual function prototypes.
Add clarifications of a few things.
Clean up the wording in a few places.
was apparently overlooked at the time the member was added. Its absence
causes some error messages from the dynamic linker to begin with
"(null):" instead of with the pathname of the dynamic linker as they
should.
I am also adding a work-around to the dynamic linker, to cope with
legacy binaries that were built with older versions of crt0.
just return errors. This removes the need for awful hacks like that in
our build of libtcl which would get link errors when linked static.
John Polstra once mentioned that this was on his "todo" list.
Note that one can use:
cc -Wl,-Bstatic -o foo foo.o
and get an executable that has it's libraries statically linked, but has
a fully functional runtime linker so the executable can call dlopen() and
have it work. (I've tested this)
