| Age | Commit message (Collapse) | Author |
|---|
|
instruction in unoptimized builds and thumb: thanks Paul Brook
|
|
The patch below fixes a bug in the new ARM _ld_linux_resolve implementation.
I'm don't know if/how the current implementation was tested, but it's
completely broken.
This patch makes the prologue and epilogue agree on how big the stack frame
is, and also makes sure EABI doubleword stack alignment is preserved.
Tested on arm-linux-gnueabi.
|
|
called to resolve DLL PLT entries. The assembler is changed to be thumb
compatible and slightly faster, the C function, _dl_linux_resolver (note
the extra r) is changed to take a byte address in place of an 8 byte
count (faster in caller and callee, and slightly easier to understand).
|
|
Put the call_via_rx code into each executable - call_via_ip cannot
possibly work if called through the PLT! ldso requires this code
too as it is not linked with the crt stuff and thumb ldso does
make calls via a register.
The patch puts the code into crti.S so that it is linked into
every normally built application (if thumb or interworking is
selected). This is only 30 extra bytes and it works - the previous
code did not because nothing both implemented and exported the
APIs (they were in libgcc, but not in the version script).
crti.S and crtn.S is also brought up to date with GCC 3.4.4 - this
is essential for thumb support because the .init and .fini sections
must use arm or thumb code to match the compilation of the libraries.
Note that code which pushes stuff into .init or .fini must be
compiled with or without -mthumb to match the uclibc compilation -
and gcc itself (which does do this) must therefore be compiled to
match.
|
|
|
|
pseudo register for PIC base so load and relocate it.
|
|
|
|
create some default macros for do_rem/do_div_10 so we dont duplicate the samething in many arch header files
|
|
|
|
|
|
This patch changes all cases where the ARM assembler mov pc,rx
instructions are used to ensure that the thumb/arm interwork change of
process more works - in essence mov pc,rx needs to become bx rc.
The ldr pc or ldm rx, {pc} instructions are not changed - this is
fine on ARM >=v5 but will fail to restore thumb mode on ARM v4T,
i.e. this code will not provide support for thumb on ARM v4T.
One mov pc is left in resolve.S, this is fixed in a different patch -
thumb-resolve.patch
The changes are protected by __THUMB_INTERWORK__ - the original
mov instruction will work on newer architectures and is required on
arch v4 (not v4t) and earlier - those which did not support thumb -
so this is safe. See gcc lib1asmfuncs for a more exact test.
|
|
|
|
the default
|
|
|
|
|
|
syscall.h should not have it's own version of __set_errno, instead syscall.h should
include errno.h(see powerpc) since errno handling depends on threads/no threads.
I leave the the fixing of syscall.h since there might be some smaller compile problems.
|
|
|
|
|
|
where possible. This will also make ldso smaller.
However the patch touches all archs and I have only tested PPC and x86.
|
|
execution of ldso.
Added new asm for MIPS to be tested.
All arches should retest.
|
|
|
|
|
|
Add elf_machine_dynamic() and elf_machine_load_address() for
all archs. elf_machine_dynamic() replaces the #ifdef mess to
get at the GOT. elf_machine_load_address() is needed to execute
ldso directly, this is not complete yet.
I probably broke one or two archs(only tested PPC) so please
try and report problems. For a report to be useful you need
to enable __SUPPORT_LD_DEBUG_EARLY__ and __SUPPORT_LD_DEBUG__
|
|
|
|
Andrew de Quincey, who has been most helpful getting this sorted
out, thanks. Thanks also to Peter Mazinger who did alot of testing.
Removed all traces of dl_parse_copy_information() since it is no longer used.
|
|
|
|
Move COPY relocs back to _dl_do_reloc().
|
|
Remove unused function parameter in ldso.
Remove echo option '-e' since it is not supported on Solaris.
|
|
is needed if ldso should use libcs malloc whenever possible.
Fix RTLD_LAZY propagation to RTLD_NOW relocation when requested by
libdl.
|
|
|
|
Hi yet again :)
in dl-startup.c when performing boot strap relocation the following test
exists to make sure that only "_dl_" symbols are relocated:
/* We only do a partial dynamic linking right now. The user
is not supposed to define any symbols that start with a
'_dl', so we can do this with confidence. */
if (!symname || !_dl_symbol(symname)) {
continue;
}
However on PPC(and the other archs as well I suspect) all symbols are
"_dl_" symbols so the test is never true. The test can be removed and the
whole loop simplified(smaller). This also makes it possible to
simplify elfinterp.c
This remove the scanning of ldso.so relocs, making relocation faster.
I have tested this on PPC and it works well.
Do you think this optimization will work for the other arches as well?
I can't see why not.
Jocke
* Tested on x86, arm, mipsel, and powerpc by Erik and works nicely
-Erik
|
|
only occur when shared libs are not compiled with -fPIC
|
|
useful syscall failure diagnostics.
|
|
|
|
Begin converting some big ugly macros to inline functions
instead
|
|
hacks. Just check for the elf magic string one byte at a time....
|
|
|
|
|
|
Hi it is me again.
This is the latest ldso patch. the NEW weak symbol handling works now
with a little special handling in _dl_find_hash(). You get to chose
if you want the new or old handling :)
There was 2 missing _dl_check_if_named_library_is_loaded() calls in _dlopen().
I then disabled the _dl_check_if_named_library_is_loaded() in dl-elf.c since
it is rendundant.
Question, why does some _dl_linux_resolver(), like i386, have 2 calls
to _dl_find_hash()? I think that is wrong, isn't it?
I really hope you can check this out soon ...
|
|
|
|
to be arch specific to work properly.
|
|
to make it easier to treat it specially while not bothering the
rest of the code with the same constraints.
|
|
_dl_parse_lazy_relocation_information() and _dl_parse_copy_information()
so they are all consistant, allowing for future consolidation.
Trim some trailing whitespace as well.
|
|
|
|
For sh64 we need implicit access to the symtab, primarily to get at the
->st_other value. This presently isn't possible, as PERFORM_BOOTSTRAP_RELOC()
is invoked as such:
PERFORM_BOOTSTRAP_RELOC(rpnt, reloc_addr, symbol_addr, load_addr);
while we can easily get the symtab_index value from rpnt->r_info, this still
doesn't buy us easy access to the actual table. As such, I've modified
PERFORM_BOOTSTRAP_RELOC() to take an additional SYMTAB argument. Most
architectures aren't going to care about this, but unfortunately we don't
have any other options for sh64.
The following patch fixes up the API for what we need for sh64, and updates
the other architectures appropriately.
|
|
|
|
-Erik
|
|
|
|
Move all configuration options into the new config system.
-Erik
|
|
how uClibc handles _init and _fini, allowing shared lib constructors and
destructors to initialize things in the correct sequence. Stefan ported the SH
architecture. I then ported x86, arm, and mips. x86 and arm are working fine,
but I don't think I quite got things correct for mips.
|
