The R_AARCH64_LDST<N>_ABS LO12_NC relocations where N is 8, 16, 32, 64 or
128 have a scaled immediate. For example R_AARCH64_LDST32_ABS_LO12_NC
shifts the calculated value right by 4. If the target symbol + relocation
addend is not aligned properly then bits of the answer will be lost.
This change adds an alignment check to the relocations to make sure the
target of the relocation is aligned properly. This matches the behavior of
GNU ld. The motivation is to catch ODR violations such as a declaration of
extern int foo, but a definition of bool foo as the compiler may use
R_AARCH64_LDST32_ABS_LO12_NC for the former, but not align the destination.
Differential Revision: https://reviews.llvm.org/D37444
llvm-svn: 312637
The fixSectionAlignments() function may alter the alignment of some
OutputSections, this is likely to alter the addresses calculated earlier
in assignAddresses(). By moving the call to fixSectionAlignments() we
make sure that assignAddresses() is consistent with the early calculation
used for RangeThunks and the final call just before writing the image.
Differential Revision: https://reviews.llvm.org/D36739
llvm-svn: 312636
The --symbol-ordering-file path was not being rewritten in the response file when
using --reproduce. This patch adds this to the list of switches that are rewritten,
so that the path is somewhere within the reproducer directory tree.
Reviewers: ruiu
Differential Revision: https://reviews.llvm.org/D37480
llvm-svn: 312626
Previously LLD did not calculate LMAOffset correctly when
AT and MEMORY were used together.
Patch fixes PR34407.
Differential revision: https://reviews.llvm.org/D37469
llvm-svn: 312625
Summary:
Previous would throw warning whenever libxml2 is not installed. Now
only give this warning if merging manifest fails.
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D37240
llvm-svn: 312604
Looks like raw_string_ostream is buffered. If we do not call `flush`
nor `str`, it is not guaranteed that a result string has all characters
that were written to it.
It wasn't failing on buildbots, but I could reproduce the issue on my
Windows workstation.
llvm-svn: 312577
std::vector::insert invalidates all iterators, so it was not safe to do
Script->Opt.Commands.insert(++I, Make(ElfSym::End1));
Script->Opt.Commands.insert(++I, Make(ElfSym::End2));
because after the first line, `I` is no longer valid.
This patch rewrites fixes the issue. I belive the new code without
higher-order functions is a bit more readable than before.
llvm-svn: 312570
Pass BSIZE and SHIFT as a function arguments to the `writeRelocation`
routine. It does not make a sense to have so many `writeRelocation's`
instances.
llvm-svn: 312495
Recently (before r312477) lib/DebugInfo incorrectly handled the case
when debug ranges uses base address of CU. Section index was not
calulcated properly and LLD crashed on provided testcase.
Differential revision: https://reviews.llvm.org/D37297
llvm-svn: 312478
Apply the simplification suggestions that Peter Collingbourne made
during the review at D37368. The returned thunk is cast to the
appropriate type in the SymbolTable, and the constant symbol's body is
not needed directly, so avoid the assignment. NFC
llvm-svn: 312391
If a symbol is locally defined and is DLL imported in another
translation unit, and the object with the locally defined version is
loaded prior to the imported version, then the linker will fail to
resolve the definition of the thunk and return the locally defined
symbol. This will then be attempted to be cast to an import thunk,
which will clearly fail.
Only return the thunk if the symbol is inserted or a thunk is created.
Otherwise, report a duplication error.
llvm-svn: 312386
We have llvm-readobj for dumping CodeView from object files, and
llvm-pdbutil has always been more focused on PDB. However,
llvm-pdbutil has a lot of useful options for summarizing debug
information in aggregate and presenting high level statistical
views. Furthermore, it's arguably better as a testing tool since
we don't have to write tests to conform to a state-machine like
structure where you match multiple lines in succession, each
depending on a previous match. llvm-pdbutil dumps much more
concisely, so it's possible to use single-line matches in many
cases where as with readobj tests you have to use multi-line
matches with an implicit state machine.
Because of this, I'm adding object file support to llvm-pdbutil.
In fact, this mirrors the cvdump tool from Microsoft, which also
supports both object files and pdb files. In the future we could
perhaps rename this tool llvm-cvutil.
In the meantime, this allows us to deep dive into object files
the same way we already can with PDB files.
llvm-svn: 312358
Previously it was called twice for .comment synthetic section.
That created 2 pieces of data, which was deduplicated anyways,
but was not clean.
llvm-svn: 312327
The problem with symbol assignments in implicit linker scripts is that
they can refer synthetic symbols such as _end, _etext or _edata. The
value of these symbols is currently fixed only after all linker script
commands are processed, so these assignments will be using non-final and
hence invalid value.
Rather than fixing the symbol values after all command processing have
finished, we instead change the logic to generate symbol assignment
commands that set the value of these symbols while processing the
commands, this ensures that the value is going to be correct by the time
any reference to these symbol is processed and is equivalent to defining
these symbols explicitly in linker script as BFD ld does.
Differential Revision: https://reviews.llvm.org/D36986
llvm-svn: 312305
writeArchive returned a pair, but the first element of the pair is always
its first argument on failure, so it doesn't make sense to return it from
the function. This patch change the return type so that it does't return it.
Differential Revision: https://reviews.llvm.org/D37313
llvm-svn: 312177
This reverts commit r312171 because it is pointed out that that's not a
correct fix (see https://bugs.llvm.org/show_bug.cgi?id=32674#c14) and
also because it broke buildbots.
llvm-svn: 312174
MSVC link.exe supports nested static libraries. That is, an .a file can
contain other .a file as its member. It is reported that MySQL actually
depends on this feature.
Fixes https://bugs.llvm.org/show_bug.cgi?id=32674
llvm-svn: 312171
Imagine we have archive file with symbols foo and bar.
And script that do foo = 1.
In that case correct behavior is not to fetch symbols from archive but
create absolute symbol.
If we have script zed = foo then symbol foo will be fetched from archive.
Currently we have Opt.ReferencedSymbols list of symbols referenced
by script and create them undefined early. That allows archives fetching logic
to work. That is what LLD already do and it seems everything is fine here.
But during writing D37059 I had to add left side of assignments to
Opt.ReferencedSymbols list because wanted to stop optimizing out
these symbols when LTO is involved (LTO also uses this list).
And testcase from this patch would have fail if we would have it before.
Differential revision: https://reviews.llvm.org/D37208
llvm-svn: 312004
This is PR32429.
We did not mention -fPIC in error about producing dynamic relocation
in readonly segment before. Patch changes that.
Differential revision: https://reviews.llvm.org/D36874
llvm-svn: 312003
Various classes have `Symtab` member variables even though we have
lld::coff::Symtab variable because previous attempts to make COFF lld's
internal structure resemble to ELF's was incomplete. This patch finishes
that job by removing member variables.
llvm-svn: 311938
Summary:
ArgParser created an instance of COFFOptTable on stack to use it to
parser command line arguments. Parsed arguments were then returned from
the function as InputArgList. This was safe because InputArgList referred
only statically-allocated InfoTable.
That is not a safe assumption after https://reviews.llvm.org/D36782,
which changes the type of its internal table from ArrayRef to std::vector.
To make lld work with that patch, we need to keep an instance of
COFFOptTable at least as long as an InputArgList is alive. This patch
does that.
Reviewers: yamaguchi
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D37217
llvm-svn: 311930
Patch by Patricio Villalobos.
I discovered that lld for darwin is generating the wrong code for lazy
bindings in the __stub_helper section (at least for osx 10.12). This is
the way i can reproduce this problem, using this program:
#include <stdio.h>
int main(int argc, char **argv) {
printf("C: printf!\n");
puts("C: puts!\n");
return 0;
}
Then I link it using i have tested it in 3.9, 4.0 and 4.1 versions:
$ clang -c hello.c
$ lld -flavor darwin hello.o -o h1 -lc
When i execute the binary h1 the system gives me the following error:
C: printf!
dyld: lazy symbol binding failed:
BIND_OPCODE_SET_SEGMENT_AND_OFFSET_ULEB
has segment 4 which is too large (0..3)
dyld: BIND_OPCODE_SET_SEGMENT_AND_OFFSET_ULEB has segment 4 which is too
large (0..3)
Trace/BPT trap: 5
Investigating the code, it seems that the problem is that the asm code
generated in the file StubPass.cpp, specifically in the line 323,when it
adds, what it seems an arbitrary number (12) to the offset into the lazy
bind opcodes section, but it should be calculated depending on the
MachONormalizedFileBinaryWrite::lazyBindingInfo result.
I confirmed this bug by patching the code manually in the binary and
writing the right offset in the asm code (__stub_helper).
This patch fixes the content of the atom that contains the assembly code
when the offset is known.
Differential Revision: https://reviews.llvm.org/D35387
llvm-svn: 311734
linker script SECTION rules. This patch extends it to use a fully specified
file name as it appears in --trace output to match agains, i.e,
"<path>/<objname>.o" or "<path>/<libname>.a(<objname>.o)".
Differential Revision: https://reviews.llvm.org/D37031
llvm-svn: 311713
Currently LLD reads the R_MIPS_HI16's addends in the `computeMipsAddend`
function, the R_MIPS_LO16's addends in both `computeMipsAddend` and
`getImplicitAddend` functions. This patch moves reading all addends to
the `getImplicitAddend` function. As a side effect it fixes a "paired"
HI16/LO16 addend calculation if "LO16" part of a pair is not found.
llvm-svn: 311711
Before this patch, lld printed out something like
error: -O: number expected, but got
After this patch, it prints out the same error message like this:
error: -O: number expected, but got ''
Fixes https://bugs.llvm.org/show_bug.cgi?id=34311
llvm-svn: 311681
Currently, LLD checks whether there's enough space for headers by
checking if headers fit below the address of the first allocated
section. However, that's always thue if the binary doesn't start
at zero which means that LLD always emits a segment for headers,
even if no other sections belong to that segment.
This is a problem in cases when linker script is being used with a
non-zero start address when we don't want to make the headers visible
by not leaving enough space for them. This pattern is common in
embedded programming but doesn't work in LLD.
This patch changes the behavior of LLD in case when linker script
is being to match the behavior of BFD ld and gold, which is to only
place headers into a segment when they're covered by some output
section.
Differential Revision: https://reviews.llvm.org/D36256
llvm-svn: 311586
Previously up to 3 errors were reported at once,
with patch we always will report only one,
just like in other linker code.
Differential revision: https://reviews.llvm.org/D37015
llvm-svn: 311537
It was broken from begining, because visibility
attributes were not applied properly to
symbols before this patch.
Differential revision: https://reviews.llvm.org/D36966
llvm-svn: 311536
Rafael Espindola