In the previous patch (r254003), I made the linker emit PT_GNU_STACK
unconditionally. But sometimes you want to have a control over the
presence of the segment. With this patch, you can omit the segment
by passing -z execstack option.
llvm-svn: 254039
In case a sysroot prefix is configured, and the filename starts with the
'/' character, and the script being processed was located inside the
sysroot prefix, the file's name will be looked for in the sysroot
prefix. Otherwise, the linker falls to the common lookup scheme.
https://www.sourceware.org/binutils/docs-2.24/ld/File-Commands.html
Differential Revision: http://reviews.llvm.org/D14916
llvm-svn: 254031
Partial (-z relro) and full (-z relro, -z now) relro cases are implemented.
Partial relro:
The ELF sections are reordered so that the ELF internal data sections (.got, .dtors, etc.) precede the program's data sections (.data and .bss).
.got is readonly, .got.plt is still writeable.
Full relro:
Supports all the features of partial RELRO, .got.plt is also readonly.
Differential revision: http://reviews.llvm.org/D14218
llvm-svn: 253967
R_X86_64_GOTTPOFF is not always requires GOT entries. Some relocations can be converted to local ones.
Differential revision: http://reviews.llvm.org/D14713
llvm-svn: 253966
With these relocations, it is now possible to build a simple "hello world"
program for AArch64 Debian.
Differential revision: http://reviews.llvm.org/D14917
llvm-svn: 253957
With this patch, lld creates PT_GNU_STACK segments only when all input
files have .note.GNU-stack sections. This is in line with other linkers
with a minor difference (we don't care about .note.GNU-stack rwx bits as
you can always remove .note.GNU-stack sections instead of setting x bit.)
At least, NetBSD loader does not understand PT_GNU_STACK segments and
reject any executables that have the section. This patch makes lld
compatible with such operating systems.
llvm-svn: 253797
This option is passed by clang driver if the target triple
is "aarch64-unknown-linux".
Differential Revision: http://reviews.llvm.org/D14831
llvm-svn: 253639
The content of reserved entries of the .got.plt section is target specific.
In particular, on x86_64 the zero entry holds the address of the .dynamic section,
but on AArch64 the same info is stored in the zero entry of the .got section.
Differential revision: http://reviews.llvm.org/D14703
llvm-svn: 253239
PT_GNU_STACK is a entry in the elf file format which contains the access rights (read, write, execute) of the stack,
it is always generated now. By default stack is not executable in this implementation.
-z execstack can be used to make executable.
Differential revision: http://reviews.llvm.org/D14571
llvm-svn: 253145
This sections can be protected with relro after resolving relocations by dynamic linker.
Differential revision: http://reviews.llvm.org/D14567
llvm-svn: 253018
The MIPS target requires specific dynamic section entries to be defined.
* DT_MIPS_RLD_VERSION and DT_MIPS_FLAGS store predefined values.
* DT_MIPS_BASE_ADDRESS holds base VA.
* DT_MIPS_LOCAL_GOTNO holds the number of local GOT entries.
* DT_MIPS_SYMTABNO holds the number of .dynsym entries.
* DT_MIPS_GOTSYM holds the index of the .dynsym entry
which corresponds to the first entry of the global part of GOT.
* DT_MIPS_RLD_MAP holds the address of the reserved space in the data segment.
* DT_MIPS_PLTGOT points to the .got.plt section if it exists.
* DT_PLTGOT holds the address of the GOT section.
See "Dynamic Section" in Chapter 5 in the following document for detailed
description: ftp://www.linux-mips.org/pub/linux/mips/doc/ABI/mipsabi.pdf
Differential revision: http://reviews.llvm.org/D14450
llvm-svn: 252857
The MIPS ABI has requirements to sort the entries in the .dyn.sym section.
Symbols which are not in the GOT have to precede the symbols which are added to
the GOT. The latter must have the same order as the corresponding GOT entries.
Since these sorting requirements contradict those of the GNU hash section,
they cannot be used together.
Differential revision: http://reviews.llvm.org/D14281
llvm-svn: 252854
This adds support for:
* Uniquing CIEs
* Dropping FDEs that point to dropped sections
It drops 657 488 bytes from the .eh_frame of a Release+Asserts clang.
The link time impact is smallish. Linking clang with a Release+Asserts
lld goes from 0.488064805 seconds to 0.504763060 seconds (1.034 X slower).
llvm-svn: 252790
GNU as can give it type SHT_PROGBITS or SHT_X86_64_UNWIND depending on
teh construct.
MC gives it type SHT_X86_64_UNWIND.
The linker has to canonicalize to one or the other so that there is only
one .eh_frame in the end.
llvm-svn: 252757
leaq symbol@tlsld(%rip), %rdi
call __tls_get_addr@plt
symbol@tlsld (R_X86_64_TLSLD) instructs the linker to generate a tls_index entry (two GOT slots) in the GOT for the entire module (shared object or executable) with an offset of 0. The symbol for this GOT entry doesn't matter (as long as it's either local to the module or null), and gold doesn't put a symbol in the dynamic R_X86_64_DTPMOD64 relocation for the GOT entry.
All other platforms defined in http://www.akkadia.org/drepper/tls.pdf except for Itanium use a similar model where global and local dynamic GOT entries take up 2 contiguous GOT slots, so we can handle this in a unified manner if we don't care about Itanium.
While scanning relocations we need to identify local dynamic relocations and generate a single tls_index entry in the GOT for the module and store the address of it somewhere so we can later statically resolve the offset for R_X86_64_TLSLD relocations. We also need to generate a R_X86_64_DTPMOD64 relocation in the RelaDyn relocation section.
This implementation is a bit hacky. It side steps the issue of GotSection and RelocationSection only handling SymbolBody entries by relying on a specific relocation type. The alternative to this seemed to be completely rewriting how GotSection and RelocationSection work, or using a different hacky signaling method.
llvm-svn: 252682
This is cleaner than computing relocations as if we had done it.
While at it, keep a single Phdr variable instead of multiple fields of it.
llvm-svn: 252352
This patch implements R_MIPS_GOT16 relocation for global symbols in order to
generate some entries in GOT. Only reserved and global entries are supported
for now. For the detailed description about GOT in MIPS, see "Global Offset
Table" in Chapter 5 in the followin document:
ftp://www.linux-mips.org/pub/linux/mips/doc/ABI/mipsabi.pdf
In addition, the platform specific symbol "_gp" is added, see "Global Data
Symbols" in Chapter 6 in the aforementioned document.
Differential revision: http://reviews.llvm.org/D14211
llvm-svn: 252275
For x86-64 the initial executable TLS block is placed directly before the
thread specific data register so compilers can directly access it via
R_X86_64_TPOFF32. Generate the correct (negative) offset for this case.
llvm-svn: 252131
This does not support TPOFF32 relocations to local symbols as the address calculations are separate. Support for this will be a separate patch.
llvm-svn: 251998
This is a case where there is inconsistency among ELF linkers:
* The spec says nothing special about empty sections.
* BFD ld removes them.
* Gold handles them like regular sections.
We were outputting them but sometimes ignoring them. This would create
odd looking outputs where a rw section could be in a ro segment for example.
The bfd way of doing things is also strange for the case where a symbol
points to the empty section.
Now we match gold and what seems to be the intention of the spec.
llvm-svn: 251988
It is required to fill up the GNU hash table section before its
finalize() method is called.
Differential Revision: http://reviews.llvm.org/D14196
llvm-svn: 251789
MachODefinedCustomSectionAtom.
The section names for these atoms are initialized from temporaries (e.g.
segName + "/" + sectName), so we can't use StringRef here.
llvm-svn: 251610
It is the GNU hash table section that should be reaponsible for storing its own
data and applying its requirements for the order to dynamic symbols.
Differential Revision: http://reviews.llvm.org/D14084
llvm-svn: 251502
GNU linkers accept both variants and at least for MIPS target gcc passes
joined variant of the '-m' option.
Differential Revision: http://reviews.llvm.org/D14133
llvm-svn: 251497
This matches ld.bfd and ld.gold behavior. The change is simple enough
and avoid trouble to consumers (they don't have to change their Makefiles).
Side note: found while trying to build FreeBSD base system with lld.
llvm-svn: 251408
There was a threading issue in the ICF code for COFF. That seems like
a venign bug in the sense that it doesn't produce an incorrect output,
but it oftentimes misses reducible sections. As a result, mergeable
sections could remain in outputs, which makes the output nondeterministic.
Basically the algorithm we are using for ICF is this: We group sections
so that identical sections will eventually be in the same group. Initially,
all sections are in one group. We split the group by relocation targets
until we get a convergence (if relocation targets are in different gruops,
the sections are different). Once a group is split, they will never be
merged.
Each section has a group ID. That variable itself is atomic, so there's
no threading issue at the level that we can use thread sanitizer.
The point is, when we split a group, we re-assign new group IDs to group
of sections. That are multiple separate writes to atomic varaibles.
Thus, splitting a group is not an atomic operation, and there's a small
chance that the other thread observes inconsistent group IDs.
Over-splitting is always "safe", so it will never create incorrect output.
I suspect that the nondeterminism stems from that point. However, I
cannot prove or fix that at this moment, so I'm going to avoid using
threads here.
llvm-svn: 251300
getFileOff functions defined for other classes return an offset
from beginning of the file. StringTableSection's getFileOff however
returned an offset from beginning of the section. That was confusing.
llvm-svn: 251192
This patch is an attempt to simplify assignAddresses function by splitting
it and using less variables. I tried to split the code to create PHDRs from
the code to assign addresses, but it didn't make this code simpler, so I
didn't do that in this patch.
llvm-svn: 251152
relocateOne is a function to apply a relocation. Previously, that
function took a pointer to Elf_Rel or Elf_Rela in addition to other
information that can be derived from the relocation entry. This patch
simplifies the parameter list. The new parameters, P or SA, are used
in the ELF spec to describe each relocation. These names make
relocateOne look like a mechanical, direct translation of the ELF spec.
llvm-svn: 251090
.eh_frame sections need to be preserved if they refer to live sections.
So the liveness relation is reverse for eh_frame sections. For now,
we simply preserve all .eh_frame sections. Thanks Rafael for pointing
this out. .jcr are kept for the same reason.
llvm-svn: 251068
Section garbage collection is a feature to remove unused sections
from outputs. Unused sections are sections that cannot be reachable
from known GC-root symbols or sections. Naturally the feature is
implemented as a mark-sweep garbage collector.
In this patch, I added Live bit to InputSectionBase. If and only
if Live bit is on, the section will be written to the output.
Starting from GC-root symbols or sections, a new function, markLive(),
visits all reachable sections and sets their Live bits. Writer then
ignores sections whose Live bit is off, so that such sections are
excluded from the output.
This change has small negative impact on performance if you use
the feature because making sections means more work. The time to
link Clang changes from 0.356s to 0.386s, or +8%.
It reduces Clang size from 57,764,984 bytes to 55,296,600 bytes.
That is 4.3% reduction.
http://reviews.llvm.org/D13950
llvm-svn: 251043
This patch implements --hash-style command line switch.
* By default, or with "sysv" or "both" parameters, the linker generates
a standard ELF hash section.
* With "gnu" or "both", it produces a GNU-style hash section.
That section requires the symbols in the dynamic symbol table section, which
are referenced in the GNU hash section, to be placed after not hashed ones and
to be sorted to correspond the order of hash buckets in the GNU Hash section.
The division function, as well as estimations for the section's parameters,
are just the first rough attempt and the subjects for further adjustments.
Differential Revision: http://reviews.llvm.org/D13815
llvm-svn: 251000
These classes are partially written, so almost all features
are FIXMEs. We do not want to add new FIXMEs to the classes
when we add new features to other non-stub classes.
llvm-svn: 250947
* Move the responsibility to call SymbolBody::setDynamicSymbolTableIndex()
from the hash table to the dynamic symbol table.
* Hash table is not longer responsible for filling the dynamic symbol table.
* The final order of symbols of both symbol tables is set before writing
phase starts.
* Remove repeaded scan of the symbol table during writting SymbolTableSection.
Differential Revision: http://reviews.llvm.org/D13911
llvm-svn: 250864
The section header table index of the entry that is associated with the section name string table.
Differential Revision: http://reviews.llvm.org/D13904
llvm-svn: 250836
Target has supportsLazyRelocations() method which can switch lazy relocations on/off (currently all targets are OFF except x64 which is ON). So no any other targets are affected now.
Differential Revision: http://reviews.llvm.org/D13856?id=37726
llvm-svn: 250808
The option now just sets NOW bit in DT_FLAGS_1 but some loaders
seem to require also BIND_NOW bit to be set in DT_FLAGS. This is,
also, what ld.bfd and gold do.
Differential Revision: http://reviews.llvm.org/D13883
llvm-svn: 250799
The two names are similar enough that they might lead to confusion.
The output of readobj clarifies but I missed it when I originally
committed this. Found while linking FreeBSD userland with lld.
llvm-svn: 250739
Given the name, it is natural for this function to compute the full target.
This will simplify SHF_MERGE handling by allowing getLocalRelTarget to
centralize the addend logic.
llvm-svn: 250731
The reason of collecting all undefines in vector is that during reading files we already need to have Symtab created. Or like was done in that patch - to put undefines from scripts somewhere to delay Symtab.addUndefinedOpt() call.
Differential Revision: http://reviews.llvm.org/D13870
llvm-svn: 250711
Hans Wennborg