.plt and .branch_lt have the type of SHT_NOBITS and may be relocated by dynamic
relocations with non-zero addends. They should be skipped for the
--check-dynamic-relocations check, as --apply-dynamic-relocs does not apply.
A side effect is that -z rel does not work for the two sections.
Added two --apply-dynamic-relocs --check-dynamic-relocations tests. Also checked
linking a PPC64 clang.
The import library thunk name suffix uses the stem of the file. We
currently would attempt to trim the suffix by dropping the trailing 4
characters (under the assumption that the output name was `.lib`). This
now uses the `llvm::sys::path` API for computing the stem. This avoids
an assertion failure when the name is less the 4 characters and
assertions are enabled.
Close#57618: currently we align the end of PT_GNU_RELRO to a
common-page-size
boundary, but do not align the end of the associated PT_LOAD. This is
benign
when runtime_page_size >= common-page-size.
However, when runtime_page_size < common-page-size, it is possible that
`alignUp(end(PT_LOAD), page_size) < alignDown(end(PT_GNU_RELRO),
page_size)`.
In this case, rtld's mprotect call for PT_GNU_RELRO will apply to
unmapped
regions and lead to an error, e.g.
```
error while loading shared libraries: cannot apply additional memory protection after relocation: Cannot allocate memory
```
To fix the issue, add a padding section .relro_padding like mold, which
is contained in the PT_GNU_RELRO segment and the associated PT_LOAD
segment. The section also prevents strip from corrupting PT_LOAD program
headers.
.relro_padding has the largest `sortRank` among RELRO sections.
Therefore, it is naturally placed at the end of `PT_GNU_RELRO` segment
in the absence of `PHDRS`/`SECTIONS` commands.
In the presence of `SECTIONS` commands, we place .relro_padding
immediately before a symbol assignment using DATA_SEGMENT_RELRO_END (see
also https://reviews.llvm.org/D124656), if present.
DATA_SEGMENT_RELRO_END is changed to align to max-page-size instead of
common-page-size.
Some edge cases worth mentioning:
* ppc64-toc-addis-nop.s: when PHDRS is present, do not append
.relro_padding
* avoid-empty-program-headers.s: when the only RELRO section is .tbss,
it is not part of PT_LOAD segment, therefore we do not append
.relro_padding.
---
Close#65002: GNU ld from 2.39 onwards aligns the end of PT_GNU_RELRO to
a
max-page-size boundary (https://sourceware.org/PR28824) so that the last
page is
protected even if runtime_page_size > common-page-size.
In my opinion, losing protection for the last page when the runtime page
size is
larger than common-page-size is not really an issue. Double mapping a
page of up
to max-common-page for the protection could cause undesired VM waste.
Internally
we had users complaining about 2MiB max-page-size applying to shared
objects.
Therefore, the end of .relro_padding is padded to a common-page-size
boundary. Users who are really anxious can set common-page-size to match
their runtime page size.
---
17 tests need updating as there are lots of change detectors.
Fix#64600: the currently implementation is minimal (see
https://reviews.llvm.org/D83758), and an assignment like
`__TEXT_REGION_ORIGIN__ = DEFINED(__TEXT_REGION_ORIGIN__) ? __TEXT_REGION_ORIGIN__ : 0;`
(used by avr-ld[1]) leads to a value of zero (default value in `declareSymbol`),
which is unexpected.
Assign orders to symbol assignments and references so that
for a script-defined symbol, the `DEFINED` results match users'
expectation. I am unclear about GNU ld's exact behavior, but this hopefully
matches its behavior in the majority of cases.
[1]: https://sourceware.org/git/?p=binutils-gdb.git;a=blob;f=ld/scripttempl/avr.sc
When there are multiple catch-all patterns (i.e. a single `*`), GNU ld and gold
select the last pattern. Match their behavior.
This change was inspired by a correction made by Michael Kerrisk to a blog post
I wrote at https://maskray.me/blog/2020-11-26-all-about-symbol-versioning ,
following the current lld rules.
Note: GNU ld prefers global: patterns to local: patterns, which might seem
awkward (https://www.airs.com/blog/archives/300). gold doesn't follow this
behavior, and we do not either.
Linker relaxation may change relocations (offsets and types). However,
when --emit-relocs is used, relocations are simply copied from the input
section causing a mismatch with the corresponding (relaxed) code
section.
This patch fixes this as follows: for non-relocatable RISC-V binaries,
`InputSection::copyRelocations` reads relocations from the relocated
section's `relocations` array (since this gets updated by the relaxation
code). For all other cases, relocations are read from the input section
directly as before.
In order to reuse as much code as possible, and to keep the diff small,
the original `InputSection::copyRelocations` is changed to accept the
relocations as a range of `Relocation` objects. This means that, in the
general case when reading from the input section, raw relocations need
to be converted to `Relocation`s first, which introduces quite a bit of
boiler plate. It also means there's a slight code size increase due to
the extra instantiations of `copyRelocations` (for both range types).
Reviewed By: MaskRay
Differential Revision: https://reviews.llvm.org/D159082
Details:
We often use wildcard symbols in the exported_symbols list, and sometimes they match autohide symbols, which triggers these "cannot export hidden symbols" warnings that can be a bit noisy.
It'd be more user-friendly if LLD could truncate these.
Differential Revision: https://reviews.llvm.org/D159095
Issue Details:
When building up line information for CodeView debug info, LLVM attempts to gather the "range" of instructions within a function as these are printed together in a single record. If there is an inlined function, then those lines are attributed to the original function to enable generating `S_INLINESITE` records. However, this thus requires there to be instructions from the inlining function after the inlined function otherwise the instruction range would not include the inlined function.
Fix Details:
Include any inlined functions when finding the extent of a function in `getFunctionLineEntries`
Reviewed By: rnk
Differential Revision: https://reviews.llvm.org/D159226
The module paths string table mapped to both an id sequentially assigned
during LTO linking, and the module hash. The former is leftover from
before the module hash was added for caching and subsequently replaced
use of the module id when renaming promoted symbols (to avoid affects
due to link order changes). The sequentially assigned module id was not
removed, however, as it was still a convenience when serializing to/from
bitcode and assembly.
This patch removes the module id from this table, since it isn't
strictly needed and can lead to confusion on when it is appropriate to
use (e.g. see fix in D156525). It also takes a (likely not significant)
amount of overhead. Where an integer module id is needed (e.g. bitcode
writing), one is assigned on the fly.
There are a couple of test changes since the paths are now sorted
alphanumerically when assigning ids on the fly during assembly writing,
in order to ensure deterministic behavior.
Differential Revision: https://reviews.llvm.org/D156730
D153645 added additional X-Form load/stores that can be generated for TLS accesses.
However, these added instructions have not been accounted for in lld. As a result,
lld does not know how to handle them and cannot relax initial-exec to local-exec
when the initial-exec sequence contains these additional load/stores.
This patch aims to resolve https://github.com/llvm/llvm-project/issues/64424.
Differential Revision: https://reviews.llvm.org/D158197
Many of these options can be passed to the linker by the Clang
driver based on other options passed to Clang, after
a23bf1786b. Before commit
5c92c9f34a, these were ignored by
lld, but now we're erroring out on the unrecognized options.
The ELF linker has even more LTO options available, but not
all of these are currently settable via the lld-link option
interface, and some aren't set automatically by Clang but only
if the user manually passes them - and thus probably aren't in
wide use so far. (Previously LLD/MinGW would have accepted them
silently but ignored them though.)
Differential Revision: https://reviews.llvm.org/D158887
LLD resolves symbols before performing LTO compilation, assuming that the symbols in question are resolved by the resulting object files from LTO. However, there is a scenario where the prevailing symbols might be resolved incorrectly due to specific assembly symbols not appearing in the symbol table of the bitcode. This patch deals with such a scenario by generating an error instead of silently allowing a mis-linkage.
If a prevailing symbol is resolved through post-loaded archives via LC linker options, a warning will now be issued.
Reviewed By: #lld-macho, thevinster
Differential Revision: https://reviews.llvm.org/D158003
This adds a new -Bsymbolic option that directly binds all non-weak
symbols. There's a couple of reasons motivating this:
* The new flag will match the default behavior on Mach-O, so you can get
consistent behavior across platforms.
* We have use cases for which making weak data preemptible is useful,
but we don't want to pessimize access to non-weak data. (For a large
internal app, we measured 2000+ data symbols whose accesses would be
unnecessarily pessimized by `-Bsymbolic-functions`.)
Reviewed By: MaskRay
Differential Revision: https://reviews.llvm.org/D158322
This patch adds support to lld for --fat-lto-objects. We add a new
--fat-lto-objects option to LLD, and slightly change how it chooses input
files in the driver when the option is set.
Fat LTO objects contain both LTO compatible IR, as well as generated object
code. This allows users to defer the choice of whether to use LTO or not to
link-time. This is a feature available in GCC for some time, and makes the
existing -ffat-lto-objects option functional in the same way as GCC's.
If the --fat-lto-objects option is passed to LLD and the input files are fat
object files, then the linker will chose the LTO compatible bitcode sections
embedded within the fat object and link them together using LTO. Otherwise,
standard object file linking is done using the assembly section in the object
files.
The previous version of this patch had a missing `REQUIRES: x86` line in
`fatlto.invalid.s`. Additionally, it was reported that this patch caused
a test failure in `export-dynamic-symbols.s`, however,
29112a9946 disabled the
`export-dynamic-symbols.s` test on Windows due to a quotation difference
between platforms, unrelated to this patch.
Original RFC: https://discourse.llvm.org/t/rfc-ffat-lto-objects-support/63977
Reviewed By: MaskRay
Differential Revision: https://reviews.llvm.org/D146778
This change writes the module name to the name section of the wasm
binary. We use the `-soname` argument to determine the name and we
default the output file basename if this option is not specified.
In the future we will likely want to embed the soname in the dylink
section too, but this the first step in supporting `-soname`.
Differential Revision: https://reviews.llvm.org/D158001
LLD resolves symbols regardless of LTO modes early when reading and parsing input files in order. The object files built from LTO passes are appended later.
Because LLD eagerly resolves the LC linker options while parsing a new object file (and its chain of dependent libraries), the prior decision on pending prevailing symbols (belonging to some bitcode files) can change to ones in those native libraries that are just loaded.
This patch delays processing LC linker options until all the native object files are added after LTO is done, similar to LD64. This way we preserve the decision on prevailing symbols LLD made, regardless of LTO modes.
- When parsing a new object file in `parseLinkerOptions()`, it just parses LC linker options in the header, and saves those contents to `unprocessedLCLinkerOptions`.
- After LTO is finished, `resolveLCLinkerOptions()` is called to recursively load dependent libraries, starting with initial linker options collected in `unprocessedLCLinkerOptions` (which also updates during recursions)
Reviewed By: #lld-macho, int3
Differential Revision: https://reviews.llvm.org/D157716
When stub libraries trigger the fetching of new object files we can
potentially introduce new undefined symbols so process the stub in
loop until no new objects are pulled in.
Differential Revision: https://reviews.llvm.org/D153466
Details:
calling getMemoryBufferRef() on an empty archive can trigger segfault so the code should check before calling this.
this seems like a bug in the Archive API but that can be fixed separately.
P.S: follow up to D156468
Differential Revision: https://reviews.llvm.org/D157300
Update this diagnostic to mention the reason (unmatched '['), matching
the other diagnostic about stray '\'. The original pattern is omitted,
as some users may mention the original pattern in another position, not
repeating it.
`R_LARCH_PCREL20_S2` is a new added relocation type in LoongArch ELF
psABI v2.10 [1] which is not corvered by D138135 except `R_LARCH_64_PCREL`.
A motivation to support `R_LARCH_PCREL20_S2` in lld is to build the
runtime of .NET core (a.k.a `CoreCLR`) in which strict PC-relative
semantics need to be guaranteed [2]. The normal `pcalau12i + addi.d`
approach doesn't work because the code will be copied to other places
with different "page" and offsets. To achieve this, we can use `pcaddi`
with explicit `R_LARCH_PCREL20_S2` reloc to address +-2MB PC-relative
range with 4-bytes aligned.
[1]: https://github.com/loongson/la-abi-specs/releases/tag/v2.10
[2]: https://github.com/dotnet/runtime/blob/release/7.0/src/coreclr/vm/loongarch64/asmhelpers.S#L307
Reviewed By: xen0n, MaskRay
Differential Revision: https://reviews.llvm.org/D156772
Two changes:
- Avoid crashing in predicate functions.
Querying the property of the Symbols via these is*() functions shouldn't crash the program - the answer should just be "false".
Currently, having them throw UNREACHABLE already (incorrectly) crash certain code paths involving macho::validateSymbolRelocation() .
- Simply ignore input archives with incompatible arch (changes from PRESIDENT810@)
Differential Revision: https://reviews.llvm.org/D156468
As per the ABI at
https://github.com/ARM-software/abi-aa/blob/main/memtagabielf64/memtagabielf64.rst,
this patch interprets the SHT_AARCH64_MEMTAG_GLOBALS_STATIC section,
which contains R_NONE relocations to tagged globals, and emits a
SHT_AARCH64_MEMTAG_GLOBALS_DYNAMIC section, with the correct
DT_AARCH64_MEMTAG_GLOBALS and DT_AARCH64_MEMTAG_GLOBALSSZ dynamic
entries. This section describes, in a uleb-encoded stream, global memory
ranges that should be tagged with MTE.
We are also out of bits to spare in the LLD Symbol class. As a result,
I've reused the 'needsTocRestore' bit, which is a PPC64 only feature.
Now, it's also used for 'isTagged' on AArch64.
An entry in SHT_AARCH64_MEMTAG_GLOBALS_STATIC is practically a guarantee
from an objfile that all references to the linked symbol are through the
GOT, and meet correct alignment requirements. As a result, we go through
all symbols and make sure that, for all symbols $SYM, all object files
that reference $SYM also have a SHT_AARCH64_MEMTAG_GLOBALS_STATIC entry
for $SYM. If this isn't the case, we demote the symbol to being
untagged. Symbols that are imported from other DSOs should always be
fine, as they're GOT-referenced (and thus the GOT entry either has the
correct tag or not, depending on whether it's tagged in the defining DSO
or not).
Additionally hand-tested by building {libc, libm, libc++, libm, and libnetd}
on Android with some experimental MTE globals support in the
linker/libc.
Reviewed By: MaskRay, peter.smith
Differential Revision: https://reviews.llvm.org/D152921
This is similar to D143540 for import libraries. ARM64EC will need it
for EC symbol table, but it should be fine for other targets as well
and it improves MSVC compatibility. I left mingw case unchanged to be
safe, although I think that it wouldn't hurt to change that as well.
The visible effect in tests is a sorted symbol map.
Revieved By: mstorsjo
Differential Revision: https://reviews.llvm.org/D156473
Mapping symbols [1] are special ELF symbols that can be inserted to
indicate regions of code or data. A sequence of data bytes is indicated
by a `$d` (or `$d.<any>`) symbol pointing to its start while a sequence
of instructions uses a `$x` (or `$x.<any>`) symbol. This can be used,
for example, to assist disassembling a memory region containing both
data and code.
This patch implements mapping symbols for RISC-V, copying the
implementation mostly from the AArch64 target.
Note that the `$x<ISA>` mapping symbol, indicating an instruction
sequence with a specific ISA extension, is not implemented by this
patch. As far as I can tell, binutils doesn't implement this yet either.
Note that this patch uses the same symbol naming convention as the
AArch64 target: always use `$x.i` and `$d.i` (where `i` is a
monotonically increasing counter). This differs from binutils where all
symbols are named `$x` or `$d` (causing multiple symbol having the same
name). I'm not not sure whether it makes more sense to avoid duplicate
symbol names or be consistent with binutils.
Note that the handling of nop-slides inserted for alignment differs from
binutils: binutils always marks the nops as instructions (`$x`) while
this patch doesn't insert a symbol for the nops (so the last inserted
symbol is used). I believe binutil's behavior makes most sense but this
seems difficult to implement in LLVM as the insertion of nops is handled
by `RISCVAsmBackend`. At this point, inserting ELF symbols seems
impossible. Any ideas for how to handle this would be appreciated.
[1]: https://github.com/riscv-non-isa/riscv-elf-psabi-doc/blob/master/riscv-elf.adoc#mapping-symbol
Depends on D156190 and D156236
Differential Revision: https://reviews.llvm.org/D153260
D108961 introduced relaxation for out-of-range conditional branches.
However, relaxation was only performed when the branch target could be
resolved. I believe this has two undesired consequences:
- `b<cc> ... foo`, where `foo` is undefined, would not be relaxed
although there is no guarantee the offset to `foo` will fit;
- Conditional branches are never relaxed with `-mattr=+relax` because MC
considers fixups where `shouldForceRelocation` returns true (which
will be the case with `+relax`) to be unresolved.
Note that binutils performs conditional branch relaxation in both cases.
This patch proposes to perform conditional branch relaxation even when
the target cannot be resolved.
Note on llvm/test/MC/RISCV/long-conditional-jump.s: I've removed the
`.p2align` because this causes alignment nops to be inserted for the
`+relax` tests. This in turn causes all the branch targets to change
compared to the non-`+relax` tests. Since `+relax` shouldn't change
these offsets, I found this confusing and hence chose to remove the
alignment.
Reviewed By: asb, MaskRay, reames
Differential Revision: https://reviews.llvm.org/D154958
Similar to D96617 for llvm-symbolizer.
This patch matches the GNU objdump -d behavior to suppress printing
labels for mapping symbols. Mapping symbol names don't convey much
information.
When --show-all-symbols (not in GNU) is specified, we still print
mapping symbols.
Note: the `for (size_t SI = 0, SE = Symbols.size(); SI != SE;)` loops
needs to iterate all mapping symbols, even if they are not displayed.
We use the new field `IsMappingSymbol` to recognize mapping symbols.
This field also enables simplification after D139131.
ELF/ARM/disassemble-all-mapping-symbols.s is enhanced to add `.space 2`.
If `End = std::min(End, Symbols[SI].Addr);` is not correctly set, we
would print a `.word`.
Reviewed By: jhenderson, jobnoorman, peter.smith
Differential Revision: https://reviews.llvm.org/D156190
Initially the .rodata section came before .text, hence sharing its
segment with the program header sitting at a small offset, pushing the
output file size to ~72GiB (the file was sparse though, so not much is
really written). This breaks on 32-bit platforms and is irrelevant to
the feature being tested, so re-order the two sections so .text gets
processed first, and both sections get their own segment.
This addresses the issue found by the clang-armv8-lld-2stage builder:
https://lab.llvm.org/buildbot/#/builders/178/builds/5340
Reviewed By: SixWeining, xry111
Differential Revision: https://reviews.llvm.org/D156293
This adds support for the LoongArch ELF psABI v2.00 [1] relocation
model to LLD. The deprecated stack-machine-based psABI v1 relocs are not
supported.
The code is tested by successfully bootstrapping a Gentoo/LoongArch
stage3, complete with common GNU userland tools and both the LLVM and
GNU toolchains (GNU toolchain is present only for building glibc,
LLVM+Clang+LLD are used for the rest). Large programs like QEMU are
tested to work as well.
[1]: https://loongson.github.io/LoongArch-Documentation/LoongArch-ELF-ABI-EN.html
Reviewed By: MaskRay, SixWeining
Differential Revision: https://reviews.llvm.org/D138135
There is a strange issue happening with command line processing though. The
command line argument
--export-dynamic-symbol 'f*'
does not have the single quotes stripped on some Windows targets (but not
all). This causes the glob matching to fail, which means the test fails on
some Windows bots and passes on others.
This is expected to be a temporary measure to get bots back to green. I've not
found a commit that has caused a behavioral change that could be reverted
instead, so this could be an issue with lit or test machine configuration.
Similar to recent changes to ELF (e.g., D154813), Mach-O, and COFF to
improve hashing performance.
Reviewed By: dschuff
Differential Revision: https://reviews.llvm.org/D155752
For a relocation, we don't differentiate the two cases:
* the symbol index is 0
* the symbol index is non zero, the type is not STT_SECTION, and the name is empty. Clang generates such local symbols for RISC-V linker relaxation.
So we may print
```
Offset Info Type Symbol's Value Symbol's Name + Addend
000000000000001c 0000000100000039 R_RISCV_32_PCREL 0000000000000000 0
// llvm-readobj
0x1C R_RISCV_32_PCREL - 0x0
```
while GNU readelf prints "<null>", which is clearer. Let's match the GNU behavior.
Related to https://reviews.llvm.org/D81842
```
000000000000001c 0000000100000039 R_RISCV_32_PCREL 0000000000000000 <null> + 0
// llvm-readobj
0x1C R_RISCV_32_PCREL <null> 0x0
```
Reviewed By: jhenderson, kito-cheng
Differential Revision: https://reviews.llvm.org/D155353
This reverts commit c9953d9891 and a
forward fix in 3a45b843de.
D14677 causes some failure on windows bots that the forward fix did not
address. Thus I'm reverting until the underlying cause can me triaged.
