This reverts commit ccdebbae4d77d3efc236af92c22941de5d437e01.
Causes test failures in the presence of Android runtime libraries in resource-dir.
See comments on https://github.com/llvm/llvm-project/pull/87866.
Follow-up to #81037.
ToolChain::LibraryPaths holds the new compiler-rt library directory
(e.g. `/tmp/Debug/lib/clang/19/lib/x86_64-unknown-linux-gnu`). However,
it might be empty when the directory does not exist (due to the `if
(getVFS().exists(P))` change in https://reviews.llvm.org/D158475).
If neither the old/new compiler-rt library directories exists, we would
suggest the undesired old compiler-rt file name:
```
% /tmp/Debug/bin/clang++ a.cc -fsanitize=memory -o a
ld.lld: error: cannot open /tmp/Debug/lib/clang/19/lib/linux/libclang_rt.msan-x86_64.a: No such file or directory
clang++: error: linker command failed with exit code 1 (use -v to see invocation)
```
With this change, we will correctly suggest the new compiler-rt file name.
Fix#87150
Pull Request: https://github.com/llvm/llvm-project/pull/87866
This reverts commit bbc0f99f3bc96f1db16f649fc21dd18e5b0918f6
(https://reviews.llvm.org/D157663).
With this change, `-g` for the next major release 19.1 will generate
R_RISCV_SET_ULEB128/R_RISCV_SUB_ULEB128 relocations, which require
lld>=18 or binutils>=2.41.
binutils 2.41 is relatively new, but GCC has been producing
R_RISCV_SET_ULEB128/R_RISCV_SUB_ULEB128 for some time now.
BSD, Linux, and z/OS enable `LLVM_ENABLE_PER_TARGET_RUNTIME_DIR` by
default.
When a compiler-rt library is not found, we currently report an
incorrect filename `libclang_rt.XXX-$arch.a`
```
% /tmp/Debug/bin/clang++ a.cc -fsanitize=address -o a
ld.lld: error: cannot open /tmp/Debug/lib/clang/19/lib/linux/libclang_rt.asan-x86_64.a: No such file or directory
clang++: error: linker command failed with exit code 1 (use -v to see invocation)
```
With this change, we will correctly report:
```
% /tmp/Debug/bin/clang++ a.cc -fsanitize=address -o a
ld.lld: error: cannot open /tmp/Debug/lib/clang/19/lib/x86_64-unknown-linux-gnu/libclang_rt.asan.a: No such file or directory
clang++: error: linker command failed with exit code 1 (use -v to see invocation)
```
Link: https://discourse.llvm.org/t/runtime-directory-fallback/76860
The deduced "-march=" option always started with aarch64, which is not a
valid value. There was also no way to distinguish between armv8-r and
armv8-a. After this commit, the deduced "-march=" option will start with
greatest available "armv*-a" value or "armv8-r".
Installapi has important distinctions when compared to the clang driver,
so much that, it doesn't make much sense to try to integrate into it.
This patch partially reverts the CC1 action & driver support to replace
with its own driver as a clang tool.
For distribution, we could use `LLVM_TOOL_LLVM_DRIVER_BUILD` mechanism
for integrating the functionality into clang such that the toolchain
size is less impacted.
This introduces a basic outline of installapi as a clang driver option.
It captures relevant information as cc1 args, which are common arguments
already passed to the linker to encode into TBD file outputs. This is
effectively an upstream for what already exists as `tapi installapi` in
Xcode toolchains, but directly in Clang. This patch does not handle any
AST traversing on input yet.
InstallAPI is broadly an operation that takes a series of header files
that represent a single dynamic library and generates a TBD file out of
it which represents all the linkable symbols and necessary attributes
for statically linking in clients. It is the linkable object in all
Apple SDKs and when building dylibs in Xcode. `clang -installapi` also
will support verification where it compares all the information recorded
for the TBD files against the already built binary, to catch possible
mismatches like when a declaration is missing a definition for an
exported symbol.
This patch replaces uses of StringRef::{starts,ends}with with
StringRef::{starts,ends}_with for consistency with
std::{string,string_view}::{starts,ends}_with in C++20.
I'm planning to deprecate and eventually remove
StringRef::{starts,ends}with.
Android triples include a version number, which makes direct triple
comparisons for per-target runtime directory searching not always work.
Instead, look for the triple with the highest compatible version number
and use that per-target runtime directory instead. This maintains the
existing fallback to a triple without any version number, but I'm hoping
we can remove that in the future. https://discourse.llvm.org/t/62717
discusses this further.
The one remaining triple mismatch after this is that Android armv7
triples usually have an environment of `androideabi`, which Clang
normalizes to `android`. If you use the `androideabi` triple when
building the runtimes with a per-target runtimes dir, the directory will
get created with `androideabi` in its name, but Clang's triple search
uses the normalized triple and will look for an `android` directory
instead. https://reviews.llvm.org/D140925 will fix that by normalizing
triples when creating the per-target runtimes directories as well.
Reviewed By: phosek, pirama
Differential Revision: https://reviews.llvm.org/D158476
Searching for target-specific standard library header and library paths
should perform fallback searches for targets, the same way searching for
the runtime libraries does. It's important for the header and library
searches to be consistent, otherwise we could end up using mismatched
headers and libraries. (See also https://reviews.llvm.org/D146664.)
Reviewed By: phosek
Differential Revision: https://reviews.llvm.org/D159293
Update DeviceRTL and the AMDGPU plugin to support code
object version 5. Default is code object version 4.
CodeGen for __builtin_amdgpu_workgroup_size generates code
for cov4 as well as cov5 if -mcode-object-version=none
is specified. DeviceRTL compilation passes this argument
via Xclang option to generate abi-agnostic code.
Generated code for the above builtin uses a clang
control constant "llvm.amdgcn.abi.version" to branch on
the abi version, which is available during linking of
user's OpenMP code. Load of this constant gets eliminated
during linking.
AMDGPU plugin queries the ELF for code object version
and then prepares various implicitargs accordingly.
Differential Revision: https://reviews.llvm.org/D139730
Reviewed By: jhuber6, yaxunl
This used to be getRuntimePath till https://reviews.llvm.org/D115049
added a fallback search path for Android. As far as I can tell, the
intent has always been to use the first existing path though instead of
actually supporting multiple runtime paths. We can move the existence
checks into getRuntimePath and have it return std::optional, which also
makes the `--print-runtime-dir` behavior much cleaner.
The motivation is a follow-up change to Android runtime path searches,
which is much nicer with this in place.
Reviewed By: phosek, MaskRay
Differential Revision: https://reviews.llvm.org/D158475
This adds a RISC-V special case to ToolChain::GetDefaultDwarfVersion,
affecting Linux/Haiku/RISCVToolChain.
DWARF v5 .debug_loclists/.debug_rnglists's
DW_LLE_offset_pair/DW_RLE_offset_pair entry kinds utilitize `.uleb128 A-B`
directives where A and B reference local labels in code sections.
When A and B are separated by a RISC-V linker-relaxable instruction,
A-B is incorrectly folded without a relocation, causing incorrect debug
information.
```
void ext(void);
int foo(int x) {ext(); return 0;}
// DW_AT_location [DW_FORM_loclistx] of a DW_TAG_formal_parameter references a DW_LLE_offset_pair that can be incorrect after linker relaxation.
int ext(void);
void foo() { {
int ret = ext();
if (__builtin_expect(ret, 0))
ext();
} }
// DW_AT_ranges [DW_FORM_rnglistx] of a DW_TAG_lexical_block references a DW_RLE_offset_pair that can be incorrect after linker relaxation.
```
D157657 will implement R_RISCV_SET_ULEB128/R_RISCV_SUB_ULEB128
relocations, fixing the issue, but the relocation is only supported by
bleeding-edge binutils 2.41 and not by lld/ELF yet.
The goal is to make the emitted DWARF correct after linking.
Many users don't care about the default DWARF version, but a linker
error will be unacceptable. Let's just downgrade the default DWARF
version, before binutils>=2.41 is more widely available.
An alternative compatibility option is to add a toggle to DwarfDebug.cpp,
but that doesn't seem like a good idea.
Reviewed By: asb, kito-cheng
Differential Revision: https://reviews.llvm.org/D157663
In preparation for removing the `#include "llvm/ADT/StringExtras.h"`
from the header to source file of `llvm/Support/Error.h`, first add in
all the missing includes that were previously included transitively
through this header.
This is fixing all files missed in b0abd4893fa1.
Differential Revision: https://reviews.llvm.org/D154543
This will enable layering multilibs on top of each other.
For example a multilib containing only a no-exceptions libc++ could be
layered on top of a multilib containing C libs. This avoids the need
to duplicate the C library for every libc++ variant.
This change doesn't expose the functionality externally, it only opens
the functionality up to be potentially used by ToolChain classes.
Differential Revision: https://reviews.llvm.org/D143059
The default location for multilib.yaml is lib/clang-runtimes, without
any target-specific suffix. This will allow multilibs for different
architectures to share a common include directory.
To avoid breaking the arm-execute-only.c CHECK-NO-EXECUTE-ONLY-ASM
test, add a ForMultilib argument to getARMTargetFeatures.
Since the presence of multilib.yaml can change the exact location of a
library, relax the baremetal.cpp test.
Differential Revision: https://reviews.llvm.org/D142986
This option causes the flags used for selecting multilibs to be printed.
This is an experimental feature that is documented in detail in D143587.
Differential Revision: https://reviews.llvm.org/D142933
Change to system assembler to compile assembly files even
-fintegrated-as is specified. We don't have a good Clang as
for now for assembly files on AIX.
Reviewed By: qiucf
Differential Revision: https://reviews.llvm.org/D148490
Functions instrumented with -fsanitize=function have two words before
the function label: a signature and a RTTI proxy.
Instrumented call sites check the signature first to skip checks
for uninstrumented callees.
The code is generic and works for all targets supporting C++ RTTI.
Change clangDriver to allow all targets. Add tests for Armv8.5
Branch Target Identification and `-fpatchable-function-entry=`.
Reviewed By: peter.smith
Differential Revision: https://reviews.llvm.org/D148573
The compiler-rt library path can be either {resource_dir}/lib/{triple}
or {resource_dir}/lib/{OS}/{arch} depending on whether
LLVM_ENABLE_PER_TARGET_RUNTIME_DIR_default is ON or OFF.
Currently, the rpath added by -rtlib-add-rpath only adds
the latter. This patch checks both and adds the one that exists.
Reviewed by: Fangrui Song
Differential Revision: https://reviews.llvm.org/D146686
New option --dxv-path is added for dxc mode to set the installation path for dxv.
If cannot find dxv, a warning will be report.
dxv will be executed with command line dxv file_name -o file_name.
It will validate and sign the file and overwrite it.
Differential Revision: https://reviews.llvm.org/D141705
The orginal single folder layout produced libraries in the form:
lib/linux/<libname>-<archname>.a
That archname for Arm depended on whether you had hard or soft float.
This is sometimes shown as "arm" (soft) vs. "armhf" (hard).
If this is set in a triple we do it in the final portion, the ABI.
"gnueabi" for soft float, "gnueabihf" for hard float.
Meaning that the expected triple is:
arm-unknown-linux-gnueabihf
Not this:
armhf-unknown-linux-gnueabihf
For the per target layout I have decided to rely on the ABI portion
of the triple instead of the arch name used for the old layout
(doing that would produce the invalid triple above).
This means that building with triple:
armv8l-unknown-linux-gnueabihf
Will result in libraries in:
lib/arm-unknown-linux-gnueabihf/
And clang will now know to look for "arm" instead of "armv8l".
Meaning that we can share libraries between an armv8 and armv7 build
as we did with the previous layout. In addition to handling spelling
differences e.g. "armv8l" with an "l" on some Linux distros.
compiler-rt will autodetect that the "armhf" and/or "arm" architecture
can be built. We then replace the given triple's architecture with that.
Then if the triple's float ABI doesn't match, we change that. That new
triple is then used as the folder name.
If you select to build only the given triple, with COMPILER_RT_DEFAULT_TARGET_ONLY,
compiler-rt will not autodetect the architecture and for that I assume you
know what you're doing. In that case the library path will use the unomdified triple.
From what I can tell, this is how most large builds e.g Android and
Arm's Embedded Toolchain for llvm do things. I assume that big endian "armeb"
builds would end up doing this too.
Bare metal builds will not be using per target runtime dirs so they
remain as they were.
Depends on D139536
Reviewed By: MaskRay, phosek
Differential Revision: https://reviews.llvm.org/D140011
Clang currently supports the `-Xarch_host` and `-Xarch_device` variants
to handle passing arguments to only one part of the offloading
toolchain. This was previously only supported fully for HIP / CUDA This
patch simple updates the logic to make it work for any offloading kind.
Fixes#59799
Reviewed By: tianshilei1992
Differential Revision: https://reviews.llvm.org/D141935
This patch adds basic support for `--offload-arch=native` to CUDA. This
is done using the `nvptx-arch` tool that was introduced previously. Some
of the logic for handling executing these tools was factored into a
common helper as well. This patch does not add support for OpenMP or the
"new" driver. That will be done later.
Reviewed By: yaxunl
Differential Revision: https://reviews.llvm.org/D141051
A drive-by change in 53c98d85a8a609552448043d5512e70313b1eb1b made
-stdlib++-isystem be suppressed by -nostdinc and -nostdlibinc in
addition to -nostdinc++. However, that's contrary to the intent of the
flag. It's common to provide your own C++ headers (e.g. when building
libc++ by itself or as a compiler-rt dependency) but rely on the system
C headers, and having -stdlib++-isystem only look at -nostdinc++ allows
us to customize both the C header path (via -nostdinc or -nostdlibinc)
and the C++ header path (via -stdlib++-isystem) at the toolchain level
but still let users of the toolchain provide their own C++ headers. Add
a comment explaining the rationale to make it clearer.
The KCFI sanitizer emits "kcfi" operand bundles to indirect
call instructions, which the LLVM back-end lowers into an
architecture-specific type check with a known machine instruction
sequence. Currently, KCFI operand bundle lowering is supported only
on 64-bit X86 and AArch64 architectures.
As a lightweight forward-edge CFI implementation that doesn't
require LTO is also useful for non-Linux low-level targets on
other machine architectures, add a generic KCFI operand bundle
lowering pass that's only used when back-end lowering support is not
available and allows -fsanitize=kcfi to be enabled in Clang on all
architectures.
This relands commit eb2a57ebc7aaad551af30462097a9e06c96db925 with
fixes.
Reviewed By: nickdesaulniers, MaskRay
Differential Revision: https://reviews.llvm.org/D135411
This reverts commit eb2a57ebc7aaad551af30462097a9e06c96db925.
llvm/include/llvm/Transforms/Instrumentation/KCFI.h including
llvm/CodeGen is a layering violation. We should use an approach where
Instrumementation/ doesn't need to include CodeGen/.
Sorry for not spotting this in the review.
The KCFI sanitizer emits "kcfi" operand bundles to indirect
call instructions, which the LLVM back-end lowers into an
architecture-specific type check with a known machine instruction
sequence. Currently, KCFI operand bundle lowering is supported only
on 64-bit X86 and AArch64 architectures.
As a lightweight forward-edge CFI implementation that doesn't
require LTO is also useful for non-Linux low-level targets on
other machine architectures, add a generic KCFI operand bundle
lowering pass that's only used when back-end lowering support is not
available and allows -fsanitize=kcfi to be enabled in Clang on all
architectures.
Reviewed By: nickdesaulniers, MaskRay
Differential Revision: https://reviews.llvm.org/D135411
The gcc compatible driver has support for overriding the default
target based on the driver's executable name, for instance
x86_64-pc-linux-gnu-clang will set the default target to
x86_64-pc-linux-gnu.
Previously, this failed when the target contained a minor version, for
example x86_64-pc-freebsd13.1, so instead of finding the file's
stem, use the whole file name, but strip off any '.exe' from the tail.
Reviewed By: MaskRay
Differential Revision: https://reviews.llvm.org/D135284
D121868 provided support for -darwin-target-variant-triple, but the
support for -darwin-target-variant-sdk-version was still missing for
cc1as. These changes build upon the previous and provides such support.
- Extracted the common code to handle -darwin-target-variant-triple and
-darwin-target-variant-sdk-version in the Darwin toolchain to a method
that can be used for both the cc1 and the cc1as job construction.
cc1as does not support some of the parameters that were provided to
cc1, so the same code cannot be used for both.
- Invoke that new common code when constructing a cc1as invocation.
- Parse the new -darwin-target-variant-sdk-version in the cc1as driver.
Apply its value to the MCObjectFileInfo to generate the right values
in the object files.
- Includes two new tests that check that cc1as uses the provided values
in -darwin-target-variant-sdk and that the Clang driver creates the
jobs with the correct arguments.
Differential Revision: https://reviews.llvm.org/D135729
This allows using --ld-path= to set a custom linker path, while
still informing clang that the binary at that path is an lld built
at the same revision as clang, so that clang can make assumptions
about the flags it supports, its output format, etc.
This currently only has an observable effect on Darwin.
Differential Revision: https://reviews.llvm.org/D134063
Previously, we linked in the ROCm device libraries which provide math
and other utility functions late. This is not stricly correct as this
library contains several flags that are only set per-TU, such as fast
math or denormalization. This patch changes this to pass the bitcode
libraries per-TU using the same method we use for the CUDA libraries.
This has the advantage that we correctly propagate attributes making
this implementation more correct. Additionally, many annoying unused
functions were not being fully removed during LTO. This lead to
erroneous warning messages and remarks on unused functions.
I am not sure if not finding these libraries should be a hard error. let
me know if it should be demoted to a warning saying that some device
utilities will not work without them.
Reviewed By: JonChesterfield
Differential Revision: https://reviews.llvm.org/D133726
Recently OpenMP has transitioned to using the "new" driver which
primarily merges the device and host linking phases into a single
wrapper that handles both at the same time. This replaced a few tools
that were only used for OpenMP offloading, such as the
`clang-offload-wrapper` and `clang-nvlink-wrapper`. The new driver
carries some marked benefits compared to the old driver that is now
being deprecated. Things like device-side LTO, static library
support, and more compatible tooling. As such, we should be able to
completely deprecate the old driver, at least for OpenMP. The old driver
support will still exist for CUDA and HIP, although both of these can
currently be compiled on Linux with `--offload-new-driver` to use the new
method.
Note that this does not deprecate the `clang-offload-bundler`, although
it is unused by OpenMP now, it is still used by the HIP toolchain both
as their device binary format and object format.
When I proposed deprecating this code I heard some vendors voice
concernes about needing to update their code in their fork. They should
be able to just revert this commit if it lands.
Reviewed By: jdoerfert, MaskRay, ye-luo
Differential Revision: https://reviews.llvm.org/D130020
The KCFI sanitizer, enabled with `-fsanitize=kcfi`, implements a
forward-edge control flow integrity scheme for indirect calls. It
uses a !kcfi_type metadata node to attach a type identifier for each
function and injects verification code before indirect calls.
Unlike the current CFI schemes implemented in LLVM, KCFI does not
require LTO, does not alter function references to point to a jump
table, and never breaks function address equality. KCFI is intended
to be used in low-level code, such as operating system kernels,
where the existing schemes can cause undue complications because
of the aforementioned properties. However, unlike the existing
schemes, KCFI is limited to validating only function pointers and is
not compatible with executable-only memory.
KCFI does not provide runtime support, but always traps when a
type mismatch is encountered. Users of the scheme are expected
to handle the trap. With `-fsanitize=kcfi`, Clang emits a `kcfi`
operand bundle to indirect calls, and LLVM lowers this to a
known architecture-specific sequence of instructions for each
callsite to make runtime patching easier for users who require this
functionality.
A KCFI type identifier is a 32-bit constant produced by taking the
lower half of xxHash64 from a C++ mangled typename. If a program
contains indirect calls to assembly functions, they must be
manually annotated with the expected type identifiers to prevent
errors. To make this easier, Clang generates a weak SHN_ABS
`__kcfi_typeid_<function>` symbol for each address-taken function
declaration, which can be used to annotate functions in assembly
as long as at least one C translation unit linked into the program
takes the function address. For example on AArch64, we might have
the following code:
```
.c:
int f(void);
int (*p)(void) = f;
p();
.s:
.4byte __kcfi_typeid_f
.global f
f:
...
```
Note that X86 uses a different preamble format for compatibility
with Linux kernel tooling. See the comments in
`X86AsmPrinter::emitKCFITypeId` for details.
As users of KCFI may need to locate trap locations for binary
validation and error handling, LLVM can additionally emit the
locations of traps to a `.kcfi_traps` section.
Similarly to other sanitizers, KCFI checking can be disabled for a
function with a `no_sanitize("kcfi")` function attribute.
Relands 67504c95494ff05be2a613129110c9bcf17f6c13 with a fix for
32-bit builds.
Reviewed By: nickdesaulniers, kees, joaomoreira, MaskRay
Differential Revision: https://reviews.llvm.org/D119296
The KCFI sanitizer, enabled with `-fsanitize=kcfi`, implements a
forward-edge control flow integrity scheme for indirect calls. It
uses a !kcfi_type metadata node to attach a type identifier for each
function and injects verification code before indirect calls.
Unlike the current CFI schemes implemented in LLVM, KCFI does not
require LTO, does not alter function references to point to a jump
table, and never breaks function address equality. KCFI is intended
to be used in low-level code, such as operating system kernels,
where the existing schemes can cause undue complications because
of the aforementioned properties. However, unlike the existing
schemes, KCFI is limited to validating only function pointers and is
not compatible with executable-only memory.
KCFI does not provide runtime support, but always traps when a
type mismatch is encountered. Users of the scheme are expected
to handle the trap. With `-fsanitize=kcfi`, Clang emits a `kcfi`
operand bundle to indirect calls, and LLVM lowers this to a
known architecture-specific sequence of instructions for each
callsite to make runtime patching easier for users who require this
functionality.
A KCFI type identifier is a 32-bit constant produced by taking the
lower half of xxHash64 from a C++ mangled typename. If a program
contains indirect calls to assembly functions, they must be
manually annotated with the expected type identifiers to prevent
errors. To make this easier, Clang generates a weak SHN_ABS
`__kcfi_typeid_<function>` symbol for each address-taken function
declaration, which can be used to annotate functions in assembly
as long as at least one C translation unit linked into the program
takes the function address. For example on AArch64, we might have
the following code:
```
.c:
int f(void);
int (*p)(void) = f;
p();
.s:
.4byte __kcfi_typeid_f
.global f
f:
...
```
Note that X86 uses a different preamble format for compatibility
with Linux kernel tooling. See the comments in
`X86AsmPrinter::emitKCFITypeId` for details.
As users of KCFI may need to locate trap locations for binary
validation and error handling, LLVM can additionally emit the
locations of traps to a `.kcfi_traps` section.
Similarly to other sanitizers, KCFI checking can be disabled for a
function with a `no_sanitize("kcfi")` function attribute.
Reviewed By: nickdesaulniers, kees, joaomoreira, MaskRay
Differential Revision: https://reviews.llvm.org/D119296
Refactor baremetal driver code to reduce the bespoke
additions and base class overrides.
This lets us use the per target runtimes like other clang
targets. E.g. clang -target armv7m-cros-none-eabi will now
be able to use the runtimes installed at
<resource_dir>/lib/armv7m-cros-none-eabi instead of the hardcoded
path <resource_dir>/lib/baremetal.
The older code paths should still continue to work as before if
<resource_dir>/lib/<tuple> does not exist.
Reviewed By: MaskRay, barannikov88
Differential Revision: https://reviews.llvm.org/D131225
Based on the discussion at [1], this patch adds a Clang flag called
-fexperimental-library that controls whether experimental library
features are provided in libc++. In essence, it links against the
experimental static archive provided by libc++ and defines a feature
that can be picked up by libc++ to enable experimental features.
This ensures that users don't start depending on experimental
(and hence unstable) features unknowingly.
[1]: https://discourse.llvm.org/t/rfc-a-compiler-flag-to-enable-experimental-unstable-language-and-library-features
Differential Revision: https://reviews.llvm.org/D121141
Currently if `--sysroot /` is passed to the Clang driver, the include paths generated by the Clang driver will start with a double slash: `//usr/include/...`.
If VFS is used to inject files into the include paths (for example, the Swift compiler does this), VFS will get confused and the injected files won't be visible.
This change makes sure that the include paths start with a single slash.
Fixes#28283.
Differential Revision: https://reviews.llvm.org/D126289
