This fixes unwinding in boundary cases on ARM with SEH.
In the case of ARM/Thumb, disp->ControlPc points at the following
instruction, with the thumb bit set. Thus by decrementing 1,
it still points at the next instruction. To achieve the desired
effect of pointing at the previous instruction, one first has to strip
out the thumb bit, then do the decrement by 1 to reach the previous
instruction.
When libcxxabi looks for call site ranges, it already does
`_Unwind_GetIP(context) - 1` (in `scan_eh_tab` in
libcxxabi/src/cxa_personality.cpp), so we shouldn't do the
corresponding `- 1` multiple times.
In the case of libcxxabi on Thumb, `funcStart` (still in `scan_eh_tab`)
may have the thumb bit set. If the program counter address is
decremented both in libunwind (first removing the thumb bit, then
decremented), and then libcxxabi decrements it further, and compares
with a `funcStart` with the thumb bit set, it could point to one byte
before the start of the call site.
Thus: This modification makes libunwind with SEH work with libcxxabi
on Thumb, in settings where libunwind and libcxxabi worked fine with
Dwarf before.
For existing cases with libunwind with SEH (on x86_64 and aarch64),
this modification doesn't break any of my testcases.
Differential Revision: https://reviews.llvm.org/D126869
There's no such corresponding code for ARM64 (which has been working
in production for years). The SEH version of the Unwind functions
(e.g. `_Unwind_GetLanguageSpecificData`) doesn't use these fields.
The `_Unwind_ForcedUnwind` function would need these bits though,
but that's not used in normal C++ exception unwinding.
Differential Revision: https://reviews.llvm.org/D126868
Check `__SEH__` when checking if ARM EHABI should be implied,
similarly to 4a3722a2c3dff1fe885cc38bf43d3c095c9851e7 / D126866.
Fix a warning by using the right format specifier (PRIxPTR instead
of PRIx64), and add a double->float cast in a codepath that hasn't
been built so far.
This is enough to make SEH unwinding of itanium ABI exceptions on
ARM mostly work - one specific issue is fixed in a separate follow-up
patch.
Differential Revision: https://reviews.llvm.org/D126867
We've observed segfaults in libunwind when attempting to check for the
Linux aarch64 sigreturn frame, presumably because of bad unwind info
leading to an incorrect PC that we attempt to read from. Use
process_vm_readv to read the memory safely instead.
The s390x code path should likely follow suit, but I don't have the
hardware to be able to test that, so I didn't modify it here either.
Reviewed By: MaskRay, rprichard, #libunwind
Differential Revision: https://reviews.llvm.org/D126343
Create a macro for this instead of duplicating the architecture checks
everywhere. (It's a little redundant to use it when we're checking for a
specific architecture, but I'm also applying it there for consistency.)
Reviewed By: rprichard, MaskRay, #libunwind
Differential Revision: https://reviews.llvm.org/D126342
Program may set the RA_SIGN_STATE pseudo register by expressions.
Libunwind expected only the DW_CFA_AARCH64_negate_ra_state could change the value
of the register which leads to runtime errors on PAC enabled systems.
In the recent version of the aadwarf64[1] a limitation is added[2] to forbid the mixing the
DW_CFA_AARCH64_negate_ra_state with other DWARF Register Rule Instructions.
[1] https://github.com/ARM-software/abi-aa/releases/tag/2022Q1
[2] https://github.com/ARM-software/abi-aa/pull/129
Reviewed By: #libunwind, MaskRay
Differential Revision: https://reviews.llvm.org/D123692
Reland: test moved because it depends on exceptions.
This is a variant of D116689 rebased on top of the new (proposed) ABI
refactoring in D120727. It should conserve the basic properties of the
original patch by @phosek, except it also allows cleaning up the merging
of libc++abi into libc++ from the libc++ side.
Differential Revision: https://reviews.llvm.org/D125393
Program may set the RA_SIGN_STATE pseudo register by expressions.
Libunwind expected only the DW_CFA_AARCH64_negate_ra_state could change the value
of the register which leads to runtime errors on PAC enabled systems.
In the recent version of the aadwarf64[1] a limitation is added[2] to forbid the mixing the
DW_CFA_AARCH64_negate_ra_state with other DWARF Register Rule Instructions.
[1] https://github.com/ARM-software/abi-aa/releases/tag/2022Q1
[2] https://github.com/ARM-software/abi-aa/pull/129
Reviewed By: #libunwind, MaskRay
Differential Revision: https://reviews.llvm.org/D123692
Unwinding out of signal handlers currently does not work since
the sigreturn trampoline is not annotated with CFI data.
Fix this by detecting the sigreturn trampoline during unwinding
and providing appropriate unwind data manually. This follows
closely the approach used by existing code for the AArch64 target.
Reviewed by: MaskRay
Differential Revision: https://reviews.llvm.org/D124765
Add support for the SystemZ (s390x) architecture to libunwind.
Support should be feature-complete with the exception of
unwinding from signal handlers (to be added later).
Reviewed by: MaskRay
Differential Revision: https://reviews.llvm.org/D124248
When the sigreturn trampoline is found the unw_proc_info_t.end_ip need to be set to
indicate a stack frame is found.
Reviewed By: cjdb, #libunwind, MaskRay
Differential Revision: https://reviews.llvm.org/D124522
Avoid repeating CMake checks across runtimes by unifying names of
variables used for results to leverage CMake caching.
Differential Revision: https://reviews.llvm.org/D110005
Summary:
This is an add-on patch to address comments.
- Replace #elif in file <assembly.h> with #else as suggested;
- Reversed the indentation changes in the main patch.
Differential Revision: https://reviews.llvm.org/D100132
Summary:
This patch contains the implementation of the unwinder for IBM AIX.
AIX does not support the eh_frame section. Instead, the traceback table located at the end of each function provides the information for stack unwinding and EH. In this patch macro _LIBUNWIND_SUPPORT_TBTAB_UNWIND is used to guard code for AIX traceback table based unwinding. Function getInfoFromTBTable() and stepWithTBTable() are added to get the EH information from the traceback table and to step up the stack respectively.
There are two kinds of LSDA information for EH on AIX, the state table and the range table. The state table is used by the previous version of the IBM XL compiler, i.e., xlC and xlclang++. The DWARF based range table is used by AIX clang++. The traceback table has flags to differentiate these cases. For the range table, relative addresses are calculated using a base of DW_EH_PE_datarel, which is the TOC base of the module where the function of the current frame belongs.
Two personality routines are employed to handle these two different LSDAs, __xlcxx_personality_v0() for the state table and __xlcxx_personality_v1() for the range table. Since the traceback table does not have the information of the personality for the state table approach, its personality __xlcxx_personality_v0() is dynamically resolved as the handler for the state table. For the range table, the locations of the LSDA and its associated personality routine are found in the traceback table.
Assembly code for 32- and 64-bit PowerPC in UnwindRegistersRestore.S and UnwindRegistersSave.S are modified so that it can be consumed by the GNU flavor assembler and the AIX assembler. The restoration of vector registers does not check VRSAVE on AIX because VRSAVE is not used in the AIX ABI.
Reviewed by: MaskRay, compnerd, cebowleratibm, sfertile, libunwind
Differential Revision: https://reviews.llvm.org/D100132
We should not assume that the cet.h header exists just because
we're on x86 linux. Only include it if __CET__ is defined. This
makes the code more similar to what compiler-rt does in
ee423d93ea/compiler-rt/lib/builtins/assembly.h (L17)
(though that one also has a __has_include() check -- I've not found
that to be necessary).
Differential Revision: https://reviews.llvm.org/D119697
The warning was introduced with the recently merged SPARCv9
support in 2b9554b8850192bdd86c02eb671de1d866df8d87.
The cast matches the existing surrounding cases.
Differential Revision: https://reviews.llvm.org/D119353
Add SPARC to the list of platforms for which we provide a full
unwind implementation which leads to _Unwind_Backtrace being defined within
libunwind.so.
Likewise for PPC (see D118320 for background).
Reviewed By: #libunwind, MaskRay, Arfrever
Differential Revision: https://reviews.llvm.org/D119068
Adds libunwind support for SPARCv9 (aka sparc64). This is a rebase of @kettenis' patch D32450, which I created (with his permission) because the original review has become inactive.
The changes are of a cosmetic nature to make it fit better with the new code style, and to reuse the existing SPARCv8 code, whenever possible.
Please let me know if I posted this on the wrong place. Also, the summary of the original review is reproduced below:
> This adds unwinder support for 64-bit SPARC (aka SPARCv9). The implementation was done on OpenBSD/sparc64, so it takes StackGhost into account:
>
> https://www.usenix.org/legacy/publications/library/proceedings/sec01/full_papers/frantzen/frantzen_html/index.html
>
> Since StackGhost xor's return addresses with a random cookie before storing them on the stack, the unwinder has to do some extra work to recover those. This is done by introducing a new kRegisterInCFADecrypt "location" type that is used to implement the DW_CFA_GNU_window_save opcode. That implementation is SPARC-specific, but should work for 32-bit SPARC as well. DW_CFA_GNU_window_save is only ever generated on SPARC as far as I know.
Co-authored-by: Mark Kettenis
Reviewed By: #libunwind, thesamesam, MaskRay, Arfrever
Differential Revision: https://reviews.llvm.org/D116857
Originally reported downstream in Gentoo: https://bugs.gentoo.org/832140
```
/var/tmp/portage/sys-libs/llvm-libunwind-13.0.0/work/libunwind/src/libunwind.cpp:77:3: error: #error Architecture not supported
77 | # error Architecture not supported
| ^~~~~
[...]
/var/tmp/portage/sys-libs/llvm-libunwind-13.0.0/work/libunwind/src/libunwind.cpp: In function ‘int __unw_init_local(unw_cursor_t*, unw_context_t*)’:
/var/tmp/portage/sys-libs/llvm-libunwind-13.0.0/work/libunwind/src/libunwind.cpp:80:57: error: ‘REGISTER_KIND’ was not declared in this scope
80 | new (reinterpret_cast<UnwindCursor<LocalAddressSpace, REGISTER_KIND> *>(cursor))
| ^~~~~~~~~~~~~
[...]
```
PPC is actually a supported architecture, but GCC (tested with 11.2.0)
on powerpc32 seems to only define: `__PPC__, _ARCH_PPC, __PPC,
__powerpc` and //not// `__ppc__`.
This instead uses `__powerpc__` which should be around on PPC32
and PPC64 (but we check it after PPC64, so it's fine).
Signed-off-by: Sam James <sam@gentoo.org>
Differential Revision: https://reviews.llvm.org/D118320
These two are not conceptually the same; the former is a pointer shoved
in an integer, the latter is an offset or length. On the architectures
supported by libunwind, these two have the same underlying type, namely
unsigned int on ILP32, unsigned long on LP64 and unsigned long long on
LLP64. However, on CHERI, and thus Arm's Morello, they are not the same,
as pointers are hardware capabilities that carry additional metadata
including bounds and permissions, which is preserved in uintptr_t but
not in size_t. Thus, fix all length variables to be of type size_t not
uintptr_t, as we have done downstream for a while in CHERI LLVM but did
not get round to upstreaming.
Note that dyld_unwind_sections is currently defined in Apple's headers
as genuinely using uintptr_t to represent lengths. This is a bad API and
should be fixed, which would be totally API and ABI compatible due to
size_t and uintptr_t being the same type on all supported Apple systems,
but our definition is left matching theirs until such a time as they fix
their bogus types.
This is intended to be an NFC change on all architectures supported by
LLVM upstream, only being a functional change for CHERI downstream in
CHERI LLVM.
This adds a CMake option (defaults to OFF to not be intrusive) to activate
2 new targets `install-unwind-headers` and `install-unwind-headers-stripped`.
So, for example:
cmake -S runtimes -B build -G Ninja \
-DLLVM_ENABLE_RUNTIMES='libunwind' \
-DLIBUNWIND_INSTALL_HEADERS=ON
And then, `ninja -C build install-unwind` would install headers in addition
to good ol' dylibs and archives, i.e., targets `install-unwind*` `DEPENDS`
on `install-unwind-headers*`. On the other hand,
`ninja -C build install-unwind-headers` gives you headers only.
Differential Revision: https://reviews.llvm.org/D115535
There's a lot of history behind this, so here's a summary:
1. I stopped forcing -fPIC when building the runtimes in 30f305efe279,
before the LLVM 9 release back in 2019.
2. Someone complained that libc++.a couldn't be used in shared libraries
built without -fPIC (http://llvm.org/PR43604) since the LLVM 9 release.
This had been caused by my removal of -fPIC when building libc++.a in (1).
3. I suggested two ways of fixing the issue, the first being to force
-fPIC back unconditionally (http://llvm.org/D104328), and the second
being to specify that option explicitly when building the LLVM release
(http://llvm.org/D104327). We converged on the first solution.
4. I landed D104328, which forced building the runtimes with -fPIC.
This was included in the LLVM 13.0 release.
5. People complained about that and requested that we be able to
customize this setting (basically we should have done the second
solution).
This patch makes it such that the LLVM release script will specifically
ask for building with -fPIC using CMAKE_POSITION_INDEPENDENT_CODE,
however by default the runtimes will not force that option onto users.
This patch has the unintended effect that Clang and the LLVM libraries
(not only the runtime ones like libc++) will also be built with -fPIC
in the release. It would be better if we could specify that -fPIC is to
be used only when building the runtimes, however this is left as a
future improvement. The release should probably be using a bootstrapping
build and passing those options to the stage that builds the runtimes
only, see https://reviews.llvm.org/D112748 for that change.
Differential Revision: https://reviews.llvm.org/D110261
This patch implements the following:
- Emit PACBTI-M build attributes in libunwind asm files
- Authenticate LR in DWARF32 using PACBTI
Use Armv8.1-M.Main PACBTI extension to authenticate the return address
(stored in the LR register) before moving it to the PC (IP) register.
The AUTG instruction is used with the candidate return address, the CFA,
and the authentication code that is retrieved from the saved
pseudo-register RA_AUTH_CODE.
- Authenticate LR in EHABI using PACBTI
Authenticate the contents of the LR register using Armv8.1-M.Main PACBTI
extension.
A new frame unwinding instruction is introduced (0xb4). This
instruction pops out of the stack the return address authentication
code, which is then used in conjunction with the SP and the next-to-be
instruction pointer to perform authentication.
This authentication code is popped into a new register,
UNW_ARM_PSEUDO_PAC, which is a pseudo-register.
This patch is part of a series that adds support for the PACBTI-M extension of
the Armv8.1-M architecture, as detailed here:
https://community.arm.com/arm-community-blogs/b/architectures-and-processors-blog/posts/armv8-1-m-pointer-authentication-and-branch-target-identification-extension
The PACBTI-M specification can be found in the Armv8-M Architecture Reference
Manual:
https://developer.arm.com/documentation/ddi0553/latest
The following people contributed to this patch:
- Momchil Velikov
- Victor Campos
- Ties Stuij
Reviewed By: #libunwind, danielkiss, mstorsjo
Differential Revision: https://reviews.llvm.org/D112430
When unwind step reaches the end of the stack that means the force unwind should notify the stop function.
This is not an error, it could mean just the thread is cleaned up completely.
Reviewed By: #libunwind, mstorsjo
Differential Revision: https://reviews.llvm.org/D109856
The libgcc runtime library provides __register_frame and
__deregister_frame functions, which can be used by dynamic code
generators to register an .eh_frame section, which contains one or
more Call Frame Information records, each consisting of a Common
Information Entry record followed by one or more Frame Description
Entry records. This libunwind library also provides __register_frame
and __deregister_frame functions, but they are aliases for
__unw_add_dynamic_fde and __unw_remove_dynamic_fde and thus can only
take a single FDE.
This patch adds __unw_add_dynamic_eh_frame_section and
__unw_remove_dynamic_eh_frame_section functions which explicitly use
the .eh_frame format. Clients such as the ORCv2 platform and runtime
can check for these functions and use them if unwinding is being
provided by libunwind, or fall back to __register_frame and
__deregister_frame if unwinding is provided by libgcc.
Reviewed By: lhames
Differential Revision: https://reviews.llvm.org/D111863
We've stopped doing it in libc++ for a while now because these names
would end up rotting as we move things around and copy/paste stuff.
This cleans up all the existing files so as to stop the spreading
as people copy-paste headers around.
If Clang is set up to link directly against libunwind (via the
--unwindlib option, or the corresponding builtin default option),
configuring libunwind will fail while bootstrapping (before the
initial libunwind is built), because every cmake test will
fail due to -lunwind not being found, and linking the shared library
will fail similarly.
Check if --unwindlib=none is supported, and add it in that case.
Using check_c_compiler_flag on its own doesn't work, because that only
adds the tested flag to the compilation command, and if -lunwind is
missing, the linking step would still fail - instead try adding it
to CMAKE_REQUIRED_FLAGS and restore the variable if it doesn't work.
This avoids having to pass --unwindlib=none while building libunwind.
Differential Revision: https://reviews.llvm.org/D112126
Fixes D110144.
registers.getFloatRegister is not const in ARM therefor can't be called here.
Reviewed By: mstorsjo, #libunwind
Differential Revision: https://reviews.llvm.org/D110731
During a backtrace the `.cfi_undefined` for a float register causes an assert in libunwind.
Reviewed By: MaskRay
Differential Revision: https://reviews.llvm.org/D110144
Control-flow Enforcement Technology (CET), published by Intel,
introduces shadow stack feature aiming to ensure a return from
a function is directed to where the function was called.
In a CET enabled system, each function call will push return
address into normal stack and shadow stack, when the function
returns, the address stored in shadow stack will be popped and
compared with the return address, program will fail if the 2
addresses don't match.
In exception handling, the control flow may skip some stack frames
and we must adjust shadow stack to avoid violating CET restriction.
In order to achieve this, we count the number of stack frames skipped
and adjust shadow stack by this number before jumping to landing pad.
Reviewed By: hjl.tools, compnerd, MaskRay
Differential Revision: https://reviews.llvm.org/D105968
Signed-off-by: gejin <ge.jin@intel.com>
The original libunwind project defines UNW_AARCH64_* instead of UNW_ARM64_*.
Rename the enum members to match. This allows some applications with simple
`unw_init_local` usage to migrate to llvm-project libunwind.
Note: the canonical names of `UNW_ARM_D{0..31}` are now `UNW_AARCH64_V{0..31}`,
to match the original libunwind.
UNW_ARM64_* are kept for now for compatibility. Some may be unneeded and can be
cleaned up in the future.
Reviewed By: #libunwind, compnerd
Differential Revision: https://reviews.llvm.org/D107996
In some binaries, built with clang/lld, libunwind crashes
with "unsupported x86_64 register" for regNum == 16:
Differential Revision: https://reviews.llvm.org/D107919
-Wunused-but-set-variable triggers a warning even the block of code is effectively dead.
Reviewed By: MaskRay
Differential Revision: https://reviews.llvm.org/D107835
_Unwind_ForcedUnwind is not mandated by the EHABI but for compatibilty
reasons adding so the interface to higher layers would be the same.
Dropping EHABI specific _Unwind_Stop_Fn definition since it is not defined by EHABI.
Reviewed By: MaskRay
Differential Revision: https://reviews.llvm.org/D89570
