[cpullvm] Build Arm/AArch64 Linux libraries under the same framework as RISC-V #76
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
jonathonpenix
requested review from
Yethendra,
apazos,
navaneethshan and
zhaoshiz
jonathonpenix
commented
Jan 20, 2026
qualcomm-software/embedded/patches/musl-embedded/0001-Allow-out-of-tree-builds.patch
Outdated
Show resolved
Hide resolved
...ware/embedded/patches/musl-embedded/0001-Remove-Linux-libraries-from-component_list.sh.patch
Show resolved
Hide resolved
qualcomm-software/embedded/scripts/build_riscv_linux_runtimes.sh
Outdated
Show resolved
Hide resolved
qualcomm-software/embedded/scripts/build_riscv_linux_runtimes.sh
Outdated
Show resolved
Hide resolved
jonathonpenix
force-pushed
the
arm-linux-rework
branch
2 times, most recently
from
574aafe to
76857a5
Compare
jonathonpenix
changed the title
…as RISC-V
Arm/AArch64 now follow the same general pattern as RISC-V and are building
against our own tools (llvm tools, musl, etc.)
Note that this again isn't a 1:1 port of our old cpullvm Arm/AArch64
libraries (or what we used to do downstream) even besides the llvm/musl
changes:
* Baseline C flags used to build compiler-rt/musl/libc++ have changed
slightly--musl used to have slightly different variations, but these
have been made consistent across the projects.
* The subset of compiler-rt builtins we build for AArch64 Linux is now
slightly different--we build only the same variants as we build for
musl/libc++ (rather than sharing some of the same configs with
baremetal).
* Profile, sanitizer libraries are now building for Arm/AArch64 (they were
coincidentally disabled in cpullvm previously). There's caveats though,
see next point.
* We are disabling a few sanitizers/sanitizer-related projects in
compiler-rt due to either a) upstream issues between LLVM and musl or
b) our musl version underlying musl-embedded being particularly old. A
brief survey of the issues (see inline comments for details):
* sanitizer_common dependent projects have issues for Arm due to
disagreement about the size of time_t. I don't think this is worth
fixing, so just disable the projects.
* rtsan relies on fopencookie, which our musl-embedded is too old to
support. I don't think this is worth backporting, so disable this.
* gwp_asan relies on execinfo.h and musl (even upstream musl) does not
have this header or an equivalent. This is a known issue in upstream
LLVM. Just disable the project.
* Fix a bug (that also impacts RISC-V!) that meant the extra (non-builtins)
compiler-rt libraries weren't making it into the final install.
Signed-off-by: Jonathon Penix <jpenix@quicinc.com>
jonathonpenix
force-pushed
the
arm-linux-rework
branch
from
76857a5 to
24ff1e0
Compare
Contributor
Author
|
FWIW, test-specific flags were also removed as part of this--we weren't running those in cpullvm anyway and there's some work that will be needed to re-enable them. So, I just discarded those flags for now (and it is in our list of TODOs for the release) |
apazos
reviewed
Jan 21, 2026
| VARIANT_MUSL_CONFIGS["rv32imac_ilp32"]="" | ||
| VARIANT_MUSL_CONFIGS["rv64imac_lp64"]="" | ||
| VARIANT_MUSL_CONFIGS["rv64gc_lp64d"]="" | ||
| VARIANT_MUSL_CONFIGS["aarch64a"]="--quic-aarch64-optmem" |
Contributor
There was a problem hiding this comment.
Need clarification: if quic_aarch64_optmem is not specified, do we need to still set quic-aarch64-mark-bti?
| # * For rtsan, our musl-embedded is too old to support the fopencookie | ||
| # extension, which rtsan relies on. | ||
| # * For gwp_asan, it requires execinfo.h which is a glibc-specific | ||
| # header--no musl version ships this (or an equivalent). |
Contributor
There was a problem hiding this comment.
Double check whether we disabled rtsan in addition to gwp_asan in RISC-V build.
Contributor
Author
There was a problem hiding this comment.
Confirming rtsan isn't supported for riscv, so this should be ok--forgot about that earlier, sorry
jonathonpenix
commented
Jan 22, 2026
| pushd "${WORKSPACE}/musl-embedded" >/dev/null | ||
| git checkout "${MUSL_PINNED_COMMIT}" | ||
| popd >/dev/null | ||
| # FIXME: once the musl-embedded changes settle, we need to bump this accordingly |
Contributor
Author
There was a problem hiding this comment.
This needs to be removed and the musl-embedded commit updated, but I don't want to mess with that until support settles.
I'm just removing this until then so that test builds are unblocked.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Arm/AArch64 now follow the same general pattern as RISC-V and are building
against our own tools (llvm tools, musl, etc.)
Note that this again isn't a 1:1 port of our old cpullvm Arm/AArch64
libraries (or what we used to do downstream) even besides the llvm/musl
changes:
slightly--musl used to have slightly different variations, but these
have been made consistent across the projects.
slightly different--we build only the same variants as we build for
musl/libc++ (rather than sharing some of the same configs with
baremetal).
coincidentally disabled in cpullvm previously). There's caveats though,
see next point.
compiler-rt due to either a) upstream issues between LLVM and musl or
b) our musl version underlying musl-embedded being particularly old. A
brief survey of the issues (see inline comments for details):
disagreement about the size of time_t. I don't think this is worth
fixing, so just disable the projects.
support. I don't think this is worth backporting, so disable this.
have this header or an equivalent. This is a known issue in upstream
LLVM. Just disable the project.
compiler-rt libraries weren't making it into the final install.