Fix build issues for musl libc (#267) #268
Open
Conversation
This file contains 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
This fixes few issues, which allow to build and pass kfdtest on musl-based systems: 1) result of ROCm@daad183#diff-681b7d40f71f20573413d5072e1c381fde6c02aed03a56307afcbe82cf3f5e5eR36-R37f - `extern int hsakmt_page_shift` is under `ifndef`, while it should not ``` /var/tmp/portage/dev-libs/roct-thunk-interface-6.3.0/work/ROCR-Runtime-rocm-6.3.0/libhsakmt/src/openclose.c: In function 'init_page_size': /var/tmp/portage/dev-libs/roct-thunk-interface-6.3.0/work/ROCR-Runtime-rocm-6.3.0/libhsakmt/src/openclose.c:117:9: error: 'hsakmt_page_size' undeclared (first use in this function); did you mean 'hsakmt_page_shift'? 117 | hsakmt_page_size = sysconf(_SC_PAGESIZE); | ^~~~~~~~~~~~~~~~ | hsakmt_page_shift ``` 2) warning of `PAGE_SIZE` macro redefinition (defined previously in `limits.h`) 3) no operator to compare 0 with std::nullptr_t (null is defined as nullptr_t, following https://eel.is/c++draft/support.types.nullptr) ``` In file included from /var/tmp/portage/dev-libs/roct-thunk-interface-6.3.0/work/ROCR-Runtime-rocm-6.3.0/libhsakmt/tests/kfdtest/src/KFDTestUtil.hpp:27, from /var/tmp/portage/dev-libs/roct-thunk-interface-6.3.0/work/ROCR-Runtime-rocm-6.3.0/libhsakmt/tests/kfdtest/src/BaseQueue.hpp:28, from /var/tmp/portage/dev-libs/roct-thunk-interface-6.3.0/work/ROCR-Runtime-rocm-6.3.0/libhsakmt/tests/kfdtest/src/SDMAQueue.hpp:27, from /var/tmp/portage/dev-libs/roct-thunk-interface-6.3.0/work/ROCR-Runtime-rocm-6.3.0/libhsakmt/tests/kfdtest/src/KFDTestUtilQueue.cpp:28: /var/tmp/portage/dev-libs/roct-thunk-interface-6.3.0/work/ROCR-Runtime-rocm-6.3.0/libhsakmt/tests/kfdtest/gtest-1.6.0/gtest/gtest.h: In instantiation of 'testing::AssertionResult testing::internal::CmpHelperNE(const char*, const char*, const T1&, const T2&) [with T1 = long unsigned int; T2 = std::nullptr_t]': /var/tmp/portage/dev-libs/roct-thunk-interface-6.3.0/work/ROCR-Runtime-rocm-6.3.0/libhsakmt/tests/kfdtest/src/KFDTestUtilQueue.cpp:60:29: required from here 19072 | GTEST_ASSERT_(pred_format(#v1, #v2, v1, v2),\ | ^ /var/tmp/portage/dev-libs/roct-thunk-interface-6.3.0/work/ROCR-Runtime-rocm-6.3.0/libhsakmt/tests/kfdtest/gtest-1.6.0/gtest/gtest.h:18573:28: error: invalid operands of types 'const long unsigned int' and 'std::nullptr_t' to binary 'operator!=' 18573 | GTEST_IMPL_CMP_HELPER_(NE, !=); | ^ /var/tmp/portage/dev-libs/roct-thunk-interface-6.3.0/work/ROCR-Runtime-rocm-6.3.0/libhsakmt/tests/kfdtest/gtest-1.6.0/gtest/gtest.h:18558:12: note: in definition of macro 'GTEST_IMPL_CMP_HELPER_' 18558 | if (val1 op val2) {\ | ^~ ``` Closes: ROCm#267 Signed-off-by: Sv. Lockal <[email protected]>
This commit applies a similar change, that was reverted in 1cee865, with major issues fixed. 1) reverted commit checked if `pthread_attr_setaffinity_np` symbol exists, but this check did not work (and effectively disabled better path), because this symbol requires `#define _GNU_SOURCE` (see https://man7.org/linux/man-pages/man3/pthread_attr_setaffinity_np.3.html) 2) non-reverted part implied that `pthread_rwlockattr_setkind` function replaces `pthread_rwlockattr_setkind_np`, which is incorrect. This function is non-portable, serves as a hint and has no alternatives. Signed-off-by: Sv. Lockal <[email protected]>
The macro NULL sometimes is defined as 0, sometimes as nullptr_t-typed value (following [support.types.nullptr]), so to compare with pointers without warnings nullptr should be used
dayatsin-amd
requested review from
shwetagkhatri and
kentrussell
and removed request for
shwetagkhatri
kentrussell
approved these changes
Jan 22, 2025
kentrussell
reviewed
Jan 22, 2025
There was a problem hiding this comment.
This portion should be covered by 09d789c , right?
kentrussell
reviewed
Jan 22, 2025
There was a problem hiding this comment.
This block was intentional. If we have to pick one or the other, we want them to align. If PAGE_SIZE isn't defined, but PAGE_SHIFT=5000 , then setting just PAGE_SIZE=1<<12 won't really work.
|
!verify full |
|
!verify |
|
!verify |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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.
This fixes few issues, which allow to build and pass kfdtest on musl-based systems:
Issues with libhsakmt:
extern int hsakmt_page_shiftis underifndef, while it should not.warning of
PAGE_SIZEmacro redefinition (defined previously inlimits.h)no operator to compare 0 with std::nullptr_t (NULL is defined as nullptr_t-typed value, following support.types.nullptr)
Issues with hsa-runtime:
Attempt to fix issues with missing symbols
pthread_attr_setaffinity_npandpthread_rwlockattr_setkindwas done earlier, but suggested patch was reworked and partially reverted in 1cee865.reverted commit checked if
pthread_attr_setaffinity_npsymbol exists, but this check did not work (and effectively disabled better path), because this function requires#define _GNU_SOURCE(see https://man7.org/linux/man-pages/man3/pthread_attr_setaffinity_np.3.html)non-reverted part implied that
pthread_rwlockattr_setkindfunction replacespthread_rwlockattr_setkind_np, which is incorrect. This function is non-portable,PTHREAD_RWLOCK_PREFER_WRITER_NPserves as a hint and has no alternatives.Closes: #267