[libc-commits] [libc] [libc][mmap] force offset to long for mmap2 (PR #96522)

[via libc-commits]

llvmbot wrote:


<!--LLVM PR SUMMARY COMMENT-->

@llvm/pr-subscribers-libc

Author: Nick Desaulniers (paternity leave) (nickdesaulniers)

<details>
<summary>Changes</summary>

llvm-libc's "Large File Support" isn't really complete.  We define off_t to be
a uint64_t always, which breaks with convention in an attempt to drop
historical baggage. We might need to revisit that decision in the future, but
for now, force off_t to long when using mmap2.

This avoids having to pass a 64b value to our syscall wrappers for 32b targets,
which we do not support on arm32.  Since we're using mmap2 rather than mmap, we
don't need a 64b value anyways.

Our mmap implementation still has further room for improvement...

---
Full diff: https://github.com/llvm/llvm-project/pull/96522.diff


1 Files Affected:

- (modified) libc/src/sys/mman/linux/mmap.cpp (+2-1) 


``````````diff
diff --git a/libc/src/sys/mman/linux/mmap.cpp b/libc/src/sys/mman/linux/mmap.cpp
index 16111c66859f5..e280ac42092a8 100644
--- a/libc/src/sys/mman/linux/mmap.cpp
+++ b/libc/src/sys/mman/linux/mmap.cpp
@@ -41,7 +41,8 @@ LLVM_LIBC_FUNCTION(void *, mmap,
 
   long ret =
       LIBC_NAMESPACE::syscall_impl(syscall_number, reinterpret_cast<long>(addr),
-                                   size, prot, flags, fd, offset);
+                                   size, prot, flags, fd,
+                                   static_cast<long>(offset));
 
   // The mmap/mmap2 syscalls return negative values on error. These negative
   // values are actually the negative values of the error codes. So, fix them

``````````

</details>


https://github.com/llvm/llvm-project/pull/96522