On Thu, Aug 29, 2024 at 06:26:41PM +1000, Michael Ellerman wrote:

Charlie Jenkins [off-list ref] writes:

quoted

Some applications rely on placing data in free bits addresses allocated
by mmap. Various architectures (eg. x86, arm64, powerpc) restrict the
address returned by mmap to be less than the 48-bit address space,
unless the hint address uses more than 47 bits (the 48th bit is reserved
for the kernel address space).

To make this behavior explicit and more versatile across all
architectures, define a mmap flag that allows users to define an
arbitrary upper limit on addresses returned by mmap.

Signed-off-by: Charlie Jenkins <redacted>
---
 include/uapi/asm-generic/mman-common.h       | 1 +
 tools/include/uapi/asm-generic/mman-common.h | 1 +

  
You're not meant to update the headers in tools/ directly. There's a
mail somewhere from acme somewhere describing the proper process, but
the tldr is leave it up to him.

Oh okay, thank you.

quoted

diff --git a/include/uapi/asm-generic/mman-common.h b/include/uapi/asm-generic/mman-common.h
index 6ce1f1ceb432..03ac13d9aa37 100644
--- a/include/uapi/asm-generic/mman-common.h
+++ b/include/uapi/asm-generic/mman-common.h

@@ -32,6 +32,7 @@
 
 #define MAP_UNINITIALIZED 0x4000000	/* For anonymous mmap, memory could be
 					 * uninitialized */
+#define MAP_BELOW_HINT	  0x8000000	/* give out address that is below (inclusive) hint address */

IMHO the API would be clearer if this actually forced the address to be
below the hint. That's what the flag name implies after all.

It would also mean the application doesn't need to take into account the
length of the mapping when passing the hint.

cheers

That's a good point. The reason I did it this way was to allow mmap the
possibility of returning the same address as the hint. If it must be
strictly less than the hint then the hint address can never be returned.
Maybe that doesn't matter though.

- Charlie