On 7/20/21 2:16 PM, Andi Kleen wrote: > On 7/20/2021 10:59 AM, Dave Hansen wrote: >> On 7/20/21 10:52 AM, Kuppuswamy, Sathyanarayanan wrote: >>>> Why does this need to use the page allocator directly? >> ^^ You didn't address this question. > > The address needs to be naturally aligned, and I'm not sure all slab > allocators guarantee 64 byte alignment. So using the page allocator > seems to be safer. I guess a comment would be good. The documentation > https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/Documentation/core-api/memory-allocation.rst#n146 claims: > The address of a chunk allocated with `kmalloc` is aligned to at least > ARCH_KMALLOC_MINALIGN bytes. For sizes which are a power of two, the > alignment is also guaranteed to be at least the respective size. This is another case where care in coding these things up, writing good changelogs and expressing assumptions in comments can save back-and-forth from a reviewer. Imagine if this had been: /* * tdreport_data needs to be 64-byte aligned. * Full page alignment is more than enough. */ tdreport_data = (void *)__get_free_pages(GFP_KERNEL | __GFP_ZERO, 0); if (!tdreport_data) { ret = -ENOMEM; goto failed; }
