Re: [PATCH] x86/mm: Split vmalloc_sync_all()

[Dave Hansen <dave.hansen@intel.com>]

On 10/7/19 8:16 AM, Joerg Roedel wrote:
> @@ -318,7 +328,7 @@ static void dump_pagetable(unsigned long address)
>  
>  #else /* CONFIG_X86_64: */
>  
> -void vmalloc_sync_all(void)
> +void vmalloc_sync_mappings(void)
>  {
>  	sync_global_pgds(VMALLOC_START & PGDIR_MASK, VMALLOC_END);
>  }

FWIW, I generally detest the use of __weak. :)

In this case, it ends up letting us gloss over the fact that we have a
32/64-bit asymmetry.  It would probably be nice to actually have a
64-bit implementation that comes along with a nice comment.  Maybe this
in vmalloc_sync_mappings():

	/*
	 * 64-bit mappings might allocate new p4d/pud pages
	 * that need to be propagated to all tasks' PGDs.
	 */

which would pair nicely with:

void vmalloc_sync_unmappings(void)
{
	/*
	 * Unmappings never allocate or free p4d/pud pages.
	 * No work is required here.
	 */
}