Re: [PATCH v6 08/10] mm, vma: use percpu sheaves for vm_area_struct cache

[Lorenzo Stoakes <lorenzo.stoakes@oracle.com>]

On Wed, Sep 03, 2025 at 02:47:28PM +0200, Vlastimil Babka wrote:
> On 9/2/25 13:13, Lorenzo Stoakes wrote:
> > On Wed, Aug 27, 2025 at 10:26:40AM +0200, Vlastimil Babka wrote:
> >> Create the vm_area_struct cache with percpu sheaves of size 32 to
> >> improve its performance.
> >>
> >> Reviewed-by: Suren Baghdasaryan <surenb@google.com>
> >> Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
> >> ---
> >>  mm/vma_init.c | 1 +
> >>  1 file changed, 1 insertion(+)
> >>
> >> diff --git a/mm/vma_init.c b/mm/vma_init.c
> >> index 8e53c7943561e7324e7992946b4065dec1149b82..52c6b55fac4519e0da39ca75ad018e14449d1d95 100644
> >> --- a/mm/vma_init.c
> >> +++ b/mm/vma_init.c
> >> @@ -16,6 +16,7 @@ void __init vma_state_init(void)
> >>  	struct kmem_cache_args args = {
> >>  		.use_freeptr_offset = true,
> >>  		.freeptr_offset = offsetof(struct vm_area_struct, vm_freeptr),
> >> +		.sheaf_capacity = 32,
> >
> > This breaks the VMA tests.
> >
> > Please also update tools/testing/vma/vma_internal.h to add this field
> > (probably actually at the commit that adds the field in mainline :P)
>
> Thanks, one of Liam's patches to support maple tree tests with sheaves does
> that, we somehow missed that the vma test also needs it so will reorder
> accordingly.
> > I do wish we could have these tests be run by a bot :/ seems today I am
> > that bot :)
> Sorry :)

Thanks :) Just happy to find the problem and help it get sorted! :>)