Re: [PATCH v4 10/21] mm/mmap: Support vma == NULL in init_vma_munmap()

From: Lorenzo Stoakes
Date: Thu Jul 11 2024 - 10:29:06 EST


On Wed, Jul 10, 2024 at 03:22:39PM GMT, Liam R. Howlett wrote:
> From: "Liam R. Howlett" <Liam.Howlett@xxxxxxxxxx>
>
> Adding support for a NULL vma means the init_vma_munmap() can be
> initialized for a less error-prone process when calling
> vms_complete_munmap_vmas() later on.
>
> Signed-off-by: Liam R. Howlett <Liam.Howlett@xxxxxxxxxx>
> ---
> mm/mmap.c | 12 +++++++-----
> 1 file changed, 7 insertions(+), 5 deletions(-)
>
> diff --git a/mm/mmap.c b/mm/mmap.c
> index a1544a68558e..e2e6b3202c25 100644
> --- a/mm/mmap.c
> +++ b/mm/mmap.c
> @@ -516,10 +516,12 @@ static inline void init_vma_munmap(struct vma_munmap_struct *vms,
> bool unlock)
> {
> vms->vmi = vmi;
> - vms->vma = vma;
> - vms->mm = vma->vm_mm;
> - vms->start = start;
> - vms->end = end;
> + if (vma) {
> + vms->vma = vma;
> + vms->mm = vma->vm_mm;
> + vms->start = start;
> + vms->end = end;
> + }

Why not store start/end even if !vma? And shouldn't we have an else clause
to make sure these are initialised in this case too?

I mean also we could have vms->vma = vma outside of this clause to so it
looks something like:

vms->vma = vma;
vms->mm = vma ? vma->vm_mm : NULL;
vms->start = start;
vms->end = end;

> vms->unlock = unlock;
> vms->uf = uf;
> vms->vma_count = 0;
> @@ -2950,11 +2952,11 @@ unsigned long mmap_region(struct file *file, unsigned long addr,
>
> /* Find the first overlapping VMA */
> vma = vma_find(&vmi, end);
> + init_vma_munmap(&vms, &vmi, vma, addr, end, uf, /* unlock = */ false);
> if (vma) {
> mt_init_flags(&mt_detach, vmi.mas.tree->ma_flags & MT_FLAGS_LOCK_MASK);
> mt_on_stack(mt_detach);
> mas_init(&mas_detach, &mt_detach, /* addr = */ 0);
> - init_vma_munmap(&vms, &vmi, vma, addr, end, uf, /* unlock = */ false);
> /* Prepare to unmap any existing mapping in the area */
> if (vms_gather_munmap_vmas(&vms, &mas_detach))
> return -ENOMEM;
> --
> 2.43.0
>

I really like this approach in general though!