Ah, you're right. This individual patch looks good to me.
Brian
On Mon, Dec 28, 2020 at 11:12 AM Dmitry Safonov <d...@arista.com> wrote:
>
> On 12/28/20 6:21 PM, Brian Geffon wrote:
> > This looks good to me with a small comment.
> >
> >> if (do_munmap(mm, old_addr, old_len, uf_unmap) < 0) {
> >> /* OOM: unable to split vma, just get accounts right */
> >> - if (vm_flags & VM_ACCOUNT)
> >> + if (vm_flags & VM_ACCOUNT && !(flags & MREMAP_DONTUNMAP))
> >> vm_acct_memory(new_len >> PAGE_SHIFT);
> >
> > Checking MREMAP_DONTUNMAP in the do_munmap path is unnecessary as
> > MREMAP_DONTUNMAP will have already returned by this point.
>
> In this code it is also used as err-path. In case move_page_tables()
> fails to move all page tables or .mremap() callback fails, the new VMA
> is unmapped.
>
> IOW, MREMAP_DONTUNMAP returns under:
> : if (unlikely(!err && (flags & MREMAP_DONTUNMAP))) {
>
> --
> Dima
