Messages in this thread | | | Date | Mon, 17 Apr 2023 13:13:58 +0200 | From | David Hildenbrand <> | Subject | Re: [PATCH v3 4/7] mm/gup: introduce the FOLL_SAME_FILE GUP flag |
| |
On 15.04.23 14:09, Lorenzo Stoakes wrote: > This flag causes GUP to assert that all VMAs within the input range possess > the same vma->vm_file. If not, the operation fails. > > This is part of a patch series which eliminates the vmas parameter from the > GUP API, implementing the one remaining assertion within the entire kernel > that requires access to the VMAs associated with a GUP range. > > Signed-off-by: Lorenzo Stoakes <lstoakes@gmail.com>
[...]
> --- > &start, &nr_pages, i, > @@ -1595,7 +1603,7 @@ long faultin_vma_page_range(struct vm_area_struct *vma, unsigned long start, > * We want to report -EINVAL instead of -EFAULT for any permission > * problems or incompatible mappings. > */ > - if (check_vma_flags(vma, gup_flags)) > + if (check_vma_flags(vma, vma->vm_file, gup_flags)) > return -EINVAL;
FOLL_SAME_FILE is never set here, just pass NULL instead of vma->vm_file.
As we're not allowing to drop the mmap lock, why can't io_uring simply go over all VMAs once, after pinning succeeded, and make sure that the files match (or even before pinning)?
In most cases, we're dealing with a single VMA only, it's not like the common case is that io_uring pins accross 100s of VMAs.
So I really wonder if the GUP complexity is justified by something. (removing the VMAs is certainly a welcome surprise -- as it doesn't make any sense when used with FOLL_UNLOCKABLE).
-- Thanks,
David / dhildenb
| |