Re: [PATCH v2] binder_alloc: Add missing mmap_lock calls when using the VMA

[Ondrej Mosnacek <omosnace@redhat.com>]

On Wed, Aug 10, 2022 at 6:02 PM Liam Howlett <liam.howlett@oracle.com> wrote:
> Take the mmap_read_lock() when using the VMA in
> binder_alloc_print_pages() and when checking for a VMA in
> binder_alloc_new_buf_locked().
>
> It is worth noting binder_alloc_new_buf_locked() drops the VMA read lock
> after it verifies a VMA exists, but may be taken again deeper in the
> call stack, if necessary.
>
> Reported-by: Ondrej Mosnacek <omosnace@redhat.com>
> Reported-by: syzbot+a7b60a176ec13cafb793@syzkaller.appspotmail.com
> Fixes: a43cfc87caaf (android: binder: stop saving a pointer to the VMA)
> Signed-off-by: Liam R. Howlett <Liam.Howlett@oracle.com>
> ---
>  drivers/android/binder_alloc.c | 31 +++++++++++++++++++++----------
>  1 file changed, 21 insertions(+), 10 deletions(-)
>
> diff --git a/drivers/android/binder_alloc.c b/drivers/android/binder_alloc.c
> index f555eebceef6..c5e7c6d3a844 100644
> --- a/drivers/android/binder_alloc.c
> +++ b/drivers/android/binder_alloc.c
> @@ -395,12 +395,15 @@ static struct binder_buffer *binder_alloc_new_buf_locked(
>         size_t size, data_offsets_size;
>         int ret;
>
> +       mmap_read_lock(alloc->vma_vm_mm);
>         if (!binder_alloc_get_vma(alloc)) {
> +               mmap_read_unlock(alloc->vma_vm_mm);
>                 binder_alloc_debug(BINDER_DEBUG_USER_ERROR,
>                                    "%d: binder_alloc_buf, no vma\n",
>                                    alloc->pid);
>                 return ERR_PTR(-ESRCH);
>         }
> +       mmap_read_unlock(alloc->vma_vm_mm);
>
>         data_offsets_size = ALIGN(data_size, sizeof(void *)) +
>                 ALIGN(offsets_size, sizeof(void *));
> @@ -922,17 +925,25 @@ void binder_alloc_print_pages(struct seq_file *m,
>          * Make sure the binder_alloc is fully initialized, otherwise we might
>          * read inconsistent state.
>          */
> -       if (binder_alloc_get_vma(alloc) != NULL) {
> -               for (i = 0; i < alloc->buffer_size / PAGE_SIZE; i++) {
> -                       page = &alloc->pages[i];
> -                       if (!page->page_ptr)
> -                               free++;
> -                       else if (list_empty(&page->lru))
> -                               active++;
> -                       else
> -                               lru++;
> -               }
> +
> +       mmap_read_lock(alloc->vma_vm_mm);
> +       if (binder_alloc_get_vma(alloc) == NULL) {
> +               mmap_read_unlock(alloc->vma_vm_mm);
> +               goto uninitialized;
>         }
> +
> +       mmap_read_unlock(alloc->vma_vm_mm);
> +       for (i = 0; i < alloc->buffer_size / PAGE_SIZE; i++) {
> +               page = &alloc->pages[i];
> +               if (!page->page_ptr)
> +                       free++;
> +               else if (list_empty(&page->lru))
> +                       active++;
> +               else
> +                       lru++;
> +       }
> +
> +uninitialized:
>         mutex_unlock(&alloc->mutex);
>         seq_printf(m, "  pages: %d:%d:%d\n", active, lru, free);
>         seq_printf(m, "  pages high watermark: %zu\n", alloc->pages_high);
> --
> 2.35.1

Can confirm that with this patch the selinux-testsuite's binder test
passes again with no warnings in dmesg.

Tested-by: Ondrej Mosnacek <omosnace@redhat.com>

Thanks!

-- 
Ondrej Mosnacek
Senior Software Engineer, Linux Security - SELinux kernel
Red Hat, Inc.