Re: [PATCH 6/7] pipe: simplify round_pipe_size()

[Kees Cook]

On Sun, Jan 7, 2018 at 9:35 PM, Eric Biggers <ebiggers3@xxxxxxxxx> wrote:
> From: Eric Biggers <ebiggers@xxxxxxxxxx>
>
> round_pipe_size() calculates the number of pages the requested size
> corresponds to, then rounds the page count up to the next power of 2.
>
> However, it also rounds everything < PAGE_SIZE up to PAGE_SIZE.
> Therefore, there's no need to actually translate the size into a page
> count; we just need to round the size up to the next power of 2.
>
> We do need to verify that bit 31 isn't set, since on 32-bit systems
> roundup_pow_of_two() would be undefined in that case.  But that can just
> be combined with the UINT_MAX check which we need anyway now.
>
> Finally, also remove the check for '!nr_pages' in pipe_set_size(), since
> round_pipe_size() always returns either 0 or a multiple of PAGE_SIZE.
>
> Signed-off-by: Eric Biggers <ebiggers@xxxxxxxxxx>
> ---
>  fs/pipe.c | 15 +++------------
>  1 file changed, 3 insertions(+), 12 deletions(-)
>
> diff --git a/fs/pipe.c b/fs/pipe.c
> index f1ee1e599495..774cafd947dc 100644
> --- a/fs/pipe.c
> +++ b/fs/pipe.c
> @@ -1022,20 +1022,14 @@ const struct file_operations pipefifo_fops = {
>   */
>  unsigned int round_pipe_size(unsigned long size)
>  {
> -       unsigned long nr_pages;
> -
> -       if (size > UINT_MAX)
> +       if (size > (1U << 31))
>                 return 0;
>
>         /* Minimum pipe size, as required by POSIX */
>         if (size < PAGE_SIZE)
> -               size = PAGE_SIZE;
> -
> -       nr_pages = (size + PAGE_SIZE - 1) >> PAGE_SHIFT;
> -       if (nr_pages == 0)
> -               return 0;
> +               return PAGE_SIZE;
>
> -       return roundup_pow_of_two(nr_pages) << PAGE_SHIFT;
> +       return roundup_pow_of_two(size);
>  }
>
>  /*

Above looks good.

> @@ -1054,9 +1048,6 @@ static long pipe_set_size(struct pipe_inode_info *pipe, unsigned long arg)
>                 return -EINVAL;
>         nr_pages = size >> PAGE_SHIFT;
>
> -       if (!nr_pages)
> -               return -EINVAL;
> -

I would just leave this hunk anyway: it's defensive for any future
changes. Maybe add a comment describing why it's currently redundant?

-Kees

>         /*
>          * If trying to increase the pipe capacity, check that an
>          * unprivileged user is not trying to exceed various limits
> --
> 2.15.1
>



-- 
Kees Cook
Pixel Security