Re: [PATCH RFC 0/3] A few round_pipe_size() and pipe-max-size fixups

[Joe Lawrence <joe.lawrence@redhat.com>]

On 09/14/2017 12:57 PM, Randy Dunlap wrote:
> On 09/14/17 06:26, Michael Kerrisk (man-pages) wrote:
>> Hello Joe,
>>
>> On 5 September 2017 at 16:44, Joe Lawrence <joe.lawrence@redhat.com> wrote:
>>> While backporting Michael's "pipe: fix limit handling" [1] patchset to a
>>> distro-kernel, Mikulas noticed that current upstream pipe limit handling
>>> contains a few problems:
>>>
>>>   1 - round_pipe_size() nr_pages overflow on 32bit:  this would
>>>       subsequently try roundup_pow_of_two(0), which is undefined.
> 
> Hi,
> Sorry I missed the initial posting of this.
> 
> The man page for F_SETPIPE_SZ (http://man7.org/linux/man-pages/man2/fcntl.2.html)
> says:
> "Attempts to set the pipe capacity below the page size are
> silently rounded up to the page size."
> 
> That implies to me that setting pipe size to 0 would round up to PAGE_SIZE.
> Doesn't patch 1/3 change that to return -EINVAL?

Good catch.  How about something like this:

/*
 * Minimum pipe size, as required by POSIX
 */
unsigned int pipe_min_size = PAGE_SIZE;

...

static inline unsigned int round_pipe_size(unsigned int size)
 {
        unsigned long nr_pages;

+       if (size < pipe_min_size)
+               size = pipe_min_size;
+
        nr_pages = (size + PAGE_SIZE - 1) >> PAGE_SHIFT;
        if (nr_pages == 0)
                return 0;

> 
> Otherwise all 3 patches look good to me.

If the above is good, I can fold this into patch 1 and respin the set.

Thanks,

-- Joe