sys_tee() bug: after tee() of partial page, both pipes can merge, clobbering each other's data

sys_tee() bug: after tee() of partial page, both pipes can merge, clobbering each other's data

[Jann Horn]

Hi!

I noticed the following behavior; basically, after copying part of a
normal pipe buffer (anon_pipe_buf_ops) from pipe A to pipe B, both
pipe A and pipe B can merge new writes into the existing page,
clobbering each other's data:

============
$ cat tee_test.c
#define _GNU_SOURCE
#include <fcntl.h>
#include <unistd.h>
#include <err.h>
#include <stdio.h>

int main(void) {
  int pipe_a[2];
  if (pipe(pipe_a)) err(1, "pipe");
  int pipe_b[2];
  if (pipe(pipe_b)) err(1, "pipe");
  if (write(pipe_a[1], "abcd", 4) != 4) err(1, "write");
  if (tee(pipe_a[0], pipe_b[1], 2, 0) != 2) err(1, "tee");
  if (write(pipe_b[1], "xx", 2) != 2) err(1, "write");

  char buf[5];
  if (read(pipe_a[0], buf, 4) != 4) err(1, "read");
  buf[4] = 0;
  printf("got back: '%s'\n", buf);
}
$ gcc -o tee_test tee_test.c
$ ./tee_test
got back: 'abxx'
$
============

splice_pipe_to_pipe() probably has the same problem?

I'm not sure what the cleanest way to fix this would be. Replace
anon_pipe_buf_ops with packet_pipe_buf_ops when copying a buffer? Or
add a new buffer flag for marking a buffer as mergeable, and get rid
of buf->ops->can_merge?

Re: sys_tee() bug: after tee() of partial page, both pipes can merge, clobbering each other's data

[Jann Horn]

On Mon, Oct 15, 2018 at 4:13 AM Jann Horn <jannh@google.com> wrote:
> I noticed the following behavior; basically, after copying part of a
> normal pipe buffer (anon_pipe_buf_ops) from pipe A to pipe B, both
> pipe A and pipe B can merge new writes into the existing page,
> clobbering each other's data:
>
> ============
> $ cat tee_test.c
> #define _GNU_SOURCE
> #include <fcntl.h>
> #include <unistd.h>
> #include <err.h>
> #include <stdio.h>
>
> int main(void) {
>   int pipe_a[2];
>   if (pipe(pipe_a)) err(1, "pipe");
>   int pipe_b[2];
>   if (pipe(pipe_b)) err(1, "pipe");
>   if (write(pipe_a[1], "abcd", 4) != 4) err(1, "write");
>   if (tee(pipe_a[0], pipe_b[1], 2, 0) != 2) err(1, "tee");
>   if (write(pipe_b[1], "xx", 2) != 2) err(1, "write");
>
>   char buf[5];
>   if (read(pipe_a[0], buf, 4) != 4) err(1, "read");
>   buf[4] = 0;
>   printf("got back: '%s'\n", buf);
> }
> $ gcc -o tee_test tee_test.c
> $ ./tee_test
> got back: 'abxx'
> $
> ============
>
> splice_pipe_to_pipe() probably has the same problem?
>
> I'm not sure what the cleanest way to fix this would be. Replace
> anon_pipe_buf_ops with packet_pipe_buf_ops when copying a buffer? Or
> add a new buffer flag for marking a buffer as mergeable, and get rid
> of buf->ops->can_merge?

Actually, I'll just cook up a simple patch myself.