privsep patch, Please test

[Ben Lindstrom]

On Wed, 29 May 2002, Tim Rice wrote:

> On Wed, 29 May 2002, Ben Lindstrom wrote:
>
> >
> > Can you make the following change to just before the mmap()..  Thanks.
> >
> >         if (address == MAP_FAILED)
> > -                fatal("mmap(%lu)", (u_long)size);
> > +                fatal("mmap(%lu) %s", (u_long)size, strerror(errno));
> >
> > I need to know what mmap is setting errno to.  Otherwise I'm making a
> > guess in the dark (which is 2.2 Linux mmap is fucked, or you have a
> > version of glibc that has a bad mmap).
>
> strace tells the story
> open("/dev/zero", O_RDWR|O_LARGEFILE)   = 8
> mmap(NULL, 65536, PROT_READ|PROT_WRITE, MAP_SHARED, 8, 0) = -1 EINVAL (Invalid a
> rgument)
>
> The man page says
>     EINVAL We don't like start or length  or  offset.   (E.g.,
>            they  are  too  large, or not aligned on a PAGESIZE
>            boundary.)
>
> I don't understand why it's happy with MAP_PRIVATE but doesn't like
> MAP_SHARED.
>

What does getpagesize() say?  It should be 4096.

What bothers me is the 1.3.x Linux kernel manpage talks about MAP_ANON
being implemented.

[..]
       The  above three flags are described in POSIX.1b (formerly
       POSIX.4).  Linux also knows about MAP_DENYWRITE,  MAP_EXE-
       CUTABLE and MAP_ANON(YMOUS).
[..]


- Ben