Date: Mon, 3 Jun 1996 06:52:32 +1000 From: Bruce Evans <bde@zeta.org.au> To: bde@freebsd.org, pjf@cts.com Cc: dyson@freebsd.org, hackers@freebsd.org Subject: Re: bugs Message-ID: <199606022052.GAA06199@godzilla.zeta.org.au>
next in thread | raw e-mail | index | archive | help
>| Subject: Re: bugs
>| MMAP doesn't extend files, AFAIK it doesn't on many (if not most
>| other OSes.) To set the length of the file you need to do an ftruncate.
>| After that, then blocks will be allocated as needed.
>Sorry, I screwed up my example. Here's the real one:
>-------------------------
>#include <stdio.h>
>#include <sys/mman.h>
>#include <sys/file.h>
>main()
>{
> int fd = open("newfile", O_RDWR|O_CREAT|O_EXCL, 0660);
> char *buf;
> ftruncate(fd, 100);
> buf = mmap(NULL, 100, PROT_WRITE, MAP_SHARED, fd, 0);
> printf("%lx\n", buf);
> strcpy(buf, "hi!");
>}
>-------------------------
>This doesn't work either. The ftruncate() appears not to work; the file
>is still zero length after the program crashes. So perhaps the problem
>is with ftruncate() and not mmap().
>This works on all of our 10 existing UNIX platforms except BSDI and
>Linux. BSDI has the same problem with ftruncate, I think; Linux's
>ftruncate works, but its mmap() appears to be totally broken, at least
>in 1.2.13.
The ftruncate() works if a prototype for ftruncate() is in scope or if
the `length' arg to ftruncate has the correct type (off_t = long long).
Otherwise the top 32 bits of the length are random. ftruncate() is
prototyped in <unistd.h>. You should also include <sys/types.h> before
including <sys/mman.h>. The above happens to work because <stdio.h>
bogusly includes <sys/types.h>. You should also include <fcntl.h>
instead of <sys/file.h> except on old systems.
The strcpy() doesn't work unless PROT_WRITE is changed to
`PROT_READ | PROT_WRITE', and this isn't because strcpy() reads its
target - `buf[0] = 1' fails in the same way.
Bruce
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199606022052.GAA06199>