> /* Try to map a page at the VGA address */
>     printf("fd = %d\n", fd);
> 
>     base = (caddr_t)mmap((caddr_t)0, 4096, PROT_READ|PROT_WRITE,
> 			 MAP_FILE, fd, (off_t)0xA0000);

It should be mmap()ing at offset 0.

This has *always* been the case, since the first instant that 386BSD
had mmap() capabilities for /dev/vga.

The difference is that our VM system now understands offsets given
to device mmap() calls, and doesn't (silently) chop it to 0.

this is an X bug.


cgd

------------------------------------------------------------------------------