Re: msync() bug

[Andrea Arcangeli]

On Wed, Jul 11, 2001 at 12:03:11AM +1000, Andrew Morton wrote:
> Linus included the test for non-null page->mapping
> as well.  I wonder why.

I think it is because if you try hard you can map a dma page via
/dev/mem and then the driver can release the dma page (at rmmod for
example) before you munmap /dev/mem, and then the page goes in the
freelist and it is allocated as pagecache.

Andrea