Date: Tue, 16 Nov 2004 17:53:28 -0800
From: "David S. Miller" <davem@davemloft.net>
To: linux-kernel@vger.kernel.org
Subject: loops in get_user_pages() for VM_IO
Message-Id: <20041116175328.5e425e01.davem@davemloft.net>
Mime-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 1068
Lines: 30


Some time recently, I don't know when, the logic in get_user_pages()
appears to have been changed a bit.

The inner-most loop of this routine basically does:

	while (follow_page() returns NULL)
		try handle_mm_fault();

The problem with this is that for !pfn_valid()
(for example, VM_IO areas mapping to I/O physical
address which have no assosciated page struct) follow_page()
will _always_ return NULL.

So when X tries to mmap() the frame buffer on my
system it loops forever here now.

Is pfn_valid() supposed to return true for I/O areas
too?  If so, where does the struct page backing store
come from?

It could be argued that setting VM_LOCKED is invalid.
And not setting it in drivers/video/sbuslib.c would make
this hang go away, but the above analysis means that
make_pages_present() cannot work on VM_IO areas.
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/