[Qemu-devel] [PATCH, RFC] Configuration device: add write channel, mark pages RO

[Qemu-devel] [PATCH, RFC] Configuration device: add write channel, mark pages RO

[Blue Swirl]

[-- Attachment #1: Type: text/plain, Size: 613 bytes --]

Hi,

The first patch extend the configuration device with a guest write
capability. A callback is invoked when the buffer is full.

It's a bit odd to get completely different behavior when doing I/O to
the same port just because of the word size used in the IO. I'd rather
have two ports, one for control write or status read and another for
data read/write. Is two ports too much to ask on a PC?

The second patch uses the first one to add a new key for marking guest
pages read only. This patch was inspired due to discussion I saw on
KernelTrap:
http://kerneltrap.org/mailarchive/linux-kernel/2008/9/4/3191234

[-- Attachment #2: fw_cfg_write_channel.diff --]
[-- Type: plain/text, Size: 5120 bytes --]

[-- Attachment #3: fw_cfg_ro_page.diff --]
[-- Type: plain/text, Size: 2818 bytes --]

Re: [Qemu-devel] [PATCH, RFC] Configuration device: add write channel, mark pages RO

[Gleb Natapov]

On Sun, Sep 07, 2008 at 09:56:56PM +0300, Blue Swirl wrote:
> Hi,
> 
> The first patch extend the configuration device with a guest write
> capability. A callback is invoked when the buffer is full.
> 
Looks useful. VMware back door allows for two way communication too.

> It's a bit odd to get completely different behavior when doing I/O to
> the same port just because of the word size used in the IO. I'd rather
> have two ports, one for control write or status read and another for
> data read/write. Is two ports too much to ask on a PC?
Lets use two ports for clarity.

--
			Gleb.