Mathias Krause <mathias.kra...@secunet.com> writes:
> Hi Kevin,
>
> On 17.09.2010 12:44, Kevin Wolf wrote:
>> Hi Mathias,
>>
>> Am 17.09.2010 08:42, schrieb Mathias Krause:
>>>> Using QEMU's block devices instead of a simple file would be
>>>> more consistent with the rest of QEMU and allow reading the
>>>> CMOS data not only from a file but also from an URL or other
>>>> sources.
>>> Thanks for the hint. Since this is my first contribution to the project
>>> I'm not that familiar with the code. Looking at other places, e.g. how
>>> the -kernel option gets handled, I just see FILE everywhere. Can you
>>> give me some pointers how to use this interface?
>>
>> Have a look at block.h which contains the prototypes for the public
>> block layer interface.
>>
>> Basically, you need to create a BlockDriverState with bdrv_new() and
>> then open it with bdrv_open(). You'll want to specify the raw block
>> driver for opening the image, you get it with bdrv_find_format("raw").
>> bdrv_pread/pwrite are the right functions to access the file with byte
>> granularity (other functions work on 512 byte sectors). bdrv_delete
>> frees the the BlockDriverState when you're done.
>
> Thank you for the detailed writeup. I think I should figure out how to
> use it myself now. Albeit there seem to be not many users of this
> interface right now. Looks like it's currently only used for storage
> devices. So I'm questioning myself: What _real_ benefit would it bring
> to use the QEMU block device layer for the CMOS file?
I'd view initial CMOS contents as configuration. We don't use the block
layer to access configuration files.
