raw device property

[Soren Hansen]

On Thu, Mar 30, 2006 at 11:46:36AM -0800, Artem Kachitchkine wrote:
> >Yea, block.solaris.raw_device is probably better. Thanks!

I'm still pretty new around here, but I don't understand the point in
the solaris bit. Is raw_device ambiguous in any way without this solaris
bit?

> BTW, Soren, I agree we're in the abstraction business and I think the
> goal you set is much higher than mine, therefore needs much more work
> than I can afford right now (and there doesn't seem to be any real
> consumer on the Linux side yet).

Well... As far as I know, any block device on Linux can be opened as a
raw device with the O_DIRECT flag to the open(2) system call. This could
be exposed through a block.o_direct_raw (better naming suggestions
welcome) attribute that would simply be set for all block devices on
Linux (and any other OS with the same behaviour.  The existence of a
raw_device attribute would obviously mean that raw access should be
obtained throught the file in that attribute, while the absence of a
raw_device attribute along with the presence of the block.o_direct_raw
attribute would signify that raw access can be obtained through the
regular device file with the O_DIRECT flag set.

Wouldn't that make everybody happy?

> Sorry for taking the easy way out, but I've been running out of Advil
> too often lately :)

Hehe. :-)

-- 
| Soren Hansen    | Linux2Go                  | http://Linux2Go.dk/ |
| Seniorkonsulent | Lindholmsvej 42, 2. TH    | +45 46 90 26 42     |
| sh at linux2go.dk  | 9400 Norresundby, Denmark | GPG key: E8BDA4E3   |
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: Digital signature
Url : http://lists.freedesktop.org/archives/hal/attachments/20060330/6dcf7696/attachment.pgp