[Qemu-devel] Re: [PATCH 0/2] fix virtio_blk serial pci config breakage

[john cooper]

Anthony Liguori wrote:

> 2) Passing an ATA identity page is goofy.  We should just pass the 
> serial number and let Linux generate the identity page.

I have a hard time disagreeing with that comment.

But the rationale that got us here was from the
user's perspective this information could be
trivially exposed via a preexisting interface vs.
creating a new interface as I'd done initially.
It was also pointed out other defined fields in
the identify page exist which were potentially
useful.  So pressing a HDIO_GET_IDENTITY ioctl
into service here didn't seem unreasonable.

In terms of getting that bag of bits from qemu to
the guest userspace, having qemu cobble together
the page is arguably appropriate as that it where
the source of the information resides.  The
alternative of pushing knowledge of this structure
into the guest driver seems unnatural.  Moreover
we would still somehow need a means to get the
data from qemu to the driver even if the driver
was only packaging it up for export as an identify
page.

> Just because 
> Linux requires this as it's user space interface, that doesn't mean that 
> other guests will (like Windows).  Instead of exposing an opaque blob, 
> we should expose the information we need in a structured way.

The identify page is defined as the return of an
"identify drive" command from an ATA drive/controller.
I'd hazard it is a familiar structure in a windows
driver/userland context as well.

-john


--
address@hidden