You didn't address all of the issues. A key issue left unaddressed is
the lack of a Linux DDI. A project that aims at making Linux kernel
drivers work in user space on any system would have to deal with the
fact that Linux doesn't have a stable DDI. This means that the
programming interface the user space environment provides is itself in
flux -- it has to match a moving target. How can you do that?
My idea is to rewrite linux driver instead of just to copy them. I
suppose rewrite a working driver
will be easier then write one from scratch, especially when it is hard
to get the hardware spec.
Therefore it does not require a clear interface for linux drivers,
which they do not have.
I feel it could be useful for writing driver for some customized
hardware since development
and debug in user mode is much more convenient than in kernel mode.
This sounds like a different project to me.
I'd be somewhat in favor of a project to do something like that,
provided that:
- It doesn't promise the ability to do Linux driver porting, as that
seems an infeasible task.
- Any core system functionality delivered via user space drivers has
a clear set of performance goals.
- The project team works with some other group or person that wants
to deliver such a driver (at least one consumer; no orphans,
please).
For what it's worth, this new proposal sounds a bit like Jungo's
"WinDriver:"
http://www.jungo.com/wd-solaris.html
That's the example I want to give. Specifically, the usb one. Such as
feature will make writing solaris
usb driver much easier. As for the pci one, windriver also achieves that
with a kernel driver. Since we already
have powerful tools like mdb, dtrace. I dont think that help to much for
Sun. Although it may be
an interest idea for hardware vendor who want to write a portable
driver. Therefore, we are now make
deeper investigation on usb.
Thank you
--Freeman
_______________________________________________
networking-discuss mailing list
[email protected]
