Re: KQEMU 1.4.0pre1 for QEMU 0.10.1

[Naoya Sugioka <naoya.sugioka@xxxxxxxxx>]

Ping! Hi again,

After some code review, I just found set_phys_mem command for ioctl
was missing in kqemu_ioctl() from original kqemu-freebsd.c and my
porting kqemu-dragonfly.c
(kqemu never work correctly under *BSD ?)

Anyway, I updated kqemu module,but still kqemu does not give me any benefit.

So far, the concept of kqemu, do Linux vm86() like system calls on the
pinned kernel anonymous
memory, is not a right solution for BSD, it seems to me.

Here I attach the updated kqemu-dragonfly.c. we can build kqemu.ko and
we can load
as a module, but does not accelerates any virtualization tasks...

If you have any idea makes kqemu more effective, let me know please. thank you,
-Naoya




On Fri, Apr 10, 2009 at 11:16 PM, Naoya Sugioka <naoya.sugioka@gmail.com> wrote:
> Hi,
>
>
> I just motivated to port kqemu module since QEMU starts working good
> recently, according to this mailing list.
>
> As I attached kqemu-dragonfly.c and Makefile.dragonfly, the porting
> from FreeBSD version is straight forward, yet
> there is a problem. The kqemu module makes QEMU very *SLOW*.
>
> I tried to investigate by myself, also compared with Linux version of
> same source, but no luck.
> Here I also attach kqemutest result and messages log, dmesg log for
> both DragonFly(2.3-DEV) and Linux (2.6.29-rc8)
>
> Could someone please see them and let me know if you find any clues
> (or fix it for :-) ?
>
> thank you,
> Naoya
>

Attachment: kqemu-dragonfly.c
Description: Binary data