On 10/24/07, Enache Adrian <[EMAIL PROTECTED]> wrote:
> Drop this in ports/emulators/qemu/patches, and build the port.
>
> I'd be grateful if anybody was going to actually test all this,
> and tell me if it works.

I think the right place to wire this in is emulators/qemu, as a flavor
(or maybe a subpackage) that only shows up on amd64 and i386. Qemu
still works with full software emulation if the module is not loaded
or the user doesn't have permission to open the device; making kqemu a
subpackage rather than an entirely different flavor is probably the
best option.

tested on i386
- boots freedos OK
didn't have any serious apps to test with

- boots ubuntu 7.10 iso
with kqemu time runs at 33% speed. running top shows time ticking over
about 3 wall-seconds / emulated second. running "openssl speed" gives
speeds about 3x higher in the emulated environment than in the real.
subjectively it feels fairly responsive. without kqemu, it's much
slower, and there's much higher cpu usage in my host. typing at the
terminal feels like working over a congested entry-level dsl line, gui
operations feel like running remote-desktop over satellite... so i
guess kqemu is doing something useful.

- boots install42.iso
without kqemu, i can do a full install. not as fast as real metal but
acceptably quick. after install, openssl speed runs 12x slower than on
real metal. with kqemu, kernel crashes right after "root on rd0a swap
on rd0b dump on rd0b"
====================================
uvm_fault(0xd069f620, 0x0, 0, 3) -> e
fatal page fault (6) in supervisor mode
trap type 6 code 2 eip d02c8103 cs 8 eflags 10216 cr2 40 cpl 0
panic: trap type 6, code 2, pc=d02c810e
syncing disks... done
panic: init died (signal 4, exit 0)
====================================

a similar crash is observed with a proper GENERIC kernel.

CK

-- 
GDB has a 'break' feature; why doesn't it have 'fix' too?

Reply via email to