Hollis Blanchard hol...@penguinppc.org writes:
I am able to run qemu with the following commandline:
/usr/local/bin/qemu-system-ppcemb -enable-kvm -kernel uImage.bamboo
-nographic -M bamboo ppc440-angstrom-linux.img
However, when I try to use virtio instead, I get this segfault:
/usr/local/bin/qemu-system-ppcemb -enable-kvm -kernel uImage.bamboo
-drive file=ppc440-angstrom-linux.img,if=virtio -nographic -M bamboo
#0 0x1009864c in qbus_find_recursive (bus=0x0, name=0x0, info=0x10287238)
at /home/hollisb/work/qemu.git/hw/qdev.c:461
#1 0x10099cc4 in qdev_device_add (opts=0x108a07a0)
at /home/hollisb/work/qemu.git/hw/qdev.c:229
#2 0x101a4220 in device_init_func (opts=value optimized out,
opaque=value optimized out) at /home/hollisb/work/qemu.git/vl.c:1519
#3 0x1002baf8 in qemu_opts_foreach (list=value optimized out,
func=0x101a4204 device_init_func, opaque=0x0,
abort_on_failure=value optimized out) at qemu-option.c:978
#4 0x101a68e0 in main (argc=value optimized out,
argv=value optimized out, envp=value optimized out)
at /home/hollisb/work/qemu.git/vl.c:2890
This patch avoids the segfault, but just gives me this message: No
'PCI' bus found for device 'virtio-blk-pci'
diff --git a/hw/qdev.c b/hw/qdev.c
index e99c73f..8fe4f06 100644
--- a/hw/qdev.c
+++ b/hw/qdev.c
@@ -455,6 +455,9 @@ static BusState *qbus_find_recursive(BusState *bus, const
ch
BusState *child, *ret;
int match = 1;
+ if (!bus)
+ return NULL;
+
if (name (strcmp(bus-name, name) != 0)) {
match = 0;
}
Is your main_system_bus still null?
We should report the problem, not crash, of course.
FWIW, hw/ppc4xx_pci.c is my PCI controller. Do I need to add some qdev
magic to that file to make this work?
Yes, you need to convert the device providing your PCI bus to qdev
before you can use any qdevified PCI devices.
piix_pci.c could serve as an example. qdev works starts with commit
8a14daa5.
My KVM Forum slides might help with the basics:
http://www.linux-kvm.org/wiki/images/f/fe/2010-forum-armbru-qdev.pdf
I can try to assist and advice, just ask.
