Qemu device prototyping

QEMU AND DEVICE EMULATION
Yan Vugenﬁrer, yan@daynix.com
Daynix Computing LTD

AGENDA
• Introduction to QEMU
• Development environment
• Examples of existing devices
• Adding a new device to
QEMU

WHAT IS QEMU?
• Open source project (GPLv2.0+ license)
• Machine emulator
• Virtualizer
• Works together with KVM and Xen

WHY SHOULDYOU CARE?
• Open source
• Easy to add additional devices
• Emulates different HW architectures
• Can be used for SW development long before
HW is ready

GET IT NOW
• Project website: http://wiki.qemu.org/Main_Page
• Code repository: git clone git://git.qemu-
project.org/qemu.git

PREPARING DEVELOPMENT
ENVIRONMENT
• Installing the packages on Ubuntu
• sudo apt-get update
• sudo apt-get install git
• sudo apt-get install build-essential
• sudo apt-get install pkg-conﬁg
• sudo apt-get install zlib zlib-dev
• sudo apt-get install zlib1g-dev zlib1g
• sudo apt-get install glib2.0
• sudo apt-get install libpixman-1-0
• sudo apt-get install libtool
• sudo apt-get install dh-autoreconf
• sudo apt-get install bridge-utils

COMPILING QEMU
• git clone https://github.com/qemu/qemu.git
• cd qemu
• git submodule update --init pixman
• ./conﬁgure --disable-docs —target-list=x86_64-softmmu
• make -j 8
• Skip if development version only: make install

NETWORK CONFIGURATION
Linux host
Virtual machine
QEMU process
NIC
Linux bridge
Physical NIC

NETWORK CONFIGURATION
• On the host edit /etc/network/interfaces
# The loopback network interface
auto lo
iface lo inet loopback
# The primary network interface
#auto eth0
#iface eth0 inet dhcp
iface eth0 inet manual
auto br0
iface br0 inet dhcp
bridge_ports eth0

#!/bin/sh
switch=br0
/sbin/ifconfig $1 promisc 0.0.0.0
/sbin/brctl addif ${switch} $1
QEMU-IFUP SCRIPT
• Create qemu-ifup script with following content

STORAGE
• Download Ubuntu server - http://
www.ubuntu.com/download/server
• CreateVM image
• qemu-img create -f qcow ubuntu.qcow 8G

RUNNINGYOU FIRSTVIRTUAL
MACHINE
./qemu/x86_64-softmmu/qemu-system-x86_64
-M pc -name Test_VM -smp 2 -m 512M
-drive ﬁle=/home/qemu/images/ubuntu.qcow,if=ide
-usbdevice tablet
-boot order=cdn,once=c,menu=on
-netdev tap,id=hostnet1,script=/home/qemu/dev/qemu-
ifup,ifname=testvm_nic
-device e1000,netdev=hostnet1,mac=56:54:46:6b:64:22,bus=pci.
0,id=e1000_01
-cdrom /ISO/ubuntu-14.04.1-server-amd64.iso
-vnc :5

CONNECTINGTOVM
• Use theVNC viewer of your choice
• ssh after you conﬁgured networking

CODETREE
• cd qemu/HW

QEMU DEVICE MODEL
• QDev device model abstraction
• Tree of devices connected by buses
• Represented by DeviceState and BusState
• Devices have common API
• Devices have properties
• Check include/hw/qdev-core.h for more info

EXAMPLES OF EXISTING
DEVICES
• e1000
• qemu/hw/net/e1000.c
• virtio family -base code
• qemu/hw/virtio/virtio-pci.c, qemu/hw/virtio/virtio-rng.c
• virtio-net
• qemu/hw/net/virtio-net.c
• vmxnet3
• qemu/hw/net/vmxnet3.c

ADDING NEW DEVICE
• Basic source ﬁle for the device
• Enable the compilation of the device
• Command line options
• Step by step enhancement of the device

ADDING NEW DEVICE
• In qemu/HW/net/ let’s create devix.c
• Add CONFIG_DEVIX_PCI option to default-conﬁgs/
pci.mak
• Add devix.o to hw/net/Makeﬁle.objs

COMPILATION
diff --git a/default-configs/pci.mak b/default-configs/
pci.mak
index 91b1e92..fcf2cf2 100644
--- a/default-configs/pci.mak
+++ b/default-configs/pci.mak
@@ -30,3 +30,4 @@ CONFIG_IPACK=y
CONFIG_WDT_IB6300ESB=y
CONFIG_PCI_TESTDEV=y
CONFIG_NVME_PCI=y
+CONFIG_DEVIX_PCI=y

COMPILATION
diff --git a/hw/net/Makefile.objs b/hw/net/Makefile.objs
index ea93293..7800755 100644
--- a/hw/net/Makefile.objs
+++ b/hw/net/Makefile.objs
@@ -10,6 +10,7 @@ common-obj-$(CONFIG_E1000_PCI) += e1000.o
common-obj-$(CONFIG_RTL8139_PCI) += rtl8139.o
common-obj-$(CONFIG_VMXNET3_PCI) += vmxnet_tx_pkt.o
vmxnet_rx_pkt.o
common-obj-$(CONFIG_VMXNET3_PCI) += vmxnet3.o
+common-obj-$(CONFIG_DEVIX_PCI) += devix.o

REGISTERYOUR DEVICETYPE
static const TypeInfo devix_info = {
.name = TYPE_DEVIX,
.parent = TYPE_PCI_DEVICE,
.instance_size = sizeof(DEVIXState),
.class_init = devix_class_init,
.instance_init = devix_instance_init,
};
static void devix_register_types(void)
{
DVX_CBPRN("devix_register_types called...");
type_register_static(&devix_info);
}
type_init(devix_register_types)

DEVICETYPE INITIALIZATION
static void devix_class_init(ObjectClass *class, void *data)
{
DeviceClass *dc = DEVICE_CLASS(class);
PCIDeviceClass *c = PCI_DEVICE_CLASS(class);
c->init = devix_pci_init;
c->exit = devix_pci_uninit;
c->vendor_id = PCI_VENDOR_ID_DAYNIX;
c->device_id = PCI_DEVICE_ID_DAYNIX_DEVIX;
c->revision = PCI_DEVICE_ID_DAYNIX_DEVIX_REVISION;
c->class_id = PCI_CLASS_NETWORK_ETHERNET;
c->subsystem_vendor_id = PCI_VENDOR_ID_DAYNIX;
c->subsystem_id = PCI_DEVICE_ID_DAYNIX_DEVIX;
c->config_write = devix_write_config;
c->config_read = devix_read_config;
dc->desc = "Daynix educational device v1";
dc->reset = devix_qdev_reset;
dc->vmsd = &vmstate_devix;
dc->props = devix_properties;
set_bit(DEVICE_CATEGORY_NETWORK, dc->categories);
}

PCI CONFIGURATION SPACE
ACCESS CALLBACKS
static void uint32_t devix_read_config(PCIDevice *pci_dev,
uint32_t address, int len)
{
return pci_default_read_config(pci_dev, address, len);
}
static void
devix_write_config(PCIDevice *pci_dev, uint32_t address, uint32_t val,
int len)
{
pci_default_write_config(pci_dev, address, val, len);
}

• At this point we have PCI device
• No BARs or Interrupts… yet

VM POINT OFVIEW
00:04.0 Ethernet controller: Device 9696:1234
(rev 01)
Subsystem: Device 9696:1234
Physical Slot: 4
Control: I/O+ Mem+ BusMaster- SpecCycle-
MemWINV- VGASnoop- ParErr- Stepping- SERR+
FastB2B- DisINTx-
Status: Cap- 66MHz- UDF- FastB2B- ParErr-
DEVSEL=fast >TAbort- <TAbort- <MAbort- >SERR-
<PERR- INTx-
• sudo lspci -vv

• Register IO and memory space
• Conﬁgure interrupts
• Our own device initializations
PCI INITIALIZATION

MEMORY REGIONS
typedef struct {
PCIDevice parent_obj;
MemoryRegion bar0;
MemoryRegion bar1;
} DEVIXState;
• Add placeholders for memory regions into device
state structure

REGISTER MEMORY REGIONS
static int devix_pci_init(PCIDevice *pci_dev)
{
DeviceState *dev = DEVICE(pci_dev);
DEVIXState *s = DEVIX(pci_dev);
DVX_CBPRN("Starting init...");
memory_region_init_io(&s->bar0, OBJECT(s), &b0_ops, s,
"devix-b0", DEVIX_BAR0_SIZE);
pci_register_bar(pci_dev, DEVIX_BAR0_IDX,
PCI_BASE_ADDRESS_SPACE_MEMORY, &s->bar0);
memory_region_init_io(&s->bar1, OBJECT(s), &b1_ops, s,
"devix-b1", DEVIX_BAR1_SIZE);
pci_register_bar(pci_dev, DEVIX_BAR1_IDX,
PCI_BASE_ADDRESS_SPACE_IO, &s->bar1);
devix_net_init(s);
return 0;
}

ACCESS CALLBACKS
static const MemoryRegionOps b0_ops = {
.read = devix_io_bar0_read,
.write = devix_io_bar0_write,
.endianness = DEVICE_LITTLE_ENDIAN,
.impl = {
.min_access_size = 4,
.max_access_size = 4,
},
};

ACCESS CALLBACKS - WRITE
static void
devix_io_bar0_write(void *opaque, hwaddr addr,
uint64_t val, unsigned size)
{
DEVIXState *s = opaque;
DVX_WRPRN("BAR0 write [%" PRIx64 "] = %"
PRIx64 ", size %d",
(uint64_t) addr, val, size);
}
• Size
• Address
• Value

ACCESS CALLBACKS - READ
static uint64_t
devix_io_bar0_read(void *opaque, hwaddr addr,
unsigned size)
{
DEVIXState *s = opaque;
uint64_t ret = 0;
DVX_CBPRN("Read BAR0 [%" PRIx64 "], size
%d",(uint64_t) addr, size);
return ret; /* Returns value of the read
register */
}
• Size
• Address
• Return the
value

VM POINT OFVIEW
(rev 01)
Physical Slot: 4
Control: I/O+ Mem+ BusMaster- SpecCycle- MemWINV-
VGASnoop- ParErr- Stepping- SERR+ FastB2B- DisINTx-
DEVSEL=fast >TAbort- <TAbort- <MAbort- >SERR- <PERR-
INTx-
Region 0: Memory at febd2000 (32-bit, non-
prefetchable) [size=4K]
Region 1: I/O ports at c000 [size=512]
• sudo lspci -vv

LEGACY INTERRUPTS
• Add interrupt line in pci_init
/* Interrupt pin A */
pci_dev->config[PCI_INTERRUPT_PIN] = 0x01;

LEGACY INTERRUPTS
• Call to void pci_irq_assert(PCIDevice *pci_dev)
when you want to raise interrupt
• Call void pci_irq_deassert(PCIDevice *pci_dev) to
de-assert interrupt from one of your registers
callback depending on clear interrupt logic

VM POINT OFVIEW
(rev 01)
Physical Slot: 4
Control: I/O+ Mem+ BusMaster- SpecCycle- MemWINV-
VGASnoop- ParErr- Stepping- SERR+ FastB2B- DisINTx-
DEVSEL=fast >TAbort- <TAbort- <MAbort- >SERR- <PERR-
INTx-
Interrupt: pin A routed to IRQ 11
Region 0: Memory at febd2000 (32-bit, non-
prefetchable) [size=4K]
Region 1: I/O ports at c000 [size=512]
• sudo lspci -vv

“DMA”
• Actually guest memory access
• Synchronous
• void cpu_physical_memory_read(hwaddr addr, void *buf, int len)
• void cpu_physical_memory_write(hwaddr addr, const void *buf, int len)
• iov_xxx functions
• Check include/exec/cpu-common.h for more access functions

WHAT’S NEXT?
• Add network back end
• Add MSI and MSI-x support
• Littlebig endian
considerations

LINKS
• Project website: http://wiki.qemu.org/Main_Page
• Code repository: git clone git://git.qemu-project.org/
qemu.git
• QEMU new device model - http://www.linux-kvm.org/
wiki/images/f/fe/2010-forum-armbru-qdev.pdf
• Daynix - www.daynix.com

Qemu device prototyping

More Related Content

What's hot (20)

Similar to Qemu device prototyping (20)

More from Yan Vugenfirer (14)

Recently uploaded (20)

Qemu device prototyping