[Xen-devel] Re: [Qemu-devel] [PATCH RFC V3 12/12] xen: Add a Xen specific ACPI Implementation to target-xen

[Blue Swirl <blauwirbel@xxxxxxxxx>]

On Fri, Sep 17, 2010 at 11:15 AM,  <anthony.perard@xxxxxxxxxx> wrote:
> From: Anthony PERARD <anthony.perard@xxxxxxxxxx>
>
> Xen currently uses a different BIOS (hvmloader + rombios) therefore the
> Qemu acpi_piix4 implementation wouldn't work correctly with Xen.
> We plan on fixing this properly but at the moment we are just adding a
> new Xen specific acpi_piix4 implementation.
> This patch is optional; without it the VM boots but it cannot shutdown
> properly or go to S3.
>
> Signed-off-by: Anthony PERARD <anthony.perard@xxxxxxxxxx>
> Signed-off-by: Stefano Stabellini <stefano.stabellini@xxxxxxxxxxxxx>
> ---
> ÂMakefile.target   |  Â1 +
> Âhw/xen_acpi_piix4.c | Â405 
> +++++++++++++++++++++++++++++++++++++++++++++++++++
> Âhw/xen_common.h   |  Â3 +
> Âhw/xen_machine_fv.c | Â Â6 +-
> Â4 files changed, 410 insertions(+), 5 deletions(-)
> Âcreate mode 100644 hw/xen_acpi_piix4.c
>
> diff --git a/Makefile.target b/Makefile.target
> index ea14393..db7f96b 100644
> --- a/Makefile.target
> +++ b/Makefile.target
> @@ -189,6 +189,7 @@ obj-$(CONFIG_NO_XEN) += xen-stub.o
> Â# xen full virtualized machine
> Âobj-$(CONFIG_XEN) += xen_machine_fv.o
> Âobj-$(CONFIG_XEN) += xen_platform.o
> +obj-$(CONFIG_XEN) += xen_acpi_piix4.o
>
> Â# USB layer
> Âobj-$(CONFIG_USB_OHCI) += usb-ohci.o
> diff --git a/hw/xen_acpi_piix4.c b/hw/xen_acpi_piix4.c
> new file mode 100644
> index 0000000..f4792f2
> --- /dev/null
> +++ b/hw/xen_acpi_piix4.c
> @@ -0,0 +1,405 @@
> + /*
> + * PIIX4 ACPI controller emulation
> + *
> + * Winston liwen Wang, winston.l.wang@xxxxxxxxx
> + * Copyright (c) 2006 , Intel Corporation.
> + *
> + * Permission is hereby granted, free of charge, to any person obtaining a 
> copy
> + * of this software and associated documentation files (the "Software"), to 
> deal
> + * in the Software without restriction, including without limitation the 
> rights
> + * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
> + * copies of the Software, and to permit persons to whom the Software is
> + * furnished to do so, subject to the following conditions:
> + *
> + * The above copyright notice and this permission notice shall be included in
> + * all copies or substantial portions of the Software.
> + *
> + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
> + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
> + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
> + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
> + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING 
> FROM,
> + * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
> + * THE SOFTWARE.
> + */
> +
> +#include "hw.h"
> +#include "pc.h"
> +#include "pci.h"
> +#include "sysemu.h"
> +#include "acpi.h"
> +
> +#include "xen_backend.h"
> +#include "xen_common.h"
> +#include "qemu-log.h"
> +
> +#include <xen/hvm/ioreq.h>
> +#include <xen/hvm/params.h>
> +
> +#define PIIX4ACPI_LOG_ERROR 0
> +#define PIIX4ACPI_LOG_INFO 1
> +#define PIIX4ACPI_LOG_DEBUG 2
> +#define PIIX4ACPI_LOGLEVEL PIIX4ACPI_LOG_INFO
> +#define PIIX4ACPI_LOG(level, fmt, ...) do { if (level <= PIIX4ACPI_LOGLEVEL) 
> qemu_log(fmt, ## __VA_ARGS__); } while (0)
> +
> +/* Sleep state type codes as defined by the \_Sx objects in the DSDT. */
> +/* These must be kept in sync with the DSDT (hvmloader/acpi/dsdt.asl) */
> +#define SLP_TYP_S4 Â Â Â Â(6 << 10)
> +#define SLP_TYP_S3 Â Â Â Â(5 << 10)
> +#define SLP_TYP_S5 Â Â Â Â(7 << 10)
> +
> +#define ACPI_DBG_IO_ADDR Â0xb044
> +#define ACPI_PHP_IO_ADDR Â0x10c0
> +
> +#define PHP_EVT_ADD Â Â 0x0
> +#define PHP_EVT_REMOVE Â0x3
> +
> +/* The bit in GPE0_STS/EN to notify the pci hotplug event */
> +#define ACPI_PHP_GPE_BIT 3
> +
> +#define DEVFN_TO_PHP_SLOT_REG(devfn) (devfn >> 1)
> +#define PHP_SLOT_REG_TO_DEVFN(reg, hilo) ((reg << 1) | hilo)
> +
> +/* ioport to monitor cpu add/remove status */
> +#define PROC_BASE 0xaf00
> +
> +typedef struct PCIAcpiState {

PCIACPIState

> + Â ÂPCIDevice dev;
> + Â Âuint16_t pm1_control; /* pm1a_ECNT_BLK */
> + Â Âqemu_irq irq;
> + Â Âqemu_irq cmos_s3;
> +} PCIAcpiState;
> +
> +typedef struct GPEState {
> + Â Â/* GPE0 block */
> + Â Âuint8_t gpe0_sts[ACPI_GPE0_BLK_LEN / 2];
> + Â Âuint8_t gpe0_en[ACPI_GPE0_BLK_LEN / 2];
> +
> + Â Â/* CPU bitmap */
> + Â Âuint8_t cpus_sts[32];
> +
> + Â Â/* SCI IRQ level */
> + Â Âuint8_t sci_asserted;
> +
> +} GPEState;
> +
> +static GPEState gpe_state;
> +
> +static qemu_irq sci_irq;
> +
> +typedef struct AcpiDeviceState AcpiDeviceState;
> +AcpiDeviceState *acpi_device_table;

The above globals and static variable should be eliminated, see
ac4040955b1669f0aac5937f623d6587d5210679.

> +
> +static const VMStateDescription vmstate_acpi = {
> + Â Â.name = "PIIX4 ACPI",
> + Â Â.version_id = 1,
> +  Â.fields   Â= (VMStateField []) {
> + Â Â Â ÂVMSTATE_PCI_DEVICE(dev, PCIAcpiState),
> + Â Â Â ÂVMSTATE_UINT16(pm1_control, PCIAcpiState),
> + Â Â Â ÂVMSTATE_END_OF_LIST()
> + Â Â}
> +};
> +
> +static void acpiPm1Control_writeb(void *opaque, uint32_t addr, uint32_t val)

acpi_pm1_control_writeb().

> +{
> + Â ÂPCIAcpiState *s = opaque;
> + Â Âs->pm1_control = (s->pm1_control & 0xff00) | (val & 0xff);
> +}
> +
> +static uint32_t acpiPm1Control_readb(void *opaque, uint32_t addr)

Likewise.

> +{
> + Â ÂPCIAcpiState *s = opaque;
> + Â Â/* Mask out the write-only bits */
> + Â Âreturn (uint8_t)(s->pm1_control & 
> ~(ACPI_BITMASK_GLOBAL_LOCK_RELEASE|ACPI_BITMASK_SLEEP_ENABLE));

Please add spaces around '|' and break the line. Adding
ACPI_BITMASK_WRITEONLY may help.

> +}
> +
> +static void acpi_shutdown(PCIAcpiState *s, uint32_t val)
> +{
> + Â Âif (!(val & ACPI_BITMASK_SLEEP_ENABLE))
> + Â Â Â Âreturn;

braces

> +
> + Â Âswitch (val & ACPI_BITMASK_SLEEP_TYPE) {
> + Â Âcase SLP_TYP_S3:
> + Â Â Â Âqemu_system_reset();
> + Â Â Â Âqemu_irq_raise(s->cmos_s3);
> + Â Â Â Âxc_set_hvm_param(xen_xc, xen_domid, HVM_PARAM_ACPI_S_STATE, 3);
> + Â Â Â Âbreak;
> + Â Âcase SLP_TYP_S4:
> + Â Âcase SLP_TYP_S5:
> + Â Â Â Âqemu_system_shutdown_request();
> + Â Â Â Âbreak;
> + Â Âdefault:
> + Â Â Â Âbreak;
> + Â Â}
> +}
> +
> +static void acpiPm1ControlP1_writeb(void *opaque, uint32_t addr, uint32_t 
> val)
> +{
> + Â ÂPCIAcpiState *s = opaque;
> +
> + Â Âval <<= 8;
> + Â Âs->pm1_control = ((s->pm1_control & 0xff) | val) & 
> ~ACPI_BITMASK_SLEEP_ENABLE;
> +
> + Â Âacpi_shutdown(s, val);
> +}
> +
> +static uint32_t acpiPm1ControlP1_readb(void *opaque, uint32_t addr)
> +{
> + Â ÂPCIAcpiState *s = opaque;
> + Â Â/* Mask out the write-only bits */
> + Â Âreturn (uint8_t)((s->pm1_control & 
> ~(ACPI_BITMASK_GLOBAL_LOCK_RELEASE|ACPI_BITMASK_SLEEP_ENABLE)) >> 8);
> +}
> +
> +static void acpiPm1Control_writew(void *opaque, uint32_t addr, uint32_t val)
> +{
> + Â ÂPCIAcpiState *s = opaque;
> +
> + Â Âs->pm1_control = val & ~ACPI_BITMASK_SLEEP_ENABLE;
> +
> + Â Âacpi_shutdown(s, val);
> +}
> +
> +static uint32_t acpiPm1Control_readw(void *opaque, uint32_t addr)
> +{
> + Â ÂPCIAcpiState *s = opaque;
> + Â Â/* Mask out the write-only bits */
> + Â Âreturn (s->pm1_control & 
> ~(ACPI_BITMASK_GLOBAL_LOCK_RELEASE|ACPI_BITMASK_SLEEP_ENABLE));
> +}
> +
> +static void acpi_map(PCIDevice *pci_dev, int region_num,
> + Â Â Â Â Â Â Â Â Â Â uint32_t addr, uint32_t size, int type)
> +{
> + Â ÂPCIAcpiState *d = (PCIAcpiState *)pci_dev;
> +
> + Â Â/* Byte access */
> + Â Âregister_ioport_write(addr + 4, 1, 1, acpiPm1Control_writeb, d);
> + Â Âregister_ioport_read(addr + 4, 1, 1, acpiPm1Control_readb, d);
> + Â Âregister_ioport_write(addr + 4 + 1, 1, 1, acpiPm1ControlP1_writeb, d);
> + Â Âregister_ioport_read(addr + 4 +1, 1, 1, acpiPm1ControlP1_readb, d);
> +
> + Â Â/* Word access */
> + Â Âregister_ioport_write(addr + 4, 2, 2, acpiPm1Control_writew, d);
> + Â Âregister_ioport_read(addr + 4, 2, 2, acpiPm1Control_readw, d);
> +}
> +
> +static inline int test_bit(uint8_t *map, int bit)
> +{
> + Â Âreturn ( map[bit / 8] & (1 << (bit % 8)) );
> +}
> +
> +static inline void set_bit(uint8_t *map, int bit)
> +{
> + Â Âmap[bit / 8] |= (1 << (bit % 8));
> +}
> +
> +static inline void clear_bit(uint8_t *map, int bit)
> +{
> + Â Âmap[bit / 8] &= ~(1 << (bit % 8));
> +}
> +
> +static void acpi_dbg_writel(void *opaque, uint32_t addr, uint32_t val)
> +{
> + Â ÂPIIX4ACPI_LOG(PIIX4ACPI_LOG_DEBUG, "ACPI: DBG: 0x%08x\n", val);
> + Â ÂPIIX4ACPI_LOG(PIIX4ACPI_LOG_INFO, "ACPI:debug: write addr=0x%x, 
> val=0x%x.\n", addr, val);
> +}
> +
> +/* GPEx_STS occupy 1st half of the block, while GPEx_EN 2nd half */
> +static uint32_t gpe_sts_read(void *opaque, uint32_t addr)
> +{
> + Â ÂGPEState *s = opaque;
> +
> + Â Âreturn s->gpe0_sts[addr - ACPI_GPE0_BLK_ADDRESS];
> +}
> +
> +/* write 1 to clear specific GPE bits */
> +static void gpe_sts_write(void *opaque, uint32_t addr, uint32_t val)
> +{
> + Â ÂGPEState *s = opaque;
> + Â Âint hotplugged = 0;
> +
> + Â ÂPIIX4ACPI_LOG(PIIX4ACPI_LOG_DEBUG, "gpe_sts_write: addr=0x%x, 
> val=0x%x.\n", addr, val);
> +
> + Â Âhotplugged = test_bit(&s->gpe0_sts[0], ACPI_PHP_GPE_BIT);
> + Â Âs->gpe0_sts[addr - ACPI_GPE0_BLK_ADDRESS] &= ~val;
> + Â Âif ( s->sci_asserted &&
> + Â Â Â Â hotplugged &&
> + Â Â Â Â !test_bit(&s->gpe0_sts[0], ACPI_PHP_GPE_BIT)) {
> + Â Â Â ÂPIIX4ACPI_LOG(PIIX4ACPI_LOG_INFO, "Clear the GPE0_STS bit for ACPI 
> hotplug & deassert the IRQ.\n");
> + Â Â Â Âqemu_irq_lower(sci_irq);
> + Â Â}
> +
> +}
> +
> +static uint32_t gpe_en_read(void *opaque, uint32_t addr)
> +{
> + Â ÂGPEState *s = opaque;
> +
> + Â Âreturn s->gpe0_en[addr - (ACPI_GPE0_BLK_ADDRESS + ACPI_GPE0_BLK_LEN / 
> 2)];
> +}
> +
> +/* write 0 to clear en bit */
> +static void gpe_en_write(void *opaque, uint32_t addr, uint32_t val)
> +{
> + Â ÂGPEState *s = opaque;
> + Â Âint reg_count;
> +
> + Â ÂPIIX4ACPI_LOG(PIIX4ACPI_LOG_DEBUG, "gpe_en_write: addr=0x%x, 
> val=0x%x.\n", addr, val);
> + Â Âreg_count = addr - (ACPI_GPE0_BLK_ADDRESS + ACPI_GPE0_BLK_LEN / 2);
> + Â Âs->gpe0_en[reg_count] = val;
> + Â Â/* If disable GPE bit right after generating SCI on it,
> + Â Â * need deassert the intr to avoid redundant intrs
> + Â Â */
> + Â Âif ( s->sci_asserted &&
> + Â Â Â Â reg_count == (ACPI_PHP_GPE_BIT / 8) &&
> + Â Â Â Â !(val & (1 << (ACPI_PHP_GPE_BIT % 8))) ) {
> + Â Â Â ÂPIIX4ACPI_LOG(PIIX4ACPI_LOG_INFO, "deassert due to disable GPE 
> bit.\n");
> + Â Â Â Âs->sci_asserted = 0;
> + Â Â Â Âqemu_irq_lower(sci_irq);
> + Â Â}
> +
> +}
> +
> +static const VMStateDescription vmstate_gpe = {
> + Â Â.name = "gpe",
> + Â Â.version_id = 2,
> + Â Â.minimum_version_id = 2,
> + Â Â.minimum_version_id_old = 2,
> + Â Â.fields = (VMStateField []) {
> + Â Â Â ÂVMSTATE_BUFFER(gpe0_sts, GPEState),
> + Â Â Â ÂVMSTATE_BUFFER(gpe0_en, GPEState),
> + Â Â Â ÂVMSTATE_UINT8(sci_asserted, GPEState),
> + Â Â Â ÂVMSTATE_END_OF_LIST()
> + Â Â}
> +};
> +
> +static uint32_t gpe_cpus_readb(void *opaque, uint32_t addr)
> +{
> + Â Âuint32_t val = 0;
> + Â ÂGPEState *g = opaque;
> +
> + Â Âswitch (addr) {
> + Â Â Â Âcase PROC_BASE ... PROC_BASE+31:
> + Â Â Â Â Â Âval = g->cpus_sts[addr - PROC_BASE];

break;

> + Â Â Â Âdefault:
> + Â Â Â Â Â Âbreak;
> + Â Â}
> +
> + Â Âreturn val;
> +}
> +
> +static void gpe_cpus_writeb(void *opaque, uint32_t addr, uint32_t val)
> +{
> + Â Â/* GPEState *g = opaque; */
> +
> + Â Âswitch (addr) {
> + Â Â Â Âcase PROC_BASE ... PROC_BASE + 31:
> + Â Â Â Â Â Â/* don't allow to change cpus_sts from inside a guest */
> + Â Â Â Â Â Âbreak;
> + Â Â Â Âdefault:
> + Â Â Â Â Â Âbreak;
> + Â Â}
> +}
> +
> +static void gpe_acpi_init(void)
> +{
> + Â ÂGPEState *s = &gpe_state;
> + Â Âmemset(s, 0, sizeof(GPEState));
> +
> + Â Âs->cpus_sts[0] = 1;
> +
> + Â Âregister_ioport_read(PROC_BASE, 32, 1, Âgpe_cpus_readb, s);
> + Â Âregister_ioport_write(PROC_BASE, 32, 1, gpe_cpus_writeb, s);
> +
> + Â Âregister_ioport_read(ACPI_GPE0_BLK_ADDRESS,
> + Â Â Â Â Â Â Â Â Â Â Â Â ACPI_GPE0_BLK_LEN / 2,
> + Â Â Â Â Â Â Â Â Â Â Â Â 1,
> + Â Â Â Â Â Â Â Â Â Â Â Â gpe_sts_read,
> + Â Â Â Â Â Â Â Â Â Â Â Â s);
> + Â Âregister_ioport_read(ACPI_GPE0_BLK_ADDRESS + ACPI_GPE0_BLK_LEN / 2,
> + Â Â Â Â Â Â Â Â Â Â Â Â ACPI_GPE0_BLK_LEN / 2,
> + Â Â Â Â Â Â Â Â Â Â Â Â 1,
> + Â Â Â Â Â Â Â Â Â Â Â Â gpe_en_read,
> + Â Â Â Â Â Â Â Â Â Â Â Â s);
> +
> + Â Âregister_ioport_write(ACPI_GPE0_BLK_ADDRESS,
> + Â Â Â Â Â Â Â Â Â Â Â Â ÂACPI_GPE0_BLK_LEN / 2,
> + Â Â Â Â Â Â Â Â Â Â Â Â Â1,
> + Â Â Â Â Â Â Â Â Â Â Â Â Âgpe_sts_write,
> + Â Â Â Â Â Â Â Â Â Â Â Â Âs);
> + Â Âregister_ioport_write(ACPI_GPE0_BLK_ADDRESS + ACPI_GPE0_BLK_LEN / 2,
> + Â Â Â Â Â Â Â Â Â Â Â Â ÂACPI_GPE0_BLK_LEN / 2,
> + Â Â Â Â Â Â Â Â Â Â Â Â Â1,
> + Â Â Â Â Â Â Â Â Â Â Â Â Âgpe_en_write,
> + Â Â Â Â Â Â Â Â Â Â Â Â Âs);
> +
> + Â Âvmstate_register(NULL, 0, &vmstate_gpe, s);
> +}
> +
> +static int piix4_pm_xen_initfn(PCIDevice *dev)
> +{
> + Â ÂPCIAcpiState *s = DO_UPCAST(PCIAcpiState, dev, dev);
> + Â Âuint8_t *pci_conf;
> +
> + Â Âpci_conf = s->dev.config;
> + Â Âpci_config_set_vendor_id(pci_conf, PCI_VENDOR_ID_INTEL);
> + Â Âpci_config_set_device_id(pci_conf, PCI_DEVICE_ID_INTEL_82371AB_3);
> + Â Âpci_conf[0x08] = 0x01; Â/* B0 stepping */
> + Â Âpci_conf[0x09] = 0x00; Â/* base class */
> + Â Âpci_config_set_class(pci_conf, PCI_CLASS_BRIDGE_OTHER);
> + Â Âpci_conf[PCI_HEADER_TYPE] = PCI_HEADER_TYPE_NORMAL; /* header_type */
> + Â Âpci_conf[0x3d] = 0x01; Â/* Hardwired to PIRQA is used */
> +
> + Â Â/* PMBA POWER MANAGEMENT BASE ADDRESS, hardcoded to 0x1f40
> + Â Â * to make shutdown work for IPF, due to IPF Guest Firmware
> + Â Â * will enumerate pci devices.
> + Â Â *
> + Â Â * TODO: Âif Guest Firmware or Guest OS will change this PMBA,
> + Â Â * More logic will be added.
> + Â Â */
> + Â Âpci_conf[0x40] = 0x41; /* Special device-specific BAR at 0x40 */
> + Â Âpci_conf[0x41] = 0x1f;
> + Â Âpci_conf[0x42] = 0x00;
> + Â Âpci_conf[0x43] = 0x00;
> +
> + Â Âs->pm1_control = ACPI_BITMASK_SCI_ENABLE;

Please extract this line to a reset function.

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel