Re: [Qemu-devel] [RFC][PATCH 2/2] hw: add Embedded Controller chip emulation

["Andreas Färber" <afaerber@suse.de>]

Am 17.04.2013 09:23, schrieb liguang:
> this work implemented Embedded Controller chip emulation
> which was defined at ACPI SEPC v5 chapter 12:
> "ACPI Embedded Controller Interface Specification"
> 
> commonly Embedded Controller will emulate keyboard,
> mouse, handle ACPI defined operations and some
> low-speed devices like SMbus.
> 
> Signed-off-by: liguang <lig.fnst@cn.fujitsu.com>
> ---
>  hw/ec.c |  113 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++

hw/misc/ec.c? hw/i386/ec.c? Not directly in hw/ anymore.

>  hw/ec.h |   20 +++++++++++
>  2 files changed, 133 insertions(+), 0 deletions(-)
>  create mode 100644 hw/ec.c
>  create mode 100644 hw/ec.h
> 
> diff --git a/hw/ec.c b/hw/ec.c
> new file mode 100644
> index 0000000..69c92cf
> --- /dev/null
> +++ b/hw/ec.c
> @@ -0,0 +1,113 @@
> +#include "ec.h"
> +#include "hw/hw.h"
> +#include "hw/isa/isa.h"
> +#include "sysemu/sysemu.h"
> +
> +#define TYPE_EC_DEV

Missing string value.

> +#define EC_DEV(obj) \
> +    OBJECT_CHECK(ECState, (obj), TYPE_EC_DEV)
> +
> +static char ec_acpi_space[EC_ACPI_SPACE_SIZE] = {0};
> +
> +typedef struct ECState {
> +    ISADevice dev;

parent_obj and white line please.

> +    char cmd;
> +    char status;
> +    char data;
> +    char irq;
> +    char buf;

char seems a bit unsafe here, since it might be signed or unsigned.
Suggest uint8_t.

> +    MemoryRegion io;
> +} ECState;
> +
> +
> +static void io62_write(void *opaque, hwaddr addr, uint64_t val, unsigned size)
> +{
> +    ECState *s = opaque;
> +    char tmp = val & 0xff;
> +
> +    if (s->status & EC_ACPI_CMD) {
> +        s->buf = tmp;
> +        s->status &= ~EC_ACPI_CMD;
> +    } else {
> +        if (tmp < EC_ACPI_SPACE_SIZE) {
> +            ec_acpi_space[s->buf] = tmp;
> +        }
> +    }
> +}
> +
> +static uint64_t io62_read(void *opaque, hwaddr addr, unsigned size)
> +{
> +    return s->data;
> +}
> +
> +static void io66_write(void *opaque, hwaddr addr, uint64_t val, unsigned size)
> +{
> +    ECState *s = opaque;
> +
> +    s->status = EC_ACPI_CMD | EC_ACPI_IBF;
> +
> +    switch (val & 0xff) {
> +    case EC_ACPI_CMD_READ:
> +    case EC_ACPI_CMD_WRITE:
> +    case EC_ACPI_CMD_BURST_EN:
> +        s->statu |= EC_ACPI_BST;

s->status

> +    case EC_ACPI_CMD_BURST_DN:
> +        s->statu &= ~EC_ACPI_BST;

s->status

> +    case EC_ACPI_CMD_QUERY:
> +        s->cmd = val & 0xff;
> +    case default:
> +        break;
> +    }
> +}
> +
> +static uint64_t io66_read(void *opaque, hwaddr addr, unsigned size)
> +{
> +    ECState *s = opaque;
> +
> +    return s->status;
> +}
> +
> +static const MemoryRegionOps io62_io_ops = {
> +    .write = io62_write,
> +    .read = io62_read,
> +    .endianness = DEVICE_NATIVE_ENDIAN,
> +    .impl = {
> +        .min_access_size = 1,
> +        .max_access_size = 1,
> +    },
> +};
> +
> +static const MemoryRegionOps io66_io_ops = {
> +    .write = io66_write,
> +    .read = io66_read,
> +    .endianness = DEVICE_NATIVE_ENDIAN,
> +    .impl = {
> +        .min_access_size = 1,
> +        .max_access_size = 1,
> +    },
> +};
> +
> +static void ec_realizefn(DeviceState *dev, Error **err)
> +{
> +    ISADevice *isadev = ISA_DEVICE(dev);
> +    ECState *s = EC_DEV(dev);
> +
> +    isa_init_irq(isadev, &s->irq, 0xb);
> +

> +    memory_region_init_io(&s->io, &io62_io_ops, NULL, "ec-acpi-data", 1);
> +    isa_register_ioport(isadev, &s->io, 0x62);
> +
> +    memory_region_init_io(&s->io, &io66_io_ops, NULL, "ec-acpi-cmd", 1);
> +    isa_register_ioport(isadev, &s->io, 0x66);

Since you are not defining any properties, all of this could go into a
.instance_init function.

But I am glad to see a realize function being used. :)

> +
> +    s->status = 0;
> +    s->data = 0;

This is probably not necessary here, since all fields get
zero-initialized. It should rather go into a reset function.

> +}
> +
> +static void ec_class_initfn(ObjectClass *klass, void *data)
> +{
> +    DeviceClass *dc = DEVICE_CLASS(klass);
> +
> +    dc->realize = ec_realizefn;
> +    dc->no_user = 1;
> +}

This file ends with an unused static function, it's missing type
registration.

I figure this device should rather be a specific (real) model, reflected
in a particular name, similar to how we have different watchdog devices
rather than a "watchdog" device.

> diff --git a/hw/ec.h b/hw/ec.h
> new file mode 100644
> index 0000000..110ce04
> --- /dev/null
> +++ b/hw/ec.h
> @@ -0,0 +1,20 @@
> +#inndef __EC_H

#ifndef and no leading underscores, please.

> +#define __EC_H
> +
> +#define EC_ACPI_SPACE_SIZE 0x80
> +
> +#define EC_ACPI_CMD_PORT 0x66
> +#define EC_ACPI_DATA_PORT 0x62
> +
> +#define EC_ACPI_OBF 0x1
> +#define EC_ACPI_IBF 0x2
> +#define EC_ACPI_CMD 0x8
> +#define EC_ACPI_BST 0x10
> +
> +#define EC_ACPI_CMD_READ 0x80
> +#define EC_ACPI_CMD_WRITE 0x81
> +#define EC_ACPI_BURST_EN 0x82
> +#define EC_ACPI_BURST_DN 0x83
> +#define EC_ACPI_CMD_QUERY 0x84
> +
> +#endif

Regards,
Andreas

-- 
SUSE LINUX Products GmbH, Maxfeldstr. 5, 90409 Nürnberg, Germany
GF: Jeff Hawn, Jennifer Guild, Felix Imendörffer; HRB 16746 AG Nürnberg