Re: [PATCH 01/14] sysbus: Remove sysbus_mmio_unmap

Re: [PATCH 01/14] sysbus: Remove sysbus_mmio_unmap

[Peter Maydell]

On Thu, 23 Jun 2022 at 11:59, Peter Delevoryas <pdel@fb.com> wrote:
>
> Cedric removed usage of this function in a previous commit.
>
> Fixes: 981b1c6266c6 ("spapr/xive: rework the mapping the KVM memory regions")
> Signed-off-by: Peter Delevoryas <pdel@fb.com>

We only added this function for the XIVE in the first place
so I guess it makes sense to remove it now it's unused.
(People doing complicated stuff that needs to unmap MRs
should probably not be using sysbus_mmio_map()/unmap()
anyway...)

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>

-- PMM

Re: [PATCH 06/14] aspeed: Add system-memory QOM link to SoC

[Peter Maydell]

On Thu, 23 Jun 2022 at 12:31, Peter Delevoryas <pdel@fb.com> wrote:
>
> Right now it's just defined as the regular global system memory. If we
> migrate all the SoC code to use this property instead of directly calling
> get_system_memory(), then we can restrict the memory container for the SoC,
> which will be useful for multi-SoC machines.
>
> Signed-off-by: Peter Delevoryas <pdel@fb.com>

>  static Property aspeed_soc_properties[] = {
> +    DEFINE_PROP_LINK("system-memory", AspeedSoCState, system_memory,
> +                     TYPE_MEMORY_REGION, MemoryRegion *),

To the extent that we have a convention, we tend to call this
property on an SoC or CPU "memory", I think. (Better suggestions
welcome...)

-- PMM

Re: [PATCH 11/14] aspeed: Switch to create_unimplemented_device_in

[Peter Maydell]

On Thu, 23 Jun 2022 at 13:04, Peter Delevoryas <pdel@fb.com> wrote:
>
> Signed-off-by: Peter Delevoryas <pdel@fb.com>
> ---
>  hw/arm/aspeed_ast10x0.c | 10 ++++------
>  hw/arm/aspeed_ast2600.c | 19 ++++++++++---------
>  hw/arm/aspeed_soc.c     |  9 +++++----
>  3 files changed, 19 insertions(+), 19 deletions(-)
>
> diff --git a/hw/arm/aspeed_ast10x0.c b/hw/arm/aspeed_ast10x0.c
> index d259d30fc0..4e6688cc68 100644
> --- a/hw/arm/aspeed_ast10x0.c
> +++ b/hw/arm/aspeed_ast10x0.c
> @@ -158,12 +158,10 @@ static void aspeed_soc_ast1030_realize(DeviceState *dev_soc, Error **errp)
>      }
>
>      /* General I/O memory space to catch all unimplemented device */
> -    create_unimplemented_device("aspeed.sbc",
> -                                sc->memmap[ASPEED_DEV_SBC],
> -                                0x40000);
> -    create_unimplemented_device("aspeed.io",
> -                                sc->memmap[ASPEED_DEV_IOMEM],
> -                                ASPEED_SOC_IOMEM_SIZE);
> +    create_unimplemented_device_in("aspeed.sbc", sc->memmap[ASPEED_DEV_SBC],
> +                                   0x40000, s->system_memory);
> +    create_unimplemented_device_in("aspeed.io", sc->memmap[ASPEED_DEV_IOMEM],
> +                                   ASPEED_SOC_IOMEM_SIZE, s->system_memory);

This is SoC code, so it should probably be handling its unimplemented
devices by creating and mapping TYPE_UNIMPLEMENTED_DEVICE child
objects directly, the same way it handles all its other child devices.

thanks
-- PMM

Re: [PATCH 01/14] sysbus: Remove sysbus_mmio_unmap

[Cédric Le Goater]

On 6/23/22 11:58, Peter Delevoryas wrote:
> Cedric removed usage of this function in a previous commit.
> 
> Fixes: 981b1c6266c6 ("spapr/xive: rework the mapping the KVM memory regions")
> Signed-off-by: Peter Delevoryas <pdel@fb.com>

Reviewed-by: Cédric Le Goater <clg@kaod.org>

Thanks,

C.

> ---
>   hw/core/sysbus.c    | 10 ----------
>   include/hw/sysbus.h |  1 -
>   2 files changed, 11 deletions(-)
> 
> diff --git a/hw/core/sysbus.c b/hw/core/sysbus.c
> index 05c1da3d31..637e18f247 100644
> --- a/hw/core/sysbus.c
> +++ b/hw/core/sysbus.c
> @@ -154,16 +154,6 @@ static void sysbus_mmio_map_common(SysBusDevice *dev, int n, hwaddr addr,
>       }
>   }
>   
> -void sysbus_mmio_unmap(SysBusDevice *dev, int n)
> -{
> -    assert(n >= 0 && n < dev->num_mmio);
> -
> -    if (dev->mmio[n].addr != (hwaddr)-1) {
> -        memory_region_del_subregion(get_system_memory(), dev->mmio[n].memory);
> -        dev->mmio[n].addr = (hwaddr)-1;
> -    }
> -}
> -
>   void sysbus_mmio_map(SysBusDevice *dev, int n, hwaddr addr)
>   {
>       sysbus_mmio_map_common(dev, n, addr, false, 0);
> diff --git a/include/hw/sysbus.h b/include/hw/sysbus.h
> index 3564b7b6a2..153ef20695 100644
> --- a/include/hw/sysbus.h
> +++ b/include/hw/sysbus.h
> @@ -82,7 +82,6 @@ qemu_irq sysbus_get_connected_irq(SysBusDevice *dev, int n);
>   void sysbus_mmio_map(SysBusDevice *dev, int n, hwaddr addr);
>   void sysbus_mmio_map_overlap(SysBusDevice *dev, int n, hwaddr addr,
>                                int priority);
> -void sysbus_mmio_unmap(SysBusDevice *dev, int n);
>   void sysbus_add_io(SysBusDevice *dev, hwaddr addr,
>                      MemoryRegion *mem);
>   MemoryRegion *sysbus_address_space(SysBusDevice *dev);

Re: [PATCH 06/14] aspeed: Add system-memory QOM link to SoC

[Peter Delevoryas]

> On Jun 23, 2022, at 5:30 AM, Peter Maydell <peter.maydell@linaro.org> wrote:
> 
> On Thu, 23 Jun 2022 at 12:31, Peter Delevoryas <pdel@fb.com> wrote:
>> 
>> Right now it's just defined as the regular global system memory. If we
>> migrate all the SoC code to use this property instead of directly calling
>> get_system_memory(), then we can restrict the memory container for the SoC,
>> which will be useful for multi-SoC machines.
>> 
>> Signed-off-by: Peter Delevoryas <pdel@fb.com>
> 
>> static Property aspeed_soc_properties[] = {
>> +    DEFINE_PROP_LINK("system-memory", AspeedSoCState, system_memory,
>> +                     TYPE_MEMORY_REGION, MemoryRegion *),
> 
> To the extent that we have a convention, we tend to call this
> property on an SoC or CPU "memory", I think. (Better suggestions
> welcome...)

Oh oops, yes, in hindsight that is very obvious. I’ll change
this from “system-memory” to “memory”.

> 
> -- PMM

Re: [PATCH 06/14] aspeed: Add system-memory QOM link to SoC

[Peter Delevoryas]

> On Jun 23, 2022, at 5:30 AM, Peter Maydell <peter.maydell@linaro.org> wrote:
> 
> On Thu, 23 Jun 2022 at 12:31, Peter Delevoryas <pdel@fb.com> wrote:
>> 
>> Right now it's just defined as the regular global system memory. If we
>> migrate all the SoC code to use this property instead of directly calling
>> get_system_memory(), then we can restrict the memory container for the SoC,
>> which will be useful for multi-SoC machines.
>> 
>> Signed-off-by: Peter Delevoryas <pdel@fb.com>
> 
>> static Property aspeed_soc_properties[] = {
>> +    DEFINE_PROP_LINK("system-memory", AspeedSoCState, system_memory,
>> +                     TYPE_MEMORY_REGION, MemoryRegion *),
> 
> To the extent that we have a convention, we tend to call this
> property on an SoC or CPU "memory", I think. (Better suggestions
> welcome...)

Oh oops, yes, in hindsight that is very obvious. I’ll change
this from “system-memory” to “memory”.

> 
> -- PMM

Re: [PATCH 11/14] aspeed: Switch to create_unimplemented_device_in

[Peter Delevoryas]

> On Jun 23, 2022, at 5:56 AM, Peter Maydell <peter.maydell@linaro.org> wrote:
> 
> On Thu, 23 Jun 2022 at 13:04, Peter Delevoryas <pdel@fb.com> wrote:
>> 
>> Signed-off-by: Peter Delevoryas <pdel@fb.com>
>> ---
>> hw/arm/aspeed_ast10x0.c | 10 ++++------
>> hw/arm/aspeed_ast2600.c | 19 ++++++++++---------
>> hw/arm/aspeed_soc.c     |  9 +++++----
>> 3 files changed, 19 insertions(+), 19 deletions(-)
>> 
>> diff --git a/hw/arm/aspeed_ast10x0.c b/hw/arm/aspeed_ast10x0.c
>> index d259d30fc0..4e6688cc68 100644
>> --- a/hw/arm/aspeed_ast10x0.c
>> +++ b/hw/arm/aspeed_ast10x0.c
>> @@ -158,12 +158,10 @@ static void aspeed_soc_ast1030_realize(DeviceState *dev_soc, Error **errp)
>>     }
>> 
>>     /* General I/O memory space to catch all unimplemented device */
>> -    create_unimplemented_device("aspeed.sbc",
>> -                                sc->memmap[ASPEED_DEV_SBC],
>> -                                0x40000);
>> -    create_unimplemented_device("aspeed.io",
>> -                                sc->memmap[ASPEED_DEV_IOMEM],
>> -                                ASPEED_SOC_IOMEM_SIZE);
>> +    create_unimplemented_device_in("aspeed.sbc", sc->memmap[ASPEED_DEV_SBC],
>> +                                   0x40000, s->system_memory);
>> +    create_unimplemented_device_in("aspeed.io", sc->memmap[ASPEED_DEV_IOMEM],
>> +                                   ASPEED_SOC_IOMEM_SIZE, s->system_memory);
> 
> This is SoC code, so it should probably be handling its unimplemented
> devices by creating and mapping TYPE_UNIMPLEMENTED_DEVICE child
> objects directly, the same way it handles all its other child devices.

Ah, right. I’ll include a patch to create the device as a regular
child object, and after that it will probably make more sense to
remove “create_unimplemented_device_in” too and just do that
memory mapping directly. Especially since I’m going to remove
the “sysbus_mmio_map_overlap_in” function too.

> 
> thanks
> -- PMM

Re: [PATCH 11/14] aspeed: Switch to create_unimplemented_device_in

[Peter Delevoryas]

> On Jun 23, 2022, at 5:56 AM, Peter Maydell <peter.maydell@linaro.org> wrote:
> 
> On Thu, 23 Jun 2022 at 13:04, Peter Delevoryas <pdel@fb.com> wrote:
>> 
>> Signed-off-by: Peter Delevoryas <pdel@fb.com>
>> ---
>> hw/arm/aspeed_ast10x0.c | 10 ++++------
>> hw/arm/aspeed_ast2600.c | 19 ++++++++++---------
>> hw/arm/aspeed_soc.c     |  9 +++++----
>> 3 files changed, 19 insertions(+), 19 deletions(-)
>> 
>> diff --git a/hw/arm/aspeed_ast10x0.c b/hw/arm/aspeed_ast10x0.c
>> index d259d30fc0..4e6688cc68 100644
>> --- a/hw/arm/aspeed_ast10x0.c
>> +++ b/hw/arm/aspeed_ast10x0.c
>> @@ -158,12 +158,10 @@ static void aspeed_soc_ast1030_realize(DeviceState *dev_soc, Error **errp)
>>     }
>> 
>>     /* General I/O memory space to catch all unimplemented device */
>> -    create_unimplemented_device("aspeed.sbc",
>> -                                sc->memmap[ASPEED_DEV_SBC],
>> -                                0x40000);
>> -    create_unimplemented_device("aspeed.io",
>> -                                sc->memmap[ASPEED_DEV_IOMEM],
>> -                                ASPEED_SOC_IOMEM_SIZE);
>> +    create_unimplemented_device_in("aspeed.sbc", sc->memmap[ASPEED_DEV_SBC],
>> +                                   0x40000, s->system_memory);
>> +    create_unimplemented_device_in("aspeed.io", sc->memmap[ASPEED_DEV_IOMEM],
>> +                                   ASPEED_SOC_IOMEM_SIZE, s->system_memory);
> 
> This is SoC code, so it should probably be handling its unimplemented
> devices by creating and mapping TYPE_UNIMPLEMENTED_DEVICE child
> objects directly, the same way it handles all its other child devices.

Ah, right. I’ll include a patch to create the device as a regular
child object, and after that it will probably make more sense to
remove “create_unimplemented_device_in” too and just do that
memory mapping directly. Especially since I’m going to remove
the “sysbus_mmio_map_overlap_in” function too.

> 
> thanks
> -- PMM