qemu-devel.nongnu.org archive mirror
 help / color / mirror / Atom feed
From: "Philippe Mathieu-Daudé" <philmd@linaro.org>
To: Thomas Huth <thuth@redhat.com>, qemu-devel@nongnu.org
Cc: "Paolo Bonzini" <pbonzini@redhat.com>,
	"Richard Henderson" <richard.henderson@linaro.org>,
	"Eduardo Habkost" <eduardo@habkost.net>,
	"Peter Maydell" <peter.maydell@linaro.org>,
	"Mark Cave-Ayland" <mark.cave-ayland@ilande.co.uk>,
	"Daniel P. Berrangé" <berrange@redhat.com>,
	qemu-arm@nongnu.org, "Alex Bennée" <alex.bennee@linaro.org>,
	"Manos Pitsidianakis" <manos.pitsidianakis@linaro.org>
Subject: Re: [RFC PATCH-for-9.0 01/11] qom: Introduce the TypeInfo::can_register() handler
Date: Thu, 23 Nov 2023 18:30:56 +0100	[thread overview]
Message-ID: <2db81947-648f-421d-86f1-b48cdab55852@linaro.org> (raw)
In-Reply-To: <010c573d-17e6-43be-bb8f-6c3ff934bf4a@redhat.com>

On 23/11/23 17:24, Thomas Huth wrote:
> On 23/11/2023 17.03, Philippe Mathieu-Daudé wrote:
>> On 23/11/23 16:09, Thomas Huth wrote:
>>> On 22/11/2023 19.30, Philippe Mathieu-Daudé wrote:
>>>> Add a helper to decide at runtime whether a type can
>>>> be registered to the QOM framework or not.
>>>>
>>>> Signed-off-by: Philippe Mathieu-Daudé <philmd@linaro.org>
>>>> ---
>>>>   include/qom/object.h | 4 ++++
>>>>   qom/object.c         | 3 +++
>>>>   2 files changed, 7 insertions(+)
>>>>
>>>> diff --git a/include/qom/object.h b/include/qom/object.h
>>>> index afccd24ca7..0d42fe17de 100644
>>>> --- a/include/qom/object.h
>>>> +++ b/include/qom/object.h
>>>> @@ -372,6 +372,8 @@ struct Object
>>>>    * struct TypeInfo:
>>>>    * @name: The name of the type.
>>>>    * @parent: The name of the parent type.
>>>> + * @can_register: This optional function is called before a type is 
>>>> registered.
>>>> + *   If it exists and returns false, the type is not registered.
>>>
>>> The second sentence is quite hard to parse, since it is not quite 
>>> clear what "it" refers to (type or function) and what "registered" 
>>> means in this context (you don't mention type_register() here).
>>>
>>> Maybe rather something like:
>>>
>>> If set, type_register() uses this function to decide whether the type 
>>> can be registered or not.
>>>
>>> ?
>>>
>>>>    * @instance_size: The size of the object (derivative of 
>>>> #Object).  If
>>>>    *   @instance_size is 0, then the size of the object will be the 
>>>> size of the
>>>>    *   parent object.
>>>> @@ -414,6 +416,8 @@ struct TypeInfo
>>>>       const char *name;
>>>>       const char *parent;
>>>> +    bool (*can_register)(void);
>>>> +
>>>>       size_t instance_size;
>>>>       size_t instance_align;
>>>>       void (*instance_init)(Object *obj);
>>>> diff --git a/qom/object.c b/qom/object.c
>>>> index 95c0dc8285..f09b6b5a92 100644
>>>> --- a/qom/object.c
>>>> +++ b/qom/object.c
>>>> @@ -150,6 +150,9 @@ static TypeImpl *type_register_internal(const 
>>>> TypeInfo *info)
>>>>   TypeImpl *type_register(const TypeInfo *info)
>>>>   {
>>>>       assert(info->parent);
>>>> +    if (info->can_register && !info->can_register()) {
>>>> +        return NULL;
>>>> +    }
>>>
>>> I have to say that I don't like it too much, since you're trying to 
>>> fix a problem here in common code that clearly belongs to the code in 
>>> hw/arm/ instead.
>>>
>>> What about dropping it, and changing your last patch to replace the 
>>> DEFINE_TYPES(raspi_machine_types) in hw/arm/raspi.c with your own 
>>> implementation of type_register_static_array() that checks the 
>>> condition there?
>>
>> This isn't ARM specific, it happens I started to unify ARM/aarch64
>> binaries.
>>
>> Types can be registered depending on build-time (config/host specific)
>> definitions and runtime ones. How can we check for runtime if not via
>> this simple helper?
>>
>> Still ARM, but as example what I have then is (module meson):
> ...
>> diff --git a/target/arm/cpu64.c b/target/arm/cpu64.c
>> index 1e9c6c85ae..c3b7e5666c 100644
>> --- a/target/arm/cpu64.c
>> +++ b/target/arm/cpu64.c
>> @@ -744,7 +744,8 @@ static void aarch64_max_initfn(Object *obj)
>>   static const ARMCPUInfo aarch64_cpus[] = {
>>       { .name = "cortex-a57",         .initfn = aarch64_a57_initfn },
>>       { .name = "cortex-a53",         .initfn = aarch64_a53_initfn },
>> -    { .name = "max",                .initfn = aarch64_max_initfn },
>> +    { .name = "max",                .initfn = aarch64_max_initfn,
>> +                                    .can_register = 
>> target_aarch64_available },
>>   #if defined(CONFIG_KVM) || defined(CONFIG_HVF)
>>       { .name = "host",               .initfn = aarch64_host_initfn },
>>   #endif
> 
> Picking this one as an example, I think I'd rather modify the for-loop in
> aarch64_cpu_register_types() to check for the availability there... 
> sounds much easier to understand for me than having a callback function.

OK.

> Anyway, that's just my personal taste - if others agree with your 
> solution instead, I won't insist on my idea.

This is a RFC so let's discuss :) I think there is a need to filter
QOM types at runtime (at least in "Single Binary" or "Heterogeneous
machines"), but I might be wrong.
Maybe we can filter that elsewhere (here seemed the simplest / more
natural place). I'll keep looking.

Regards,

Phil.


  reply	other threads:[~2023-11-23 17:32 UTC|newest]

Thread overview: 25+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2023-11-22 18:30 [PATCH-for-9.0 00/11] hw/arm: Step toward building qemu-system-{arm, aarch64} altogether Philippe Mathieu-Daudé
2023-11-22 18:30 ` [RFC PATCH-for-9.0 01/11] qom: Introduce the TypeInfo::can_register() handler Philippe Mathieu-Daudé
2023-11-23 15:09   ` Thomas Huth
2023-11-23 16:03     ` Philippe Mathieu-Daudé
2023-11-23 16:24       ` Thomas Huth
2023-11-23 17:30         ` Philippe Mathieu-Daudé [this message]
2023-11-22 18:30 ` [RFC PATCH-for-9.0 02/11] target/arm: Add target_aarch64_available() helper Philippe Mathieu-Daudé
2023-11-23 10:00   ` Philippe Mathieu-Daudé
2023-11-22 18:30 ` [PATCH-for-9.0 03/11] target/arm: Declare ARM_CPU_TYPE_NAME/SUFFIX in 'cpu-qom.h' Philippe Mathieu-Daudé
2023-11-28 13:59   ` Richard Henderson
2023-11-22 18:30 ` [PATCH-for-9.0 04/11] target/arm: Move ARM_CPU_IRQ/FIQ definitions to 'cpu-qom.h' Philippe Mathieu-Daudé
2023-11-28 14:00   ` Richard Henderson
2023-11-22 18:30 ` [PATCH-for-9.0 05/11] target/arm: Move GTIMER definitions to 'cpu-defs.h' Philippe Mathieu-Daudé
2023-11-28 14:02   ` Richard Henderson
2023-11-28 16:32     ` Philippe Mathieu-Daudé
2023-11-22 18:30 ` [PATCH-for-9.0 06/11] hw/arm/bcm2836: Simplify use of 'reset-cbar' property Philippe Mathieu-Daudé
2023-11-28 14:03   ` Richard Henderson
2023-11-22 18:30 ` [PATCH-for-9.0 07/11] hw/arm/bcm2836: Simplify access to 'start-powered-off' property Philippe Mathieu-Daudé
2023-11-28 14:03   ` Richard Henderson
2023-11-22 18:30 ` [PATCH-for-9.0 08/11] hw/arm/bcm2836: Use ARM_CPU 'mp-affinity' property Philippe Mathieu-Daudé
2023-11-28 14:04   ` Richard Henderson
2023-11-22 18:30 ` [RFC PATCH-for-9.0 09/11] hw/arm/bcm2836: Allocate ARM CPU state with object_new() Philippe Mathieu-Daudé
2023-11-28 14:06   ` Richard Henderson
2023-11-22 18:30 ` [RFC PATCH-for-9.0 10/11] hw/arm/raspi: Build bcm2836.o and raspi.o objects once Philippe Mathieu-Daudé
2023-11-22 18:30 ` [RFC PATCH-for-9.0 11/11] hw/intc/meson: Simplify how arm_gicv3_kvm.o objects are built Philippe Mathieu-Daudé

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=2db81947-648f-421d-86f1-b48cdab55852@linaro.org \
    --to=philmd@linaro.org \
    --cc=alex.bennee@linaro.org \
    --cc=berrange@redhat.com \
    --cc=eduardo@habkost.net \
    --cc=manos.pitsidianakis@linaro.org \
    --cc=mark.cave-ayland@ilande.co.uk \
    --cc=pbonzini@redhat.com \
    --cc=peter.maydell@linaro.org \
    --cc=qemu-arm@nongnu.org \
    --cc=qemu-devel@nongnu.org \
    --cc=richard.henderson@linaro.org \
    --cc=thuth@redhat.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).