qemu-devel.nongnu.org archive mirror
 help / color / mirror / Atom feed
From: "Philippe Mathieu-Daudé" <philmd@redhat.com>
To: Thomas Huth <thuth@redhat.com>,
	Markus Armbruster <armbru@redhat.com>,
	Claudio Fontana <cfontana@suse.de>
Cc: "Laurent Vivier" <lvivier@redhat.com>,
	"Andrew Jones" <drjones@redhat.com>,
	"Daniel P.Berrangé" <berrange@redhat.com>,
	"Eduardo Habkost" <ehabkost@redhat.com>,
	"Michael S. Tsirkin" <mst@redhat.com>,
	"Alex Bennée" <alex.bennee@linaro.org>,
	"Richard Henderson" <richard.henderson@linaro.org>,
	qemu-devel@nongnu.org, qemu-arm@nongnu.org,
	"Igor Mammedov" <imammedo@redhat.com>,
	"Paolo Bonzini" <pbonzini@redhat.com>
Subject: Re: [PATCH 1/6] accel: Introduce 'query-accels' QMP command
Date: Tue, 16 Mar 2021 16:20:40 +0100	[thread overview]
Message-ID: <d7422fac-fa3d-f220-aea6-184b7d711b1f@redhat.com> (raw)
In-Reply-To: <57f03668-a73a-32f2-bce8-4d70a13ccf77@redhat.com>

On 3/16/21 1:48 PM, Thomas Huth wrote:
> On 16/03/2021 13.41, Markus Armbruster wrote:
>> Philippe Mathieu-Daudé <philmd@redhat.com> writes:
>>
>>> On 3/16/21 11:26 AM, Philippe Mathieu-Daudé wrote:
>>>> On 3/16/21 10:02 AM, Philippe Mathieu-Daudé wrote:
>>>>> On 3/16/21 7:51 AM, Markus Armbruster wrote:
>>>>>> Eric Blake <eblake@redhat.com> writes:
>>>>>>
>>>>>>> On 3/11/21 5:11 PM, Philippe Mathieu-Daudé wrote:
>>>>>> [...]
>>>>>>>> diff --git a/accel/accel-qmp.c b/accel/accel-qmp.c
>>>>>>>> new file mode 100644
>>>>>>>> index 00000000000..f16e49b8956
>>>>>>>> --- /dev/null
>>>>>>>> +++ b/accel/accel-qmp.c
>>>>>>>> @@ -0,0 +1,47 @@
>>>>>>>> +/*
>>>>>>>> + * QEMU accelerators, QMP commands
>>>>>>>> + *
>>>>>>>> + * Copyright (c) 2021 Red Hat Inc.
>>>>>>>> + *
>>>>>>>> + * SPDX-License-Identifier: GPL-2.0-or-later
>>>>>>>> + */
>>>>>>>> +
>>>>>>>> +#include "qemu/osdep.h"
>>>>>>>> +#include "qapi/qapi-commands-machine.h"
>>>>>>>> +
>>>>>>>> +static const Accelerator accel_list[] = {
>>>>>>>> +    ACCELERATOR_QTEST,
>>>>>>>> +#ifdef CONFIG_TCG
>>>>>>>> +    ACCELERATOR_TCG,
>>>>>>>> +#endif
>>>>>>>> +#ifdef CONFIG_KVM
>>>>>>>> +    ACCELERATOR_KVM,
>>>>>>>> +#endif
>>>>>>>
>>>>>>> ...would it be worth compiling the enum to only list enum values
>>>>>>> that
>>>>>>> were actually compiled in?  That would change it to:
>>>>>>>
>>>>>>> { 'enum': 'Accelerator',
>>>>>>>    'data': [ 'qtest',
>>>>>>>              { 'name': 'tcg', 'if': 'defined(CONFIG_TCG)' },
>>>>>>> ...
>>>>>
>>>>> These accelerator definitions are supposed to be poisoned in generic
>>>>> code... But I like the simplicity of your suggestion, so I'll give it
>>>>> a try and see what happens with removing the poisoned definitions.
>>>>
>>>> This is actually quite interesting :) Accelerator definitions are
>>>> declared in config-target.h, but acceleration is host specific...
>>>
>>> Thomas, I guess I hit Claudio's reported bug again...
>>>
>>> 1/ generic libqemuutil.a is built without any CONFIG_accel definition.
>>>
>>> So this qapi-generated enum ... :
>>>
>>> typedef enum Accelerator {
>>>      ACCELERATOR_QTEST,
>>> #if defined(CONFIG_TCG)
>>>      ACCELERATOR_TCG,
>>> #endif /* defined(CONFIG_TCG) */
>>> #if defined(CONFIG_KVM)
>>>      ACCELERATOR_KVM,
>>> #endif /* defined(CONFIG_KVM) */
>>> #if defined(CONFIG_HAX)
>>>      ACCELERATOR_HAX,
>>> #endif /* defined(CONFIG_HAX) */
>>> #if defined(CONFIG_HVF)
>>>      ACCELERATOR_HVF,
>>> #endif /* defined(CONFIG_HVF) */
>>> #if defined(CONFIG_WHPX)
>>>      ACCELERATOR_WHPX,
>>> #endif /* defined(CONFIG_WHPX) */
>>> #if defined(CONFIG_XEN_BACKEND)
>>>      ACCELERATOR_XEN,
>>> #endif /* defined(CONFIG_XEN_BACKEND) */
>>>      ACCELERATOR__MAX,
>>> } Accelerator;
>>>
>>> ... is expanded to:
>>>
>>> typedef enum Accelerator {
>>>      ACCELERATOR_QTEST,
>>>      ACCELERATOR__MAX,
>>> } Accelerator;
>>
>> CONFIG_KVM, CONFIG_TCG, ...  are defined in ${target}-config-target.h,
>> and may only be used in target-specific code.
>>
>> If the enum ends up in libqemuutil.a, there are uses outside
>> target-specific code.
>>
>> exec/poison.h lacks CONFIG_KVM, CONFIG_TCG, ...  Should they be added?
> 
> CONFIG_KVM is already in poison.h, and CONFIG_TCG cannot be added there
> since it is also defined in config-host.h. But the other accelerator
> switches should be marked as poisoned, too.

Maybe we hit "exec/poison.h" limit. Maybe it is too wide / generic,
and need split, or multiple ones.

Should we redefine on which code area we want these definitions
poisoned?

AFAIK accel/ is not target specific but host specific.

My list of area / useful to poison:

- user-mode
  . non-tcg accel
  . hardware emulation

- generic hardware emulation
  . target specific
  . accel specific

For this one it is already too late:

- target acceleration
  . hardware emulation

Thoughts?



  reply	other threads:[~2021-03-16 15:22 UTC|newest]

Thread overview: 34+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-03-11 23:11 [PATCH 0/6] qtests: Check accelerator available at runtime via QMP 'query-accels' Philippe Mathieu-Daudé
2021-03-11 23:11 ` [PATCH 1/6] accel: Introduce 'query-accels' QMP command Philippe Mathieu-Daudé
2021-03-12  7:42   ` Marc-André Lureau
2021-03-12  8:11     ` Thomas Huth
2021-03-12  8:48       ` Andrew Jones
2021-03-12  8:52         ` Claudio Fontana
2021-03-12  9:09           ` Andrew Jones
2021-03-12  9:01         ` Paolo Bonzini
2021-03-12  9:17           ` Andrew Jones
2021-03-12  9:28             ` Paolo Bonzini
2021-03-12  8:46     ` Claudio Fontana
2021-03-12  9:04       ` Paolo Bonzini
2021-03-15 17:53   ` Eric Blake
2021-03-16  6:51     ` Markus Armbruster
2021-03-16  8:21       ` Paolo Bonzini
2021-03-16  9:02       ` Philippe Mathieu-Daudé
2021-03-16 10:26         ` Philippe Mathieu-Daudé
2021-03-16 10:39           ` Markus Armbruster
2021-03-16 10:55           ` Philippe Mathieu-Daudé
2021-03-16 12:41             ` Markus Armbruster
2021-03-16 12:48               ` Thomas Huth
2021-03-16 15:20                 ` Philippe Mathieu-Daudé [this message]
2021-03-11 23:11 ` [PATCH 2/6] tests/qtest: Add qtest_probe_accel() method Philippe Mathieu-Daudé
2021-03-12  8:16   ` Thomas Huth
2021-03-12  8:58     ` Andrew Jones
2021-03-11 23:11 ` [PATCH 3/6] qtest/bios-tables-test: Make test build-independent from accelerator Philippe Mathieu-Daudé
2021-03-11 23:12 ` [PATCH 4/6] qtest/arm-cpu-features: Check KVM availability at runtime Philippe Mathieu-Daudé
2021-03-12  9:05   ` Paolo Bonzini
2021-03-12  9:32     ` Philippe Mathieu-Daudé
2021-03-12  9:50       ` Paolo Bonzini
2021-03-11 23:12 ` [PATCH 5/6] qtest/arm-cpu-features: Check TCG " Philippe Mathieu-Daudé
2021-03-12  9:05   ` Paolo Bonzini
2021-03-11 23:12 ` [PATCH 6/6] tests/qtest: Do not restrict bios-tables-test to Aarch64 hosts anymore Philippe Mathieu-Daudé
2021-03-12  9:06   ` Paolo Bonzini

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=d7422fac-fa3d-f220-aea6-184b7d711b1f@redhat.com \
    --to=philmd@redhat.com \
    --cc=alex.bennee@linaro.org \
    --cc=armbru@redhat.com \
    --cc=berrange@redhat.com \
    --cc=cfontana@suse.de \
    --cc=drjones@redhat.com \
    --cc=ehabkost@redhat.com \
    --cc=imammedo@redhat.com \
    --cc=lvivier@redhat.com \
    --cc=mst@redhat.com \
    --cc=pbonzini@redhat.com \
    --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).