qemu-devel.nongnu.org archive mirror
 help / color / mirror / Atom feed
From: Laurent Vivier <lvivier@redhat.com>
To: "Daniel P. Berrangé" <berrange@redhat.com>, qemu-devel@nongnu.org
Cc: Kevin Wolf <kwolf@redhat.com>,
	Eduardo Habkost <eduardo@habkost.net>,
	Thomas Huth <thuth@redhat.com>, Peter Krempa <pkrempa@redhat.com>,
	Markus Armbruster <armbru@redhat.com>,
	Paolo Bonzini <pbonzini@redhat.com>,
	Eric Blake <eblake@redhat.com>
Subject: Re: [PATCH 1/1] softmmu: fix device deletion events with -device JSON syntax
Date: Wed, 5 Jan 2022 15:49:12 +0100	[thread overview]
Message-ID: <a2a94c4c-190f-9be6-eadf-bd1404a2e272@redhat.com> (raw)
In-Reply-To: <20220105123847.4047954-2-berrange@redhat.com>

On 05/01/2022 13:38, Daniel P. Berrangé wrote:
> The -device JSON syntax impl leaks a reference on the created
> DeviceState instance. As a result when you hot-unplug the
> device, the device_finalize method won't be called and thus
> it will fail to emit the required DEVICE_DELETED event.
> 
> A 'json-cli' feature was previously added against the
> 'device_add' QMP command QAPI schema to indicated to mgmt
> apps that -device supported JSON syntax. Given the hotplug
> bug that feature flag is no unusable for its purpose, so

Not sure to understand: do you mean "now unusable"?

> we add a new 'json-cli-hotplug' feature to indicate the
> -device supports JSON without breaking hotplug.
> 
> Fixes: https://gitlab.com/qemu-project/qemu/-/issues/802
> Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
> ---
>   qapi/qdev.json                 |  5 ++++-
>   softmmu/vl.c                   |  4 +++-
>   tests/qtest/device-plug-test.c | 19 +++++++++++++++++++
>   3 files changed, 26 insertions(+), 2 deletions(-)
> 
> diff --git a/qapi/qdev.json b/qapi/qdev.json
> index 69656b14df..26cd10106b 100644
> --- a/qapi/qdev.json
> +++ b/qapi/qdev.json
> @@ -44,6 +44,9 @@
>   # @json-cli: If present, the "-device" command line option supports JSON
>   #            syntax with a structure identical to the arguments of this
>   #            command.
> +# @json-cli-hotplug: If present, the "-device" command line option supports JSON
> +#                    syntax without the reference counting leak that broke
> +#                    hot-unplug
>   #
>   # Notes:
>   #
> @@ -74,7 +77,7 @@
>   { 'command': 'device_add',
>     'data': {'driver': 'str', '*bus': 'str', '*id': 'str'},
>     'gen': false, # so we can get the additional arguments
> -  'features': ['json-cli'] }
> +  'features': ['json-cli', 'json-cli-hotplug'] }
>   
>   ##
>   # @device_del:
> diff --git a/softmmu/vl.c b/softmmu/vl.c
> index d9e4c619d3..b1fc7da104 100644
> --- a/softmmu/vl.c
> +++ b/softmmu/vl.c
> @@ -2684,6 +2684,7 @@ static void qemu_create_cli_devices(void)
>       qemu_opts_foreach(qemu_find_opts("device"),
>                         device_init_func, NULL, &error_fatal);
>       QTAILQ_FOREACH(opt, &device_opts, next) {
> +        DeviceState *dev;
>           loc_push_restore(&opt->loc);
>           /*
>            * TODO Eventually we should call qmp_device_add() here to make sure it
> @@ -2692,7 +2693,8 @@ static void qemu_create_cli_devices(void)
>            * from the start, so call qdev_device_add_from_qdict() directly for
>            * now.
>            */
> -        qdev_device_add_from_qdict(opt->opts, true, &error_fatal);
> +        dev = qdev_device_add_from_qdict(opt->opts, true, &error_fatal);
> +        object_unref(OBJECT(dev));
>           loc_pop(&opt->loc);
>       }
>       rom_reset_order_override();
> diff --git a/tests/qtest/device-plug-test.c b/tests/qtest/device-plug-test.c
> index 559d47727a..ad79bd4c14 100644
> --- a/tests/qtest/device-plug-test.c
> +++ b/tests/qtest/device-plug-test.c
> @@ -77,6 +77,23 @@ static void test_pci_unplug_request(void)
>       qtest_quit(qtest);
>   }
>   
> +static void test_pci_unplug_json_request(void)
> +{
> +    QTestState *qtest = qtest_initf(
> +        "-device '{\"driver\": \"virtio-mouse-pci\", \"id\": \"dev0\"}'");
> +
> +    /*
> +     * Request device removal. As the guest is not running, the request won't
> +     * be processed. However during system reset, the removal will be
> +     * handled, removing the device.
> +     */
> +    device_del(qtest, "dev0");
> +    system_reset(qtest);

You can use qpci_unplug_acpi_device_test() to process the request... but I see this is 
done like that too in test_pci_unplug_request()

> +    wait_device_deleted_event(qtest, "dev0");
> +
> +    qtest_quit(qtest);
> +}
> +
>   static void test_ccw_unplug(void)
>   {
>       QTestState *qtest = qtest_initf("-device virtio-balloon-ccw,id=dev0");
> @@ -145,6 +162,8 @@ int main(int argc, char **argv)
>        */
>       qtest_add_func("/device-plug/pci-unplug-request",
>                      test_pci_unplug_request);
> +    qtest_add_func("/device-plug/pci-unplug-json-request",
> +                   test_pci_unplug_json_request);
>   
>       if (!strcmp(arch, "s390x")) {
>           qtest_add_func("/device-plug/ccw-unplug",

Reviewed-by: Laurent Vivier <lvivier@redhat.com>



  parent reply	other threads:[~2022-01-05 15:06 UTC|newest]

Thread overview: 12+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2022-01-05 12:38 [PATCH 0/1] Fix -device JSON support wrt hotplug Daniel P. Berrangé
2022-01-05 12:38 ` [PATCH 1/1] softmmu: fix device deletion events with -device JSON syntax Daniel P. Berrangé
2022-01-05 12:49   ` Thomas Huth
2022-01-05 14:37   ` Ján Tomko
2022-01-05 14:49   ` Laurent Vivier [this message]
2022-01-05 14:55     ` Daniel P. Berrangé
2022-01-05 15:00       ` Laurent Vivier
2022-01-05 15:18         ` Daniel P. Berrangé
2022-01-05 15:19           ` Thomas Huth
2022-01-05 15:31           ` Laurent Vivier
2022-01-14 12:22   ` Markus Armbruster
2022-01-11 15:21 ` [PATCH 0/1] Fix -device JSON support wrt hotplug Kevin Wolf

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=a2a94c4c-190f-9be6-eadf-bd1404a2e272@redhat.com \
    --to=lvivier@redhat.com \
    --cc=armbru@redhat.com \
    --cc=berrange@redhat.com \
    --cc=eblake@redhat.com \
    --cc=eduardo@habkost.net \
    --cc=kwolf@redhat.com \
    --cc=pbonzini@redhat.com \
    --cc=pkrempa@redhat.com \
    --cc=qemu-devel@nongnu.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).