Re: [RFC PATCH] subprojects: add a wrapper for libvirglrenderer

["Alex Bennée" <alex.bennee@linaro.org>]

Akihiko Odaki <akihiko.odaki@daynix.com> writes:

> On 2024/06/05 22:35, Alex Bennée wrote:
>> As the latest features for virtio-gpu need a pretty recent version of
>> libvirglrenderer. When it is not available on the system we can use a
>> meson wrapper and provide it when --download is specified in
>> configure.
>> We have to take some additional care as currently QEMU will hang
>> libvirglrenderer fails to exec the render server. As the error isn't
>> back propagated we make sure we at least test we have a path to an
>> executable before tweaking the environment.
>
> Hi,
>
> The intent of this patch sounds good to me. It is the responsibility
> of users to set up virglrenderer in principle, but we can just be kind
> for them.
>
>> Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
>> Cc: Manos Pitsidianakis <manos.pitsidianakis@linaro.org>
>> Cc: Dmitry Osipenko <dmitry.osipenko@collabora.com>
>> Cc: Akihiko Odaki <akihiko.odaki@daynix.com>
>> ---
>>   meson.build                    |  7 ++++++-
>>   hw/display/virtio-gpu-virgl.c  | 24 ++++++++++++++++++++++++
>>   subprojects/virglrenderer.wrap |  6 ++++++
>>   3 files changed, 36 insertions(+), 1 deletion(-)
>>   create mode 100644 subprojects/virglrenderer.wrap
>> diff --git a/meson.build b/meson.build
>> index 1d7346b703..e4e270df78 100644
>> --- a/meson.build
>> +++ b/meson.build
>> @@ -1203,7 +1203,8 @@ have_vhost_user_gpu = have_tools and host_os == 'linux' and pixman.found()
>>   if not get_option('virglrenderer').auto() or have_system or have_vhost_user_gpu
>>     virgl = dependency('virglrenderer',
>>                        method: 'pkg-config',
>> -                     required: get_option('virglrenderer'))
>> +                     required: get_option('virglrenderer'),
>> +                     default_options: ['default_library=static', 'render-server=true', 'venus=true'])
>
> meson_options.txt of virglrenderer says:
>> DEPRECATED: render server is enabled by venus automatically
>
> I'm also a bit concerned to enable Venus by default when the upstream
> virglrenderer doesn't. Why is it disabled by the upstream? Perhaps is
> it time for upstream to enable it by default?
>
>>   endif
>>   rutabaga = not_found
>>   if not get_option('rutabaga_gfx').auto() or have_system or have_vhost_user_gpu
>> @@ -2314,6 +2315,10 @@ if virgl.version().version_compare('>=1.0.0')
>>     config_host_data.set('HAVE_VIRGL_RESOURCE_BLOB', 1)
>>     config_host_data.set('HAVE_VIRGL_VENUS', 1)
>>   endif
>> +if virgl.type_name().contains('internal')
>> +  config_host_data.set('HAVE_BUNDLED_VIRGL_SERVER', 1)
>> +endif
>> +
>>   config_host_data.set('CONFIG_VIRTFS', have_virtfs)
>>   config_host_data.set('CONFIG_VTE', vte.found())
>>   config_host_data.set('CONFIG_XKBCOMMON', xkbcommon.found())
>> diff --git a/hw/display/virtio-gpu-virgl.c b/hw/display/virtio-gpu-virgl.c
>> index c9d20a8a60..53d6742e79 100644
>> --- a/hw/display/virtio-gpu-virgl.c
>> +++ b/hw/display/virtio-gpu-virgl.c
>> @@ -14,6 +14,7 @@
>>   #include "qemu/osdep.h"
>>   #include "qemu/error-report.h"
>>   #include "qemu/iov.h"
>> +#include "qemu/cutils.h"
>>   #include "trace.h"
>>   #include "hw/virtio/virtio.h"
>>   #include "hw/virtio/virtio-gpu.h"
>> @@ -1122,6 +1123,26 @@ void virtio_gpu_virgl_reset(VirtIOGPU *g)
>>       virgl_renderer_reset();
>>   }
>>   +/*
>> + * If we fail to spawn the render server things tend to hang so it is
>> + * important to do our due diligence before then. If QEMU has bundled
>> + * the virgl server we want to ensure we can run it from the build
>> + * directory and if installed.
>
> This comment sounds a bit misleading. The following code does not
> ensure the render server exists; it just opportunistically sets the
> path to the bundled render server or let it fail otherwise.
>
> It also sounds like virgl_set_render_env() does an extra step to
> prevent hangs, but it is actually mandatory for relocated scenarios;
> the lack of render server always results in a non-functional Venus
> setup even if the hang is fixed.
>
> The hang is better to be noted in subprojects/virglrenderer.wrap since
> that is the reason we would want to wrap the project.
>
>> + *
>> + * The principle way we can override the libvirglrenders behaviour is
>> + * by setting environment variables.
>> + */
>> +static void virgl_set_render_env(void)
>> +{
>> +#ifdef HAVE_BUNDLED_VIRGL_SERVER
>> +    g_autofree char *file = get_relocated_path(CONFIG_QEMU_HELPERDIR "/virgl_render_server");
>> +    if (g_file_test(file, G_FILE_TEST_EXISTS | G_FILE_TEST_IS_EXECUTABLE)) {
>
> I think this g_file_test() should be removed; we would not want to let
> virglrenderer pick a random render server when the bundled server
> exists since the ABI between them can be different in theory.

I was thinking mainly of validating it was built, maybe that should be
an assert instead because if we failed to build the server we got the
bundling wrong?

>
>> +        g_setenv("RENDER_SERVER_EXEC_PATH", file, false);
>> +    }
>> +#endif
>> +}
>> +
>> +
>>   int virtio_gpu_virgl_init(VirtIOGPU *g)
>>   {
>>       int ret;
>> @@ -1145,6 +1166,9 @@ int virtio_gpu_virgl_init(VirtIOGPU *g)
>>       }
>>   #endif
>>   +    /* Ensure we can find the render server */
>> +    virgl_set_render_env();
>> +
>>       ret = virgl_renderer_init(g, flags, &virtio_gpu_3d_cbs);
>>       if (ret != 0) {
>>           error_report("virgl could not be initialized: %d", ret);
>> diff --git a/subprojects/virglrenderer.wrap b/subprojects/virglrenderer.wrap
>> new file mode 100644
>> index 0000000000..3656a478c4
>> --- /dev/null
>> +++ b/subprojects/virglrenderer.wrap
>> @@ -0,0 +1,6 @@
>> +[wrap-git]
>> +url = https://gitlab.freedesktop.org/virgl/virglrenderer.git
>> +revision = virglrenderer-1.0.1
>> +
>> +[provide]
>> +virglrenderer = libvirglrenderer_dep

-- 
Alex Bennée
Virtualisation Tech Lead @ Linaro