Re: [RFC PATCH 2/3] drm: define new accel major and register it

[Jeffrey Hugo <quic_jhugo@quicinc.com>]

On 10/24/2022 1:52 AM, Dave Airlie wrote:
> On Mon, 24 Oct 2022 at 17:23, Oded Gabbay <ogabbay@kernel.org> wrote:
>>
>> On Sun, Oct 23, 2022 at 3:40 PM Greg Kroah-Hartman
>> <gregkh@linuxfoundation.org> wrote:
>>>
>>> On Sun, Oct 23, 2022 at 12:46:21AM +0300, Oded Gabbay wrote:
>>>> The accelerator devices will be exposed to the user space with a new,
>>>> dedicated major number - 261.
>>>>
>>>> The drm core registers the new major number as a char device and create
>>>> corresponding sysfs and debugfs root entries, same as for the drm major.
>>>>
>>>> In case CONFIG_ACCEL is not selected, this code is not compiled in.
>>>>
>>>> Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
>>>> ---
>>>>   Documentation/admin-guide/devices.txt |  5 +++
>>>>   drivers/gpu/drm/drm_drv.c             | 45 +++++++++++++++++++++++
>>>>   drivers/gpu/drm/drm_internal.h        |  3 ++
>>>>   drivers/gpu/drm/drm_sysfs.c           | 52 +++++++++++++++++++++++++++
>>>>   include/drm/drm_ioctl.h               |  1 +
>>>>   5 files changed, 106 insertions(+)
>>>>
>>>> diff --git a/Documentation/admin-guide/devices.txt b/Documentation/admin-guide/devices.txt
>>>> index 9764d6edb189..06c525e01ea5 100644
>>>> --- a/Documentation/admin-guide/devices.txt
>>>> +++ b/Documentation/admin-guide/devices.txt
>>>> @@ -3080,6 +3080,11 @@
>>>>                  ...
>>>>                  255 = /dev/osd255     256th OSD Device
>>>>
>>>> + 261 char    Compute Acceleration Devices
>>>> +               0 = /dev/accel/accel0 First acceleration device
>>>> +               1 = /dev/accel/accel1 Second acceleration device
>>>> +                 ...
>>>> +
>>>>    384-511 char        RESERVED FOR DYNAMIC ASSIGNMENT
>>>>                Character devices that request a dynamic allocation of major
>>>>                number will take numbers starting from 511 and downward,
>>>> diff --git a/drivers/gpu/drm/drm_drv.c b/drivers/gpu/drm/drm_drv.c
>>>> index 8214a0b1ab7f..b58ffb1433d6 100644
>>>> --- a/drivers/gpu/drm/drm_drv.c
>>>> +++ b/drivers/gpu/drm/drm_drv.c
>>>> @@ -67,6 +67,10 @@ static bool drm_core_init_complete;
>>>>
>>>>   static struct dentry *drm_debugfs_root;
>>>>
>>>> +#ifdef CONFIG_ACCEL
>>>> +static struct dentry *accel_debugfs_root;
>>>> +#endif
>>>> +
>>>>   DEFINE_STATIC_SRCU(drm_unplug_srcu);
>>>>
>>>>   /*
>>>> @@ -1031,9 +1035,19 @@ static const struct file_operations drm_stub_fops = {
>>>>        .llseek = noop_llseek,
>>>>   };
>>>>
>>>> +static void accel_core_exit(void)
>>>> +{
>>>> +#ifdef CONFIG_ACCEL
>>>> +     unregister_chrdev(ACCEL_MAJOR, "accel");
>>>> +     debugfs_remove(accel_debugfs_root);
>>>> +     accel_sysfs_destroy();
>>>> +#endif
>>>> +}
>>>
>>> Why is all of this in drm_drv.c?
>>>
>>> Why not put it in drm/accel/accel.c or something like that?  Then put
>>> the proper stuff into a .h file and then you have no #ifdef in the .c
>>> files.
>> I thought about that, adding an accel.c in drivers/accel/ and putting
>> this code there.
>> Eventually I thought that for two functions it's not worth it, but I
>> guess that in addition to the reason you gave, one can argue that
>> there will probably be more code in that file anyway, so why not open
>> it now.
>> I'll change this if no one else thinks otherwise.
> 
> Seems like a good idea to start doing it now, might make things easier
> to keep separated.

I agree.  I was a bit confused going through this patch, and envisioning 
how an accel driver would use the interface.  I think an 
accel_internal.h would be clearer.

-Jeff