Re: [RFC PATCH v2 2/3] accel: add dedicated minor for accelerator devices

[Oded Gabbay <ogabbay@kernel.org>]

On Sun, Nov 6, 2022 at 12:54 PM Oded Gabbay <ogabbay@kernel.org> wrote:
>
> On Thu, Nov 3, 2022 at 7:26 AM Jiho Chu <jiho.chu@samsung.com> wrote:
> >
> > On Wed,  2 Nov 2022 22:34:04 +0200
> > Oded Gabbay <ogabbay@kernel.org> wrote:
> >
> > > +/**
> > > + * accel_open - open method for ACCEL file
> > > + * @inode: device inode
> > > + * @filp: file pointer.
> > > + *
> > > + * This function must be used by drivers as their &file_operations.open method.
> > > + * It looks up the correct ACCEL device and instantiates all the per-file
> > > + * resources for it. It also calls the &drm_driver.open driver callback.
> > > + *
> > > + * Return: 0 on success or negative errno value on failure.
> > > + */
> > > +int accel_open(struct inode *inode, struct file *filp)
> > > +{
> > > +     struct drm_device *dev;
> > > +     struct drm_minor *minor;
> > > +     int retcode;
> > > +
> > > +     minor = accel_minor_acquire(iminor(inode));
> > > +     if (IS_ERR(minor))
> > > +             return PTR_ERR(minor);
> > > +
> > > +     dev = minor->dev;
> > > +
> > > +     atomic_fetch_inc(&dev->open_count);
> > > +
> >
> > Hi,
> > It needs to consider drm_global_mutex to access open_count.
> > please check doxy of open_count.
> Now that I'm changing the code back to be part of drm.ko, I can return
> all the code that is in drm_copy which I removed for this to compile.
I take it back. All the code that I omitted was for legacy drivers.
If you look inside drm_dev_needs_global_mutex(), you will see 3 cases
where you need to take the global mutex, and all 3 are only relevant
for legacy drivers and/or drivers that use deprecated features.
So, I disagree with your original comment here.
Moreover, open_count is atomic, so I don't need to take the mutex to
increment it, and as you can see in drm_open(), the function
increments it regardless of whether it takes
drm_dev_needs_global_mutex.
Oded

>
> >
> >
> > > +     /* share address_space across all char-devs of a single device */
> > > +     filp->f_mapping = dev->anon_inode->i_mapping;
> > > +
> > > +     retcode = drm_open_helper(filp, minor);
> > > +     if (retcode)
> > > +             goto err_undo;
> > > +
> > > +     return 0;
> > > +
> > > +err_undo:
> > > +     atomic_dec(&dev->open_count);
> > > +     accel_minor_release(minor);
> > > +     return retcode;
> > > +}
> > > +EXPORT_SYMBOL_GPL(accel_open);
> > > +
> > >  static int accel_stub_open(struct inode *inode, struct file *filp)
> > >  {
> > > -     DRM_DEBUG("Operation not supported");
> > > +     const struct file_operations *new_fops;
> > > +     struct drm_minor *minor;
> > > +     int err;
> > > +
> > > +     DRM_DEBUG("\n");
> >
> > It seems useless.
> Correct, I removed it in v3.
> Thanks,
> Oded
> >
> > Thanks.
> > Jiho Chu