On Wed, Nov 6, 2019 at 11:43 AM Chris Wilson chris@chris-wilson.co.uk wrote:
Quoting Chris Wilson (2019-11-06 10:26:48)
Quoting Daniel Vetter (2019-11-06 10:19:50)
On Wed, Nov 06, 2019 at 10:07:14AM +0000, Chris Wilson wrote:
Sometimes we need to create a struct file to wrap a drm_device, as it the user were to have opened /dev/dri/card0 but to do so anonymously (i.e. for internal use). Provide a utility method to create a struct file with the drm_device->driver.fops, that wrap the drm_device.
Signed-off-by: Chris Wilson chris@chris-wilson.co.uk
For proper internal access we already have drm_client_open, so I think this has limited (but good use) in selftests only. So EXPORT_SYMBOL_FOR_TESTS_ONLY plus maybe a clearer name for the intended use like drm_file_mock_open?
I found the example in drm_gem_prime_mmap() that was doing the same trick, and the trick of being able to instantiate new struct file and install a fd whenever seems like it will come in handy... Just lacking the third user at the moment to claim generality.
The closest example I found in the spirit of creating a new drm_device struct file and installing it is drm_mode_create_lease_ioctl() that uses file_clone_open() for this purpose. The argument there would be whether cloning the (file->f_path, file->f_flags, file->f_cred) is appropriate versus an anonymous inode. I think cloning the credentials seems correct for leasing.
Hm ... I think we want the clone for this one here too. Otherwise we get the wrong inode, and then unmap_mapping_range wont work correctly, and we cant use this for selftests. That's the only case where I think we do actually need the file/inode to be functional. For anything else the drm_client internal api gets away without the file/inode stuff. -Daniel