Never mind, it was my own fault. Was essentially returning a positive errno value (bug in new code) and hadn't realized probe had actually succeeded due to that instead of failed. Thanks, Roderick On Tue, Sep 27, 2022 at 9:53 PM Roderick Colenbrander <thunderbird2k@xxxxxxxxx> wrote: > > Hi Benjamin and Jiri, > > I'm doing some work around hid-playstation and as part of that I > stumbled on some kernel oopses, which I at first couldn't explain. I > believe I understand the problem now, but I don't know what the best > fix is and if it is completely my driver. > > The basic summary is that hid-playstation relies on associating custom > data with the device using hid_set_drvdata during probe (in > dualsense_create to be exact). When a probe call fails (one of the sub > paths of dualsense_create), there is currently an undefined pointer > associated with the hid_device. > > What seems to happen is that raw_event (ps_raw_event) still gets > called while in this undefined state. I don't fully understand this > well, but it is because the driver has to perform some get feature > requests and had to open the device. > > Is my understanding so far correct? If it is, my driver has a race > condition. The ultimate issue is raw_event, would it need to do > something special (e.g. looking at some flags). I can also delay > calling 'hid_set_drvdata' closer to successful setup (a little hacky > perhaps). Or is the issue and fix something else? > > Thanks, > Roderick
