On Wed, Jun 05, 2024 at 10:36:30AM +0800, sicong wrote: > > > At 2024-06-04 21:11:57, "Greg KH" <gregkh@xxxxxxxxxxxxxxxxxxx> wrote: > >On Mon, May 13, 2024 at 10:19:48AM +0800, Sicong Huang wrote: > >> This bug exists in drivers/usb/cdns3/cdns3-gadget.c. Function > >> __cdns3_gadget_init will call cdns3_gadget_start to do futher jobs > >> during the initialization proccess of cdns3 gadget. In cdns3_gadget_start, > >> &priv_dev->pending_status_wq is bound with cdns3_pending_setup_status_handler. > >> Then this work will be added to system_freezable_wq in cdns3_gadget_ep0_queue. > >> Here is the code. > >> queue_work(system_freezable_wq, &priv_dev->pending_status_wq); > >> > >> If we call cdns3_gadget_exit to remove the device and make cleanup, > >> there are some unfinished works. This function will call cdns3_free_all_eps to > >> free all the endpoints. However, if cdns3_pending_setup_status_handler is > >> scheduled to run after the free job, it will cause use-after-free error as > >> cdns3_pending_setup_status_handler will use the endpoint in the following code. > >> request->complete(&priv_dev->eps[0]->endpoint, request); > >> > >> The possible execution flow that may lead to this issue is as follows: > >> CPU0 CPU1 > >> | __cdns3_gadget_init > >> | cdns3_gadget_start > >> cdns3_gadget_exit | > >> cdns3_free_all_eps | > >> devm_kfree (free) | > >> | cdns3_pending_setup_status_handler > >> | &priv_dev->eps[0]->endpoint (use) > >> > >> Fix it by cleaning the work in cdns3_gadget_exit. > >> > >> Signed-off-by: Sicong Huang <congei42@xxxxxxx> > >> --- > >> drivers/usb/cdns3/cdns3-gadget.c | 3 +++ > >> 1 file changed, 3 insertions(+) > >> > >> diff --git a/drivers/usb/cdns3/cdns3-gadget.c b/drivers/usb/cdns3/cdns3-gadget.c > >> index fd1beb10bba7..0f2e143bd17a 100644 > >> --- a/drivers/usb/cdns3/cdns3-gadget.c > >> +++ b/drivers/usb/cdns3/cdns3-gadget.c > >> @@ -3252,6 +3252,9 @@ static void cdns3_gadget_exit(struct cdns *cdns) > >> pm_runtime_mark_last_busy(cdns->dev); > >> pm_runtime_put_autosuspend(cdns->dev); > >> > >> + cancel_work_sync(&priv_dev->pending_status_wq); > >> + cancel_work_sync(&priv_dev->aligned_buf_wq); > >> + > >> usb_del_gadget(&priv_dev->gadget); > >> devm_free_irq(cdns->dev, cdns->dev_irq, priv_dev); > >> > >> -- > >> 2.34.1 > > > >What commit id does this fix? > > > >thanks, > > > > >greg k-h > > > I made the changes based on the following commit information. > commit 51474ab44abf907023a8a875e799b07de461e466 (origin/usb-testing, origin/usb-next) That is not what I meant, sorry. I mean, "this is a bugfix, so what commit caused the problem that this is fixing?" thanks, greg k-h
