From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: X-Spam-Checker-Version: SpamAssassin 3.4.0 (2014-02-07) on aws-us-west-2-korg-lkml-1.web.codeaurora.org X-Spam-Level: X-Spam-Status: No, score=-15.2 required=3.0 tests=BAYES_00, HEADER_FROM_DIFFERENT_DOMAINS,INCLUDES_CR_TRAILER,INCLUDES_PATCH, MAILING_LIST_MULTI,SPF_HELO_NONE,SPF_PASS,URIBL_BLOCKED,USER_AGENT_SANE_1 autolearn=unavailable autolearn_force=no version=3.4.0 Received: from mail.kernel.org (mail.kernel.org [198.145.29.99]) by smtp.lore.kernel.org (Postfix) with ESMTP id CF9C5C433E0 for ; Tue, 30 Mar 2021 15:50:11 +0000 (UTC) Received: from vger.kernel.org (vger.kernel.org [23.128.96.18]) by mail.kernel.org (Postfix) with ESMTP id 9C8B5619DF for ; Tue, 30 Mar 2021 15:50:11 +0000 (UTC) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S231967AbhC3Pth (ORCPT ); Tue, 30 Mar 2021 11:49:37 -0400 Received: from netrider.rowland.org ([192.131.102.5]:60125 "HELO netrider.rowland.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with SMTP id S231622AbhC3PtW (ORCPT ); Tue, 30 Mar 2021 11:49:22 -0400 Received: (qmail 980943 invoked by uid 1000); 30 Mar 2021 11:49:21 -0400 Date: Tue, 30 Mar 2021 11:49:21 -0400 From: Alan Stern To: Heikki Krogerus Cc: Greg Kroah-Hartman , Benson Leung , Prashant Malani , Guenter Roeck , linux-usb@vger.kernel.org, linux-kernel@vger.kernel.org Subject: Re: [PATCH v2 5/6] usb: Iterator for ports Message-ID: <20210330154921.GC980260@rowland.harvard.edu> References: <20210329084426.78138-1-heikki.krogerus@linux.intel.com> <20210329084426.78138-6-heikki.krogerus@linux.intel.com> <20210329184946.GA944482@rowland.harvard.edu> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: User-Agent: Mutt/1.10.1 (2018-07-13) Precedence: bulk List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Tue, Mar 30, 2021 at 12:07:35PM +0300, Heikki Krogerus wrote: > On Mon, Mar 29, 2021 at 02:49:46PM -0400, Alan Stern wrote: > > On Mon, Mar 29, 2021 at 11:44:25AM +0300, Heikki Krogerus wrote: > > > Introducing usb_for_each_port(). It works the same way as > > > usb_for_each_dev(), but instead of going through every USB > > > device in the system, it walks through the USB ports in the > > > system. > > > > > > Signed-off-by: Heikki Krogerus > > > --- > > > drivers/usb/core/usb.c | 46 ++++++++++++++++++++++++++++++++++++++++++ > > > include/linux/usb.h | 1 + > > > 2 files changed, 47 insertions(+) > > > > > > diff --git a/drivers/usb/core/usb.c b/drivers/usb/core/usb.c > > > index 2ce3667ec6fae..62368c4ed37af 100644 > > > --- a/drivers/usb/core/usb.c > > > +++ b/drivers/usb/core/usb.c > > > @@ -398,6 +398,52 @@ int usb_for_each_dev(void *data, int (*fn)(struct usb_device *, void *)) > > > } > > > EXPORT_SYMBOL_GPL(usb_for_each_dev); > > > > > > +struct each_hub_arg { > > > + void *data; > > > + int (*fn)(struct device *, void *); > > > +}; > > > + > > > +static int __each_hub(struct usb_device *hdev, void *data) > > > +{ > > > + struct each_hub_arg *arg = (struct each_hub_arg *)data; > > > + struct usb_hub *hub; > > > + int ret = 0; > > > + int i; > > > + > > > + hub = usb_hub_to_struct_hub(hdev); > > > + if (!hub) > > > + return 0; > > > > What happens if the hub is removed exactly now? Although hdev is > > reference-counted (and the loop iterator does take a reference to it), > > usb_hub_to_struct_hub doesn't take a reference to hub. And hub->ports > > isn't refcounted at all. > > If the hub is removed right now, and if hub_disconnect() also manages > to remove the ports before we have time to take the lock below, then > hdev->maxchild will be 0 by the time we can take the lock. In that > case nothing happens here. Okay, good. > If on the other hand we manage to acquire the usb_port_peer_mutex > before hub_disconnect(), then hub_disconnect() will simply have to > wait until we are done, and only after that remove the ports. > > > > + mutex_lock(&usb_port_peer_mutex); > > > + > > > + for (i = 0; i < hdev->maxchild; i++) { > > > + ret = arg->fn(&hub->ports[i]->dev, arg->data); > > > + if (ret) > > > + break; > > > + } > > > + > > > + mutex_unlock(&usb_port_peer_mutex); > > > > I have a feeling that it would be better to take and release this mutex > > in usb_for_each_port (or its caller), so that it is held over the whole > > loop. > > I disagree. The lock is for the ports, not the hubs. We should take > the lock when we are going through the ports of a hub, but release it > between the hubs. Otherwise we will be only keeping things on hold for > a long period of time for no good reason (I for example have to > evaluate the _PLD of every single port which takes a lot of time). We > don't need to prevent other things from happening to the hubs at the > same time. All right, you convinced me. Acked-by: Alan Stern Alan Stern