Return-Path: Date: Sun, 1 Jul 2018 17:19:41 +0200 From: Sebastian Andrzej Siewior To: Marcel Holtmann Cc: Alan Stern , Greg Kroah-Hartman , "open list:BLUETOOTH DRIVERS" , linux-usb@vger.kernel.org, Thomas Gleixner , Johan Hedberg Subject: Re: [PATCH] Bluetooth: btusb: use irqsave() in URB's complete callback Message-ID: <20180701151940.cbtbgpv3vncwugqz@linutronix.de> References: <20180621125207.7xtz2jpnp2vdvaau@linutronix.de> MIME-Version: 1.0 Content-Type: text/plain; charset=utf-8 In-Reply-To: List-ID: On 2018-06-21 11:34:15 [-0400], Alan Stern wrote: > On Thu, 21 Jun 2018, Sebastian Andrzej Siewior wrote: > > > On 2018-06-21 14:43:41 [+0200], Marcel Holtmann wrote: > > > can I get an ACK from someone ensuring that this is the direction we are going with the USB host controllers? > > +Alan. > > > > EHCI completes in BH since v3.12-rc1. In order to get rid of that > > local_irq_save() in USB core code I need to make sure that the USB > > device driver(s) use irqsave primitives. See > > https://lkml.kernel.org/r/Pine.LNX.4.44L0.1806011629140.1404-100000@iolanthe.rowland.org > > Hi, Marcel! Hi Marcel, > Yes, Sebastian is right. We are aiming to make it possible for the USB > core to invoke URB completion handlers with interrupts enabled, in > order to reduce latency (since USB interrupt processing can take a > fairly long time). And of course, this means completion handlers have > to work correctly regardless of whether interrupts are enabled or > disabled. I don't see this patch in linux-next. Do you still need some kind of confirmation or has this been resolved? > Currently ehci-hcd supports this possibility. Other host controller > drivers may follow along; I'd like to see xhci-hcd do this too. > > Alan Stern Sebastian