From mboxrd@z Thu Jan 1 00:00:00 1970 From: Tony Lindgren Subject: Re: [PATCHv2] usb: musb: Fix unbalanced platform_disable Date: Sun, 18 Sep 2016 08:19:02 -0700 Message-ID: <20160918151901.k7go65s4jauldxcl@atomide.com> References: <20160912153947.k4gnggur6usyujii@atomide.com> <14745757.ahNt4Zeunj@avalon> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Return-path: Content-Disposition: inline In-Reply-To: <14745757.ahNt4Zeunj@avalon> Sender: linux-usb-owner-u79uwXL29TY76Z2rM5mHXA@public.gmane.org To: Laurent Pinchart Cc: Bin Liu , Greg Kroah-Hartman , Kishon Vijay Abraham I , Andreas Kemnade , linux-usb-u79uwXL29TY76Z2rM5mHXA@public.gmane.org, linux-omap-u79uwXL29TY76Z2rM5mHXA@public.gmane.org List-Id: linux-omap@vger.kernel.org * Laurent Pinchart [160918 05:13]: > > FYI, while this patch allows me to boot my Panda board with NFS over usbnet, > it only works with cold boots. A warm reboot results in the following warning, > and no ethernet traffic going through. The USB device is detected by the host > though. Yeah I noticed too that we still have issues. For example doing rmmod of omap2430 with gadget configured and connected will produce a hardirq-safe hardirq-unsafe lock order error. That also happens with reboot with gadget configured and connected. > I'm not sure whether this is a regression introduced by commit a83e17d0f73b > ("usb: musb: Improve PM runtime and phy handling for 2430 glue layer") or an > entirely different issue; ... > [ 5.711303] [] (_raw_spin_unlock_irqrestore) from [] > (musb_gadget_queue+0x128/0x4ac) > [ 5.711303] [] (musb_gadget_queue) from [] > (usb_ep_queue+0x38/0x1d4) > [ 5.729766] [] (usb_ep_queue) from [] > (rx_submit+0xc8/0x19c) > [ 5.737548] [] (rx_submit) from [] (rx_fill+0x7c/0xa0) > [ 5.737548] [] (rx_fill) from [] (eth_start+0x28/0x48) > [ 5.751983] [] (eth_start) from [] (eth_open+0x6c/0x7c) > [ 5.751983] [] (eth_open) from [] > (__dev_open+0x9c/0x104) This could be something else though. Care to email me your .config, maybe this is related to legacy g_ether being built in? Anyways, please also give the following patch a try. The reason why we currently have no chance of getting musb_platform_enable/disable balanced is because we need to try to set the musb devctl session bit in various places and possibly retry too. So musb_start should really be called musb_try_start_session() or something. So I think the only sane thing to do at this point is to revert the changes trying to enable/disable USB PHY from omap2430_musb_enable/disable. The other fixes are OK too as they get us a bit closer to making the platform glue calls balanced. Regards, Tony 8< ---------------------- diff --git a/drivers/usb/musb/omap2430.c b/drivers/usb/musb/omap2430.c --- a/drivers/usb/musb/omap2430.c +++ b/drivers/usb/musb/omap2430.c @@ -337,6 +337,7 @@ static int omap2430_musb_init(struct musb *musb) } musb->isr = omap2430_musb_interrupt; phy_init(musb->phy); + phy_power_on(musb->phy); l = musb_readl(musb->mregs, OTG_INTERFSEL); @@ -373,9 +374,6 @@ static void omap2430_musb_enable(struct musb *musb) struct musb_hdrc_platform_data *pdata = dev_get_platdata(dev); struct omap_musb_board_data *data = pdata->board_data; - if (!WARN_ON(!musb->phy)) - phy_power_on(musb->phy); - omap2430_set_power(musb, true, glue->cable_connected); switch (glue->status) { @@ -413,9 +411,6 @@ static void omap2430_musb_disable(struct musb *musb) struct device *dev = musb->controller; struct omap2430_glue *glue = dev_get_drvdata(dev->parent); - if (!WARN_ON(!musb->phy)) - phy_power_off(musb->phy);