Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S934052AbbDVHdv (ORCPT ); Wed, 22 Apr 2015 03:33:51 -0400 Received: from arroyo.ext.ti.com ([192.94.94.40]:50984 "EHLO arroyo.ext.ti.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1755389AbbDVHdr (ORCPT ); Wed, 22 Apr 2015 03:33:47 -0400 Message-ID: <55374EC4.40107@ti.com> Date: Wed, 22 Apr 2015 10:33:24 +0300 From: Roger Quadros User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:31.0) Gecko/20100101 Thunderbird/31.6.0 MIME-Version: 1.0 To: Peter Chen CC: "gregkh@linuxfoundation.org" , "balbi@ti.com" , "stern@rowland.harvard.edu" , "dan.j.williams@intel.com" , "Jun.Li@freescale.com" , "mathias.nyman@linux.intel.com" , "tony@atomide.com" , "linux-usb@vger.kernel.org" , "linux-kernel@vger.kernel.org" , "linux-omap@vger.kernel.org" , "macpaul@gmail.com" , "tony@atomide.com" Subject: Re: [RFC][PATCH v2 00/13] USB: OTG/DRD Core functionality References: <1429008120-5395-1-git-send-email-rogerq@ti.com> <20150420030531.GA30878@shlinux2> <5534AA8C.1070400@ti.com> <5535FD69.6070900@ti.com> <20150422021723.GB2680@shlinux2> In-Reply-To: <20150422021723.GB2680@shlinux2> Content-Type: text/plain; charset="windows-1252" Content-Transfer-Encoding: 8bit Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Content-Length: 6902 Lines: 194 On 22/04/15 05:17, Peter Chen wrote: > On Tue, Apr 21, 2015 at 10:34:01AM +0300, Roger Quadros wrote: >> On 21/04/15 09:04, Peter Chen wrote: >>> >>>> >>>> On 20/04/15 06:05, Peter Chen wrote: >>>>> On Tue, Apr 14, 2015 at 01:41:47PM +0300, Roger Quadros wrote: >>>>>> This is an attempt to centralize OTG/Dual-role functionality in the kernel. >>>>>> As of now I've got Dual-role functionality working pretty reliably on >>>>>> dra7-evm. xhci side of things for OTG/DRD use are fixed in >>>>>> http://thread.gmane.org/gmane.linux.kernel/1923161 >>>>> >>>>> Hi Roger, >>>>> >>>>> Currently, there are two main problems for DRD/OTG framework. >>>>> >>>>> - For multi-platform supports, we may define CONFIG_USB_OTG, but the >>>>> gadget should not add its otg descriptor to its configuration >>>>> descriptors if it does not support one of otg features (SRP/HNP/ADP). >>>>> Macpaul Lin's patch set [1] is the right way to do it. >>>> >>>> Agreed. That check (whether OTG descriptor can be added and which version >>>> of it) has to be done at runtime and it must be added only if hardware supports >>>> OTG _and_ kernel OTG support is enabled. >>>> >>> >>> Ok, let's put this patch set in mainline first, since your patch set may need some >>> information from it. >>> >>>>> - We are lack of framework to handle OTG (DRD) switch, it is great you >>>>> are designing it. The main problem for this framework is how to handle >>>>> DRD/OTG FSM unify. My thought is we add two paths for them separate. >>>>> For easy, I suggest if the platform supports one of otg features, then >>>>> it goes to fully otg fsm, else it goes to simply otg fsm (like your >>>>> drd fsm). If you agree with it too, you may not need to add another >>>> "dr_mode" >>>>> value. >>>> >>>> It would be nice that way but unfortunately it does't work in all cases. >>>> e.g. What if the SoC itself supports all OTG features but the board is not >>>> designed for OTG. Or the product designer simply is not interested in full OTG >>>> support but just dual-role. So we need some flexibility for the device >>>> tree/platform-data to specify that. This is where a new "dr_mode" == "dual- >>>> role" is needed. >>>> >>> >>> Since "dr_mode" has been widely used now, if we add a new property for it, we >>> need to change all drivers. >>> >>> Your OTG/DRD framework needs to (partial) use otg fsm, and we will not teach old >>> driver to use it since there are some driver related stuffs. >> >> fair enough. Let's not change dr_mode then and decide based on other parameters. >> >>> >>> SRP/HNP/ADP support can be board level capabilities, and we may consider the >>> otg device which does not support otg fsm (hardware finishes fsm). So I suggest >>> we have below properties at dts: >>> >>> - otg-support /* fully otg support */ >>> - otg-fsm-support /* fully otg fsm support */ >> >> what is the difference between otg-support and otg-fsm-support? > > Like I mentioned at above, the hardware finishes HNP/SRP which does not > use otg fsm code (usb-otg-fsm.c), most of legacy otg platforms (musb?) > use this way, for these platforms, only need to set otg-support = 1 So dr_mode = "otg" _and_ otg-support = 1? Again wouldn't this involve changes to dts for musb like platforms supporting full OTG? Instead we could add a new field saying otg-fsm-type "otg-hw", "otg-sw", "drd-sw" If the field is absent it defaults to "otg-hw". This also means we don't need otg-fsm-support flag. Now the pseudo-code to decide fsm is if (dr_mode == "otg" && CONFIG_USB_OTG) if (otg-fsm-type == "otg-sw") { if (CONFIG_USB_OTG_FSM) full otg fsm support via sw else error "CONFIG_USB_OTG_FSM" not set } else if (otg-fsm-type == "drd-sw") { dual role fsm support } else { full otg support via hw } if (otg-fsm-type == "otg else error "CONFIG_USB_OTG" not set > > For platforms which software finishes HNP/SRP using otg fsm code, need > to set both flags. > > For platforms which only do role switch through id pin, do not need to > set both. > OK. I get it now. >> >>> - otg-ver /* eh & otg supplement version */ >> >> we can get otg version from the OTG controller. What exactly is the >> otg-ver in dts for? > > Since most of otg stuffs are software's, eg, for otg descriptor, we will > only use otg 2.0 descriptor when both CONFIG_USB_OTG20 and otg-ver = 20 > are set. CONFIG_USB_OTG20 is redundant now. Plus I mentioned in the respective thread that it is not suitable for booting single image on different platforms. As of now I can see 2 inputs regarding otg-ver - One comes from otg-ver DT property or platform data. - Second that may come from OTG controller registers. e.g. It might support OTG v3.0 but system designer wants to limit to OTG v2.0 via otg-ver. Controller driver can decide among the 2 and set the appropriate otg version in the data structure. > >> >>> - adp-support /* board adp support */ >>> - srp-support /* board srp support */ >>> - hnp-support /* board hnp support */ >> >> So if these options are not provided in DTS but the OTG core supports them then >> we keep the respective feature disabled? > > Yes. > >> Won't this need dts change for existing boards? > > Does you know any dts based platform supports hnp/srp? I'm the wrong person to ask this. Maybe Felipe/Tony can comment. Irrespective of whether any dts platforms supports hsn/srp or not we must assume that till date dr_mode = otg implies full otg support and we cannot change its meaning. We can add new fields to indicate dual-role mode which is a new feature. > For chipidea platform, currently, we depends on kernel > configurations (CONFIG_USB_OTG_FSM), but it is incorrect way. > >> >> Instead how about having disable flags instead. >> - adp-disable /* board doesn't support adp */ >> - srp-disable /* board doesn't support srp */ >> - hnp-disable /* board doesn't support hnp */ >> >> Now, if the flags are not provided in dts we use the OTG core's flags. >> > > How the OTG core's know if it supports these? By OTG core, I meant OTG controller core. At least the dwc3 OTG controller has register bits to identify if it supports adp/srp/hnp. But we also need to keep in mind that adp feature can be provided separately even if the OTG controller core doesn't support it. cheers, -roger > >>> >>> Currently, if CONFIG_USB_OTG and CONFIG_USB_OTG_FSM are enabled, we will >>> have otg fsm code (usb-otg-fsm.c). >>> >>> if (otg-support & otg-fsm-support) >>> this device has fully otg support, and will follow full otg fsm transitions. >>> else >>> this device is drd, and will follow simple otg fsm transtions. >>> >> >> cheers, >> -roger >> > -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/