Return-Path: MIME-Version: 1.0 In-Reply-To: <545CD2CB.3050500@LilandIT.com> References: <5457A17D.9060404@LilandIT.com> <5458C0CD.9010907@LilandIT.com> <545A5CA3.1000807@LilandIT.com> <545CD2CB.3050500@LilandIT.com> Date: Fri, 7 Nov 2014 12:32:07 -0800 Message-ID: Subject: Re: How to get started on a GATT Server ? From: Arman Uguray To: "Urbani, Edmund" Cc: "linux-bluetooth@vger.kernel.org" Content-Type: text/plain; charset=UTF-8 Sender: linux-bluetooth-owner@vger.kernel.org List-ID: Hi Edmund, > On Fri, Nov 7, 2014 at 6:10 AM, Urbani, Edmund wrote: > On 11/05/2014 06:47 PM, Arman Uguray wrote: >> >> Hi Edmund, >> >>> On Wed, Nov 5, 2014 at 9:21 AM, Urbani, Edmund >>> wrote: >>> On 11/04/2014 08:11 PM, Arman Uguray wrote: >>>> >>>> Hi Edmund, >>>> >>>> On Tue, Nov 4, 2014 at 4:04 AM, Urbani, Edmund >>>> wrote: >>>>> >>>>> On 11/03/2014 08:32 PM, Arman Uguray wrote: >>>>>> >>>>>> Hi Edmund, >>>>>> >>>>>>> On Mon, Nov 3, 2014 at 7:38 AM, Urbani, Edmund >>>>>>> wrote: >>>>>>> Hello all, >>>>>>> >>>>>>> I am trying to write a simple GATT-based service of my own. Right now >>>>>>> I >>>>>>> don't really know which docs I should be looking at. I have skimmed >>>>>>> through >>>>>>> bluez sources (5.23) and found gatt-service.h and gatt.h, as well as >>>>>>> the >>>>>>> gatt-example plugin. >>>>>>> >>>>>>> So far I am still trying to figure out whether I should be going the >>>>>>> C >>>>>>> API >>>>>>> route or maybe use the D-Bus interface. Also gatt-service.c seems to >>>>>>> suggest >>>>>>> that there is a way to register a GATT service over D-Bus but running >>>>>>> it >>>>>>> gives me this: >>>>>>> RegisterService: Method "RegisterService" with signature "oa{sv}" on >>>>>>> interface "org.bluez.GattManager1" doesn't exist" >>>>>> >>>>>> The reason you're not seeing GattManager1 is because it's hidden >>>>>> behind the "experimental" flag. So you need to configure and run bluez >>>>>> with the --enable-experimental and the --experimental flags >>>>>> respectively. >>>>>> >>>>>> That said, that API implementation isn't complete and isn't in fact >>>>>> even wired up with the internal attrib-server. We're working towards >>>>>> rewriting all of the C API and finally implementing the proposed D-Bus >>>>>> API but until then, I would ignore doc/gatt-api.txt (it's only a >>>>>> proposal at the moment and isn't even included in release tarballs). >>>>>> So for now you have to use the C API route. There will probably be an >>>>>> announcement once that D-Bus API is finally supported. >>>>> >>>>> >>>>>>> Any pointers to help me get started would be appreciated. Thanks! >>>>>>> >>>>>>> Kind regards, >>>>>>> Edmund >>>>>> >>>>>> Cheers, >>>>>> Arman >>>>> >>>>> >>>>> Thanks, I won't bother with the D-Bus interface then. But if the C API >>>>> is >>>>> being rewritten now, that sounds like there is currently no stable API >>>>> available at all. :/ Or is it just the internals being rewritten? >>>>> >>>> The C API for GATT interactions is being rewritten, i.e. the code in >>>> attrib/* is being deprecated in favor of code in src/shared/gatt*. You >>>> can still keep using the existing APIs though at some point we will >>>> rip out the existing attrib-server to be a combination of >>>> shared/gatt-db and shared/gatt-server. So, make sure you keep track of >>>> the patches going in. >>>> >>>> Keep in mind that if you're developing your code on top of the master >>>> branch then things are meant to be in flight and unstable there, so >>>> you should follow releases anyway. I think we'll figure out a way so >>>> that the C based GATT server/client won't be entirely broken for >>>> people in between official BlueZ releases. >>>> >>>>> So I assume, the fastest route the get a prototype that does something, >>>>> would be to get gatt-example.c to run and then modify that, right? >>>>> >>>> Yes, that is the easiest way right now. When the new C APIs are mature >>>> enough we will convert gatt-example to use the new code but that's >>>> still a few months down the line. >>>> >>>>> Kind regards, >>>>> Edmund >>>>> >>>>> >>>>> -- >>>>> To unsubscribe from this list: send the line "unsubscribe >>>>> linux-bluetooth" >>>>> in >>>>> the body of a message to majordomo@vger.kernel.org >>>>> More majordomo info at http://vger.kernel.org/majordomo-info.html >>>> >>>> Cheers, >>>> Arman >>> >>> >>> Thanks, Arman. I now have a basic service with one characteristic and a >>> read-only value up and running (I can see and read its value from another >>> system using gatttool... though my iOS app still fails to discover the >>> service :/ ). >>> >>> What I actually need however is a read/write attribute and a way to react >>> to >>> write requests. The examples in gatt-service.c all seem to be read-only. >>> How >>> would I get a call-back for an incoming write request? >>> >> It's not obvious from reading the existing code but the >> "gatt_service_add" function defined attrib/gatt-service.h allows you >> to specify read/write handlers using options passed via its variable >> length argument list. You can look at profiles/proximity/linkloss.c >> for an example of this. Looking at the link_loss_register function, I >> see the profile register both a read and a write callback for the >> Alert Level characteristic (the functions link_loss_alert_lvl_read and >> link_loss_alert_lvl_write). I think this is what you're looking for. >> >> That said, the new API will hopefully make this more readable by doing >> away with the variable argument list and having explicit functions and >> arguments for inserting characteristics into the database with >> read/write handlers. >> >>> Kind regards, >>> Edmund >>> >>> -- >>> To unsubscribe from this list: send the line "unsubscribe >>> linux-bluetooth" >>> in >>> the body of a message to majordomo@vger.kernel.org >>> More majordomo info at http://vger.kernel.org/majordomo-info.html >> >> Cheers, >> Arman > > Alright, I have a GATT service up and running with handlers for read and > write requests, and an iOS app that acts as the client/central. > > The only thing that is still now working the way it should is service > discovery. Currently the client finds the service by connecting to any > device which advertises itself as connectable and then queries its service > UUIDs. Apparently my service UUID (currently still 16 bit) is not > advertised, and so I have to resort to this kind of slow, battery-consuming > service discovery. How do I get BlueZ to include the service UUID in LE > advertisements? > As far as I know, there is a mgmt command for this which you could invoke directly from your plugin. I'm not sure if bluetoothd actually advertises any of the UUIDs registered in profiles. There is definitely no D-Bus API for this but others might be able to answer the current state affairs regarding hosted GATT services and if the daemon can advertise them. So, using the mgmt API directly might be your best bet for now. > This is the last obstacle to a demo-ready prototype for my project. Thanks > for everything so far! > > And as for future development and production use - any guesses when a BlueZ > version with a stable GATT service interface over D-Bus might be released? > I can't speak for certain regarding a specific BlueZ release. My hope is to have a D-Bus API for GATT client role in the upcoming month (at least as experimental) and one for GATT server role, along with potential D-Bus APIs to put the device into advertising mode with certain service UUIDs being advertised in the following months. At least by the end of the year, we should have clean and easy to use C APIs in src/shared for both client and server roles finished. So, keep following the mailing list for patches and discussions that fly by; there's certainly a lot of activity around completing GATT API support now that we have most of the shared code in place. > Kind regards, > Edmund Urbani > > > -- > To unsubscribe from this list: send the line "unsubscribe linux-bluetooth" > in > the body of a message to majordomo@vger.kernel.org > More majordomo info at http://vger.kernel.org/majordomo-info.html Cheers, Arman