Return-Path: Message-ID: <4EB95698.4070402@codeaurora.org> Date: Tue, 08 Nov 2011 08:19:36 -0800 From: Brian Gix MIME-Version: 1.0 To: Anderson Lizardo CC: linux-bluetooth@vger.kernel.org Subject: Re: [PATCH 3/3] Bluetooth: Add address type fields to mgmt messages that need them References: <1320700419-4665-1-git-send-email-johan.hedberg@gmail.com> <1320700419-4665-3-git-send-email-johan.hedberg@gmail.com> <4EB86A27.1030403@codeaurora.org> <20111108085256.GA9733@fusion.localdomain> In-Reply-To: Content-Type: text/plain; charset=ISO-8859-1; format=flowed Sender: linux-bluetooth-owner@vger.kernel.org List-ID: Hi Anderson/Johan, On 11/8/2011 3:39 AM, Anderson Lizardo wrote: > Hi Johan/Brian, > > On Tue, Nov 8, 2011 at 4:52 AM, Johan Hedberg wrote: >> Hi Brian, >> >> On Mon, Nov 07, 2011, Brian Gix wrote: >>>> +#define MGMT_ADDR_BREDR 0x00 >>>> +#define MGMT_ADDR_LE 0x01 >>>> +#define MGMT_ADDR_BREDR_LE 0x02 >>>> +#define MGMT_ADDR_INVALID 0xff >>> >>> What would you think about overloading this Address type with the >>> Public/Random flag? >>> >>> We are already seeing devices in the marketplace with Random >>> Addresses, effectively giving LE addresses 7 significant octets of >>> address information, rather than the standard 6 octet "MAC" >>> addresses. >>> >>> >>>> + >>>> +struct mgmt_addr_info { >>>> + bdaddr_t bdaddr; >>>> + __u8 type; >>> >>> I would also support adding an additional octet here, which would be >>> essentially the "Address Type" as used in the HCI LE Connect >>> Command, and in the SMP LE Pairing protocol. > > Brian: note that this address type is being used only on mgmt > messages. So it will be mostly useful (in LE context) for discovery > command and events. > > For creating connections, we still do not have a way to indicate the > address type from userspace (currently we rely on the advertising > cache). In the solution we have been working on here, we are storing the (Public/Random) address type with the LTK, both in the BlueZ database, and with the keys that we send back to the kernel which on the reworked MGMT interface is the command "Load Long Term Keys". We still rely on the Advertising cache when pairing, but once paired can then reconnect to LE devices (and know their Public/Random setting) via the LTK list. Also remember that when we get around to supporting the LE GAP Privacy/Reconnection (Vol 3, Part C, Sections 12.3 & 12.4), and SMP Identity information (Vol 3, Part H, Sections 3.6.4 & 3.6.5) we will need to track this (Public/Random) information as part of the overall identity of remote peripherals. We will be unable to connect to "Private" devices without the Public/Random info, because they will be invisible to an LE Discovery scan. Plus Also: other than for Pairing, I think that forcing an LE Discovery each time you want to connect to a previously known device to be inelegant. > >>> I have stated elsewhere, that I think this to be crucial information >>> to store with Long Term KEYs (LTKs) as well as future LE Signing >>> keys, and in future Address Resolution solutions. The earlier we get >>> this bit of information into the MGMT interface, the better IMO. >> >> This is the same impression I got also when looking into this several >> months ago. Thanks for reminding me! >> >> The BREDR_LE option is there for the updated start_discovery command. >> You'll be able to specify whether you want BR/EDR-only, LE-only or >> interleaved discovery. I wouldn't add a completely new octet for public >> vs random information though but reuse the existing one instead. To be >> able to reuse our address type definitions for all purposes, how about >> making each value orthogonal to the others, e.g.: >> >> ADDR_BREDR 0x01 >> ADDR_LE_PUBLIC 0x02 >> ADDR_LE_RANDOM 0x04 >> >> For events like connected, found, etc you'd only have one of the bits >> set whereas for the start_discovery you could have any (but at least >> one). > > Johan: how would link_to_mgmt() in your code work in this case? > > I see this will help simplifying how do we identify LE devices on > BlueZ (currently it is done by parsing AD flags, but broken devices > may lack the correct "BR/EDR not supported" flag, and thus we > incorrectly identify then as BR/EDR). Unfortunately, it will still not > allow us to correctly connect to LE devices without relying on the > kernel's advertising cache. > > Regards, -- Brian Gix bgix@codeaurora.org Employee of Qualcomm Innovation Center, Inc. Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum