Return-Path: MIME-Version: 1.0 In-Reply-To: <4C18BDDF.7030002@Atheros.com> References: <35c90d960912081950t135e3f10m8848e54fde1e596f@mail.gmail.com> <1260335175.2901.20.camel@violet> <35c90d960912082213s26fb0ebse75ce85d43213d9@mail.gmail.com> <1260482634.2901.70.camel@violet> <35c90d960912161359u2b3f9b2fi875288896a7a8478@mail.gmail.com> <35c90d961003091207u66571bt789461dcc7972693@mail.gmail.com> <1268167524.3712.61.camel@localhost.localdomain> <4C18BDDF.7030002@Atheros.com> Date: Wed, 16 Jun 2010 18:14:56 +0300 Message-ID: Subject: Re: RFC: Allow Bluez to select flushable or non-flushable ACL packets with L2CAP_LM_RELIABLE From: Luiz Augusto von Dentz To: Suraj Cc: Marcel Holtmann , Nick Pelly , "linux-bluetooth@vger.kernel.org" Content-Type: text/plain; charset=ISO-8859-1 List-ID: Hi, On Wed, Jun 16, 2010 at 3:04 PM, Suraj wrote: > Hi Luis, > > On 6/16/2010 5:10 PM, Luiz Augusto von Dentz wrote: >> >> Hi, >> >> On Tue, Mar 9, 2010 at 11:45 PM, Marcel Holtmann >> =A0wrote: >>> >>> Hi Nick, >>> >>>>>>>>> Right now Bluez always requests flushable ACL packets (but does n= ot >>>>>>>>> set a flush timeout, so effectively they are non-flushable): >>>>>>>>> >>>>>>>>> However it is desirable to use an ACL flush timeout on A2DP packe= ts >>>>>>>>> so >>>>>>>>> that if the ACL packets block for some reason then the LM can flu= sh >>>>>>>>> them to make room for newer packets. >>>>>>>>> >>>>>>>>> Is it reasonable for Bluez to use the 0x00 ACL packet boundary fl= ag >>>>>>>>> by >>>>>>>>> default (non-flushable packet), and let userspace request flushab= le >>>>>>>>> packets on A2DP L2CAP sockets with the socket option >>>>>>>>> L2CAP_LM_RELIABLE. >>>>>>>> >>>>>>>> the reliable option has a different meaning. It comes back from th= e >>>>>>>> old >>>>>>>> Bluetooth 1.1 qualification days where we had to tests on L2CAP th= at >>>>>>>> had >>>>>>>> to confirm that we can detect malformed packets and report them. >>>>>>>> These >>>>>>>> days it is just fine to drop them. >>>>>>> >>>>>>> Got it, how about introducing >>>>>>> >>>>>>> #define L2CAP_LM_FLUSHABLE 0x0040 >>>>>> >>>>>> that l2cap_sock_setsockopt_old() sets this didn't give you a hint th= at >>>>>> we might wanna deprecate this socket options ;) >>>>>> >>>>>> I need to read up on the flushable stuff, but in the end it deserves >>>>>> its >>>>>> own socket option. Also an ioctl() to actually trigger Enhanced flus= h >>>>>> might be needed. >>>>>> >>>>>>> struct l2cap_pinfo { >>>>>>> =A0 =A0... >>>>>>> =A0 =A0__u8 flushable; >>>>>>> } >>>>>> >>>>>> Sure. In the long run we need to turn this into a bitmask. We are ju= st >>>>>> wasting memory here. >>>>> >>>>> Attached is an updated patch, that checks the LMP features bitmask >>>>> before using the new non-flushable packet type. >>>>> >>>>> I am still using L2CAP_LM_FLUSHABLE socket option in >>>>> l2cap_sock_setsockopt_old(), which I don't think you are happy with. >>>>> So how about a new option: >>>>> >>>>> SOL_L2CAP, L2CAP_ACL_FLUSH >>>>> which has a default value of 0, and can be set to 1 to make the ACL >>>>> data sent by this L2CAP socket flushable. >>>>> >>>>> In a later commit we would then add >>>>> SOL_ACL, ACL_FLUSH_TIMEOUT >>>>> That is used to set an automatic flush timeout for the ACL link on a >>>>> L2CAP socket. Note that SOL_ACL is new. >>>>> >>>>> But maybe this is not what you had in mind, so I'm looking for your >>>>> advice before I implement this. >>>> >>>> Attached an updated patch for 2.6.32 kernel. We've been using this >>>> patch successfully on production devices. >>> >>> can see anything wrong with that patch. However we need to use >>> SOL_BLUETOOTH for it of course. So we need to come up with something to >>> make this simple. >>> >>> An additional change I like to see is to use flags for booleans like >>> flushable in the structures. Can you work on changing that. >>> >>> Also do we have decoding support for this in hcidump. It might be nice >>> to include some really simple examples in the commit message. >>> >>> Regards >> >> I would like to play a little bit with this, so is there any missing >> updates? >> > This is not exactly something related to your question, but there is anot= her > side effect for the current implementation. > > Assume you have 2 ACL links, FTP and A2DP. A2DP streaming and FTP doing F= TP > Put. > When the A2DP packets start blocking, it effectively start blocking the > packets available for FTP too. But, the host has no idea about it and kee= p > pumping in A2DP data until all available buffers are blocked. Effectively > blocking both A2DP and FTP. > > So at the user level you will see your FTP connection stalling as long yo= ur > A2DP connection is stalled (out of range). FTP will restart as soon as A2= DP > comes back in range. > > I had raised this issue sometime before, but could not follow it up. I got the impression that we can still control which packets are Automatically-Flushable and which are not, so even thought we set the timeout in a per ACL link fashion we can still mark which packets should be flushable in a per socket basis. Is that correct, Nick? --=20 Luiz Augusto von Dentz Computer Engineer