Hi, su, 2023-03-12 kello 22:36 -0700, Luiz Augusto von Dentz kirjoitti: > Hi Pauli, Frederic, > > On Fri, Mar 10, 2023 at 4:40 PM <patchwork-bot+bluetooth@xxxxxxxxxx> wrote: > > > > Hello: > > > > This series was applied to bluetooth/bluez.git (master) > > by Luiz Augusto von Dentz <luiz.von.dentz@xxxxxxxxx>: > > > > On Tue, 7 Mar 2023 14:24:11 -0800 you wrote: > > > From: Luiz Augusto von Dentz <luiz.von.dentz@xxxxxxxxx> > > > > > > This adds bt_crypto_sirk which attempts to generate a unique SIRK using > > > the following steps: > > > > > > - Generate a hash (k) using the str as input > > > - Generate a hash (sirk) using vendor, product, version and source as input > > > - Encrypt sirk using k as LTK with sef function. > > > > > > [...] > > > > Here is the summary with links: > > - [RFC,v2,01/12] shared/crypto: Add bt_crypto_sirk > > https://git.kernel.org/pub/scm/bluetooth/bluez.git/?id=c1dd94cc7f81 > > - [RFC,v2,02/12] shared/ad: Add RSI data type > > https://git.kernel.org/pub/scm/bluetooth/bluez.git/?id=c2e99aefd337 > > - [RFC,v2,03/12] doc: Add set-api > > https://git.kernel.org/pub/scm/bluetooth/bluez.git/?id=6477522e92e3 > > - [RFC,v2,04/12] device-api: Add Set property > > https://git.kernel.org/pub/scm/bluetooth/bluez.git/?id=5bac4cddb191 > > - [RFC,v2,05/12] core: Add initial implementation of DeviceSet interface > > (no matching commit) > > - [RFC,v2,06/12] core: Check if device has RSI > > https://git.kernel.org/pub/scm/bluetooth/bluez.git/?id=f95ffcc8b1fe > > - [RFC,v2,07/12] main.conf: Add CSIP profile configurable options > > (no matching commit) > > - [RFC,v2,08/12] shared/csip: Add initial code for handling CSIP > > https://git.kernel.org/pub/scm/bluetooth/bluez.git/?id=d297a03b7a61 > > - [RFC,v2,09/12] profiles: Add initial code for csip plugin > > https://git.kernel.org/pub/scm/bluetooth/bluez.git/?id=9e1eb0a62b3f > > - [RFC,v2,10/12] tools: Add support to generate RSI using SIRK > > https://git.kernel.org/pub/scm/bluetooth/bluez.git/?id=c446a64d461b > > - [RFC,v2,11/12] client: Add support for DeviceSet proxy > > https://git.kernel.org/pub/scm/bluetooth/bluez.git/?id=373bafc34ce6 > > - [RFC,v2,12/12] client: Use AdvertisingFlags when available > > https://git.kernel.org/pub/scm/bluetooth/bluez.git/?id=815f779aa8e4 > > > > You are awesome, thank you! > > -- > > Deet-doot-dot, I am a bot. > > https://korg.docs.kernel.org/patchwork/pwbot.html > > Let me know if you guys are happy with this interface to detect > Coordinated Sets, it still experimental so we can experiment with it > until we think it is stable, regarding the implementation of the > transports one major difference here is that we will need to encode > left and right separately, not sure how hard it is to do that in > pipewire? As far as the device set DBus interface is concerned, it seems to work fine for me currently (in wip implementation for PW [0]). Don't right now see something that would need to be added/changed in it. Channel splitting/merging is generally easy in PW. How the playback synchronization is going to work on socket level may determine a bit at what level in PW it is convenient to do though. --- Laundry list for PW related to this: * How to do TX syncronization properly with the ISO sockets needs still some thinking. I have some wip patches [2] that add the timestamps and other socket API that provide timing information to allow synchronization to the Number of Completed packets events. Corresponding Pipewire implementation [3] rate matches to keep the time difference between those events and our audio reference time fixed at e.g. 25ms (2 packets in controller). Not really clear yet if this is a right thing to do to help the controller send packets at the right time. Here I see LE Read ISO TX Sync with Intel AX210 returning only zero values in Command Complete in btmon for running CIS, so that command doesn't seem to help here. * BlueZ doesn't seem to pass on the PAC audio location it reads via read_sink/source_pac_loc, probably very easy to fix. * The CIS in a CIG cannot be started one by one, or connected to same destination. The kernel appears to wait until all CIS sockets in same CIG go to connect state before proceeding to create CIS. The spec does not seem to require this (I have some pre-rfc patches to make it more flexible [1].) * PW currently does transport acquires synchronously and fails because of that with multiple CIS, but it probably should do them async. [0] https://gitlab.freedesktop.org/pipewire/pipewire/-/merge_requests/1564 [1] https://github.com/pv/linux/commits/iso-fix-multicis [2] https://github.com/pv/linux/commits/iso-timestamp [3] https://gitlab.freedesktop.org/pvir/pipewire/-/commits/iso-timestamp-test -- Pauli Virtanen
