| Message ID | 20240117222317.1792594-2-luiz.dentz@gmail.com (mailing list archive) |
|---|---|
| State | Accepted |
| Commit | e108c744fa8dfa1c4f54257532f3433a47179869 |
| Headers | show |
| Series | [BlueZ,v1,1/2] client/player: Add support for printing A2DP codec details | expand |
| Context | Check | Description |
|---|---|---|
| tedd_an/pre-ci_am | success | Success |
| tedd_an/CheckPatch | success | CheckPatch PASS |
| tedd_an/GitLint | success | Gitlint PASS |
| tedd_an/IncrementalBuild | success | Incremental Build PASS |
Hi Pauli, On Wed, Jan 17, 2024 at 5:23 PM Luiz Augusto von Dentz <luiz.dentz@gmail.com> wrote: > > From: Luiz Augusto von Dentz <luiz.von.dentz@intel.com> > > This prints the owner name when it exits/quits/crash or releases a > transport. > --- > - Looks like there is a bug/race on codec switch it appears the likes of > pipewire attempts to call Release in the process which stops the > acquire/resume: > > bluetoothd[1774429]: profiles/audio/transport.c:transport_set_state() State > changed /org/bluez/hci0/dev_94_DB_56_F7_F2_88/sep1/fd8: TRANSPORT_G > bluetoothd[1774429]: profiles/audio/transport.c:media_request_create() > Request created: method=Acquire id=50 > bluetoothd[1774429]: profiles/audio/transport.c:media_owner_add() Owner > :1.133105 Request Acquire > bluetoothd[1774429]: profiles/audio/transport.c:media_transport_set_owner() > Transport /org/bluez/hci0/dev_94_DB_56_F7_F2_88/sep1/fd8 Owner :1.5 > bluetoothd[1774429]: profiles/audio/transport.c:release() Owner :1.133105 > bluetoothd[1774429]: profiles/audio/transport.c:media_owner_remove() Owner > :1.133105 Request Acquire I wonder if we have a regression on BlueZ or PW for A2DP, since I recall this used to work just fine while switching codecs back and forth, but today it didn't work at all for me and I had to reconnect a couple of times to get it working: >pipewire --version pipewire Compiled with libpipewire 1.0.0 Linked with libpipewire 1.0.0 pw.node: (bluez_output.94_DB_56_F7_F2_88.1-70) running -> error (Received error event) > profiles/audio/transport.c | 4 ++++ > 1 file changed, 4 insertions(+) > > diff --git a/profiles/audio/transport.c b/profiles/audio/transport.c > index a4696154aba5..5395985b990f 100644 > --- a/profiles/audio/transport.c > +++ b/profiles/audio/transport.c > @@ -531,6 +531,8 @@ static void media_owner_exit(DBusConnection *connection, void *user_data) > { > struct media_owner *owner = user_data; > > + DBG("Owner %s", owner->name); > + > owner->watch = 0; > > media_owner_remove(owner); > @@ -742,6 +744,8 @@ static DBusMessage *release(DBusConnection *conn, DBusMessage *msg, > if (owner == NULL || g_strcmp0(owner->name, sender) != 0) > return btd_error_not_authorized(msg); > > + DBG("Owner %s", owner->name); > + > if (owner->pending) { > const char *member; > > -- > 2.43.0 It seems you PW calls Release after Acquire, not sure if it didn't like our response to Acquire or something.
Hi, ke, 2024-01-17 kello 17:36 -0500, Luiz Augusto von Dentz kirjoitti: > On Wed, Jan 17, 2024 at 5:23 PM Luiz Augusto von Dentz > <luiz.dentz@gmail.com> wrote: > > > > From: Luiz Augusto von Dentz <luiz.von.dentz@intel.com> > > > > This prints the owner name when it exits/quits/crash or releases a > > transport. > > --- > > - Looks like there is a bug/race on codec switch it appears the likes of > > pipewire attempts to call Release in the process which stops the > > acquire/resume: > > > > bluetoothd[1774429]: profiles/audio/transport.c:transport_set_state() State > > changed /org/bluez/hci0/dev_94_DB_56_F7_F2_88/sep1/fd8: TRANSPORT_G > > bluetoothd[1774429]: profiles/audio/transport.c:media_request_create() > > Request created: method=Acquire id=50 > > bluetoothd[1774429]: profiles/audio/transport.c:media_owner_add() Owner > > :1.133105 Request Acquire > > bluetoothd[1774429]: profiles/audio/transport.c:media_transport_set_owner() > > Transport /org/bluez/hci0/dev_94_DB_56_F7_F2_88/sep1/fd8 Owner :1.5 > > bluetoothd[1774429]: profiles/audio/transport.c:release() Owner :1.133105 > > bluetoothd[1774429]: profiles/audio/transport.c:media_owner_remove() Owner > > :1.133105 Request Acquire > > I wonder if we have a regression on BlueZ or PW for A2DP, since I > recall this used to work just fine while switching codecs back and > forth, but today it didn't work at all for me and I had to reconnect a > couple of times to get it working: > > > pipewire --version > pipewire > Compiled with libpipewire 1.0.0 > Linked with libpipewire 1.0.0 I don't think there are any recent related changes on Pipewire side recently. I've also not been able to reproduce the above (w/ bluez 5.72 or current master) or seen other similar reports. Maybe there is some old bug that surfaces only in some cases. From the above log I don't get what is going wrong, maybe if you have the full bluetoothd log that shows what happened after connect with timestamps that could help. > pw.node: (bluez_output.94_DB_56_F7_F2_88.1-70) running -> error > (Received error event) This should only appear if: 1. Acquire DBus call returns error 2. Acquire DBus call results to NoReply timeout error 3. BlueZ initiates a transition back to inactive state for a transport that was previously acquired Due to the last one, the full log would be useful. > > > profiles/audio/transport.c | 4 ++++ > > 1 file changed, 4 insertions(+) > > > > diff --git a/profiles/audio/transport.c b/profiles/audio/transport.c > > index a4696154aba5..5395985b990f 100644 > > --- a/profiles/audio/transport.c > > +++ b/profiles/audio/transport.c > > @@ -531,6 +531,8 @@ static void media_owner_exit(DBusConnection *connection, void *user_data) > > { > > struct media_owner *owner = user_data; > > > > + DBG("Owner %s", owner->name); > > + > > owner->watch = 0; > > > > media_owner_remove(owner); > > @@ -742,6 +744,8 @@ static DBusMessage *release(DBusConnection *conn, DBusMessage *msg, > > if (owner == NULL || g_strcmp0(owner->name, sender) != 0) > > return btd_error_not_authorized(msg); > > > > + DBG("Owner %s", owner->name); > > + > > if (owner->pending) { > > const char *member; > > > > -- > > 2.43.0 > > It seems you PW calls Release after Acquire, not sure if it didn't > like our response to Acquire or something. >
diff --git a/profiles/audio/transport.c b/profiles/audio/transport.c index a4696154aba5..5395985b990f 100644 --- a/profiles/audio/transport.c +++ b/profiles/audio/transport.c @@ -531,6 +531,8 @@ static void media_owner_exit(DBusConnection *connection, void *user_data) { struct media_owner *owner = user_data; + DBG("Owner %s", owner->name); + owner->watch = 0; media_owner_remove(owner); @@ -742,6 +744,8 @@ static DBusMessage *release(DBusConnection *conn, DBusMessage *msg, if (owner == NULL || g_strcmp0(owner->name, sender) != 0) return btd_error_not_authorized(msg); + DBG("Owner %s", owner->name); + if (owner->pending) { const char *member;
From: Luiz Augusto von Dentz <luiz.von.dentz@intel.com> This prints the owner name when it exits/quits/crash or releases a transport. --- - Looks like there is a bug/race on codec switch it appears the likes of pipewire attempts to call Release in the process which stops the acquire/resume: bluetoothd[1774429]: profiles/audio/transport.c:transport_set_state() State changed /org/bluez/hci0/dev_94_DB_56_F7_F2_88/sep1/fd8: TRANSPORT_G bluetoothd[1774429]: profiles/audio/transport.c:media_request_create() Request created: method=Acquire id=50 bluetoothd[1774429]: profiles/audio/transport.c:media_owner_add() Owner :1.133105 Request Acquire bluetoothd[1774429]: profiles/audio/transport.c:media_transport_set_owner() Transport /org/bluez/hci0/dev_94_DB_56_F7_F2_88/sep1/fd8 Owner :1.5 bluetoothd[1774429]: profiles/audio/transport.c:release() Owner :1.133105 bluetoothd[1774429]: profiles/audio/transport.c:media_owner_remove() Owner :1.133105 Request Acquire profiles/audio/transport.c | 4 ++++ 1 file changed, 4 insertions(+)