btdispatch crashed in some builders |
|||||
Issue descriptionSome test from some builders failed because unexpected btdispatch crash: https://luci-milo.appspot.com/buildbot/chromeos/guado-paladin/3234 https://luci-milo.appspot.com/buildbot/chromeos/tidus-paladin/2776
,
Apr 20 2018
btdispatch[1803]: Check failed: 1 == dbus_message_get_type(raw_message) (1 vs. 4)#012/usr/lib64/libbase-core-395517.so(_ZN4base5debug10StackTraceC1Ev+0x13) [0x7e5cdbc1f693]#012 2018-04-20T21:24:26.074031+00:00 CRIT btdispatch[4694]: Check failed: 1 == dbus_message_get_type(raw_message) (1 vs. 4)#012/usr/lib64/libbase-core-395517.so(_ZN4base5debug10StackTraceC1Ev+0x13) [0x7ae5d5d9f473]#012 Both found such message in log
,
Apr 24 2018
,
Apr 24 2018
,
Apr 25 2018
,
Apr 25 2018
In https://luci-milo.appspot.com/buildbot/chromeos/guado-paladin/3234 It said "enterprise_CFM_ConfigurableCfmTestSanity.verifications: FAIL: Unhandled AssertionError: New crash files detected: ['btdispatch.20180420.144856.9022.meta ']" Then click "Test logs", then sysinfo -> var -> spool -> crash, click any dmp.txt, it will give stack trace: So crash happened at 4 libbase-dbus-395517.so!dbus::ExportedObject::HandleMessage(DBusConnection*, DBusMessage*) [exported_object.cc : 192 + 0x2b] That's line 192 for exported_object.c I think it should be this line, since METHOD_CALL == 1 https://chromium.googlesource.com/chromium/src.git/+/68.0.3400.0/dbus/exported_object.cc#194
,
Apr 25 2018
The following revision refers to this bug: https://chromium.googlesource.com/chromium/src.git/+/a3f2862214239a61f3c8557d80a17f26f84e2133 commit a3f2862214239a61f3c8557d80a17f26f84e2133 Author: Sonny Sasaka <sonnysasaka@chromium.org> Date: Wed Apr 25 17:53:37 2018 dbus: Don't DCHECK unexpected message type, but ignore it In dbus::ExportedObject, an assumption was made that no signal could be received for an exported object, so DCHECK was put there. Now that there is a use case of D-Bus forwarding by Bluetooth dispatcher, it is fine that an exported object receives D-Bus signal messages as it is a property change signal that is handled by dbus::ObjectManager. So we can just ignore message types other than method in dbus::ExportedObject rather than DCHECK it. BUG= 835457 Change-Id: Icae03881052b3d7365a9017e5f2b987a058cd356 Reviewed-on: https://chromium-review.googlesource.com/1027081 Reviewed-by: Ryo Hashimoto <hashimoto@chromium.org> Reviewed-by: Satoru Takabayashi <satorux@google.com> Commit-Queue: Sonny Sasaka <sonnysasaka@chromium.org> Cr-Commit-Position: refs/heads/master@{#553639} [modify] https://crrev.com/a3f2862214239a61f3c8557d80a17f26f84e2133/dbus/exported_object.cc
,
Apr 26 2018
The following revision refers to this bug: https://chromium.googlesource.com/chromiumos/overlays/chromiumos-overlay/+/1e8ad444d7f4267c97a36bf4530e26a940b45f9d commit 1e8ad444d7f4267c97a36bf4530e26a940b45f9d Author: Sonny Sasaka <sonnysasaka@chromium.org> Date: Thu Apr 26 06:00:27 2018 libchrome: Don't DCHECK unexpected message type, but ignore it In dbus::ExportedObject, an assumption was made that no signal could be received for an exported object, so DCHECK was put there. Now that there is a use case of D-Bus forwarding by Bluetooth dispatcher, it is fine that an exported object receives D-Bus signal messages as it is a property change signal that is handled by dbus::ObjectManager. So we can just ignore message types other than method in dbus::ExportedObject rather than DCHECK it. This is a cherry-pick of CL:1027081. BUG=chromium:812468, chromium:835457 TEST=Build and platform2/bluetooth no longer crashes Change-Id: I2dfc7c195bc5e72d2760e923d3f62174abc3fdfc Reviewed-on: https://chromium-review.googlesource.com/1028712 Commit-Ready: Sonny Sasaka <sonnysasaka@chromium.org> Tested-by: Sonny Sasaka <sonnysasaka@chromium.org> Reviewed-by: Ben Chan <benchan@chromium.org> [add] https://crrev.com/1e8ad444d7f4267c97a36bf4530e26a940b45f9d/chromeos-base/libchrome/files/libchrome-395517-dbus-Don-t-DCHECK-unexpected-message-type-but-ignore.patch [rename] https://crrev.com/1e8ad444d7f4267c97a36bf4530e26a940b45f9d/chromeos-base/libchrome/libchrome-395517-r29.ebuild [modify] https://crrev.com/1e8ad444d7f4267c97a36bf4530e26a940b45f9d/chromeos-base/libchrome/libchrome-395517.ebuild
,
Aug 3
This bug has an owner, thus, it's been triaged. Changing status to "assigned".
,
Aug 3
|
|||||
►
Sign in to add a comment |
|||||
Comment 1 by lepton@chromium.org
, Apr 20 2018