| Message ID | 20240723035203.7260-1-leo.lin@canonical.com |
|---|---|
| Headers | show |
| Series | Fix L2CAP/LE/CPU/BI-02-C bluetooth certification failure | expand |
On Tue, Jul 23, 2024 at 11:52:02AM +0800, Yo-Jung (Leo) Lin wrote: > BugLink: https://bugs.launchpad.net/bugs/2072858 > > [Impact] > > Noble failed the L2CAP/LE/CPU/BI-02-C test in the Porfile Tuning Suite (PTS), > which Jammy previously could pass. > > This is due to new behavior introduced in e4b019515f950b4e6e5b74b2e1bb03a90cb33039. > (Bluetooth: Enforce validation on max value of connection interval). The kernel > only accept Connection Parameter Update Requests whose incoming conn_max_interval > are lower than the current conn_max_interval, and adjust the newer > conn_max_interval to that received max value. > > However, this behavior means that conn_max_interval can only decrease, but never > increase. This could potentially make the conditions for connection parameters > narrower over time, causing subsequent connections failed on some devices. > See the issue 847 in bluez upstream[1]. The patch 806a5198c05987b748b50f3d0c0cfb3d417381a4 > (Bluetooth: L2CAP: Fix rejecting L2CAP_CONN_PARAM_UPDATE_REQ) in the linux-next > fixed this by accepting connection parameter unconditionally. The relavent test > procedure is also on the mailing list[2]. > > [1] https://github.com/bluez/bluez/issues/847 > [2] https://linuxlists.cc/l/15/linux-bluetooth/t/5350289/(patch_v3)_bluetooth:_l2cap:_fix_rejecting_l2cap_conn_param_update_req#post5352326 > > [Fix] > Backport the from commit 806a5198c05987b748b50f3d0c0cfb3d417381a4 > (Bluetooth: L2CAP: Fix rejecting L2CAP_CONN_PARAM_UPDATE_REQ), which > already in the linux-next. > > [1] https://patchwork.kernel.org/project/bluetooth/patch/20240521143521.1568672-1-luiz.dentz@gmail.com/ > > [Test Case] > 1. Install the kernel with the backported patch > 2. Run the following test case in the PTS: > L2CAP/LE/CPU/BI-02-C > GAP/CONN/CPUP/BV-05-C > > [Where problems could occur] > This essentially revert the behavior of accepting L2CAP connection parameters > back to its original state before e4b019515f950b4e6e5b74b2e1bb03a90cb33039 > (Bluetooth: Enforce validation on max value of connection interval). > > Note that implementing restriction to the conenction parameters may take greater > effort than just adding a few checks in the kernel. The user space, notably the > bluetoothd may also need adjustments[1]. So in this case, removing the half-done > boundary checks in kernel may still do greater good if there's no plan to make > those additional changes. > > [1] https://github.com/bluez/bluez/issues/717#issuecomment-1885719058 > > Luiz Augusto von Dentz (1): > Bluetooth: L2CAP: Fix rejecting L2CAP_CONN_PARAM_UPDATE_REQ > > include/net/bluetooth/hci_core.h | 36 ++++++++++++++++++++++++++++---- > net/bluetooth/l2cap_core.c | 8 +------ > 2 files changed, 33 insertions(+), 11 deletions(-) > > -- > 2.34.1 > > > -- > kernel-team mailing list > kernel-team@lists.ubuntu.com > https://lists.ubuntu.com/mailman/listinfo/kernel-team Acked-by: Noah Wager <noah.wager@canonical.com>
On Tue, Jul 23, 2024 at 11:52:02AM +0800, Yo-Jung (Leo) Lin wrote: > BugLink: https://bugs.launchpad.net/bugs/2072858 > > [Impact] > > Noble failed the L2CAP/LE/CPU/BI-02-C test in the Porfile Tuning Suite (PTS), > which Jammy previously could pass. > > This is due to new behavior introduced in e4b019515f950b4e6e5b74b2e1bb03a90cb33039. > (Bluetooth: Enforce validation on max value of connection interval). The kernel > only accept Connection Parameter Update Requests whose incoming conn_max_interval > are lower than the current conn_max_interval, and adjust the newer > conn_max_interval to that received max value. > > However, this behavior means that conn_max_interval can only decrease, but never > increase. This could potentially make the conditions for connection parameters > narrower over time, causing subsequent connections failed on some devices. > See the issue 847 in bluez upstream[1]. The patch 806a5198c05987b748b50f3d0c0cfb3d417381a4 > (Bluetooth: L2CAP: Fix rejecting L2CAP_CONN_PARAM_UPDATE_REQ) in the linux-next > fixed this by accepting connection parameter unconditionally. The relavent test > procedure is also on the mailing list[2]. > > [1] https://github.com/bluez/bluez/issues/847 > [2] https://linuxlists.cc/l/15/linux-bluetooth/t/5350289/(patch_v3)_bluetooth:_l2cap:_fix_rejecting_l2cap_conn_param_update_req#post5352326 > > [Fix] > Backport the from commit 806a5198c05987b748b50f3d0c0cfb3d417381a4 > (Bluetooth: L2CAP: Fix rejecting L2CAP_CONN_PARAM_UPDATE_REQ), which > already in the linux-next. > > [1] https://patchwork.kernel.org/project/bluetooth/patch/20240521143521.1568672-1-luiz.dentz@gmail.com/ > > [Test Case] > 1. Install the kernel with the backported patch > 2. Run the following test case in the PTS: > L2CAP/LE/CPU/BI-02-C > GAP/CONN/CPUP/BV-05-C > > [Where problems could occur] > This essentially revert the behavior of accepting L2CAP connection parameters > back to its original state before e4b019515f950b4e6e5b74b2e1bb03a90cb33039 > (Bluetooth: Enforce validation on max value of connection interval). > > Note that implementing restriction to the conenction parameters may take greater > effort than just adding a few checks in the kernel. The user space, notably the > bluetoothd may also need adjustments[1]. So in this case, removing the half-done > boundary checks in kernel may still do greater good if there's no plan to make > those additional changes. > > [1] https://github.com/bluez/bluez/issues/717#issuecomment-1885719058 > > Luiz Augusto von Dentz (1): > Bluetooth: L2CAP: Fix rejecting L2CAP_CONN_PARAM_UPDATE_REQ > > include/net/bluetooth/hci_core.h | 36 ++++++++++++++++++++++++++++---- > net/bluetooth/l2cap_core.c | 8 +------ > 2 files changed, 33 insertions(+), 11 deletions(-) > > -- > 2.34.1 > > > -- > kernel-team mailing list > kernel-team@lists.ubuntu.com > https://lists.ubuntu.com/mailman/listinfo/kernel-team Acked-by: Manuel Diewald <manuel.diewald@canonical.com>
On Mon, Jul 22, 2024 at 11:52 PM Yo-Jung (Leo) Lin <leo.lin@canonical.com> wrote: > > BugLink: https://bugs.launchpad.net/bugs/2072858 > > [Impact] > > Noble failed the L2CAP/LE/CPU/BI-02-C test in the Porfile Tuning Suite (PTS), > which Jammy previously could pass. > > This is due to new behavior introduced in e4b019515f950b4e6e5b74b2e1bb03a90cb33039. > (Bluetooth: Enforce validation on max value of connection interval). The kernel > only accept Connection Parameter Update Requests whose incoming conn_max_interval > are lower than the current conn_max_interval, and adjust the newer > conn_max_interval to that received max value. > > However, this behavior means that conn_max_interval can only decrease, but never > increase. This could potentially make the conditions for connection parameters > narrower over time, causing subsequent connections failed on some devices. > See the issue 847 in bluez upstream[1]. The patch 806a5198c05987b748b50f3d0c0cfb3d417381a4 > (Bluetooth: L2CAP: Fix rejecting L2CAP_CONN_PARAM_UPDATE_REQ) in the linux-next > fixed this by accepting connection parameter unconditionally. The relavent test > procedure is also on the mailing list[2]. > > [1] https://github.com/bluez/bluez/issues/847 > [2] https://linuxlists.cc/l/15/linux-bluetooth/t/5350289/(patch_v3)_bluetooth:_l2cap:_fix_rejecting_l2cap_conn_param_update_req#post5352326 > > [Fix] > Backport the from commit 806a5198c05987b748b50f3d0c0cfb3d417381a4 > (Bluetooth: L2CAP: Fix rejecting L2CAP_CONN_PARAM_UPDATE_REQ), which > already in the linux-next. > > [1] https://patchwork.kernel.org/project/bluetooth/patch/20240521143521.1568672-1-luiz.dentz@gmail.com/ > > [Test Case] > 1. Install the kernel with the backported patch > 2. Run the following test case in the PTS: > L2CAP/LE/CPU/BI-02-C > GAP/CONN/CPUP/BV-05-C > > [Where problems could occur] > This essentially revert the behavior of accepting L2CAP connection parameters > back to its original state before e4b019515f950b4e6e5b74b2e1bb03a90cb33039 > (Bluetooth: Enforce validation on max value of connection interval). > > Note that implementing restriction to the conenction parameters may take greater > effort than just adding a few checks in the kernel. The user space, notably the > bluetoothd may also need adjustments[1]. So in this case, removing the half-done > boundary checks in kernel may still do greater good if there's no plan to make > those additional changes. > > [1] https://github.com/bluez/bluez/issues/717#issuecomment-1885719058 > > Luiz Augusto von Dentz (1): > Bluetooth: L2CAP: Fix rejecting L2CAP_CONN_PARAM_UPDATE_REQ > > include/net/bluetooth/hci_core.h | 36 ++++++++++++++++++++++++++++---- > net/bluetooth/l2cap_core.c | 8 +------ > 2 files changed, 33 insertions(+), 11 deletions(-) > > -- Acked-by: Kevin Becker <kevin.becker@canonical.com>
On 23/07/2024 05:52, Yo-Jung (Leo) Lin wrote: > BugLink: https://bugs.launchpad.net/bugs/2072858 > > [Impact] > > Noble failed the L2CAP/LE/CPU/BI-02-C test in the Porfile Tuning Suite (PTS), > which Jammy previously could pass. > > This is due to new behavior introduced in e4b019515f950b4e6e5b74b2e1bb03a90cb33039. > (Bluetooth: Enforce validation on max value of connection interval). The kernel > only accept Connection Parameter Update Requests whose incoming conn_max_interval > are lower than the current conn_max_interval, and adjust the newer > conn_max_interval to that received max value. > > However, this behavior means that conn_max_interval can only decrease, but never > increase. This could potentially make the conditions for connection parameters > narrower over time, causing subsequent connections failed on some devices. > See the issue 847 in bluez upstream[1]. The patch 806a5198c05987b748b50f3d0c0cfb3d417381a4 > (Bluetooth: L2CAP: Fix rejecting L2CAP_CONN_PARAM_UPDATE_REQ) in the linux-next > fixed this by accepting connection parameter unconditionally. The relavent test > procedure is also on the mailing list[2]. > > [1] https://github.com/bluez/bluez/issues/847 > [2] https://linuxlists.cc/l/15/linux-bluetooth/t/5350289/(patch_v3)_bluetooth:_l2cap:_fix_rejecting_l2cap_conn_param_update_req#post5352326 > > [Fix] > Backport the from commit 806a5198c05987b748b50f3d0c0cfb3d417381a4 > (Bluetooth: L2CAP: Fix rejecting L2CAP_CONN_PARAM_UPDATE_REQ), which > already in the linux-next. > > [1] https://patchwork.kernel.org/project/bluetooth/patch/20240521143521.1568672-1-luiz.dentz@gmail.com/ > > [Test Case] > 1. Install the kernel with the backported patch > 2. Run the following test case in the PTS: > L2CAP/LE/CPU/BI-02-C > GAP/CONN/CPUP/BV-05-C > > [Where problems could occur] > This essentially revert the behavior of accepting L2CAP connection parameters > back to its original state before e4b019515f950b4e6e5b74b2e1bb03a90cb33039 > (Bluetooth: Enforce validation on max value of connection interval). > > Note that implementing restriction to the conenction parameters may take greater > effort than just adding a few checks in the kernel. The user space, notably the > bluetoothd may also need adjustments[1]. So in this case, removing the half-done > boundary checks in kernel may still do greater good if there's no plan to make > those additional changes. > > [1] https://github.com/bluez/bluez/issues/717#issuecomment-1885719058 > > Luiz Augusto von Dentz (1): > Bluetooth: L2CAP: Fix rejecting L2CAP_CONN_PARAM_UPDATE_REQ > > include/net/bluetooth/hci_core.h | 36 ++++++++++++++++++++++++++++---- > net/bluetooth/l2cap_core.c | 8 +------ > 2 files changed, 33 insertions(+), 11 deletions(-) > They were already applied from upstream stable, but I included the buglinks to keep track of this bug in launchpad.