Return-Path: MIME-Version: 1.0 In-Reply-To: References: <1318469712-6639-1-git-send-email-kengyu@ubuntu.com> Date: Thu, 13 Oct 2011 14:23:01 +0300 Message-ID: Subject: Re: [RFC] Bluez: Fix the adapter is not initialised after resume From: Luiz Augusto von Dentz To: Keng-Yu Lin Cc: linux-bluetooth@vger.kernel.org Content-Type: text/plain; charset=ISO-8859-1 Sender: linux-bluetooth-owner@vger.kernel.org List-ID: Hi, On Thu, Oct 13, 2011 at 10:35 AM, Luiz Augusto von Dentz wrote: > > Thanks for the patch, but I suspect this may cause some problems if we > didn't receive all the pending information, do you have bluetoothd > logs showing the sequence of the events leading to this? > > -- Ive been able to reproduce it myself, it is not all the time that it happens but when it does happen it is because we missed a command (might be a kernel bug). We have a workaround for this already but it doesn't work most of the times because the adapter is not up so it doesn't accept commands, the following should fix the problem: diff --git a/plugins/hciops.c b/plugins/hciops.c index 9fb28c8..d0ba3ad 100644 --- a/plugins/hciops.c +++ b/plugins/hciops.c @@ -1482,21 +1482,6 @@ static void read_local_name_complete(int index, read_local_name_rp *rp) DBG("Got name for hci%d", index); - /* Even though it shouldn't happen (assuming the kernel behaves - * properly) it seems like we might miss the very first - * initialization commands that the kernel sends. So check for - * it here (since read_local_name is one of the last init - * commands) and resend the first ones if we haven't seen - * their results yet */ - - if (hci_test_bit(PENDING_FEATURES, &dev->pending)) - hci_send_cmd(dev->sk, OGF_INFO_PARAM, - OCF_READ_LOCAL_FEATURES, 0, NULL); - - if (hci_test_bit(PENDING_VERSION, &dev->pending)) - hci_send_cmd(dev->sk, OGF_INFO_PARAM, - OCF_READ_LOCAL_VERSION, 0, NULL); - if (!dev->pending && dev->up) init_adapter(index); } @@ -2502,8 +2487,31 @@ static void device_devup_setup(int index) hci_send_cmd(dev->sk, OGF_HOST_CTL, OCF_READ_STORED_LINK_KEY, READ_STORED_LINK_KEY_CP_SIZE, &cp); - if (!dev->pending) + if (!dev->pending) { init_adapter(index); + return; + } + + /* Even though it shouldn't happen (assuming the kernel behaves + * properly) it seems like we might miss the very first + * initialization commands that the kernel sends. So check for + * it here and resend the ones we haven't seen their results yet */ + + if (hci_test_bit(PENDING_FEATURES, &dev->pending)) + hci_send_cmd(dev->sk, OGF_INFO_PARAM, + OCF_READ_LOCAL_FEATURES, 0, NULL); + + if (hci_test_bit(PENDING_VERSION, &dev->pending)) + hci_send_cmd(dev->sk, OGF_INFO_PARAM, + OCF_READ_LOCAL_VERSION, 0, NULL); + + if (hci_test_bit(PENDING_NAME, &dev->pending)) + hci_send_cmd(dev->sk, OGF_HOST_CTL, + OCF_READ_LOCAL_NAME, 0, 0); + + if (hci_test_bit(PENDING_BDADDR, &dev->pending)) + hci_send_cmd(dev->sk, OGF_INFO_PARAM, + OCF_READ_BD_ADDR, 0, NULL); } static void init_pending(int index) -- Luiz Augusto von Dentz