From mboxrd@z Thu Jan 1 00:00:00 1970 From: Marek Szyprowski Subject: Re: Odroid U3 mutex deadlock. Date: Mon, 04 Jan 2016 11:38:29 +0100 Message-ID: <568A4BA5.3090203@samsung.com> References: <566E0319.6040100@samsung.com> Mime-Version: 1.0 Content-Type: text/plain; charset=utf-8; format=flowed Content-Transfer-Encoding: 7bit Return-path: Received: from mailout4.w1.samsung.com ([210.118.77.14]:63633 "EHLO mailout4.w1.samsung.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752841AbcADKif (ORCPT ); Mon, 4 Jan 2016 05:38:35 -0500 In-reply-to: Sender: linux-samsung-soc-owner@vger.kernel.org List-Id: linux-samsung-soc@vger.kernel.org To: Anand Moon , Krzysztof Kozlowski Cc: "linux-samsung-soc@vger.kernel.org" , Linux Kernel , Javier Martinez Canillas , Kukjin Kim , linux-arm-kernel@lists.infradead.org Hi Anand, On 2016-01-02 06:14, Anand Moon wrote: > Hi Krzysztof > > On 14 December 2015 at 05:15, Krzysztof Kozlowski > wrote: >> On 12.12.2015 13:32, Anand Moon wrote: >>> Hi Krzysztof, >>> >>> I am just observing this deadlock om my Odroid U3. >> This is not a deadlock yet, just a report from lockdep. Could be a real >> issue, could be false positive, maybe some locks miss nesting annotations. >> >> Typical information for bug report would be useful, like the exact >> version (it is mentioned in lockdep report but is it really correct?) >> and reproducibility. And of course git-bisect would be nice to have, see: >> Documentation/BUG-HUNTING >> >> Best regards, >> Krzysztof >> >> > Thanks for your inputs. I narrowed down to this commit. > ---------------------------------------------------------------------------------- > commit ec459c0c77faca53cf161830cb264e51bb1abba6 > Author: Marek Szyprowski > Date: Wed Feb 4 23:44:15 2015 +0900 > > ARM: dts: add dependency between TV and LCD0 power domains for exynos4 > > TV Mixer needs both TV and LCD0 domains enabled to be fully operational. > This dependency is modelled by making TV power domains a sub-domain of > LCD0 power domain. > > Signed-off-by: Marek Szyprowski > Signed-off-by: Kukjin Kim > ------------------------------------------------------------------------------------- > After reverting this I am not observing below lockdep warning report. > But after reverting this patch I observer another bug. The mentioned lockdep warning is a false positive. Such warning is reported for every subdomain present in a system. As you noticed that removing dependency between power domains also removes this warning, but it will introduce real problems with power on/off sequences of graphics devices. I will send a patch for genpd core fixing the deplock warning. Thanks for reporting the issue and pointing a good place to start digging in the source :) Best regards -- Marek Szyprowski, PhD Samsung R&D Institute Poland From mboxrd@z Thu Jan 1 00:00:00 1970 From: m.szyprowski@samsung.com (Marek Szyprowski) Date: Mon, 04 Jan 2016 11:38:29 +0100 Subject: Odroid U3 mutex deadlock. In-Reply-To: References: <566E0319.6040100@samsung.com> Message-ID: <568A4BA5.3090203@samsung.com> To: linux-arm-kernel@lists.infradead.org List-Id: linux-arm-kernel.lists.infradead.org Hi Anand, On 2016-01-02 06:14, Anand Moon wrote: > Hi Krzysztof > > On 14 December 2015 at 05:15, Krzysztof Kozlowski > wrote: >> On 12.12.2015 13:32, Anand Moon wrote: >>> Hi Krzysztof, >>> >>> I am just observing this deadlock om my Odroid U3. >> This is not a deadlock yet, just a report from lockdep. Could be a real >> issue, could be false positive, maybe some locks miss nesting annotations. >> >> Typical information for bug report would be useful, like the exact >> version (it is mentioned in lockdep report but is it really correct?) >> and reproducibility. And of course git-bisect would be nice to have, see: >> Documentation/BUG-HUNTING >> >> Best regards, >> Krzysztof >> >> > Thanks for your inputs. I narrowed down to this commit. > ---------------------------------------------------------------------------------- > commit ec459c0c77faca53cf161830cb264e51bb1abba6 > Author: Marek Szyprowski > Date: Wed Feb 4 23:44:15 2015 +0900 > > ARM: dts: add dependency between TV and LCD0 power domains for exynos4 > > TV Mixer needs both TV and LCD0 domains enabled to be fully operational. > This dependency is modelled by making TV power domains a sub-domain of > LCD0 power domain. > > Signed-off-by: Marek Szyprowski > Signed-off-by: Kukjin Kim > ------------------------------------------------------------------------------------- > After reverting this I am not observing below lockdep warning report. > But after reverting this patch I observer another bug. The mentioned lockdep warning is a false positive. Such warning is reported for every subdomain present in a system. As you noticed that removing dependency between power domains also removes this warning, but it will introduce real problems with power on/off sequences of graphics devices. I will send a patch for genpd core fixing the deplock warning. Thanks for reporting the issue and pointing a good place to start digging in the source :) Best regards -- Marek Szyprowski, PhD Samsung R&D Institute Poland