From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mailout1.samsung.com ([203.254.224.24]:53244 "EHLO mailout1.samsung.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1750914AbcL0IYk (ORCPT ); Tue, 27 Dec 2016 03:24:40 -0500 MIME-version: 1.0 Content-type: text/plain; charset=utf-8 Message-id: <58622544.406@samsung.com> Date: Tue, 27 Dec 2016 17:24:36 +0900 From: Chanwoo Choi To: Marek Szyprowski , cwchoi00@gmail.com Cc: linux-clk@vger.kernel.org, linux-samsung-soc , Sylwester Nawrocki , Krzysztof Kozlowski , Bartlomiej Zolnierkiewicz Subject: Re: [PATCH] clk/samsung: exynos542x: mark some clocks as critical In-reply-to: <35dcf8f3-13e2-bc38-64b2-69a42b1f8d9c@samsung.com> References: <1482399870-18563-1-git-send-email-m.szyprowski@samsung.com> <35dcf8f3-13e2-bc38-64b2-69a42b1f8d9c@samsung.com> Sender: linux-clk-owner@vger.kernel.org List-ID: Dear Marek, On 2016년 12월 27일 17:14, Marek Szyprowski wrote: > Dear Chanwoo, > > > On 2016-12-23 18:00, Chanwoo Choi wrote: >> Dear Marek, >> >> I have a question. I agree some clocks should be always on with CLK_IS_CRITICAL. >> CLK_ACLK333 is only used on exynos5420.dtsi. So, I understand to add >> CLK_IS_CRITICAL flag to CLK_ACLK333. >> >> Except for CLK_ACLK333, the remaining clocks with CLK_IS_CRITICAL were >> not used on mainline kernel (arch/arm/boot/dts/*). It means that the >> remaining clocks might be ON before applying this patch. So, >> CLK_IGNORE_UNUSED might be enough to maintain their ON state. >> >> If we don't add the CLK_IGNORE_UNUSED flag to clocks except for >> CLK_ACLK333, are there any problem? >> But, if bootloader don't turn on these clocks, this patch is necessary. > > Please note that CLK_IGNORE_UNUSED doesn't protect clock from turning off as > a result of turn off operation on one of the child-clocks. This patch fixes Right. It was my mistake that I thought the 'aclk_*' clocks are leaf clock. > the boot hang caused by changes in SYSMMU driver, which resulted in turning > off some clocks for G2D domain. Similar issues might happen for other domains > depending on the probe order. Till now it worked only by a luck, because > there was child clock enabled early enough, which protected one of those > clocks from turning off. This patch makes sense. Reviewed-by: Chanwoo Choi Regards, Chanwoo Choi [snip]