From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: From: Heiko Stuebner To: Doug Anderson Cc: Michael Turquette , Stephen Boyd , Xing Zheng , Tomeu Vizoso , zhangqing , "open list:ARM/Rockchip SoC..." , linux-clk Subject: Re: [RFC PATCH 2/3] clk: adjust clocks to their requested rate after parent changes Date: Sun, 08 May 2016 22:34:41 +0200 Message-ID: <3005442.vcrD3GKCxP@phil> In-Reply-To: References: <1462206982-10444-1-git-send-email-heiko@sntech.de> MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" List-ID: Am Donnerstag, 5. Mai 2016, 17:49:39 schrieb Doug Anderson: > Hi, > > On Thu, May 5, 2016 at 5:35 PM, Doug Anderson wrote: > > I guess we don't care about errors here? > > > > ...maybe (?) we could ignore errors if we validated this rate change > > with PRE_RATE_CHANGE before attempting to change the parent clock, but > > I don't see the code doing this unless I missed it. > > One other related thought: it seems like there should be code > _somewhere_ that decides whether to adjust the child dividers before > or after the parent clock changes. Specifically if the parent clock > is increasing in speed we probably want to slow down the child clock > ahead of the parent's increase. I don't see such code here, but again > I'm pretty good at missing things unless they are slapping me in the > face. ;) yep, but that problem of a divider exceeding a requested rate temporarily exist currently already, even before this change ;-) Recently we did fix it on a small-scale for composite-clocks though [0]. Heiko [0] https://git.kernel.org/cgit/linux/kernel/git/clk/linux.git/commit/drivers/clk/clk-composite.c?h=clk-next&id=9e52cec04fd3b9b686f9256151b47fe61f7c28ef