From mboxrd@z Thu Jan 1 00:00:00 1970 From: martin@kaiser.cx (Martin Kaiser) Date: Sun, 11 Mar 2018 17:39:46 +0100 Subject: [PATCH] clk: imx25: set correct parents for ssi ipg clocks In-Reply-To: References: <1520373735-13699-1-git-send-email-martin@kaiser.cx> <20180308140851.GA25601@botnar.kaiser.cx> <20180308164654.GA31199@botnar.kaiser.cx> Message-ID: <20180311163946.GA11507@botnar.kaiser.cx> To: linux-arm-kernel@lists.infradead.org List-Id: linux-arm-kernel.lists.infradead.org Hi Fabio & Lothar, Thus wrote Fabio Estevam (festevam at gmail.com): > On Thu, Mar 8, 2018 at 1:46 PM, Martin Kaiser wrote: > >> I can get audio to work fine without your patch on a mx25pdk. > > this is surprising. How come the ssi1_ipg_per clock is not turned off by > > clk_disable_unused()? Where is it used? Do you have > > <&clks 55> > > anywhere in your DT? > No, I don't. imx25-pdk board operates SSI in slave mode. > > (My codec chip does not use a dedicated clock line. It takes the bit clock that > > is the output of SSI. Are you maybe using ssi1_ipg_per for your codec and > > enable it there?) > The difference between our boards is that you use SSI in master mode > and mx25pdk in slave mode. > > In my first mail, I was wondering about imx25 uart1, where we also have > > uart1_ipg and uart_ipg_per and the clock seeting is > > clk[uart1_ipg] = imx_clk_gate("uart1_ipg", "ipg", ccm(CCM_CGCR2), 14); > > In this case, both uart1 and uart_ipg_per are listed in the device tree > > uart1: serial at 43f90000 { > > ... > > clocks = <&clks 120>, <&clks 57>; > > clock-names = "ipg", "per"; > > }; > > Documentation/devicetree/bindings/clock/imx25-clock.txt > > uart_ipg_per 57 > > uart1_ipg 120 > > and the driver enables both clocks explicitly. So they are not unused. > > Doing something like this is not an option for ssi, this will not work with > > imx31, 35 etc. > The solution to this is passing the "baud" clock as Lothar pointed out. ok, I got your point now. The ssi1_ipg clock does not depend on ssi1_ipg_per. This is in line with the clock generation scheme in the reference manual. And there's configurations where ssi1_ipg should be switched on, but not ssi1_ipg_per. I'll look into using the baud clock on my board. Thanks to both of you for taking the time to explain this. Best regards, Martin