From mboxrd@z Thu Jan 1 00:00:00 1970 From: Sjoerd Simons Subject: Re: [PATCH] pwm: Add clk enable/disable for pwm_samsung_enable/pwm_samsung_disable Date: Wed, 15 Apr 2015 10:34:07 +0200 Message-ID: <1429086847.22057.159.camel@collabora.co.uk> References: <1429034708-4537-1-git-send-email-linux.amoon@gmail.com> Mime-Version: 1.0 Content-Type: text/plain; charset="UTF-8" Content-Transfer-Encoding: 8BIT Return-path: Received: from bhuna.collabora.co.uk ([93.93.135.160]:36034 "EHLO bhuna.collabora.co.uk" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1754746AbbDOIeK convert rfc822-to-8bit (ORCPT ); Wed, 15 Apr 2015 04:34:10 -0400 In-Reply-To: <1429034708-4537-1-git-send-email-linux.amoon@gmail.com> Sender: linux-pwm-owner@vger.kernel.org List-Id: linux-pwm@vger.kernel.org To: Anand Moon Cc: Thierry Reding , linux-pwm@vger.kernel.org, linux-kernel@vger.kernel.org, linux-samsung-soc@vger.kernel.org On Wed, 2015-04-15 at 03:35 +0930, Anand Moon wrote: > diff --git a/drivers/pwm/pwm-samsung.c b/drivers/pwm/pwm-samsung.c > index 3e9b583..b579753 100644 > --- a/drivers/pwm/pwm-samsung.c > +++ b/drivers/pwm/pwm-samsung.c > @@ -247,6 +247,7 @@ static int pwm_samsung_enable(struct pwm_chip *chip, struct pwm_device *pwm) > tcon &= ~TCON_MANUALUPDATE(tcon_chan); > tcon |= TCON_START(tcon_chan) | TCON_AUTORELOAD(tcon_chan); > writel(tcon, our_chip->base + REG_TCON); > + clk_prepare_enable(our_chip->base_clk); > > spin_unlock_irqrestore(&samsung_pwm_lock, flags); > > @@ -265,6 +266,7 @@ static void pwm_samsung_disable(struct pwm_chip *chip, struct pwm_device *pwm) > tcon = readl(our_chip->base + REG_TCON); > tcon &= ~TCON_AUTORELOAD(tcon_chan); > writel(tcon, our_chip->base + REG_TCON); > + clk_disable_unprepare(our_chip->base_clk); > > spin_unlock_irqrestore(&samsung_pwm_lock, flags); > } As far as i can tell this code doesn't have any effect. clk_enable is refcounted, so the clock will stay enabled for as long as the driver is loaded (as it's enabled in _probe). Your code above just raises and lowers the clocks enabled refcount, but won't actually ever cause it to be disabled. With respect to trying to disabling the clocks on pwm_disable, that will need some more work to ensure the output signal has the expected level when you turn of the clock. Specifically, when disabling from a non-100% duty state the driver relies on the PWM turning the output signal low at the end of a duty cycle. However if you turn off the clock at the start of a duty cycle while the output signal is still high it will unexpectedly remain high. -- Sjoerd Simons Collabora Ltd.