From mboxrd@z Thu Jan 1 00:00:00 1970 From: sboyd@codeaurora.org (Stephen Boyd) Date: Tue, 19 May 2015 17:22:34 -0700 Subject: [PATCHv3 2/4] clk: socfpga: add a clock driver for the Arria 10 platform In-Reply-To: <555BC352.4000302@opensource.altera.com> References: <1431011523-10049-1-git-send-email-dinguyen@opensource.altera.com> <1431011523-10049-3-git-send-email-dinguyen@opensource.altera.com> <20150516005235.GP31753@codeaurora.org> <555B64FE.40104@opensource.altera.com> <555BB016.8000506@codeaurora.org> <555BC352.4000302@opensource.altera.com> Message-ID: <555BD3CA.6000202@codeaurora.org> To: linux-arm-kernel@lists.infradead.org List-Id: linux-arm-kernel.lists.infradead.org On 05/19/15 16:12, Dinh Nguyen wrote: > > On 5/19/15 4:50 PM, Stephen Boyd wrote: >> On 05/19/15 09:29, Dinh Nguyen wrote: >>> On 5/15/15 7:52 PM, Stephen Boyd wrote: >>>> On 05/07, dinguyen at opensource.altera.com wrote: >>>>> + >>>>> +static int socfpga_clk_prepare(struct clk_hw *hwclk) >>>>> +{ >>>>> + struct socfpga_gate_clk *socfpgaclk = to_socfpga_gate_clk(hwclk); >>>>> + struct regmap *sys_mgr_base_addr; >>>>> + int i; >>>>> + u32 hs_timing; >>>>> + u32 clk_phase[2]; >>>>> + >>>>> + if (socfpgaclk->clk_phase[0] || socfpgaclk->clk_phase[1]) { >>>>> + sys_mgr_base_addr = syscon_regmap_lookup_by_compatible("altr,sys-mgr"); >>>>> + if (IS_ERR(sys_mgr_base_addr)) { >>>> Is there a reason the syscon is grabbed lazily in prepare? Why >>>> not get it before registering this clock? >>> This syscon node is only associated with clocks that have a clk-phase >>> property, which on the SoCFPGA platform, is the SD/MMC clocks. The way >>> to implement this went through quite a few rounds of discussion for the >>> Cyclone5/Arria5 platform before settling to this method. >>> >>> The reason why syscon is grabbed here is that the setting of the clock >>> phase must be done before enabling of the clock, so it seem that prepare >>> was a good place. Should this be move moved to the socfpga_gate_init() >>> instead? >> I was expecting the regmap to be found before the clock is registered >> and stored away into the socfpga_gate_clk structure. Getting the regmap >> during prepare is akin to ioremapping a register region during prepare, >> which doesn't sound right at all. Maybe there's some good reason in the >> earlier discussions? Any hints? >> > Ah okay, the earlier discussions revolve mainly around moving the regmap > from the SD/MMC driver into the clock driver. But there weren't any > issue raised for putting the regmap in the prepare function. > > If you're curious, here are the links to the discussion for adding the > clk-phase to the driver: > > http://archive.arm.linux.org.uk/lurker/message/20131212.203042.d37c8ee9.en.html > > http://archive.arm.linux.org.uk/lurker/message/20140109.213116.1f13b27a.en.html > > But perhaps putting the regmap lookup in the init function is the > correct way to do this? Yes that would seem more appropriate. I suspect this lazy approach is done because syscon isn't ready when of_clk_init() runs though. If this was written to be a proper device driver with probe defer support this wouldn't be a problem. -- Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, a Linux Foundation Collaborative Project