Received: by 2002:a25:8b91:0:0:0:0:0 with SMTP id j17csp5612473ybl; Sun, 22 Dec 2019 09:59:48 -0800 (PST) X-Google-Smtp-Source: APXvYqwDsSsuN0POOWnp6j3FMLRrpkmldrNJR3qm/PUwApam6bjVie2eaiExZ7Kb+SbRZcHh2ABN X-Received: by 2002:a05:6830:121a:: with SMTP id r26mr21091996otp.225.1577037588005; Sun, 22 Dec 2019 09:59:48 -0800 (PST) ARC-Seal: i=1; a=rsa-sha256; t=1577037588; cv=none; d=google.com; s=arc-20160816; b=FG8OR47B5VFm7T+UapRQDPYPbyd9z1FAT6YzoTX9KlcBhY2eyHuBh0HnPFx2ARDLZ1 KjKaXk1B085dLhBw/Ch9DPe3ZGQ/OdC8+febs8SD7QDu0B8QPpj8dZH0BNHOPsEzQeB0 eXjEPsjteRXjiborMZiZCdgmE/szRwzwhR4RpNMFdd3LrS7VDl2lUvjEb/cMlQv/yb17 qdsm2KQcX2m28RLFsPqzQjNyS351EcVNMLd9F7v8osKB03zYEHaN3MvNBjyfKxvbJSxH xwq3Yzt2ZdtdGkyVnEAxJWCgvNxwR/Ff3AhQcjdzcbfnST2G4ogzOfsJ+Qlk1a0MMKYe p8dA== ARC-Message-Signature: i=1; a=rsa-sha256; c=relaxed/relaxed; d=google.com; s=arc-20160816; h=list-id:precedence:sender:content-transfer-encoding:mime-version :references:in-reply-to:message-id:subject:cc:to:from:date :dkim-signature; bh=cwc6+UtUN2GeYUuuZYyjwu/DcC0sBDmqqp4keF3O/o0=; b=ZltR0No2yByuXVG70j/kcBMa/iyH+yMwZ12OXgq2CldhebBtTiY6GJOPbGQYWYaI8L OexXD84nDQZ8mAxHf0GApZXw1uYr+G0DqwoP1OCn7lcEmRcqC2U2HipyRJ2sIilyoFFv Ix0GmQdJSjypLeuzWN/bEbh1BfPoYWyeXuwbX0AeTWiuCyipTTd91eTBsb1K0OiIty46 g0yfpaQIHmvmoKRh5ai7kHZYdeLUrN9E93/BjBWRiF79nJgP3xbXdTe+J1SDgbg1TNDx qY4QKwqXSzVXtBhYPWFv4eMjbVuXrSb+tmGAGBvat9+xdy7cWDaAJoZwmate9ZYlqEyz B/6g== ARC-Authentication-Results: i=1; mx.google.com; dkim=pass header.i=@kernel.org header.s=default header.b=OyNgvQeQ; spf=pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) smtp.mailfrom=linux-kernel-owner@vger.kernel.org; dmarc=pass (p=NONE sp=NONE dis=NONE) header.from=kernel.org Return-Path: Received: from vger.kernel.org (vger.kernel.org. [209.132.180.67]) by mx.google.com with ESMTP id l6si8928951oti.249.2019.12.22.09.59.34; Sun, 22 Dec 2019 09:59:47 -0800 (PST) Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67; Authentication-Results: mx.google.com; dkim=pass header.i=@kernel.org header.s=default header.b=OyNgvQeQ; spf=pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) smtp.mailfrom=linux-kernel-owner@vger.kernel.org; dmarc=pass (p=NONE sp=NONE dis=NONE) header.from=kernel.org Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1726139AbfLVR6z (ORCPT + 99 others); Sun, 22 Dec 2019 12:58:55 -0500 Received: from mail.kernel.org ([198.145.29.99]:39904 "EHLO mail.kernel.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1725922AbfLVR6z (ORCPT ); Sun, 22 Dec 2019 12:58:55 -0500 Received: from archlinux (cpc149474-cmbg20-2-0-cust94.5-4.cable.virginm.net [82.4.196.95]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by mail.kernel.org (Postfix) with ESMTPSA id EDA2E206B7; Sun, 22 Dec 2019 17:58:50 +0000 (UTC) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/simple; d=kernel.org; s=default; t=1577037532; bh=1jmsPtm7NjVpHetqaHWBeEv7jqOxvg8KP2OXtwePE3Q=; h=Date:From:To:Cc:Subject:In-Reply-To:References:From; b=OyNgvQeQA0AWFMLoK3cY9y0w2+HzT5pIMF+Yg2CR5bmxjxjtQ4CbKSMfP5SCkMyUs n84yztL+iPmmqF/wCXZwbJsUk3E00gZsMqifxUXOC4N7eJuqEFlwnqFxEQ90+Cfut3 ALIyV4T0o2RG6OXBSV1UtkRcVaRBHv4itxaibIrQ= Date: Sun, 22 Dec 2019 17:58:47 +0000 From: Jonathan Cameron To: Mircea Caprioru Cc: , , , , , , , Subject: Re: [PATCH 2/3] iio: frequency: ltc6952: Add support for LTC6952 Message-ID: <20191222175847.0b7f8da9@archlinux> In-Reply-To: <20191219134810.6677-2-mircea.caprioru@analog.com> References: <20191219134810.6677-1-mircea.caprioru@analog.com> <20191219134810.6677-2-mircea.caprioru@analog.com> X-Mailer: Claws Mail 3.17.4 (GTK+ 2.24.32; x86_64-pc-linux-gnu) MIME-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit Sender: linux-kernel-owner@vger.kernel.org Precedence: bulk List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Thu, 19 Dec 2019 15:48:09 +0200 Mircea Caprioru wrote: > LTC6952 is a high performance, ultralow jitter, JESD204B/C clock generation > and distribution IC. It includes a Phase Locked Loop (PLL) core, consisting > of a reference divider, phase-frequency detector (PFD) with a phase-lock > indicator, ultralow noise charge pump and integer feedback divider. > > The LTC6952's 11 outputs can be configured as up to 5 JESD204B/C subclass 1 > device clock/SYSREF pairs plus one general purpose output, or simply 11 > general purpose clock outputs for non-JESD204B/C applications. > > Datasheet: https://www.analog.com/media/en/technical-documentation/data-sheets/ltc6952.pdf > > Signed-off-by: Mircea Caprioru Hi Mircea, I think we need a clean answer to whether a clock generation device should have a driver in IIO because it's targeted at usecases that are going to involve other IIO components (I would imagine JESD204B/C ADC/DAC) The answer is probably that it does make sense for these to be in IIO, but it's not entirely obvious so the patch set should set out the arguments for IIO vs clock framework only (perhaps with an IIO bridge driver). This also needs to go to an appropriate clock list and maintainers for review. Various other bits inline. Please don't do the ret |= thing in particular. If you get a series of errors in one of those sequences you end up with really odd error codes and a great deal of confusion from any users who hit them! Thanks, Jonathan > --- > drivers/iio/frequency/Kconfig | 12 + > drivers/iio/frequency/Makefile | 1 + > drivers/iio/frequency/ltc6952.c | 896 ++++++++++++++++++++++++++++++++ > 3 files changed, 909 insertions(+) > create mode 100644 drivers/iio/frequency/ltc6952.c > > diff --git a/drivers/iio/frequency/Kconfig b/drivers/iio/frequency/Kconfig > index 240b81502512..3ee1295d1139 100644 > --- a/drivers/iio/frequency/Kconfig > +++ b/drivers/iio/frequency/Kconfig > @@ -21,6 +21,18 @@ config AD9523 > To compile this driver as a module, choose M here: the > module will be called ad9523. > > +config LTC6952 > + tristate "Analog Devices LTC6952 Clock Ultralow Jitter with JESD204B/C" > + depends on SPI > + depends on COMMON_CLK > + help > + Say yes here to build support for Analog Devices LTC6952 Clock Ultralow > + Jitter Attenuator with JESD204B/C. The driver provides direct access > + via sysfs. > + > + To compile this driver as a module, choose M here: the > + module will be called ltc6952. > + > endmenu > > # > diff --git a/drivers/iio/frequency/Makefile b/drivers/iio/frequency/Makefile > index 518b1e50caef..ccc20904a327 100644 > --- a/drivers/iio/frequency/Makefile > +++ b/drivers/iio/frequency/Makefile > @@ -7,3 +7,4 @@ > obj-$(CONFIG_AD9523) += ad9523.o > obj-$(CONFIG_ADF4350) += adf4350.o > obj-$(CONFIG_ADF4371) += adf4371.o > +bj-$(CONFIG_LTC6952) += ltc6952.o > diff --git a/drivers/iio/frequency/ltc6952.c b/drivers/iio/frequency/ltc6952.c > new file mode 100644 > index 000000000000..d9b7250febbe > --- /dev/null > +++ b/drivers/iio/frequency/ltc6952.c > @@ -0,0 +1,896 @@ > +// SPDX-License-Identifier: GPL-2.0 > +/* > + * Driver for LTC6952 ultralow jitter, JESD204B/C clock generation IC. > + * > + * Copyright 2019 Analog Devices Inc. > + */ > + > +#include > +#include > +#include > +#include > +#include > +#include > +#include > + > +#include > +#include > + > +#include > +#include > + > +/* Register address macro */ > +#define LTC6952_REG(x) (x) > + > +/* LTC6952_REG0 */ > +#define LTC6952_UNLOCK_MSK BIT(6) > +#define LTC6952_LOCK_MSK BIT(4) > +#define LTC6952_VCOOK_MSK BIT(2) > +#define LTC6952_REFOK_MSK BIT(0) > + > +/* LTC6952_REG1 */ > +#define LTC6952_INVSTAT_MSK BIT(7) > +#define LTC6952_INVSTAT(x) FIELD_PREP(LTC6952_INVSTAT_MSK, x) > +#define LTC6952_STAT_OUT_MSK GENMASK(6, 0) > +#define LTC6952_STAT_OUT(x) FIELD_PREP(LTC6952_STAT_OUT_MSK, x) > + > +/* LTC6952_REG2 */ > +#define LTC6952_PDALL_MSK BIT(7) > +#define LTC6952_PDALL(x) FIELD_PREP(LTC6952_PDALL_MSK, x) > +#define LTC6952_PDPLL_MSK BIT(6) > +#define LTC6952_PDPLL(x) FIELD_PREP(LTC6952_PDPLL_MSK, x) > +#define LTC6952_PDVCOPK_MSK BIT(5) > +#define LTC6952_PDVCOPK(x) FIELD_PREP(LTC6952_PDVCOPK_MSK, x) > +#define LTC6952_PDREFPK_MSK BIT(4) > +#define LTC6952_PDREFPK(x) FIELD_PREP(LTC6952_PDREFPK_MSK, x) > +#define LTC6952_BST_MSK BIT(3) > +#define LTC6952_BST(x) FIELD_PREP(LTC6952_BST_MSK, x) > +#define LTC6952_FILTR_MSK BIT(2) > +#define LTC6952_FILTR(x) FIELD_PREP(LTC6952_FILTR_MSK, x) > +#define LTC6952_FILTV_MSK BIT(1) > +#define LTC6952_FILTV(x) FIELD_PREP(LTC6952_FILTV_MSK, x) > +#define LTC6952_POR_MSK BIT(0) > +#define LTC6952_POR(x) FIELD_PREP(LTC6952_POR_MSK, x) > + > +/* LTC6952_REG3 */ > +#define LTC6952_PD3_MSK GENMASK(7, 6) > +#define LTC6952_PD3(x) FIELD_PREP(LTC6952_PD3_MSK, x) > +#define LTC6952_PD2_MSK GENMASK(5, 4) > +#define LTC6952_PD2(x) FIELD_PREP(LTC6952_PD2_MSK, x) > +#define LTC6952_PD1_MSK GENMASK(3, 2) > +#define LTC6952_PD1(x) FIELD_PREP(LTC6952_PD1_MSK, x) > +#define LTC6952_PD0_MSK GENMASK(1, 0) > +#define LTC6952_PD0(x) FIELD_PREP(LTC6952_PD0_MSK, x) > + > +/* LTC6952_REG4 */ > +#define LTC6952_PD7_MSK GENMASK(7, 6) > +#define LTC6952_PD7(x) FIELD_PREP(LTC6952_PD7_MSK, x) > +#define LTC6952_PD6_MSK GENMASK(5, 4) > +#define LTC6952_PD6(x) FIELD_PREP(LTC6952_PD6_MSK, x) > +#define LTC6952_PD5_MSK GENMASK(3, 2) > +#define LTC6952_PD5(x) FIELD_PREP(LTC6952_PD5_MSK, x) > +#define LTC6952_PD4_MSK GENMASK(1, 0) > +#define LTC6952_PD4(x) FIELD_PREP(LTC6952_PD4_MSK, x) > + > +/* LTC6952_REG5 */ > +#define LTC6952_TEMPO_MSK BIT(7) > +#define LTC6952_TEMPO(x) FIELD_PREP(LTC6952_TEMPO_MSK, x) > +#define LTC6952_PD10_MSK GENMASK(5, 4) > +#define LTC6952_PD10(x) FIELD_PREP(LTC6952_PD10_MSK, x) > +#define LTC6952_PD9_MSK GENMASK(3, 2) > +#define LTC6952_PD9(x) FIELD_PREP(LTC6952_PD9_MSK, x) > +#define LTC6952_PD8_MSK GENMASK(1, 0) > +#define LTC6952_PD8(x) FIELD_PREP(LTC6952_PD8_MSK, x) > + > +#define LTC6952_PD_MSK(ch) GENMASK(((ch) & 0x03) * 2 + 1, ((ch) & 0x03) * 2) > +#define LTC6952_PD(ch, x) ((x) << ((ch) & 0x03)) > + > +/* LTC6952_REG6 */ > +#define LTC6952_RAO_MSK BIT(7) > +#define LTC6952_RAO(x) FIELD_PREP(LTC6952_RAO_MSK, x) > +#define LTC6952_PARSYNC_MSK BIT(6) > +#define LTC6952_PARSYNC(x) FIELD_PREP(LTC6952_PARSYNC_MSK, x) > +#define LTC6952_LKWIN_MSK BIT(4) > +#define LTC6952_LKWIN(x) FIELD_PREP(LTC6952_LKWIN_MSK, x) > +#define LTC6952_LKCT_MSK GENMASK(3, 2) > +#define LTC6952_LKCT(x) FIELD_PREP(LTC6952_LKCT_MSK, x) > +#define LTC6952_RD_HIGH_MSK GENMASK(1, 0) > +#define LTC6952_RD_HIGH(x) FIELD_PREP(LTC6952_RD_HIGH_MSK, x) > + > +/* LTC6952_REG7 */ > +#define LTC6952_RD_LOW_MSK GENMASK(7, 0) > +#define LTC6952_RD_LOW(x) FIELD_PREP(LTC6952_RD_LOW_MSK, x) In general I'm in favour of using FIELD_PREP, but when you end up just mapping to the whole register it seems a bit odd. > + > +/* LTC6952_REG8 */ > +#define LTC6952_ND_HIGH_MSK GENMASK(7, 0) > +#define LTC6952_ND_HIGH(x) FIELD_PREP(LTC6952_ND_HIGH_MSK, x) > + > +/* LTC6952_REG9 */ > +#define LTC6952_ND_LOW_MSK GENMASK(7, 0) > +#define LTC6952_ND_LOW(x) FIELD_PREP(LTC6952_ND_LOW_MSK, x) > + > +/* LTC6952_REG10 */ > +#define LTC6952_CPRST_MSK BIT(7) > +#define LTC6952_CPRST(x) FIELD_PREP(LTC6952_CPRST_MSK, x) > +#define LTC6952_CPUP_MSK BIT(6) > +#define LTC6952_CPUP(x) FIELD_PREP(LTC6952_CPUP_MSK, x) > +#define LTC6952_CPDN_MSK BIT(5) > +#define LTC6952_CPDN(x) FIELD_PREP(LTC6952_CPDN_MSK, x) > +#define LTC6952_CP_MSK GENMASK(4, 0) > +#define LTC6952_CP(x) FIELD_PREP(LTC6952_CP_MSK, x) > + > +/* LTC6952_REG11 */ > +#define LTC6952_CPMID_MSK BIT(7) > +#define LTC6952_CPMID(x) FIELD_PREP(LTC6952_CPMID_MSK, x) > +#define LTC6952_CPWIDE_MSK BIT(6) > +#define LTC6952_CPWIDE(x) FIELD_PREP(LTC6952_CPWIDE_MSK, x) > +#define LTC6952_CPINV_MSK BIT(5) > +#define LTC6952_CPINV(x) FIELD_PREP(LTC6952_CPINV_MSK, x) > +#define LTC6952_EZMD_MSK BIT(4) > +#define LTC6952_EZMD(x) FIELD_PREP(LTC6952_EZMD_MSK, x) > +#define LTC6952_SRQMD_MSK BIT(3) > +#define LTC6952_SRQMD(x) FIELD_PREP(LTC6952_SRQMD_MSK, x) > +#define LTC6952_SYSCT_MSK GENMASK(2, 1) > +#define LTC6952_SYSCT(x) FIELD_PREP(LTC6952_SYSCT_MSK, x) > +#define LTC6952_SSRQ_MSK BIT(0) > +#define LTC6952_SSRQ(x) FIELD_PREP(LTC6952_SSRQ_MSK, x) > + > +/* LTC6952_REG12,16,20,24,28,32,36,40,44,48,52 */ > +#define LTC6952_MP_MSK GENMASK(7, 3) > +#define LTC6952_MP(x) FIELD_PREP(LTC6952_MP_MSK, x) > +#define LTC6952_MD_MSK GENMASK(2, 0) > +#define LTC6952_MD(x) FIELD_PREP(LTC6952_MD_MSK, x) > + > +/* LTC6952_REG13,17,21,25,29,33,37,41,45,49,53 */ > +#define LTC6952_SRQEN_MSK BIT(7) > +#define LTC6952_SRQEN(x) FIELD_PREP(LTC6952_SRQEN_MSK, x) > +#define LTC6952_MODE_MSK GENMASK(6, 5) > +#define LTC6952_MODE(x) FIELD_PREP(LTC6952_MODE_MSK, x) > +#define LTC6952_OINV_MSK BIT(4) > +#define LTC6952_OINV(x) FIELD_PREP(LTC6952_OINV_MSK, x) > +#define LTC6952_DDEL_HIGH_MSK GENMASK(3, 0) > +#define LTC6952_DDEL_HIGH(x) FIELD_PREP(LTC6952_DDEL_HIGH_MSK, x) > + > +/* LTC6952_REG14,18,22,26,30,34,38,42,46,50,54 */ > +#define LTC6952_DDEL_LOW_MSK GENMASK(7, 0) > +#define LTC6952_DDEL_LOW(x) FIELD_PREP(LTC6952_DDEL_LOW_MSK, x) > + > +/* LTC6952_REG15,19,23,27,31,35,39,43,47,51,55 */ > +#define LTC6952_ADEL_MSK GENMASK(5, 0) > +#define LTC6952_ADEL(x) FIELD_PREP(LTC6952_ADEL_MSK, x) > + > +/* LTC6952_REG56 */ > +#define LTC6952_REV_MSK GENMASK(7, 4) > +#define LTC6952_PART_MSK GENMASK(3, 0) > + > +#define LTC6952_CMD_READ 0x1 > +#define LTC6952_CMD_WRITE 0x0 > +#define LTC6952_CMD_ADDR(x) ((x) << 1) > + > +#define LTC6952_NUM_CHAN 11 > + > +#define LTC6952_N_MAX 65535 N and R could be more descriptive perhaps? Also as these are 2^N - 1, it might be useful to describe them instead in number of bits... > +#define LTC6952_R_MAX 1023 > + > +#define LTC6952_PFD_FREQ_MAX 167000 > + > +#define LTC6952_OUT_DIV_MIN 1 > +#define LTC6952_OUT_DIV_MAX 1048576 > + > +#define LTC6952_CH_OFFSET(ch) (0x04 * (ch)) > + > +struct ltc6952_output { > + unsigned int address; > + struct clk_hw hw; > + struct iio_dev *indio_dev; > +}; > + > +struct ltc6952_chan_spec { > + unsigned int num; > + unsigned int out_divider; > + unsigned int mp; > + unsigned int md; > + unsigned int digital_delay; > + unsigned int analog_delay; > + unsigned int sysref_mode; > + unsigned int power_down_mode; > + const char *extended_name; I moan about extended_name below... > +}; > + > +struct ltc6952_state { > + struct spi_device *spi; > + struct mutex lock; The scope of any lock should always be documented. > + u32 ref_freq; > + u32 vco_freq; > + u32 pfd_freq; > + bool sysref_request; > + unsigned int ref_divider; > + unsigned int vco_divider; > + const char *clk_out_names[LTC6952_NUM_CHAN]; > + unsigned int num_channels; > + struct ltc6952_chan_spec *channels; > + struct iio_chan_spec iio_channels[LTC6952_NUM_CHAN]; > + struct ltc6952_output outputs[LTC6952_NUM_CHAN]; > + struct clk *clks[LTC6952_NUM_CHAN]; > + struct clk_onecell_data clk_data; > +}; > + > +#define to_output(_hw) container_of(_hw, struct ltc6952_output, hw) > + > +static int ltc6952_write(struct iio_dev *indio_dev, > + unsigned int reg, > + unsigned int val) Given val is a u8 in all writes, just make the parameter a u8. That will make it easier for a compiler to spot truncation etc. > +{ > + struct ltc6952_state *st = iio_priv(indio_dev); > + u8 buf[2]; > + u8 cmd; > + > + cmd = LTC6952_CMD_WRITE | LTC6952_CMD_ADDR(reg); > + buf[0] = cmd; > + buf[1] = val; > + > + return spi_write(st->spi, buf, ARRAY_SIZE(buf)); spi_write can (IIRC) do dma with buf direction. Hence buf needs to be DMA safe, which means it needs not to share a cache line with any other data. You can't guarantee that with variables on the stack... Not spi_write_then_read always uses a bounce buffer, so you could just use that with a read length of 0. Otherwise put the buffer inside state and use the normal __cache_line_aligned magic. > +} > + > +static int ltc6952_read(struct iio_dev *indio_dev, > + unsigned int reg, > + unsigned int *val) As above, val as a u8 would be easier to follow I think. > +{ > + struct ltc6952_state *st = iio_priv(indio_dev); > + u8 cmd; > + > + cmd = LTC6952_CMD_READ | LTC6952_CMD_ADDR(reg); > + > + return spi_write_then_read(st->spi, &cmd, 1, val, 1); > +} > + > +static int ltc6952_write_mask(struct iio_dev *indio_dev, > + unsigned int addr, > + unsigned long mask, > + unsigned int val) > +{ > + int readval, ret; > + > + ret = ltc6952_read(indio_dev, addr, &readval); > + if (ret < 0) > + return ret; > + > + readval &= ~mask; > + readval |= val; > + > + return ltc6952_write(indio_dev, addr, readval); > +} > + > +static unsigned int ltc6952_calc_out_div(unsigned long parent_rate, > + unsigned long rate) > +{ > + unsigned int div; > + > + div = DIV_ROUND_CLOSEST(parent_rate, rate); > + > + div = clamp_t(unsigned int, > + div, > + LTC6952_OUT_DIV_MIN, > + LTC6952_OUT_DIV_MAX); > + > + return div; > +} > + > +static int ltc6952_calculate_divider(struct ltc6952_chan_spec *chan) > +{ > + int mp = 0, md = 0; mp always set below, so don't do it here. md also I think... > + unsigned int out_divider; > + > + /* M(x) = (MP(x) + 1)2^MD(x) */ > + chan->md = 0; > + for (mp = 0; mp < 32; mp++) { > + if (chan->out_divider == (mp + 1)) { > + chan->mp = mp; > + chan->md = 0; > + return 0; > + } > + > + /* MD works only if MP is greater than 15 */ > + if (mp <= 15) > + continue; > + for (md = 0; md < 7; md++) { > + out_divider = (mp + 1) << md; > + if (chan->out_divider == out_divider) { > + chan->mp = mp; > + chan->md = md; > + return 0; > + } > + if (chan->out_divider < out_divider) > + break; > + } > + } > + > + return -EINVAL; > +} > + > +static struct attribute *ltc6952_attributes[] = { > + NULL, ? Why add this at all. It's optional. > +}; > + > +static const struct attribute_group ltc6952_attribute_group = { > + .attrs = ltc6952_attributes, > +}; > + > +static int ltc6952_get_phase(struct iio_dev *indio_dev, > + struct ltc6952_chan_spec *ch, > + unsigned int *val) > +{ > + struct ltc6952_state *st = iio_priv(indio_dev); > + unsigned int tmp1, tmp2; > + int ret; > + > + mutex_lock(&st->lock); > + ret = ltc6952_read(indio_dev, LTC6952_REG(0x0D) + > + LTC6952_CH_OFFSET(ch->num), > + &tmp1); > + if (ret < 0) > + goto err_unlock; > + > + ret = ltc6952_read(indio_dev, LTC6952_REG(0x0E) + > + LTC6952_CH_OFFSET(ch->num), > + &tmp2); > + if (ret < 0) > + goto err_unlock; > + > + *val = ((tmp1 & 0x0F) << 8) + (tmp2 & 0xFF); > + > +err_unlock: > + mutex_unlock(&st->lock); > + return ret; > +} > + > +static int ltc6952_read_raw(struct iio_dev *indio_dev, > + struct iio_chan_spec const *chan, > + int *val, > + int *val2, > + long mask) > +{ > + struct ltc6952_state *st = iio_priv(indio_dev); > + struct ltc6952_chan_spec *ch; > + unsigned int tmp, code; > + int ret; > + > + if (chan->address >= LTC6952_NUM_CHAN) > + return -EINVAL; > + > + ch = &st->channels[chan->address]; > + > + switch (mask) { > + case IIO_CHAN_INFO_FREQUENCY: > + *val = st->vco_freq / ch->out_divider; > + return IIO_VAL_INT; > + case IIO_CHAN_INFO_PHASE: > + ret = ltc6952_get_phase(indio_dev, ch, &tmp); > + if (ret < 0) > + return ret; > + > + code = DIV_ROUND_CLOSEST(tmp * 3141592, > + ch->out_divider); > + *val = code / 1000000; > + *val2 = code % 1000000; > + return IIO_VAL_INT_PLUS_MICRO; > + default: > + return -EINVAL; > + } > +}; > + > +static int ltc6952_set_phase(struct iio_dev *indio_dev, > + struct ltc6952_chan_spec *ch, > + unsigned int val) > +{ > + struct ltc6952_state *st = iio_priv(indio_dev); > + int ret; > + > + mutex_lock(&st->lock); > + ret = ltc6952_write_mask(indio_dev, LTC6952_REG(0x0D) + > + LTC6952_CH_OFFSET(ch->num), > + LTC6952_DDEL_HIGH_MSK, > + LTC6952_DDEL_HIGH((val & 0xF00) >> 8)); > + if (ret < 0) > + goto err_unlock; > + > + ret = ltc6952_write(indio_dev, LTC6952_REG(0x0E) + > + LTC6952_CH_OFFSET(ch->num), > + LTC6952_DDEL_LOW(val)); > + if (ret < 0) > + goto err_unlock; > + > + /* Phase Syncronization */ > + ret = ltc6952_write_mask(indio_dev, LTC6952_REG(0x0B), > + LTC6952_SSRQ_MSK, LTC6952_SSRQ(1)); > + if (ret < 0) > + goto err_unlock; > + > + usleep_range(1000, 1001); /* sleep > 1000us */ > + ret = ltc6952_write_mask(indio_dev, LTC6952_REG(0x0B), > + LTC6952_SSRQ_MSK, LTC6952_SSRQ(0)); > +err_unlock: > + mutex_unlock(&st->lock); > + return ret; > +} > + > +static int ltc6952_write_raw(struct iio_dev *indio_dev, > + struct iio_chan_spec const *chan, > + int val, > + int val2, > + long mask) > +{ > + struct ltc6952_state *st = iio_priv(indio_dev); > + struct ltc6952_chan_spec *ch; > + unsigned int code, tmp; > + int ret; > + > + if (chan->address >= st->num_channels) > + return -EINVAL; > + > + ch = &st->channels[chan->address]; > + > + switch (mask) { > + case IIO_CHAN_INFO_FREQUENCY: > + ch->out_divider = ltc6952_calc_out_div(st->vco_freq, val); > + ret = ltc6952_calculate_divider(ch); > + if (ret < 0) > + return ret; > + > + mutex_lock(&st->lock); > + tmp = LTC6952_MP(ch->mp) | LTC6952_MD(ch->md); > + ret = ltc6952_write(indio_dev, LTC6952_REG(0x0C) + > + LTC6952_CH_OFFSET(ch->num), > + tmp); > + mutex_unlock(&st->lock); > + if (ret < 0) > + return ret; > + break; > + case IIO_CHAN_INFO_PHASE: > + code = val * 1000000 + val2 % 1000000; > + tmp = DIV_ROUND_CLOSEST(code * ch->out_divider, 3141592); > + tmp = clamp_t(unsigned int, tmp, 0, 4095); > + ret = ltc6952_set_phase(indio_dev, ch, tmp); > + if (ret < 0) > + return ret; > + > + break; > + default: > + return -EINVAL; > + } > + > + return 0; > +} > + > +static int ltc6952_reg_access(struct iio_dev *indio_dev, unsigned int reg, > + unsigned int writeval, unsigned int *readval) > +{ > + struct ltc6952_state *st = iio_priv(indio_dev); > + int ret; > + > + mutex_lock(&st->lock); > + if (readval) > + ret = ltc6952_read(indio_dev, reg, readval); > + else > + ret = ltc6952_write(indio_dev, reg, writeval); > + mutex_unlock(&st->lock); > + > + return ret; > +} > + > +static const struct iio_info ltc6952_iio_info = { > + .read_raw = <c6952_read_raw, > + .write_raw = <c6952_write_raw, > + .debugfs_reg_access = <c6952_reg_access, > + .attrs = <c6952_attribute_group, > +}; > + > +static long ltc6952_get_clk_attr(struct clk_hw *hw, > + long mask) > +{ > + struct iio_dev *indio_dev = to_output(hw)->indio_dev; > + struct ltc6952_state *st = iio_priv(indio_dev); > + struct iio_chan_spec *chan; > + unsigned int address; > + int val, ret; > + > + address = to_output(hw)->address; > + if (address >= st->num_channels) > + return -EINVAL; > + > + chan = &st->iio_channels[address]; > + > + ret = ltc6952_read_raw(indio_dev, chan, &val, NULL, mask); > + > + if (ret == IIO_VAL_INT) > + return val; > + > + return ret; > +} > + > +static long ltc6952_set_clk_attr(struct clk_hw *hw, > + long mask, > + unsigned long val) > +{ > + struct iio_dev *indio_dev = to_output(hw)->indio_dev; > + struct ltc6952_state *st = iio_priv(indio_dev); > + struct iio_chan_spec *chan; > + unsigned int address; > + > + address = to_output(hw)->address; > + if (address >= st->num_channels) > + return -EINVAL; > + > + chan = &st->iio_channels[address]; > + > + return ltc6952_write_raw(indio_dev, chan, val, 0, mask); > +} > + > +static unsigned long ltc6952_clk_recalc_rate(struct clk_hw *hw, > + unsigned long parent_rate) > +{ > + return ltc6952_get_clk_attr(hw, IIO_CHAN_INFO_FREQUENCY); > +} > + > +static long ltc6952_clk_round_rate(struct clk_hw *hw, > + unsigned long rate, > + unsigned long *parent_rate) > +{ > + struct ltc6952_output *out = to_output(hw); > + struct iio_dev *indio_dev = out->indio_dev; > + struct ltc6952_state *st = iio_priv(indio_dev); > + unsigned int div; > + > + div = ltc6952_calc_out_div(st->vco_freq, rate); > + > + return DIV_ROUND_CLOSEST(st->vco_freq, div); > +} > + > +static int ltc6952_clk_set_rate(struct clk_hw *hw, > + unsigned long rate, > + unsigned long parent_rate) > +{ > + return ltc6952_set_clk_attr(hw, IIO_CHAN_INFO_FREQUENCY, rate); > +} > + > +static const struct clk_ops ltc6952_clk_ops = { > + .recalc_rate = ltc6952_clk_recalc_rate, > + .round_rate = ltc6952_clk_round_rate, > + .set_rate = ltc6952_clk_set_rate, > +}; > + > +static int ltc6952_clk_register(struct iio_dev *indio_dev, > + unsigned int num, > + unsigned int address) > +{ > + struct ltc6952_state *st = iio_priv(indio_dev); > + struct clk_init_data init; > + struct clk *clk; > + > + init.name = st->clk_out_names[num]; > + init.ops = <c6952_clk_ops; > + init.flags = 0; > + init.num_parents = 0; > + > + st->outputs[num].hw.init = &init; > + st->outputs[num].indio_dev = indio_dev; > + st->outputs[num].address = address; > + > + clk = devm_clk_register(&st->spi->dev, &st->outputs[num].hw); > + if (IS_ERR(clk)) > + return PTR_ERR(clk); > + > + st->clks[num] = clk; > + > + return 0; > +} > + > +static int ltc6952_setup(struct iio_dev *indio_dev) > +{ > + struct ltc6952_state *st = iio_priv(indio_dev); > + struct ltc6952_chan_spec *chan; > + unsigned long vco_freq, ref_freq; > + unsigned long pfd_freq; > + unsigned long n, r; > + unsigned int i, tmp; > + int ret; > + > + vco_freq = st->vco_freq / 1000; > + ref_freq = st->ref_freq / 1000; > + > + /* fVCO / N = fREF / R */ > + rational_best_approximation(vco_freq, ref_freq, > + LTC6952_N_MAX, LTC6952_R_MAX, > + &n, &r); > + > + pfd_freq = vco_freq / n; > + while ((pfd_freq > LTC6952_PFD_FREQ_MAX) && > + (n <= LTC6952_N_MAX / 2) && > + (r <= LTC6952_R_MAX / 2)) { > + pfd_freq /= 2; > + n *= 2; > + r *= 2; > + } > + > + mutex_lock(&st->lock); > + /* Resets all registers to default values */ > + ret = ltc6952_write_mask(indio_dev, LTC6952_REG(0x02), > + LTC6952_POR_MSK, LTC6952_POR(1)); > + if (ret < 0) > + goto err_unlock; > + > + /* Program the dividers */ > + ret |= ltc6952_write_mask(indio_dev, LTC6952_REG(0x06), > + LTC6952_RD_HIGH_MSK, > + LTC6952_RD_HIGH((r & 0x300) >> 8)); Stacking up return values like this is almost always a bad idea. You can end up with two different errors being combined and completely obscure the actual return value. Please have a separate error check for each one. > + ret |= ltc6952_write(indio_dev, LTC6952_REG(0x07), LTC6952_RD_LOW(r)); > + ret |= ltc6952_write(indio_dev, LTC6952_REG(0x08), > + LTC6952_ND_HIGH(n >> 8)); > + ret |= ltc6952_write(indio_dev, LTC6952_REG(0x09), LTC6952_ND_LOW(n)); These seemed a bit odd as you are mapping bytes to bytes via a macro.. > + if (ret < 0) > + goto err_unlock; > + > + /* PLL lock cycle count */ > + ret = ltc6952_write_mask(indio_dev, LTC6952_REG(0x06), LTC6952_LKCT_MSK, > + LTC6952_LKCT(0x03)); > + if (ret < 0) > + goto err_unlock; > + > + /* Disable CP Hi-Z */ > + ret = ltc6952_write_mask(indio_dev, LTC6952_REG(0x0A), LTC6952_CPRST_MSK, > + LTC6952_CPRST(0x0)); > + if (ret < 0) > + goto err_unlock; > + > + /* Program the output channels */ > + for (i = 0; i < st->num_channels; i++) { > + chan = &st->channels[i]; > + > + if (chan->num >= LTC6952_NUM_CHAN) > + continue; > + > + /* Enable channel */ > + tmp = LTC6952_PD(i, 0); > + ret |= ltc6952_write_mask(indio_dev, > + LTC6952_REG(0x03) + (chan->num >> 2), > + LTC6952_PD_MSK(chan->num), tmp); > + > + ret |= ltc6952_calculate_divider(chan); > + ret |= ltc6952_write(indio_dev, LTC6952_REG(0x0c) + > + LTC6952_CH_OFFSET(chan->num), > + LTC6952_MP(chan->mp) | LTC6952_MD(chan->md)); > + > + /* Enable sync or sysref */ > + ret |= ltc6952_write_mask(indio_dev, LTC6952_REG(0x0D) + > + LTC6952_CH_OFFSET(chan->num), > + LTC6952_SRQEN_MSK, > + LTC6952_SRQEN(1)); > + > + /* Set channel delay */ > + ret |= ltc6952_write_mask(indio_dev, LTC6952_REG(0x0D) + > + LTC6952_CH_OFFSET(chan->num), > + LTC6952_DDEL_HIGH_MSK, > + LTC6952_DDEL_HIGH((chan->digital_delay & > + 0xF00) >> 8)); > + ret |= ltc6952_write(indio_dev, LTC6952_REG(0x0E) + > + LTC6952_CH_OFFSET(chan->num), > + LTC6952_DDEL_LOW(chan->digital_delay)); > + ret |= ltc6952_write(indio_dev, LTC6952_REG(0x0F) + > + LTC6952_CH_OFFSET(chan->num), > + LTC6952_ADEL(chan->analog_delay)); > + if (ret < 0) > + goto err_unlock; > + > + st->iio_channels[i].type = IIO_ALTVOLTAGE; Hmm. Is this an appropriate channel type for a clock? I'm not convinced. For most of the frequency drivers there is the option of a sinusoid. Here I don't think there is? > + st->iio_channels[i].output = 1; > + st->iio_channels[i].indexed = 1; > + st->iio_channels[i].channel = chan->num; > + st->iio_channels[i].address = i; > + st->iio_channels[i].extend_name = chan->extended_name; See below. Definitely don't do this. extend_name defines part of the ABI it cannot come from DT as then the ABI would not be defined. > + st->iio_channels[i].info_mask_separate = > + BIT(IIO_CHAN_INFO_FREQUENCY) | > + BIT(IIO_CHAN_INFO_PHASE); > + } > + > + /* Phase Syncronization */ > + ret |= ltc6952_write_mask(indio_dev, LTC6952_REG(0x0B), > + LTC6952_SSRQ_MSK, LTC6952_SSRQ(1)); > + usleep_range(1000, 1001); /* sleep > 1000us */ > + ret |= ltc6952_write_mask(indio_dev, LTC6952_REG(0x0B), > + LTC6952_SSRQ_MSK, LTC6952_SSRQ(0)); > + if (ret < 0) > + goto err_unlock; > + > + mutex_unlock(&st->lock); > + There are several different activities in this function. It would be worth thinking about breaking it up. This clock related section for example could be moved out of here. > + /* Configure clocks */ > + for (i = 0; i < st->num_channels; i++) { > + chan = &st->channels[i]; > + > + if (chan->num >= LTC6952_NUM_CHAN) > + continue; > + > + ret = ltc6952_clk_register(indio_dev, chan->num, i); > + if (ret) > + return ret; > + } > + > + st->clk_data.clks = st->clks; > + st->clk_data.clk_num = LTC6952_NUM_CHAN; > + > + return of_clk_add_provider(st->spi->dev.of_node, > + of_clk_src_onecell_get, > + &st->clk_data); > +err_unlock: > + mutex_unlock(&st->lock); > + return ret; > +} > + > +static int ltc6952_parse_dt(struct device *dev, > + struct ltc6952_state *st) > +{ > + struct device_node *np = dev->of_node, *chan_np; > + unsigned int cnt = 0; > + int ret; > + > + ret = of_property_read_u32(np, "adi,ref-frequency-hz", > + &st->ref_freq); > + if (ret < 0) > + return ret; This handling doesn't make sense for an optional property and you don't have this listed as required (I think) in the dt binding. > + > + ret = of_property_read_u32(np, "adi,vco-frequency-hz", > + &st->vco_freq); > + if (ret < 0) > + return ret; > + > + ret = of_property_read_string_array(np, "clock-output-names", > + st->clk_out_names, ARRAY_SIZE(st->clk_out_names)); > + if (ret < 0) > + return ret; > + > + st->num_channels = of_get_available_child_count(np); > + if (st->num_channels > LTC6952_NUM_CHAN) > + return -EINVAL; > + > + st->channels = devm_kzalloc(dev, > + sizeof(struct ltc6952_chan_spec) * st->num_channels, > + GFP_KERNEL); > + if (!st->channels) > + return -ENOMEM; > + > + for_each_child_of_node(np, chan_np) { > + st->channels[cnt].num = cnt; > + of_property_read_u32(chan_np, "reg", > + &st->channels[cnt].num); > + > + if (of_property_read_u32(chan_np, "adi,divider", > + &st->channels[cnt].out_divider)) > + st->channels[cnt].out_divider = 4; > + > + of_property_read_u32(chan_np, "adi,digital-delay", > + &st->channels[cnt].digital_delay); > + > + of_property_read_u32(chan_np, "adi,analog-delay", > + &st->channels[cnt].analog_delay); > + > + of_property_read_string(chan_np, "adi,extended-name", > + &st->channels[cnt].extended_name); Extended names form part of the abi. As such they can never come from devicetree. If you want to do this, we have talked about allowing labels to be assigned to channels that appear as a separate sysfs file. That route is flexible and provides the same information whilst allowing code that isn't aware of it to work. > + > + cnt++; > + } > + > + return 0; > +} > + > +static int ltc6952_status_show(struct seq_file *file, void *offset) > +{ > + struct iio_dev *indio_dev = spi_get_drvdata(file->private); > + int ret; > + u32 status; > + > + ret = ltc6952_read(indio_dev, 0x00, &status); > + if (ret < 0) > + return ret; > + > + seq_printf(file, > + "SYSREF Status:\t%s\nVCO Status:\t%s\nLock Status:\t%s\n", > + status & LTC6952_REFOK_MSK ? > + "Valid" : "Invalid", > + status & LTC6952_VCOOK_MSK ? > + "Valid" : "Invalid", > + status & LTC6952_LOCK_MSK ? > + "PLL Locked" : "Unlocked"); > + > + return 0; > +} > + > +static int ltc6952_probe(struct spi_device *spi) > +{ > + struct iio_dev *indio_dev; > + struct ltc6952_state *st; > + int ret; > + > + indio_dev = devm_iio_device_alloc(&spi->dev, sizeof(*st)); > + if (!indio_dev) > + return -ENOMEM; > + > + st = iio_priv(indio_dev); > + > + mutex_init(&st->lock); > + > + spi_set_drvdata(spi, indio_dev); > + > + st->spi = spi; > + > + ret = ltc6952_parse_dt(&spi->dev, st); > + if (ret < 0) > + return ret; > + > + indio_dev->dev.parent = &spi->dev; > + indio_dev->name = spi_get_device_id(spi)->name; > + indio_dev->info = <c6952_iio_info; > + indio_dev->modes = INDIO_DIRECT_MODE; > + indio_dev->channels = st->iio_channels; > + indio_dev->num_channels = st->num_channels; > + > + ret = ltc6952_setup(indio_dev); > + if (ret) > + return ret; > + > + ret = iio_device_register(indio_dev); > + if (ret < 0) { > + of_clk_del_provider(spi->dev.of_node); > + return ret; > + } > + > + if (iio_get_debugfs_dentry(indio_dev)) { > + struct dentry *stats; > + > + stats = debugfs_create_devm_seqfile(&spi->dev, "status", > + iio_get_debugfs_dentry(indio_dev), > + ltc6952_status_show); > + if (PTR_ERR_OR_ZERO(stats)) > + dev_err(&spi->dev, > + "Failed to create debugfs entry"); > + } > + > + return ret; > +} > + > +static int ltc6952_remove(struct spi_device *spi) > +{ > + struct iio_dev *indio_dev = spi_get_drvdata(spi); > + > + iio_device_unregister(indio_dev); > + > + of_clk_del_provider(spi->dev.of_node); Could we have a function in here to match against _setup, so that its obvious that this call onwinds stuff from in there? > + > + return 0; > +} > + > +static const struct spi_device_id ltc6952_id[] = { > + {"ltc6952", 0}, > + {} > +}; > +MODULE_DEVICE_TABLE(spi, ltc6952_id); > + > +static const struct of_device_id ltc6952_of_match[] = { > + { .compatible = "adi,ltc6952" }, > + { }, > +}; > +MODULE_DEVICE_TABLE(of, ltc6952_of_match); > + > +static struct spi_driver ltc6952_driver = { > + .driver = { > + .name = "ltc6952", > + .of_match_table = ltc6952_of_match, > + }, > + .probe = ltc6952_probe, > + .remove = ltc6952_remove, > + .id_table = ltc6952_id, > +}; > +module_spi_driver(ltc6952_driver); > + > +MODULE_AUTHOR("Mircea Caprioru "); > +MODULE_DESCRIPTION("Analog Devices LTC6952 driver"); > +MODULE_LICENSE("GPL v2");