]> git.kernelconcepts.de Git - karo-tx-linux.git/log
karo-tx-linux.git
9 years agoclk: check ->determine/round_rate() return value in clk_calc_new_rates
Boris Brezillon [Sun, 29 Mar 2015 01:48:48 +0000 (03:48 +0200)]
clk: check ->determine/round_rate() return value in clk_calc_new_rates

->determine_rate() and ->round_rate() can return the closest rate to the
requested one or an error code.
clk_calc_new_rates is assuming these functions can't return a negative
value, which leads to a undefined behavior when the clk implementation
returns such an error code.
Fix this by returning NULL in case ->determine_rate() or ->round_rate()
returned an error code.

Signed-off-by: Boris Brezillon <boris.brezillon@free-electrons.com>
Signed-off-by: Michael Turquette <mturquette@linaro.org>
9 years agoclk: at91: usb: propagate rate modification to the parent clk
Boris Brezillon [Sun, 29 Mar 2015 01:45:33 +0000 (03:45 +0200)]
clk: at91: usb: propagate rate modification to the parent clk

The at91sam9n12 and at91sam9x5 usb clocks do not propagate rate
modification requests to their parents.
This causes a bug when the PLLB is left uninitialized by the bootloader
(PLL multiplier set to 0, or in other words, PLL rate = 0 Hz).

Implement the determinate_rate method and propagate the change rate
request to the parent clk.

Cc: <stable@vger.kernel.org> # v3.14+
Signed-off-by: Boris Brezillon <boris.brezillon@free-electrons.com>
Reported-by: Bo Shen <voice.shen@atmel.com>
Tested-by: Bo Shen <voice.shen@atmel.com>
Signed-off-by: Michael Turquette <mturquette@linaro.org>
9 years agoclk: samsung: exynos4: Disable ARMCLK down feature on Exynos4210 SoC
Bartlomiej Zolnierkiewicz [Fri, 27 Mar 2015 16:27:10 +0000 (17:27 +0100)]
clk: samsung: exynos4: Disable ARMCLK down feature on Exynos4210 SoC

Commit 42773b28e71d ("clk: samsung: exynos4: Enable ARMCLK
down feature") enabled ARMCLK down feature on all Exynos4
SoCs.  Unfortunately on Exynos4210 SoC ARMCLK down feature
causes a lockup when ondemand cpufreq governor is used.
Fix it by limiting ARMCLK down feature to Exynos4x12 SoCs.

This patch was tested on:
- Exynos4210 SoC based Trats board
- Exynos4210 SoC based Origen board
- Exynos4412 SoC based Trats2 board
- Exynos4412 SoC based Odroid-U3 board

Cc: Daniel Drake <drake@endlessm.com>
Cc: Tomasz Figa <t.figa@samsung.com>
Cc: Kukjin Kim <kgene@kernel.org>
Fixes: 42773b28e71d ("clk: samsung: exynos4: Enable ARMCLK down feature")
Cc: <stable@vger.kernel.org> # v3.17+
Reviewed-by: Krzysztof Kozlowski <k.kozlowski@samsung.com>
Signed-off-by: Bartlomiej Zolnierkiewicz <b.zolnierkie@samsung.com>
Signed-off-by: Michael Turquette <mturquette@linaro.org>
9 years agoclk: don't use __initconst for non-const arrays
Uwe Kleine-König [Wed, 18 Feb 2015 09:59:45 +0000 (10:59 +0100)]
clk: don't use __initconst for non-const arrays

The statement

static const char *name[];

defines a modifiable array of pointers to constant chars. That is

*name[0] = 'f';

is forbidden, but

name[0] = "f";

is not. So marking an array that is defined as above with __initconst is
wrong. Either an additional const must be added such that the whole
definition reads:

static const char *const name[] __initconst;

or where this is not possible __initdata must be used.

Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Signed-off-by: Michael Turquette <mturquette@linaro.org>
9 years agoclk: at91: change to using endian agnositc IO
Ben Dooks [Thu, 26 Mar 2015 13:07:29 +0000 (13:07 +0000)]
clk: at91: change to using endian agnositc IO

Change to using endian agnostic _relaxed IO accessors instead of __raw

Signed-off-by: Ben Dooks <ben.dooks@codethink.co.uk>
--
CC: Andrew Victor <linux@maxim.org.za>
CC: Nicolas Ferre <nicolas.ferre@atmel.com>
CC: Jean-Christophe Plagniol-Villard <plagnioj@jcrosoft.com>
CC: Mike Turquette <mturquette@linaro.org> (maintainer:COMMON CLK FRAMEWORK)
CC: Stephen Boyd <sboyd@codeaurora.org> (maintainer:COMMON CLK FRAMEWORK)
CC: linux-kernel@vger.kernel.org (open list:COMMON CLK FRAMEWORK)
Signed-off-by: Michael Turquette <mturquette@linaro.org>
9 years agoMerge tag 'clk/for-4.1' of git://git.kernel.org/pub/scm/linux/kernel/git/tegra/linux...
Michael Turquette [Sat, 11 Apr 2015 01:06:55 +0000 (18:06 -0700)]
Merge tag 'clk/for-4.1' of git://git.kernel.org/pub/scm/linux/kernel/git/tegra/linux into clk-next

clk/tegra: Changes for v4.1-rc1

These are mostly cleanups that I've been carrying in my local tree for
far too long. In addition to those, there are some preparatory patches
for the upcoming Tegra210 support and a patch to enable clocks needed
for HDMI audio support.

9 years agoclk: clk-gpio-gate: Fix active low
Martin Fuzzey [Wed, 18 Mar 2015 13:53:17 +0000 (14:53 +0100)]
clk: clk-gpio-gate: Fix active low

The active low flag in the DT cell is currently ignored.

This occurs because of_get_named_gpio_flags() does not apply the flags
to the underlying struct gpio_desc so the test in clk_register_gpio_gate()
was bogus.

Note that this patch changes the internal kernel API for
clk_register_gpio_gate() but there are currently no other users.

Signed-off-by: Martin Fuzzey <mfuzzey@parkeon.com>
Acked-by: Jyri Sarha <jsarha@ti.com>
Signed-off-by: Michael Turquette <mturquette@linaro.org>
9 years agoclk: Add PWM clock driver
Philipp Zabel [Fri, 13 Feb 2015 19:18:52 +0000 (20:18 +0100)]
clk: Add PWM clock driver

Some board designers, when running out of clock output pads, decide to
(mis)use PWM output pads to provide a clock to external components.
This driver supports this practice by providing an adapter between the
PWM and clock bindings in the device tree. As the PWM bindings specify
the period in the device tree, this is a fixed clock.

Tested-by: Janusz Uzycki <j.uzycki@elproma.com.pl>
Signed-off-by: Philipp Zabel <p.zabel@pengutronix.de>
Signed-off-by: Michael Turquette <mturquette@linaro.org>
9 years agoclk: Add clock driver for mb86s7x
Jassi Brar [Wed, 4 Mar 2015 11:04:03 +0000 (19:04 +0800)]
clk: Add clock driver for mb86s7x

 The CRG11 clock controller is managed by remote f/w.
This driver simply maps Linux CLK ops onto mailbox api.

Signed-off-by: Andy Green <andy.green@linaro.org>
Signed-off-by: Vincent Yang <vincent.yang@socionext.com>
Signed-off-by: Tetsuya Nuriya <nuriya.tetsuya@socionext.com>
Signed-off-by: Michael Turquette <mturquette@linaro.org>
9 years agoclk: pxa: pxa3xx: add missing os timer clock
Robert Jarzmik [Sat, 14 Feb 2015 14:54:58 +0000 (15:54 +0100)]
clk: pxa: pxa3xx: add missing os timer clock

The pxa3xx scheduler relies on the pxa-timer, which requires a clock for
its rate. As the clock handling will be taken over by the clock
framework, add this missing clock.

The miss was discovered by attempting to run a zylonite platform in a
device-tree configuration, with the future patch to shift clocks
handling to clock framework applied.

Signed-off-by: Robert Jarzmik <robert.jarzmik@free.fr>
Signed-off-by: Michael Turquette <mturquette@linaro.org>
9 years agoclk: tegra: Use the proper parent for plld_dsi
Thierry Reding [Thu, 26 Mar 2015 16:53:01 +0000 (17:53 +0100)]
clk: tegra: Use the proper parent for plld_dsi

The current parent, plld_out0, does not exist. The proper name is
pll_d_out0. While at it, rename the plld_dsi clock to pll_d_dsi_out to
be more consistent with other clock names.

Fixes: b270491eb9a0 ("clk: tegra: Define PLLD_DSI and remove dsia(b)_mux")
Signed-off-by: Thierry Reding <treding@nvidia.com>
9 years agoclk: tegra: Use generic tegra_osc_clk_init() on Tegra114
Thierry Reding [Thu, 26 Mar 2015 16:50:06 +0000 (17:50 +0100)]
clk: tegra: Use generic tegra_osc_clk_init() on Tegra114

There is no reason why Tegra114 cannot use the same generic code to set
up the oscillator, clk_m and pll_ref clocks. The only effective change
that this causes is that the CLK_SET_PARENT_RATE flag is dropped, but
since these clocks are all fixed it is not needed anyway.

Signed-off-by: Thierry Reding <treding@nvidia.com>
9 years agoclk: tegra: Model oscillator as clock
Thierry Reding [Thu, 26 Mar 2015 16:43:56 +0000 (17:43 +0100)]
clk: tegra: Model oscillator as clock

Currently the Tegra clock driver simplifies the clock tree somewhat by
taking advantage of the fact that clk_m runs at the same frequency as
the oscillator. While that's true on all currently supported SoCs, it
does not apply to Tegra210 anymore. On Tegra210 clk_m is typically
divided down from the oscillator frequency. To support that setup, add
a separate clock for the oscillator that both clk_m and pll_ref derive
from.

Modify the tegra_osc_clk_init() function to take an additional divider
parameter for clk_m. Existing SoCs always pass in 1, whereas Tegra210
will read the divider from a register in the clock & reset controller.

Signed-off-by: Thierry Reding <treding@nvidia.com>
9 years agoclk: tegra: Add peripheral registers for bank Y
Thierry Reding [Mon, 23 Mar 2015 09:52:45 +0000 (10:52 +0100)]
clk: tegra: Add peripheral registers for bank Y

Tegra210 has an extra bank of peripheral clock registers. Add it to the
generic peripheral clock code.

Cc: Peter De Schrijver <pdeschrijver@nvidia.com>
Cc: Prashant Gaikwad <pgaikwad@nvidia.com>
Reviewed-by: Paul Walmsley <paul@pwsan.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
9 years agoclk: tegra: Register the proper number of resets
Thierry Reding [Mon, 23 Mar 2015 09:57:46 +0000 (10:57 +0100)]
clk: tegra: Register the proper number of resets

The number of resets controls is 32 times the number of peripheral
register banks rather than 32 times the number of clocks. This reduces
(drastically) the number of reset controls registered from 10080 (315
clocks * 32) to 224 (6 peripheral register banks * 32).

This also fixes a potential crash because trying to use any of the
excess reset controls (224-10079) would have caused accesses beyond
the array bounds of the peripheral register banks definition array.

Cc: Peter De Schrijver <pdeschrijver@nvidia.com>
Cc: Prashant Gaikwad <pgaikwad@nvidia.com>
Fixes: 6d5b988e7dc5 ("clk: tegra: implement a reset driver")
Cc: stable@vger.kernel.org # 3.14+
Signed-off-by: Thierry Reding <treding@nvidia.com>
9 years agoclk: tegra: Remove needless initializations
Thierry Reding [Wed, 18 Feb 2015 15:25:16 +0000 (16:25 +0100)]
clk: tegra: Remove needless initializations

The ret variable is often explicitly initialized to 0, but there is no
need to do so in many cases because it will immediately be overwritten
with the return value from a function.

Signed-off-by: Thierry Reding <treding@nvidia.com>
9 years agoclk: tegra: Use consistent indentation
Thierry Reding [Fri, 23 Jan 2015 08:42:33 +0000 (09:42 +0100)]
clk: tegra: Use consistent indentation

Some of the .dev_id entries in the devclks table were oddly indented.
Make them consistent with the rest of the table.

Reviewed-by: Paul Walmsley <paul@pwsan.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
9 years agoclk: tegra: Various whitespace cleanups
Thierry Reding [Fri, 1 Aug 2014 08:44:20 +0000 (10:44 +0200)]
clk: tegra: Various whitespace cleanups

Make usage of blank lines as separators more consistent.

Signed-off-by: Thierry Reding <treding@nvidia.com>
9 years agoclk: tegra: Enable HDA to HDMI clocks on Tegra124
Dylan Reid [Tue, 20 May 2014 02:17:23 +0000 (19:17 -0700)]
clk: tegra: Enable HDA to HDMI clocks on Tegra124

Add the clocks used for HDMI audio played through the HDA controller.
Initialize the codec clock to 48Mhz and the HDA clock to 102MHz per
the TRM.

Signed-off-by: Dylan Reid <dgreid@chromium.org>
Acked-by: Stephen Warren <swarren@nvidia.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
9 years agoclk: tegra: Fix a bunch of sparse warnings
Thierry Reding [Mon, 21 Jul 2014 11:16:36 +0000 (13:16 +0200)]
clk: tegra: Fix a bunch of sparse warnings

The second to last parameter of the TEGRA_CLK_PERIPH macro denotes a
table and should therefore users should pass in NULL instead of 0.

Fixes a bunch of sparse warnings like this:

warning: Using plain integer as NULL pointer

Reviewed-by: Paul Walmsley <paul@pwsan.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
9 years agoclk: tegra: Fix typo tabel -> table
Thierry Reding [Tue, 5 Aug 2014 11:26:12 +0000 (13:26 +0200)]
clk: tegra: Fix typo tabel -> table

The clock initialization structure is named struct clk_init_table.
Update the kerneldoc comment to use the correct name.

Reviewed-by: Paul Walmsley <paul@pwsan.com>
Signed-off-by: Thierry Reding <treding@nvidia.com>
9 years agoclk: versatile: test returned value
Julia Lawall [Sat, 4 Apr 2015 14:59:29 +0000 (16:59 +0200)]
clk: versatile: test returned value

Put NULL test on the result of the previous call instead on one of its
arguments.  A simplified version of the semantic match that finds this
problem is as follows (http://coccinelle.lip6.fr/):

// <smpl>
r@
expression *e1;
expression *e2;
identifier f;
statement S1,S2;
@@

e1 = f(...,e2,...);
(
if (e1 == NULL || ...) S1 else S2
|
*if (e2 == NULL || ...) S1 else S2
)
// </smpl>

Signed-off-by: Julia Lawall <Julia.Lawall@lip6.fr>
Acked-by: Linus Walleij <linus.walleij@linaro.org>
Signed-off-by: Stephen Boyd <sboyd@codeaurora.org>
9 years agoclk: qcom: Fix parent_map translations
Georgi Djakov [Tue, 7 Apr 2015 14:14:51 +0000 (17:14 +0300)]
clk: qcom: Fix parent_map translations

When we introduced the parent_map tables, we missed to update
some of the functions where mapping is translated. Fix this.

Signed-off-by: Georgi Djakov <georgi.djakov@linaro.org>
Tested-by: Nicolas Dechesne <nicolas.dechesne@linaro.org>
Signed-off-by: Stephen Boyd <sboyd@codeaurora.org>
9 years agoclk: qcom: fix driver dependencies
Bartlomiej Zolnierkiewicz [Tue, 7 Apr 2015 12:12:04 +0000 (14:12 +0200)]
clk: qcom: fix driver dependencies

Support for Qualcomm's clock controllers should be available only
on Qualcomm platforms.

Signed-off-by: Bartlomiej Zolnierkiewicz <b.zolnierkie@samsung.com>
Acked-by: Kyungmin Park <kyungmin.park@samsung.com>
Cc: Mike Turquette <mturquette@linaro.org>
Signed-off-by: Stephen Boyd <sboyd@codeaurora.org>
9 years agoclk: qcom: Add EBI2 clocks for IPQ806x
Archit Taneja [Thu, 2 Apr 2015 05:50:41 +0000 (11:20 +0530)]
clk: qcom: Add EBI2 clocks for IPQ806x

The NAND controller within EBI2 requires EBI2_CLK and
EBI2_ALWAYS_ON_CLK clocks.  Create structs for these clocks so
that they can be used by the NAND controller driver. Add an entry
for EBI2_AON_CLK in the gcc-ipq806x DT binding document.

Signed-off-by: Archit Taneja <architt@codeaurora.org>
Signed-off-by: Stephen Boyd <sboyd@codeaurora.org>
9 years agoclk: constify of_device_id array
Fabian Frederick [Tue, 31 Mar 2015 18:50:42 +0000 (20:50 +0200)]
clk: constify of_device_id array

of_device_id is always used as const.
(See driver.of_match_table and open firmware functions)

__initdata updated to __initconst for
static const struct of_device_id ti_clkdm_match_table[]

Signed-off-by: Fabian Frederick <fabf@skynet.be>
Signed-off-by: Stephen Boyd <sboyd@codeaurora.org>
9 years agoMAINTAINERS: clk: discuss on the linux-clk@vger.kernel.org list
Paul Walmsley [Wed, 25 Mar 2015 18:32:59 +0000 (18:32 +0000)]
MAINTAINERS: clk: discuss on the linux-clk@vger.kernel.org list

Most Linux clock framework discussions take place on the
linux-kernel@vger.kernel.org or linux-arm-kernel@lists.infradead.org
mailing lists.  The volume of unrelated messages on these lists makes
it difficult for non-maintainers to follow along with discussions.
Switch the discussion list for clock framework discussions to
linux-clk@vger.kernel.org.  Also, add linux-clk@vger.kernel.org as a
mailing list for clock API discussions.

Signed-off-by: Paul Walmsley <paul@pwsan.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Sylwester Nawrocki <s.nawrocki@samsung.com>
Cc: Tomasz Figa <tomasz.figa@gmail.com>
Cc: Viresh Kumar <viresh.linux@gmail.com>
Cc: Peter De Schrijver <pdeschrijver@nvidia.com>
Cc: Prashant Gaikwad <pgaikwad@nvidia.com>
Cc: Tero Kristo <t-kristo@ti.com>
Cc: Russell King <linux@arm.linux.org.uk>
Acked-by: Stephen Boyd <sboyd@codeaurora.org>
Signed-off-by: Michael Turquette <mturquette@linaro.org>
9 years agoclk: samsung: exynos3250: Add driver for CMU_ISP clock domain
Tomasz Figa [Mon, 2 Mar 2015 13:17:29 +0000 (14:17 +0100)]
clk: samsung: exynos3250: Add driver for CMU_ISP clock domain

Add clock controller for CMU ISP clock domain on Exynos3250,
providing clocks for FIMC-IS subsystem.

[b.michalska: use samsung_cmu_register_one to register
 the provider; updated DT binding documentation]

Signed-off-by: Tomasz Figa <t.figa@samsung.com>
Signed-off-by: Chanwoo Choi <cw00.choi@samsung.com>
Signed-off-by: Beata Michalska <b.michalska@samsung.com>
Acked-by: Kyungmin Park <kyungmin.park@samsung.com>
[s.nawrocki: added __init attribute which was missing in function
 exynos3250_cmu_platform_init() in function, which has been]
Reported-by: kbuild test robot <fengguang.wu@intel.com>
Signed-off-by: Sylwester Nawrocki <s.nawrocki@samsung.com>
Signed-off-by: Michael Turquette <mturquette@linaro.org>
9 years agoclk: samsung: Add CLKOUT driver support for Exynos3250 SoC
Inha Song [Mon, 2 Mar 2015 02:15:00 +0000 (11:15 +0900)]
clk: samsung: Add CLKOUT driver support for Exynos3250 SoC

This patch add CLKOUT driver support for Exynos3250 SoC.

Exynos3250 SoC PMU_DEBUG is the same with Exynos4's PMU_DEBUG
including CLKOUT mux. So we can use the exynos4's clkout init
function for Exynos3250 without the need to add new function.

Signed-off-by: Inha Song <ideal.song@samsung.com>
Acked-by: Stephen Boyd <sboyd@codeaurora.org>
Signed-off-by: Sylwester Nawrocki <s.nawrocki@samsung.com>
Signed-off-by: Michael Turquette <mturquette@linaro.org>
9 years agoclk: cdce706: Constify struct regmap_config
Krzysztof Kozlowski [Fri, 20 Mar 2015 11:34:12 +0000 (12:34 +0100)]
clk: cdce706: Constify struct regmap_config

The regmap_config struct may be const because it is not modified by the
driver and regmap_init() accepts pointer to const.

Signed-off-by: Krzysztof Kozlowski <k.kozlowski@samsung.com>
Acked-by: Max Filippov <jcmvbkbc@gmail.com>
Signed-off-by: Stephen Boyd <sboyd@codeaurora.org>
9 years agoclk: si570: Constify struct regmap_config
Krzysztof Kozlowski [Fri, 20 Mar 2015 11:34:11 +0000 (12:34 +0100)]
clk: si570: Constify struct regmap_config

The regmap_config struct may be const because it is not modified by the
driver and regmap_init() accepts pointer to const.

Signed-off-by: Krzysztof Kozlowski <k.kozlowski@samsung.com>
Reviewed-by: Max Filippov <jcmvbkbc@gmail.com>
Signed-off-by: Stephen Boyd <sboyd@codeaurora.org>
9 years agoclk: si5351: Constify clock names and struct regmap_config
Krzysztof Kozlowski [Fri, 20 Mar 2015 11:34:10 +0000 (12:34 +0100)]
clk: si5351: Constify clock names and struct regmap_config

The regmap_config struct may be const because it is not modified by the
driver and regmap_init() accepts pointer to const.

Replace doubled const in the arrays of clock names with proper const
pointer to const data. This fixes the warnings:

drivers/clk/clk-si5351.c:71:25: warning: duplicate const
drivers/clk/clk-si5351.c:74:25: warning: duplicate const
drivers/clk/clk-si5351.c:77:25: warning: duplicate const
drivers/clk/clk-si5351.c:80:25: warning: duplicate const

Signed-off-by: Krzysztof Kozlowski <k.kozlowski@samsung.com>
Reviewed-by: Max Filippov <jcmvbkbc@gmail.com>
Signed-off-by: Stephen Boyd <sboyd@codeaurora.org>
9 years agoclk: qcom: fix simple_return.cocci warnings
Fengguang Wu [Fri, 28 Nov 2014 18:01:38 +0000 (02:01 +0800)]
clk: qcom: fix simple_return.cocci warnings

drivers/clk/qcom/clk-pll.c:74:1-4: WARNING: end returns can be simpified

 Simplify a trivial if-return sequence.  Possibly combine with a
 preceding function call.
Generated by: scripts/coccinelle/misc/simple_return.cocci

Signed-off-by: Fengguang Wu <fengguang.wu@intel.com>
Signed-off-by: Stephen Boyd <sboyd@codeaurora.org>
9 years agoclk: qcom: Fix ipq806x LCC frequency tables
Stephen Boyd [Fri, 27 Feb 2015 03:34:35 +0000 (19:34 -0800)]
clk: qcom: Fix ipq806x LCC frequency tables

These frequency tables list the wrong rates. Either they don't
have the correct frequency at all, or they're specified in kHz
instead of Hz. Fix it.

Fixes: c99e515a92e9 "clk: qcom: Add IPQ806X LPASS clock controller (LCC) driver"
Tested-by: Kenneth Westfield <kwestfie@codeaurora.org>
Signed-off-by: Stephen Boyd <sboyd@codeaurora.org>
9 years agoMerge tag 'sunxi-clocks-for-4.1' of https://git.kernel.org/pub/scm/linux/kernel/git...
Michael Turquette [Wed, 25 Mar 2015 21:51:40 +0000 (14:51 -0700)]
Merge tag 'sunxi-clocks-for-4.1' of https://git.kernel.org/pub/scm/linux/kernel/git/mripard/linux into clk-next

Allwinner clocks changes for 4.1

The usual round of clock changes for the Allwinner SoCs.

There is nothing really standing out here, but a few changes and fixes, most
notably to allow the AHB clock to be parented to a PLL, instead of the CPU
clock to avoid any AHB rate change due to cpufreq.

9 years agoclk: sunxi: Add pll6 / 4 clock output to sun4i-a10-pll6
Chen-Yu Tsai [Tue, 24 Mar 2015 17:22:08 +0000 (01:22 +0800)]
clk: sunxi: Add pll6 / 4 clock output to sun4i-a10-pll6

The pll6 has a /4 output that is used as an input to the ahb mux clock.

Signed-off-by: Chen-Yu Tsai <wens@csie.org>
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
9 years agoclk: sunxi: Make divs clocks specify which output is the base factor clock
Chen-Yu Tsai [Tue, 24 Mar 2015 17:22:07 +0000 (01:22 +0800)]
clk: sunxi: Make divs clocks specify which output is the base factor clock

The current sunxi clock driver has the base factor clock of divs clocks
as the last clock output of the clock node. This makes it rather difficult
to add new outputs, such as fixed dividers, which were previously unknown.

This patch makes the divs clocks data structure specify which output is
the factor clock, and updates all current divs clocks accordingly.

We can then add new outputs after the factor clocks, at least not breaking
backward compatibility with regards to the devicetree bindings.

Also replace kzalloc with kcalloc in sunxi_divs_clk_setup().

Signed-off-by: Chen-Yu Tsai <wens@csie.org>
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
9 years agoMerge branch 'for-4.1-clk-ti' of github.com:t-kristo/linux-pm into clk-next
Michael Turquette [Tue, 24 Mar 2015 23:33:33 +0000 (16:33 -0700)]
Merge branch 'for-4.1-clk-ti' of github.com:t-kristo/linux-pm into clk-next

9 years agoclk: ti: Implement FAPLL set_rate for the PLL
Tony Lindgren [Sun, 22 Mar 2015 22:35:26 +0000 (15:35 -0700)]
clk: ti: Implement FAPLL set_rate for the PLL

Since we have a fractional divider for the synthesizer, just implement
a simple multiply logic for the PLL.

It seems the PLL divider needs to have also the multiplier set for the PLL
to lock. At least I have not yet figured out if divided rates are doable.

So let's just ignore the PLL divider for now as the synthesizer has both
integer and fractional dividers so we don't even need to use the PLL
divider for the rates we know work with PLL locking.

Cc: Brian Hutchinson <b.hutchman@gmail.com>
Cc: Matthijs van Duin <matthijsvanduin@gmail.com>
Cc: Tero Kristo <t-kristo@ti.com>
Signed-off-by: Tony Lindgren <tony@atomide.com>
Signed-off-by: Tero Kristo <t-kristo@ti.com>
9 years agoclk: ti: Implement FAPLL set_rate for the synthesizer
Tony Lindgren [Tue, 17 Mar 2015 01:04:20 +0000 (18:04 -0700)]
clk: ti: Implement FAPLL set_rate for the synthesizer

We can pretty much get any rate out of the FAPLL because of the fractional
divider. Let's first try just adjusting the post divider, and if that is
not enough, then reprogram both the fractional divider and the post divider.

Let's also add a define for the fixed SYNTH_PHASE_K instead of using 8.

Cc: Brian Hutchinson <b.hutchman@gmail.com>
Cc: Matthijs van Duin <matthijsvanduin@gmail.com>
Cc: Tero Kristo <t-kristo@ti.com>
Signed-off-by: Tony Lindgren <tony@atomide.com>
Signed-off-by: Tero Kristo <t-kristo@ti.com>
9 years agoclk: ti: Fix FAPLL recalc_rate for rounding errors
Tony Lindgren [Sun, 22 Mar 2015 22:35:24 +0000 (15:35 -0700)]
clk: ti: Fix FAPLL recalc_rate for rounding errors

We need to round the calculated value to have it match the requested rate.

While at it, let's fix a typo and use a define for SYNTH_MAX_DIV_M as we
will need it in later patches for set_rate.

And let's remove two unused includes.

Cc: Brian Hutchinson <b.hutchman@gmail.com>
Cc: Matthijs van Duin <matthijsvanduin@gmail.com>
Cc: Tero Kristo <t-kristo@ti.com>
Signed-off-by: Tony Lindgren <tony@atomide.com>
Signed-off-by: Tero Kristo <t-kristo@ti.com>
9 years agoclk: ti: DRA7: Add timer_sys_ck aliases for Timers 13 through 16
Suman Anna [Fri, 13 Mar 2015 22:58:38 +0000 (17:58 -0500)]
clk: ti: DRA7: Add timer_sys_ck aliases for Timers 13 through 16

The OMAP DMTimer API, omap_dm_timer_set_source(), uses the clock name
timer_sys_ck for setting a timer's clock source for the source index
OMAP_TIMER_SRC_SYS_CLK. There is currently no clock alias data for
the Timers 13 through 16 for this clock name, so add the same.

Signed-off-by: Suman Anna <s-anna@ti.com>
Signed-off-by: Tero Kristo <t-kristo@ti.com>
9 years agoclk: ti: DRA7: Correct timer_sys_ck clock aliases for Timers
Suman Anna [Fri, 13 Mar 2015 22:58:37 +0000 (17:58 -0500)]
clk: ti: DRA7: Correct timer_sys_ck clock aliases for Timers

The OMAP DMTimer API, omap_dm_timer_set_source(), can set the parent
of a timer node using 3 different values that use fixed parent names
for the clocks. The parent name, timer_sys_ck, is used for setting the
parent when used with the source index OMAP_TIMER_SRC_SYS_CLK. This
should point to the TIMER_SYS_CLK and not the SYSCLKIN2, so correct
the clock aliases appropriately. SYSCLKIN2 is not a mandatory clock
input.

Signed-off-by: Suman Anna <s-anna@ti.com>
Signed-off-by: Tero Kristo <t-kristo@ti.com>
9 years agoclk: ti: OMAP5: Correct the DT clock aliases for timers
Suman Anna [Fri, 13 Mar 2015 22:58:36 +0000 (17:58 -0500)]
clk: ti: OMAP5: Correct the DT clock aliases for timers

The DT clock aliases for Timers use the legacy (non-DT) device
names and a source clock named sys_ck. OMAP5 is DT-boot only,
so correct the DT clock aliases to use the DT device names
instead. Also, the source clock name is corrected from 'sys_ck'
to 'timer_sys_ck', the name used by the OMAP dmtimer driver.

Signed-off-by: Suman Anna <s-anna@ti.com>
Signed-off-by: Tero Kristo <t-kristo@ti.com>
9 years agoclk: ti: OMAP4: Remove the legacy timer DT clock aliases
Suman Anna [Fri, 13 Mar 2015 22:58:35 +0000 (17:58 -0500)]
clk: ti: OMAP4: Remove the legacy timer DT clock aliases

The DT clock aliases for timers using the legacy OMAP timer
device names have been cleaned up. These device names reflect
the names used in legacy boot, and are no longer applicable
as OMAP4 is DT boot only now.

Signed-off-by: Suman Anna <s-anna@ti.com>
Signed-off-by: Tero Kristo <t-kristo@ti.com>
9 years agoclk: ti: clk-3xxx-legacy: Correct McBSP related clock aliases
Peter Ujfalusi [Mon, 16 Mar 2015 10:40:58 +0000 (12:40 +0200)]
clk: ti: clk-3xxx-legacy: Correct McBSP related clock aliases

Correct the McBSP2/4 ick mapping (they were 2->4 and 4->2).
Add missing mcbsp clock aliases.
Collect the McBSP clock definition in one location at the same time.

Fixes the following warning on boot:
[    0.307739] omap_hwmod: mcbsp2: _wait_target_ready failed: -16
[    0.307769] omap_hwmod: mcbsp2: cannot be enabled for reset (3)

Signed-off-by: Peter Ujfalusi <peter.ujfalusi@ti.com>
Signed-off-by: Tero Kristo <t-kristo@ti.com>
9 years agoclk: ti: clk-3xxx: Correct McBSP related DT clock definitions
Peter Ujfalusi [Mon, 16 Mar 2015 10:40:57 +0000 (12:40 +0200)]
clk: ti: clk-3xxx: Correct McBSP related DT clock definitions

In DT boot we do not have devices named as omap-mcbsp.X.
Correct the McBSP2/4 ick mapping (they were 2->4 and 4->2).
Collect the McBSP clock definition in one location at the same time.

Fixes the following warning on boot:
[    0.307739] omap_hwmod: mcbsp2: _wait_target_ready failed: -16
[    0.307769] omap_hwmod: mcbsp2: cannot be enabled for reset (3)

Signed-off-by: Peter Ujfalusi <peter.ujfalusi@ti.com>
Signed-off-by: Tero Kristo <t-kristo@ti.com>
9 years agoclk: ti: fix ti_clk_get_reg_addr error handling
Tero Kristo [Mon, 23 Feb 2015 19:06:08 +0000 (21:06 +0200)]
clk: ti: fix ti_clk_get_reg_addr error handling

There is a case where NULL can be a valid return value for
ti_clk_get_reg_addr, specifically the case where both the provider index
and register offsets are zero. In this case, the current error checking
against a NULL pointer will fail. Thus, change the API to return a
ERR_PTR value in an error case, and change all the users of this API to
check against IS_ERR instead.

Signed-off-by: Tero Kristo <t-kristo@ti.com>
Acked-by: Michael Turquette <mturquette@linaro.org>
9 years agoclk: qcom: Add MSM8916 Global Clock Controller support
Georgi Djakov [Wed, 18 Mar 2015 14:08:22 +0000 (16:08 +0200)]
clk: qcom: Add MSM8916 Global Clock Controller support

This patch adds support for the global clock controller found on the MSM8916
based devices. It allows the various device drivers to probe and control
their clocks and resets.

Signed-off-by: Georgi Djakov <georgi.djakov@linaro.org>
[sboyd@codeaurora.org: Removed NULL entry from parent_maps]
Signed-off-by: Stephen Boyd <sboyd@codeaurora.org>
9 years agodt-bindings: Add #defines for MSM8916 clocks and resets
Georgi Djakov [Wed, 18 Mar 2015 14:08:21 +0000 (16:08 +0200)]
dt-bindings: Add #defines for MSM8916 clocks and resets

Add clocks/resets defines for the global clock controller
found on Qualcomm MSM8916 SoCs.

Signed-off-by: Georgi Djakov <georgi.djakov@linaro.org>
Signed-off-by: Stephen Boyd <sboyd@codeaurora.org>
9 years agoclk: qcom: Introduce parent_map tables
Georgi Djakov [Fri, 20 Mar 2015 16:30:26 +0000 (18:30 +0200)]
clk: qcom: Introduce parent_map tables

In the current parent mapping code, we can get duplicate or inconsistent
indexes, which leads to discrepancy between the number of elements in the
array and the number of parents. Until now, this was solved with some
reordering but this is not always possible.

This patch introduces index tables that are used to define the relations
between the PLL source and the hardware mux configuration value.
To accomplish this, here we do the following:
 - Define a parent_map struct to map the relations between PLL source index
 and register configuration value.
 - Add a qcom_find_src_index() function for finding the index of a clock
 matching the specific PLL configuration.
 - Update the {set,get}_parent RCG functions use the newly introduced
 parent_map struct.
 - Convert all existing drivers to the new parent_map tables.

Signed-off-by: Georgi Djakov <georgi.djakov@linaro.org>
Signed-off-by: Stephen Boyd <sboyd@codeaurora.org>
9 years agoclk: qcom: Do some error handling in configure_bank()
Georgi Djakov [Fri, 20 Mar 2015 16:30:25 +0000 (18:30 +0200)]
clk: qcom: Do some error handling in configure_bank()

Currently configure_bank() returns void. Add some error
checking on the regmap calls and propagate if there is
any error.

Signed-off-by: Georgi Djakov <georgi.djakov@linaro.org>
Signed-off-by: Stephen Boyd <sboyd@codeaurora.org>
9 years agoclk: qcom: Fix clk_get_parent function return value
Georgi Djakov [Fri, 20 Mar 2015 16:30:24 +0000 (18:30 +0200)]
clk: qcom: Fix clk_get_parent function return value

According to the common clock framework API, the clk_get_parent() function
should return u8. Currently we are returning negative values on error. Fix
this and use the default parent in case of an error.

Signed-off-by: Georgi Djakov <georgi.djakov@linaro.org>
Signed-off-by: Stephen Boyd <sboyd@codeaurora.org>
9 years agoclk: qcom: fix RCG M/N counter configuration
Archit Taneja [Wed, 4 Mar 2015 09:49:35 +0000 (15:19 +0530)]
clk: qcom: fix RCG M/N counter configuration

Currently, a RCG's M/N counter (used for fraction division) is
set to either 'bypass' (counter disabled) or 'dual edge' (counter
enabled) based on whether the corresponding rcg struct has a mnd
field specified and a non-zero N.

In the case where M and N are the same value, the M/N counter is
still enabled by code even though no division takes place.
Leaving the RCG in such a state can result in improper behavior.
This was observed with the DSI pixel clock RCG when M and N were
both set to 1.

Add an additional check (M != N) to enable the M/N counter only
when it's needed for fraction division.

Signed-off-by: Archit Taneja <architt@codeaurora.org>
Fixes: bcd61c0f535a (clk: qcom: Add support for root clock
generators (RCGs))
Signed-off-by: Stephen Boyd <sboyd@codeaurora.org>
9 years agoMerge tag 'mvebu-clk-4.1' of git://git.infradead.org/linux-mvebu into clk-next
Michael Turquette [Sun, 22 Mar 2015 17:11:39 +0000 (10:11 -0700)]
Merge tag 'mvebu-clk-4.1' of git://git.infradead.org/linux-mvebu into clk-next

clock changes for mvebu for v4.1

- Add clock support for Armada 39x

9 years agoclk: sunxi: Register divs clocks before factor clocks
Chen-Yu Tsai [Thu, 19 Mar 2015 17:19:05 +0000 (01:19 +0800)]
clk: sunxi: Register divs clocks before factor clocks

We want to reparent AHB clock to PLL6 on sun5i/sun7i using the assigned
clocks properties. AHB is a factor clock, while PLL6 is a divs clock.

Register divs clocks before factor clocks so reparenting works. This
is only needed because we do the reparenting on the clock provider.

The proper way to fix this is to split out all the old sunxi clocks
into separate CLK_OF_DECLARE statements, like we are doing for sun9i.

Signed-off-by: Chen-Yu Tsai <wens@csie.org>
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
9 years agoclk: sunxi: Add "cpu" to list of protected clocks for sun5i
Chen-Yu Tsai [Thu, 19 Mar 2015 17:19:04 +0000 (01:19 +0800)]
clk: sunxi: Add "cpu" to list of protected clocks for sun5i

Now that the ahb clock on sun5i/sun7i is muxable, ahb is no longer
guaranteed to be a child of the cpu clock. Add the cpu clock to
the list of protected clocks so it doesn't get disabled.

Signed-off-by: Chen-Yu Tsai <wens@csie.org>
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
9 years agoclk: sunxi: Add muxable ahb factors clock for sun5i and sun7i
Chen-Yu Tsai [Thu, 19 Mar 2015 17:19:03 +0000 (01:19 +0800)]
clk: sunxi: Add muxable ahb factors clock for sun5i and sun7i

The AHB clock on sun5i and sun7i are muxable divider clocks.
Use a factors clock to support them.

Signed-off-by: Chen-Yu Tsai <wens@csie.org>
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
9 years agoMerge branch 'clk-fixes' into clk-next
Stephen Boyd [Thu, 12 Mar 2015 00:07:37 +0000 (17:07 -0700)]
Merge branch 'clk-fixes' into clk-next

9 years agoclk: Replace of_clk_get_by_clkspec() with of_clk_get_from_provider()
Stephen Boyd [Thu, 5 Feb 2015 23:39:11 +0000 (15:39 -0800)]
clk: Replace of_clk_get_by_clkspec() with of_clk_get_from_provider()

of_clk_get_by_clkspec() has the same function signature as
of_clk_get_from_provider()

 struct clk *of_clk_get_by_clkspec(struct of_phandle_args
 *clkspec)
 struct clk *of_clk_get_from_provider(struct of_phandle_args
 *clkspec)

except of_clk_get_by_clkspec() checks to make sure clkspec is not
NULL. Let's remove of_clk_get_by_clkspec() and replace the
callers of it (clkconf.c) with of_clk_get_from_provider().

Cc: Sylwester Nawrocki <s.nawrocki@samsung.com>
Reviewed-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Signed-off-by: Stephen Boyd <sboyd@codeaurora.org>
9 years agoclk: Rename child_node to clks_node to avoid confusion
Stephen Boyd [Fri, 6 Feb 2015 19:42:44 +0000 (11:42 -0800)]
clk: Rename child_node to clks_node to avoid confusion

The child_node member of struct clk is named the same as the
child_node member of struct clk_core. Let's rename the struct
clk's member to clks_node to avoid getting confused with the
child_node member of struct clk_core and to match the name of the
list head, clks.

Reviewed-by: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Cc: Alban Browaeys <alban.browaeys@gmail.com>
Signed-off-by: Stephen Boyd <sboyd@codeaurora.org>
9 years agoclk: qcom: Properly change rates for ahbix clock
Stephen Boyd [Fri, 6 Mar 2015 23:41:53 +0000 (15:41 -0800)]
clk: qcom: Properly change rates for ahbix clock

The ahbix clock can never be turned off in practice. To change the
rates we need to switch the mux off the M/N counter to an always on
source (XO), reprogram the M/N counter to get the rate we want and
finally switch back to the M/N counter. Add a new ops structure
for this type of clock so that we can set the rate properly.

Fixes: c99e515a92e9 "clk: qcom: Add IPQ806X LPASS clock controller (LCC) driver"
Tested-by: Kenneth Westfield <kwestfie@codeaurora.org>
Signed-off-by: Stephen Boyd <sboyd@codeaurora.org>
9 years agoMerge tag 'v3.20-exynos5433-clk' of git://linuxtv.org/snawrocki/samsung into clk...
Michael Turquette [Fri, 6 Mar 2015 19:21:13 +0000 (11:21 -0800)]
Merge tag 'v3.20-exynos5433-clk' of git://linuxtv.org/snawrocki/samsung into clk-next

Clock controller driver for Exynos 5433 SoC.

9 years agoclk: qcom: Fix i2c frequency table
Stephen Boyd [Mon, 23 Feb 2015 21:30:28 +0000 (13:30 -0800)]
clk: qcom: Fix i2c frequency table

PXO is 25MHz, not 27MHz. Fix the table.

Fixes: 24d8fba44af3 "clk: qcom: Add support for IPQ8064's global
clock controller (GCC)"

Signed-off-by: Stephen Boyd <sboyd@codeaurora.org>
Reviewed-by: Andy Gross <agross@codeaurora.org>
Tested-by: Andy Gross <agross@codeaurora.org>
Signed-off-by: Michael Turquette <mturquette@linaro.org>
9 years agoclk: clk_set_parent() with current parent shouldn't fail
Stephen Boyd [Mon, 2 Feb 2015 22:11:25 +0000 (14:11 -0800)]
clk: clk_set_parent() with current parent shouldn't fail

If a driver calls clk_set_parent(clk, parent) and parent is the
current parent of clk we shouldn't fail in any case.
Unfortunately if clk is a read-only mux we return -ENOSYS
because we think we can't change the parent, except for in this
special case where we don't actually need to change the parent at
all. Return 0 in such a situation.

Signed-off-by: Stephen Boyd <sboyd@codeaurora.org>
Acked-by: Philipp Zabel <p.zabel@pengutronix.de>
Signed-off-by: Michael Turquette <mturquette@linaro.org>
9 years agoclk: Add tracepoints for hardware operations
Stephen Boyd [Mon, 2 Feb 2015 22:37:41 +0000 (14:37 -0800)]
clk: Add tracepoints for hardware operations

It's useful to have tracepoints around operations that change the
hardware state so that we can debug clock hardware performance
and operations. Four basic types of events are supported: on/off
events for enable, disable, prepare, unprepare that only record
an event and a clock name, rate changing events for
clk_set_{min_,max_}rate{_range}(), phase changing events for
clk_set_phase() and parent changing events for clk_set_parent().

Cc: Steven Rostedt <rostedt@goodmis.org>
Signed-off-by: Stephen Boyd <sboyd@codeaurora.org>
Signed-off-by: Michael Turquette <mturquette@linaro.org>
9 years agoclk: Missing set_phase op is an error
Stephen Boyd [Mon, 2 Feb 2015 22:09:43 +0000 (14:09 -0800)]
clk: Missing set_phase op is an error

If a clock's clk_ops doesn't have the set_phase op set we should
return an error from clk_set_phase(). This way clock consumers
know that when they tried to set a phase it didn't work, as
opposed to the current behavior where the return value is 0
meaning success.

Signed-off-by: Stephen Boyd <sboyd@codeaurora.org>
Signed-off-by: Michael Turquette <mturquette@linaro.org>
9 years agoclk: fractional-divider: support for divider bypassing
Heikki Krogerus [Mon, 2 Feb 2015 13:37:04 +0000 (15:37 +0200)]
clk: fractional-divider: support for divider bypassing

If the divider or multiplier values are 0 in the register, bypassing the
divider and returning the parent clock rate in clk_fd_recalc_rate().

Signed-off-by: Heikki Krogerus <heikki.krogerus@linux.intel.com>
Reviewed-by: Stephen Boyd <sboyd@codeaurora.org>
Signed-off-by: Michael Turquette <mturquette@linaro.org>
[mturquette@linaro.org: fixed commitlog typo]

9 years agoclk: Use lockdep asserts to find missing hold of prepare_lock
Krzysztof Kozlowski [Fri, 9 Jan 2015 08:28:10 +0000 (09:28 +0100)]
clk: Use lockdep asserts to find missing hold of prepare_lock

Add lockdep asserts for holding the prepare_lock to all functions
marking this as a requirement in description. Add this to private and
exported functions so all locking misuse could be detected during
debugging.

Signed-off-by: Krzysztof Kozlowski <k.kozlowski@samsung.com>
Signed-off-by: Michael Turquette <mturquette@linaro.org>
9 years agoASoC: kirkwood: fix struct clk pointer comparing
Shawn Guo [Wed, 25 Feb 2015 14:53:38 +0000 (22:53 +0800)]
ASoC: kirkwood: fix struct clk pointer comparing

Since commit 035a61c314eb ("clk: Make clk API return per-user struct clk
instances"), clk API users can no longer check if two struct clk
pointers are pointing to the same hardware clock, i.e. struct clk_hw, by
simply comparing two pointers.  That's because with the per-user clk
change, a brand new struct clk is created whenever clients try to look
up the clock by calling clk_get() or sister functions like clk_get_sys()
and of_clk_get().  This changes the original behavior where the struct
clk is only created for once when clock driver registers the clock to
CCF in the first place.  The net change here is before commit
035a61c314eb the struct clk pointer is unique for given hardware
clock, while after the commit the pointers returned by clk lookup calls
become different for the same hardware clock.

That said, the struct clk pointer comparing in the code doesn't work any
more.  Call helper function clk_is_match() instead to fix the problem.

Signed-off-by: Shawn Guo <shawn.guo@linaro.org>
Signed-off-by: Michael Turquette <mturquette@linaro.org>
Signed-off-by: Stephen Boyd <sboyd@codeaurora.org>
9 years agoASoC: fsl_spdif: fix struct clk pointer comparing
Shawn Guo [Wed, 25 Feb 2015 14:53:37 +0000 (22:53 +0800)]
ASoC: fsl_spdif: fix struct clk pointer comparing

Since commit 035a61c314eb ("clk: Make clk API return per-user struct clk
instances"), clk API users can no longer check if two struct clk
pointers are pointing to the same hardware clock, i.e. struct clk_hw, by
simply comparing two pointers.  That's because with the per-user clk
change, a brand new struct clk is created whenever clients try to look
up the clock by calling clk_get() or sister functions like clk_get_sys()
and of_clk_get().  This changes the original behavior where the struct
clk is only created for once when clock driver registers the clock to
CCF in the first place.  The net change here is before commit
035a61c314eb the struct clk pointer is unique for given hardware
clock, while after the commit the pointers returned by clk lookup calls
become different for the same hardware clock.

That said, the struct clk pointer comparing in the code doesn't work any
more.  Call helper function clk_is_match() instead to fix the problem.

Signed-off-by: Shawn Guo <shawn.guo@linaro.org>
Signed-off-by: Michael Turquette <mturquette@linaro.org>
Signed-off-by: Stephen Boyd <sboyd@codeaurora.org>
9 years agoARM: imx: fix struct clk pointer comparing
Shawn Guo [Wed, 25 Feb 2015 14:53:32 +0000 (22:53 +0800)]
ARM: imx: fix struct clk pointer comparing

Since commit 035a61c314eb ("clk: Make clk API return per-user struct clk
instances"), clk API users can no longer check if two struct clk
pointers are pointing to the same hardware clock, i.e. struct clk_hw, by
simply comparing two pointers.  That's because with the per-user clk
change, a brand new struct clk is created whenever clients try to look
up the clock by calling clk_get() or sister functions like clk_get_sys()
and of_clk_get().  This changes the original behavior where the struct
clk is only created for once when clock driver registers the clock to
CCF in the first place.  The net change here is before commit
035a61c314eb the struct clk pointer is unique for given hardware
clock, while after the commit the pointers returned by clk lookup calls
become different for the same hardware clock.

That said, the struct clk pointer comparing in the code doesn't work any
more.  Call helper function clk_is_match() instead to fix the problem.

Signed-off-by: Shawn Guo <shawn.guo@linaro.org>
Signed-off-by: Michael Turquette <mturquette@linaro.org>
Signed-off-by: Stephen Boyd <sboyd@codeaurora.org>
9 years agoclk: introduce clk_is_match
Michael Turquette [Wed, 25 Feb 2015 17:11:01 +0000 (09:11 -0800)]
clk: introduce clk_is_match

Some drivers compare struct clk pointers as a means of knowing
if the two pointers reference the same clock hardware. This behavior is
dubious (drivers must not dereference struct clk), but did not cause any
regressions until the per-user struct clk patch was merged. Now the test
for matching clk's will always fail with per-user struct clk's.

clk_is_match is introduced to fix the regression and prevent drivers
from comparing the pointers manually.

Fixes: 035a61c314eb ("clk: Make clk API return per-user struct clk instances")
Cc: Russell King <linux@arm.linux.org.uk>
Cc: Shawn Guo <shawn.guo@linaro.org>
Cc: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Signed-off-by: Michael Turquette <mturquette@linaro.org>
[arnd@arndb.de: Fix COMMON_CLK=N && HAS_CLK=Y config]
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
[sboyd@codeaurora.org: const arguments to clk_is_match() and
remove unnecessary ternary operation]
Signed-off-by: Stephen Boyd <sboyd@codeaurora.org>
9 years agoclk: don't export static symbol
Julia Lawall [Wed, 11 Mar 2015 16:56:31 +0000 (17:56 +0100)]
clk: don't export static symbol

The semantic patch that fixes this problem is as follows:
(http://coccinelle.lip6.fr/)

// <smpl>
@r@
type T;
identifier f;
@@

static T f (...) { ... }

@@
identifier r.f;
declarer name EXPORT_SYMBOL_GPL;
@@

-EXPORT_SYMBOL_GPL(f);
// </smpl>

Signed-off-by: Julia Lawall <Julia.Lawall@lip6.fr>
Fixes: 035a61c314eb "clk: Make clk API return per-user struct clk instances"
Signed-off-by: Stephen Boyd <sboyd@codeaurora.org>
9 years agoclk: divider: fix calculation of initial best divider when rounding to closest
Uwe Kleine-König [Sat, 21 Feb 2015 10:40:25 +0000 (11:40 +0100)]
clk: divider: fix calculation of initial best divider when rounding to closest

Similar to the reasoning for the previous commit

DIV_ROUND_CLOSEST(parent_rate, rate)

might not be the best integer divisor to get a good approximation for
rate from parent_rate (given the metric for CLK_DIVIDER_ROUND_CLOSEST).

For example assume a parent rate of 1000 Hz and a target rate of 700.
Using DIV_ROUND_CLOSEST the suggested divisor gets calculated to 1
resulting in a target rate of 1000 with a delta of 300 to the desired
rate. With choosing 2 as divisor however the resulting rate is 500 which
is nearer to 700.

Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Acked-by: Sascha Hauer <s.hauer@pengutronix.de>
Acked-by: Maxime Coquelin <maxime.coquelin@st.com>
Signed-off-by: Michael Turquette <mturquette@linaro.org>
9 years agoclk: divider: fix selection of divider when rounding to closest
Uwe Kleine-König [Sat, 21 Feb 2015 10:40:24 +0000 (11:40 +0100)]
clk: divider: fix selection of divider when rounding to closest

It's an invalid approach to assume that among two divider values
the one nearer the exact divider is the better one.

Assume a parent rate of 1000 Hz, a divider with CLK_DIVIDER_POWER_OF_TWO
and a target rate of 89 Hz. The exact divider is ~ 11.236 so 8 and 16
are the candidates to choose from yielding rates 125 Hz and 62.5 Hz
respectivly. While 8 is nearer to 11.236 than 16 is, the latter is still
the better divider as 62.5 is nearer to 89 than 125 is.

Fixes: 774b514390b1 (clk: divider: Add round to closest divider)
Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Acked-by: Sascha Hauer <s.hauer@pengutronix.de>
Acked-by: Maxime Coquelin <maxime.coquelin@st.com>
Signed-off-by: Michael Turquette <mturquette@linaro.org>
9 years agoclk: divider: fix calculation of maximal parent rate for a given divider
Uwe Kleine-König [Sat, 21 Feb 2015 10:40:23 +0000 (11:40 +0100)]
clk: divider: fix calculation of maximal parent rate for a given divider

The rate provided at the output of a clk-divider is calculated as:

DIV_ROUND_UP(parent_rate, div)

since commit b11d282dbea2 (clk: divider: fix rate calculation for
fractional rates). So to yield a rate not bigger than r parent_rate
must be <= r * div.

The effect of choosing a parent rate that is too big as was done before
this patch results in wrongly ruling out good dividers.

Note that this is not a complete fix as __clk_round_rate might return a
value >= its 2nd parameter. Also for dividers with
CLK_DIVIDER_ROUND_CLOSEST set the calculation is not accurate. But this
fixes the test case by Sascha Hauer that uses a chain of three dividers
under a fixed clock.

Fixes: b11d282dbea2 (clk: divider: fix rate calculation for fractional rates)
Suggested-by: Sascha Hauer <s.hauer@pengutronix.de>
Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Acked-by: Sascha Hauer <s.hauer@pengutronix.de>
Signed-off-by: Michael Turquette <mturquette@linaro.org>
9 years agoclk: divider: return real rate instead of divider value
Heiko Stübner [Tue, 24 Feb 2015 10:39:25 +0000 (11:39 +0100)]
clk: divider: return real rate instead of divider value

Commit bca9690b9426 ("clk: divider: Make generic for usage elsewhere")
returned only the divider value for read-only dividers instead of the
actual rate.

Fixes: bca9690b9426 ("clk: divider: Make generic for usage elsewhere")
Signed-off-by: Heiko Stuebner <heiko@sntech.de>
Reviewed-by: James Hogan <james.hogan@imgtec.com>
Tested-by: James Hogan <james.hogan@imgtec.com>
Acked-by: Stephen Boyd <sboyd@codeaurora.org>
Signed-off-by: Michael Turquette <mturquette@linaro.org>
9 years agoclk: mvebu: add Marvell Armada 39x driver
Thomas Petazzoni [Tue, 3 Mar 2015 14:41:09 +0000 (15:41 +0100)]
clk: mvebu: add Marvell Armada 39x driver

This commit adds a new clock driver for the Marvell Armada 39x family
of processors. This driver is fairly similar to the ones already used
on other Marvell EBU processors, with the following main differences:

 * Different set of ratios
 * Different set of core clocks
 * Configurable reference clock in frequency

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Gregory CLEMENT <gregory.clement@free-electrons.com>
9 years agoclk: mvebu: extend common code to allow an optional refclk
Thomas Petazzoni [Tue, 3 Mar 2015 14:41:08 +0000 (15:41 +0100)]
clk: mvebu: extend common code to allow an optional refclk

The Armada 39x, contrary to its predecessor, has a configurable
reference clock frequency, of either 25 Mhz, or 40 Mhz. For the
previous SoCs, it was fixed to 25 Mhz and described directly as such
in the Device Tree.

For Armada 39x, we need to read certain registers to know whether the
frequency is 25 or 40 Mhz. Therefore, this commit extends the common
mvebu clock code to allow the SoC-specific code to say it wants to
register a reference clock, by giving a non-NULL ->get_refclk_freq()
function pointer in its coreclk_soc_desc structure.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Gregory CLEMENT <gregory.clement@free-electrons.com>
9 years agodevicetree: bindings: update DT bindings for Marvell EBU clock support
Thomas Petazzoni [Tue, 3 Mar 2015 14:41:05 +0000 (15:41 +0100)]
devicetree: bindings: update DT bindings for Marvell EBU clock support

With the introduction of the Marvell Armada 39x SoC, the DT bindings
for Marvell EBU clocks need to be extended. This commit include the
corresponding update to the Device Tree bindings documentation.

Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Gregory CLEMENT <gregory.clement@free-electrons.com>
9 years agoclk: qcom: fix platform_no_drv_owner.cocci warnings
kbuild test robot [Wed, 28 Jan 2015 07:11:34 +0000 (15:11 +0800)]
clk: qcom: fix platform_no_drv_owner.cocci warnings

drivers/clk/qcom/lcc-msm8960.c:577:3-8: No need to set .owner here. The core will do it.

 Remove .owner field if calls are used which set it automatically

Generated by: scripts/coccinelle/api/platform_no_drv_owner.cocci

Signed-off-by: Fengguang Wu <fengguang.wu@intel.com>
Reviewed-by: Stephen Boyd <sboyd@codeaurora.org>
Signed-off-by: Michael Turquette <mturquette@linaro.org>
9 years agoclk: qcom: fix platform_no_drv_owner.cocci warnings
kbuild test robot [Wed, 28 Jan 2015 07:00:51 +0000 (15:00 +0800)]
clk: qcom: fix platform_no_drv_owner.cocci warnings

drivers/clk/qcom/lcc-ipq806x.c:465:3-8: No need to set .owner here. The core will do it.

 Remove .owner field if calls are used which set it automatically

Generated by: scripts/coccinelle/api/platform_no_drv_owner.cocci

CC: Rajendra Nayak <rnayak@codeaurora.org>
Signed-off-by: Fengguang Wu <fengguang.wu@intel.com>
Reviewed-by: Stephen Boyd <sboyd@codeaurora.org>
Signed-off-by: Michael Turquette <mturquette@linaro.org>
9 years agoclk: qcom: Add PLL4 vote clock
Stephen Boyd [Thu, 29 Jan 2015 23:38:13 +0000 (15:38 -0800)]
clk: qcom: Add PLL4 vote clock

This clock is needed for most audio clock frequencies. Add it.

Signed-off-by: Stephen Boyd <sboyd@codeaurora.org>
Signed-off-by: Michael Turquette <mturquette@linaro.org>
9 years agoclk: qcom: lcc-msm8960: Fix PLL rate detection
Stephen Boyd [Thu, 29 Jan 2015 23:38:12 +0000 (15:38 -0800)]
clk: qcom: lcc-msm8960: Fix PLL rate detection

regmap_read() returns 0 on success, not the value of the register
that is read. Fix it so we properly detect the frequency plan.

Fixes: b82875ee07e5 "clk: qcom: Add MSM8960/APQ8064 LPASS clock
controller (LCC) driver"
Signed-off-by: Stephen Boyd <sboyd@codeaurora.org>
Signed-off-by: Michael Turquette <mturquette@linaro.org>
9 years agoclk: qcom: Fix slimbus n and m val offsets
Stephen Boyd [Thu, 29 Jan 2015 23:38:11 +0000 (15:38 -0800)]
clk: qcom: Fix slimbus n and m val offsets

These shifts were copy/pasted from the pcm which is a different
size RCG. Use the correct offsets so that slimbus rates are
correct.

Fixes: b82875ee07e5 "clk: qcom: Add MSM8960/APQ8064 LPASS clock controller (LCC) driver"
Signed-off-by: Stephen Boyd <sboyd@codeaurora.org>
Signed-off-by: Michael Turquette <mturquette@linaro.org>
9 years agoclk: ti: Fix FAPLL parent enable bit handling
Tony Lindgren [Wed, 28 Jan 2015 17:00:49 +0000 (09:00 -0800)]
clk: ti: Fix FAPLL parent enable bit handling

Commit 163152cbbe32 ("clk: ti: Add support for FAPLL on dm816x")
added basic support for the FAPLL on dm818x, but has a bug for the
parent PLL enable bit. The FAPLL_MAIN_PLLEN is defined as BIT(3)
but the code is doing a shift on it.

This means the parent PLL won't get disabled even if all it's child
synthesizers are disabled.

Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Cc: Brian Hutchinson <b.hutchman@gmail.com>
Signed-off-by: Tony Lindgren <tony@atomide.com>
Signed-off-by: Michael Turquette <mturquette@linaro.org>
9 years agoclk: sunxi: Add support for sun9i A80 USB clocks and resets
Chen-Yu Tsai [Tue, 27 Jan 2015 19:54:07 +0000 (03:54 +0800)]
clk: sunxi: Add support for sun9i A80 USB clocks and resets

The USB controller/phy clocks and reset controls are in a separate
address block, unlike previous SoCs where they were in the clock
controller. Also, access to the address block is controlled by a
clock gate to AHB.

Add support for resets requiring a clock to be enabled when
asserting/deasserting the reset controls, and add the sun9i USB
clocks.

Signed-off-by: Chen-Yu Tsai <wens@csie.org>
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
9 years agoclk: sunxi: Move USB clocks to separate file
Chen-Yu Tsai [Tue, 27 Jan 2015 19:54:06 +0000 (03:54 +0800)]
clk: sunxi: Move USB clocks to separate file

The USB clocks originally shared code with the gates clocks, but had
additional reset controllers. Move these to a separate file. This will
allow us to add new support for slightly different USB clocks, such as
on the A80, without affecting gates clocks, and also facilitate the
migration of gates clocks to a generic solution.

This also cleans up the USB clocks code slightly, such as adding
newlines, getting rid of the unused clkdev call, using a simple
u32 instead of BITMAP for the clock masks, using BIT() macro to
declare the clock bitmasks, and using of_io_request_and_map() to
get the I/O address.

Signed-off-by: Chen-Yu Tsai <wens@csie.org>
Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
9 years agoLinux 4.0-rc1 v4.0-rc1
Linus Torvalds [Mon, 23 Feb 2015 02:21:14 +0000 (18:21 -0800)]
Linux 4.0-rc1

.. after extensive statistical analysis of my G+ polling, I've come to
the inescapable conclusion that internet polls are bad.

Big surprise.

But "Hurr durr I'ma sheep" trounced "I like online polls" by a 62-to-38%
margin, in a poll that people weren't even supposed to participate in.
Who can argue with solid numbers like that? 5,796 votes from people who
can't even follow the most basic directions?

In contrast, "v4.0" beat out "v3.20" by a slimmer margin of 56-to-44%,
but with a total of 29,110 votes right now.

Now, arguably, that vote spread is only about 3,200 votes, which is less
than the almost six thousand votes that the "please ignore" poll got, so
it could be considered noise.

But hey, I asked, so I'll honor the votes.

9 years agoMerge tag 'ext4_for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tytso...
Linus Torvalds [Mon, 23 Feb 2015 02:05:13 +0000 (18:05 -0800)]
Merge tag 'ext4_for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tytso/ext4

Pull ext4 fixes from Ted Ts'o:
 "Ext4 bug fixes.

  We also reserved code points for encryption and read-only images (for
  which the implementation is mostly just the reserved code point for a
  read-only feature :-)"

* tag 'ext4_for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tytso/ext4:
  ext4: fix indirect punch hole corruption
  ext4: ignore journal checksum on remount; don't fail
  ext4: remove duplicate remount check for JOURNAL_CHECKSUM change
  ext4: fix mmap data corruption in nodelalloc mode when blocksize < pagesize
  ext4: support read-only images
  ext4: change to use setup_timer() instead of init_timer()
  ext4: reserve codepoints used by the ext4 encryption feature
  jbd2: complain about descriptor block checksum errors

9 years agoMerge branch 'for-linus-2' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs
Linus Torvalds [Mon, 23 Feb 2015 01:42:14 +0000 (17:42 -0800)]
Merge branch 'for-linus-2' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs

Pull more vfs updates from Al Viro:
 "Assorted stuff from this cycle.  The big ones here are multilayer
  overlayfs from Miklos and beginning of sorting ->d_inode accesses out
  from David"

* 'for-linus-2' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs: (51 commits)
  autofs4 copy_dev_ioctl(): keep the value of ->size we'd used for allocation
  procfs: fix race between symlink removals and traversals
  debugfs: leave freeing a symlink body until inode eviction
  Documentation/filesystems/Locking: ->get_sb() is long gone
  trylock_super(): replacement for grab_super_passive()
  fanotify: Fix up scripted S_ISDIR/S_ISREG/S_ISLNK conversions
  Cachefiles: Fix up scripted S_ISDIR/S_ISREG/S_ISLNK conversions
  VFS: (Scripted) Convert S_ISLNK/DIR/REG(dentry->d_inode) to d_is_*(dentry)
  SELinux: Use d_is_positive() rather than testing dentry->d_inode
  Smack: Use d_is_positive() rather than testing dentry->d_inode
  TOMOYO: Use d_is_dir() rather than d_inode and S_ISDIR()
  Apparmor: Use d_is_positive/negative() rather than testing dentry->d_inode
  Apparmor: mediated_filesystem() should use dentry->d_sb not inode->i_sb
  VFS: Split DCACHE_FILE_TYPE into regular and special types
  VFS: Add a fallthrough flag for marking virtual dentries
  VFS: Add a whiteout dentry type
  VFS: Introduce inode-getting helpers for layered/unioned fs environments
  Infiniband: Fix potential NULL d_inode dereference
  posix_acl: fix reference leaks in posix_acl_create
  autofs4: Wrong format for printing dentry
  ...

9 years agoMerge branch 'fixes' of git://ftp.arm.linux.org.uk/~rmk/linux-arm
Linus Torvalds [Sun, 22 Feb 2015 17:57:16 +0000 (09:57 -0800)]
Merge branch 'fixes' of git://ftp.arm.linux.org.uk/~rmk/linux-arm

Pull ARM fix from Russell King:
 "Just one fix this time around.  __iommu_alloc_buffer() can cause a
  BUG() if dma_alloc_coherent() is called with either __GFP_DMA32 or
  __GFP_HIGHMEM set.  The patch from Alexandre addresses this"

* 'fixes' of git://ftp.arm.linux.org.uk/~rmk/linux-arm:
  ARM: 8305/1: DMA: Fix kzalloc flags in __iommu_alloc_buffer()

9 years agoautofs4 copy_dev_ioctl(): keep the value of ->size we'd used for allocation
Al Viro [Sun, 22 Feb 2015 03:19:57 +0000 (22:19 -0500)]
autofs4 copy_dev_ioctl(): keep the value of ->size we'd used for allocation

X-Coverup: just ask spender
Cc: stable@vger.kernel.org
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
9 years agoprocfs: fix race between symlink removals and traversals
Al Viro [Sun, 22 Feb 2015 03:16:11 +0000 (22:16 -0500)]
procfs: fix race between symlink removals and traversals

use_pde()/unuse_pde() in ->follow_link()/->put_link() resp.

Cc: stable@vger.kernel.org
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
9 years agodebugfs: leave freeing a symlink body until inode eviction
Al Viro [Sun, 22 Feb 2015 03:05:11 +0000 (22:05 -0500)]
debugfs: leave freeing a symlink body until inode eviction

As it is, we have debugfs_remove() racing with symlink traversals.
Supply ->evict_inode() and do freeing there - inode will remain
pinned until we are done with the symlink body.

And rip the idiocy with checking if dentry is positive right after
we'd verified debugfs_positive(), which is a stronger check...

Cc: stable@vger.kernel.org
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
9 years agoDocumentation/filesystems/Locking: ->get_sb() is long gone
Al Viro [Sat, 21 Feb 2015 09:59:02 +0000 (04:59 -0500)]
Documentation/filesystems/Locking: ->get_sb() is long gone

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
9 years agotrylock_super(): replacement for grab_super_passive()
Konstantin Khlebnikov [Thu, 19 Feb 2015 17:19:35 +0000 (20:19 +0300)]
trylock_super(): replacement for grab_super_passive()

I've noticed significant locking contention in memory reclaimer around
sb_lock inside grab_super_passive(). Grab_super_passive() is called from
two places: in icache/dcache shrinkers (function super_cache_scan) and
from writeback (function __writeback_inodes_wb). Both are required for
progress in memory allocator.

Grab_super_passive() acquires sb_lock to increment sb->s_count and check
sb->s_instances. It seems sb->s_umount locked for read is enough here:
super-block deactivation always runs under sb->s_umount locked for write.
Protecting super-block itself isn't a problem: in super_cache_scan() sb
is protected by shrinker_rwsem: it cannot be freed if its slab shrinkers
are still active. Inside writeback super-block comes from inode from bdi
writeback list under wb->list_lock.

This patch removes locking sb_lock and checks s_instances under s_umount:
generic_shutdown_super() unlinks it under sb->s_umount locked for write.
New variant is called trylock_super() and since it only locks semaphore,
callers must call up_read(&sb->s_umount) instead of drop_super(sb) when
they're done.

Signed-off-by: Konstantin Khlebnikov <khlebnikov@yandex-team.ru>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
9 years agofanotify: Fix up scripted S_ISDIR/S_ISREG/S_ISLNK conversions
David Howells [Thu, 29 Jan 2015 12:02:36 +0000 (12:02 +0000)]
fanotify: Fix up scripted S_ISDIR/S_ISREG/S_ISLNK conversions

Fanotify probably doesn't want to watch autodirs so make it use d_can_lookup()
rather than d_is_dir() when checking a dir watch and give an error on fake
directories.

Signed-off-by: David Howells <dhowells@redhat.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
9 years agoCachefiles: Fix up scripted S_ISDIR/S_ISREG/S_ISLNK conversions
David Howells [Thu, 29 Jan 2015 12:02:36 +0000 (12:02 +0000)]
Cachefiles: Fix up scripted S_ISDIR/S_ISREG/S_ISLNK conversions

Fix up the following scripted S_ISDIR/S_ISREG/S_ISLNK conversions (or lack
thereof) in cachefiles:

 (1) Cachefiles mostly wants to use d_can_lookup() rather than d_is_dir() as
     it doesn't want to deal with automounts in its cache.

 (2) Coccinelle didn't find S_IS* expressions in ASSERT() statements in
     cachefiles.

Signed-off-by: David Howells <dhowells@redhat.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>