1
linux/arch/arm/mach-omap1/clock.c

841 lines
21 KiB
C
Raw Normal View History

// SPDX-License-Identifier: GPL-2.0-only
/*
* linux/arch/arm/mach-omap1/clock.c
*
* Copyright (C) 2004 - 2005, 2009-2010 Nokia Corporation
* Written by Tuukka Tikkanen <tuukka.tikkanen@elektrobit.com>
*
* Modified to use omap shared clock framework by
* Tony Lindgren <tony@atomide.com>
*/
#include <linux/kernel.h>
#include <linux/export.h>
#include <linux/list.h>
#include <linux/errno.h>
#include <linux/err.h>
#include <linux/io.h>
#include <linux/clk.h>
#include <linux/clkdev.h>
ARM: OMAP1: clock: Convert to CCF OMAP1 still uses its own implementation of standard clock API defined in include/linux/clk.h. Internals of that implementation are not visible outside OMAP1 directory. As a consequence, device drivers are not able to register clocks potentially provided by peripheral devices. Drop OMAP1 implementation of the clock API and enable common clock framework. Modify the remaining low level code to be compatible with clock provider API and register the clocks with CCF. Move initialisation of clocks to omap1_timer_init() to avoid memory allocation issues at early setup phase from where omap1_init_early() is called. Register the clocks after initialization of clock I/O registers, local clock pointers used by OMAP1 clock ops, and local .rate fields of clocks with no local implementation of .recalc ops, so CCF structures are populated with correct data during clock registration. Instead of enabling some of the registered clocks, flag them for CCF as critical. Introduce .is_enabled op using code that verifies hardware status of clock enablement, split out from implementation of .disable_unused op, so the latter is actually called by CCF for not requested but hardware enabled clocks. Add .round_rate ops where missing so .set_rate ops are called by CCF as expected. Since CCF allows parallel execution of .enable/.disable and .set_rate ops, protect registers shared among those groups of ops from concurrent access with spinlocks. Drop local debugfs support in favor of that provided by CCF. v2: flag tc2_ck as CLK_IS_CRITICAL (Aaro) v3: rebase on top of soc/omap1-multiplatform-5.18, - drop no longer needed includes from arch/arm/mach-omap1/io.c Signed-off-by: Janusz Krzysztofik <jmkrzyszt@gmail.com> Acked-by: Tony Lindgren <tony@atomide.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2022-04-10 06:07:57 -07:00
#include <linux/clk-provider.h>
#include <linux/soc/ti/omap1-io.h>
ARM: OMAP1: clock: Convert to CCF OMAP1 still uses its own implementation of standard clock API defined in include/linux/clk.h. Internals of that implementation are not visible outside OMAP1 directory. As a consequence, device drivers are not able to register clocks potentially provided by peripheral devices. Drop OMAP1 implementation of the clock API and enable common clock framework. Modify the remaining low level code to be compatible with clock provider API and register the clocks with CCF. Move initialisation of clocks to omap1_timer_init() to avoid memory allocation issues at early setup phase from where omap1_init_early() is called. Register the clocks after initialization of clock I/O registers, local clock pointers used by OMAP1 clock ops, and local .rate fields of clocks with no local implementation of .recalc ops, so CCF structures are populated with correct data during clock registration. Instead of enabling some of the registered clocks, flag them for CCF as critical. Introduce .is_enabled op using code that verifies hardware status of clock enablement, split out from implementation of .disable_unused op, so the latter is actually called by CCF for not requested but hardware enabled clocks. Add .round_rate ops where missing so .set_rate ops are called by CCF as expected. Since CCF allows parallel execution of .enable/.disable and .set_rate ops, protect registers shared among those groups of ops from concurrent access with spinlocks. Drop local debugfs support in favor of that provided by CCF. v2: flag tc2_ck as CLK_IS_CRITICAL (Aaro) v3: rebase on top of soc/omap1-multiplatform-5.18, - drop no longer needed includes from arch/arm/mach-omap1/io.c Signed-off-by: Janusz Krzysztofik <jmkrzyszt@gmail.com> Acked-by: Tony Lindgren <tony@atomide.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2022-04-10 06:07:57 -07:00
#include <linux/spinlock.h>
#include <asm/mach-types.h>
#include "hardware.h"
#include "soc.h"
#include "iomap.h"
#include "clock.h"
#include "opp.h"
#include "sram.h"
__u32 arm_idlect1_mask;
ARM: OMAP1: clock: Convert to CCF OMAP1 still uses its own implementation of standard clock API defined in include/linux/clk.h. Internals of that implementation are not visible outside OMAP1 directory. As a consequence, device drivers are not able to register clocks potentially provided by peripheral devices. Drop OMAP1 implementation of the clock API and enable common clock framework. Modify the remaining low level code to be compatible with clock provider API and register the clocks with CCF. Move initialisation of clocks to omap1_timer_init() to avoid memory allocation issues at early setup phase from where omap1_init_early() is called. Register the clocks after initialization of clock I/O registers, local clock pointers used by OMAP1 clock ops, and local .rate fields of clocks with no local implementation of .recalc ops, so CCF structures are populated with correct data during clock registration. Instead of enabling some of the registered clocks, flag them for CCF as critical. Introduce .is_enabled op using code that verifies hardware status of clock enablement, split out from implementation of .disable_unused op, so the latter is actually called by CCF for not requested but hardware enabled clocks. Add .round_rate ops where missing so .set_rate ops are called by CCF as expected. Since CCF allows parallel execution of .enable/.disable and .set_rate ops, protect registers shared among those groups of ops from concurrent access with spinlocks. Drop local debugfs support in favor of that provided by CCF. v2: flag tc2_ck as CLK_IS_CRITICAL (Aaro) v3: rebase on top of soc/omap1-multiplatform-5.18, - drop no longer needed includes from arch/arm/mach-omap1/io.c Signed-off-by: Janusz Krzysztofik <jmkrzyszt@gmail.com> Acked-by: Tony Lindgren <tony@atomide.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2022-04-10 06:07:57 -07:00
/* provide direct internal access (not via clk API) to some clocks */
struct omap1_clk *api_ck_p, *ck_dpll1_p, *ck_ref_p;
ARM: OMAP1: clock: Convert to CCF OMAP1 still uses its own implementation of standard clock API defined in include/linux/clk.h. Internals of that implementation are not visible outside OMAP1 directory. As a consequence, device drivers are not able to register clocks potentially provided by peripheral devices. Drop OMAP1 implementation of the clock API and enable common clock framework. Modify the remaining low level code to be compatible with clock provider API and register the clocks with CCF. Move initialisation of clocks to omap1_timer_init() to avoid memory allocation issues at early setup phase from where omap1_init_early() is called. Register the clocks after initialization of clock I/O registers, local clock pointers used by OMAP1 clock ops, and local .rate fields of clocks with no local implementation of .recalc ops, so CCF structures are populated with correct data during clock registration. Instead of enabling some of the registered clocks, flag them for CCF as critical. Introduce .is_enabled op using code that verifies hardware status of clock enablement, split out from implementation of .disable_unused op, so the latter is actually called by CCF for not requested but hardware enabled clocks. Add .round_rate ops where missing so .set_rate ops are called by CCF as expected. Since CCF allows parallel execution of .enable/.disable and .set_rate ops, protect registers shared among those groups of ops from concurrent access with spinlocks. Drop local debugfs support in favor of that provided by CCF. v2: flag tc2_ck as CLK_IS_CRITICAL (Aaro) v3: rebase on top of soc/omap1-multiplatform-5.18, - drop no longer needed includes from arch/arm/mach-omap1/io.c Signed-off-by: Janusz Krzysztofik <jmkrzyszt@gmail.com> Acked-by: Tony Lindgren <tony@atomide.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2022-04-10 06:07:57 -07:00
/* protect registeres shared among clk_enable/disable() and clk_set_rate() operations */
static DEFINE_SPINLOCK(arm_ckctl_lock);
static DEFINE_SPINLOCK(arm_idlect2_lock);
static DEFINE_SPINLOCK(mod_conf_ctrl_0_lock);
static DEFINE_SPINLOCK(mod_conf_ctrl_1_lock);
static DEFINE_SPINLOCK(swd_clk_div_ctrl_sel_lock);
/*
* Omap1 specific clock functions
*/
ARM: OMAP1: clock: Convert to CCF OMAP1 still uses its own implementation of standard clock API defined in include/linux/clk.h. Internals of that implementation are not visible outside OMAP1 directory. As a consequence, device drivers are not able to register clocks potentially provided by peripheral devices. Drop OMAP1 implementation of the clock API and enable common clock framework. Modify the remaining low level code to be compatible with clock provider API and register the clocks with CCF. Move initialisation of clocks to omap1_timer_init() to avoid memory allocation issues at early setup phase from where omap1_init_early() is called. Register the clocks after initialization of clock I/O registers, local clock pointers used by OMAP1 clock ops, and local .rate fields of clocks with no local implementation of .recalc ops, so CCF structures are populated with correct data during clock registration. Instead of enabling some of the registered clocks, flag them for CCF as critical. Introduce .is_enabled op using code that verifies hardware status of clock enablement, split out from implementation of .disable_unused op, so the latter is actually called by CCF for not requested but hardware enabled clocks. Add .round_rate ops where missing so .set_rate ops are called by CCF as expected. Since CCF allows parallel execution of .enable/.disable and .set_rate ops, protect registers shared among those groups of ops from concurrent access with spinlocks. Drop local debugfs support in favor of that provided by CCF. v2: flag tc2_ck as CLK_IS_CRITICAL (Aaro) v3: rebase on top of soc/omap1-multiplatform-5.18, - drop no longer needed includes from arch/arm/mach-omap1/io.c Signed-off-by: Janusz Krzysztofik <jmkrzyszt@gmail.com> Acked-by: Tony Lindgren <tony@atomide.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2022-04-10 06:07:57 -07:00
unsigned long omap1_uart_recalc(struct omap1_clk *clk, unsigned long p_rate)
{
unsigned int val = __raw_readl(clk->enable_reg);
return val & 1 << clk->enable_bit ? 48000000 : 12000000;
}
ARM: OMAP1: clock: Convert to CCF OMAP1 still uses its own implementation of standard clock API defined in include/linux/clk.h. Internals of that implementation are not visible outside OMAP1 directory. As a consequence, device drivers are not able to register clocks potentially provided by peripheral devices. Drop OMAP1 implementation of the clock API and enable common clock framework. Modify the remaining low level code to be compatible with clock provider API and register the clocks with CCF. Move initialisation of clocks to omap1_timer_init() to avoid memory allocation issues at early setup phase from where omap1_init_early() is called. Register the clocks after initialization of clock I/O registers, local clock pointers used by OMAP1 clock ops, and local .rate fields of clocks with no local implementation of .recalc ops, so CCF structures are populated with correct data during clock registration. Instead of enabling some of the registered clocks, flag them for CCF as critical. Introduce .is_enabled op using code that verifies hardware status of clock enablement, split out from implementation of .disable_unused op, so the latter is actually called by CCF for not requested but hardware enabled clocks. Add .round_rate ops where missing so .set_rate ops are called by CCF as expected. Since CCF allows parallel execution of .enable/.disable and .set_rate ops, protect registers shared among those groups of ops from concurrent access with spinlocks. Drop local debugfs support in favor of that provided by CCF. v2: flag tc2_ck as CLK_IS_CRITICAL (Aaro) v3: rebase on top of soc/omap1-multiplatform-5.18, - drop no longer needed includes from arch/arm/mach-omap1/io.c Signed-off-by: Janusz Krzysztofik <jmkrzyszt@gmail.com> Acked-by: Tony Lindgren <tony@atomide.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2022-04-10 06:07:57 -07:00
unsigned long omap1_sossi_recalc(struct omap1_clk *clk, unsigned long p_rate)
{
u32 div = omap_readl(MOD_CONF_CTRL_1);
div = (div >> 17) & 0x7;
div++;
ARM: OMAP1: clock: Convert to CCF OMAP1 still uses its own implementation of standard clock API defined in include/linux/clk.h. Internals of that implementation are not visible outside OMAP1 directory. As a consequence, device drivers are not able to register clocks potentially provided by peripheral devices. Drop OMAP1 implementation of the clock API and enable common clock framework. Modify the remaining low level code to be compatible with clock provider API and register the clocks with CCF. Move initialisation of clocks to omap1_timer_init() to avoid memory allocation issues at early setup phase from where omap1_init_early() is called. Register the clocks after initialization of clock I/O registers, local clock pointers used by OMAP1 clock ops, and local .rate fields of clocks with no local implementation of .recalc ops, so CCF structures are populated with correct data during clock registration. Instead of enabling some of the registered clocks, flag them for CCF as critical. Introduce .is_enabled op using code that verifies hardware status of clock enablement, split out from implementation of .disable_unused op, so the latter is actually called by CCF for not requested but hardware enabled clocks. Add .round_rate ops where missing so .set_rate ops are called by CCF as expected. Since CCF allows parallel execution of .enable/.disable and .set_rate ops, protect registers shared among those groups of ops from concurrent access with spinlocks. Drop local debugfs support in favor of that provided by CCF. v2: flag tc2_ck as CLK_IS_CRITICAL (Aaro) v3: rebase on top of soc/omap1-multiplatform-5.18, - drop no longer needed includes from arch/arm/mach-omap1/io.c Signed-off-by: Janusz Krzysztofik <jmkrzyszt@gmail.com> Acked-by: Tony Lindgren <tony@atomide.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2022-04-10 06:07:57 -07:00
return p_rate / div;
}
ARM: OMAP1: clock: Convert to CCF OMAP1 still uses its own implementation of standard clock API defined in include/linux/clk.h. Internals of that implementation are not visible outside OMAP1 directory. As a consequence, device drivers are not able to register clocks potentially provided by peripheral devices. Drop OMAP1 implementation of the clock API and enable common clock framework. Modify the remaining low level code to be compatible with clock provider API and register the clocks with CCF. Move initialisation of clocks to omap1_timer_init() to avoid memory allocation issues at early setup phase from where omap1_init_early() is called. Register the clocks after initialization of clock I/O registers, local clock pointers used by OMAP1 clock ops, and local .rate fields of clocks with no local implementation of .recalc ops, so CCF structures are populated with correct data during clock registration. Instead of enabling some of the registered clocks, flag them for CCF as critical. Introduce .is_enabled op using code that verifies hardware status of clock enablement, split out from implementation of .disable_unused op, so the latter is actually called by CCF for not requested but hardware enabled clocks. Add .round_rate ops where missing so .set_rate ops are called by CCF as expected. Since CCF allows parallel execution of .enable/.disable and .set_rate ops, protect registers shared among those groups of ops from concurrent access with spinlocks. Drop local debugfs support in favor of that provided by CCF. v2: flag tc2_ck as CLK_IS_CRITICAL (Aaro) v3: rebase on top of soc/omap1-multiplatform-5.18, - drop no longer needed includes from arch/arm/mach-omap1/io.c Signed-off-by: Janusz Krzysztofik <jmkrzyszt@gmail.com> Acked-by: Tony Lindgren <tony@atomide.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2022-04-10 06:07:57 -07:00
static void omap1_clk_allow_idle(struct omap1_clk *clk)
{
struct arm_idlect1_clk * iclk = (struct arm_idlect1_clk *)clk;
if (!(clk->flags & CLOCK_IDLE_CONTROL))
return;
if (iclk->no_idle_count > 0 && !(--iclk->no_idle_count))
arm_idlect1_mask |= 1 << iclk->idlect_shift;
}
ARM: OMAP1: clock: Convert to CCF OMAP1 still uses its own implementation of standard clock API defined in include/linux/clk.h. Internals of that implementation are not visible outside OMAP1 directory. As a consequence, device drivers are not able to register clocks potentially provided by peripheral devices. Drop OMAP1 implementation of the clock API and enable common clock framework. Modify the remaining low level code to be compatible with clock provider API and register the clocks with CCF. Move initialisation of clocks to omap1_timer_init() to avoid memory allocation issues at early setup phase from where omap1_init_early() is called. Register the clocks after initialization of clock I/O registers, local clock pointers used by OMAP1 clock ops, and local .rate fields of clocks with no local implementation of .recalc ops, so CCF structures are populated with correct data during clock registration. Instead of enabling some of the registered clocks, flag them for CCF as critical. Introduce .is_enabled op using code that verifies hardware status of clock enablement, split out from implementation of .disable_unused op, so the latter is actually called by CCF for not requested but hardware enabled clocks. Add .round_rate ops where missing so .set_rate ops are called by CCF as expected. Since CCF allows parallel execution of .enable/.disable and .set_rate ops, protect registers shared among those groups of ops from concurrent access with spinlocks. Drop local debugfs support in favor of that provided by CCF. v2: flag tc2_ck as CLK_IS_CRITICAL (Aaro) v3: rebase on top of soc/omap1-multiplatform-5.18, - drop no longer needed includes from arch/arm/mach-omap1/io.c Signed-off-by: Janusz Krzysztofik <jmkrzyszt@gmail.com> Acked-by: Tony Lindgren <tony@atomide.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2022-04-10 06:07:57 -07:00
static void omap1_clk_deny_idle(struct omap1_clk *clk)
{
struct arm_idlect1_clk * iclk = (struct arm_idlect1_clk *)clk;
if (!(clk->flags & CLOCK_IDLE_CONTROL))
return;
if (iclk->no_idle_count++ == 0)
arm_idlect1_mask &= ~(1 << iclk->idlect_shift);
}
static __u16 verify_ckctl_value(__u16 newval)
{
/* This function checks for following limitations set
* by the hardware (all conditions must be true):
* DSPMMU_CK == DSP_CK or DSPMMU_CK == DSP_CK/2
* ARM_CK >= TC_CK
* DSP_CK >= TC_CK
* DSPMMU_CK >= TC_CK
*
* In addition following rules are enforced:
* LCD_CK <= TC_CK
* ARMPER_CK <= TC_CK
*
* However, maximum frequencies are not checked for!
*/
__u8 per_exp;
__u8 lcd_exp;
__u8 arm_exp;
__u8 dsp_exp;
__u8 tc_exp;
__u8 dspmmu_exp;
per_exp = (newval >> CKCTL_PERDIV_OFFSET) & 3;
lcd_exp = (newval >> CKCTL_LCDDIV_OFFSET) & 3;
arm_exp = (newval >> CKCTL_ARMDIV_OFFSET) & 3;
dsp_exp = (newval >> CKCTL_DSPDIV_OFFSET) & 3;
tc_exp = (newval >> CKCTL_TCDIV_OFFSET) & 3;
dspmmu_exp = (newval >> CKCTL_DSPMMUDIV_OFFSET) & 3;
if (dspmmu_exp < dsp_exp)
dspmmu_exp = dsp_exp;
if (dspmmu_exp > dsp_exp+1)
dspmmu_exp = dsp_exp+1;
if (tc_exp < arm_exp)
tc_exp = arm_exp;
if (tc_exp < dspmmu_exp)
tc_exp = dspmmu_exp;
if (tc_exp > lcd_exp)
lcd_exp = tc_exp;
if (tc_exp > per_exp)
per_exp = tc_exp;
newval &= 0xf000;
newval |= per_exp << CKCTL_PERDIV_OFFSET;
newval |= lcd_exp << CKCTL_LCDDIV_OFFSET;
newval |= arm_exp << CKCTL_ARMDIV_OFFSET;
newval |= dsp_exp << CKCTL_DSPDIV_OFFSET;
newval |= tc_exp << CKCTL_TCDIV_OFFSET;
newval |= dspmmu_exp << CKCTL_DSPMMUDIV_OFFSET;
return newval;
}
ARM: OMAP1: clock: Convert to CCF OMAP1 still uses its own implementation of standard clock API defined in include/linux/clk.h. Internals of that implementation are not visible outside OMAP1 directory. As a consequence, device drivers are not able to register clocks potentially provided by peripheral devices. Drop OMAP1 implementation of the clock API and enable common clock framework. Modify the remaining low level code to be compatible with clock provider API and register the clocks with CCF. Move initialisation of clocks to omap1_timer_init() to avoid memory allocation issues at early setup phase from where omap1_init_early() is called. Register the clocks after initialization of clock I/O registers, local clock pointers used by OMAP1 clock ops, and local .rate fields of clocks with no local implementation of .recalc ops, so CCF structures are populated with correct data during clock registration. Instead of enabling some of the registered clocks, flag them for CCF as critical. Introduce .is_enabled op using code that verifies hardware status of clock enablement, split out from implementation of .disable_unused op, so the latter is actually called by CCF for not requested but hardware enabled clocks. Add .round_rate ops where missing so .set_rate ops are called by CCF as expected. Since CCF allows parallel execution of .enable/.disable and .set_rate ops, protect registers shared among those groups of ops from concurrent access with spinlocks. Drop local debugfs support in favor of that provided by CCF. v2: flag tc2_ck as CLK_IS_CRITICAL (Aaro) v3: rebase on top of soc/omap1-multiplatform-5.18, - drop no longer needed includes from arch/arm/mach-omap1/io.c Signed-off-by: Janusz Krzysztofik <jmkrzyszt@gmail.com> Acked-by: Tony Lindgren <tony@atomide.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2022-04-10 06:07:57 -07:00
static int calc_dsor_exp(unsigned long rate, unsigned long realrate)
{
/* Note: If target frequency is too low, this function will return 4,
* which is invalid value. Caller must check for this value and act
* accordingly.
*
* Note: This function does not check for following limitations set
* by the hardware (all conditions must be true):
* DSPMMU_CK == DSP_CK or DSPMMU_CK == DSP_CK/2
* ARM_CK >= TC_CK
* DSP_CK >= TC_CK
* DSPMMU_CK >= TC_CK
*/
unsigned dsor_exp;
ARM: OMAP1: clock: Convert to CCF OMAP1 still uses its own implementation of standard clock API defined in include/linux/clk.h. Internals of that implementation are not visible outside OMAP1 directory. As a consequence, device drivers are not able to register clocks potentially provided by peripheral devices. Drop OMAP1 implementation of the clock API and enable common clock framework. Modify the remaining low level code to be compatible with clock provider API and register the clocks with CCF. Move initialisation of clocks to omap1_timer_init() to avoid memory allocation issues at early setup phase from where omap1_init_early() is called. Register the clocks after initialization of clock I/O registers, local clock pointers used by OMAP1 clock ops, and local .rate fields of clocks with no local implementation of .recalc ops, so CCF structures are populated with correct data during clock registration. Instead of enabling some of the registered clocks, flag them for CCF as critical. Introduce .is_enabled op using code that verifies hardware status of clock enablement, split out from implementation of .disable_unused op, so the latter is actually called by CCF for not requested but hardware enabled clocks. Add .round_rate ops where missing so .set_rate ops are called by CCF as expected. Since CCF allows parallel execution of .enable/.disable and .set_rate ops, protect registers shared among those groups of ops from concurrent access with spinlocks. Drop local debugfs support in favor of that provided by CCF. v2: flag tc2_ck as CLK_IS_CRITICAL (Aaro) v3: rebase on top of soc/omap1-multiplatform-5.18, - drop no longer needed includes from arch/arm/mach-omap1/io.c Signed-off-by: Janusz Krzysztofik <jmkrzyszt@gmail.com> Acked-by: Tony Lindgren <tony@atomide.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2022-04-10 06:07:57 -07:00
if (unlikely(realrate == 0))
return -EIO;
for (dsor_exp=0; dsor_exp<4; dsor_exp++) {
if (realrate <= rate)
break;
realrate /= 2;
}
return dsor_exp;
}
ARM: OMAP1: clock: Convert to CCF OMAP1 still uses its own implementation of standard clock API defined in include/linux/clk.h. Internals of that implementation are not visible outside OMAP1 directory. As a consequence, device drivers are not able to register clocks potentially provided by peripheral devices. Drop OMAP1 implementation of the clock API and enable common clock framework. Modify the remaining low level code to be compatible with clock provider API and register the clocks with CCF. Move initialisation of clocks to omap1_timer_init() to avoid memory allocation issues at early setup phase from where omap1_init_early() is called. Register the clocks after initialization of clock I/O registers, local clock pointers used by OMAP1 clock ops, and local .rate fields of clocks with no local implementation of .recalc ops, so CCF structures are populated with correct data during clock registration. Instead of enabling some of the registered clocks, flag them for CCF as critical. Introduce .is_enabled op using code that verifies hardware status of clock enablement, split out from implementation of .disable_unused op, so the latter is actually called by CCF for not requested but hardware enabled clocks. Add .round_rate ops where missing so .set_rate ops are called by CCF as expected. Since CCF allows parallel execution of .enable/.disable and .set_rate ops, protect registers shared among those groups of ops from concurrent access with spinlocks. Drop local debugfs support in favor of that provided by CCF. v2: flag tc2_ck as CLK_IS_CRITICAL (Aaro) v3: rebase on top of soc/omap1-multiplatform-5.18, - drop no longer needed includes from arch/arm/mach-omap1/io.c Signed-off-by: Janusz Krzysztofik <jmkrzyszt@gmail.com> Acked-by: Tony Lindgren <tony@atomide.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2022-04-10 06:07:57 -07:00
unsigned long omap1_ckctl_recalc(struct omap1_clk *clk, unsigned long p_rate)
{
/* Calculate divisor encoded as 2-bit exponent */
int dsor = 1 << (3 & (omap_readw(ARM_CKCTL) >> clk->rate_offset));
ARM: OMAP1: clock: Convert to CCF OMAP1 still uses its own implementation of standard clock API defined in include/linux/clk.h. Internals of that implementation are not visible outside OMAP1 directory. As a consequence, device drivers are not able to register clocks potentially provided by peripheral devices. Drop OMAP1 implementation of the clock API and enable common clock framework. Modify the remaining low level code to be compatible with clock provider API and register the clocks with CCF. Move initialisation of clocks to omap1_timer_init() to avoid memory allocation issues at early setup phase from where omap1_init_early() is called. Register the clocks after initialization of clock I/O registers, local clock pointers used by OMAP1 clock ops, and local .rate fields of clocks with no local implementation of .recalc ops, so CCF structures are populated with correct data during clock registration. Instead of enabling some of the registered clocks, flag them for CCF as critical. Introduce .is_enabled op using code that verifies hardware status of clock enablement, split out from implementation of .disable_unused op, so the latter is actually called by CCF for not requested but hardware enabled clocks. Add .round_rate ops where missing so .set_rate ops are called by CCF as expected. Since CCF allows parallel execution of .enable/.disable and .set_rate ops, protect registers shared among those groups of ops from concurrent access with spinlocks. Drop local debugfs support in favor of that provided by CCF. v2: flag tc2_ck as CLK_IS_CRITICAL (Aaro) v3: rebase on top of soc/omap1-multiplatform-5.18, - drop no longer needed includes from arch/arm/mach-omap1/io.c Signed-off-by: Janusz Krzysztofik <jmkrzyszt@gmail.com> Acked-by: Tony Lindgren <tony@atomide.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2022-04-10 06:07:57 -07:00
/* update locally maintained rate, required by arm_ck for omap1_show_rates() */
clk->rate = p_rate / dsor;
return clk->rate;
}
ARM: OMAP1: clock: Convert to CCF OMAP1 still uses its own implementation of standard clock API defined in include/linux/clk.h. Internals of that implementation are not visible outside OMAP1 directory. As a consequence, device drivers are not able to register clocks potentially provided by peripheral devices. Drop OMAP1 implementation of the clock API and enable common clock framework. Modify the remaining low level code to be compatible with clock provider API and register the clocks with CCF. Move initialisation of clocks to omap1_timer_init() to avoid memory allocation issues at early setup phase from where omap1_init_early() is called. Register the clocks after initialization of clock I/O registers, local clock pointers used by OMAP1 clock ops, and local .rate fields of clocks with no local implementation of .recalc ops, so CCF structures are populated with correct data during clock registration. Instead of enabling some of the registered clocks, flag them for CCF as critical. Introduce .is_enabled op using code that verifies hardware status of clock enablement, split out from implementation of .disable_unused op, so the latter is actually called by CCF for not requested but hardware enabled clocks. Add .round_rate ops where missing so .set_rate ops are called by CCF as expected. Since CCF allows parallel execution of .enable/.disable and .set_rate ops, protect registers shared among those groups of ops from concurrent access with spinlocks. Drop local debugfs support in favor of that provided by CCF. v2: flag tc2_ck as CLK_IS_CRITICAL (Aaro) v3: rebase on top of soc/omap1-multiplatform-5.18, - drop no longer needed includes from arch/arm/mach-omap1/io.c Signed-off-by: Janusz Krzysztofik <jmkrzyszt@gmail.com> Acked-by: Tony Lindgren <tony@atomide.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2022-04-10 06:07:57 -07:00
static int omap1_clk_is_enabled(struct clk_hw *hw)
{
ARM: OMAP1: clock: Convert to CCF OMAP1 still uses its own implementation of standard clock API defined in include/linux/clk.h. Internals of that implementation are not visible outside OMAP1 directory. As a consequence, device drivers are not able to register clocks potentially provided by peripheral devices. Drop OMAP1 implementation of the clock API and enable common clock framework. Modify the remaining low level code to be compatible with clock provider API and register the clocks with CCF. Move initialisation of clocks to omap1_timer_init() to avoid memory allocation issues at early setup phase from where omap1_init_early() is called. Register the clocks after initialization of clock I/O registers, local clock pointers used by OMAP1 clock ops, and local .rate fields of clocks with no local implementation of .recalc ops, so CCF structures are populated with correct data during clock registration. Instead of enabling some of the registered clocks, flag them for CCF as critical. Introduce .is_enabled op using code that verifies hardware status of clock enablement, split out from implementation of .disable_unused op, so the latter is actually called by CCF for not requested but hardware enabled clocks. Add .round_rate ops where missing so .set_rate ops are called by CCF as expected. Since CCF allows parallel execution of .enable/.disable and .set_rate ops, protect registers shared among those groups of ops from concurrent access with spinlocks. Drop local debugfs support in favor of that provided by CCF. v2: flag tc2_ck as CLK_IS_CRITICAL (Aaro) v3: rebase on top of soc/omap1-multiplatform-5.18, - drop no longer needed includes from arch/arm/mach-omap1/io.c Signed-off-by: Janusz Krzysztofik <jmkrzyszt@gmail.com> Acked-by: Tony Lindgren <tony@atomide.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2022-04-10 06:07:57 -07:00
struct omap1_clk *clk = to_omap1_clk(hw);
bool api_ck_was_enabled = true;
__u32 regval32;
int ret;
if (!clk->ops) /* no gate -- always enabled */
return 1;
if (clk->ops == &clkops_dspck) {
api_ck_was_enabled = omap1_clk_is_enabled(&api_ck_p->hw);
if (!api_ck_was_enabled)
if (api_ck_p->ops->enable(api_ck_p) < 0)
return 0;
}
if (clk->flags & ENABLE_REG_32BIT)
regval32 = __raw_readl(clk->enable_reg);
else
regval32 = __raw_readw(clk->enable_reg);
ret = regval32 & (1 << clk->enable_bit);
if (!api_ck_was_enabled)
api_ck_p->ops->disable(api_ck_p);
return ret;
}
unsigned long omap1_ckctl_recalc_dsp_domain(struct omap1_clk *clk, unsigned long p_rate)
{
bool api_ck_was_enabled;
int dsor;
/* Calculate divisor encoded as 2-bit exponent
*
* The clock control bits are in DSP domain,
* so api_ck is needed for access.
* Note that DSP_CKCTL virt addr = phys addr, so
* we must use __raw_readw() instead of omap_readw().
*/
ARM: OMAP1: clock: Convert to CCF OMAP1 still uses its own implementation of standard clock API defined in include/linux/clk.h. Internals of that implementation are not visible outside OMAP1 directory. As a consequence, device drivers are not able to register clocks potentially provided by peripheral devices. Drop OMAP1 implementation of the clock API and enable common clock framework. Modify the remaining low level code to be compatible with clock provider API and register the clocks with CCF. Move initialisation of clocks to omap1_timer_init() to avoid memory allocation issues at early setup phase from where omap1_init_early() is called. Register the clocks after initialization of clock I/O registers, local clock pointers used by OMAP1 clock ops, and local .rate fields of clocks with no local implementation of .recalc ops, so CCF structures are populated with correct data during clock registration. Instead of enabling some of the registered clocks, flag them for CCF as critical. Introduce .is_enabled op using code that verifies hardware status of clock enablement, split out from implementation of .disable_unused op, so the latter is actually called by CCF for not requested but hardware enabled clocks. Add .round_rate ops where missing so .set_rate ops are called by CCF as expected. Since CCF allows parallel execution of .enable/.disable and .set_rate ops, protect registers shared among those groups of ops from concurrent access with spinlocks. Drop local debugfs support in favor of that provided by CCF. v2: flag tc2_ck as CLK_IS_CRITICAL (Aaro) v3: rebase on top of soc/omap1-multiplatform-5.18, - drop no longer needed includes from arch/arm/mach-omap1/io.c Signed-off-by: Janusz Krzysztofik <jmkrzyszt@gmail.com> Acked-by: Tony Lindgren <tony@atomide.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2022-04-10 06:07:57 -07:00
api_ck_was_enabled = omap1_clk_is_enabled(&api_ck_p->hw);
if (!api_ck_was_enabled)
api_ck_p->ops->enable(api_ck_p);
dsor = 1 << (3 & (__raw_readw(DSP_CKCTL) >> clk->rate_offset));
ARM: OMAP1: clock: Convert to CCF OMAP1 still uses its own implementation of standard clock API defined in include/linux/clk.h. Internals of that implementation are not visible outside OMAP1 directory. As a consequence, device drivers are not able to register clocks potentially provided by peripheral devices. Drop OMAP1 implementation of the clock API and enable common clock framework. Modify the remaining low level code to be compatible with clock provider API and register the clocks with CCF. Move initialisation of clocks to omap1_timer_init() to avoid memory allocation issues at early setup phase from where omap1_init_early() is called. Register the clocks after initialization of clock I/O registers, local clock pointers used by OMAP1 clock ops, and local .rate fields of clocks with no local implementation of .recalc ops, so CCF structures are populated with correct data during clock registration. Instead of enabling some of the registered clocks, flag them for CCF as critical. Introduce .is_enabled op using code that verifies hardware status of clock enablement, split out from implementation of .disable_unused op, so the latter is actually called by CCF for not requested but hardware enabled clocks. Add .round_rate ops where missing so .set_rate ops are called by CCF as expected. Since CCF allows parallel execution of .enable/.disable and .set_rate ops, protect registers shared among those groups of ops from concurrent access with spinlocks. Drop local debugfs support in favor of that provided by CCF. v2: flag tc2_ck as CLK_IS_CRITICAL (Aaro) v3: rebase on top of soc/omap1-multiplatform-5.18, - drop no longer needed includes from arch/arm/mach-omap1/io.c Signed-off-by: Janusz Krzysztofik <jmkrzyszt@gmail.com> Acked-by: Tony Lindgren <tony@atomide.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2022-04-10 06:07:57 -07:00
if (!api_ck_was_enabled)
api_ck_p->ops->disable(api_ck_p);
ARM: OMAP1: clock: Convert to CCF OMAP1 still uses its own implementation of standard clock API defined in include/linux/clk.h. Internals of that implementation are not visible outside OMAP1 directory. As a consequence, device drivers are not able to register clocks potentially provided by peripheral devices. Drop OMAP1 implementation of the clock API and enable common clock framework. Modify the remaining low level code to be compatible with clock provider API and register the clocks with CCF. Move initialisation of clocks to omap1_timer_init() to avoid memory allocation issues at early setup phase from where omap1_init_early() is called. Register the clocks after initialization of clock I/O registers, local clock pointers used by OMAP1 clock ops, and local .rate fields of clocks with no local implementation of .recalc ops, so CCF structures are populated with correct data during clock registration. Instead of enabling some of the registered clocks, flag them for CCF as critical. Introduce .is_enabled op using code that verifies hardware status of clock enablement, split out from implementation of .disable_unused op, so the latter is actually called by CCF for not requested but hardware enabled clocks. Add .round_rate ops where missing so .set_rate ops are called by CCF as expected. Since CCF allows parallel execution of .enable/.disable and .set_rate ops, protect registers shared among those groups of ops from concurrent access with spinlocks. Drop local debugfs support in favor of that provided by CCF. v2: flag tc2_ck as CLK_IS_CRITICAL (Aaro) v3: rebase on top of soc/omap1-multiplatform-5.18, - drop no longer needed includes from arch/arm/mach-omap1/io.c Signed-off-by: Janusz Krzysztofik <jmkrzyszt@gmail.com> Acked-by: Tony Lindgren <tony@atomide.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2022-04-10 06:07:57 -07:00
return p_rate / dsor;
}
/* MPU virtual clock functions */
ARM: OMAP1: clock: Convert to CCF OMAP1 still uses its own implementation of standard clock API defined in include/linux/clk.h. Internals of that implementation are not visible outside OMAP1 directory. As a consequence, device drivers are not able to register clocks potentially provided by peripheral devices. Drop OMAP1 implementation of the clock API and enable common clock framework. Modify the remaining low level code to be compatible with clock provider API and register the clocks with CCF. Move initialisation of clocks to omap1_timer_init() to avoid memory allocation issues at early setup phase from where omap1_init_early() is called. Register the clocks after initialization of clock I/O registers, local clock pointers used by OMAP1 clock ops, and local .rate fields of clocks with no local implementation of .recalc ops, so CCF structures are populated with correct data during clock registration. Instead of enabling some of the registered clocks, flag them for CCF as critical. Introduce .is_enabled op using code that verifies hardware status of clock enablement, split out from implementation of .disable_unused op, so the latter is actually called by CCF for not requested but hardware enabled clocks. Add .round_rate ops where missing so .set_rate ops are called by CCF as expected. Since CCF allows parallel execution of .enable/.disable and .set_rate ops, protect registers shared among those groups of ops from concurrent access with spinlocks. Drop local debugfs support in favor of that provided by CCF. v2: flag tc2_ck as CLK_IS_CRITICAL (Aaro) v3: rebase on top of soc/omap1-multiplatform-5.18, - drop no longer needed includes from arch/arm/mach-omap1/io.c Signed-off-by: Janusz Krzysztofik <jmkrzyszt@gmail.com> Acked-by: Tony Lindgren <tony@atomide.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2022-04-10 06:07:57 -07:00
int omap1_select_table_rate(struct omap1_clk *clk, unsigned long rate, unsigned long p_rate)
{
/* Find the highest supported frequency <= rate and switch to it */
struct mpu_rate * ptr;
ARM: OMAP2+: clean up some cppcheck warnings Resolve some warnings identified by cppcheck in arch/arm/mach-omap2: [arch/arm/mach-omap2/usb-tusb6010.c:129]: (style) Checking if unsigned variable 'tmp' is less than zero. [arch/arm/mach-omap2/prm_common.c:241]: (error) Possible null pointer dereference: irq_setup - otherwise it is redundant to check if irq_setup is null at line 247 [arch/arm/mach-omap2/pm34xx.c:790]: (style) Variable 'per_clkdm' is assigned a value that is never used [arch/arm/mach-omap2/pm34xx.c:790]: (style) Variable 'core_clkdm' is assigned a value that is never used [arch/arm/mach-omap2/pm24xx.c:185]: (style) Variable 'only_idle' is assigned a value that is never used [arch/arm/mach-omap2/mux.c:254]: (error) Possible null pointer dereference: mux [arch/arm/mach-omap2/mux.c:258]: (error) Possible null pointer dereference: mux [arch/arm/mach-omap2/gpmc-onenand.c:178]: (style) Variable 'tick_ns' is assigned a value that is never used [arch/arm/mach-omap2/gpio.c:56]: (error) Possible null pointer dereference: pdata - otherwise it is redundant to check if pdata is null at line 57 [arch/arm/mach-omap2/devices.c:45]: (style) Variable 'l' is assigned a value that is never used [arch/arm/mach-omap2/board-omap3evm.c:641] -> [arch/arm/mach-omap2/board-omap3evm.c:639]: (style) Found duplicate branches for if and else. [arch/arm/mach-omap2/am35xx-emac.c:95]: (style) Variable 'regval' is assigned a value that is never used [arch/arm/mach-omap2/devices.c:74]: (style) Variable 'l' is assigned a value that is never used [arch/arm/mach-omap2/pm34xx.c:277]: (style) Variable 'per_prev_state' is assigned a value that is never used [arch/arm/plat-omap/dmtimer.c:352]: (error) Possible null pointer dereference: timer - otherwise it is redundant to check if timer is null at line 354 [arch/arm/plat-omap/omap_device.c:478]: (style) Variable 'c' is assigned a value that is never used [arch/arm/plat-omap/usb.c:42]: (style) Variable 'status' is assigned a value that is never used [arch/arm/mach-omap1/clock.c:197]: (style) Variable 'dpll1_rate' is assigned a value that is never used [arch/arm/mach-omap1/lcd_dma.c:60]: (style) struct or union member 'lcd_dma_info::size' is never used [arch/arm/mach-omap1/pm.c:572]: (style) Variable 'entry' is assigned a value that is never used Some of them are pretty good catches, such as gpio.c:56 and usb-tusb6010.c:129. Thanks to Jarkko Nikula for some comments on the sscanf() warnings. It seems that the kernel sscanf() ignores the field width anyway for the %d format, so those changes have been dropped from this second version. Thanks to Daniel Marjamäki <daniel.marjamaki@gmail.com> for pointing out that a variable was unnecessarily marked static in the board-omap3evm.c change. Signed-off-by: Paul Walmsley <paul@pwsan.com> Cc: Felipe Balbi <balbi@ti.com> Cc: Tony Lindgren <tony@atomide.com> Cc: Kevin Hilman <khilman@ti.com> Cc: Peter Ujfalusi <peter.ujfalusi@ti.com> Cc: Jarkko Nikula <jarkko.nikula@bitmer.com> Cc: Charulatha Varadarajan <charu@ti.com> Cc: Daniel Marjamäki <daniel.marjamaki@gmail.com> Cc: Tarun Kanti DebBarma <tarun.kanti@ti.com> Reviewed-by: Charulatha Varadarajan <charu@ti.com> # for gpio.c
2012-04-13 05:34:32 -07:00
unsigned long ref_rate;
ref_rate = ck_ref_p->rate;
for (ptr = omap1_rate_table; ptr->rate; ptr++) {
if (!(ptr->flags & cpu_mask))
continue;
if (ptr->xtal != ref_rate)
continue;
/* Can check only after xtal frequency check */
if (ptr->rate <= rate)
break;
}
if (!ptr->rate)
return -EINVAL;
/*
* In most cases we should not need to reprogram DPLL.
* Reprogramming the DPLL is tricky, it must be done from SRAM.
*/
omap_sram_reprogram_clock(ptr->dpllctl_val, ptr->ckctl_val);
/* XXX Do we need to recalculate the tree below DPLL1 at this point? */
ck_dpll1_p->rate = ptr->pll_rate;
return 0;
}
ARM: OMAP1: clock: Convert to CCF OMAP1 still uses its own implementation of standard clock API defined in include/linux/clk.h. Internals of that implementation are not visible outside OMAP1 directory. As a consequence, device drivers are not able to register clocks potentially provided by peripheral devices. Drop OMAP1 implementation of the clock API and enable common clock framework. Modify the remaining low level code to be compatible with clock provider API and register the clocks with CCF. Move initialisation of clocks to omap1_timer_init() to avoid memory allocation issues at early setup phase from where omap1_init_early() is called. Register the clocks after initialization of clock I/O registers, local clock pointers used by OMAP1 clock ops, and local .rate fields of clocks with no local implementation of .recalc ops, so CCF structures are populated with correct data during clock registration. Instead of enabling some of the registered clocks, flag them for CCF as critical. Introduce .is_enabled op using code that verifies hardware status of clock enablement, split out from implementation of .disable_unused op, so the latter is actually called by CCF for not requested but hardware enabled clocks. Add .round_rate ops where missing so .set_rate ops are called by CCF as expected. Since CCF allows parallel execution of .enable/.disable and .set_rate ops, protect registers shared among those groups of ops from concurrent access with spinlocks. Drop local debugfs support in favor of that provided by CCF. v2: flag tc2_ck as CLK_IS_CRITICAL (Aaro) v3: rebase on top of soc/omap1-multiplatform-5.18, - drop no longer needed includes from arch/arm/mach-omap1/io.c Signed-off-by: Janusz Krzysztofik <jmkrzyszt@gmail.com> Acked-by: Tony Lindgren <tony@atomide.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2022-04-10 06:07:57 -07:00
int omap1_clk_set_rate_dsp_domain(struct omap1_clk *clk, unsigned long rate, unsigned long p_rate)
{
int dsor_exp;
u16 regval;
ARM: OMAP1: clock: Convert to CCF OMAP1 still uses its own implementation of standard clock API defined in include/linux/clk.h. Internals of that implementation are not visible outside OMAP1 directory. As a consequence, device drivers are not able to register clocks potentially provided by peripheral devices. Drop OMAP1 implementation of the clock API and enable common clock framework. Modify the remaining low level code to be compatible with clock provider API and register the clocks with CCF. Move initialisation of clocks to omap1_timer_init() to avoid memory allocation issues at early setup phase from where omap1_init_early() is called. Register the clocks after initialization of clock I/O registers, local clock pointers used by OMAP1 clock ops, and local .rate fields of clocks with no local implementation of .recalc ops, so CCF structures are populated with correct data during clock registration. Instead of enabling some of the registered clocks, flag them for CCF as critical. Introduce .is_enabled op using code that verifies hardware status of clock enablement, split out from implementation of .disable_unused op, so the latter is actually called by CCF for not requested but hardware enabled clocks. Add .round_rate ops where missing so .set_rate ops are called by CCF as expected. Since CCF allows parallel execution of .enable/.disable and .set_rate ops, protect registers shared among those groups of ops from concurrent access with spinlocks. Drop local debugfs support in favor of that provided by CCF. v2: flag tc2_ck as CLK_IS_CRITICAL (Aaro) v3: rebase on top of soc/omap1-multiplatform-5.18, - drop no longer needed includes from arch/arm/mach-omap1/io.c Signed-off-by: Janusz Krzysztofik <jmkrzyszt@gmail.com> Acked-by: Tony Lindgren <tony@atomide.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2022-04-10 06:07:57 -07:00
dsor_exp = calc_dsor_exp(rate, p_rate);
if (dsor_exp > 3)
dsor_exp = -EINVAL;
if (dsor_exp < 0)
return dsor_exp;
regval = __raw_readw(DSP_CKCTL);
regval &= ~(3 << clk->rate_offset);
regval |= dsor_exp << clk->rate_offset;
__raw_writew(regval, DSP_CKCTL);
ARM: OMAP1: clock: Convert to CCF OMAP1 still uses its own implementation of standard clock API defined in include/linux/clk.h. Internals of that implementation are not visible outside OMAP1 directory. As a consequence, device drivers are not able to register clocks potentially provided by peripheral devices. Drop OMAP1 implementation of the clock API and enable common clock framework. Modify the remaining low level code to be compatible with clock provider API and register the clocks with CCF. Move initialisation of clocks to omap1_timer_init() to avoid memory allocation issues at early setup phase from where omap1_init_early() is called. Register the clocks after initialization of clock I/O registers, local clock pointers used by OMAP1 clock ops, and local .rate fields of clocks with no local implementation of .recalc ops, so CCF structures are populated with correct data during clock registration. Instead of enabling some of the registered clocks, flag them for CCF as critical. Introduce .is_enabled op using code that verifies hardware status of clock enablement, split out from implementation of .disable_unused op, so the latter is actually called by CCF for not requested but hardware enabled clocks. Add .round_rate ops where missing so .set_rate ops are called by CCF as expected. Since CCF allows parallel execution of .enable/.disable and .set_rate ops, protect registers shared among those groups of ops from concurrent access with spinlocks. Drop local debugfs support in favor of that provided by CCF. v2: flag tc2_ck as CLK_IS_CRITICAL (Aaro) v3: rebase on top of soc/omap1-multiplatform-5.18, - drop no longer needed includes from arch/arm/mach-omap1/io.c Signed-off-by: Janusz Krzysztofik <jmkrzyszt@gmail.com> Acked-by: Tony Lindgren <tony@atomide.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2022-04-10 06:07:57 -07:00
clk->rate = p_rate / (1 << dsor_exp);
return 0;
}
ARM: OMAP1: clock: Convert to CCF OMAP1 still uses its own implementation of standard clock API defined in include/linux/clk.h. Internals of that implementation are not visible outside OMAP1 directory. As a consequence, device drivers are not able to register clocks potentially provided by peripheral devices. Drop OMAP1 implementation of the clock API and enable common clock framework. Modify the remaining low level code to be compatible with clock provider API and register the clocks with CCF. Move initialisation of clocks to omap1_timer_init() to avoid memory allocation issues at early setup phase from where omap1_init_early() is called. Register the clocks after initialization of clock I/O registers, local clock pointers used by OMAP1 clock ops, and local .rate fields of clocks with no local implementation of .recalc ops, so CCF structures are populated with correct data during clock registration. Instead of enabling some of the registered clocks, flag them for CCF as critical. Introduce .is_enabled op using code that verifies hardware status of clock enablement, split out from implementation of .disable_unused op, so the latter is actually called by CCF for not requested but hardware enabled clocks. Add .round_rate ops where missing so .set_rate ops are called by CCF as expected. Since CCF allows parallel execution of .enable/.disable and .set_rate ops, protect registers shared among those groups of ops from concurrent access with spinlocks. Drop local debugfs support in favor of that provided by CCF. v2: flag tc2_ck as CLK_IS_CRITICAL (Aaro) v3: rebase on top of soc/omap1-multiplatform-5.18, - drop no longer needed includes from arch/arm/mach-omap1/io.c Signed-off-by: Janusz Krzysztofik <jmkrzyszt@gmail.com> Acked-by: Tony Lindgren <tony@atomide.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2022-04-10 06:07:57 -07:00
long omap1_clk_round_rate_ckctl_arm(struct omap1_clk *clk, unsigned long rate,
unsigned long *p_rate)
{
ARM: OMAP1: clock: Convert to CCF OMAP1 still uses its own implementation of standard clock API defined in include/linux/clk.h. Internals of that implementation are not visible outside OMAP1 directory. As a consequence, device drivers are not able to register clocks potentially provided by peripheral devices. Drop OMAP1 implementation of the clock API and enable common clock framework. Modify the remaining low level code to be compatible with clock provider API and register the clocks with CCF. Move initialisation of clocks to omap1_timer_init() to avoid memory allocation issues at early setup phase from where omap1_init_early() is called. Register the clocks after initialization of clock I/O registers, local clock pointers used by OMAP1 clock ops, and local .rate fields of clocks with no local implementation of .recalc ops, so CCF structures are populated with correct data during clock registration. Instead of enabling some of the registered clocks, flag them for CCF as critical. Introduce .is_enabled op using code that verifies hardware status of clock enablement, split out from implementation of .disable_unused op, so the latter is actually called by CCF for not requested but hardware enabled clocks. Add .round_rate ops where missing so .set_rate ops are called by CCF as expected. Since CCF allows parallel execution of .enable/.disable and .set_rate ops, protect registers shared among those groups of ops from concurrent access with spinlocks. Drop local debugfs support in favor of that provided by CCF. v2: flag tc2_ck as CLK_IS_CRITICAL (Aaro) v3: rebase on top of soc/omap1-multiplatform-5.18, - drop no longer needed includes from arch/arm/mach-omap1/io.c Signed-off-by: Janusz Krzysztofik <jmkrzyszt@gmail.com> Acked-by: Tony Lindgren <tony@atomide.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2022-04-10 06:07:57 -07:00
int dsor_exp = calc_dsor_exp(rate, *p_rate);
if (dsor_exp < 0)
return dsor_exp;
if (dsor_exp > 3)
dsor_exp = 3;
ARM: OMAP1: clock: Convert to CCF OMAP1 still uses its own implementation of standard clock API defined in include/linux/clk.h. Internals of that implementation are not visible outside OMAP1 directory. As a consequence, device drivers are not able to register clocks potentially provided by peripheral devices. Drop OMAP1 implementation of the clock API and enable common clock framework. Modify the remaining low level code to be compatible with clock provider API and register the clocks with CCF. Move initialisation of clocks to omap1_timer_init() to avoid memory allocation issues at early setup phase from where omap1_init_early() is called. Register the clocks after initialization of clock I/O registers, local clock pointers used by OMAP1 clock ops, and local .rate fields of clocks with no local implementation of .recalc ops, so CCF structures are populated with correct data during clock registration. Instead of enabling some of the registered clocks, flag them for CCF as critical. Introduce .is_enabled op using code that verifies hardware status of clock enablement, split out from implementation of .disable_unused op, so the latter is actually called by CCF for not requested but hardware enabled clocks. Add .round_rate ops where missing so .set_rate ops are called by CCF as expected. Since CCF allows parallel execution of .enable/.disable and .set_rate ops, protect registers shared among those groups of ops from concurrent access with spinlocks. Drop local debugfs support in favor of that provided by CCF. v2: flag tc2_ck as CLK_IS_CRITICAL (Aaro) v3: rebase on top of soc/omap1-multiplatform-5.18, - drop no longer needed includes from arch/arm/mach-omap1/io.c Signed-off-by: Janusz Krzysztofik <jmkrzyszt@gmail.com> Acked-by: Tony Lindgren <tony@atomide.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2022-04-10 06:07:57 -07:00
return *p_rate / (1 << dsor_exp);
}
ARM: OMAP1: clock: Convert to CCF OMAP1 still uses its own implementation of standard clock API defined in include/linux/clk.h. Internals of that implementation are not visible outside OMAP1 directory. As a consequence, device drivers are not able to register clocks potentially provided by peripheral devices. Drop OMAP1 implementation of the clock API and enable common clock framework. Modify the remaining low level code to be compatible with clock provider API and register the clocks with CCF. Move initialisation of clocks to omap1_timer_init() to avoid memory allocation issues at early setup phase from where omap1_init_early() is called. Register the clocks after initialization of clock I/O registers, local clock pointers used by OMAP1 clock ops, and local .rate fields of clocks with no local implementation of .recalc ops, so CCF structures are populated with correct data during clock registration. Instead of enabling some of the registered clocks, flag them for CCF as critical. Introduce .is_enabled op using code that verifies hardware status of clock enablement, split out from implementation of .disable_unused op, so the latter is actually called by CCF for not requested but hardware enabled clocks. Add .round_rate ops where missing so .set_rate ops are called by CCF as expected. Since CCF allows parallel execution of .enable/.disable and .set_rate ops, protect registers shared among those groups of ops from concurrent access with spinlocks. Drop local debugfs support in favor of that provided by CCF. v2: flag tc2_ck as CLK_IS_CRITICAL (Aaro) v3: rebase on top of soc/omap1-multiplatform-5.18, - drop no longer needed includes from arch/arm/mach-omap1/io.c Signed-off-by: Janusz Krzysztofik <jmkrzyszt@gmail.com> Acked-by: Tony Lindgren <tony@atomide.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2022-04-10 06:07:57 -07:00
int omap1_clk_set_rate_ckctl_arm(struct omap1_clk *clk, unsigned long rate, unsigned long p_rate)
{
ARM: OMAP1: clock: Convert to CCF OMAP1 still uses its own implementation of standard clock API defined in include/linux/clk.h. Internals of that implementation are not visible outside OMAP1 directory. As a consequence, device drivers are not able to register clocks potentially provided by peripheral devices. Drop OMAP1 implementation of the clock API and enable common clock framework. Modify the remaining low level code to be compatible with clock provider API and register the clocks with CCF. Move initialisation of clocks to omap1_timer_init() to avoid memory allocation issues at early setup phase from where omap1_init_early() is called. Register the clocks after initialization of clock I/O registers, local clock pointers used by OMAP1 clock ops, and local .rate fields of clocks with no local implementation of .recalc ops, so CCF structures are populated with correct data during clock registration. Instead of enabling some of the registered clocks, flag them for CCF as critical. Introduce .is_enabled op using code that verifies hardware status of clock enablement, split out from implementation of .disable_unused op, so the latter is actually called by CCF for not requested but hardware enabled clocks. Add .round_rate ops where missing so .set_rate ops are called by CCF as expected. Since CCF allows parallel execution of .enable/.disable and .set_rate ops, protect registers shared among those groups of ops from concurrent access with spinlocks. Drop local debugfs support in favor of that provided by CCF. v2: flag tc2_ck as CLK_IS_CRITICAL (Aaro) v3: rebase on top of soc/omap1-multiplatform-5.18, - drop no longer needed includes from arch/arm/mach-omap1/io.c Signed-off-by: Janusz Krzysztofik <jmkrzyszt@gmail.com> Acked-by: Tony Lindgren <tony@atomide.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2022-04-10 06:07:57 -07:00
unsigned long flags;
int dsor_exp;
u16 regval;
ARM: OMAP1: clock: Convert to CCF OMAP1 still uses its own implementation of standard clock API defined in include/linux/clk.h. Internals of that implementation are not visible outside OMAP1 directory. As a consequence, device drivers are not able to register clocks potentially provided by peripheral devices. Drop OMAP1 implementation of the clock API and enable common clock framework. Modify the remaining low level code to be compatible with clock provider API and register the clocks with CCF. Move initialisation of clocks to omap1_timer_init() to avoid memory allocation issues at early setup phase from where omap1_init_early() is called. Register the clocks after initialization of clock I/O registers, local clock pointers used by OMAP1 clock ops, and local .rate fields of clocks with no local implementation of .recalc ops, so CCF structures are populated with correct data during clock registration. Instead of enabling some of the registered clocks, flag them for CCF as critical. Introduce .is_enabled op using code that verifies hardware status of clock enablement, split out from implementation of .disable_unused op, so the latter is actually called by CCF for not requested but hardware enabled clocks. Add .round_rate ops where missing so .set_rate ops are called by CCF as expected. Since CCF allows parallel execution of .enable/.disable and .set_rate ops, protect registers shared among those groups of ops from concurrent access with spinlocks. Drop local debugfs support in favor of that provided by CCF. v2: flag tc2_ck as CLK_IS_CRITICAL (Aaro) v3: rebase on top of soc/omap1-multiplatform-5.18, - drop no longer needed includes from arch/arm/mach-omap1/io.c Signed-off-by: Janusz Krzysztofik <jmkrzyszt@gmail.com> Acked-by: Tony Lindgren <tony@atomide.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2022-04-10 06:07:57 -07:00
dsor_exp = calc_dsor_exp(rate, p_rate);
if (dsor_exp > 3)
dsor_exp = -EINVAL;
if (dsor_exp < 0)
return dsor_exp;
ARM: OMAP1: clock: Convert to CCF OMAP1 still uses its own implementation of standard clock API defined in include/linux/clk.h. Internals of that implementation are not visible outside OMAP1 directory. As a consequence, device drivers are not able to register clocks potentially provided by peripheral devices. Drop OMAP1 implementation of the clock API and enable common clock framework. Modify the remaining low level code to be compatible with clock provider API and register the clocks with CCF. Move initialisation of clocks to omap1_timer_init() to avoid memory allocation issues at early setup phase from where omap1_init_early() is called. Register the clocks after initialization of clock I/O registers, local clock pointers used by OMAP1 clock ops, and local .rate fields of clocks with no local implementation of .recalc ops, so CCF structures are populated with correct data during clock registration. Instead of enabling some of the registered clocks, flag them for CCF as critical. Introduce .is_enabled op using code that verifies hardware status of clock enablement, split out from implementation of .disable_unused op, so the latter is actually called by CCF for not requested but hardware enabled clocks. Add .round_rate ops where missing so .set_rate ops are called by CCF as expected. Since CCF allows parallel execution of .enable/.disable and .set_rate ops, protect registers shared among those groups of ops from concurrent access with spinlocks. Drop local debugfs support in favor of that provided by CCF. v2: flag tc2_ck as CLK_IS_CRITICAL (Aaro) v3: rebase on top of soc/omap1-multiplatform-5.18, - drop no longer needed includes from arch/arm/mach-omap1/io.c Signed-off-by: Janusz Krzysztofik <jmkrzyszt@gmail.com> Acked-by: Tony Lindgren <tony@atomide.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2022-04-10 06:07:57 -07:00
/* protect ARM_CKCTL register from concurrent access via clk_enable/disable() */
spin_lock_irqsave(&arm_ckctl_lock, flags);
regval = omap_readw(ARM_CKCTL);
regval &= ~(3 << clk->rate_offset);
regval |= dsor_exp << clk->rate_offset;
regval = verify_ckctl_value(regval);
omap_writew(regval, ARM_CKCTL);
ARM: OMAP1: clock: Convert to CCF OMAP1 still uses its own implementation of standard clock API defined in include/linux/clk.h. Internals of that implementation are not visible outside OMAP1 directory. As a consequence, device drivers are not able to register clocks potentially provided by peripheral devices. Drop OMAP1 implementation of the clock API and enable common clock framework. Modify the remaining low level code to be compatible with clock provider API and register the clocks with CCF. Move initialisation of clocks to omap1_timer_init() to avoid memory allocation issues at early setup phase from where omap1_init_early() is called. Register the clocks after initialization of clock I/O registers, local clock pointers used by OMAP1 clock ops, and local .rate fields of clocks with no local implementation of .recalc ops, so CCF structures are populated with correct data during clock registration. Instead of enabling some of the registered clocks, flag them for CCF as critical. Introduce .is_enabled op using code that verifies hardware status of clock enablement, split out from implementation of .disable_unused op, so the latter is actually called by CCF for not requested but hardware enabled clocks. Add .round_rate ops where missing so .set_rate ops are called by CCF as expected. Since CCF allows parallel execution of .enable/.disable and .set_rate ops, protect registers shared among those groups of ops from concurrent access with spinlocks. Drop local debugfs support in favor of that provided by CCF. v2: flag tc2_ck as CLK_IS_CRITICAL (Aaro) v3: rebase on top of soc/omap1-multiplatform-5.18, - drop no longer needed includes from arch/arm/mach-omap1/io.c Signed-off-by: Janusz Krzysztofik <jmkrzyszt@gmail.com> Acked-by: Tony Lindgren <tony@atomide.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2022-04-10 06:07:57 -07:00
clk->rate = p_rate / (1 << dsor_exp);
spin_unlock_irqrestore(&arm_ckctl_lock, flags);
return 0;
}
ARM: OMAP1: clock: Convert to CCF OMAP1 still uses its own implementation of standard clock API defined in include/linux/clk.h. Internals of that implementation are not visible outside OMAP1 directory. As a consequence, device drivers are not able to register clocks potentially provided by peripheral devices. Drop OMAP1 implementation of the clock API and enable common clock framework. Modify the remaining low level code to be compatible with clock provider API and register the clocks with CCF. Move initialisation of clocks to omap1_timer_init() to avoid memory allocation issues at early setup phase from where omap1_init_early() is called. Register the clocks after initialization of clock I/O registers, local clock pointers used by OMAP1 clock ops, and local .rate fields of clocks with no local implementation of .recalc ops, so CCF structures are populated with correct data during clock registration. Instead of enabling some of the registered clocks, flag them for CCF as critical. Introduce .is_enabled op using code that verifies hardware status of clock enablement, split out from implementation of .disable_unused op, so the latter is actually called by CCF for not requested but hardware enabled clocks. Add .round_rate ops where missing so .set_rate ops are called by CCF as expected. Since CCF allows parallel execution of .enable/.disable and .set_rate ops, protect registers shared among those groups of ops from concurrent access with spinlocks. Drop local debugfs support in favor of that provided by CCF. v2: flag tc2_ck as CLK_IS_CRITICAL (Aaro) v3: rebase on top of soc/omap1-multiplatform-5.18, - drop no longer needed includes from arch/arm/mach-omap1/io.c Signed-off-by: Janusz Krzysztofik <jmkrzyszt@gmail.com> Acked-by: Tony Lindgren <tony@atomide.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2022-04-10 06:07:57 -07:00
long omap1_round_to_table_rate(struct omap1_clk *clk, unsigned long rate, unsigned long *p_rate)
{
/* Find the highest supported frequency <= rate */
struct mpu_rate * ptr;
long highest_rate;
unsigned long ref_rate;
ref_rate = ck_ref_p->rate;
highest_rate = -EINVAL;
for (ptr = omap1_rate_table; ptr->rate; ptr++) {
if (!(ptr->flags & cpu_mask))
continue;
if (ptr->xtal != ref_rate)
continue;
highest_rate = ptr->rate;
/* Can check only after xtal frequency check */
if (ptr->rate <= rate)
break;
}
return highest_rate;
}
static unsigned calc_ext_dsor(unsigned long rate)
{
unsigned dsor;
/* MCLK and BCLK divisor selection is not linear:
* freq = 96MHz / dsor
*
* RATIO_SEL range: dsor <-> RATIO_SEL
* 0..6: (RATIO_SEL+2) <-> (dsor-2)
* 6..48: (8+(RATIO_SEL-6)*2) <-> ((dsor-8)/2+6)
* Minimum dsor is 2 and maximum is 96. Odd divisors starting from 9
* can not be used.
*/
for (dsor = 2; dsor < 96; ++dsor) {
if ((dsor & 1) && dsor > 8)
continue;
if (rate >= 96000000 / dsor)
break;
}
return dsor;
}
/* XXX Only needed on 1510 */
ARM: OMAP1: clock: Convert to CCF OMAP1 still uses its own implementation of standard clock API defined in include/linux/clk.h. Internals of that implementation are not visible outside OMAP1 directory. As a consequence, device drivers are not able to register clocks potentially provided by peripheral devices. Drop OMAP1 implementation of the clock API and enable common clock framework. Modify the remaining low level code to be compatible with clock provider API and register the clocks with CCF. Move initialisation of clocks to omap1_timer_init() to avoid memory allocation issues at early setup phase from where omap1_init_early() is called. Register the clocks after initialization of clock I/O registers, local clock pointers used by OMAP1 clock ops, and local .rate fields of clocks with no local implementation of .recalc ops, so CCF structures are populated with correct data during clock registration. Instead of enabling some of the registered clocks, flag them for CCF as critical. Introduce .is_enabled op using code that verifies hardware status of clock enablement, split out from implementation of .disable_unused op, so the latter is actually called by CCF for not requested but hardware enabled clocks. Add .round_rate ops where missing so .set_rate ops are called by CCF as expected. Since CCF allows parallel execution of .enable/.disable and .set_rate ops, protect registers shared among those groups of ops from concurrent access with spinlocks. Drop local debugfs support in favor of that provided by CCF. v2: flag tc2_ck as CLK_IS_CRITICAL (Aaro) v3: rebase on top of soc/omap1-multiplatform-5.18, - drop no longer needed includes from arch/arm/mach-omap1/io.c Signed-off-by: Janusz Krzysztofik <jmkrzyszt@gmail.com> Acked-by: Tony Lindgren <tony@atomide.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2022-04-10 06:07:57 -07:00
long omap1_round_uart_rate(struct omap1_clk *clk, unsigned long rate, unsigned long *p_rate)
{
ARM: OMAP1: clock: Convert to CCF OMAP1 still uses its own implementation of standard clock API defined in include/linux/clk.h. Internals of that implementation are not visible outside OMAP1 directory. As a consequence, device drivers are not able to register clocks potentially provided by peripheral devices. Drop OMAP1 implementation of the clock API and enable common clock framework. Modify the remaining low level code to be compatible with clock provider API and register the clocks with CCF. Move initialisation of clocks to omap1_timer_init() to avoid memory allocation issues at early setup phase from where omap1_init_early() is called. Register the clocks after initialization of clock I/O registers, local clock pointers used by OMAP1 clock ops, and local .rate fields of clocks with no local implementation of .recalc ops, so CCF structures are populated with correct data during clock registration. Instead of enabling some of the registered clocks, flag them for CCF as critical. Introduce .is_enabled op using code that verifies hardware status of clock enablement, split out from implementation of .disable_unused op, so the latter is actually called by CCF for not requested but hardware enabled clocks. Add .round_rate ops where missing so .set_rate ops are called by CCF as expected. Since CCF allows parallel execution of .enable/.disable and .set_rate ops, protect registers shared among those groups of ops from concurrent access with spinlocks. Drop local debugfs support in favor of that provided by CCF. v2: flag tc2_ck as CLK_IS_CRITICAL (Aaro) v3: rebase on top of soc/omap1-multiplatform-5.18, - drop no longer needed includes from arch/arm/mach-omap1/io.c Signed-off-by: Janusz Krzysztofik <jmkrzyszt@gmail.com> Acked-by: Tony Lindgren <tony@atomide.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2022-04-10 06:07:57 -07:00
return rate > 24000000 ? 48000000 : 12000000;
}
int omap1_set_uart_rate(struct omap1_clk *clk, unsigned long rate, unsigned long p_rate)
{
unsigned long flags;
unsigned int val;
if (rate == 12000000)
ARM: OMAP1: clock: Convert to CCF OMAP1 still uses its own implementation of standard clock API defined in include/linux/clk.h. Internals of that implementation are not visible outside OMAP1 directory. As a consequence, device drivers are not able to register clocks potentially provided by peripheral devices. Drop OMAP1 implementation of the clock API and enable common clock framework. Modify the remaining low level code to be compatible with clock provider API and register the clocks with CCF. Move initialisation of clocks to omap1_timer_init() to avoid memory allocation issues at early setup phase from where omap1_init_early() is called. Register the clocks after initialization of clock I/O registers, local clock pointers used by OMAP1 clock ops, and local .rate fields of clocks with no local implementation of .recalc ops, so CCF structures are populated with correct data during clock registration. Instead of enabling some of the registered clocks, flag them for CCF as critical. Introduce .is_enabled op using code that verifies hardware status of clock enablement, split out from implementation of .disable_unused op, so the latter is actually called by CCF for not requested but hardware enabled clocks. Add .round_rate ops where missing so .set_rate ops are called by CCF as expected. Since CCF allows parallel execution of .enable/.disable and .set_rate ops, protect registers shared among those groups of ops from concurrent access with spinlocks. Drop local debugfs support in favor of that provided by CCF. v2: flag tc2_ck as CLK_IS_CRITICAL (Aaro) v3: rebase on top of soc/omap1-multiplatform-5.18, - drop no longer needed includes from arch/arm/mach-omap1/io.c Signed-off-by: Janusz Krzysztofik <jmkrzyszt@gmail.com> Acked-by: Tony Lindgren <tony@atomide.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2022-04-10 06:07:57 -07:00
val = 0;
else if (rate == 48000000)
ARM: OMAP1: clock: Convert to CCF OMAP1 still uses its own implementation of standard clock API defined in include/linux/clk.h. Internals of that implementation are not visible outside OMAP1 directory. As a consequence, device drivers are not able to register clocks potentially provided by peripheral devices. Drop OMAP1 implementation of the clock API and enable common clock framework. Modify the remaining low level code to be compatible with clock provider API and register the clocks with CCF. Move initialisation of clocks to omap1_timer_init() to avoid memory allocation issues at early setup phase from where omap1_init_early() is called. Register the clocks after initialization of clock I/O registers, local clock pointers used by OMAP1 clock ops, and local .rate fields of clocks with no local implementation of .recalc ops, so CCF structures are populated with correct data during clock registration. Instead of enabling some of the registered clocks, flag them for CCF as critical. Introduce .is_enabled op using code that verifies hardware status of clock enablement, split out from implementation of .disable_unused op, so the latter is actually called by CCF for not requested but hardware enabled clocks. Add .round_rate ops where missing so .set_rate ops are called by CCF as expected. Since CCF allows parallel execution of .enable/.disable and .set_rate ops, protect registers shared among those groups of ops from concurrent access with spinlocks. Drop local debugfs support in favor of that provided by CCF. v2: flag tc2_ck as CLK_IS_CRITICAL (Aaro) v3: rebase on top of soc/omap1-multiplatform-5.18, - drop no longer needed includes from arch/arm/mach-omap1/io.c Signed-off-by: Janusz Krzysztofik <jmkrzyszt@gmail.com> Acked-by: Tony Lindgren <tony@atomide.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2022-04-10 06:07:57 -07:00
val = 1 << clk->enable_bit;
else
return -EINVAL;
ARM: OMAP1: clock: Convert to CCF OMAP1 still uses its own implementation of standard clock API defined in include/linux/clk.h. Internals of that implementation are not visible outside OMAP1 directory. As a consequence, device drivers are not able to register clocks potentially provided by peripheral devices. Drop OMAP1 implementation of the clock API and enable common clock framework. Modify the remaining low level code to be compatible with clock provider API and register the clocks with CCF. Move initialisation of clocks to omap1_timer_init() to avoid memory allocation issues at early setup phase from where omap1_init_early() is called. Register the clocks after initialization of clock I/O registers, local clock pointers used by OMAP1 clock ops, and local .rate fields of clocks with no local implementation of .recalc ops, so CCF structures are populated with correct data during clock registration. Instead of enabling some of the registered clocks, flag them for CCF as critical. Introduce .is_enabled op using code that verifies hardware status of clock enablement, split out from implementation of .disable_unused op, so the latter is actually called by CCF for not requested but hardware enabled clocks. Add .round_rate ops where missing so .set_rate ops are called by CCF as expected. Since CCF allows parallel execution of .enable/.disable and .set_rate ops, protect registers shared among those groups of ops from concurrent access with spinlocks. Drop local debugfs support in favor of that provided by CCF. v2: flag tc2_ck as CLK_IS_CRITICAL (Aaro) v3: rebase on top of soc/omap1-multiplatform-5.18, - drop no longer needed includes from arch/arm/mach-omap1/io.c Signed-off-by: Janusz Krzysztofik <jmkrzyszt@gmail.com> Acked-by: Tony Lindgren <tony@atomide.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2022-04-10 06:07:57 -07:00
/* protect MOD_CONF_CTRL_0 register from concurrent access via clk_enable/disable() */
spin_lock_irqsave(&mod_conf_ctrl_0_lock, flags);
val |= __raw_readl(clk->enable_reg) & ~(1 << clk->enable_bit);
__raw_writel(val, clk->enable_reg);
ARM: OMAP1: clock: Convert to CCF OMAP1 still uses its own implementation of standard clock API defined in include/linux/clk.h. Internals of that implementation are not visible outside OMAP1 directory. As a consequence, device drivers are not able to register clocks potentially provided by peripheral devices. Drop OMAP1 implementation of the clock API and enable common clock framework. Modify the remaining low level code to be compatible with clock provider API and register the clocks with CCF. Move initialisation of clocks to omap1_timer_init() to avoid memory allocation issues at early setup phase from where omap1_init_early() is called. Register the clocks after initialization of clock I/O registers, local clock pointers used by OMAP1 clock ops, and local .rate fields of clocks with no local implementation of .recalc ops, so CCF structures are populated with correct data during clock registration. Instead of enabling some of the registered clocks, flag them for CCF as critical. Introduce .is_enabled op using code that verifies hardware status of clock enablement, split out from implementation of .disable_unused op, so the latter is actually called by CCF for not requested but hardware enabled clocks. Add .round_rate ops where missing so .set_rate ops are called by CCF as expected. Since CCF allows parallel execution of .enable/.disable and .set_rate ops, protect registers shared among those groups of ops from concurrent access with spinlocks. Drop local debugfs support in favor of that provided by CCF. v2: flag tc2_ck as CLK_IS_CRITICAL (Aaro) v3: rebase on top of soc/omap1-multiplatform-5.18, - drop no longer needed includes from arch/arm/mach-omap1/io.c Signed-off-by: Janusz Krzysztofik <jmkrzyszt@gmail.com> Acked-by: Tony Lindgren <tony@atomide.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2022-04-10 06:07:57 -07:00
spin_unlock_irqrestore(&mod_conf_ctrl_0_lock, flags);
clk->rate = rate;
return 0;
}
/* External clock (MCLK & BCLK) functions */
ARM: OMAP1: clock: Convert to CCF OMAP1 still uses its own implementation of standard clock API defined in include/linux/clk.h. Internals of that implementation are not visible outside OMAP1 directory. As a consequence, device drivers are not able to register clocks potentially provided by peripheral devices. Drop OMAP1 implementation of the clock API and enable common clock framework. Modify the remaining low level code to be compatible with clock provider API and register the clocks with CCF. Move initialisation of clocks to omap1_timer_init() to avoid memory allocation issues at early setup phase from where omap1_init_early() is called. Register the clocks after initialization of clock I/O registers, local clock pointers used by OMAP1 clock ops, and local .rate fields of clocks with no local implementation of .recalc ops, so CCF structures are populated with correct data during clock registration. Instead of enabling some of the registered clocks, flag them for CCF as critical. Introduce .is_enabled op using code that verifies hardware status of clock enablement, split out from implementation of .disable_unused op, so the latter is actually called by CCF for not requested but hardware enabled clocks. Add .round_rate ops where missing so .set_rate ops are called by CCF as expected. Since CCF allows parallel execution of .enable/.disable and .set_rate ops, protect registers shared among those groups of ops from concurrent access with spinlocks. Drop local debugfs support in favor of that provided by CCF. v2: flag tc2_ck as CLK_IS_CRITICAL (Aaro) v3: rebase on top of soc/omap1-multiplatform-5.18, - drop no longer needed includes from arch/arm/mach-omap1/io.c Signed-off-by: Janusz Krzysztofik <jmkrzyszt@gmail.com> Acked-by: Tony Lindgren <tony@atomide.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2022-04-10 06:07:57 -07:00
int omap1_set_ext_clk_rate(struct omap1_clk *clk, unsigned long rate, unsigned long p_rate)
{
ARM: OMAP1: clock: Convert to CCF OMAP1 still uses its own implementation of standard clock API defined in include/linux/clk.h. Internals of that implementation are not visible outside OMAP1 directory. As a consequence, device drivers are not able to register clocks potentially provided by peripheral devices. Drop OMAP1 implementation of the clock API and enable common clock framework. Modify the remaining low level code to be compatible with clock provider API and register the clocks with CCF. Move initialisation of clocks to omap1_timer_init() to avoid memory allocation issues at early setup phase from where omap1_init_early() is called. Register the clocks after initialization of clock I/O registers, local clock pointers used by OMAP1 clock ops, and local .rate fields of clocks with no local implementation of .recalc ops, so CCF structures are populated with correct data during clock registration. Instead of enabling some of the registered clocks, flag them for CCF as critical. Introduce .is_enabled op using code that verifies hardware status of clock enablement, split out from implementation of .disable_unused op, so the latter is actually called by CCF for not requested but hardware enabled clocks. Add .round_rate ops where missing so .set_rate ops are called by CCF as expected. Since CCF allows parallel execution of .enable/.disable and .set_rate ops, protect registers shared among those groups of ops from concurrent access with spinlocks. Drop local debugfs support in favor of that provided by CCF. v2: flag tc2_ck as CLK_IS_CRITICAL (Aaro) v3: rebase on top of soc/omap1-multiplatform-5.18, - drop no longer needed includes from arch/arm/mach-omap1/io.c Signed-off-by: Janusz Krzysztofik <jmkrzyszt@gmail.com> Acked-by: Tony Lindgren <tony@atomide.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2022-04-10 06:07:57 -07:00
unsigned long flags;
unsigned dsor;
__u16 ratio_bits;
dsor = calc_ext_dsor(rate);
clk->rate = 96000000 / dsor;
if (dsor > 8)
ratio_bits = ((dsor - 8) / 2 + 6) << 2;
else
ratio_bits = (dsor - 2) << 2;
ARM: OMAP1: clock: Convert to CCF OMAP1 still uses its own implementation of standard clock API defined in include/linux/clk.h. Internals of that implementation are not visible outside OMAP1 directory. As a consequence, device drivers are not able to register clocks potentially provided by peripheral devices. Drop OMAP1 implementation of the clock API and enable common clock framework. Modify the remaining low level code to be compatible with clock provider API and register the clocks with CCF. Move initialisation of clocks to omap1_timer_init() to avoid memory allocation issues at early setup phase from where omap1_init_early() is called. Register the clocks after initialization of clock I/O registers, local clock pointers used by OMAP1 clock ops, and local .rate fields of clocks with no local implementation of .recalc ops, so CCF structures are populated with correct data during clock registration. Instead of enabling some of the registered clocks, flag them for CCF as critical. Introduce .is_enabled op using code that verifies hardware status of clock enablement, split out from implementation of .disable_unused op, so the latter is actually called by CCF for not requested but hardware enabled clocks. Add .round_rate ops where missing so .set_rate ops are called by CCF as expected. Since CCF allows parallel execution of .enable/.disable and .set_rate ops, protect registers shared among those groups of ops from concurrent access with spinlocks. Drop local debugfs support in favor of that provided by CCF. v2: flag tc2_ck as CLK_IS_CRITICAL (Aaro) v3: rebase on top of soc/omap1-multiplatform-5.18, - drop no longer needed includes from arch/arm/mach-omap1/io.c Signed-off-by: Janusz Krzysztofik <jmkrzyszt@gmail.com> Acked-by: Tony Lindgren <tony@atomide.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2022-04-10 06:07:57 -07:00
/* protect SWD_CLK_DIV_CTRL_SEL register from concurrent access via clk_enable/disable() */
spin_lock_irqsave(&swd_clk_div_ctrl_sel_lock, flags);
ratio_bits |= __raw_readw(clk->enable_reg) & ~0xfd;
__raw_writew(ratio_bits, clk->enable_reg);
ARM: OMAP1: clock: Convert to CCF OMAP1 still uses its own implementation of standard clock API defined in include/linux/clk.h. Internals of that implementation are not visible outside OMAP1 directory. As a consequence, device drivers are not able to register clocks potentially provided by peripheral devices. Drop OMAP1 implementation of the clock API and enable common clock framework. Modify the remaining low level code to be compatible with clock provider API and register the clocks with CCF. Move initialisation of clocks to omap1_timer_init() to avoid memory allocation issues at early setup phase from where omap1_init_early() is called. Register the clocks after initialization of clock I/O registers, local clock pointers used by OMAP1 clock ops, and local .rate fields of clocks with no local implementation of .recalc ops, so CCF structures are populated with correct data during clock registration. Instead of enabling some of the registered clocks, flag them for CCF as critical. Introduce .is_enabled op using code that verifies hardware status of clock enablement, split out from implementation of .disable_unused op, so the latter is actually called by CCF for not requested but hardware enabled clocks. Add .round_rate ops where missing so .set_rate ops are called by CCF as expected. Since CCF allows parallel execution of .enable/.disable and .set_rate ops, protect registers shared among those groups of ops from concurrent access with spinlocks. Drop local debugfs support in favor of that provided by CCF. v2: flag tc2_ck as CLK_IS_CRITICAL (Aaro) v3: rebase on top of soc/omap1-multiplatform-5.18, - drop no longer needed includes from arch/arm/mach-omap1/io.c Signed-off-by: Janusz Krzysztofik <jmkrzyszt@gmail.com> Acked-by: Tony Lindgren <tony@atomide.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2022-04-10 06:07:57 -07:00
spin_unlock_irqrestore(&swd_clk_div_ctrl_sel_lock, flags);
return 0;
}
ARM: OMAP1: clock: Convert to CCF OMAP1 still uses its own implementation of standard clock API defined in include/linux/clk.h. Internals of that implementation are not visible outside OMAP1 directory. As a consequence, device drivers are not able to register clocks potentially provided by peripheral devices. Drop OMAP1 implementation of the clock API and enable common clock framework. Modify the remaining low level code to be compatible with clock provider API and register the clocks with CCF. Move initialisation of clocks to omap1_timer_init() to avoid memory allocation issues at early setup phase from where omap1_init_early() is called. Register the clocks after initialization of clock I/O registers, local clock pointers used by OMAP1 clock ops, and local .rate fields of clocks with no local implementation of .recalc ops, so CCF structures are populated with correct data during clock registration. Instead of enabling some of the registered clocks, flag them for CCF as critical. Introduce .is_enabled op using code that verifies hardware status of clock enablement, split out from implementation of .disable_unused op, so the latter is actually called by CCF for not requested but hardware enabled clocks. Add .round_rate ops where missing so .set_rate ops are called by CCF as expected. Since CCF allows parallel execution of .enable/.disable and .set_rate ops, protect registers shared among those groups of ops from concurrent access with spinlocks. Drop local debugfs support in favor of that provided by CCF. v2: flag tc2_ck as CLK_IS_CRITICAL (Aaro) v3: rebase on top of soc/omap1-multiplatform-5.18, - drop no longer needed includes from arch/arm/mach-omap1/io.c Signed-off-by: Janusz Krzysztofik <jmkrzyszt@gmail.com> Acked-by: Tony Lindgren <tony@atomide.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2022-04-10 06:07:57 -07:00
static int calc_div_sossi(unsigned long rate, unsigned long p_rate)
{
int div;
/* Round towards slower frequency */
div = (p_rate + rate - 1) / rate;
ARM: OMAP1: clock: Convert to CCF OMAP1 still uses its own implementation of standard clock API defined in include/linux/clk.h. Internals of that implementation are not visible outside OMAP1 directory. As a consequence, device drivers are not able to register clocks potentially provided by peripheral devices. Drop OMAP1 implementation of the clock API and enable common clock framework. Modify the remaining low level code to be compatible with clock provider API and register the clocks with CCF. Move initialisation of clocks to omap1_timer_init() to avoid memory allocation issues at early setup phase from where omap1_init_early() is called. Register the clocks after initialization of clock I/O registers, local clock pointers used by OMAP1 clock ops, and local .rate fields of clocks with no local implementation of .recalc ops, so CCF structures are populated with correct data during clock registration. Instead of enabling some of the registered clocks, flag them for CCF as critical. Introduce .is_enabled op using code that verifies hardware status of clock enablement, split out from implementation of .disable_unused op, so the latter is actually called by CCF for not requested but hardware enabled clocks. Add .round_rate ops where missing so .set_rate ops are called by CCF as expected. Since CCF allows parallel execution of .enable/.disable and .set_rate ops, protect registers shared among those groups of ops from concurrent access with spinlocks. Drop local debugfs support in favor of that provided by CCF. v2: flag tc2_ck as CLK_IS_CRITICAL (Aaro) v3: rebase on top of soc/omap1-multiplatform-5.18, - drop no longer needed includes from arch/arm/mach-omap1/io.c Signed-off-by: Janusz Krzysztofik <jmkrzyszt@gmail.com> Acked-by: Tony Lindgren <tony@atomide.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2022-04-10 06:07:57 -07:00
return --div;
}
long omap1_round_sossi_rate(struct omap1_clk *clk, unsigned long rate, unsigned long *p_rate)
{
int div;
div = calc_div_sossi(rate, *p_rate);
if (div < 0)
div = 0;
else if (div > 7)
div = 7;
return *p_rate / (div + 1);
}
int omap1_set_sossi_rate(struct omap1_clk *clk, unsigned long rate, unsigned long p_rate)
{
unsigned long flags;
u32 l;
int div;
div = calc_div_sossi(rate, p_rate);
if (div < 0 || div > 7)
return -EINVAL;
ARM: OMAP1: clock: Convert to CCF OMAP1 still uses its own implementation of standard clock API defined in include/linux/clk.h. Internals of that implementation are not visible outside OMAP1 directory. As a consequence, device drivers are not able to register clocks potentially provided by peripheral devices. Drop OMAP1 implementation of the clock API and enable common clock framework. Modify the remaining low level code to be compatible with clock provider API and register the clocks with CCF. Move initialisation of clocks to omap1_timer_init() to avoid memory allocation issues at early setup phase from where omap1_init_early() is called. Register the clocks after initialization of clock I/O registers, local clock pointers used by OMAP1 clock ops, and local .rate fields of clocks with no local implementation of .recalc ops, so CCF structures are populated with correct data during clock registration. Instead of enabling some of the registered clocks, flag them for CCF as critical. Introduce .is_enabled op using code that verifies hardware status of clock enablement, split out from implementation of .disable_unused op, so the latter is actually called by CCF for not requested but hardware enabled clocks. Add .round_rate ops where missing so .set_rate ops are called by CCF as expected. Since CCF allows parallel execution of .enable/.disable and .set_rate ops, protect registers shared among those groups of ops from concurrent access with spinlocks. Drop local debugfs support in favor of that provided by CCF. v2: flag tc2_ck as CLK_IS_CRITICAL (Aaro) v3: rebase on top of soc/omap1-multiplatform-5.18, - drop no longer needed includes from arch/arm/mach-omap1/io.c Signed-off-by: Janusz Krzysztofik <jmkrzyszt@gmail.com> Acked-by: Tony Lindgren <tony@atomide.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2022-04-10 06:07:57 -07:00
/* protect MOD_CONF_CTRL_1 register from concurrent access via clk_enable/disable() */
spin_lock_irqsave(&mod_conf_ctrl_1_lock, flags);
l = omap_readl(MOD_CONF_CTRL_1);
l &= ~(7 << 17);
l |= div << 17;
omap_writel(l, MOD_CONF_CTRL_1);
clk->rate = p_rate / (div + 1);
ARM: OMAP1: clock: Convert to CCF OMAP1 still uses its own implementation of standard clock API defined in include/linux/clk.h. Internals of that implementation are not visible outside OMAP1 directory. As a consequence, device drivers are not able to register clocks potentially provided by peripheral devices. Drop OMAP1 implementation of the clock API and enable common clock framework. Modify the remaining low level code to be compatible with clock provider API and register the clocks with CCF. Move initialisation of clocks to omap1_timer_init() to avoid memory allocation issues at early setup phase from where omap1_init_early() is called. Register the clocks after initialization of clock I/O registers, local clock pointers used by OMAP1 clock ops, and local .rate fields of clocks with no local implementation of .recalc ops, so CCF structures are populated with correct data during clock registration. Instead of enabling some of the registered clocks, flag them for CCF as critical. Introduce .is_enabled op using code that verifies hardware status of clock enablement, split out from implementation of .disable_unused op, so the latter is actually called by CCF for not requested but hardware enabled clocks. Add .round_rate ops where missing so .set_rate ops are called by CCF as expected. Since CCF allows parallel execution of .enable/.disable and .set_rate ops, protect registers shared among those groups of ops from concurrent access with spinlocks. Drop local debugfs support in favor of that provided by CCF. v2: flag tc2_ck as CLK_IS_CRITICAL (Aaro) v3: rebase on top of soc/omap1-multiplatform-5.18, - drop no longer needed includes from arch/arm/mach-omap1/io.c Signed-off-by: Janusz Krzysztofik <jmkrzyszt@gmail.com> Acked-by: Tony Lindgren <tony@atomide.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2022-04-10 06:07:57 -07:00
spin_unlock_irqrestore(&mod_conf_ctrl_1_lock, flags);
return 0;
}
ARM: OMAP1: clock: Convert to CCF OMAP1 still uses its own implementation of standard clock API defined in include/linux/clk.h. Internals of that implementation are not visible outside OMAP1 directory. As a consequence, device drivers are not able to register clocks potentially provided by peripheral devices. Drop OMAP1 implementation of the clock API and enable common clock framework. Modify the remaining low level code to be compatible with clock provider API and register the clocks with CCF. Move initialisation of clocks to omap1_timer_init() to avoid memory allocation issues at early setup phase from where omap1_init_early() is called. Register the clocks after initialization of clock I/O registers, local clock pointers used by OMAP1 clock ops, and local .rate fields of clocks with no local implementation of .recalc ops, so CCF structures are populated with correct data during clock registration. Instead of enabling some of the registered clocks, flag them for CCF as critical. Introduce .is_enabled op using code that verifies hardware status of clock enablement, split out from implementation of .disable_unused op, so the latter is actually called by CCF for not requested but hardware enabled clocks. Add .round_rate ops where missing so .set_rate ops are called by CCF as expected. Since CCF allows parallel execution of .enable/.disable and .set_rate ops, protect registers shared among those groups of ops from concurrent access with spinlocks. Drop local debugfs support in favor of that provided by CCF. v2: flag tc2_ck as CLK_IS_CRITICAL (Aaro) v3: rebase on top of soc/omap1-multiplatform-5.18, - drop no longer needed includes from arch/arm/mach-omap1/io.c Signed-off-by: Janusz Krzysztofik <jmkrzyszt@gmail.com> Acked-by: Tony Lindgren <tony@atomide.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2022-04-10 06:07:57 -07:00
long omap1_round_ext_clk_rate(struct omap1_clk *clk, unsigned long rate, unsigned long *p_rate)
{
return 96000000 / calc_ext_dsor(rate);
}
ARM: OMAP1: clock: Convert to CCF OMAP1 still uses its own implementation of standard clock API defined in include/linux/clk.h. Internals of that implementation are not visible outside OMAP1 directory. As a consequence, device drivers are not able to register clocks potentially provided by peripheral devices. Drop OMAP1 implementation of the clock API and enable common clock framework. Modify the remaining low level code to be compatible with clock provider API and register the clocks with CCF. Move initialisation of clocks to omap1_timer_init() to avoid memory allocation issues at early setup phase from where omap1_init_early() is called. Register the clocks after initialization of clock I/O registers, local clock pointers used by OMAP1 clock ops, and local .rate fields of clocks with no local implementation of .recalc ops, so CCF structures are populated with correct data during clock registration. Instead of enabling some of the registered clocks, flag them for CCF as critical. Introduce .is_enabled op using code that verifies hardware status of clock enablement, split out from implementation of .disable_unused op, so the latter is actually called by CCF for not requested but hardware enabled clocks. Add .round_rate ops where missing so .set_rate ops are called by CCF as expected. Since CCF allows parallel execution of .enable/.disable and .set_rate ops, protect registers shared among those groups of ops from concurrent access with spinlocks. Drop local debugfs support in favor of that provided by CCF. v2: flag tc2_ck as CLK_IS_CRITICAL (Aaro) v3: rebase on top of soc/omap1-multiplatform-5.18, - drop no longer needed includes from arch/arm/mach-omap1/io.c Signed-off-by: Janusz Krzysztofik <jmkrzyszt@gmail.com> Acked-by: Tony Lindgren <tony@atomide.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2022-04-10 06:07:57 -07:00
int omap1_init_ext_clk(struct omap1_clk *clk)
{
unsigned dsor;
__u16 ratio_bits;
/* Determine current rate and ensure clock is based on 96MHz APLL */
ratio_bits = __raw_readw(clk->enable_reg) & ~1;
__raw_writew(ratio_bits, clk->enable_reg);
ratio_bits = (ratio_bits & 0xfc) >> 2;
if (ratio_bits > 6)
dsor = (ratio_bits - 6) * 2 + 8;
else
dsor = ratio_bits + 2;
clk-> rate = 96000000 / dsor;
ARM: OMAP1: clock: Convert to CCF OMAP1 still uses its own implementation of standard clock API defined in include/linux/clk.h. Internals of that implementation are not visible outside OMAP1 directory. As a consequence, device drivers are not able to register clocks potentially provided by peripheral devices. Drop OMAP1 implementation of the clock API and enable common clock framework. Modify the remaining low level code to be compatible with clock provider API and register the clocks with CCF. Move initialisation of clocks to omap1_timer_init() to avoid memory allocation issues at early setup phase from where omap1_init_early() is called. Register the clocks after initialization of clock I/O registers, local clock pointers used by OMAP1 clock ops, and local .rate fields of clocks with no local implementation of .recalc ops, so CCF structures are populated with correct data during clock registration. Instead of enabling some of the registered clocks, flag them for CCF as critical. Introduce .is_enabled op using code that verifies hardware status of clock enablement, split out from implementation of .disable_unused op, so the latter is actually called by CCF for not requested but hardware enabled clocks. Add .round_rate ops where missing so .set_rate ops are called by CCF as expected. Since CCF allows parallel execution of .enable/.disable and .set_rate ops, protect registers shared among those groups of ops from concurrent access with spinlocks. Drop local debugfs support in favor of that provided by CCF. v2: flag tc2_ck as CLK_IS_CRITICAL (Aaro) v3: rebase on top of soc/omap1-multiplatform-5.18, - drop no longer needed includes from arch/arm/mach-omap1/io.c Signed-off-by: Janusz Krzysztofik <jmkrzyszt@gmail.com> Acked-by: Tony Lindgren <tony@atomide.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2022-04-10 06:07:57 -07:00
return 0;
}
ARM: OMAP1: clock: Convert to CCF OMAP1 still uses its own implementation of standard clock API defined in include/linux/clk.h. Internals of that implementation are not visible outside OMAP1 directory. As a consequence, device drivers are not able to register clocks potentially provided by peripheral devices. Drop OMAP1 implementation of the clock API and enable common clock framework. Modify the remaining low level code to be compatible with clock provider API and register the clocks with CCF. Move initialisation of clocks to omap1_timer_init() to avoid memory allocation issues at early setup phase from where omap1_init_early() is called. Register the clocks after initialization of clock I/O registers, local clock pointers used by OMAP1 clock ops, and local .rate fields of clocks with no local implementation of .recalc ops, so CCF structures are populated with correct data during clock registration. Instead of enabling some of the registered clocks, flag them for CCF as critical. Introduce .is_enabled op using code that verifies hardware status of clock enablement, split out from implementation of .disable_unused op, so the latter is actually called by CCF for not requested but hardware enabled clocks. Add .round_rate ops where missing so .set_rate ops are called by CCF as expected. Since CCF allows parallel execution of .enable/.disable and .set_rate ops, protect registers shared among those groups of ops from concurrent access with spinlocks. Drop local debugfs support in favor of that provided by CCF. v2: flag tc2_ck as CLK_IS_CRITICAL (Aaro) v3: rebase on top of soc/omap1-multiplatform-5.18, - drop no longer needed includes from arch/arm/mach-omap1/io.c Signed-off-by: Janusz Krzysztofik <jmkrzyszt@gmail.com> Acked-by: Tony Lindgren <tony@atomide.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2022-04-10 06:07:57 -07:00
static int omap1_clk_enable(struct clk_hw *hw)
{
ARM: OMAP1: clock: Convert to CCF OMAP1 still uses its own implementation of standard clock API defined in include/linux/clk.h. Internals of that implementation are not visible outside OMAP1 directory. As a consequence, device drivers are not able to register clocks potentially provided by peripheral devices. Drop OMAP1 implementation of the clock API and enable common clock framework. Modify the remaining low level code to be compatible with clock provider API and register the clocks with CCF. Move initialisation of clocks to omap1_timer_init() to avoid memory allocation issues at early setup phase from where omap1_init_early() is called. Register the clocks after initialization of clock I/O registers, local clock pointers used by OMAP1 clock ops, and local .rate fields of clocks with no local implementation of .recalc ops, so CCF structures are populated with correct data during clock registration. Instead of enabling some of the registered clocks, flag them for CCF as critical. Introduce .is_enabled op using code that verifies hardware status of clock enablement, split out from implementation of .disable_unused op, so the latter is actually called by CCF for not requested but hardware enabled clocks. Add .round_rate ops where missing so .set_rate ops are called by CCF as expected. Since CCF allows parallel execution of .enable/.disable and .set_rate ops, protect registers shared among those groups of ops from concurrent access with spinlocks. Drop local debugfs support in favor of that provided by CCF. v2: flag tc2_ck as CLK_IS_CRITICAL (Aaro) v3: rebase on top of soc/omap1-multiplatform-5.18, - drop no longer needed includes from arch/arm/mach-omap1/io.c Signed-off-by: Janusz Krzysztofik <jmkrzyszt@gmail.com> Acked-by: Tony Lindgren <tony@atomide.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2022-04-10 06:07:57 -07:00
struct omap1_clk *clk = to_omap1_clk(hw), *parent = to_omap1_clk(clk_hw_get_parent(hw));
int ret = 0;
ARM: OMAP1: clock: Convert to CCF OMAP1 still uses its own implementation of standard clock API defined in include/linux/clk.h. Internals of that implementation are not visible outside OMAP1 directory. As a consequence, device drivers are not able to register clocks potentially provided by peripheral devices. Drop OMAP1 implementation of the clock API and enable common clock framework. Modify the remaining low level code to be compatible with clock provider API and register the clocks with CCF. Move initialisation of clocks to omap1_timer_init() to avoid memory allocation issues at early setup phase from where omap1_init_early() is called. Register the clocks after initialization of clock I/O registers, local clock pointers used by OMAP1 clock ops, and local .rate fields of clocks with no local implementation of .recalc ops, so CCF structures are populated with correct data during clock registration. Instead of enabling some of the registered clocks, flag them for CCF as critical. Introduce .is_enabled op using code that verifies hardware status of clock enablement, split out from implementation of .disable_unused op, so the latter is actually called by CCF for not requested but hardware enabled clocks. Add .round_rate ops where missing so .set_rate ops are called by CCF as expected. Since CCF allows parallel execution of .enable/.disable and .set_rate ops, protect registers shared among those groups of ops from concurrent access with spinlocks. Drop local debugfs support in favor of that provided by CCF. v2: flag tc2_ck as CLK_IS_CRITICAL (Aaro) v3: rebase on top of soc/omap1-multiplatform-5.18, - drop no longer needed includes from arch/arm/mach-omap1/io.c Signed-off-by: Janusz Krzysztofik <jmkrzyszt@gmail.com> Acked-by: Tony Lindgren <tony@atomide.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2022-04-10 06:07:57 -07:00
if (parent && clk->flags & CLOCK_NO_IDLE_PARENT)
omap1_clk_deny_idle(parent);
ARM: OMAP1: clock: Convert to CCF OMAP1 still uses its own implementation of standard clock API defined in include/linux/clk.h. Internals of that implementation are not visible outside OMAP1 directory. As a consequence, device drivers are not able to register clocks potentially provided by peripheral devices. Drop OMAP1 implementation of the clock API and enable common clock framework. Modify the remaining low level code to be compatible with clock provider API and register the clocks with CCF. Move initialisation of clocks to omap1_timer_init() to avoid memory allocation issues at early setup phase from where omap1_init_early() is called. Register the clocks after initialization of clock I/O registers, local clock pointers used by OMAP1 clock ops, and local .rate fields of clocks with no local implementation of .recalc ops, so CCF structures are populated with correct data during clock registration. Instead of enabling some of the registered clocks, flag them for CCF as critical. Introduce .is_enabled op using code that verifies hardware status of clock enablement, split out from implementation of .disable_unused op, so the latter is actually called by CCF for not requested but hardware enabled clocks. Add .round_rate ops where missing so .set_rate ops are called by CCF as expected. Since CCF allows parallel execution of .enable/.disable and .set_rate ops, protect registers shared among those groups of ops from concurrent access with spinlocks. Drop local debugfs support in favor of that provided by CCF. v2: flag tc2_ck as CLK_IS_CRITICAL (Aaro) v3: rebase on top of soc/omap1-multiplatform-5.18, - drop no longer needed includes from arch/arm/mach-omap1/io.c Signed-off-by: Janusz Krzysztofik <jmkrzyszt@gmail.com> Acked-by: Tony Lindgren <tony@atomide.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2022-04-10 06:07:57 -07:00
if (clk->ops && !(WARN_ON(!clk->ops->enable)))
ret = clk->ops->enable(clk);
return ret;
}
ARM: OMAP1: clock: Convert to CCF OMAP1 still uses its own implementation of standard clock API defined in include/linux/clk.h. Internals of that implementation are not visible outside OMAP1 directory. As a consequence, device drivers are not able to register clocks potentially provided by peripheral devices. Drop OMAP1 implementation of the clock API and enable common clock framework. Modify the remaining low level code to be compatible with clock provider API and register the clocks with CCF. Move initialisation of clocks to omap1_timer_init() to avoid memory allocation issues at early setup phase from where omap1_init_early() is called. Register the clocks after initialization of clock I/O registers, local clock pointers used by OMAP1 clock ops, and local .rate fields of clocks with no local implementation of .recalc ops, so CCF structures are populated with correct data during clock registration. Instead of enabling some of the registered clocks, flag them for CCF as critical. Introduce .is_enabled op using code that verifies hardware status of clock enablement, split out from implementation of .disable_unused op, so the latter is actually called by CCF for not requested but hardware enabled clocks. Add .round_rate ops where missing so .set_rate ops are called by CCF as expected. Since CCF allows parallel execution of .enable/.disable and .set_rate ops, protect registers shared among those groups of ops from concurrent access with spinlocks. Drop local debugfs support in favor of that provided by CCF. v2: flag tc2_ck as CLK_IS_CRITICAL (Aaro) v3: rebase on top of soc/omap1-multiplatform-5.18, - drop no longer needed includes from arch/arm/mach-omap1/io.c Signed-off-by: Janusz Krzysztofik <jmkrzyszt@gmail.com> Acked-by: Tony Lindgren <tony@atomide.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2022-04-10 06:07:57 -07:00
static void omap1_clk_disable(struct clk_hw *hw)
{
ARM: OMAP1: clock: Convert to CCF OMAP1 still uses its own implementation of standard clock API defined in include/linux/clk.h. Internals of that implementation are not visible outside OMAP1 directory. As a consequence, device drivers are not able to register clocks potentially provided by peripheral devices. Drop OMAP1 implementation of the clock API and enable common clock framework. Modify the remaining low level code to be compatible with clock provider API and register the clocks with CCF. Move initialisation of clocks to omap1_timer_init() to avoid memory allocation issues at early setup phase from where omap1_init_early() is called. Register the clocks after initialization of clock I/O registers, local clock pointers used by OMAP1 clock ops, and local .rate fields of clocks with no local implementation of .recalc ops, so CCF structures are populated with correct data during clock registration. Instead of enabling some of the registered clocks, flag them for CCF as critical. Introduce .is_enabled op using code that verifies hardware status of clock enablement, split out from implementation of .disable_unused op, so the latter is actually called by CCF for not requested but hardware enabled clocks. Add .round_rate ops where missing so .set_rate ops are called by CCF as expected. Since CCF allows parallel execution of .enable/.disable and .set_rate ops, protect registers shared among those groups of ops from concurrent access with spinlocks. Drop local debugfs support in favor of that provided by CCF. v2: flag tc2_ck as CLK_IS_CRITICAL (Aaro) v3: rebase on top of soc/omap1-multiplatform-5.18, - drop no longer needed includes from arch/arm/mach-omap1/io.c Signed-off-by: Janusz Krzysztofik <jmkrzyszt@gmail.com> Acked-by: Tony Lindgren <tony@atomide.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2022-04-10 06:07:57 -07:00
struct omap1_clk *clk = to_omap1_clk(hw), *parent = to_omap1_clk(clk_hw_get_parent(hw));
if (clk->ops && !(WARN_ON(!clk->ops->disable)))
clk->ops->disable(clk);
ARM: OMAP1: clock: Convert to CCF OMAP1 still uses its own implementation of standard clock API defined in include/linux/clk.h. Internals of that implementation are not visible outside OMAP1 directory. As a consequence, device drivers are not able to register clocks potentially provided by peripheral devices. Drop OMAP1 implementation of the clock API and enable common clock framework. Modify the remaining low level code to be compatible with clock provider API and register the clocks with CCF. Move initialisation of clocks to omap1_timer_init() to avoid memory allocation issues at early setup phase from where omap1_init_early() is called. Register the clocks after initialization of clock I/O registers, local clock pointers used by OMAP1 clock ops, and local .rate fields of clocks with no local implementation of .recalc ops, so CCF structures are populated with correct data during clock registration. Instead of enabling some of the registered clocks, flag them for CCF as critical. Introduce .is_enabled op using code that verifies hardware status of clock enablement, split out from implementation of .disable_unused op, so the latter is actually called by CCF for not requested but hardware enabled clocks. Add .round_rate ops where missing so .set_rate ops are called by CCF as expected. Since CCF allows parallel execution of .enable/.disable and .set_rate ops, protect registers shared among those groups of ops from concurrent access with spinlocks. Drop local debugfs support in favor of that provided by CCF. v2: flag tc2_ck as CLK_IS_CRITICAL (Aaro) v3: rebase on top of soc/omap1-multiplatform-5.18, - drop no longer needed includes from arch/arm/mach-omap1/io.c Signed-off-by: Janusz Krzysztofik <jmkrzyszt@gmail.com> Acked-by: Tony Lindgren <tony@atomide.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2022-04-10 06:07:57 -07:00
if (likely(parent) && clk->flags & CLOCK_NO_IDLE_PARENT)
omap1_clk_allow_idle(parent);
}
ARM: OMAP1: clock: Convert to CCF OMAP1 still uses its own implementation of standard clock API defined in include/linux/clk.h. Internals of that implementation are not visible outside OMAP1 directory. As a consequence, device drivers are not able to register clocks potentially provided by peripheral devices. Drop OMAP1 implementation of the clock API and enable common clock framework. Modify the remaining low level code to be compatible with clock provider API and register the clocks with CCF. Move initialisation of clocks to omap1_timer_init() to avoid memory allocation issues at early setup phase from where omap1_init_early() is called. Register the clocks after initialization of clock I/O registers, local clock pointers used by OMAP1 clock ops, and local .rate fields of clocks with no local implementation of .recalc ops, so CCF structures are populated with correct data during clock registration. Instead of enabling some of the registered clocks, flag them for CCF as critical. Introduce .is_enabled op using code that verifies hardware status of clock enablement, split out from implementation of .disable_unused op, so the latter is actually called by CCF for not requested but hardware enabled clocks. Add .round_rate ops where missing so .set_rate ops are called by CCF as expected. Since CCF allows parallel execution of .enable/.disable and .set_rate ops, protect registers shared among those groups of ops from concurrent access with spinlocks. Drop local debugfs support in favor of that provided by CCF. v2: flag tc2_ck as CLK_IS_CRITICAL (Aaro) v3: rebase on top of soc/omap1-multiplatform-5.18, - drop no longer needed includes from arch/arm/mach-omap1/io.c Signed-off-by: Janusz Krzysztofik <jmkrzyszt@gmail.com> Acked-by: Tony Lindgren <tony@atomide.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2022-04-10 06:07:57 -07:00
static int omap1_clk_enable_generic(struct omap1_clk *clk)
{
ARM: OMAP1: clock: Convert to CCF OMAP1 still uses its own implementation of standard clock API defined in include/linux/clk.h. Internals of that implementation are not visible outside OMAP1 directory. As a consequence, device drivers are not able to register clocks potentially provided by peripheral devices. Drop OMAP1 implementation of the clock API and enable common clock framework. Modify the remaining low level code to be compatible with clock provider API and register the clocks with CCF. Move initialisation of clocks to omap1_timer_init() to avoid memory allocation issues at early setup phase from where omap1_init_early() is called. Register the clocks after initialization of clock I/O registers, local clock pointers used by OMAP1 clock ops, and local .rate fields of clocks with no local implementation of .recalc ops, so CCF structures are populated with correct data during clock registration. Instead of enabling some of the registered clocks, flag them for CCF as critical. Introduce .is_enabled op using code that verifies hardware status of clock enablement, split out from implementation of .disable_unused op, so the latter is actually called by CCF for not requested but hardware enabled clocks. Add .round_rate ops where missing so .set_rate ops are called by CCF as expected. Since CCF allows parallel execution of .enable/.disable and .set_rate ops, protect registers shared among those groups of ops from concurrent access with spinlocks. Drop local debugfs support in favor of that provided by CCF. v2: flag tc2_ck as CLK_IS_CRITICAL (Aaro) v3: rebase on top of soc/omap1-multiplatform-5.18, - drop no longer needed includes from arch/arm/mach-omap1/io.c Signed-off-by: Janusz Krzysztofik <jmkrzyszt@gmail.com> Acked-by: Tony Lindgren <tony@atomide.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2022-04-10 06:07:57 -07:00
unsigned long flags;
__u16 regval16;
__u32 regval32;
if (unlikely(clk->enable_reg == NULL)) {
printk(KERN_ERR "clock.c: Enable for %s without enable code\n",
ARM: OMAP1: clock: Convert to CCF OMAP1 still uses its own implementation of standard clock API defined in include/linux/clk.h. Internals of that implementation are not visible outside OMAP1 directory. As a consequence, device drivers are not able to register clocks potentially provided by peripheral devices. Drop OMAP1 implementation of the clock API and enable common clock framework. Modify the remaining low level code to be compatible with clock provider API and register the clocks with CCF. Move initialisation of clocks to omap1_timer_init() to avoid memory allocation issues at early setup phase from where omap1_init_early() is called. Register the clocks after initialization of clock I/O registers, local clock pointers used by OMAP1 clock ops, and local .rate fields of clocks with no local implementation of .recalc ops, so CCF structures are populated with correct data during clock registration. Instead of enabling some of the registered clocks, flag them for CCF as critical. Introduce .is_enabled op using code that verifies hardware status of clock enablement, split out from implementation of .disable_unused op, so the latter is actually called by CCF for not requested but hardware enabled clocks. Add .round_rate ops where missing so .set_rate ops are called by CCF as expected. Since CCF allows parallel execution of .enable/.disable and .set_rate ops, protect registers shared among those groups of ops from concurrent access with spinlocks. Drop local debugfs support in favor of that provided by CCF. v2: flag tc2_ck as CLK_IS_CRITICAL (Aaro) v3: rebase on top of soc/omap1-multiplatform-5.18, - drop no longer needed includes from arch/arm/mach-omap1/io.c Signed-off-by: Janusz Krzysztofik <jmkrzyszt@gmail.com> Acked-by: Tony Lindgren <tony@atomide.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2022-04-10 06:07:57 -07:00
clk_hw_get_name(&clk->hw));
return -EINVAL;
}
ARM: OMAP1: clock: Convert to CCF OMAP1 still uses its own implementation of standard clock API defined in include/linux/clk.h. Internals of that implementation are not visible outside OMAP1 directory. As a consequence, device drivers are not able to register clocks potentially provided by peripheral devices. Drop OMAP1 implementation of the clock API and enable common clock framework. Modify the remaining low level code to be compatible with clock provider API and register the clocks with CCF. Move initialisation of clocks to omap1_timer_init() to avoid memory allocation issues at early setup phase from where omap1_init_early() is called. Register the clocks after initialization of clock I/O registers, local clock pointers used by OMAP1 clock ops, and local .rate fields of clocks with no local implementation of .recalc ops, so CCF structures are populated with correct data during clock registration. Instead of enabling some of the registered clocks, flag them for CCF as critical. Introduce .is_enabled op using code that verifies hardware status of clock enablement, split out from implementation of .disable_unused op, so the latter is actually called by CCF for not requested but hardware enabled clocks. Add .round_rate ops where missing so .set_rate ops are called by CCF as expected. Since CCF allows parallel execution of .enable/.disable and .set_rate ops, protect registers shared among those groups of ops from concurrent access with spinlocks. Drop local debugfs support in favor of that provided by CCF. v2: flag tc2_ck as CLK_IS_CRITICAL (Aaro) v3: rebase on top of soc/omap1-multiplatform-5.18, - drop no longer needed includes from arch/arm/mach-omap1/io.c Signed-off-by: Janusz Krzysztofik <jmkrzyszt@gmail.com> Acked-by: Tony Lindgren <tony@atomide.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2022-04-10 06:07:57 -07:00
/* protect clk->enable_reg from concurrent access via clk_set_rate() */
if (clk->enable_reg == OMAP1_IO_ADDRESS(ARM_CKCTL))
spin_lock_irqsave(&arm_ckctl_lock, flags);
else if (clk->enable_reg == OMAP1_IO_ADDRESS(ARM_IDLECT2))
spin_lock_irqsave(&arm_idlect2_lock, flags);
else if (clk->enable_reg == OMAP1_IO_ADDRESS(MOD_CONF_CTRL_0))
spin_lock_irqsave(&mod_conf_ctrl_0_lock, flags);
else if (clk->enable_reg == OMAP1_IO_ADDRESS(MOD_CONF_CTRL_1))
spin_lock_irqsave(&mod_conf_ctrl_1_lock, flags);
else if (clk->enable_reg == OMAP1_IO_ADDRESS(SWD_CLK_DIV_CTRL_SEL))
spin_lock_irqsave(&swd_clk_div_ctrl_sel_lock, flags);
if (clk->flags & ENABLE_REG_32BIT) {
regval32 = __raw_readl(clk->enable_reg);
regval32 |= (1 << clk->enable_bit);
__raw_writel(regval32, clk->enable_reg);
} else {
regval16 = __raw_readw(clk->enable_reg);
regval16 |= (1 << clk->enable_bit);
__raw_writew(regval16, clk->enable_reg);
}
ARM: OMAP1: clock: Convert to CCF OMAP1 still uses its own implementation of standard clock API defined in include/linux/clk.h. Internals of that implementation are not visible outside OMAP1 directory. As a consequence, device drivers are not able to register clocks potentially provided by peripheral devices. Drop OMAP1 implementation of the clock API and enable common clock framework. Modify the remaining low level code to be compatible with clock provider API and register the clocks with CCF. Move initialisation of clocks to omap1_timer_init() to avoid memory allocation issues at early setup phase from where omap1_init_early() is called. Register the clocks after initialization of clock I/O registers, local clock pointers used by OMAP1 clock ops, and local .rate fields of clocks with no local implementation of .recalc ops, so CCF structures are populated with correct data during clock registration. Instead of enabling some of the registered clocks, flag them for CCF as critical. Introduce .is_enabled op using code that verifies hardware status of clock enablement, split out from implementation of .disable_unused op, so the latter is actually called by CCF for not requested but hardware enabled clocks. Add .round_rate ops where missing so .set_rate ops are called by CCF as expected. Since CCF allows parallel execution of .enable/.disable and .set_rate ops, protect registers shared among those groups of ops from concurrent access with spinlocks. Drop local debugfs support in favor of that provided by CCF. v2: flag tc2_ck as CLK_IS_CRITICAL (Aaro) v3: rebase on top of soc/omap1-multiplatform-5.18, - drop no longer needed includes from arch/arm/mach-omap1/io.c Signed-off-by: Janusz Krzysztofik <jmkrzyszt@gmail.com> Acked-by: Tony Lindgren <tony@atomide.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2022-04-10 06:07:57 -07:00
if (clk->enable_reg == OMAP1_IO_ADDRESS(ARM_CKCTL))
spin_unlock_irqrestore(&arm_ckctl_lock, flags);
else if (clk->enable_reg == OMAP1_IO_ADDRESS(ARM_IDLECT2))
spin_unlock_irqrestore(&arm_idlect2_lock, flags);
else if (clk->enable_reg == OMAP1_IO_ADDRESS(MOD_CONF_CTRL_0))
spin_unlock_irqrestore(&mod_conf_ctrl_0_lock, flags);
else if (clk->enable_reg == OMAP1_IO_ADDRESS(MOD_CONF_CTRL_1))
spin_unlock_irqrestore(&mod_conf_ctrl_1_lock, flags);
else if (clk->enable_reg == OMAP1_IO_ADDRESS(SWD_CLK_DIV_CTRL_SEL))
spin_unlock_irqrestore(&swd_clk_div_ctrl_sel_lock, flags);
return 0;
}
ARM: OMAP1: clock: Convert to CCF OMAP1 still uses its own implementation of standard clock API defined in include/linux/clk.h. Internals of that implementation are not visible outside OMAP1 directory. As a consequence, device drivers are not able to register clocks potentially provided by peripheral devices. Drop OMAP1 implementation of the clock API and enable common clock framework. Modify the remaining low level code to be compatible with clock provider API and register the clocks with CCF. Move initialisation of clocks to omap1_timer_init() to avoid memory allocation issues at early setup phase from where omap1_init_early() is called. Register the clocks after initialization of clock I/O registers, local clock pointers used by OMAP1 clock ops, and local .rate fields of clocks with no local implementation of .recalc ops, so CCF structures are populated with correct data during clock registration. Instead of enabling some of the registered clocks, flag them for CCF as critical. Introduce .is_enabled op using code that verifies hardware status of clock enablement, split out from implementation of .disable_unused op, so the latter is actually called by CCF for not requested but hardware enabled clocks. Add .round_rate ops where missing so .set_rate ops are called by CCF as expected. Since CCF allows parallel execution of .enable/.disable and .set_rate ops, protect registers shared among those groups of ops from concurrent access with spinlocks. Drop local debugfs support in favor of that provided by CCF. v2: flag tc2_ck as CLK_IS_CRITICAL (Aaro) v3: rebase on top of soc/omap1-multiplatform-5.18, - drop no longer needed includes from arch/arm/mach-omap1/io.c Signed-off-by: Janusz Krzysztofik <jmkrzyszt@gmail.com> Acked-by: Tony Lindgren <tony@atomide.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2022-04-10 06:07:57 -07:00
static void omap1_clk_disable_generic(struct omap1_clk *clk)
{
ARM: OMAP1: clock: Convert to CCF OMAP1 still uses its own implementation of standard clock API defined in include/linux/clk.h. Internals of that implementation are not visible outside OMAP1 directory. As a consequence, device drivers are not able to register clocks potentially provided by peripheral devices. Drop OMAP1 implementation of the clock API and enable common clock framework. Modify the remaining low level code to be compatible with clock provider API and register the clocks with CCF. Move initialisation of clocks to omap1_timer_init() to avoid memory allocation issues at early setup phase from where omap1_init_early() is called. Register the clocks after initialization of clock I/O registers, local clock pointers used by OMAP1 clock ops, and local .rate fields of clocks with no local implementation of .recalc ops, so CCF structures are populated with correct data during clock registration. Instead of enabling some of the registered clocks, flag them for CCF as critical. Introduce .is_enabled op using code that verifies hardware status of clock enablement, split out from implementation of .disable_unused op, so the latter is actually called by CCF for not requested but hardware enabled clocks. Add .round_rate ops where missing so .set_rate ops are called by CCF as expected. Since CCF allows parallel execution of .enable/.disable and .set_rate ops, protect registers shared among those groups of ops from concurrent access with spinlocks. Drop local debugfs support in favor of that provided by CCF. v2: flag tc2_ck as CLK_IS_CRITICAL (Aaro) v3: rebase on top of soc/omap1-multiplatform-5.18, - drop no longer needed includes from arch/arm/mach-omap1/io.c Signed-off-by: Janusz Krzysztofik <jmkrzyszt@gmail.com> Acked-by: Tony Lindgren <tony@atomide.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2022-04-10 06:07:57 -07:00
unsigned long flags;
__u16 regval16;
__u32 regval32;
if (clk->enable_reg == NULL)
return;
ARM: OMAP1: clock: Convert to CCF OMAP1 still uses its own implementation of standard clock API defined in include/linux/clk.h. Internals of that implementation are not visible outside OMAP1 directory. As a consequence, device drivers are not able to register clocks potentially provided by peripheral devices. Drop OMAP1 implementation of the clock API and enable common clock framework. Modify the remaining low level code to be compatible with clock provider API and register the clocks with CCF. Move initialisation of clocks to omap1_timer_init() to avoid memory allocation issues at early setup phase from where omap1_init_early() is called. Register the clocks after initialization of clock I/O registers, local clock pointers used by OMAP1 clock ops, and local .rate fields of clocks with no local implementation of .recalc ops, so CCF structures are populated with correct data during clock registration. Instead of enabling some of the registered clocks, flag them for CCF as critical. Introduce .is_enabled op using code that verifies hardware status of clock enablement, split out from implementation of .disable_unused op, so the latter is actually called by CCF for not requested but hardware enabled clocks. Add .round_rate ops where missing so .set_rate ops are called by CCF as expected. Since CCF allows parallel execution of .enable/.disable and .set_rate ops, protect registers shared among those groups of ops from concurrent access with spinlocks. Drop local debugfs support in favor of that provided by CCF. v2: flag tc2_ck as CLK_IS_CRITICAL (Aaro) v3: rebase on top of soc/omap1-multiplatform-5.18, - drop no longer needed includes from arch/arm/mach-omap1/io.c Signed-off-by: Janusz Krzysztofik <jmkrzyszt@gmail.com> Acked-by: Tony Lindgren <tony@atomide.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2022-04-10 06:07:57 -07:00
/* protect clk->enable_reg from concurrent access via clk_set_rate() */
if (clk->enable_reg == OMAP1_IO_ADDRESS(ARM_CKCTL))
spin_lock_irqsave(&arm_ckctl_lock, flags);
else if (clk->enable_reg == OMAP1_IO_ADDRESS(ARM_IDLECT2))
spin_lock_irqsave(&arm_idlect2_lock, flags);
else if (clk->enable_reg == OMAP1_IO_ADDRESS(MOD_CONF_CTRL_0))
spin_lock_irqsave(&mod_conf_ctrl_0_lock, flags);
else if (clk->enable_reg == OMAP1_IO_ADDRESS(MOD_CONF_CTRL_1))
spin_lock_irqsave(&mod_conf_ctrl_1_lock, flags);
else if (clk->enable_reg == OMAP1_IO_ADDRESS(SWD_CLK_DIV_CTRL_SEL))
spin_lock_irqsave(&swd_clk_div_ctrl_sel_lock, flags);
if (clk->flags & ENABLE_REG_32BIT) {
regval32 = __raw_readl(clk->enable_reg);
regval32 &= ~(1 << clk->enable_bit);
__raw_writel(regval32, clk->enable_reg);
} else {
regval16 = __raw_readw(clk->enable_reg);
regval16 &= ~(1 << clk->enable_bit);
__raw_writew(regval16, clk->enable_reg);
}
ARM: OMAP1: clock: Convert to CCF OMAP1 still uses its own implementation of standard clock API defined in include/linux/clk.h. Internals of that implementation are not visible outside OMAP1 directory. As a consequence, device drivers are not able to register clocks potentially provided by peripheral devices. Drop OMAP1 implementation of the clock API and enable common clock framework. Modify the remaining low level code to be compatible with clock provider API and register the clocks with CCF. Move initialisation of clocks to omap1_timer_init() to avoid memory allocation issues at early setup phase from where omap1_init_early() is called. Register the clocks after initialization of clock I/O registers, local clock pointers used by OMAP1 clock ops, and local .rate fields of clocks with no local implementation of .recalc ops, so CCF structures are populated with correct data during clock registration. Instead of enabling some of the registered clocks, flag them for CCF as critical. Introduce .is_enabled op using code that verifies hardware status of clock enablement, split out from implementation of .disable_unused op, so the latter is actually called by CCF for not requested but hardware enabled clocks. Add .round_rate ops where missing so .set_rate ops are called by CCF as expected. Since CCF allows parallel execution of .enable/.disable and .set_rate ops, protect registers shared among those groups of ops from concurrent access with spinlocks. Drop local debugfs support in favor of that provided by CCF. v2: flag tc2_ck as CLK_IS_CRITICAL (Aaro) v3: rebase on top of soc/omap1-multiplatform-5.18, - drop no longer needed includes from arch/arm/mach-omap1/io.c Signed-off-by: Janusz Krzysztofik <jmkrzyszt@gmail.com> Acked-by: Tony Lindgren <tony@atomide.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2022-04-10 06:07:57 -07:00
if (clk->enable_reg == OMAP1_IO_ADDRESS(ARM_CKCTL))
spin_unlock_irqrestore(&arm_ckctl_lock, flags);
else if (clk->enable_reg == OMAP1_IO_ADDRESS(ARM_IDLECT2))
spin_unlock_irqrestore(&arm_idlect2_lock, flags);
else if (clk->enable_reg == OMAP1_IO_ADDRESS(MOD_CONF_CTRL_0))
spin_unlock_irqrestore(&mod_conf_ctrl_0_lock, flags);
else if (clk->enable_reg == OMAP1_IO_ADDRESS(MOD_CONF_CTRL_1))
spin_unlock_irqrestore(&mod_conf_ctrl_1_lock, flags);
else if (clk->enable_reg == OMAP1_IO_ADDRESS(SWD_CLK_DIV_CTRL_SEL))
spin_unlock_irqrestore(&swd_clk_div_ctrl_sel_lock, flags);
}
const struct clkops clkops_generic = {
.enable = omap1_clk_enable_generic,
.disable = omap1_clk_disable_generic,
};
ARM: OMAP1: clock: Convert to CCF OMAP1 still uses its own implementation of standard clock API defined in include/linux/clk.h. Internals of that implementation are not visible outside OMAP1 directory. As a consequence, device drivers are not able to register clocks potentially provided by peripheral devices. Drop OMAP1 implementation of the clock API and enable common clock framework. Modify the remaining low level code to be compatible with clock provider API and register the clocks with CCF. Move initialisation of clocks to omap1_timer_init() to avoid memory allocation issues at early setup phase from where omap1_init_early() is called. Register the clocks after initialization of clock I/O registers, local clock pointers used by OMAP1 clock ops, and local .rate fields of clocks with no local implementation of .recalc ops, so CCF structures are populated with correct data during clock registration. Instead of enabling some of the registered clocks, flag them for CCF as critical. Introduce .is_enabled op using code that verifies hardware status of clock enablement, split out from implementation of .disable_unused op, so the latter is actually called by CCF for not requested but hardware enabled clocks. Add .round_rate ops where missing so .set_rate ops are called by CCF as expected. Since CCF allows parallel execution of .enable/.disable and .set_rate ops, protect registers shared among those groups of ops from concurrent access with spinlocks. Drop local debugfs support in favor of that provided by CCF. v2: flag tc2_ck as CLK_IS_CRITICAL (Aaro) v3: rebase on top of soc/omap1-multiplatform-5.18, - drop no longer needed includes from arch/arm/mach-omap1/io.c Signed-off-by: Janusz Krzysztofik <jmkrzyszt@gmail.com> Acked-by: Tony Lindgren <tony@atomide.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2022-04-10 06:07:57 -07:00
static int omap1_clk_enable_dsp_domain(struct omap1_clk *clk)
{
ARM: OMAP1: clock: Convert to CCF OMAP1 still uses its own implementation of standard clock API defined in include/linux/clk.h. Internals of that implementation are not visible outside OMAP1 directory. As a consequence, device drivers are not able to register clocks potentially provided by peripheral devices. Drop OMAP1 implementation of the clock API and enable common clock framework. Modify the remaining low level code to be compatible with clock provider API and register the clocks with CCF. Move initialisation of clocks to omap1_timer_init() to avoid memory allocation issues at early setup phase from where omap1_init_early() is called. Register the clocks after initialization of clock I/O registers, local clock pointers used by OMAP1 clock ops, and local .rate fields of clocks with no local implementation of .recalc ops, so CCF structures are populated with correct data during clock registration. Instead of enabling some of the registered clocks, flag them for CCF as critical. Introduce .is_enabled op using code that verifies hardware status of clock enablement, split out from implementation of .disable_unused op, so the latter is actually called by CCF for not requested but hardware enabled clocks. Add .round_rate ops where missing so .set_rate ops are called by CCF as expected. Since CCF allows parallel execution of .enable/.disable and .set_rate ops, protect registers shared among those groups of ops from concurrent access with spinlocks. Drop local debugfs support in favor of that provided by CCF. v2: flag tc2_ck as CLK_IS_CRITICAL (Aaro) v3: rebase on top of soc/omap1-multiplatform-5.18, - drop no longer needed includes from arch/arm/mach-omap1/io.c Signed-off-by: Janusz Krzysztofik <jmkrzyszt@gmail.com> Acked-by: Tony Lindgren <tony@atomide.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2022-04-10 06:07:57 -07:00
bool api_ck_was_enabled;
int retval = 0;
api_ck_was_enabled = omap1_clk_is_enabled(&api_ck_p->hw);
if (!api_ck_was_enabled)
retval = api_ck_p->ops->enable(api_ck_p);
if (!retval) {
retval = omap1_clk_enable_generic(clk);
ARM: OMAP1: clock: Convert to CCF OMAP1 still uses its own implementation of standard clock API defined in include/linux/clk.h. Internals of that implementation are not visible outside OMAP1 directory. As a consequence, device drivers are not able to register clocks potentially provided by peripheral devices. Drop OMAP1 implementation of the clock API and enable common clock framework. Modify the remaining low level code to be compatible with clock provider API and register the clocks with CCF. Move initialisation of clocks to omap1_timer_init() to avoid memory allocation issues at early setup phase from where omap1_init_early() is called. Register the clocks after initialization of clock I/O registers, local clock pointers used by OMAP1 clock ops, and local .rate fields of clocks with no local implementation of .recalc ops, so CCF structures are populated with correct data during clock registration. Instead of enabling some of the registered clocks, flag them for CCF as critical. Introduce .is_enabled op using code that verifies hardware status of clock enablement, split out from implementation of .disable_unused op, so the latter is actually called by CCF for not requested but hardware enabled clocks. Add .round_rate ops where missing so .set_rate ops are called by CCF as expected. Since CCF allows parallel execution of .enable/.disable and .set_rate ops, protect registers shared among those groups of ops from concurrent access with spinlocks. Drop local debugfs support in favor of that provided by CCF. v2: flag tc2_ck as CLK_IS_CRITICAL (Aaro) v3: rebase on top of soc/omap1-multiplatform-5.18, - drop no longer needed includes from arch/arm/mach-omap1/io.c Signed-off-by: Janusz Krzysztofik <jmkrzyszt@gmail.com> Acked-by: Tony Lindgren <tony@atomide.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2022-04-10 06:07:57 -07:00
if (!api_ck_was_enabled)
api_ck_p->ops->disable(api_ck_p);
}
return retval;
}
ARM: OMAP1: clock: Convert to CCF OMAP1 still uses its own implementation of standard clock API defined in include/linux/clk.h. Internals of that implementation are not visible outside OMAP1 directory. As a consequence, device drivers are not able to register clocks potentially provided by peripheral devices. Drop OMAP1 implementation of the clock API and enable common clock framework. Modify the remaining low level code to be compatible with clock provider API and register the clocks with CCF. Move initialisation of clocks to omap1_timer_init() to avoid memory allocation issues at early setup phase from where omap1_init_early() is called. Register the clocks after initialization of clock I/O registers, local clock pointers used by OMAP1 clock ops, and local .rate fields of clocks with no local implementation of .recalc ops, so CCF structures are populated with correct data during clock registration. Instead of enabling some of the registered clocks, flag them for CCF as critical. Introduce .is_enabled op using code that verifies hardware status of clock enablement, split out from implementation of .disable_unused op, so the latter is actually called by CCF for not requested but hardware enabled clocks. Add .round_rate ops where missing so .set_rate ops are called by CCF as expected. Since CCF allows parallel execution of .enable/.disable and .set_rate ops, protect registers shared among those groups of ops from concurrent access with spinlocks. Drop local debugfs support in favor of that provided by CCF. v2: flag tc2_ck as CLK_IS_CRITICAL (Aaro) v3: rebase on top of soc/omap1-multiplatform-5.18, - drop no longer needed includes from arch/arm/mach-omap1/io.c Signed-off-by: Janusz Krzysztofik <jmkrzyszt@gmail.com> Acked-by: Tony Lindgren <tony@atomide.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2022-04-10 06:07:57 -07:00
static void omap1_clk_disable_dsp_domain(struct omap1_clk *clk)
{
ARM: OMAP1: clock: Convert to CCF OMAP1 still uses its own implementation of standard clock API defined in include/linux/clk.h. Internals of that implementation are not visible outside OMAP1 directory. As a consequence, device drivers are not able to register clocks potentially provided by peripheral devices. Drop OMAP1 implementation of the clock API and enable common clock framework. Modify the remaining low level code to be compatible with clock provider API and register the clocks with CCF. Move initialisation of clocks to omap1_timer_init() to avoid memory allocation issues at early setup phase from where omap1_init_early() is called. Register the clocks after initialization of clock I/O registers, local clock pointers used by OMAP1 clock ops, and local .rate fields of clocks with no local implementation of .recalc ops, so CCF structures are populated with correct data during clock registration. Instead of enabling some of the registered clocks, flag them for CCF as critical. Introduce .is_enabled op using code that verifies hardware status of clock enablement, split out from implementation of .disable_unused op, so the latter is actually called by CCF for not requested but hardware enabled clocks. Add .round_rate ops where missing so .set_rate ops are called by CCF as expected. Since CCF allows parallel execution of .enable/.disable and .set_rate ops, protect registers shared among those groups of ops from concurrent access with spinlocks. Drop local debugfs support in favor of that provided by CCF. v2: flag tc2_ck as CLK_IS_CRITICAL (Aaro) v3: rebase on top of soc/omap1-multiplatform-5.18, - drop no longer needed includes from arch/arm/mach-omap1/io.c Signed-off-by: Janusz Krzysztofik <jmkrzyszt@gmail.com> Acked-by: Tony Lindgren <tony@atomide.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2022-04-10 06:07:57 -07:00
bool api_ck_was_enabled;
api_ck_was_enabled = omap1_clk_is_enabled(&api_ck_p->hw);
if (!api_ck_was_enabled)
if (api_ck_p->ops->enable(api_ck_p) < 0)
return;
omap1_clk_disable_generic(clk);
if (!api_ck_was_enabled)
api_ck_p->ops->disable(api_ck_p);
}
const struct clkops clkops_dspck = {
.enable = omap1_clk_enable_dsp_domain,
.disable = omap1_clk_disable_dsp_domain,
};
/* XXX SYSC register handling does not belong in the clock framework */
ARM: OMAP1: clock: Convert to CCF OMAP1 still uses its own implementation of standard clock API defined in include/linux/clk.h. Internals of that implementation are not visible outside OMAP1 directory. As a consequence, device drivers are not able to register clocks potentially provided by peripheral devices. Drop OMAP1 implementation of the clock API and enable common clock framework. Modify the remaining low level code to be compatible with clock provider API and register the clocks with CCF. Move initialisation of clocks to omap1_timer_init() to avoid memory allocation issues at early setup phase from where omap1_init_early() is called. Register the clocks after initialization of clock I/O registers, local clock pointers used by OMAP1 clock ops, and local .rate fields of clocks with no local implementation of .recalc ops, so CCF structures are populated with correct data during clock registration. Instead of enabling some of the registered clocks, flag them for CCF as critical. Introduce .is_enabled op using code that verifies hardware status of clock enablement, split out from implementation of .disable_unused op, so the latter is actually called by CCF for not requested but hardware enabled clocks. Add .round_rate ops where missing so .set_rate ops are called by CCF as expected. Since CCF allows parallel execution of .enable/.disable and .set_rate ops, protect registers shared among those groups of ops from concurrent access with spinlocks. Drop local debugfs support in favor of that provided by CCF. v2: flag tc2_ck as CLK_IS_CRITICAL (Aaro) v3: rebase on top of soc/omap1-multiplatform-5.18, - drop no longer needed includes from arch/arm/mach-omap1/io.c Signed-off-by: Janusz Krzysztofik <jmkrzyszt@gmail.com> Acked-by: Tony Lindgren <tony@atomide.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2022-04-10 06:07:57 -07:00
static int omap1_clk_enable_uart_functional_16xx(struct omap1_clk *clk)
{
int ret;
struct uart_clk *uclk;
ret = omap1_clk_enable_generic(clk);
if (ret == 0) {
/* Set smart idle acknowledgement mode */
uclk = (struct uart_clk *)clk;
omap_writeb((omap_readb(uclk->sysc_addr) & ~0x10) | 8,
uclk->sysc_addr);
}
return ret;
}
/* XXX SYSC register handling does not belong in the clock framework */
ARM: OMAP1: clock: Convert to CCF OMAP1 still uses its own implementation of standard clock API defined in include/linux/clk.h. Internals of that implementation are not visible outside OMAP1 directory. As a consequence, device drivers are not able to register clocks potentially provided by peripheral devices. Drop OMAP1 implementation of the clock API and enable common clock framework. Modify the remaining low level code to be compatible with clock provider API and register the clocks with CCF. Move initialisation of clocks to omap1_timer_init() to avoid memory allocation issues at early setup phase from where omap1_init_early() is called. Register the clocks after initialization of clock I/O registers, local clock pointers used by OMAP1 clock ops, and local .rate fields of clocks with no local implementation of .recalc ops, so CCF structures are populated with correct data during clock registration. Instead of enabling some of the registered clocks, flag them for CCF as critical. Introduce .is_enabled op using code that verifies hardware status of clock enablement, split out from implementation of .disable_unused op, so the latter is actually called by CCF for not requested but hardware enabled clocks. Add .round_rate ops where missing so .set_rate ops are called by CCF as expected. Since CCF allows parallel execution of .enable/.disable and .set_rate ops, protect registers shared among those groups of ops from concurrent access with spinlocks. Drop local debugfs support in favor of that provided by CCF. v2: flag tc2_ck as CLK_IS_CRITICAL (Aaro) v3: rebase on top of soc/omap1-multiplatform-5.18, - drop no longer needed includes from arch/arm/mach-omap1/io.c Signed-off-by: Janusz Krzysztofik <jmkrzyszt@gmail.com> Acked-by: Tony Lindgren <tony@atomide.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2022-04-10 06:07:57 -07:00
static void omap1_clk_disable_uart_functional_16xx(struct omap1_clk *clk)
{
struct uart_clk *uclk;
/* Set force idle acknowledgement mode */
uclk = (struct uart_clk *)clk;
omap_writeb((omap_readb(uclk->sysc_addr) & ~0x18), uclk->sysc_addr);
omap1_clk_disable_generic(clk);
}
/* XXX SYSC register handling does not belong in the clock framework */
const struct clkops clkops_uart_16xx = {
.enable = omap1_clk_enable_uart_functional_16xx,
.disable = omap1_clk_disable_uart_functional_16xx,
};
ARM: OMAP1: clock: Convert to CCF OMAP1 still uses its own implementation of standard clock API defined in include/linux/clk.h. Internals of that implementation are not visible outside OMAP1 directory. As a consequence, device drivers are not able to register clocks potentially provided by peripheral devices. Drop OMAP1 implementation of the clock API and enable common clock framework. Modify the remaining low level code to be compatible with clock provider API and register the clocks with CCF. Move initialisation of clocks to omap1_timer_init() to avoid memory allocation issues at early setup phase from where omap1_init_early() is called. Register the clocks after initialization of clock I/O registers, local clock pointers used by OMAP1 clock ops, and local .rate fields of clocks with no local implementation of .recalc ops, so CCF structures are populated with correct data during clock registration. Instead of enabling some of the registered clocks, flag them for CCF as critical. Introduce .is_enabled op using code that verifies hardware status of clock enablement, split out from implementation of .disable_unused op, so the latter is actually called by CCF for not requested but hardware enabled clocks. Add .round_rate ops where missing so .set_rate ops are called by CCF as expected. Since CCF allows parallel execution of .enable/.disable and .set_rate ops, protect registers shared among those groups of ops from concurrent access with spinlocks. Drop local debugfs support in favor of that provided by CCF. v2: flag tc2_ck as CLK_IS_CRITICAL (Aaro) v3: rebase on top of soc/omap1-multiplatform-5.18, - drop no longer needed includes from arch/arm/mach-omap1/io.c Signed-off-by: Janusz Krzysztofik <jmkrzyszt@gmail.com> Acked-by: Tony Lindgren <tony@atomide.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2022-04-10 06:07:57 -07:00
static unsigned long omap1_clk_recalc_rate(struct clk_hw *hw, unsigned long p_rate)
{
ARM: OMAP1: clock: Convert to CCF OMAP1 still uses its own implementation of standard clock API defined in include/linux/clk.h. Internals of that implementation are not visible outside OMAP1 directory. As a consequence, device drivers are not able to register clocks potentially provided by peripheral devices. Drop OMAP1 implementation of the clock API and enable common clock framework. Modify the remaining low level code to be compatible with clock provider API and register the clocks with CCF. Move initialisation of clocks to omap1_timer_init() to avoid memory allocation issues at early setup phase from where omap1_init_early() is called. Register the clocks after initialization of clock I/O registers, local clock pointers used by OMAP1 clock ops, and local .rate fields of clocks with no local implementation of .recalc ops, so CCF structures are populated with correct data during clock registration. Instead of enabling some of the registered clocks, flag them for CCF as critical. Introduce .is_enabled op using code that verifies hardware status of clock enablement, split out from implementation of .disable_unused op, so the latter is actually called by CCF for not requested but hardware enabled clocks. Add .round_rate ops where missing so .set_rate ops are called by CCF as expected. Since CCF allows parallel execution of .enable/.disable and .set_rate ops, protect registers shared among those groups of ops from concurrent access with spinlocks. Drop local debugfs support in favor of that provided by CCF. v2: flag tc2_ck as CLK_IS_CRITICAL (Aaro) v3: rebase on top of soc/omap1-multiplatform-5.18, - drop no longer needed includes from arch/arm/mach-omap1/io.c Signed-off-by: Janusz Krzysztofik <jmkrzyszt@gmail.com> Acked-by: Tony Lindgren <tony@atomide.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2022-04-10 06:07:57 -07:00
struct omap1_clk *clk = to_omap1_clk(hw);
if (clk->recalc)
return clk->recalc(clk, p_rate);
return clk->rate;
}
ARM: OMAP1: clock: Convert to CCF OMAP1 still uses its own implementation of standard clock API defined in include/linux/clk.h. Internals of that implementation are not visible outside OMAP1 directory. As a consequence, device drivers are not able to register clocks potentially provided by peripheral devices. Drop OMAP1 implementation of the clock API and enable common clock framework. Modify the remaining low level code to be compatible with clock provider API and register the clocks with CCF. Move initialisation of clocks to omap1_timer_init() to avoid memory allocation issues at early setup phase from where omap1_init_early() is called. Register the clocks after initialization of clock I/O registers, local clock pointers used by OMAP1 clock ops, and local .rate fields of clocks with no local implementation of .recalc ops, so CCF structures are populated with correct data during clock registration. Instead of enabling some of the registered clocks, flag them for CCF as critical. Introduce .is_enabled op using code that verifies hardware status of clock enablement, split out from implementation of .disable_unused op, so the latter is actually called by CCF for not requested but hardware enabled clocks. Add .round_rate ops where missing so .set_rate ops are called by CCF as expected. Since CCF allows parallel execution of .enable/.disable and .set_rate ops, protect registers shared among those groups of ops from concurrent access with spinlocks. Drop local debugfs support in favor of that provided by CCF. v2: flag tc2_ck as CLK_IS_CRITICAL (Aaro) v3: rebase on top of soc/omap1-multiplatform-5.18, - drop no longer needed includes from arch/arm/mach-omap1/io.c Signed-off-by: Janusz Krzysztofik <jmkrzyszt@gmail.com> Acked-by: Tony Lindgren <tony@atomide.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2022-04-10 06:07:57 -07:00
static long omap1_clk_round_rate(struct clk_hw *hw, unsigned long rate, unsigned long *p_rate)
{
struct omap1_clk *clk = to_omap1_clk(hw);
if (clk->round_rate != NULL)
return clk->round_rate(clk, rate, p_rate);
return omap1_clk_recalc_rate(hw, *p_rate);
}
static int omap1_clk_set_rate(struct clk_hw *hw, unsigned long rate, unsigned long p_rate)
{
ARM: OMAP1: clock: Convert to CCF OMAP1 still uses its own implementation of standard clock API defined in include/linux/clk.h. Internals of that implementation are not visible outside OMAP1 directory. As a consequence, device drivers are not able to register clocks potentially provided by peripheral devices. Drop OMAP1 implementation of the clock API and enable common clock framework. Modify the remaining low level code to be compatible with clock provider API and register the clocks with CCF. Move initialisation of clocks to omap1_timer_init() to avoid memory allocation issues at early setup phase from where omap1_init_early() is called. Register the clocks after initialization of clock I/O registers, local clock pointers used by OMAP1 clock ops, and local .rate fields of clocks with no local implementation of .recalc ops, so CCF structures are populated with correct data during clock registration. Instead of enabling some of the registered clocks, flag them for CCF as critical. Introduce .is_enabled op using code that verifies hardware status of clock enablement, split out from implementation of .disable_unused op, so the latter is actually called by CCF for not requested but hardware enabled clocks. Add .round_rate ops where missing so .set_rate ops are called by CCF as expected. Since CCF allows parallel execution of .enable/.disable and .set_rate ops, protect registers shared among those groups of ops from concurrent access with spinlocks. Drop local debugfs support in favor of that provided by CCF. v2: flag tc2_ck as CLK_IS_CRITICAL (Aaro) v3: rebase on top of soc/omap1-multiplatform-5.18, - drop no longer needed includes from arch/arm/mach-omap1/io.c Signed-off-by: Janusz Krzysztofik <jmkrzyszt@gmail.com> Acked-by: Tony Lindgren <tony@atomide.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2022-04-10 06:07:57 -07:00
struct omap1_clk *clk = to_omap1_clk(hw);
int ret = -EINVAL;
if (clk->set_rate)
ARM: OMAP1: clock: Convert to CCF OMAP1 still uses its own implementation of standard clock API defined in include/linux/clk.h. Internals of that implementation are not visible outside OMAP1 directory. As a consequence, device drivers are not able to register clocks potentially provided by peripheral devices. Drop OMAP1 implementation of the clock API and enable common clock framework. Modify the remaining low level code to be compatible with clock provider API and register the clocks with CCF. Move initialisation of clocks to omap1_timer_init() to avoid memory allocation issues at early setup phase from where omap1_init_early() is called. Register the clocks after initialization of clock I/O registers, local clock pointers used by OMAP1 clock ops, and local .rate fields of clocks with no local implementation of .recalc ops, so CCF structures are populated with correct data during clock registration. Instead of enabling some of the registered clocks, flag them for CCF as critical. Introduce .is_enabled op using code that verifies hardware status of clock enablement, split out from implementation of .disable_unused op, so the latter is actually called by CCF for not requested but hardware enabled clocks. Add .round_rate ops where missing so .set_rate ops are called by CCF as expected. Since CCF allows parallel execution of .enable/.disable and .set_rate ops, protect registers shared among those groups of ops from concurrent access with spinlocks. Drop local debugfs support in favor of that provided by CCF. v2: flag tc2_ck as CLK_IS_CRITICAL (Aaro) v3: rebase on top of soc/omap1-multiplatform-5.18, - drop no longer needed includes from arch/arm/mach-omap1/io.c Signed-off-by: Janusz Krzysztofik <jmkrzyszt@gmail.com> Acked-by: Tony Lindgren <tony@atomide.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2022-04-10 06:07:57 -07:00
ret = clk->set_rate(clk, rate, p_rate);
return ret;
}
/*
* Omap1 clock reset and init functions
*/
ARM: OMAP1: clock: Convert to CCF OMAP1 still uses its own implementation of standard clock API defined in include/linux/clk.h. Internals of that implementation are not visible outside OMAP1 directory. As a consequence, device drivers are not able to register clocks potentially provided by peripheral devices. Drop OMAP1 implementation of the clock API and enable common clock framework. Modify the remaining low level code to be compatible with clock provider API and register the clocks with CCF. Move initialisation of clocks to omap1_timer_init() to avoid memory allocation issues at early setup phase from where omap1_init_early() is called. Register the clocks after initialization of clock I/O registers, local clock pointers used by OMAP1 clock ops, and local .rate fields of clocks with no local implementation of .recalc ops, so CCF structures are populated with correct data during clock registration. Instead of enabling some of the registered clocks, flag them for CCF as critical. Introduce .is_enabled op using code that verifies hardware status of clock enablement, split out from implementation of .disable_unused op, so the latter is actually called by CCF for not requested but hardware enabled clocks. Add .round_rate ops where missing so .set_rate ops are called by CCF as expected. Since CCF allows parallel execution of .enable/.disable and .set_rate ops, protect registers shared among those groups of ops from concurrent access with spinlocks. Drop local debugfs support in favor of that provided by CCF. v2: flag tc2_ck as CLK_IS_CRITICAL (Aaro) v3: rebase on top of soc/omap1-multiplatform-5.18, - drop no longer needed includes from arch/arm/mach-omap1/io.c Signed-off-by: Janusz Krzysztofik <jmkrzyszt@gmail.com> Acked-by: Tony Lindgren <tony@atomide.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2022-04-10 06:07:57 -07:00
static int omap1_clk_init_op(struct clk_hw *hw)
{
struct omap1_clk *clk = to_omap1_clk(hw);
if (clk->init)
return clk->init(clk);
return 0;
}
#ifdef CONFIG_OMAP_RESET_CLOCKS
ARM: OMAP1: clock: Convert to CCF OMAP1 still uses its own implementation of standard clock API defined in include/linux/clk.h. Internals of that implementation are not visible outside OMAP1 directory. As a consequence, device drivers are not able to register clocks potentially provided by peripheral devices. Drop OMAP1 implementation of the clock API and enable common clock framework. Modify the remaining low level code to be compatible with clock provider API and register the clocks with CCF. Move initialisation of clocks to omap1_timer_init() to avoid memory allocation issues at early setup phase from where omap1_init_early() is called. Register the clocks after initialization of clock I/O registers, local clock pointers used by OMAP1 clock ops, and local .rate fields of clocks with no local implementation of .recalc ops, so CCF structures are populated with correct data during clock registration. Instead of enabling some of the registered clocks, flag them for CCF as critical. Introduce .is_enabled op using code that verifies hardware status of clock enablement, split out from implementation of .disable_unused op, so the latter is actually called by CCF for not requested but hardware enabled clocks. Add .round_rate ops where missing so .set_rate ops are called by CCF as expected. Since CCF allows parallel execution of .enable/.disable and .set_rate ops, protect registers shared among those groups of ops from concurrent access with spinlocks. Drop local debugfs support in favor of that provided by CCF. v2: flag tc2_ck as CLK_IS_CRITICAL (Aaro) v3: rebase on top of soc/omap1-multiplatform-5.18, - drop no longer needed includes from arch/arm/mach-omap1/io.c Signed-off-by: Janusz Krzysztofik <jmkrzyszt@gmail.com> Acked-by: Tony Lindgren <tony@atomide.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2022-04-10 06:07:57 -07:00
static void omap1_clk_disable_unused(struct clk_hw *hw)
{
ARM: OMAP1: clock: Convert to CCF OMAP1 still uses its own implementation of standard clock API defined in include/linux/clk.h. Internals of that implementation are not visible outside OMAP1 directory. As a consequence, device drivers are not able to register clocks potentially provided by peripheral devices. Drop OMAP1 implementation of the clock API and enable common clock framework. Modify the remaining low level code to be compatible with clock provider API and register the clocks with CCF. Move initialisation of clocks to omap1_timer_init() to avoid memory allocation issues at early setup phase from where omap1_init_early() is called. Register the clocks after initialization of clock I/O registers, local clock pointers used by OMAP1 clock ops, and local .rate fields of clocks with no local implementation of .recalc ops, so CCF structures are populated with correct data during clock registration. Instead of enabling some of the registered clocks, flag them for CCF as critical. Introduce .is_enabled op using code that verifies hardware status of clock enablement, split out from implementation of .disable_unused op, so the latter is actually called by CCF for not requested but hardware enabled clocks. Add .round_rate ops where missing so .set_rate ops are called by CCF as expected. Since CCF allows parallel execution of .enable/.disable and .set_rate ops, protect registers shared among those groups of ops from concurrent access with spinlocks. Drop local debugfs support in favor of that provided by CCF. v2: flag tc2_ck as CLK_IS_CRITICAL (Aaro) v3: rebase on top of soc/omap1-multiplatform-5.18, - drop no longer needed includes from arch/arm/mach-omap1/io.c Signed-off-by: Janusz Krzysztofik <jmkrzyszt@gmail.com> Acked-by: Tony Lindgren <tony@atomide.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2022-04-10 06:07:57 -07:00
struct omap1_clk *clk = to_omap1_clk(hw);
const char *name = clk_hw_get_name(hw);
/* Clocks in the DSP domain need api_ck. Just assume bootloader
* has not enabled any DSP clocks */
if (clk->enable_reg == DSP_IDLECT2) {
ARM: OMAP1: clock: Convert to CCF OMAP1 still uses its own implementation of standard clock API defined in include/linux/clk.h. Internals of that implementation are not visible outside OMAP1 directory. As a consequence, device drivers are not able to register clocks potentially provided by peripheral devices. Drop OMAP1 implementation of the clock API and enable common clock framework. Modify the remaining low level code to be compatible with clock provider API and register the clocks with CCF. Move initialisation of clocks to omap1_timer_init() to avoid memory allocation issues at early setup phase from where omap1_init_early() is called. Register the clocks after initialization of clock I/O registers, local clock pointers used by OMAP1 clock ops, and local .rate fields of clocks with no local implementation of .recalc ops, so CCF structures are populated with correct data during clock registration. Instead of enabling some of the registered clocks, flag them for CCF as critical. Introduce .is_enabled op using code that verifies hardware status of clock enablement, split out from implementation of .disable_unused op, so the latter is actually called by CCF for not requested but hardware enabled clocks. Add .round_rate ops where missing so .set_rate ops are called by CCF as expected. Since CCF allows parallel execution of .enable/.disable and .set_rate ops, protect registers shared among those groups of ops from concurrent access with spinlocks. Drop local debugfs support in favor of that provided by CCF. v2: flag tc2_ck as CLK_IS_CRITICAL (Aaro) v3: rebase on top of soc/omap1-multiplatform-5.18, - drop no longer needed includes from arch/arm/mach-omap1/io.c Signed-off-by: Janusz Krzysztofik <jmkrzyszt@gmail.com> Acked-by: Tony Lindgren <tony@atomide.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2022-04-10 06:07:57 -07:00
pr_info("Skipping reset check for DSP domain clock \"%s\"\n", name);
return;
}
ARM: OMAP1: clock: Convert to CCF OMAP1 still uses its own implementation of standard clock API defined in include/linux/clk.h. Internals of that implementation are not visible outside OMAP1 directory. As a consequence, device drivers are not able to register clocks potentially provided by peripheral devices. Drop OMAP1 implementation of the clock API and enable common clock framework. Modify the remaining low level code to be compatible with clock provider API and register the clocks with CCF. Move initialisation of clocks to omap1_timer_init() to avoid memory allocation issues at early setup phase from where omap1_init_early() is called. Register the clocks after initialization of clock I/O registers, local clock pointers used by OMAP1 clock ops, and local .rate fields of clocks with no local implementation of .recalc ops, so CCF structures are populated with correct data during clock registration. Instead of enabling some of the registered clocks, flag them for CCF as critical. Introduce .is_enabled op using code that verifies hardware status of clock enablement, split out from implementation of .disable_unused op, so the latter is actually called by CCF for not requested but hardware enabled clocks. Add .round_rate ops where missing so .set_rate ops are called by CCF as expected. Since CCF allows parallel execution of .enable/.disable and .set_rate ops, protect registers shared among those groups of ops from concurrent access with spinlocks. Drop local debugfs support in favor of that provided by CCF. v2: flag tc2_ck as CLK_IS_CRITICAL (Aaro) v3: rebase on top of soc/omap1-multiplatform-5.18, - drop no longer needed includes from arch/arm/mach-omap1/io.c Signed-off-by: Janusz Krzysztofik <jmkrzyszt@gmail.com> Acked-by: Tony Lindgren <tony@atomide.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2022-04-10 06:07:57 -07:00
pr_info("Disabling unused clock \"%s\"... ", name);
omap1_clk_disable(hw);
printk(" done\n");
}
#endif
ARM: OMAP1: clock: Convert to CCF OMAP1 still uses its own implementation of standard clock API defined in include/linux/clk.h. Internals of that implementation are not visible outside OMAP1 directory. As a consequence, device drivers are not able to register clocks potentially provided by peripheral devices. Drop OMAP1 implementation of the clock API and enable common clock framework. Modify the remaining low level code to be compatible with clock provider API and register the clocks with CCF. Move initialisation of clocks to omap1_timer_init() to avoid memory allocation issues at early setup phase from where omap1_init_early() is called. Register the clocks after initialization of clock I/O registers, local clock pointers used by OMAP1 clock ops, and local .rate fields of clocks with no local implementation of .recalc ops, so CCF structures are populated with correct data during clock registration. Instead of enabling some of the registered clocks, flag them for CCF as critical. Introduce .is_enabled op using code that verifies hardware status of clock enablement, split out from implementation of .disable_unused op, so the latter is actually called by CCF for not requested but hardware enabled clocks. Add .round_rate ops where missing so .set_rate ops are called by CCF as expected. Since CCF allows parallel execution of .enable/.disable and .set_rate ops, protect registers shared among those groups of ops from concurrent access with spinlocks. Drop local debugfs support in favor of that provided by CCF. v2: flag tc2_ck as CLK_IS_CRITICAL (Aaro) v3: rebase on top of soc/omap1-multiplatform-5.18, - drop no longer needed includes from arch/arm/mach-omap1/io.c Signed-off-by: Janusz Krzysztofik <jmkrzyszt@gmail.com> Acked-by: Tony Lindgren <tony@atomide.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2022-04-10 06:07:57 -07:00
const struct clk_ops omap1_clk_gate_ops = {
.enable = omap1_clk_enable,
.disable = omap1_clk_disable,
.is_enabled = omap1_clk_is_enabled,
#ifdef CONFIG_OMAP_RESET_CLOCKS
.disable_unused = omap1_clk_disable_unused,
#endif
};
ARM: OMAP1: clock: Convert to CCF OMAP1 still uses its own implementation of standard clock API defined in include/linux/clk.h. Internals of that implementation are not visible outside OMAP1 directory. As a consequence, device drivers are not able to register clocks potentially provided by peripheral devices. Drop OMAP1 implementation of the clock API and enable common clock framework. Modify the remaining low level code to be compatible with clock provider API and register the clocks with CCF. Move initialisation of clocks to omap1_timer_init() to avoid memory allocation issues at early setup phase from where omap1_init_early() is called. Register the clocks after initialization of clock I/O registers, local clock pointers used by OMAP1 clock ops, and local .rate fields of clocks with no local implementation of .recalc ops, so CCF structures are populated with correct data during clock registration. Instead of enabling some of the registered clocks, flag them for CCF as critical. Introduce .is_enabled op using code that verifies hardware status of clock enablement, split out from implementation of .disable_unused op, so the latter is actually called by CCF for not requested but hardware enabled clocks. Add .round_rate ops where missing so .set_rate ops are called by CCF as expected. Since CCF allows parallel execution of .enable/.disable and .set_rate ops, protect registers shared among those groups of ops from concurrent access with spinlocks. Drop local debugfs support in favor of that provided by CCF. v2: flag tc2_ck as CLK_IS_CRITICAL (Aaro) v3: rebase on top of soc/omap1-multiplatform-5.18, - drop no longer needed includes from arch/arm/mach-omap1/io.c Signed-off-by: Janusz Krzysztofik <jmkrzyszt@gmail.com> Acked-by: Tony Lindgren <tony@atomide.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2022-04-10 06:07:57 -07:00
const struct clk_ops omap1_clk_rate_ops = {
.recalc_rate = omap1_clk_recalc_rate,
.round_rate = omap1_clk_round_rate,
.set_rate = omap1_clk_set_rate,
.init = omap1_clk_init_op,
};
ARM: OMAP1: clock: Convert to CCF OMAP1 still uses its own implementation of standard clock API defined in include/linux/clk.h. Internals of that implementation are not visible outside OMAP1 directory. As a consequence, device drivers are not able to register clocks potentially provided by peripheral devices. Drop OMAP1 implementation of the clock API and enable common clock framework. Modify the remaining low level code to be compatible with clock provider API and register the clocks with CCF. Move initialisation of clocks to omap1_timer_init() to avoid memory allocation issues at early setup phase from where omap1_init_early() is called. Register the clocks after initialization of clock I/O registers, local clock pointers used by OMAP1 clock ops, and local .rate fields of clocks with no local implementation of .recalc ops, so CCF structures are populated with correct data during clock registration. Instead of enabling some of the registered clocks, flag them for CCF as critical. Introduce .is_enabled op using code that verifies hardware status of clock enablement, split out from implementation of .disable_unused op, so the latter is actually called by CCF for not requested but hardware enabled clocks. Add .round_rate ops where missing so .set_rate ops are called by CCF as expected. Since CCF allows parallel execution of .enable/.disable and .set_rate ops, protect registers shared among those groups of ops from concurrent access with spinlocks. Drop local debugfs support in favor of that provided by CCF. v2: flag tc2_ck as CLK_IS_CRITICAL (Aaro) v3: rebase on top of soc/omap1-multiplatform-5.18, - drop no longer needed includes from arch/arm/mach-omap1/io.c Signed-off-by: Janusz Krzysztofik <jmkrzyszt@gmail.com> Acked-by: Tony Lindgren <tony@atomide.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2022-04-10 06:07:57 -07:00
const struct clk_ops omap1_clk_full_ops = {
.enable = omap1_clk_enable,
.disable = omap1_clk_disable,
.is_enabled = omap1_clk_is_enabled,
#ifdef CONFIG_OMAP_RESET_CLOCKS
.disable_unused = omap1_clk_disable_unused,
#endif
.recalc_rate = omap1_clk_recalc_rate,
.round_rate = omap1_clk_round_rate,
.set_rate = omap1_clk_set_rate,
.init = omap1_clk_init_op,
};
/*
* OMAP specific clock functions shared between omap1 and omap2
*/
/* Used for clocks that always have same value as the parent clock */
ARM: OMAP1: clock: Convert to CCF OMAP1 still uses its own implementation of standard clock API defined in include/linux/clk.h. Internals of that implementation are not visible outside OMAP1 directory. As a consequence, device drivers are not able to register clocks potentially provided by peripheral devices. Drop OMAP1 implementation of the clock API and enable common clock framework. Modify the remaining low level code to be compatible with clock provider API and register the clocks with CCF. Move initialisation of clocks to omap1_timer_init() to avoid memory allocation issues at early setup phase from where omap1_init_early() is called. Register the clocks after initialization of clock I/O registers, local clock pointers used by OMAP1 clock ops, and local .rate fields of clocks with no local implementation of .recalc ops, so CCF structures are populated with correct data during clock registration. Instead of enabling some of the registered clocks, flag them for CCF as critical. Introduce .is_enabled op using code that verifies hardware status of clock enablement, split out from implementation of .disable_unused op, so the latter is actually called by CCF for not requested but hardware enabled clocks. Add .round_rate ops where missing so .set_rate ops are called by CCF as expected. Since CCF allows parallel execution of .enable/.disable and .set_rate ops, protect registers shared among those groups of ops from concurrent access with spinlocks. Drop local debugfs support in favor of that provided by CCF. v2: flag tc2_ck as CLK_IS_CRITICAL (Aaro) v3: rebase on top of soc/omap1-multiplatform-5.18, - drop no longer needed includes from arch/arm/mach-omap1/io.c Signed-off-by: Janusz Krzysztofik <jmkrzyszt@gmail.com> Acked-by: Tony Lindgren <tony@atomide.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2022-04-10 06:07:57 -07:00
unsigned long followparent_recalc(struct omap1_clk *clk, unsigned long p_rate)
{
ARM: OMAP1: clock: Convert to CCF OMAP1 still uses its own implementation of standard clock API defined in include/linux/clk.h. Internals of that implementation are not visible outside OMAP1 directory. As a consequence, device drivers are not able to register clocks potentially provided by peripheral devices. Drop OMAP1 implementation of the clock API and enable common clock framework. Modify the remaining low level code to be compatible with clock provider API and register the clocks with CCF. Move initialisation of clocks to omap1_timer_init() to avoid memory allocation issues at early setup phase from where omap1_init_early() is called. Register the clocks after initialization of clock I/O registers, local clock pointers used by OMAP1 clock ops, and local .rate fields of clocks with no local implementation of .recalc ops, so CCF structures are populated with correct data during clock registration. Instead of enabling some of the registered clocks, flag them for CCF as critical. Introduce .is_enabled op using code that verifies hardware status of clock enablement, split out from implementation of .disable_unused op, so the latter is actually called by CCF for not requested but hardware enabled clocks. Add .round_rate ops where missing so .set_rate ops are called by CCF as expected. Since CCF allows parallel execution of .enable/.disable and .set_rate ops, protect registers shared among those groups of ops from concurrent access with spinlocks. Drop local debugfs support in favor of that provided by CCF. v2: flag tc2_ck as CLK_IS_CRITICAL (Aaro) v3: rebase on top of soc/omap1-multiplatform-5.18, - drop no longer needed includes from arch/arm/mach-omap1/io.c Signed-off-by: Janusz Krzysztofik <jmkrzyszt@gmail.com> Acked-by: Tony Lindgren <tony@atomide.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2022-04-10 06:07:57 -07:00
return p_rate;
}
/*
* Used for clocks that have the same value as the parent clock,
* divided by some factor
*/
ARM: OMAP1: clock: Convert to CCF OMAP1 still uses its own implementation of standard clock API defined in include/linux/clk.h. Internals of that implementation are not visible outside OMAP1 directory. As a consequence, device drivers are not able to register clocks potentially provided by peripheral devices. Drop OMAP1 implementation of the clock API and enable common clock framework. Modify the remaining low level code to be compatible with clock provider API and register the clocks with CCF. Move initialisation of clocks to omap1_timer_init() to avoid memory allocation issues at early setup phase from where omap1_init_early() is called. Register the clocks after initialization of clock I/O registers, local clock pointers used by OMAP1 clock ops, and local .rate fields of clocks with no local implementation of .recalc ops, so CCF structures are populated with correct data during clock registration. Instead of enabling some of the registered clocks, flag them for CCF as critical. Introduce .is_enabled op using code that verifies hardware status of clock enablement, split out from implementation of .disable_unused op, so the latter is actually called by CCF for not requested but hardware enabled clocks. Add .round_rate ops where missing so .set_rate ops are called by CCF as expected. Since CCF allows parallel execution of .enable/.disable and .set_rate ops, protect registers shared among those groups of ops from concurrent access with spinlocks. Drop local debugfs support in favor of that provided by CCF. v2: flag tc2_ck as CLK_IS_CRITICAL (Aaro) v3: rebase on top of soc/omap1-multiplatform-5.18, - drop no longer needed includes from arch/arm/mach-omap1/io.c Signed-off-by: Janusz Krzysztofik <jmkrzyszt@gmail.com> Acked-by: Tony Lindgren <tony@atomide.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2022-04-10 06:07:57 -07:00
unsigned long omap_fixed_divisor_recalc(struct omap1_clk *clk, unsigned long p_rate)
{
WARN_ON(!clk->fixed_div);
ARM: OMAP1: clock: Convert to CCF OMAP1 still uses its own implementation of standard clock API defined in include/linux/clk.h. Internals of that implementation are not visible outside OMAP1 directory. As a consequence, device drivers are not able to register clocks potentially provided by peripheral devices. Drop OMAP1 implementation of the clock API and enable common clock framework. Modify the remaining low level code to be compatible with clock provider API and register the clocks with CCF. Move initialisation of clocks to omap1_timer_init() to avoid memory allocation issues at early setup phase from where omap1_init_early() is called. Register the clocks after initialization of clock I/O registers, local clock pointers used by OMAP1 clock ops, and local .rate fields of clocks with no local implementation of .recalc ops, so CCF structures are populated with correct data during clock registration. Instead of enabling some of the registered clocks, flag them for CCF as critical. Introduce .is_enabled op using code that verifies hardware status of clock enablement, split out from implementation of .disable_unused op, so the latter is actually called by CCF for not requested but hardware enabled clocks. Add .round_rate ops where missing so .set_rate ops are called by CCF as expected. Since CCF allows parallel execution of .enable/.disable and .set_rate ops, protect registers shared among those groups of ops from concurrent access with spinlocks. Drop local debugfs support in favor of that provided by CCF. v2: flag tc2_ck as CLK_IS_CRITICAL (Aaro) v3: rebase on top of soc/omap1-multiplatform-5.18, - drop no longer needed includes from arch/arm/mach-omap1/io.c Signed-off-by: Janusz Krzysztofik <jmkrzyszt@gmail.com> Acked-by: Tony Lindgren <tony@atomide.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2022-04-10 06:07:57 -07:00
return p_rate / clk->fixed_div;
}
/* Propagate rate to children */
ARM: OMAP1: clock: Convert to CCF OMAP1 still uses its own implementation of standard clock API defined in include/linux/clk.h. Internals of that implementation are not visible outside OMAP1 directory. As a consequence, device drivers are not able to register clocks potentially provided by peripheral devices. Drop OMAP1 implementation of the clock API and enable common clock framework. Modify the remaining low level code to be compatible with clock provider API and register the clocks with CCF. Move initialisation of clocks to omap1_timer_init() to avoid memory allocation issues at early setup phase from where omap1_init_early() is called. Register the clocks after initialization of clock I/O registers, local clock pointers used by OMAP1 clock ops, and local .rate fields of clocks with no local implementation of .recalc ops, so CCF structures are populated with correct data during clock registration. Instead of enabling some of the registered clocks, flag them for CCF as critical. Introduce .is_enabled op using code that verifies hardware status of clock enablement, split out from implementation of .disable_unused op, so the latter is actually called by CCF for not requested but hardware enabled clocks. Add .round_rate ops where missing so .set_rate ops are called by CCF as expected. Since CCF allows parallel execution of .enable/.disable and .set_rate ops, protect registers shared among those groups of ops from concurrent access with spinlocks. Drop local debugfs support in favor of that provided by CCF. v2: flag tc2_ck as CLK_IS_CRITICAL (Aaro) v3: rebase on top of soc/omap1-multiplatform-5.18, - drop no longer needed includes from arch/arm/mach-omap1/io.c Signed-off-by: Janusz Krzysztofik <jmkrzyszt@gmail.com> Acked-by: Tony Lindgren <tony@atomide.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2022-04-10 06:07:57 -07:00
void propagate_rate(struct omap1_clk *tclk)
{
struct clk *clkp;
ARM: OMAP1: clock: Convert to CCF OMAP1 still uses its own implementation of standard clock API defined in include/linux/clk.h. Internals of that implementation are not visible outside OMAP1 directory. As a consequence, device drivers are not able to register clocks potentially provided by peripheral devices. Drop OMAP1 implementation of the clock API and enable common clock framework. Modify the remaining low level code to be compatible with clock provider API and register the clocks with CCF. Move initialisation of clocks to omap1_timer_init() to avoid memory allocation issues at early setup phase from where omap1_init_early() is called. Register the clocks after initialization of clock I/O registers, local clock pointers used by OMAP1 clock ops, and local .rate fields of clocks with no local implementation of .recalc ops, so CCF structures are populated with correct data during clock registration. Instead of enabling some of the registered clocks, flag them for CCF as critical. Introduce .is_enabled op using code that verifies hardware status of clock enablement, split out from implementation of .disable_unused op, so the latter is actually called by CCF for not requested but hardware enabled clocks. Add .round_rate ops where missing so .set_rate ops are called by CCF as expected. Since CCF allows parallel execution of .enable/.disable and .set_rate ops, protect registers shared among those groups of ops from concurrent access with spinlocks. Drop local debugfs support in favor of that provided by CCF. v2: flag tc2_ck as CLK_IS_CRITICAL (Aaro) v3: rebase on top of soc/omap1-multiplatform-5.18, - drop no longer needed includes from arch/arm/mach-omap1/io.c Signed-off-by: Janusz Krzysztofik <jmkrzyszt@gmail.com> Acked-by: Tony Lindgren <tony@atomide.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2022-04-10 06:07:57 -07:00
/* depend on CCF ability to recalculate new rates across whole clock subtree */
if (WARN_ON(!(clk_hw_get_flags(&tclk->hw) & CLK_GET_RATE_NOCACHE)))
return;
ARM: OMAP1: clock: Convert to CCF OMAP1 still uses its own implementation of standard clock API defined in include/linux/clk.h. Internals of that implementation are not visible outside OMAP1 directory. As a consequence, device drivers are not able to register clocks potentially provided by peripheral devices. Drop OMAP1 implementation of the clock API and enable common clock framework. Modify the remaining low level code to be compatible with clock provider API and register the clocks with CCF. Move initialisation of clocks to omap1_timer_init() to avoid memory allocation issues at early setup phase from where omap1_init_early() is called. Register the clocks after initialization of clock I/O registers, local clock pointers used by OMAP1 clock ops, and local .rate fields of clocks with no local implementation of .recalc ops, so CCF structures are populated with correct data during clock registration. Instead of enabling some of the registered clocks, flag them for CCF as critical. Introduce .is_enabled op using code that verifies hardware status of clock enablement, split out from implementation of .disable_unused op, so the latter is actually called by CCF for not requested but hardware enabled clocks. Add .round_rate ops where missing so .set_rate ops are called by CCF as expected. Since CCF allows parallel execution of .enable/.disable and .set_rate ops, protect registers shared among those groups of ops from concurrent access with spinlocks. Drop local debugfs support in favor of that provided by CCF. v2: flag tc2_ck as CLK_IS_CRITICAL (Aaro) v3: rebase on top of soc/omap1-multiplatform-5.18, - drop no longer needed includes from arch/arm/mach-omap1/io.c Signed-off-by: Janusz Krzysztofik <jmkrzyszt@gmail.com> Acked-by: Tony Lindgren <tony@atomide.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2022-04-10 06:07:57 -07:00
clkp = clk_get_sys(NULL, clk_hw_get_name(&tclk->hw));
if (WARN_ON(!clkp))
return;
ARM: OMAP1: clock: Convert to CCF OMAP1 still uses its own implementation of standard clock API defined in include/linux/clk.h. Internals of that implementation are not visible outside OMAP1 directory. As a consequence, device drivers are not able to register clocks potentially provided by peripheral devices. Drop OMAP1 implementation of the clock API and enable common clock framework. Modify the remaining low level code to be compatible with clock provider API and register the clocks with CCF. Move initialisation of clocks to omap1_timer_init() to avoid memory allocation issues at early setup phase from where omap1_init_early() is called. Register the clocks after initialization of clock I/O registers, local clock pointers used by OMAP1 clock ops, and local .rate fields of clocks with no local implementation of .recalc ops, so CCF structures are populated with correct data during clock registration. Instead of enabling some of the registered clocks, flag them for CCF as critical. Introduce .is_enabled op using code that verifies hardware status of clock enablement, split out from implementation of .disable_unused op, so the latter is actually called by CCF for not requested but hardware enabled clocks. Add .round_rate ops where missing so .set_rate ops are called by CCF as expected. Since CCF allows parallel execution of .enable/.disable and .set_rate ops, protect registers shared among those groups of ops from concurrent access with spinlocks. Drop local debugfs support in favor of that provided by CCF. v2: flag tc2_ck as CLK_IS_CRITICAL (Aaro) v3: rebase on top of soc/omap1-multiplatform-5.18, - drop no longer needed includes from arch/arm/mach-omap1/io.c Signed-off-by: Janusz Krzysztofik <jmkrzyszt@gmail.com> Acked-by: Tony Lindgren <tony@atomide.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2022-04-10 06:07:57 -07:00
clk_get_rate(clkp);
clk_put(clkp);
}
ARM: OMAP1: clock: Convert to CCF OMAP1 still uses its own implementation of standard clock API defined in include/linux/clk.h. Internals of that implementation are not visible outside OMAP1 directory. As a consequence, device drivers are not able to register clocks potentially provided by peripheral devices. Drop OMAP1 implementation of the clock API and enable common clock framework. Modify the remaining low level code to be compatible with clock provider API and register the clocks with CCF. Move initialisation of clocks to omap1_timer_init() to avoid memory allocation issues at early setup phase from where omap1_init_early() is called. Register the clocks after initialization of clock I/O registers, local clock pointers used by OMAP1 clock ops, and local .rate fields of clocks with no local implementation of .recalc ops, so CCF structures are populated with correct data during clock registration. Instead of enabling some of the registered clocks, flag them for CCF as critical. Introduce .is_enabled op using code that verifies hardware status of clock enablement, split out from implementation of .disable_unused op, so the latter is actually called by CCF for not requested but hardware enabled clocks. Add .round_rate ops where missing so .set_rate ops are called by CCF as expected. Since CCF allows parallel execution of .enable/.disable and .set_rate ops, protect registers shared among those groups of ops from concurrent access with spinlocks. Drop local debugfs support in favor of that provided by CCF. v2: flag tc2_ck as CLK_IS_CRITICAL (Aaro) v3: rebase on top of soc/omap1-multiplatform-5.18, - drop no longer needed includes from arch/arm/mach-omap1/io.c Signed-off-by: Janusz Krzysztofik <jmkrzyszt@gmail.com> Acked-by: Tony Lindgren <tony@atomide.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2022-04-10 06:07:57 -07:00
const struct clk_ops omap1_clk_null_ops = {
};
/*
* Dummy clock
*
* Used for clock aliases that are needed on some OMAPs, but not others
*/
ARM: OMAP1: clock: Convert to CCF OMAP1 still uses its own implementation of standard clock API defined in include/linux/clk.h. Internals of that implementation are not visible outside OMAP1 directory. As a consequence, device drivers are not able to register clocks potentially provided by peripheral devices. Drop OMAP1 implementation of the clock API and enable common clock framework. Modify the remaining low level code to be compatible with clock provider API and register the clocks with CCF. Move initialisation of clocks to omap1_timer_init() to avoid memory allocation issues at early setup phase from where omap1_init_early() is called. Register the clocks after initialization of clock I/O registers, local clock pointers used by OMAP1 clock ops, and local .rate fields of clocks with no local implementation of .recalc ops, so CCF structures are populated with correct data during clock registration. Instead of enabling some of the registered clocks, flag them for CCF as critical. Introduce .is_enabled op using code that verifies hardware status of clock enablement, split out from implementation of .disable_unused op, so the latter is actually called by CCF for not requested but hardware enabled clocks. Add .round_rate ops where missing so .set_rate ops are called by CCF as expected. Since CCF allows parallel execution of .enable/.disable and .set_rate ops, protect registers shared among those groups of ops from concurrent access with spinlocks. Drop local debugfs support in favor of that provided by CCF. v2: flag tc2_ck as CLK_IS_CRITICAL (Aaro) v3: rebase on top of soc/omap1-multiplatform-5.18, - drop no longer needed includes from arch/arm/mach-omap1/io.c Signed-off-by: Janusz Krzysztofik <jmkrzyszt@gmail.com> Acked-by: Tony Lindgren <tony@atomide.com> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
2022-04-10 06:07:57 -07:00
struct omap1_clk dummy_ck __refdata = {
.hw.init = CLK_HW_INIT_NO_PARENT("dummy", &omap1_clk_null_ops, 0),
};