OpenCloudOS-Kernel/drivers/clk/clk-qoriq.c

1286 lines
29 KiB
C
Raw Normal View History

/*
* Copyright 2013 Freescale Semiconductor, Inc.
*
* This program is free software; you can redistribute it and/or modify
* it under the terms of the GNU General Public License version 2 as
* published by the Free Software Foundation.
*
* clock driver for Freescale QorIQ SoCs.
*/
#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
clk: qoriq: Move chip-specific knowledge into driver The device tree should describe the chips (or chip-like subblocks) in the system, but it generally does not describe individual registers -- it should identify, rather than describe, a programming interface. This has not been the case with the QorIQ clockgen nodes. The knowledge of what each bit setting of CLKCnCSR means is encoded in three places (binding, pll node, and mux node), and the last also needs to know which options are valid on a particular chip. All three of these locations are considered stable ABI, making it difficult to fix mistakes (of which I have found several), much less refactor the abstraction to be able to address problems, limitations, or new chips. Under the current binding, a pll clock specifier of 2 means that the PLL is divided by 4 -- and the driver implements this, unless there happen to be four clock-output-names rather than 3, in which case it interprets it as PLL divided by 3. This does not appear in the binding documentation at all. That hack is now considered stable ABI. The current device tree nodes contain errors, such as saying that T1040 can set a core clock to PLL/4 when only PLL and PLL/2 are options. The current binding also ignores some restrictions on clock selection, such as p5020's requirement that if a core uses the "wrong" PLL, that PLL must be clocked lower than the "correct" PLL and be at most 80% of the rated CPU frequency. Possibly because of the lack of the ability to express such nuance in the binding, some valid options are omitted from the device trees, such as the ability on p4080 to run cores 0-3 from PLL3 and cores 4-7 from PLL1 (again, only if they are at most 80% of rated CPU frequency). This omission, combined with excessive caution in the cpufreq driver (addressed in a subsequent patch), means that currently on a 1500 MHz p4080 with typical PLL configuration, cpufreq can lower the frequency to 1200 MHz on half the CPUs and do nothing on the others. With this patchset, all CPUs can be lowered to 1200 MHz on a rev2 p4080, and on a rev3 p4080 half can be lowered to 750 MHz and the other half to 600 MHz. The current binding only deals with CPU clocks. To describe FMan in the device tree, we need to describe its clock. Some chips have additional muxes that work like the CPU muxes, but are not described in the device tree. Others require inspecting the Reset Control Word to determine which PLL is used. Rather than continue to extend this mess, replace it. Have the driver bind to the chip-specific clockgen compatible, and keep the detailed description of quirky chip variations in the driver, where it can be easily fixed, refactored, and extended. Older device trees will continue to work (including a workaround for old ls1021a device trees that are missing compatible and reg in the clockgen node, which even the old binding required). The pll/mux details in old device trees will be ignored, but "clocks" properties pointing at the old nodes will still work, and be directed at the corresponding new clock. Signed-off-by: Scott Wood <scottwood@freescale.com> Acked-by: Stephen Boyd <sboyd@codeaurora.org>
2015-09-20 12:29:54 +08:00
#include <linux/clk.h>
#include <linux/clk-provider.h>
clk: qoriq: Move chip-specific knowledge into driver The device tree should describe the chips (or chip-like subblocks) in the system, but it generally does not describe individual registers -- it should identify, rather than describe, a programming interface. This has not been the case with the QorIQ clockgen nodes. The knowledge of what each bit setting of CLKCnCSR means is encoded in three places (binding, pll node, and mux node), and the last also needs to know which options are valid on a particular chip. All three of these locations are considered stable ABI, making it difficult to fix mistakes (of which I have found several), much less refactor the abstraction to be able to address problems, limitations, or new chips. Under the current binding, a pll clock specifier of 2 means that the PLL is divided by 4 -- and the driver implements this, unless there happen to be four clock-output-names rather than 3, in which case it interprets it as PLL divided by 3. This does not appear in the binding documentation at all. That hack is now considered stable ABI. The current device tree nodes contain errors, such as saying that T1040 can set a core clock to PLL/4 when only PLL and PLL/2 are options. The current binding also ignores some restrictions on clock selection, such as p5020's requirement that if a core uses the "wrong" PLL, that PLL must be clocked lower than the "correct" PLL and be at most 80% of the rated CPU frequency. Possibly because of the lack of the ability to express such nuance in the binding, some valid options are omitted from the device trees, such as the ability on p4080 to run cores 0-3 from PLL3 and cores 4-7 from PLL1 (again, only if they are at most 80% of rated CPU frequency). This omission, combined with excessive caution in the cpufreq driver (addressed in a subsequent patch), means that currently on a 1500 MHz p4080 with typical PLL configuration, cpufreq can lower the frequency to 1200 MHz on half the CPUs and do nothing on the others. With this patchset, all CPUs can be lowered to 1200 MHz on a rev2 p4080, and on a rev3 p4080 half can be lowered to 750 MHz and the other half to 600 MHz. The current binding only deals with CPU clocks. To describe FMan in the device tree, we need to describe its clock. Some chips have additional muxes that work like the CPU muxes, but are not described in the device tree. Others require inspecting the Reset Control Word to determine which PLL is used. Rather than continue to extend this mess, replace it. Have the driver bind to the chip-specific clockgen compatible, and keep the detailed description of quirky chip variations in the driver, where it can be easily fixed, refactored, and extended. Older device trees will continue to work (including a workaround for old ls1021a device trees that are missing compatible and reg in the clockgen node, which even the old binding required). The pll/mux details in old device trees will be ignored, but "clocks" properties pointing at the old nodes will still work, and be directed at the corresponding new clock. Signed-off-by: Scott Wood <scottwood@freescale.com> Acked-by: Stephen Boyd <sboyd@codeaurora.org>
2015-09-20 12:29:54 +08:00
#include <linux/fsl/guts.h>
#include <linux/io.h>
#include <linux/kernel.h>
#include <linux/module.h>
#include <linux/of_address.h>
#include <linux/of_platform.h>
#include <linux/of.h>
#include <linux/slab.h>
clk: qoriq: Move chip-specific knowledge into driver The device tree should describe the chips (or chip-like subblocks) in the system, but it generally does not describe individual registers -- it should identify, rather than describe, a programming interface. This has not been the case with the QorIQ clockgen nodes. The knowledge of what each bit setting of CLKCnCSR means is encoded in three places (binding, pll node, and mux node), and the last also needs to know which options are valid on a particular chip. All three of these locations are considered stable ABI, making it difficult to fix mistakes (of which I have found several), much less refactor the abstraction to be able to address problems, limitations, or new chips. Under the current binding, a pll clock specifier of 2 means that the PLL is divided by 4 -- and the driver implements this, unless there happen to be four clock-output-names rather than 3, in which case it interprets it as PLL divided by 3. This does not appear in the binding documentation at all. That hack is now considered stable ABI. The current device tree nodes contain errors, such as saying that T1040 can set a core clock to PLL/4 when only PLL and PLL/2 are options. The current binding also ignores some restrictions on clock selection, such as p5020's requirement that if a core uses the "wrong" PLL, that PLL must be clocked lower than the "correct" PLL and be at most 80% of the rated CPU frequency. Possibly because of the lack of the ability to express such nuance in the binding, some valid options are omitted from the device trees, such as the ability on p4080 to run cores 0-3 from PLL3 and cores 4-7 from PLL1 (again, only if they are at most 80% of rated CPU frequency). This omission, combined with excessive caution in the cpufreq driver (addressed in a subsequent patch), means that currently on a 1500 MHz p4080 with typical PLL configuration, cpufreq can lower the frequency to 1200 MHz on half the CPUs and do nothing on the others. With this patchset, all CPUs can be lowered to 1200 MHz on a rev2 p4080, and on a rev3 p4080 half can be lowered to 750 MHz and the other half to 600 MHz. The current binding only deals with CPU clocks. To describe FMan in the device tree, we need to describe its clock. Some chips have additional muxes that work like the CPU muxes, but are not described in the device tree. Others require inspecting the Reset Control Word to determine which PLL is used. Rather than continue to extend this mess, replace it. Have the driver bind to the chip-specific clockgen compatible, and keep the detailed description of quirky chip variations in the driver, where it can be easily fixed, refactored, and extended. Older device trees will continue to work (including a workaround for old ls1021a device trees that are missing compatible and reg in the clockgen node, which even the old binding required). The pll/mux details in old device trees will be ignored, but "clocks" properties pointing at the old nodes will still work, and be directed at the corresponding new clock. Signed-off-by: Scott Wood <scottwood@freescale.com> Acked-by: Stephen Boyd <sboyd@codeaurora.org>
2015-09-20 12:29:54 +08:00
#define PLL_DIV1 0
#define PLL_DIV2 1
#define PLL_DIV3 2
#define PLL_DIV4 3
#define PLATFORM_PLL 0
#define CGA_PLL1 1
#define CGA_PLL2 2
#define CGA_PLL3 3
#define CGA_PLL4 4 /* only on clockgen-1.0, which lacks CGB */
#define CGB_PLL1 4
#define CGB_PLL2 5
struct clockgen_pll_div {
struct clk *clk;
char name[32];
};
struct clockgen_pll {
struct clockgen_pll_div div[4];
};
#define CLKSEL_VALID 1
#define CLKSEL_80PCT 2 /* Only allowed if PLL <= 80% of max cpu freq */
struct clockgen_sourceinfo {
u32 flags; /* CLKSEL_xxx */
int pll; /* CGx_PLLn */
int div; /* PLL_DIVn */
};
#define NUM_MUX_PARENTS 16
struct clockgen_muxinfo {
struct clockgen_sourceinfo clksel[NUM_MUX_PARENTS];
};
#define NUM_HWACCEL 5
#define NUM_CMUX 8
struct clockgen;
/*
* cmux freq must be >= platform pll.
* If not set, cmux freq must be >= platform pll/2
*/
#define CG_CMUX_GE_PLAT 1
clk: qoriq: Move chip-specific knowledge into driver The device tree should describe the chips (or chip-like subblocks) in the system, but it generally does not describe individual registers -- it should identify, rather than describe, a programming interface. This has not been the case with the QorIQ clockgen nodes. The knowledge of what each bit setting of CLKCnCSR means is encoded in three places (binding, pll node, and mux node), and the last also needs to know which options are valid on a particular chip. All three of these locations are considered stable ABI, making it difficult to fix mistakes (of which I have found several), much less refactor the abstraction to be able to address problems, limitations, or new chips. Under the current binding, a pll clock specifier of 2 means that the PLL is divided by 4 -- and the driver implements this, unless there happen to be four clock-output-names rather than 3, in which case it interprets it as PLL divided by 3. This does not appear in the binding documentation at all. That hack is now considered stable ABI. The current device tree nodes contain errors, such as saying that T1040 can set a core clock to PLL/4 when only PLL and PLL/2 are options. The current binding also ignores some restrictions on clock selection, such as p5020's requirement that if a core uses the "wrong" PLL, that PLL must be clocked lower than the "correct" PLL and be at most 80% of the rated CPU frequency. Possibly because of the lack of the ability to express such nuance in the binding, some valid options are omitted from the device trees, such as the ability on p4080 to run cores 0-3 from PLL3 and cores 4-7 from PLL1 (again, only if they are at most 80% of rated CPU frequency). This omission, combined with excessive caution in the cpufreq driver (addressed in a subsequent patch), means that currently on a 1500 MHz p4080 with typical PLL configuration, cpufreq can lower the frequency to 1200 MHz on half the CPUs and do nothing on the others. With this patchset, all CPUs can be lowered to 1200 MHz on a rev2 p4080, and on a rev3 p4080 half can be lowered to 750 MHz and the other half to 600 MHz. The current binding only deals with CPU clocks. To describe FMan in the device tree, we need to describe its clock. Some chips have additional muxes that work like the CPU muxes, but are not described in the device tree. Others require inspecting the Reset Control Word to determine which PLL is used. Rather than continue to extend this mess, replace it. Have the driver bind to the chip-specific clockgen compatible, and keep the detailed description of quirky chip variations in the driver, where it can be easily fixed, refactored, and extended. Older device trees will continue to work (including a workaround for old ls1021a device trees that are missing compatible and reg in the clockgen node, which even the old binding required). The pll/mux details in old device trees will be ignored, but "clocks" properties pointing at the old nodes will still work, and be directed at the corresponding new clock. Signed-off-by: Scott Wood <scottwood@freescale.com> Acked-by: Stephen Boyd <sboyd@codeaurora.org>
2015-09-20 12:29:54 +08:00
#define CG_PLL_8BIT 2 /* PLLCnGSR[CFG] is 8 bits, not 6 */
#define CG_VER3 4 /* version 3 cg: reg layout different */
#define CG_LITTLE_ENDIAN 8
clk: qoriq: Move chip-specific knowledge into driver The device tree should describe the chips (or chip-like subblocks) in the system, but it generally does not describe individual registers -- it should identify, rather than describe, a programming interface. This has not been the case with the QorIQ clockgen nodes. The knowledge of what each bit setting of CLKCnCSR means is encoded in three places (binding, pll node, and mux node), and the last also needs to know which options are valid on a particular chip. All three of these locations are considered stable ABI, making it difficult to fix mistakes (of which I have found several), much less refactor the abstraction to be able to address problems, limitations, or new chips. Under the current binding, a pll clock specifier of 2 means that the PLL is divided by 4 -- and the driver implements this, unless there happen to be four clock-output-names rather than 3, in which case it interprets it as PLL divided by 3. This does not appear in the binding documentation at all. That hack is now considered stable ABI. The current device tree nodes contain errors, such as saying that T1040 can set a core clock to PLL/4 when only PLL and PLL/2 are options. The current binding also ignores some restrictions on clock selection, such as p5020's requirement that if a core uses the "wrong" PLL, that PLL must be clocked lower than the "correct" PLL and be at most 80% of the rated CPU frequency. Possibly because of the lack of the ability to express such nuance in the binding, some valid options are omitted from the device trees, such as the ability on p4080 to run cores 0-3 from PLL3 and cores 4-7 from PLL1 (again, only if they are at most 80% of rated CPU frequency). This omission, combined with excessive caution in the cpufreq driver (addressed in a subsequent patch), means that currently on a 1500 MHz p4080 with typical PLL configuration, cpufreq can lower the frequency to 1200 MHz on half the CPUs and do nothing on the others. With this patchset, all CPUs can be lowered to 1200 MHz on a rev2 p4080, and on a rev3 p4080 half can be lowered to 750 MHz and the other half to 600 MHz. The current binding only deals with CPU clocks. To describe FMan in the device tree, we need to describe its clock. Some chips have additional muxes that work like the CPU muxes, but are not described in the device tree. Others require inspecting the Reset Control Word to determine which PLL is used. Rather than continue to extend this mess, replace it. Have the driver bind to the chip-specific clockgen compatible, and keep the detailed description of quirky chip variations in the driver, where it can be easily fixed, refactored, and extended. Older device trees will continue to work (including a workaround for old ls1021a device trees that are missing compatible and reg in the clockgen node, which even the old binding required). The pll/mux details in old device trees will be ignored, but "clocks" properties pointing at the old nodes will still work, and be directed at the corresponding new clock. Signed-off-by: Scott Wood <scottwood@freescale.com> Acked-by: Stephen Boyd <sboyd@codeaurora.org>
2015-09-20 12:29:54 +08:00
struct clockgen_chipinfo {
const char *compat, *guts_compat;
const struct clockgen_muxinfo *cmux_groups[2];
const struct clockgen_muxinfo *hwaccel[NUM_HWACCEL];
void (*init_periph)(struct clockgen *cg);
int cmux_to_group[NUM_CMUX]; /* -1 terminates if fewer than NUM_CMUX */
u32 pll_mask; /* 1 << n bit set if PLL n is valid */
u32 flags; /* CG_xxx */
};
struct clockgen {
struct device_node *node;
void __iomem *regs;
struct clockgen_chipinfo info; /* mutable copy */
struct clk *sysclk;
struct clockgen_pll pll[6];
struct clk *cmux[NUM_CMUX];
struct clk *hwaccel[NUM_HWACCEL];
struct clk *fman[2];
struct ccsr_guts __iomem *guts;
};
static struct clockgen clockgen;
static void cg_out(struct clockgen *cg, u32 val, u32 __iomem *reg)
{
if (cg->info.flags & CG_LITTLE_ENDIAN)
iowrite32(val, reg);
else
iowrite32be(val, reg);
}
static u32 cg_in(struct clockgen *cg, u32 __iomem *reg)
{
u32 val;
if (cg->info.flags & CG_LITTLE_ENDIAN)
val = ioread32(reg);
else
val = ioread32be(reg);
return val;
}
clk: qoriq: Move chip-specific knowledge into driver The device tree should describe the chips (or chip-like subblocks) in the system, but it generally does not describe individual registers -- it should identify, rather than describe, a programming interface. This has not been the case with the QorIQ clockgen nodes. The knowledge of what each bit setting of CLKCnCSR means is encoded in three places (binding, pll node, and mux node), and the last also needs to know which options are valid on a particular chip. All three of these locations are considered stable ABI, making it difficult to fix mistakes (of which I have found several), much less refactor the abstraction to be able to address problems, limitations, or new chips. Under the current binding, a pll clock specifier of 2 means that the PLL is divided by 4 -- and the driver implements this, unless there happen to be four clock-output-names rather than 3, in which case it interprets it as PLL divided by 3. This does not appear in the binding documentation at all. That hack is now considered stable ABI. The current device tree nodes contain errors, such as saying that T1040 can set a core clock to PLL/4 when only PLL and PLL/2 are options. The current binding also ignores some restrictions on clock selection, such as p5020's requirement that if a core uses the "wrong" PLL, that PLL must be clocked lower than the "correct" PLL and be at most 80% of the rated CPU frequency. Possibly because of the lack of the ability to express such nuance in the binding, some valid options are omitted from the device trees, such as the ability on p4080 to run cores 0-3 from PLL3 and cores 4-7 from PLL1 (again, only if they are at most 80% of rated CPU frequency). This omission, combined with excessive caution in the cpufreq driver (addressed in a subsequent patch), means that currently on a 1500 MHz p4080 with typical PLL configuration, cpufreq can lower the frequency to 1200 MHz on half the CPUs and do nothing on the others. With this patchset, all CPUs can be lowered to 1200 MHz on a rev2 p4080, and on a rev3 p4080 half can be lowered to 750 MHz and the other half to 600 MHz. The current binding only deals with CPU clocks. To describe FMan in the device tree, we need to describe its clock. Some chips have additional muxes that work like the CPU muxes, but are not described in the device tree. Others require inspecting the Reset Control Word to determine which PLL is used. Rather than continue to extend this mess, replace it. Have the driver bind to the chip-specific clockgen compatible, and keep the detailed description of quirky chip variations in the driver, where it can be easily fixed, refactored, and extended. Older device trees will continue to work (including a workaround for old ls1021a device trees that are missing compatible and reg in the clockgen node, which even the old binding required). The pll/mux details in old device trees will be ignored, but "clocks" properties pointing at the old nodes will still work, and be directed at the corresponding new clock. Signed-off-by: Scott Wood <scottwood@freescale.com> Acked-by: Stephen Boyd <sboyd@codeaurora.org>
2015-09-20 12:29:54 +08:00
static const struct clockgen_muxinfo p2041_cmux_grp1 = {
{
[0] = { CLKSEL_VALID, CGA_PLL1, PLL_DIV1 },
[1] = { CLKSEL_VALID, CGA_PLL1, PLL_DIV2 },
[4] = { CLKSEL_VALID, CGA_PLL2, PLL_DIV1 },
}
};
static const struct clockgen_muxinfo p2041_cmux_grp2 = {
{
[0] = { CLKSEL_VALID, CGA_PLL1, PLL_DIV1 },
[4] = { CLKSEL_VALID, CGA_PLL2, PLL_DIV1 },
[5] = { CLKSEL_VALID, CGA_PLL2, PLL_DIV2 },
clk: qoriq: Move chip-specific knowledge into driver The device tree should describe the chips (or chip-like subblocks) in the system, but it generally does not describe individual registers -- it should identify, rather than describe, a programming interface. This has not been the case with the QorIQ clockgen nodes. The knowledge of what each bit setting of CLKCnCSR means is encoded in three places (binding, pll node, and mux node), and the last also needs to know which options are valid on a particular chip. All three of these locations are considered stable ABI, making it difficult to fix mistakes (of which I have found several), much less refactor the abstraction to be able to address problems, limitations, or new chips. Under the current binding, a pll clock specifier of 2 means that the PLL is divided by 4 -- and the driver implements this, unless there happen to be four clock-output-names rather than 3, in which case it interprets it as PLL divided by 3. This does not appear in the binding documentation at all. That hack is now considered stable ABI. The current device tree nodes contain errors, such as saying that T1040 can set a core clock to PLL/4 when only PLL and PLL/2 are options. The current binding also ignores some restrictions on clock selection, such as p5020's requirement that if a core uses the "wrong" PLL, that PLL must be clocked lower than the "correct" PLL and be at most 80% of the rated CPU frequency. Possibly because of the lack of the ability to express such nuance in the binding, some valid options are omitted from the device trees, such as the ability on p4080 to run cores 0-3 from PLL3 and cores 4-7 from PLL1 (again, only if they are at most 80% of rated CPU frequency). This omission, combined with excessive caution in the cpufreq driver (addressed in a subsequent patch), means that currently on a 1500 MHz p4080 with typical PLL configuration, cpufreq can lower the frequency to 1200 MHz on half the CPUs and do nothing on the others. With this patchset, all CPUs can be lowered to 1200 MHz on a rev2 p4080, and on a rev3 p4080 half can be lowered to 750 MHz and the other half to 600 MHz. The current binding only deals with CPU clocks. To describe FMan in the device tree, we need to describe its clock. Some chips have additional muxes that work like the CPU muxes, but are not described in the device tree. Others require inspecting the Reset Control Word to determine which PLL is used. Rather than continue to extend this mess, replace it. Have the driver bind to the chip-specific clockgen compatible, and keep the detailed description of quirky chip variations in the driver, where it can be easily fixed, refactored, and extended. Older device trees will continue to work (including a workaround for old ls1021a device trees that are missing compatible and reg in the clockgen node, which even the old binding required). The pll/mux details in old device trees will be ignored, but "clocks" properties pointing at the old nodes will still work, and be directed at the corresponding new clock. Signed-off-by: Scott Wood <scottwood@freescale.com> Acked-by: Stephen Boyd <sboyd@codeaurora.org>
2015-09-20 12:29:54 +08:00
}
};
static const struct clockgen_muxinfo p5020_cmux_grp1 = {
{
[0] = { CLKSEL_VALID, CGA_PLL1, PLL_DIV1 },
[1] = { CLKSEL_VALID, CGA_PLL1, PLL_DIV2 },
[4] = { CLKSEL_VALID | CLKSEL_80PCT, CGA_PLL2, PLL_DIV1 },
}
};
static const struct clockgen_muxinfo p5020_cmux_grp2 = {
{
[0] = { CLKSEL_VALID | CLKSEL_80PCT, CGA_PLL1, PLL_DIV1 },
[4] = { CLKSEL_VALID, CGA_PLL2, PLL_DIV1 },
[5] = { CLKSEL_VALID, CGA_PLL2, PLL_DIV2 },
}
};
static const struct clockgen_muxinfo p5040_cmux_grp1 = {
{
[0] = { CLKSEL_VALID, CGA_PLL1, PLL_DIV1 },
[1] = { CLKSEL_VALID, CGA_PLL1, PLL_DIV2 },
[4] = { CLKSEL_VALID | CLKSEL_80PCT, CGA_PLL2, PLL_DIV1 },
[5] = { CLKSEL_VALID | CLKSEL_80PCT, CGA_PLL2, PLL_DIV2 },
}
};
static const struct clockgen_muxinfo p5040_cmux_grp2 = {
{
[0] = { CLKSEL_VALID | CLKSEL_80PCT, CGA_PLL1, PLL_DIV1 },
[1] = { CLKSEL_VALID | CLKSEL_80PCT, CGA_PLL1, PLL_DIV2 },
[4] = { CLKSEL_VALID, CGA_PLL2, PLL_DIV1 },
[5] = { CLKSEL_VALID, CGA_PLL2, PLL_DIV2 },
}
};
static const struct clockgen_muxinfo p4080_cmux_grp1 = {
{
[0] = { CLKSEL_VALID, CGA_PLL1, PLL_DIV1 },
[1] = { CLKSEL_VALID, CGA_PLL1, PLL_DIV2 },
[4] = { CLKSEL_VALID, CGA_PLL2, PLL_DIV1 },
[5] = { CLKSEL_VALID, CGA_PLL2, PLL_DIV2 },
[8] = { CLKSEL_VALID | CLKSEL_80PCT, CGA_PLL3, PLL_DIV1 },
}
};
static const struct clockgen_muxinfo p4080_cmux_grp2 = {
{
[0] = { CLKSEL_VALID | CLKSEL_80PCT, CGA_PLL1, PLL_DIV1 },
[8] = { CLKSEL_VALID, CGA_PLL3, PLL_DIV1 },
[9] = { CLKSEL_VALID, CGA_PLL3, PLL_DIV2 },
[12] = { CLKSEL_VALID, CGA_PLL4, PLL_DIV1 },
[13] = { CLKSEL_VALID, CGA_PLL4, PLL_DIV2 },
}
};
static const struct clockgen_muxinfo t1023_cmux = {
{
[0] = { CLKSEL_VALID, CGA_PLL1, PLL_DIV1 },
[1] = { CLKSEL_VALID, CGA_PLL1, PLL_DIV2 },
}
};
static const struct clockgen_muxinfo t1040_cmux = {
{
[0] = { CLKSEL_VALID, CGA_PLL1, PLL_DIV1 },
[1] = { CLKSEL_VALID, CGA_PLL1, PLL_DIV2 },
[4] = { CLKSEL_VALID, CGA_PLL2, PLL_DIV1 },
[5] = { CLKSEL_VALID, CGA_PLL2, PLL_DIV2 },
}
};
static const struct clockgen_muxinfo clockgen2_cmux_cga = {
{
{ CLKSEL_VALID, CGA_PLL1, PLL_DIV1 },
{ CLKSEL_VALID, CGA_PLL1, PLL_DIV2 },
{ CLKSEL_VALID, CGA_PLL1, PLL_DIV4 },
{},
{ CLKSEL_VALID, CGA_PLL2, PLL_DIV1 },
{ CLKSEL_VALID, CGA_PLL2, PLL_DIV2 },
{ CLKSEL_VALID, CGA_PLL2, PLL_DIV4 },
{},
{ CLKSEL_VALID, CGA_PLL3, PLL_DIV1 },
{ CLKSEL_VALID, CGA_PLL3, PLL_DIV2 },
{ CLKSEL_VALID, CGA_PLL3, PLL_DIV4 },
},
};
static const struct clockgen_muxinfo clockgen2_cmux_cga12 = {
{
{ CLKSEL_VALID, CGA_PLL1, PLL_DIV1 },
{ CLKSEL_VALID, CGA_PLL1, PLL_DIV2 },
{ CLKSEL_VALID, CGA_PLL1, PLL_DIV4 },
{},
{ CLKSEL_VALID, CGA_PLL2, PLL_DIV1 },
{ CLKSEL_VALID, CGA_PLL2, PLL_DIV2 },
{ CLKSEL_VALID, CGA_PLL2, PLL_DIV4 },
},
};
static const struct clockgen_muxinfo clockgen2_cmux_cgb = {
{
{ CLKSEL_VALID, CGB_PLL1, PLL_DIV1 },
{ CLKSEL_VALID, CGB_PLL1, PLL_DIV2 },
{ CLKSEL_VALID, CGB_PLL1, PLL_DIV4 },
{},
{ CLKSEL_VALID, CGB_PLL2, PLL_DIV1 },
{ CLKSEL_VALID, CGB_PLL2, PLL_DIV2 },
{ CLKSEL_VALID, CGB_PLL2, PLL_DIV4 },
},
};
static const struct clockgen_muxinfo ls1043a_hwa1 = {
{
{},
{},
{ CLKSEL_VALID, CGA_PLL1, PLL_DIV2 },
{ CLKSEL_VALID, CGA_PLL1, PLL_DIV3 },
{},
{},
{ CLKSEL_VALID, CGA_PLL2, PLL_DIV2 },
{ CLKSEL_VALID, CGA_PLL2, PLL_DIV3 },
},
};
static const struct clockgen_muxinfo ls1043a_hwa2 = {
{
{},
{ CLKSEL_VALID, CGA_PLL2, PLL_DIV1 },
{},
{ CLKSEL_VALID, CGA_PLL2, PLL_DIV3 },
},
};
clk: qoriq: Move chip-specific knowledge into driver The device tree should describe the chips (or chip-like subblocks) in the system, but it generally does not describe individual registers -- it should identify, rather than describe, a programming interface. This has not been the case with the QorIQ clockgen nodes. The knowledge of what each bit setting of CLKCnCSR means is encoded in three places (binding, pll node, and mux node), and the last also needs to know which options are valid on a particular chip. All three of these locations are considered stable ABI, making it difficult to fix mistakes (of which I have found several), much less refactor the abstraction to be able to address problems, limitations, or new chips. Under the current binding, a pll clock specifier of 2 means that the PLL is divided by 4 -- and the driver implements this, unless there happen to be four clock-output-names rather than 3, in which case it interprets it as PLL divided by 3. This does not appear in the binding documentation at all. That hack is now considered stable ABI. The current device tree nodes contain errors, such as saying that T1040 can set a core clock to PLL/4 when only PLL and PLL/2 are options. The current binding also ignores some restrictions on clock selection, such as p5020's requirement that if a core uses the "wrong" PLL, that PLL must be clocked lower than the "correct" PLL and be at most 80% of the rated CPU frequency. Possibly because of the lack of the ability to express such nuance in the binding, some valid options are omitted from the device trees, such as the ability on p4080 to run cores 0-3 from PLL3 and cores 4-7 from PLL1 (again, only if they are at most 80% of rated CPU frequency). This omission, combined with excessive caution in the cpufreq driver (addressed in a subsequent patch), means that currently on a 1500 MHz p4080 with typical PLL configuration, cpufreq can lower the frequency to 1200 MHz on half the CPUs and do nothing on the others. With this patchset, all CPUs can be lowered to 1200 MHz on a rev2 p4080, and on a rev3 p4080 half can be lowered to 750 MHz and the other half to 600 MHz. The current binding only deals with CPU clocks. To describe FMan in the device tree, we need to describe its clock. Some chips have additional muxes that work like the CPU muxes, but are not described in the device tree. Others require inspecting the Reset Control Word to determine which PLL is used. Rather than continue to extend this mess, replace it. Have the driver bind to the chip-specific clockgen compatible, and keep the detailed description of quirky chip variations in the driver, where it can be easily fixed, refactored, and extended. Older device trees will continue to work (including a workaround for old ls1021a device trees that are missing compatible and reg in the clockgen node, which even the old binding required). The pll/mux details in old device trees will be ignored, but "clocks" properties pointing at the old nodes will still work, and be directed at the corresponding new clock. Signed-off-by: Scott Wood <scottwood@freescale.com> Acked-by: Stephen Boyd <sboyd@codeaurora.org>
2015-09-20 12:29:54 +08:00
static const struct clockgen_muxinfo t1023_hwa1 = {
{
{},
{ CLKSEL_VALID, CGA_PLL1, PLL_DIV1 },
{ CLKSEL_VALID, CGA_PLL1, PLL_DIV2 },
{ CLKSEL_VALID, CGA_PLL1, PLL_DIV3 },
},
};
static const struct clockgen_muxinfo t1023_hwa2 = {
{
[6] = { CLKSEL_VALID, CGA_PLL1, PLL_DIV2 },
},
};
static const struct clockgen_muxinfo t2080_hwa1 = {
{
{},
{ CLKSEL_VALID, CGA_PLL1, PLL_DIV1 },
{ CLKSEL_VALID, CGA_PLL1, PLL_DIV2 },
{ CLKSEL_VALID, CGA_PLL1, PLL_DIV3 },
{ CLKSEL_VALID, CGA_PLL1, PLL_DIV4 },
{ CLKSEL_VALID, PLATFORM_PLL, PLL_DIV1 },
{ CLKSEL_VALID, CGA_PLL2, PLL_DIV2 },
{ CLKSEL_VALID, CGA_PLL2, PLL_DIV3 },
},
};
static const struct clockgen_muxinfo t2080_hwa2 = {
{
{},
{ CLKSEL_VALID, CGA_PLL2, PLL_DIV1 },
{ CLKSEL_VALID, CGA_PLL2, PLL_DIV2 },
{ CLKSEL_VALID, CGA_PLL2, PLL_DIV3 },
{ CLKSEL_VALID, CGA_PLL2, PLL_DIV4 },
{ CLKSEL_VALID, PLATFORM_PLL, PLL_DIV1 },
{ CLKSEL_VALID, CGA_PLL1, PLL_DIV2 },
{ CLKSEL_VALID, CGA_PLL1, PLL_DIV3 },
},
};
static const struct clockgen_muxinfo t4240_hwa1 = {
{
{ CLKSEL_VALID, PLATFORM_PLL, PLL_DIV2 },
{ CLKSEL_VALID, CGA_PLL1, PLL_DIV1 },
{ CLKSEL_VALID, CGA_PLL1, PLL_DIV2 },
{ CLKSEL_VALID, CGA_PLL1, PLL_DIV3 },
{ CLKSEL_VALID, CGA_PLL1, PLL_DIV4 },
{},
{ CLKSEL_VALID, CGA_PLL2, PLL_DIV2 },
{ CLKSEL_VALID, CGA_PLL2, PLL_DIV3 },
},
};
static const struct clockgen_muxinfo t4240_hwa4 = {
{
[2] = { CLKSEL_VALID, CGB_PLL1, PLL_DIV2 },
[3] = { CLKSEL_VALID, CGB_PLL1, PLL_DIV3 },
[4] = { CLKSEL_VALID, CGB_PLL1, PLL_DIV4 },
[5] = { CLKSEL_VALID, PLATFORM_PLL, PLL_DIV1 },
[6] = { CLKSEL_VALID, CGB_PLL2, PLL_DIV2 },
},
};
static const struct clockgen_muxinfo t4240_hwa5 = {
{
[2] = { CLKSEL_VALID, CGB_PLL2, PLL_DIV2 },
[3] = { CLKSEL_VALID, CGB_PLL2, PLL_DIV3 },
[4] = { CLKSEL_VALID, CGB_PLL2, PLL_DIV4 },
[5] = { CLKSEL_VALID, PLATFORM_PLL, PLL_DIV1 },
[6] = { CLKSEL_VALID, CGB_PLL1, PLL_DIV2 },
[7] = { CLKSEL_VALID, CGB_PLL1, PLL_DIV3 },
},
};
#define RCWSR7_FM1_CLK_SEL 0x40000000
#define RCWSR7_FM2_CLK_SEL 0x20000000
#define RCWSR7_HWA_ASYNC_DIV 0x04000000
static void __init p2041_init_periph(struct clockgen *cg)
{
u32 reg;
reg = ioread32be(&cg->guts->rcwsr[7]);
if (reg & RCWSR7_FM1_CLK_SEL)
cg->fman[0] = cg->pll[CGA_PLL2].div[PLL_DIV2].clk;
else
cg->fman[0] = cg->pll[PLATFORM_PLL].div[PLL_DIV2].clk;
}
static void __init p4080_init_periph(struct clockgen *cg)
{
u32 reg;
reg = ioread32be(&cg->guts->rcwsr[7]);
if (reg & RCWSR7_FM1_CLK_SEL)
cg->fman[0] = cg->pll[CGA_PLL3].div[PLL_DIV2].clk;
else
cg->fman[0] = cg->pll[PLATFORM_PLL].div[PLL_DIV2].clk;
if (reg & RCWSR7_FM2_CLK_SEL)
cg->fman[1] = cg->pll[CGA_PLL3].div[PLL_DIV2].clk;
else
cg->fman[1] = cg->pll[PLATFORM_PLL].div[PLL_DIV2].clk;
}
static void __init p5020_init_periph(struct clockgen *cg)
{
u32 reg;
int div = PLL_DIV2;
reg = ioread32be(&cg->guts->rcwsr[7]);
if (reg & RCWSR7_HWA_ASYNC_DIV)
div = PLL_DIV4;
if (reg & RCWSR7_FM1_CLK_SEL)
cg->fman[0] = cg->pll[CGA_PLL2].div[div].clk;
else
cg->fman[0] = cg->pll[PLATFORM_PLL].div[PLL_DIV2].clk;
}
static void __init p5040_init_periph(struct clockgen *cg)
{
u32 reg;
int div = PLL_DIV2;
reg = ioread32be(&cg->guts->rcwsr[7]);
if (reg & RCWSR7_HWA_ASYNC_DIV)
div = PLL_DIV4;
if (reg & RCWSR7_FM1_CLK_SEL)
cg->fman[0] = cg->pll[CGA_PLL3].div[div].clk;
else
cg->fman[0] = cg->pll[PLATFORM_PLL].div[PLL_DIV2].clk;
if (reg & RCWSR7_FM2_CLK_SEL)
cg->fman[1] = cg->pll[CGA_PLL3].div[div].clk;
else
cg->fman[1] = cg->pll[PLATFORM_PLL].div[PLL_DIV2].clk;
}
static void __init t1023_init_periph(struct clockgen *cg)
{
cg->fman[0] = cg->hwaccel[1];
}
static void __init t1040_init_periph(struct clockgen *cg)
{
cg->fman[0] = cg->pll[PLATFORM_PLL].div[PLL_DIV1].clk;
}
static void __init t2080_init_periph(struct clockgen *cg)
{
cg->fman[0] = cg->hwaccel[0];
}
static void __init t4240_init_periph(struct clockgen *cg)
{
cg->fman[0] = cg->hwaccel[3];
cg->fman[1] = cg->hwaccel[4];
}
static const struct clockgen_chipinfo chipinfo[] = {
{
.compat = "fsl,b4420-clockgen",
.guts_compat = "fsl,b4860-device-config",
.init_periph = t2080_init_periph,
.cmux_groups = {
&clockgen2_cmux_cga12, &clockgen2_cmux_cgb
},
.hwaccel = {
&t2080_hwa1
},
.cmux_to_group = {
0, 1, 1, 1, -1
},
.pll_mask = 0x3f,
.flags = CG_PLL_8BIT,
},
{
.compat = "fsl,b4860-clockgen",
.guts_compat = "fsl,b4860-device-config",
.init_periph = t2080_init_periph,
.cmux_groups = {
&clockgen2_cmux_cga12, &clockgen2_cmux_cgb
},
.hwaccel = {
&t2080_hwa1
},
.cmux_to_group = {
0, 1, 1, 1, -1
},
.pll_mask = 0x3f,
.flags = CG_PLL_8BIT,
},
{
.compat = "fsl,ls1021a-clockgen",
.cmux_groups = {
&t1023_cmux
},
.cmux_to_group = {
0, -1
},
.pll_mask = 0x03,
},
{
.compat = "fsl,ls1043a-clockgen",
.init_periph = t2080_init_periph,
.cmux_groups = {
&t1040_cmux
},
.hwaccel = {
&ls1043a_hwa1, &ls1043a_hwa2
},
.cmux_to_group = {
0, -1
},
.pll_mask = 0x07,
.flags = CG_PLL_8BIT,
},
{
.compat = "fsl,ls2080a-clockgen",
.cmux_groups = {
&clockgen2_cmux_cga12, &clockgen2_cmux_cgb
},
.cmux_to_group = {
0, 0, 1, 1, -1
},
.pll_mask = 0x37,
.flags = CG_VER3 | CG_LITTLE_ENDIAN,
},
clk: qoriq: Move chip-specific knowledge into driver The device tree should describe the chips (or chip-like subblocks) in the system, but it generally does not describe individual registers -- it should identify, rather than describe, a programming interface. This has not been the case with the QorIQ clockgen nodes. The knowledge of what each bit setting of CLKCnCSR means is encoded in three places (binding, pll node, and mux node), and the last also needs to know which options are valid on a particular chip. All three of these locations are considered stable ABI, making it difficult to fix mistakes (of which I have found several), much less refactor the abstraction to be able to address problems, limitations, or new chips. Under the current binding, a pll clock specifier of 2 means that the PLL is divided by 4 -- and the driver implements this, unless there happen to be four clock-output-names rather than 3, in which case it interprets it as PLL divided by 3. This does not appear in the binding documentation at all. That hack is now considered stable ABI. The current device tree nodes contain errors, such as saying that T1040 can set a core clock to PLL/4 when only PLL and PLL/2 are options. The current binding also ignores some restrictions on clock selection, such as p5020's requirement that if a core uses the "wrong" PLL, that PLL must be clocked lower than the "correct" PLL and be at most 80% of the rated CPU frequency. Possibly because of the lack of the ability to express such nuance in the binding, some valid options are omitted from the device trees, such as the ability on p4080 to run cores 0-3 from PLL3 and cores 4-7 from PLL1 (again, only if they are at most 80% of rated CPU frequency). This omission, combined with excessive caution in the cpufreq driver (addressed in a subsequent patch), means that currently on a 1500 MHz p4080 with typical PLL configuration, cpufreq can lower the frequency to 1200 MHz on half the CPUs and do nothing on the others. With this patchset, all CPUs can be lowered to 1200 MHz on a rev2 p4080, and on a rev3 p4080 half can be lowered to 750 MHz and the other half to 600 MHz. The current binding only deals with CPU clocks. To describe FMan in the device tree, we need to describe its clock. Some chips have additional muxes that work like the CPU muxes, but are not described in the device tree. Others require inspecting the Reset Control Word to determine which PLL is used. Rather than continue to extend this mess, replace it. Have the driver bind to the chip-specific clockgen compatible, and keep the detailed description of quirky chip variations in the driver, where it can be easily fixed, refactored, and extended. Older device trees will continue to work (including a workaround for old ls1021a device trees that are missing compatible and reg in the clockgen node, which even the old binding required). The pll/mux details in old device trees will be ignored, but "clocks" properties pointing at the old nodes will still work, and be directed at the corresponding new clock. Signed-off-by: Scott Wood <scottwood@freescale.com> Acked-by: Stephen Boyd <sboyd@codeaurora.org>
2015-09-20 12:29:54 +08:00
{
.compat = "fsl,p2041-clockgen",
.guts_compat = "fsl,qoriq-device-config-1.0",
.init_periph = p2041_init_periph,
.cmux_groups = {
&p2041_cmux_grp1, &p2041_cmux_grp2
},
.cmux_to_group = {
0, 0, 1, 1, -1
},
.pll_mask = 0x07,
},
{
.compat = "fsl,p3041-clockgen",
.guts_compat = "fsl,qoriq-device-config-1.0",
.init_periph = p2041_init_periph,
.cmux_groups = {
&p2041_cmux_grp1, &p2041_cmux_grp2
},
.cmux_to_group = {
0, 0, 1, 1, -1
},
.pll_mask = 0x07,
},
{
.compat = "fsl,p4080-clockgen",
.guts_compat = "fsl,qoriq-device-config-1.0",
.init_periph = p4080_init_periph,
.cmux_groups = {
&p4080_cmux_grp1, &p4080_cmux_grp2
},
.cmux_to_group = {
0, 0, 0, 0, 1, 1, 1, 1
},
.pll_mask = 0x1f,
},
{
.compat = "fsl,p5020-clockgen",
.guts_compat = "fsl,qoriq-device-config-1.0",
.init_periph = p5020_init_periph,
.cmux_groups = {
&p2041_cmux_grp1, &p2041_cmux_grp2
},
.cmux_to_group = {
0, 1, -1
},
.pll_mask = 0x07,
},
{
.compat = "fsl,p5040-clockgen",
.guts_compat = "fsl,p5040-device-config",
.init_periph = p5040_init_periph,
.cmux_groups = {
&p5040_cmux_grp1, &p5040_cmux_grp2
},
.cmux_to_group = {
0, 0, 1, 1, -1
},
.pll_mask = 0x0f,
},
{
.compat = "fsl,t1023-clockgen",
.guts_compat = "fsl,t1023-device-config",
.init_periph = t1023_init_periph,
.cmux_groups = {
&t1023_cmux
},
.hwaccel = {
&t1023_hwa1, &t1023_hwa2
},
.cmux_to_group = {
0, 0, -1
},
.pll_mask = 0x03,
.flags = CG_PLL_8BIT,
},
{
.compat = "fsl,t1040-clockgen",
.guts_compat = "fsl,t1040-device-config",
.init_periph = t1040_init_periph,
.cmux_groups = {
&t1040_cmux
},
.cmux_to_group = {
0, 0, 0, 0, -1
},
.pll_mask = 0x07,
.flags = CG_PLL_8BIT,
},
{
.compat = "fsl,t2080-clockgen",
.guts_compat = "fsl,t2080-device-config",
.init_periph = t2080_init_periph,
.cmux_groups = {
&clockgen2_cmux_cga12
},
.hwaccel = {
&t2080_hwa1, &t2080_hwa2
},
.cmux_to_group = {
0, -1
},
.pll_mask = 0x07,
.flags = CG_PLL_8BIT,
},
{
.compat = "fsl,t4240-clockgen",
.guts_compat = "fsl,t4240-device-config",
.init_periph = t4240_init_periph,
.cmux_groups = {
&clockgen2_cmux_cga, &clockgen2_cmux_cgb
},
.hwaccel = {
&t4240_hwa1, NULL, NULL, &t4240_hwa4, &t4240_hwa5
},
.cmux_to_group = {
0, 0, 1, -1
},
.pll_mask = 0x3f,
.flags = CG_PLL_8BIT,
},
{},
};
struct mux_hwclock {
struct clk_hw hw;
clk: qoriq: Move chip-specific knowledge into driver The device tree should describe the chips (or chip-like subblocks) in the system, but it generally does not describe individual registers -- it should identify, rather than describe, a programming interface. This has not been the case with the QorIQ clockgen nodes. The knowledge of what each bit setting of CLKCnCSR means is encoded in three places (binding, pll node, and mux node), and the last also needs to know which options are valid on a particular chip. All three of these locations are considered stable ABI, making it difficult to fix mistakes (of which I have found several), much less refactor the abstraction to be able to address problems, limitations, or new chips. Under the current binding, a pll clock specifier of 2 means that the PLL is divided by 4 -- and the driver implements this, unless there happen to be four clock-output-names rather than 3, in which case it interprets it as PLL divided by 3. This does not appear in the binding documentation at all. That hack is now considered stable ABI. The current device tree nodes contain errors, such as saying that T1040 can set a core clock to PLL/4 when only PLL and PLL/2 are options. The current binding also ignores some restrictions on clock selection, such as p5020's requirement that if a core uses the "wrong" PLL, that PLL must be clocked lower than the "correct" PLL and be at most 80% of the rated CPU frequency. Possibly because of the lack of the ability to express such nuance in the binding, some valid options are omitted from the device trees, such as the ability on p4080 to run cores 0-3 from PLL3 and cores 4-7 from PLL1 (again, only if they are at most 80% of rated CPU frequency). This omission, combined with excessive caution in the cpufreq driver (addressed in a subsequent patch), means that currently on a 1500 MHz p4080 with typical PLL configuration, cpufreq can lower the frequency to 1200 MHz on half the CPUs and do nothing on the others. With this patchset, all CPUs can be lowered to 1200 MHz on a rev2 p4080, and on a rev3 p4080 half can be lowered to 750 MHz and the other half to 600 MHz. The current binding only deals with CPU clocks. To describe FMan in the device tree, we need to describe its clock. Some chips have additional muxes that work like the CPU muxes, but are not described in the device tree. Others require inspecting the Reset Control Word to determine which PLL is used. Rather than continue to extend this mess, replace it. Have the driver bind to the chip-specific clockgen compatible, and keep the detailed description of quirky chip variations in the driver, where it can be easily fixed, refactored, and extended. Older device trees will continue to work (including a workaround for old ls1021a device trees that are missing compatible and reg in the clockgen node, which even the old binding required). The pll/mux details in old device trees will be ignored, but "clocks" properties pointing at the old nodes will still work, and be directed at the corresponding new clock. Signed-off-by: Scott Wood <scottwood@freescale.com> Acked-by: Stephen Boyd <sboyd@codeaurora.org>
2015-09-20 12:29:54 +08:00
struct clockgen *cg;
const struct clockgen_muxinfo *info;
u32 __iomem *reg;
u8 parent_to_clksel[NUM_MUX_PARENTS];
s8 clksel_to_parent[NUM_MUX_PARENTS];
int num_parents;
};
clk: qoriq: Move chip-specific knowledge into driver The device tree should describe the chips (or chip-like subblocks) in the system, but it generally does not describe individual registers -- it should identify, rather than describe, a programming interface. This has not been the case with the QorIQ clockgen nodes. The knowledge of what each bit setting of CLKCnCSR means is encoded in three places (binding, pll node, and mux node), and the last also needs to know which options are valid on a particular chip. All three of these locations are considered stable ABI, making it difficult to fix mistakes (of which I have found several), much less refactor the abstraction to be able to address problems, limitations, or new chips. Under the current binding, a pll clock specifier of 2 means that the PLL is divided by 4 -- and the driver implements this, unless there happen to be four clock-output-names rather than 3, in which case it interprets it as PLL divided by 3. This does not appear in the binding documentation at all. That hack is now considered stable ABI. The current device tree nodes contain errors, such as saying that T1040 can set a core clock to PLL/4 when only PLL and PLL/2 are options. The current binding also ignores some restrictions on clock selection, such as p5020's requirement that if a core uses the "wrong" PLL, that PLL must be clocked lower than the "correct" PLL and be at most 80% of the rated CPU frequency. Possibly because of the lack of the ability to express such nuance in the binding, some valid options are omitted from the device trees, such as the ability on p4080 to run cores 0-3 from PLL3 and cores 4-7 from PLL1 (again, only if they are at most 80% of rated CPU frequency). This omission, combined with excessive caution in the cpufreq driver (addressed in a subsequent patch), means that currently on a 1500 MHz p4080 with typical PLL configuration, cpufreq can lower the frequency to 1200 MHz on half the CPUs and do nothing on the others. With this patchset, all CPUs can be lowered to 1200 MHz on a rev2 p4080, and on a rev3 p4080 half can be lowered to 750 MHz and the other half to 600 MHz. The current binding only deals with CPU clocks. To describe FMan in the device tree, we need to describe its clock. Some chips have additional muxes that work like the CPU muxes, but are not described in the device tree. Others require inspecting the Reset Control Word to determine which PLL is used. Rather than continue to extend this mess, replace it. Have the driver bind to the chip-specific clockgen compatible, and keep the detailed description of quirky chip variations in the driver, where it can be easily fixed, refactored, and extended. Older device trees will continue to work (including a workaround for old ls1021a device trees that are missing compatible and reg in the clockgen node, which even the old binding required). The pll/mux details in old device trees will be ignored, but "clocks" properties pointing at the old nodes will still work, and be directed at the corresponding new clock. Signed-off-by: Scott Wood <scottwood@freescale.com> Acked-by: Stephen Boyd <sboyd@codeaurora.org>
2015-09-20 12:29:54 +08:00
#define to_mux_hwclock(p) container_of(p, struct mux_hwclock, hw)
#define CLKSEL_MASK 0x78000000
#define CLKSEL_SHIFT 27
clk: qoriq: Move chip-specific knowledge into driver The device tree should describe the chips (or chip-like subblocks) in the system, but it generally does not describe individual registers -- it should identify, rather than describe, a programming interface. This has not been the case with the QorIQ clockgen nodes. The knowledge of what each bit setting of CLKCnCSR means is encoded in three places (binding, pll node, and mux node), and the last also needs to know which options are valid on a particular chip. All three of these locations are considered stable ABI, making it difficult to fix mistakes (of which I have found several), much less refactor the abstraction to be able to address problems, limitations, or new chips. Under the current binding, a pll clock specifier of 2 means that the PLL is divided by 4 -- and the driver implements this, unless there happen to be four clock-output-names rather than 3, in which case it interprets it as PLL divided by 3. This does not appear in the binding documentation at all. That hack is now considered stable ABI. The current device tree nodes contain errors, such as saying that T1040 can set a core clock to PLL/4 when only PLL and PLL/2 are options. The current binding also ignores some restrictions on clock selection, such as p5020's requirement that if a core uses the "wrong" PLL, that PLL must be clocked lower than the "correct" PLL and be at most 80% of the rated CPU frequency. Possibly because of the lack of the ability to express such nuance in the binding, some valid options are omitted from the device trees, such as the ability on p4080 to run cores 0-3 from PLL3 and cores 4-7 from PLL1 (again, only if they are at most 80% of rated CPU frequency). This omission, combined with excessive caution in the cpufreq driver (addressed in a subsequent patch), means that currently on a 1500 MHz p4080 with typical PLL configuration, cpufreq can lower the frequency to 1200 MHz on half the CPUs and do nothing on the others. With this patchset, all CPUs can be lowered to 1200 MHz on a rev2 p4080, and on a rev3 p4080 half can be lowered to 750 MHz and the other half to 600 MHz. The current binding only deals with CPU clocks. To describe FMan in the device tree, we need to describe its clock. Some chips have additional muxes that work like the CPU muxes, but are not described in the device tree. Others require inspecting the Reset Control Word to determine which PLL is used. Rather than continue to extend this mess, replace it. Have the driver bind to the chip-specific clockgen compatible, and keep the detailed description of quirky chip variations in the driver, where it can be easily fixed, refactored, and extended. Older device trees will continue to work (including a workaround for old ls1021a device trees that are missing compatible and reg in the clockgen node, which even the old binding required). The pll/mux details in old device trees will be ignored, but "clocks" properties pointing at the old nodes will still work, and be directed at the corresponding new clock. Signed-off-by: Scott Wood <scottwood@freescale.com> Acked-by: Stephen Boyd <sboyd@codeaurora.org>
2015-09-20 12:29:54 +08:00
static int mux_set_parent(struct clk_hw *hw, u8 idx)
{
clk: qoriq: Move chip-specific knowledge into driver The device tree should describe the chips (or chip-like subblocks) in the system, but it generally does not describe individual registers -- it should identify, rather than describe, a programming interface. This has not been the case with the QorIQ clockgen nodes. The knowledge of what each bit setting of CLKCnCSR means is encoded in three places (binding, pll node, and mux node), and the last also needs to know which options are valid on a particular chip. All three of these locations are considered stable ABI, making it difficult to fix mistakes (of which I have found several), much less refactor the abstraction to be able to address problems, limitations, or new chips. Under the current binding, a pll clock specifier of 2 means that the PLL is divided by 4 -- and the driver implements this, unless there happen to be four clock-output-names rather than 3, in which case it interprets it as PLL divided by 3. This does not appear in the binding documentation at all. That hack is now considered stable ABI. The current device tree nodes contain errors, such as saying that T1040 can set a core clock to PLL/4 when only PLL and PLL/2 are options. The current binding also ignores some restrictions on clock selection, such as p5020's requirement that if a core uses the "wrong" PLL, that PLL must be clocked lower than the "correct" PLL and be at most 80% of the rated CPU frequency. Possibly because of the lack of the ability to express such nuance in the binding, some valid options are omitted from the device trees, such as the ability on p4080 to run cores 0-3 from PLL3 and cores 4-7 from PLL1 (again, only if they are at most 80% of rated CPU frequency). This omission, combined with excessive caution in the cpufreq driver (addressed in a subsequent patch), means that currently on a 1500 MHz p4080 with typical PLL configuration, cpufreq can lower the frequency to 1200 MHz on half the CPUs and do nothing on the others. With this patchset, all CPUs can be lowered to 1200 MHz on a rev2 p4080, and on a rev3 p4080 half can be lowered to 750 MHz and the other half to 600 MHz. The current binding only deals with CPU clocks. To describe FMan in the device tree, we need to describe its clock. Some chips have additional muxes that work like the CPU muxes, but are not described in the device tree. Others require inspecting the Reset Control Word to determine which PLL is used. Rather than continue to extend this mess, replace it. Have the driver bind to the chip-specific clockgen compatible, and keep the detailed description of quirky chip variations in the driver, where it can be easily fixed, refactored, and extended. Older device trees will continue to work (including a workaround for old ls1021a device trees that are missing compatible and reg in the clockgen node, which even the old binding required). The pll/mux details in old device trees will be ignored, but "clocks" properties pointing at the old nodes will still work, and be directed at the corresponding new clock. Signed-off-by: Scott Wood <scottwood@freescale.com> Acked-by: Stephen Boyd <sboyd@codeaurora.org>
2015-09-20 12:29:54 +08:00
struct mux_hwclock *hwc = to_mux_hwclock(hw);
u32 clksel;
clk: qoriq: Move chip-specific knowledge into driver The device tree should describe the chips (or chip-like subblocks) in the system, but it generally does not describe individual registers -- it should identify, rather than describe, a programming interface. This has not been the case with the QorIQ clockgen nodes. The knowledge of what each bit setting of CLKCnCSR means is encoded in three places (binding, pll node, and mux node), and the last also needs to know which options are valid on a particular chip. All three of these locations are considered stable ABI, making it difficult to fix mistakes (of which I have found several), much less refactor the abstraction to be able to address problems, limitations, or new chips. Under the current binding, a pll clock specifier of 2 means that the PLL is divided by 4 -- and the driver implements this, unless there happen to be four clock-output-names rather than 3, in which case it interprets it as PLL divided by 3. This does not appear in the binding documentation at all. That hack is now considered stable ABI. The current device tree nodes contain errors, such as saying that T1040 can set a core clock to PLL/4 when only PLL and PLL/2 are options. The current binding also ignores some restrictions on clock selection, such as p5020's requirement that if a core uses the "wrong" PLL, that PLL must be clocked lower than the "correct" PLL and be at most 80% of the rated CPU frequency. Possibly because of the lack of the ability to express such nuance in the binding, some valid options are omitted from the device trees, such as the ability on p4080 to run cores 0-3 from PLL3 and cores 4-7 from PLL1 (again, only if they are at most 80% of rated CPU frequency). This omission, combined with excessive caution in the cpufreq driver (addressed in a subsequent patch), means that currently on a 1500 MHz p4080 with typical PLL configuration, cpufreq can lower the frequency to 1200 MHz on half the CPUs and do nothing on the others. With this patchset, all CPUs can be lowered to 1200 MHz on a rev2 p4080, and on a rev3 p4080 half can be lowered to 750 MHz and the other half to 600 MHz. The current binding only deals with CPU clocks. To describe FMan in the device tree, we need to describe its clock. Some chips have additional muxes that work like the CPU muxes, but are not described in the device tree. Others require inspecting the Reset Control Word to determine which PLL is used. Rather than continue to extend this mess, replace it. Have the driver bind to the chip-specific clockgen compatible, and keep the detailed description of quirky chip variations in the driver, where it can be easily fixed, refactored, and extended. Older device trees will continue to work (including a workaround for old ls1021a device trees that are missing compatible and reg in the clockgen node, which even the old binding required). The pll/mux details in old device trees will be ignored, but "clocks" properties pointing at the old nodes will still work, and be directed at the corresponding new clock. Signed-off-by: Scott Wood <scottwood@freescale.com> Acked-by: Stephen Boyd <sboyd@codeaurora.org>
2015-09-20 12:29:54 +08:00
if (idx >= hwc->num_parents)
return -EINVAL;
clksel = hwc->parent_to_clksel[idx];
cg_out(hwc->cg, (clksel << CLKSEL_SHIFT) & CLKSEL_MASK, hwc->reg);
return 0;
}
clk: qoriq: Move chip-specific knowledge into driver The device tree should describe the chips (or chip-like subblocks) in the system, but it generally does not describe individual registers -- it should identify, rather than describe, a programming interface. This has not been the case with the QorIQ clockgen nodes. The knowledge of what each bit setting of CLKCnCSR means is encoded in three places (binding, pll node, and mux node), and the last also needs to know which options are valid on a particular chip. All three of these locations are considered stable ABI, making it difficult to fix mistakes (of which I have found several), much less refactor the abstraction to be able to address problems, limitations, or new chips. Under the current binding, a pll clock specifier of 2 means that the PLL is divided by 4 -- and the driver implements this, unless there happen to be four clock-output-names rather than 3, in which case it interprets it as PLL divided by 3. This does not appear in the binding documentation at all. That hack is now considered stable ABI. The current device tree nodes contain errors, such as saying that T1040 can set a core clock to PLL/4 when only PLL and PLL/2 are options. The current binding also ignores some restrictions on clock selection, such as p5020's requirement that if a core uses the "wrong" PLL, that PLL must be clocked lower than the "correct" PLL and be at most 80% of the rated CPU frequency. Possibly because of the lack of the ability to express such nuance in the binding, some valid options are omitted from the device trees, such as the ability on p4080 to run cores 0-3 from PLL3 and cores 4-7 from PLL1 (again, only if they are at most 80% of rated CPU frequency). This omission, combined with excessive caution in the cpufreq driver (addressed in a subsequent patch), means that currently on a 1500 MHz p4080 with typical PLL configuration, cpufreq can lower the frequency to 1200 MHz on half the CPUs and do nothing on the others. With this patchset, all CPUs can be lowered to 1200 MHz on a rev2 p4080, and on a rev3 p4080 half can be lowered to 750 MHz and the other half to 600 MHz. The current binding only deals with CPU clocks. To describe FMan in the device tree, we need to describe its clock. Some chips have additional muxes that work like the CPU muxes, but are not described in the device tree. Others require inspecting the Reset Control Word to determine which PLL is used. Rather than continue to extend this mess, replace it. Have the driver bind to the chip-specific clockgen compatible, and keep the detailed description of quirky chip variations in the driver, where it can be easily fixed, refactored, and extended. Older device trees will continue to work (including a workaround for old ls1021a device trees that are missing compatible and reg in the clockgen node, which even the old binding required). The pll/mux details in old device trees will be ignored, but "clocks" properties pointing at the old nodes will still work, and be directed at the corresponding new clock. Signed-off-by: Scott Wood <scottwood@freescale.com> Acked-by: Stephen Boyd <sboyd@codeaurora.org>
2015-09-20 12:29:54 +08:00
static u8 mux_get_parent(struct clk_hw *hw)
{
clk: qoriq: Move chip-specific knowledge into driver The device tree should describe the chips (or chip-like subblocks) in the system, but it generally does not describe individual registers -- it should identify, rather than describe, a programming interface. This has not been the case with the QorIQ clockgen nodes. The knowledge of what each bit setting of CLKCnCSR means is encoded in three places (binding, pll node, and mux node), and the last also needs to know which options are valid on a particular chip. All three of these locations are considered stable ABI, making it difficult to fix mistakes (of which I have found several), much less refactor the abstraction to be able to address problems, limitations, or new chips. Under the current binding, a pll clock specifier of 2 means that the PLL is divided by 4 -- and the driver implements this, unless there happen to be four clock-output-names rather than 3, in which case it interprets it as PLL divided by 3. This does not appear in the binding documentation at all. That hack is now considered stable ABI. The current device tree nodes contain errors, such as saying that T1040 can set a core clock to PLL/4 when only PLL and PLL/2 are options. The current binding also ignores some restrictions on clock selection, such as p5020's requirement that if a core uses the "wrong" PLL, that PLL must be clocked lower than the "correct" PLL and be at most 80% of the rated CPU frequency. Possibly because of the lack of the ability to express such nuance in the binding, some valid options are omitted from the device trees, such as the ability on p4080 to run cores 0-3 from PLL3 and cores 4-7 from PLL1 (again, only if they are at most 80% of rated CPU frequency). This omission, combined with excessive caution in the cpufreq driver (addressed in a subsequent patch), means that currently on a 1500 MHz p4080 with typical PLL configuration, cpufreq can lower the frequency to 1200 MHz on half the CPUs and do nothing on the others. With this patchset, all CPUs can be lowered to 1200 MHz on a rev2 p4080, and on a rev3 p4080 half can be lowered to 750 MHz and the other half to 600 MHz. The current binding only deals with CPU clocks. To describe FMan in the device tree, we need to describe its clock. Some chips have additional muxes that work like the CPU muxes, but are not described in the device tree. Others require inspecting the Reset Control Word to determine which PLL is used. Rather than continue to extend this mess, replace it. Have the driver bind to the chip-specific clockgen compatible, and keep the detailed description of quirky chip variations in the driver, where it can be easily fixed, refactored, and extended. Older device trees will continue to work (including a workaround for old ls1021a device trees that are missing compatible and reg in the clockgen node, which even the old binding required). The pll/mux details in old device trees will be ignored, but "clocks" properties pointing at the old nodes will still work, and be directed at the corresponding new clock. Signed-off-by: Scott Wood <scottwood@freescale.com> Acked-by: Stephen Boyd <sboyd@codeaurora.org>
2015-09-20 12:29:54 +08:00
struct mux_hwclock *hwc = to_mux_hwclock(hw);
u32 clksel;
clk: qoriq: Move chip-specific knowledge into driver The device tree should describe the chips (or chip-like subblocks) in the system, but it generally does not describe individual registers -- it should identify, rather than describe, a programming interface. This has not been the case with the QorIQ clockgen nodes. The knowledge of what each bit setting of CLKCnCSR means is encoded in three places (binding, pll node, and mux node), and the last also needs to know which options are valid on a particular chip. All three of these locations are considered stable ABI, making it difficult to fix mistakes (of which I have found several), much less refactor the abstraction to be able to address problems, limitations, or new chips. Under the current binding, a pll clock specifier of 2 means that the PLL is divided by 4 -- and the driver implements this, unless there happen to be four clock-output-names rather than 3, in which case it interprets it as PLL divided by 3. This does not appear in the binding documentation at all. That hack is now considered stable ABI. The current device tree nodes contain errors, such as saying that T1040 can set a core clock to PLL/4 when only PLL and PLL/2 are options. The current binding also ignores some restrictions on clock selection, such as p5020's requirement that if a core uses the "wrong" PLL, that PLL must be clocked lower than the "correct" PLL and be at most 80% of the rated CPU frequency. Possibly because of the lack of the ability to express such nuance in the binding, some valid options are omitted from the device trees, such as the ability on p4080 to run cores 0-3 from PLL3 and cores 4-7 from PLL1 (again, only if they are at most 80% of rated CPU frequency). This omission, combined with excessive caution in the cpufreq driver (addressed in a subsequent patch), means that currently on a 1500 MHz p4080 with typical PLL configuration, cpufreq can lower the frequency to 1200 MHz on half the CPUs and do nothing on the others. With this patchset, all CPUs can be lowered to 1200 MHz on a rev2 p4080, and on a rev3 p4080 half can be lowered to 750 MHz and the other half to 600 MHz. The current binding only deals with CPU clocks. To describe FMan in the device tree, we need to describe its clock. Some chips have additional muxes that work like the CPU muxes, but are not described in the device tree. Others require inspecting the Reset Control Word to determine which PLL is used. Rather than continue to extend this mess, replace it. Have the driver bind to the chip-specific clockgen compatible, and keep the detailed description of quirky chip variations in the driver, where it can be easily fixed, refactored, and extended. Older device trees will continue to work (including a workaround for old ls1021a device trees that are missing compatible and reg in the clockgen node, which even the old binding required). The pll/mux details in old device trees will be ignored, but "clocks" properties pointing at the old nodes will still work, and be directed at the corresponding new clock. Signed-off-by: Scott Wood <scottwood@freescale.com> Acked-by: Stephen Boyd <sboyd@codeaurora.org>
2015-09-20 12:29:54 +08:00
s8 ret;
clksel = (cg_in(hwc->cg, hwc->reg) & CLKSEL_MASK) >> CLKSEL_SHIFT;
clk: qoriq: Move chip-specific knowledge into driver The device tree should describe the chips (or chip-like subblocks) in the system, but it generally does not describe individual registers -- it should identify, rather than describe, a programming interface. This has not been the case with the QorIQ clockgen nodes. The knowledge of what each bit setting of CLKCnCSR means is encoded in three places (binding, pll node, and mux node), and the last also needs to know which options are valid on a particular chip. All three of these locations are considered stable ABI, making it difficult to fix mistakes (of which I have found several), much less refactor the abstraction to be able to address problems, limitations, or new chips. Under the current binding, a pll clock specifier of 2 means that the PLL is divided by 4 -- and the driver implements this, unless there happen to be four clock-output-names rather than 3, in which case it interprets it as PLL divided by 3. This does not appear in the binding documentation at all. That hack is now considered stable ABI. The current device tree nodes contain errors, such as saying that T1040 can set a core clock to PLL/4 when only PLL and PLL/2 are options. The current binding also ignores some restrictions on clock selection, such as p5020's requirement that if a core uses the "wrong" PLL, that PLL must be clocked lower than the "correct" PLL and be at most 80% of the rated CPU frequency. Possibly because of the lack of the ability to express such nuance in the binding, some valid options are omitted from the device trees, such as the ability on p4080 to run cores 0-3 from PLL3 and cores 4-7 from PLL1 (again, only if they are at most 80% of rated CPU frequency). This omission, combined with excessive caution in the cpufreq driver (addressed in a subsequent patch), means that currently on a 1500 MHz p4080 with typical PLL configuration, cpufreq can lower the frequency to 1200 MHz on half the CPUs and do nothing on the others. With this patchset, all CPUs can be lowered to 1200 MHz on a rev2 p4080, and on a rev3 p4080 half can be lowered to 750 MHz and the other half to 600 MHz. The current binding only deals with CPU clocks. To describe FMan in the device tree, we need to describe its clock. Some chips have additional muxes that work like the CPU muxes, but are not described in the device tree. Others require inspecting the Reset Control Word to determine which PLL is used. Rather than continue to extend this mess, replace it. Have the driver bind to the chip-specific clockgen compatible, and keep the detailed description of quirky chip variations in the driver, where it can be easily fixed, refactored, and extended. Older device trees will continue to work (including a workaround for old ls1021a device trees that are missing compatible and reg in the clockgen node, which even the old binding required). The pll/mux details in old device trees will be ignored, but "clocks" properties pointing at the old nodes will still work, and be directed at the corresponding new clock. Signed-off-by: Scott Wood <scottwood@freescale.com> Acked-by: Stephen Boyd <sboyd@codeaurora.org>
2015-09-20 12:29:54 +08:00
ret = hwc->clksel_to_parent[clksel];
if (ret < 0) {
pr_err("%s: mux at %p has bad clksel\n", __func__, hwc->reg);
return 0;
}
return ret;
}
static const struct clk_ops cmux_ops = {
clk: qoriq: Move chip-specific knowledge into driver The device tree should describe the chips (or chip-like subblocks) in the system, but it generally does not describe individual registers -- it should identify, rather than describe, a programming interface. This has not been the case with the QorIQ clockgen nodes. The knowledge of what each bit setting of CLKCnCSR means is encoded in three places (binding, pll node, and mux node), and the last also needs to know which options are valid on a particular chip. All three of these locations are considered stable ABI, making it difficult to fix mistakes (of which I have found several), much less refactor the abstraction to be able to address problems, limitations, or new chips. Under the current binding, a pll clock specifier of 2 means that the PLL is divided by 4 -- and the driver implements this, unless there happen to be four clock-output-names rather than 3, in which case it interprets it as PLL divided by 3. This does not appear in the binding documentation at all. That hack is now considered stable ABI. The current device tree nodes contain errors, such as saying that T1040 can set a core clock to PLL/4 when only PLL and PLL/2 are options. The current binding also ignores some restrictions on clock selection, such as p5020's requirement that if a core uses the "wrong" PLL, that PLL must be clocked lower than the "correct" PLL and be at most 80% of the rated CPU frequency. Possibly because of the lack of the ability to express such nuance in the binding, some valid options are omitted from the device trees, such as the ability on p4080 to run cores 0-3 from PLL3 and cores 4-7 from PLL1 (again, only if they are at most 80% of rated CPU frequency). This omission, combined with excessive caution in the cpufreq driver (addressed in a subsequent patch), means that currently on a 1500 MHz p4080 with typical PLL configuration, cpufreq can lower the frequency to 1200 MHz on half the CPUs and do nothing on the others. With this patchset, all CPUs can be lowered to 1200 MHz on a rev2 p4080, and on a rev3 p4080 half can be lowered to 750 MHz and the other half to 600 MHz. The current binding only deals with CPU clocks. To describe FMan in the device tree, we need to describe its clock. Some chips have additional muxes that work like the CPU muxes, but are not described in the device tree. Others require inspecting the Reset Control Word to determine which PLL is used. Rather than continue to extend this mess, replace it. Have the driver bind to the chip-specific clockgen compatible, and keep the detailed description of quirky chip variations in the driver, where it can be easily fixed, refactored, and extended. Older device trees will continue to work (including a workaround for old ls1021a device trees that are missing compatible and reg in the clockgen node, which even the old binding required). The pll/mux details in old device trees will be ignored, but "clocks" properties pointing at the old nodes will still work, and be directed at the corresponding new clock. Signed-off-by: Scott Wood <scottwood@freescale.com> Acked-by: Stephen Boyd <sboyd@codeaurora.org>
2015-09-20 12:29:54 +08:00
.get_parent = mux_get_parent,
.set_parent = mux_set_parent,
};
clk: qoriq: Move chip-specific knowledge into driver The device tree should describe the chips (or chip-like subblocks) in the system, but it generally does not describe individual registers -- it should identify, rather than describe, a programming interface. This has not been the case with the QorIQ clockgen nodes. The knowledge of what each bit setting of CLKCnCSR means is encoded in three places (binding, pll node, and mux node), and the last also needs to know which options are valid on a particular chip. All three of these locations are considered stable ABI, making it difficult to fix mistakes (of which I have found several), much less refactor the abstraction to be able to address problems, limitations, or new chips. Under the current binding, a pll clock specifier of 2 means that the PLL is divided by 4 -- and the driver implements this, unless there happen to be four clock-output-names rather than 3, in which case it interprets it as PLL divided by 3. This does not appear in the binding documentation at all. That hack is now considered stable ABI. The current device tree nodes contain errors, such as saying that T1040 can set a core clock to PLL/4 when only PLL and PLL/2 are options. The current binding also ignores some restrictions on clock selection, such as p5020's requirement that if a core uses the "wrong" PLL, that PLL must be clocked lower than the "correct" PLL and be at most 80% of the rated CPU frequency. Possibly because of the lack of the ability to express such nuance in the binding, some valid options are omitted from the device trees, such as the ability on p4080 to run cores 0-3 from PLL3 and cores 4-7 from PLL1 (again, only if they are at most 80% of rated CPU frequency). This omission, combined with excessive caution in the cpufreq driver (addressed in a subsequent patch), means that currently on a 1500 MHz p4080 with typical PLL configuration, cpufreq can lower the frequency to 1200 MHz on half the CPUs and do nothing on the others. With this patchset, all CPUs can be lowered to 1200 MHz on a rev2 p4080, and on a rev3 p4080 half can be lowered to 750 MHz and the other half to 600 MHz. The current binding only deals with CPU clocks. To describe FMan in the device tree, we need to describe its clock. Some chips have additional muxes that work like the CPU muxes, but are not described in the device tree. Others require inspecting the Reset Control Word to determine which PLL is used. Rather than continue to extend this mess, replace it. Have the driver bind to the chip-specific clockgen compatible, and keep the detailed description of quirky chip variations in the driver, where it can be easily fixed, refactored, and extended. Older device trees will continue to work (including a workaround for old ls1021a device trees that are missing compatible and reg in the clockgen node, which even the old binding required). The pll/mux details in old device trees will be ignored, but "clocks" properties pointing at the old nodes will still work, and be directed at the corresponding new clock. Signed-off-by: Scott Wood <scottwood@freescale.com> Acked-by: Stephen Boyd <sboyd@codeaurora.org>
2015-09-20 12:29:54 +08:00
/*
* Don't allow setting for now, as the clock options haven't been
* sanitized for additional restrictions.
*/
static const struct clk_ops hwaccel_ops = {
.get_parent = mux_get_parent,
};
static const struct clockgen_pll_div *get_pll_div(struct clockgen *cg,
struct mux_hwclock *hwc,
int idx)
{
clk: qoriq: Move chip-specific knowledge into driver The device tree should describe the chips (or chip-like subblocks) in the system, but it generally does not describe individual registers -- it should identify, rather than describe, a programming interface. This has not been the case with the QorIQ clockgen nodes. The knowledge of what each bit setting of CLKCnCSR means is encoded in three places (binding, pll node, and mux node), and the last also needs to know which options are valid on a particular chip. All three of these locations are considered stable ABI, making it difficult to fix mistakes (of which I have found several), much less refactor the abstraction to be able to address problems, limitations, or new chips. Under the current binding, a pll clock specifier of 2 means that the PLL is divided by 4 -- and the driver implements this, unless there happen to be four clock-output-names rather than 3, in which case it interprets it as PLL divided by 3. This does not appear in the binding documentation at all. That hack is now considered stable ABI. The current device tree nodes contain errors, such as saying that T1040 can set a core clock to PLL/4 when only PLL and PLL/2 are options. The current binding also ignores some restrictions on clock selection, such as p5020's requirement that if a core uses the "wrong" PLL, that PLL must be clocked lower than the "correct" PLL and be at most 80% of the rated CPU frequency. Possibly because of the lack of the ability to express such nuance in the binding, some valid options are omitted from the device trees, such as the ability on p4080 to run cores 0-3 from PLL3 and cores 4-7 from PLL1 (again, only if they are at most 80% of rated CPU frequency). This omission, combined with excessive caution in the cpufreq driver (addressed in a subsequent patch), means that currently on a 1500 MHz p4080 with typical PLL configuration, cpufreq can lower the frequency to 1200 MHz on half the CPUs and do nothing on the others. With this patchset, all CPUs can be lowered to 1200 MHz on a rev2 p4080, and on a rev3 p4080 half can be lowered to 750 MHz and the other half to 600 MHz. The current binding only deals with CPU clocks. To describe FMan in the device tree, we need to describe its clock. Some chips have additional muxes that work like the CPU muxes, but are not described in the device tree. Others require inspecting the Reset Control Word to determine which PLL is used. Rather than continue to extend this mess, replace it. Have the driver bind to the chip-specific clockgen compatible, and keep the detailed description of quirky chip variations in the driver, where it can be easily fixed, refactored, and extended. Older device trees will continue to work (including a workaround for old ls1021a device trees that are missing compatible and reg in the clockgen node, which even the old binding required). The pll/mux details in old device trees will be ignored, but "clocks" properties pointing at the old nodes will still work, and be directed at the corresponding new clock. Signed-off-by: Scott Wood <scottwood@freescale.com> Acked-by: Stephen Boyd <sboyd@codeaurora.org>
2015-09-20 12:29:54 +08:00
int pll, div;
clk: qoriq: Move chip-specific knowledge into driver The device tree should describe the chips (or chip-like subblocks) in the system, but it generally does not describe individual registers -- it should identify, rather than describe, a programming interface. This has not been the case with the QorIQ clockgen nodes. The knowledge of what each bit setting of CLKCnCSR means is encoded in three places (binding, pll node, and mux node), and the last also needs to know which options are valid on a particular chip. All three of these locations are considered stable ABI, making it difficult to fix mistakes (of which I have found several), much less refactor the abstraction to be able to address problems, limitations, or new chips. Under the current binding, a pll clock specifier of 2 means that the PLL is divided by 4 -- and the driver implements this, unless there happen to be four clock-output-names rather than 3, in which case it interprets it as PLL divided by 3. This does not appear in the binding documentation at all. That hack is now considered stable ABI. The current device tree nodes contain errors, such as saying that T1040 can set a core clock to PLL/4 when only PLL and PLL/2 are options. The current binding also ignores some restrictions on clock selection, such as p5020's requirement that if a core uses the "wrong" PLL, that PLL must be clocked lower than the "correct" PLL and be at most 80% of the rated CPU frequency. Possibly because of the lack of the ability to express such nuance in the binding, some valid options are omitted from the device trees, such as the ability on p4080 to run cores 0-3 from PLL3 and cores 4-7 from PLL1 (again, only if they are at most 80% of rated CPU frequency). This omission, combined with excessive caution in the cpufreq driver (addressed in a subsequent patch), means that currently on a 1500 MHz p4080 with typical PLL configuration, cpufreq can lower the frequency to 1200 MHz on half the CPUs and do nothing on the others. With this patchset, all CPUs can be lowered to 1200 MHz on a rev2 p4080, and on a rev3 p4080 half can be lowered to 750 MHz and the other half to 600 MHz. The current binding only deals with CPU clocks. To describe FMan in the device tree, we need to describe its clock. Some chips have additional muxes that work like the CPU muxes, but are not described in the device tree. Others require inspecting the Reset Control Word to determine which PLL is used. Rather than continue to extend this mess, replace it. Have the driver bind to the chip-specific clockgen compatible, and keep the detailed description of quirky chip variations in the driver, where it can be easily fixed, refactored, and extended. Older device trees will continue to work (including a workaround for old ls1021a device trees that are missing compatible and reg in the clockgen node, which even the old binding required). The pll/mux details in old device trees will be ignored, but "clocks" properties pointing at the old nodes will still work, and be directed at the corresponding new clock. Signed-off-by: Scott Wood <scottwood@freescale.com> Acked-by: Stephen Boyd <sboyd@codeaurora.org>
2015-09-20 12:29:54 +08:00
if (!(hwc->info->clksel[idx].flags & CLKSEL_VALID))
return NULL;
clk: qoriq: Move chip-specific knowledge into driver The device tree should describe the chips (or chip-like subblocks) in the system, but it generally does not describe individual registers -- it should identify, rather than describe, a programming interface. This has not been the case with the QorIQ clockgen nodes. The knowledge of what each bit setting of CLKCnCSR means is encoded in three places (binding, pll node, and mux node), and the last also needs to know which options are valid on a particular chip. All three of these locations are considered stable ABI, making it difficult to fix mistakes (of which I have found several), much less refactor the abstraction to be able to address problems, limitations, or new chips. Under the current binding, a pll clock specifier of 2 means that the PLL is divided by 4 -- and the driver implements this, unless there happen to be four clock-output-names rather than 3, in which case it interprets it as PLL divided by 3. This does not appear in the binding documentation at all. That hack is now considered stable ABI. The current device tree nodes contain errors, such as saying that T1040 can set a core clock to PLL/4 when only PLL and PLL/2 are options. The current binding also ignores some restrictions on clock selection, such as p5020's requirement that if a core uses the "wrong" PLL, that PLL must be clocked lower than the "correct" PLL and be at most 80% of the rated CPU frequency. Possibly because of the lack of the ability to express such nuance in the binding, some valid options are omitted from the device trees, such as the ability on p4080 to run cores 0-3 from PLL3 and cores 4-7 from PLL1 (again, only if they are at most 80% of rated CPU frequency). This omission, combined with excessive caution in the cpufreq driver (addressed in a subsequent patch), means that currently on a 1500 MHz p4080 with typical PLL configuration, cpufreq can lower the frequency to 1200 MHz on half the CPUs and do nothing on the others. With this patchset, all CPUs can be lowered to 1200 MHz on a rev2 p4080, and on a rev3 p4080 half can be lowered to 750 MHz and the other half to 600 MHz. The current binding only deals with CPU clocks. To describe FMan in the device tree, we need to describe its clock. Some chips have additional muxes that work like the CPU muxes, but are not described in the device tree. Others require inspecting the Reset Control Word to determine which PLL is used. Rather than continue to extend this mess, replace it. Have the driver bind to the chip-specific clockgen compatible, and keep the detailed description of quirky chip variations in the driver, where it can be easily fixed, refactored, and extended. Older device trees will continue to work (including a workaround for old ls1021a device trees that are missing compatible and reg in the clockgen node, which even the old binding required). The pll/mux details in old device trees will be ignored, but "clocks" properties pointing at the old nodes will still work, and be directed at the corresponding new clock. Signed-off-by: Scott Wood <scottwood@freescale.com> Acked-by: Stephen Boyd <sboyd@codeaurora.org>
2015-09-20 12:29:54 +08:00
pll = hwc->info->clksel[idx].pll;
div = hwc->info->clksel[idx].div;
clk: qoriq: Move chip-specific knowledge into driver The device tree should describe the chips (or chip-like subblocks) in the system, but it generally does not describe individual registers -- it should identify, rather than describe, a programming interface. This has not been the case with the QorIQ clockgen nodes. The knowledge of what each bit setting of CLKCnCSR means is encoded in three places (binding, pll node, and mux node), and the last also needs to know which options are valid on a particular chip. All three of these locations are considered stable ABI, making it difficult to fix mistakes (of which I have found several), much less refactor the abstraction to be able to address problems, limitations, or new chips. Under the current binding, a pll clock specifier of 2 means that the PLL is divided by 4 -- and the driver implements this, unless there happen to be four clock-output-names rather than 3, in which case it interprets it as PLL divided by 3. This does not appear in the binding documentation at all. That hack is now considered stable ABI. The current device tree nodes contain errors, such as saying that T1040 can set a core clock to PLL/4 when only PLL and PLL/2 are options. The current binding also ignores some restrictions on clock selection, such as p5020's requirement that if a core uses the "wrong" PLL, that PLL must be clocked lower than the "correct" PLL and be at most 80% of the rated CPU frequency. Possibly because of the lack of the ability to express such nuance in the binding, some valid options are omitted from the device trees, such as the ability on p4080 to run cores 0-3 from PLL3 and cores 4-7 from PLL1 (again, only if they are at most 80% of rated CPU frequency). This omission, combined with excessive caution in the cpufreq driver (addressed in a subsequent patch), means that currently on a 1500 MHz p4080 with typical PLL configuration, cpufreq can lower the frequency to 1200 MHz on half the CPUs and do nothing on the others. With this patchset, all CPUs can be lowered to 1200 MHz on a rev2 p4080, and on a rev3 p4080 half can be lowered to 750 MHz and the other half to 600 MHz. The current binding only deals with CPU clocks. To describe FMan in the device tree, we need to describe its clock. Some chips have additional muxes that work like the CPU muxes, but are not described in the device tree. Others require inspecting the Reset Control Word to determine which PLL is used. Rather than continue to extend this mess, replace it. Have the driver bind to the chip-specific clockgen compatible, and keep the detailed description of quirky chip variations in the driver, where it can be easily fixed, refactored, and extended. Older device trees will continue to work (including a workaround for old ls1021a device trees that are missing compatible and reg in the clockgen node, which even the old binding required). The pll/mux details in old device trees will be ignored, but "clocks" properties pointing at the old nodes will still work, and be directed at the corresponding new clock. Signed-off-by: Scott Wood <scottwood@freescale.com> Acked-by: Stephen Boyd <sboyd@codeaurora.org>
2015-09-20 12:29:54 +08:00
return &cg->pll[pll].div[div];
}
clk: qoriq: Move chip-specific knowledge into driver The device tree should describe the chips (or chip-like subblocks) in the system, but it generally does not describe individual registers -- it should identify, rather than describe, a programming interface. This has not been the case with the QorIQ clockgen nodes. The knowledge of what each bit setting of CLKCnCSR means is encoded in three places (binding, pll node, and mux node), and the last also needs to know which options are valid on a particular chip. All three of these locations are considered stable ABI, making it difficult to fix mistakes (of which I have found several), much less refactor the abstraction to be able to address problems, limitations, or new chips. Under the current binding, a pll clock specifier of 2 means that the PLL is divided by 4 -- and the driver implements this, unless there happen to be four clock-output-names rather than 3, in which case it interprets it as PLL divided by 3. This does not appear in the binding documentation at all. That hack is now considered stable ABI. The current device tree nodes contain errors, such as saying that T1040 can set a core clock to PLL/4 when only PLL and PLL/2 are options. The current binding also ignores some restrictions on clock selection, such as p5020's requirement that if a core uses the "wrong" PLL, that PLL must be clocked lower than the "correct" PLL and be at most 80% of the rated CPU frequency. Possibly because of the lack of the ability to express such nuance in the binding, some valid options are omitted from the device trees, such as the ability on p4080 to run cores 0-3 from PLL3 and cores 4-7 from PLL1 (again, only if they are at most 80% of rated CPU frequency). This omission, combined with excessive caution in the cpufreq driver (addressed in a subsequent patch), means that currently on a 1500 MHz p4080 with typical PLL configuration, cpufreq can lower the frequency to 1200 MHz on half the CPUs and do nothing on the others. With this patchset, all CPUs can be lowered to 1200 MHz on a rev2 p4080, and on a rev3 p4080 half can be lowered to 750 MHz and the other half to 600 MHz. The current binding only deals with CPU clocks. To describe FMan in the device tree, we need to describe its clock. Some chips have additional muxes that work like the CPU muxes, but are not described in the device tree. Others require inspecting the Reset Control Word to determine which PLL is used. Rather than continue to extend this mess, replace it. Have the driver bind to the chip-specific clockgen compatible, and keep the detailed description of quirky chip variations in the driver, where it can be easily fixed, refactored, and extended. Older device trees will continue to work (including a workaround for old ls1021a device trees that are missing compatible and reg in the clockgen node, which even the old binding required). The pll/mux details in old device trees will be ignored, but "clocks" properties pointing at the old nodes will still work, and be directed at the corresponding new clock. Signed-off-by: Scott Wood <scottwood@freescale.com> Acked-by: Stephen Boyd <sboyd@codeaurora.org>
2015-09-20 12:29:54 +08:00
static struct clk * __init create_mux_common(struct clockgen *cg,
struct mux_hwclock *hwc,
const struct clk_ops *ops,
unsigned long min_rate,
unsigned long pct80_rate,
const char *fmt, int idx)
{
struct clk_init_data init = {};
struct clk *clk;
const struct clockgen_pll_div *div;
const char *parent_names[NUM_MUX_PARENTS];
char name[32];
int i, j;
clk: qoriq: Move chip-specific knowledge into driver The device tree should describe the chips (or chip-like subblocks) in the system, but it generally does not describe individual registers -- it should identify, rather than describe, a programming interface. This has not been the case with the QorIQ clockgen nodes. The knowledge of what each bit setting of CLKCnCSR means is encoded in three places (binding, pll node, and mux node), and the last also needs to know which options are valid on a particular chip. All three of these locations are considered stable ABI, making it difficult to fix mistakes (of which I have found several), much less refactor the abstraction to be able to address problems, limitations, or new chips. Under the current binding, a pll clock specifier of 2 means that the PLL is divided by 4 -- and the driver implements this, unless there happen to be four clock-output-names rather than 3, in which case it interprets it as PLL divided by 3. This does not appear in the binding documentation at all. That hack is now considered stable ABI. The current device tree nodes contain errors, such as saying that T1040 can set a core clock to PLL/4 when only PLL and PLL/2 are options. The current binding also ignores some restrictions on clock selection, such as p5020's requirement that if a core uses the "wrong" PLL, that PLL must be clocked lower than the "correct" PLL and be at most 80% of the rated CPU frequency. Possibly because of the lack of the ability to express such nuance in the binding, some valid options are omitted from the device trees, such as the ability on p4080 to run cores 0-3 from PLL3 and cores 4-7 from PLL1 (again, only if they are at most 80% of rated CPU frequency). This omission, combined with excessive caution in the cpufreq driver (addressed in a subsequent patch), means that currently on a 1500 MHz p4080 with typical PLL configuration, cpufreq can lower the frequency to 1200 MHz on half the CPUs and do nothing on the others. With this patchset, all CPUs can be lowered to 1200 MHz on a rev2 p4080, and on a rev3 p4080 half can be lowered to 750 MHz and the other half to 600 MHz. The current binding only deals with CPU clocks. To describe FMan in the device tree, we need to describe its clock. Some chips have additional muxes that work like the CPU muxes, but are not described in the device tree. Others require inspecting the Reset Control Word to determine which PLL is used. Rather than continue to extend this mess, replace it. Have the driver bind to the chip-specific clockgen compatible, and keep the detailed description of quirky chip variations in the driver, where it can be easily fixed, refactored, and extended. Older device trees will continue to work (including a workaround for old ls1021a device trees that are missing compatible and reg in the clockgen node, which even the old binding required). The pll/mux details in old device trees will be ignored, but "clocks" properties pointing at the old nodes will still work, and be directed at the corresponding new clock. Signed-off-by: Scott Wood <scottwood@freescale.com> Acked-by: Stephen Boyd <sboyd@codeaurora.org>
2015-09-20 12:29:54 +08:00
snprintf(name, sizeof(name), fmt, idx);
clk: qoriq: Move chip-specific knowledge into driver The device tree should describe the chips (or chip-like subblocks) in the system, but it generally does not describe individual registers -- it should identify, rather than describe, a programming interface. This has not been the case with the QorIQ clockgen nodes. The knowledge of what each bit setting of CLKCnCSR means is encoded in three places (binding, pll node, and mux node), and the last also needs to know which options are valid on a particular chip. All three of these locations are considered stable ABI, making it difficult to fix mistakes (of which I have found several), much less refactor the abstraction to be able to address problems, limitations, or new chips. Under the current binding, a pll clock specifier of 2 means that the PLL is divided by 4 -- and the driver implements this, unless there happen to be four clock-output-names rather than 3, in which case it interprets it as PLL divided by 3. This does not appear in the binding documentation at all. That hack is now considered stable ABI. The current device tree nodes contain errors, such as saying that T1040 can set a core clock to PLL/4 when only PLL and PLL/2 are options. The current binding also ignores some restrictions on clock selection, such as p5020's requirement that if a core uses the "wrong" PLL, that PLL must be clocked lower than the "correct" PLL and be at most 80% of the rated CPU frequency. Possibly because of the lack of the ability to express such nuance in the binding, some valid options are omitted from the device trees, such as the ability on p4080 to run cores 0-3 from PLL3 and cores 4-7 from PLL1 (again, only if they are at most 80% of rated CPU frequency). This omission, combined with excessive caution in the cpufreq driver (addressed in a subsequent patch), means that currently on a 1500 MHz p4080 with typical PLL configuration, cpufreq can lower the frequency to 1200 MHz on half the CPUs and do nothing on the others. With this patchset, all CPUs can be lowered to 1200 MHz on a rev2 p4080, and on a rev3 p4080 half can be lowered to 750 MHz and the other half to 600 MHz. The current binding only deals with CPU clocks. To describe FMan in the device tree, we need to describe its clock. Some chips have additional muxes that work like the CPU muxes, but are not described in the device tree. Others require inspecting the Reset Control Word to determine which PLL is used. Rather than continue to extend this mess, replace it. Have the driver bind to the chip-specific clockgen compatible, and keep the detailed description of quirky chip variations in the driver, where it can be easily fixed, refactored, and extended. Older device trees will continue to work (including a workaround for old ls1021a device trees that are missing compatible and reg in the clockgen node, which even the old binding required). The pll/mux details in old device trees will be ignored, but "clocks" properties pointing at the old nodes will still work, and be directed at the corresponding new clock. Signed-off-by: Scott Wood <scottwood@freescale.com> Acked-by: Stephen Boyd <sboyd@codeaurora.org>
2015-09-20 12:29:54 +08:00
for (i = 0, j = 0; i < NUM_MUX_PARENTS; i++) {
unsigned long rate;
clk: qoriq: Move chip-specific knowledge into driver The device tree should describe the chips (or chip-like subblocks) in the system, but it generally does not describe individual registers -- it should identify, rather than describe, a programming interface. This has not been the case with the QorIQ clockgen nodes. The knowledge of what each bit setting of CLKCnCSR means is encoded in three places (binding, pll node, and mux node), and the last also needs to know which options are valid on a particular chip. All three of these locations are considered stable ABI, making it difficult to fix mistakes (of which I have found several), much less refactor the abstraction to be able to address problems, limitations, or new chips. Under the current binding, a pll clock specifier of 2 means that the PLL is divided by 4 -- and the driver implements this, unless there happen to be four clock-output-names rather than 3, in which case it interprets it as PLL divided by 3. This does not appear in the binding documentation at all. That hack is now considered stable ABI. The current device tree nodes contain errors, such as saying that T1040 can set a core clock to PLL/4 when only PLL and PLL/2 are options. The current binding also ignores some restrictions on clock selection, such as p5020's requirement that if a core uses the "wrong" PLL, that PLL must be clocked lower than the "correct" PLL and be at most 80% of the rated CPU frequency. Possibly because of the lack of the ability to express such nuance in the binding, some valid options are omitted from the device trees, such as the ability on p4080 to run cores 0-3 from PLL3 and cores 4-7 from PLL1 (again, only if they are at most 80% of rated CPU frequency). This omission, combined with excessive caution in the cpufreq driver (addressed in a subsequent patch), means that currently on a 1500 MHz p4080 with typical PLL configuration, cpufreq can lower the frequency to 1200 MHz on half the CPUs and do nothing on the others. With this patchset, all CPUs can be lowered to 1200 MHz on a rev2 p4080, and on a rev3 p4080 half can be lowered to 750 MHz and the other half to 600 MHz. The current binding only deals with CPU clocks. To describe FMan in the device tree, we need to describe its clock. Some chips have additional muxes that work like the CPU muxes, but are not described in the device tree. Others require inspecting the Reset Control Word to determine which PLL is used. Rather than continue to extend this mess, replace it. Have the driver bind to the chip-specific clockgen compatible, and keep the detailed description of quirky chip variations in the driver, where it can be easily fixed, refactored, and extended. Older device trees will continue to work (including a workaround for old ls1021a device trees that are missing compatible and reg in the clockgen node, which even the old binding required). The pll/mux details in old device trees will be ignored, but "clocks" properties pointing at the old nodes will still work, and be directed at the corresponding new clock. Signed-off-by: Scott Wood <scottwood@freescale.com> Acked-by: Stephen Boyd <sboyd@codeaurora.org>
2015-09-20 12:29:54 +08:00
hwc->clksel_to_parent[i] = -1;
clk: qoriq: Move chip-specific knowledge into driver The device tree should describe the chips (or chip-like subblocks) in the system, but it generally does not describe individual registers -- it should identify, rather than describe, a programming interface. This has not been the case with the QorIQ clockgen nodes. The knowledge of what each bit setting of CLKCnCSR means is encoded in three places (binding, pll node, and mux node), and the last also needs to know which options are valid on a particular chip. All three of these locations are considered stable ABI, making it difficult to fix mistakes (of which I have found several), much less refactor the abstraction to be able to address problems, limitations, or new chips. Under the current binding, a pll clock specifier of 2 means that the PLL is divided by 4 -- and the driver implements this, unless there happen to be four clock-output-names rather than 3, in which case it interprets it as PLL divided by 3. This does not appear in the binding documentation at all. That hack is now considered stable ABI. The current device tree nodes contain errors, such as saying that T1040 can set a core clock to PLL/4 when only PLL and PLL/2 are options. The current binding also ignores some restrictions on clock selection, such as p5020's requirement that if a core uses the "wrong" PLL, that PLL must be clocked lower than the "correct" PLL and be at most 80% of the rated CPU frequency. Possibly because of the lack of the ability to express such nuance in the binding, some valid options are omitted from the device trees, such as the ability on p4080 to run cores 0-3 from PLL3 and cores 4-7 from PLL1 (again, only if they are at most 80% of rated CPU frequency). This omission, combined with excessive caution in the cpufreq driver (addressed in a subsequent patch), means that currently on a 1500 MHz p4080 with typical PLL configuration, cpufreq can lower the frequency to 1200 MHz on half the CPUs and do nothing on the others. With this patchset, all CPUs can be lowered to 1200 MHz on a rev2 p4080, and on a rev3 p4080 half can be lowered to 750 MHz and the other half to 600 MHz. The current binding only deals with CPU clocks. To describe FMan in the device tree, we need to describe its clock. Some chips have additional muxes that work like the CPU muxes, but are not described in the device tree. Others require inspecting the Reset Control Word to determine which PLL is used. Rather than continue to extend this mess, replace it. Have the driver bind to the chip-specific clockgen compatible, and keep the detailed description of quirky chip variations in the driver, where it can be easily fixed, refactored, and extended. Older device trees will continue to work (including a workaround for old ls1021a device trees that are missing compatible and reg in the clockgen node, which even the old binding required). The pll/mux details in old device trees will be ignored, but "clocks" properties pointing at the old nodes will still work, and be directed at the corresponding new clock. Signed-off-by: Scott Wood <scottwood@freescale.com> Acked-by: Stephen Boyd <sboyd@codeaurora.org>
2015-09-20 12:29:54 +08:00
div = get_pll_div(cg, hwc, i);
if (!div)
continue;
clk: qoriq: Move chip-specific knowledge into driver The device tree should describe the chips (or chip-like subblocks) in the system, but it generally does not describe individual registers -- it should identify, rather than describe, a programming interface. This has not been the case with the QorIQ clockgen nodes. The knowledge of what each bit setting of CLKCnCSR means is encoded in three places (binding, pll node, and mux node), and the last also needs to know which options are valid on a particular chip. All three of these locations are considered stable ABI, making it difficult to fix mistakes (of which I have found several), much less refactor the abstraction to be able to address problems, limitations, or new chips. Under the current binding, a pll clock specifier of 2 means that the PLL is divided by 4 -- and the driver implements this, unless there happen to be four clock-output-names rather than 3, in which case it interprets it as PLL divided by 3. This does not appear in the binding documentation at all. That hack is now considered stable ABI. The current device tree nodes contain errors, such as saying that T1040 can set a core clock to PLL/4 when only PLL and PLL/2 are options. The current binding also ignores some restrictions on clock selection, such as p5020's requirement that if a core uses the "wrong" PLL, that PLL must be clocked lower than the "correct" PLL and be at most 80% of the rated CPU frequency. Possibly because of the lack of the ability to express such nuance in the binding, some valid options are omitted from the device trees, such as the ability on p4080 to run cores 0-3 from PLL3 and cores 4-7 from PLL1 (again, only if they are at most 80% of rated CPU frequency). This omission, combined with excessive caution in the cpufreq driver (addressed in a subsequent patch), means that currently on a 1500 MHz p4080 with typical PLL configuration, cpufreq can lower the frequency to 1200 MHz on half the CPUs and do nothing on the others. With this patchset, all CPUs can be lowered to 1200 MHz on a rev2 p4080, and on a rev3 p4080 half can be lowered to 750 MHz and the other half to 600 MHz. The current binding only deals with CPU clocks. To describe FMan in the device tree, we need to describe its clock. Some chips have additional muxes that work like the CPU muxes, but are not described in the device tree. Others require inspecting the Reset Control Word to determine which PLL is used. Rather than continue to extend this mess, replace it. Have the driver bind to the chip-specific clockgen compatible, and keep the detailed description of quirky chip variations in the driver, where it can be easily fixed, refactored, and extended. Older device trees will continue to work (including a workaround for old ls1021a device trees that are missing compatible and reg in the clockgen node, which even the old binding required). The pll/mux details in old device trees will be ignored, but "clocks" properties pointing at the old nodes will still work, and be directed at the corresponding new clock. Signed-off-by: Scott Wood <scottwood@freescale.com> Acked-by: Stephen Boyd <sboyd@codeaurora.org>
2015-09-20 12:29:54 +08:00
rate = clk_get_rate(div->clk);
if (hwc->info->clksel[i].flags & CLKSEL_80PCT &&
rate > pct80_rate)
continue;
if (rate < min_rate)
continue;
parent_names[j] = div->name;
hwc->parent_to_clksel[j] = i;
hwc->clksel_to_parent[i] = j;
j++;
}
clk: qoriq: Move chip-specific knowledge into driver The device tree should describe the chips (or chip-like subblocks) in the system, but it generally does not describe individual registers -- it should identify, rather than describe, a programming interface. This has not been the case with the QorIQ clockgen nodes. The knowledge of what each bit setting of CLKCnCSR means is encoded in three places (binding, pll node, and mux node), and the last also needs to know which options are valid on a particular chip. All three of these locations are considered stable ABI, making it difficult to fix mistakes (of which I have found several), much less refactor the abstraction to be able to address problems, limitations, or new chips. Under the current binding, a pll clock specifier of 2 means that the PLL is divided by 4 -- and the driver implements this, unless there happen to be four clock-output-names rather than 3, in which case it interprets it as PLL divided by 3. This does not appear in the binding documentation at all. That hack is now considered stable ABI. The current device tree nodes contain errors, such as saying that T1040 can set a core clock to PLL/4 when only PLL and PLL/2 are options. The current binding also ignores some restrictions on clock selection, such as p5020's requirement that if a core uses the "wrong" PLL, that PLL must be clocked lower than the "correct" PLL and be at most 80% of the rated CPU frequency. Possibly because of the lack of the ability to express such nuance in the binding, some valid options are omitted from the device trees, such as the ability on p4080 to run cores 0-3 from PLL3 and cores 4-7 from PLL1 (again, only if they are at most 80% of rated CPU frequency). This omission, combined with excessive caution in the cpufreq driver (addressed in a subsequent patch), means that currently on a 1500 MHz p4080 with typical PLL configuration, cpufreq can lower the frequency to 1200 MHz on half the CPUs and do nothing on the others. With this patchset, all CPUs can be lowered to 1200 MHz on a rev2 p4080, and on a rev3 p4080 half can be lowered to 750 MHz and the other half to 600 MHz. The current binding only deals with CPU clocks. To describe FMan in the device tree, we need to describe its clock. Some chips have additional muxes that work like the CPU muxes, but are not described in the device tree. Others require inspecting the Reset Control Word to determine which PLL is used. Rather than continue to extend this mess, replace it. Have the driver bind to the chip-specific clockgen compatible, and keep the detailed description of quirky chip variations in the driver, where it can be easily fixed, refactored, and extended. Older device trees will continue to work (including a workaround for old ls1021a device trees that are missing compatible and reg in the clockgen node, which even the old binding required). The pll/mux details in old device trees will be ignored, but "clocks" properties pointing at the old nodes will still work, and be directed at the corresponding new clock. Signed-off-by: Scott Wood <scottwood@freescale.com> Acked-by: Stephen Boyd <sboyd@codeaurora.org>
2015-09-20 12:29:54 +08:00
init.name = name;
init.ops = ops;
init.parent_names = parent_names;
clk: qoriq: Move chip-specific knowledge into driver The device tree should describe the chips (or chip-like subblocks) in the system, but it generally does not describe individual registers -- it should identify, rather than describe, a programming interface. This has not been the case with the QorIQ clockgen nodes. The knowledge of what each bit setting of CLKCnCSR means is encoded in three places (binding, pll node, and mux node), and the last also needs to know which options are valid on a particular chip. All three of these locations are considered stable ABI, making it difficult to fix mistakes (of which I have found several), much less refactor the abstraction to be able to address problems, limitations, or new chips. Under the current binding, a pll clock specifier of 2 means that the PLL is divided by 4 -- and the driver implements this, unless there happen to be four clock-output-names rather than 3, in which case it interprets it as PLL divided by 3. This does not appear in the binding documentation at all. That hack is now considered stable ABI. The current device tree nodes contain errors, such as saying that T1040 can set a core clock to PLL/4 when only PLL and PLL/2 are options. The current binding also ignores some restrictions on clock selection, such as p5020's requirement that if a core uses the "wrong" PLL, that PLL must be clocked lower than the "correct" PLL and be at most 80% of the rated CPU frequency. Possibly because of the lack of the ability to express such nuance in the binding, some valid options are omitted from the device trees, such as the ability on p4080 to run cores 0-3 from PLL3 and cores 4-7 from PLL1 (again, only if they are at most 80% of rated CPU frequency). This omission, combined with excessive caution in the cpufreq driver (addressed in a subsequent patch), means that currently on a 1500 MHz p4080 with typical PLL configuration, cpufreq can lower the frequency to 1200 MHz on half the CPUs and do nothing on the others. With this patchset, all CPUs can be lowered to 1200 MHz on a rev2 p4080, and on a rev3 p4080 half can be lowered to 750 MHz and the other half to 600 MHz. The current binding only deals with CPU clocks. To describe FMan in the device tree, we need to describe its clock. Some chips have additional muxes that work like the CPU muxes, but are not described in the device tree. Others require inspecting the Reset Control Word to determine which PLL is used. Rather than continue to extend this mess, replace it. Have the driver bind to the chip-specific clockgen compatible, and keep the detailed description of quirky chip variations in the driver, where it can be easily fixed, refactored, and extended. Older device trees will continue to work (including a workaround for old ls1021a device trees that are missing compatible and reg in the clockgen node, which even the old binding required). The pll/mux details in old device trees will be ignored, but "clocks" properties pointing at the old nodes will still work, and be directed at the corresponding new clock. Signed-off-by: Scott Wood <scottwood@freescale.com> Acked-by: Stephen Boyd <sboyd@codeaurora.org>
2015-09-20 12:29:54 +08:00
init.num_parents = hwc->num_parents = j;
init.flags = 0;
clk: qoriq: Move chip-specific knowledge into driver The device tree should describe the chips (or chip-like subblocks) in the system, but it generally does not describe individual registers -- it should identify, rather than describe, a programming interface. This has not been the case with the QorIQ clockgen nodes. The knowledge of what each bit setting of CLKCnCSR means is encoded in three places (binding, pll node, and mux node), and the last also needs to know which options are valid on a particular chip. All three of these locations are considered stable ABI, making it difficult to fix mistakes (of which I have found several), much less refactor the abstraction to be able to address problems, limitations, or new chips. Under the current binding, a pll clock specifier of 2 means that the PLL is divided by 4 -- and the driver implements this, unless there happen to be four clock-output-names rather than 3, in which case it interprets it as PLL divided by 3. This does not appear in the binding documentation at all. That hack is now considered stable ABI. The current device tree nodes contain errors, such as saying that T1040 can set a core clock to PLL/4 when only PLL and PLL/2 are options. The current binding also ignores some restrictions on clock selection, such as p5020's requirement that if a core uses the "wrong" PLL, that PLL must be clocked lower than the "correct" PLL and be at most 80% of the rated CPU frequency. Possibly because of the lack of the ability to express such nuance in the binding, some valid options are omitted from the device trees, such as the ability on p4080 to run cores 0-3 from PLL3 and cores 4-7 from PLL1 (again, only if they are at most 80% of rated CPU frequency). This omission, combined with excessive caution in the cpufreq driver (addressed in a subsequent patch), means that currently on a 1500 MHz p4080 with typical PLL configuration, cpufreq can lower the frequency to 1200 MHz on half the CPUs and do nothing on the others. With this patchset, all CPUs can be lowered to 1200 MHz on a rev2 p4080, and on a rev3 p4080 half can be lowered to 750 MHz and the other half to 600 MHz. The current binding only deals with CPU clocks. To describe FMan in the device tree, we need to describe its clock. Some chips have additional muxes that work like the CPU muxes, but are not described in the device tree. Others require inspecting the Reset Control Word to determine which PLL is used. Rather than continue to extend this mess, replace it. Have the driver bind to the chip-specific clockgen compatible, and keep the detailed description of quirky chip variations in the driver, where it can be easily fixed, refactored, and extended. Older device trees will continue to work (including a workaround for old ls1021a device trees that are missing compatible and reg in the clockgen node, which even the old binding required). The pll/mux details in old device trees will be ignored, but "clocks" properties pointing at the old nodes will still work, and be directed at the corresponding new clock. Signed-off-by: Scott Wood <scottwood@freescale.com> Acked-by: Stephen Boyd <sboyd@codeaurora.org>
2015-09-20 12:29:54 +08:00
hwc->hw.init = &init;
hwc->cg = cg;
clk: qoriq: Move chip-specific knowledge into driver The device tree should describe the chips (or chip-like subblocks) in the system, but it generally does not describe individual registers -- it should identify, rather than describe, a programming interface. This has not been the case with the QorIQ clockgen nodes. The knowledge of what each bit setting of CLKCnCSR means is encoded in three places (binding, pll node, and mux node), and the last also needs to know which options are valid on a particular chip. All three of these locations are considered stable ABI, making it difficult to fix mistakes (of which I have found several), much less refactor the abstraction to be able to address problems, limitations, or new chips. Under the current binding, a pll clock specifier of 2 means that the PLL is divided by 4 -- and the driver implements this, unless there happen to be four clock-output-names rather than 3, in which case it interprets it as PLL divided by 3. This does not appear in the binding documentation at all. That hack is now considered stable ABI. The current device tree nodes contain errors, such as saying that T1040 can set a core clock to PLL/4 when only PLL and PLL/2 are options. The current binding also ignores some restrictions on clock selection, such as p5020's requirement that if a core uses the "wrong" PLL, that PLL must be clocked lower than the "correct" PLL and be at most 80% of the rated CPU frequency. Possibly because of the lack of the ability to express such nuance in the binding, some valid options are omitted from the device trees, such as the ability on p4080 to run cores 0-3 from PLL3 and cores 4-7 from PLL1 (again, only if they are at most 80% of rated CPU frequency). This omission, combined with excessive caution in the cpufreq driver (addressed in a subsequent patch), means that currently on a 1500 MHz p4080 with typical PLL configuration, cpufreq can lower the frequency to 1200 MHz on half the CPUs and do nothing on the others. With this patchset, all CPUs can be lowered to 1200 MHz on a rev2 p4080, and on a rev3 p4080 half can be lowered to 750 MHz and the other half to 600 MHz. The current binding only deals with CPU clocks. To describe FMan in the device tree, we need to describe its clock. Some chips have additional muxes that work like the CPU muxes, but are not described in the device tree. Others require inspecting the Reset Control Word to determine which PLL is used. Rather than continue to extend this mess, replace it. Have the driver bind to the chip-specific clockgen compatible, and keep the detailed description of quirky chip variations in the driver, where it can be easily fixed, refactored, and extended. Older device trees will continue to work (including a workaround for old ls1021a device trees that are missing compatible and reg in the clockgen node, which even the old binding required). The pll/mux details in old device trees will be ignored, but "clocks" properties pointing at the old nodes will still work, and be directed at the corresponding new clock. Signed-off-by: Scott Wood <scottwood@freescale.com> Acked-by: Stephen Boyd <sboyd@codeaurora.org>
2015-09-20 12:29:54 +08:00
clk = clk_register(NULL, &hwc->hw);
if (IS_ERR(clk)) {
clk: qoriq: Move chip-specific knowledge into driver The device tree should describe the chips (or chip-like subblocks) in the system, but it generally does not describe individual registers -- it should identify, rather than describe, a programming interface. This has not been the case with the QorIQ clockgen nodes. The knowledge of what each bit setting of CLKCnCSR means is encoded in three places (binding, pll node, and mux node), and the last also needs to know which options are valid on a particular chip. All three of these locations are considered stable ABI, making it difficult to fix mistakes (of which I have found several), much less refactor the abstraction to be able to address problems, limitations, or new chips. Under the current binding, a pll clock specifier of 2 means that the PLL is divided by 4 -- and the driver implements this, unless there happen to be four clock-output-names rather than 3, in which case it interprets it as PLL divided by 3. This does not appear in the binding documentation at all. That hack is now considered stable ABI. The current device tree nodes contain errors, such as saying that T1040 can set a core clock to PLL/4 when only PLL and PLL/2 are options. The current binding also ignores some restrictions on clock selection, such as p5020's requirement that if a core uses the "wrong" PLL, that PLL must be clocked lower than the "correct" PLL and be at most 80% of the rated CPU frequency. Possibly because of the lack of the ability to express such nuance in the binding, some valid options are omitted from the device trees, such as the ability on p4080 to run cores 0-3 from PLL3 and cores 4-7 from PLL1 (again, only if they are at most 80% of rated CPU frequency). This omission, combined with excessive caution in the cpufreq driver (addressed in a subsequent patch), means that currently on a 1500 MHz p4080 with typical PLL configuration, cpufreq can lower the frequency to 1200 MHz on half the CPUs and do nothing on the others. With this patchset, all CPUs can be lowered to 1200 MHz on a rev2 p4080, and on a rev3 p4080 half can be lowered to 750 MHz and the other half to 600 MHz. The current binding only deals with CPU clocks. To describe FMan in the device tree, we need to describe its clock. Some chips have additional muxes that work like the CPU muxes, but are not described in the device tree. Others require inspecting the Reset Control Word to determine which PLL is used. Rather than continue to extend this mess, replace it. Have the driver bind to the chip-specific clockgen compatible, and keep the detailed description of quirky chip variations in the driver, where it can be easily fixed, refactored, and extended. Older device trees will continue to work (including a workaround for old ls1021a device trees that are missing compatible and reg in the clockgen node, which even the old binding required). The pll/mux details in old device trees will be ignored, but "clocks" properties pointing at the old nodes will still work, and be directed at the corresponding new clock. Signed-off-by: Scott Wood <scottwood@freescale.com> Acked-by: Stephen Boyd <sboyd@codeaurora.org>
2015-09-20 12:29:54 +08:00
pr_err("%s: Couldn't register %s: %ld\n", __func__, name,
PTR_ERR(clk));
kfree(hwc);
return NULL;
}
return clk;
}
static struct clk * __init create_one_cmux(struct clockgen *cg, int idx)
{
struct mux_hwclock *hwc;
const struct clockgen_pll_div *div;
unsigned long plat_rate, min_rate;
u64 pct80_rate;
u32 clksel;
hwc = kzalloc(sizeof(*hwc), GFP_KERNEL);
if (!hwc)
return NULL;
if (cg->info.flags & CG_VER3)
hwc->reg = cg->regs + 0x70000 + 0x20 * idx;
else
hwc->reg = cg->regs + 0x20 * idx;
clk: qoriq: Move chip-specific knowledge into driver The device tree should describe the chips (or chip-like subblocks) in the system, but it generally does not describe individual registers -- it should identify, rather than describe, a programming interface. This has not been the case with the QorIQ clockgen nodes. The knowledge of what each bit setting of CLKCnCSR means is encoded in three places (binding, pll node, and mux node), and the last also needs to know which options are valid on a particular chip. All three of these locations are considered stable ABI, making it difficult to fix mistakes (of which I have found several), much less refactor the abstraction to be able to address problems, limitations, or new chips. Under the current binding, a pll clock specifier of 2 means that the PLL is divided by 4 -- and the driver implements this, unless there happen to be four clock-output-names rather than 3, in which case it interprets it as PLL divided by 3. This does not appear in the binding documentation at all. That hack is now considered stable ABI. The current device tree nodes contain errors, such as saying that T1040 can set a core clock to PLL/4 when only PLL and PLL/2 are options. The current binding also ignores some restrictions on clock selection, such as p5020's requirement that if a core uses the "wrong" PLL, that PLL must be clocked lower than the "correct" PLL and be at most 80% of the rated CPU frequency. Possibly because of the lack of the ability to express such nuance in the binding, some valid options are omitted from the device trees, such as the ability on p4080 to run cores 0-3 from PLL3 and cores 4-7 from PLL1 (again, only if they are at most 80% of rated CPU frequency). This omission, combined with excessive caution in the cpufreq driver (addressed in a subsequent patch), means that currently on a 1500 MHz p4080 with typical PLL configuration, cpufreq can lower the frequency to 1200 MHz on half the CPUs and do nothing on the others. With this patchset, all CPUs can be lowered to 1200 MHz on a rev2 p4080, and on a rev3 p4080 half can be lowered to 750 MHz and the other half to 600 MHz. The current binding only deals with CPU clocks. To describe FMan in the device tree, we need to describe its clock. Some chips have additional muxes that work like the CPU muxes, but are not described in the device tree. Others require inspecting the Reset Control Word to determine which PLL is used. Rather than continue to extend this mess, replace it. Have the driver bind to the chip-specific clockgen compatible, and keep the detailed description of quirky chip variations in the driver, where it can be easily fixed, refactored, and extended. Older device trees will continue to work (including a workaround for old ls1021a device trees that are missing compatible and reg in the clockgen node, which even the old binding required). The pll/mux details in old device trees will be ignored, but "clocks" properties pointing at the old nodes will still work, and be directed at the corresponding new clock. Signed-off-by: Scott Wood <scottwood@freescale.com> Acked-by: Stephen Boyd <sboyd@codeaurora.org>
2015-09-20 12:29:54 +08:00
hwc->info = cg->info.cmux_groups[cg->info.cmux_to_group[idx]];
/*
* Find the rate for the default clksel, and treat it as the
* maximum rated core frequency. If this is an incorrect
* assumption, certain clock options (possibly including the
* default clksel) may be inappropriately excluded on certain
* chips.
*/
clksel = (cg_in(cg, hwc->reg) & CLKSEL_MASK) >> CLKSEL_SHIFT;
clk: qoriq: Move chip-specific knowledge into driver The device tree should describe the chips (or chip-like subblocks) in the system, but it generally does not describe individual registers -- it should identify, rather than describe, a programming interface. This has not been the case with the QorIQ clockgen nodes. The knowledge of what each bit setting of CLKCnCSR means is encoded in three places (binding, pll node, and mux node), and the last also needs to know which options are valid on a particular chip. All three of these locations are considered stable ABI, making it difficult to fix mistakes (of which I have found several), much less refactor the abstraction to be able to address problems, limitations, or new chips. Under the current binding, a pll clock specifier of 2 means that the PLL is divided by 4 -- and the driver implements this, unless there happen to be four clock-output-names rather than 3, in which case it interprets it as PLL divided by 3. This does not appear in the binding documentation at all. That hack is now considered stable ABI. The current device tree nodes contain errors, such as saying that T1040 can set a core clock to PLL/4 when only PLL and PLL/2 are options. The current binding also ignores some restrictions on clock selection, such as p5020's requirement that if a core uses the "wrong" PLL, that PLL must be clocked lower than the "correct" PLL and be at most 80% of the rated CPU frequency. Possibly because of the lack of the ability to express such nuance in the binding, some valid options are omitted from the device trees, such as the ability on p4080 to run cores 0-3 from PLL3 and cores 4-7 from PLL1 (again, only if they are at most 80% of rated CPU frequency). This omission, combined with excessive caution in the cpufreq driver (addressed in a subsequent patch), means that currently on a 1500 MHz p4080 with typical PLL configuration, cpufreq can lower the frequency to 1200 MHz on half the CPUs and do nothing on the others. With this patchset, all CPUs can be lowered to 1200 MHz on a rev2 p4080, and on a rev3 p4080 half can be lowered to 750 MHz and the other half to 600 MHz. The current binding only deals with CPU clocks. To describe FMan in the device tree, we need to describe its clock. Some chips have additional muxes that work like the CPU muxes, but are not described in the device tree. Others require inspecting the Reset Control Word to determine which PLL is used. Rather than continue to extend this mess, replace it. Have the driver bind to the chip-specific clockgen compatible, and keep the detailed description of quirky chip variations in the driver, where it can be easily fixed, refactored, and extended. Older device trees will continue to work (including a workaround for old ls1021a device trees that are missing compatible and reg in the clockgen node, which even the old binding required). The pll/mux details in old device trees will be ignored, but "clocks" properties pointing at the old nodes will still work, and be directed at the corresponding new clock. Signed-off-by: Scott Wood <scottwood@freescale.com> Acked-by: Stephen Boyd <sboyd@codeaurora.org>
2015-09-20 12:29:54 +08:00
div = get_pll_div(cg, hwc, clksel);
if (!div) {
kfree(hwc);
clk: qoriq: Move chip-specific knowledge into driver The device tree should describe the chips (or chip-like subblocks) in the system, but it generally does not describe individual registers -- it should identify, rather than describe, a programming interface. This has not been the case with the QorIQ clockgen nodes. The knowledge of what each bit setting of CLKCnCSR means is encoded in three places (binding, pll node, and mux node), and the last also needs to know which options are valid on a particular chip. All three of these locations are considered stable ABI, making it difficult to fix mistakes (of which I have found several), much less refactor the abstraction to be able to address problems, limitations, or new chips. Under the current binding, a pll clock specifier of 2 means that the PLL is divided by 4 -- and the driver implements this, unless there happen to be four clock-output-names rather than 3, in which case it interprets it as PLL divided by 3. This does not appear in the binding documentation at all. That hack is now considered stable ABI. The current device tree nodes contain errors, such as saying that T1040 can set a core clock to PLL/4 when only PLL and PLL/2 are options. The current binding also ignores some restrictions on clock selection, such as p5020's requirement that if a core uses the "wrong" PLL, that PLL must be clocked lower than the "correct" PLL and be at most 80% of the rated CPU frequency. Possibly because of the lack of the ability to express such nuance in the binding, some valid options are omitted from the device trees, such as the ability on p4080 to run cores 0-3 from PLL3 and cores 4-7 from PLL1 (again, only if they are at most 80% of rated CPU frequency). This omission, combined with excessive caution in the cpufreq driver (addressed in a subsequent patch), means that currently on a 1500 MHz p4080 with typical PLL configuration, cpufreq can lower the frequency to 1200 MHz on half the CPUs and do nothing on the others. With this patchset, all CPUs can be lowered to 1200 MHz on a rev2 p4080, and on a rev3 p4080 half can be lowered to 750 MHz and the other half to 600 MHz. The current binding only deals with CPU clocks. To describe FMan in the device tree, we need to describe its clock. Some chips have additional muxes that work like the CPU muxes, but are not described in the device tree. Others require inspecting the Reset Control Word to determine which PLL is used. Rather than continue to extend this mess, replace it. Have the driver bind to the chip-specific clockgen compatible, and keep the detailed description of quirky chip variations in the driver, where it can be easily fixed, refactored, and extended. Older device trees will continue to work (including a workaround for old ls1021a device trees that are missing compatible and reg in the clockgen node, which even the old binding required). The pll/mux details in old device trees will be ignored, but "clocks" properties pointing at the old nodes will still work, and be directed at the corresponding new clock. Signed-off-by: Scott Wood <scottwood@freescale.com> Acked-by: Stephen Boyd <sboyd@codeaurora.org>
2015-09-20 12:29:54 +08:00
return NULL;
}
clk: qoriq: Move chip-specific knowledge into driver The device tree should describe the chips (or chip-like subblocks) in the system, but it generally does not describe individual registers -- it should identify, rather than describe, a programming interface. This has not been the case with the QorIQ clockgen nodes. The knowledge of what each bit setting of CLKCnCSR means is encoded in three places (binding, pll node, and mux node), and the last also needs to know which options are valid on a particular chip. All three of these locations are considered stable ABI, making it difficult to fix mistakes (of which I have found several), much less refactor the abstraction to be able to address problems, limitations, or new chips. Under the current binding, a pll clock specifier of 2 means that the PLL is divided by 4 -- and the driver implements this, unless there happen to be four clock-output-names rather than 3, in which case it interprets it as PLL divided by 3. This does not appear in the binding documentation at all. That hack is now considered stable ABI. The current device tree nodes contain errors, such as saying that T1040 can set a core clock to PLL/4 when only PLL and PLL/2 are options. The current binding also ignores some restrictions on clock selection, such as p5020's requirement that if a core uses the "wrong" PLL, that PLL must be clocked lower than the "correct" PLL and be at most 80% of the rated CPU frequency. Possibly because of the lack of the ability to express such nuance in the binding, some valid options are omitted from the device trees, such as the ability on p4080 to run cores 0-3 from PLL3 and cores 4-7 from PLL1 (again, only if they are at most 80% of rated CPU frequency). This omission, combined with excessive caution in the cpufreq driver (addressed in a subsequent patch), means that currently on a 1500 MHz p4080 with typical PLL configuration, cpufreq can lower the frequency to 1200 MHz on half the CPUs and do nothing on the others. With this patchset, all CPUs can be lowered to 1200 MHz on a rev2 p4080, and on a rev3 p4080 half can be lowered to 750 MHz and the other half to 600 MHz. The current binding only deals with CPU clocks. To describe FMan in the device tree, we need to describe its clock. Some chips have additional muxes that work like the CPU muxes, but are not described in the device tree. Others require inspecting the Reset Control Word to determine which PLL is used. Rather than continue to extend this mess, replace it. Have the driver bind to the chip-specific clockgen compatible, and keep the detailed description of quirky chip variations in the driver, where it can be easily fixed, refactored, and extended. Older device trees will continue to work (including a workaround for old ls1021a device trees that are missing compatible and reg in the clockgen node, which even the old binding required). The pll/mux details in old device trees will be ignored, but "clocks" properties pointing at the old nodes will still work, and be directed at the corresponding new clock. Signed-off-by: Scott Wood <scottwood@freescale.com> Acked-by: Stephen Boyd <sboyd@codeaurora.org>
2015-09-20 12:29:54 +08:00
pct80_rate = clk_get_rate(div->clk);
pct80_rate *= 8;
do_div(pct80_rate, 10);
plat_rate = clk_get_rate(cg->pll[PLATFORM_PLL].div[PLL_DIV1].clk);
if (cg->info.flags & CG_CMUX_GE_PLAT)
min_rate = plat_rate;
else
min_rate = plat_rate / 2;
return create_mux_common(cg, hwc, &cmux_ops, min_rate,
pct80_rate, "cg-cmux%d", idx);
}
static struct clk * __init create_one_hwaccel(struct clockgen *cg, int idx)
{
struct mux_hwclock *hwc;
hwc = kzalloc(sizeof(*hwc), GFP_KERNEL);
if (!hwc)
return NULL;
hwc->reg = cg->regs + 0x20 * idx + 0x10;
hwc->info = cg->info.hwaccel[idx];
return create_mux_common(cg, hwc, &hwaccel_ops, 0, 0,
"cg-hwaccel%d", idx);
}
static void __init create_muxes(struct clockgen *cg)
{
int i;
for (i = 0; i < ARRAY_SIZE(cg->cmux); i++) {
if (cg->info.cmux_to_group[i] < 0)
break;
if (cg->info.cmux_to_group[i] >=
ARRAY_SIZE(cg->info.cmux_groups)) {
WARN_ON_ONCE(1);
continue;
}
cg->cmux[i] = create_one_cmux(cg, i);
}
clk: qoriq: Move chip-specific knowledge into driver The device tree should describe the chips (or chip-like subblocks) in the system, but it generally does not describe individual registers -- it should identify, rather than describe, a programming interface. This has not been the case with the QorIQ clockgen nodes. The knowledge of what each bit setting of CLKCnCSR means is encoded in three places (binding, pll node, and mux node), and the last also needs to know which options are valid on a particular chip. All three of these locations are considered stable ABI, making it difficult to fix mistakes (of which I have found several), much less refactor the abstraction to be able to address problems, limitations, or new chips. Under the current binding, a pll clock specifier of 2 means that the PLL is divided by 4 -- and the driver implements this, unless there happen to be four clock-output-names rather than 3, in which case it interprets it as PLL divided by 3. This does not appear in the binding documentation at all. That hack is now considered stable ABI. The current device tree nodes contain errors, such as saying that T1040 can set a core clock to PLL/4 when only PLL and PLL/2 are options. The current binding also ignores some restrictions on clock selection, such as p5020's requirement that if a core uses the "wrong" PLL, that PLL must be clocked lower than the "correct" PLL and be at most 80% of the rated CPU frequency. Possibly because of the lack of the ability to express such nuance in the binding, some valid options are omitted from the device trees, such as the ability on p4080 to run cores 0-3 from PLL3 and cores 4-7 from PLL1 (again, only if they are at most 80% of rated CPU frequency). This omission, combined with excessive caution in the cpufreq driver (addressed in a subsequent patch), means that currently on a 1500 MHz p4080 with typical PLL configuration, cpufreq can lower the frequency to 1200 MHz on half the CPUs and do nothing on the others. With this patchset, all CPUs can be lowered to 1200 MHz on a rev2 p4080, and on a rev3 p4080 half can be lowered to 750 MHz and the other half to 600 MHz. The current binding only deals with CPU clocks. To describe FMan in the device tree, we need to describe its clock. Some chips have additional muxes that work like the CPU muxes, but are not described in the device tree. Others require inspecting the Reset Control Word to determine which PLL is used. Rather than continue to extend this mess, replace it. Have the driver bind to the chip-specific clockgen compatible, and keep the detailed description of quirky chip variations in the driver, where it can be easily fixed, refactored, and extended. Older device trees will continue to work (including a workaround for old ls1021a device trees that are missing compatible and reg in the clockgen node, which even the old binding required). The pll/mux details in old device trees will be ignored, but "clocks" properties pointing at the old nodes will still work, and be directed at the corresponding new clock. Signed-off-by: Scott Wood <scottwood@freescale.com> Acked-by: Stephen Boyd <sboyd@codeaurora.org>
2015-09-20 12:29:54 +08:00
for (i = 0; i < ARRAY_SIZE(cg->hwaccel); i++) {
if (!cg->info.hwaccel[i])
continue;
cg->hwaccel[i] = create_one_hwaccel(cg, i);
}
}
static void __init clockgen_init(struct device_node *np);
/* Legacy nodes may get probed before the parent clockgen node */
static void __init legacy_init_clockgen(struct device_node *np)
{
if (!clockgen.node)
clockgen_init(of_get_parent(np));
}
/* Legacy node */
static void __init core_mux_init(struct device_node *np)
{
struct clk *clk;
struct resource res;
int idx, rc;
legacy_init_clockgen(np);
if (of_address_to_resource(np, 0, &res))
return;
idx = (res.start & 0xf0) >> 5;
clk = clockgen.cmux[idx];
rc = of_clk_add_provider(np, of_clk_src_simple_get, clk);
if (rc) {
clk: qoriq: Move chip-specific knowledge into driver The device tree should describe the chips (or chip-like subblocks) in the system, but it generally does not describe individual registers -- it should identify, rather than describe, a programming interface. This has not been the case with the QorIQ clockgen nodes. The knowledge of what each bit setting of CLKCnCSR means is encoded in three places (binding, pll node, and mux node), and the last also needs to know which options are valid on a particular chip. All three of these locations are considered stable ABI, making it difficult to fix mistakes (of which I have found several), much less refactor the abstraction to be able to address problems, limitations, or new chips. Under the current binding, a pll clock specifier of 2 means that the PLL is divided by 4 -- and the driver implements this, unless there happen to be four clock-output-names rather than 3, in which case it interprets it as PLL divided by 3. This does not appear in the binding documentation at all. That hack is now considered stable ABI. The current device tree nodes contain errors, such as saying that T1040 can set a core clock to PLL/4 when only PLL and PLL/2 are options. The current binding also ignores some restrictions on clock selection, such as p5020's requirement that if a core uses the "wrong" PLL, that PLL must be clocked lower than the "correct" PLL and be at most 80% of the rated CPU frequency. Possibly because of the lack of the ability to express such nuance in the binding, some valid options are omitted from the device trees, such as the ability on p4080 to run cores 0-3 from PLL3 and cores 4-7 from PLL1 (again, only if they are at most 80% of rated CPU frequency). This omission, combined with excessive caution in the cpufreq driver (addressed in a subsequent patch), means that currently on a 1500 MHz p4080 with typical PLL configuration, cpufreq can lower the frequency to 1200 MHz on half the CPUs and do nothing on the others. With this patchset, all CPUs can be lowered to 1200 MHz on a rev2 p4080, and on a rev3 p4080 half can be lowered to 750 MHz and the other half to 600 MHz. The current binding only deals with CPU clocks. To describe FMan in the device tree, we need to describe its clock. Some chips have additional muxes that work like the CPU muxes, but are not described in the device tree. Others require inspecting the Reset Control Word to determine which PLL is used. Rather than continue to extend this mess, replace it. Have the driver bind to the chip-specific clockgen compatible, and keep the detailed description of quirky chip variations in the driver, where it can be easily fixed, refactored, and extended. Older device trees will continue to work (including a workaround for old ls1021a device trees that are missing compatible and reg in the clockgen node, which even the old binding required). The pll/mux details in old device trees will be ignored, but "clocks" properties pointing at the old nodes will still work, and be directed at the corresponding new clock. Signed-off-by: Scott Wood <scottwood@freescale.com> Acked-by: Stephen Boyd <sboyd@codeaurora.org>
2015-09-20 12:29:54 +08:00
pr_err("%s: Couldn't register clk provider for node %s: %d\n",
__func__, np->name, rc);
return;
}
clk: qoriq: Move chip-specific knowledge into driver The device tree should describe the chips (or chip-like subblocks) in the system, but it generally does not describe individual registers -- it should identify, rather than describe, a programming interface. This has not been the case with the QorIQ clockgen nodes. The knowledge of what each bit setting of CLKCnCSR means is encoded in three places (binding, pll node, and mux node), and the last also needs to know which options are valid on a particular chip. All three of these locations are considered stable ABI, making it difficult to fix mistakes (of which I have found several), much less refactor the abstraction to be able to address problems, limitations, or new chips. Under the current binding, a pll clock specifier of 2 means that the PLL is divided by 4 -- and the driver implements this, unless there happen to be four clock-output-names rather than 3, in which case it interprets it as PLL divided by 3. This does not appear in the binding documentation at all. That hack is now considered stable ABI. The current device tree nodes contain errors, such as saying that T1040 can set a core clock to PLL/4 when only PLL and PLL/2 are options. The current binding also ignores some restrictions on clock selection, such as p5020's requirement that if a core uses the "wrong" PLL, that PLL must be clocked lower than the "correct" PLL and be at most 80% of the rated CPU frequency. Possibly because of the lack of the ability to express such nuance in the binding, some valid options are omitted from the device trees, such as the ability on p4080 to run cores 0-3 from PLL3 and cores 4-7 from PLL1 (again, only if they are at most 80% of rated CPU frequency). This omission, combined with excessive caution in the cpufreq driver (addressed in a subsequent patch), means that currently on a 1500 MHz p4080 with typical PLL configuration, cpufreq can lower the frequency to 1200 MHz on half the CPUs and do nothing on the others. With this patchset, all CPUs can be lowered to 1200 MHz on a rev2 p4080, and on a rev3 p4080 half can be lowered to 750 MHz and the other half to 600 MHz. The current binding only deals with CPU clocks. To describe FMan in the device tree, we need to describe its clock. Some chips have additional muxes that work like the CPU muxes, but are not described in the device tree. Others require inspecting the Reset Control Word to determine which PLL is used. Rather than continue to extend this mess, replace it. Have the driver bind to the chip-specific clockgen compatible, and keep the detailed description of quirky chip variations in the driver, where it can be easily fixed, refactored, and extended. Older device trees will continue to work (including a workaround for old ls1021a device trees that are missing compatible and reg in the clockgen node, which even the old binding required). The pll/mux details in old device trees will be ignored, but "clocks" properties pointing at the old nodes will still work, and be directed at the corresponding new clock. Signed-off-by: Scott Wood <scottwood@freescale.com> Acked-by: Stephen Boyd <sboyd@codeaurora.org>
2015-09-20 12:29:54 +08:00
}
static struct clk __init
*sysclk_from_fixed(struct device_node *node, const char *name)
clk: qoriq: Move chip-specific knowledge into driver The device tree should describe the chips (or chip-like subblocks) in the system, but it generally does not describe individual registers -- it should identify, rather than describe, a programming interface. This has not been the case with the QorIQ clockgen nodes. The knowledge of what each bit setting of CLKCnCSR means is encoded in three places (binding, pll node, and mux node), and the last also needs to know which options are valid on a particular chip. All three of these locations are considered stable ABI, making it difficult to fix mistakes (of which I have found several), much less refactor the abstraction to be able to address problems, limitations, or new chips. Under the current binding, a pll clock specifier of 2 means that the PLL is divided by 4 -- and the driver implements this, unless there happen to be four clock-output-names rather than 3, in which case it interprets it as PLL divided by 3. This does not appear in the binding documentation at all. That hack is now considered stable ABI. The current device tree nodes contain errors, such as saying that T1040 can set a core clock to PLL/4 when only PLL and PLL/2 are options. The current binding also ignores some restrictions on clock selection, such as p5020's requirement that if a core uses the "wrong" PLL, that PLL must be clocked lower than the "correct" PLL and be at most 80% of the rated CPU frequency. Possibly because of the lack of the ability to express such nuance in the binding, some valid options are omitted from the device trees, such as the ability on p4080 to run cores 0-3 from PLL3 and cores 4-7 from PLL1 (again, only if they are at most 80% of rated CPU frequency). This omission, combined with excessive caution in the cpufreq driver (addressed in a subsequent patch), means that currently on a 1500 MHz p4080 with typical PLL configuration, cpufreq can lower the frequency to 1200 MHz on half the CPUs and do nothing on the others. With this patchset, all CPUs can be lowered to 1200 MHz on a rev2 p4080, and on a rev3 p4080 half can be lowered to 750 MHz and the other half to 600 MHz. The current binding only deals with CPU clocks. To describe FMan in the device tree, we need to describe its clock. Some chips have additional muxes that work like the CPU muxes, but are not described in the device tree. Others require inspecting the Reset Control Word to determine which PLL is used. Rather than continue to extend this mess, replace it. Have the driver bind to the chip-specific clockgen compatible, and keep the detailed description of quirky chip variations in the driver, where it can be easily fixed, refactored, and extended. Older device trees will continue to work (including a workaround for old ls1021a device trees that are missing compatible and reg in the clockgen node, which even the old binding required). The pll/mux details in old device trees will be ignored, but "clocks" properties pointing at the old nodes will still work, and be directed at the corresponding new clock. Signed-off-by: Scott Wood <scottwood@freescale.com> Acked-by: Stephen Boyd <sboyd@codeaurora.org>
2015-09-20 12:29:54 +08:00
{
u32 rate;
if (of_property_read_u32(node, "clock-frequency", &rate))
return ERR_PTR(-ENODEV);
return clk_register_fixed_rate(NULL, name, NULL, 0, rate);
}
clk: qoriq: Move chip-specific knowledge into driver The device tree should describe the chips (or chip-like subblocks) in the system, but it generally does not describe individual registers -- it should identify, rather than describe, a programming interface. This has not been the case with the QorIQ clockgen nodes. The knowledge of what each bit setting of CLKCnCSR means is encoded in three places (binding, pll node, and mux node), and the last also needs to know which options are valid on a particular chip. All three of these locations are considered stable ABI, making it difficult to fix mistakes (of which I have found several), much less refactor the abstraction to be able to address problems, limitations, or new chips. Under the current binding, a pll clock specifier of 2 means that the PLL is divided by 4 -- and the driver implements this, unless there happen to be four clock-output-names rather than 3, in which case it interprets it as PLL divided by 3. This does not appear in the binding documentation at all. That hack is now considered stable ABI. The current device tree nodes contain errors, such as saying that T1040 can set a core clock to PLL/4 when only PLL and PLL/2 are options. The current binding also ignores some restrictions on clock selection, such as p5020's requirement that if a core uses the "wrong" PLL, that PLL must be clocked lower than the "correct" PLL and be at most 80% of the rated CPU frequency. Possibly because of the lack of the ability to express such nuance in the binding, some valid options are omitted from the device trees, such as the ability on p4080 to run cores 0-3 from PLL3 and cores 4-7 from PLL1 (again, only if they are at most 80% of rated CPU frequency). This omission, combined with excessive caution in the cpufreq driver (addressed in a subsequent patch), means that currently on a 1500 MHz p4080 with typical PLL configuration, cpufreq can lower the frequency to 1200 MHz on half the CPUs and do nothing on the others. With this patchset, all CPUs can be lowered to 1200 MHz on a rev2 p4080, and on a rev3 p4080 half can be lowered to 750 MHz and the other half to 600 MHz. The current binding only deals with CPU clocks. To describe FMan in the device tree, we need to describe its clock. Some chips have additional muxes that work like the CPU muxes, but are not described in the device tree. Others require inspecting the Reset Control Word to determine which PLL is used. Rather than continue to extend this mess, replace it. Have the driver bind to the chip-specific clockgen compatible, and keep the detailed description of quirky chip variations in the driver, where it can be easily fixed, refactored, and extended. Older device trees will continue to work (including a workaround for old ls1021a device trees that are missing compatible and reg in the clockgen node, which even the old binding required). The pll/mux details in old device trees will be ignored, but "clocks" properties pointing at the old nodes will still work, and be directed at the corresponding new clock. Signed-off-by: Scott Wood <scottwood@freescale.com> Acked-by: Stephen Boyd <sboyd@codeaurora.org>
2015-09-20 12:29:54 +08:00
static struct clk *sysclk_from_parent(const char *name)
{
struct clk *clk;
const char *parent_name;
clk = of_clk_get(clockgen.node, 0);
if (IS_ERR(clk))
return clk;
/* Register the input clock under the desired name. */
parent_name = __clk_get_name(clk);
clk = clk_register_fixed_factor(NULL, name, parent_name,
0, 1, 1);
if (IS_ERR(clk))
pr_err("%s: Couldn't register %s: %ld\n", __func__, name,
PTR_ERR(clk));
return clk;
}
static struct clk * __init create_sysclk(const char *name)
{
struct device_node *sysclk;
struct clk *clk;
clk = sysclk_from_fixed(clockgen.node, name);
if (!IS_ERR(clk))
return clk;
clk = sysclk_from_parent(name);
if (!IS_ERR(clk))
return clk;
sysclk = of_get_child_by_name(clockgen.node, "sysclk");
if (sysclk) {
clk = sysclk_from_fixed(sysclk, name);
if (!IS_ERR(clk))
return clk;
}
pr_err("%s: No input clock\n", __func__);
return NULL;
}
/* Legacy node */
static void __init sysclk_init(struct device_node *node)
{
struct clk *clk;
legacy_init_clockgen(node);
clk = clockgen.sysclk;
if (clk)
of_clk_add_provider(node, of_clk_src_simple_get, clk);
}
#define PLL_KILL BIT(31)
static void __init create_one_pll(struct clockgen *cg, int idx)
{
clk: qoriq: Move chip-specific knowledge into driver The device tree should describe the chips (or chip-like subblocks) in the system, but it generally does not describe individual registers -- it should identify, rather than describe, a programming interface. This has not been the case with the QorIQ clockgen nodes. The knowledge of what each bit setting of CLKCnCSR means is encoded in three places (binding, pll node, and mux node), and the last also needs to know which options are valid on a particular chip. All three of these locations are considered stable ABI, making it difficult to fix mistakes (of which I have found several), much less refactor the abstraction to be able to address problems, limitations, or new chips. Under the current binding, a pll clock specifier of 2 means that the PLL is divided by 4 -- and the driver implements this, unless there happen to be four clock-output-names rather than 3, in which case it interprets it as PLL divided by 3. This does not appear in the binding documentation at all. That hack is now considered stable ABI. The current device tree nodes contain errors, such as saying that T1040 can set a core clock to PLL/4 when only PLL and PLL/2 are options. The current binding also ignores some restrictions on clock selection, such as p5020's requirement that if a core uses the "wrong" PLL, that PLL must be clocked lower than the "correct" PLL and be at most 80% of the rated CPU frequency. Possibly because of the lack of the ability to express such nuance in the binding, some valid options are omitted from the device trees, such as the ability on p4080 to run cores 0-3 from PLL3 and cores 4-7 from PLL1 (again, only if they are at most 80% of rated CPU frequency). This omission, combined with excessive caution in the cpufreq driver (addressed in a subsequent patch), means that currently on a 1500 MHz p4080 with typical PLL configuration, cpufreq can lower the frequency to 1200 MHz on half the CPUs and do nothing on the others. With this patchset, all CPUs can be lowered to 1200 MHz on a rev2 p4080, and on a rev3 p4080 half can be lowered to 750 MHz and the other half to 600 MHz. The current binding only deals with CPU clocks. To describe FMan in the device tree, we need to describe its clock. Some chips have additional muxes that work like the CPU muxes, but are not described in the device tree. Others require inspecting the Reset Control Word to determine which PLL is used. Rather than continue to extend this mess, replace it. Have the driver bind to the chip-specific clockgen compatible, and keep the detailed description of quirky chip variations in the driver, where it can be easily fixed, refactored, and extended. Older device trees will continue to work (including a workaround for old ls1021a device trees that are missing compatible and reg in the clockgen node, which even the old binding required). The pll/mux details in old device trees will be ignored, but "clocks" properties pointing at the old nodes will still work, and be directed at the corresponding new clock. Signed-off-by: Scott Wood <scottwood@freescale.com> Acked-by: Stephen Boyd <sboyd@codeaurora.org>
2015-09-20 12:29:54 +08:00
u32 __iomem *reg;
u32 mult;
clk: qoriq: Move chip-specific knowledge into driver The device tree should describe the chips (or chip-like subblocks) in the system, but it generally does not describe individual registers -- it should identify, rather than describe, a programming interface. This has not been the case with the QorIQ clockgen nodes. The knowledge of what each bit setting of CLKCnCSR means is encoded in three places (binding, pll node, and mux node), and the last also needs to know which options are valid on a particular chip. All three of these locations are considered stable ABI, making it difficult to fix mistakes (of which I have found several), much less refactor the abstraction to be able to address problems, limitations, or new chips. Under the current binding, a pll clock specifier of 2 means that the PLL is divided by 4 -- and the driver implements this, unless there happen to be four clock-output-names rather than 3, in which case it interprets it as PLL divided by 3. This does not appear in the binding documentation at all. That hack is now considered stable ABI. The current device tree nodes contain errors, such as saying that T1040 can set a core clock to PLL/4 when only PLL and PLL/2 are options. The current binding also ignores some restrictions on clock selection, such as p5020's requirement that if a core uses the "wrong" PLL, that PLL must be clocked lower than the "correct" PLL and be at most 80% of the rated CPU frequency. Possibly because of the lack of the ability to express such nuance in the binding, some valid options are omitted from the device trees, such as the ability on p4080 to run cores 0-3 from PLL3 and cores 4-7 from PLL1 (again, only if they are at most 80% of rated CPU frequency). This omission, combined with excessive caution in the cpufreq driver (addressed in a subsequent patch), means that currently on a 1500 MHz p4080 with typical PLL configuration, cpufreq can lower the frequency to 1200 MHz on half the CPUs and do nothing on the others. With this patchset, all CPUs can be lowered to 1200 MHz on a rev2 p4080, and on a rev3 p4080 half can be lowered to 750 MHz and the other half to 600 MHz. The current binding only deals with CPU clocks. To describe FMan in the device tree, we need to describe its clock. Some chips have additional muxes that work like the CPU muxes, but are not described in the device tree. Others require inspecting the Reset Control Word to determine which PLL is used. Rather than continue to extend this mess, replace it. Have the driver bind to the chip-specific clockgen compatible, and keep the detailed description of quirky chip variations in the driver, where it can be easily fixed, refactored, and extended. Older device trees will continue to work (including a workaround for old ls1021a device trees that are missing compatible and reg in the clockgen node, which even the old binding required). The pll/mux details in old device trees will be ignored, but "clocks" properties pointing at the old nodes will still work, and be directed at the corresponding new clock. Signed-off-by: Scott Wood <scottwood@freescale.com> Acked-by: Stephen Boyd <sboyd@codeaurora.org>
2015-09-20 12:29:54 +08:00
struct clockgen_pll *pll = &cg->pll[idx];
int i;
clk: qoriq: Move chip-specific knowledge into driver The device tree should describe the chips (or chip-like subblocks) in the system, but it generally does not describe individual registers -- it should identify, rather than describe, a programming interface. This has not been the case with the QorIQ clockgen nodes. The knowledge of what each bit setting of CLKCnCSR means is encoded in three places (binding, pll node, and mux node), and the last also needs to know which options are valid on a particular chip. All three of these locations are considered stable ABI, making it difficult to fix mistakes (of which I have found several), much less refactor the abstraction to be able to address problems, limitations, or new chips. Under the current binding, a pll clock specifier of 2 means that the PLL is divided by 4 -- and the driver implements this, unless there happen to be four clock-output-names rather than 3, in which case it interprets it as PLL divided by 3. This does not appear in the binding documentation at all. That hack is now considered stable ABI. The current device tree nodes contain errors, such as saying that T1040 can set a core clock to PLL/4 when only PLL and PLL/2 are options. The current binding also ignores some restrictions on clock selection, such as p5020's requirement that if a core uses the "wrong" PLL, that PLL must be clocked lower than the "correct" PLL and be at most 80% of the rated CPU frequency. Possibly because of the lack of the ability to express such nuance in the binding, some valid options are omitted from the device trees, such as the ability on p4080 to run cores 0-3 from PLL3 and cores 4-7 from PLL1 (again, only if they are at most 80% of rated CPU frequency). This omission, combined with excessive caution in the cpufreq driver (addressed in a subsequent patch), means that currently on a 1500 MHz p4080 with typical PLL configuration, cpufreq can lower the frequency to 1200 MHz on half the CPUs and do nothing on the others. With this patchset, all CPUs can be lowered to 1200 MHz on a rev2 p4080, and on a rev3 p4080 half can be lowered to 750 MHz and the other half to 600 MHz. The current binding only deals with CPU clocks. To describe FMan in the device tree, we need to describe its clock. Some chips have additional muxes that work like the CPU muxes, but are not described in the device tree. Others require inspecting the Reset Control Word to determine which PLL is used. Rather than continue to extend this mess, replace it. Have the driver bind to the chip-specific clockgen compatible, and keep the detailed description of quirky chip variations in the driver, where it can be easily fixed, refactored, and extended. Older device trees will continue to work (including a workaround for old ls1021a device trees that are missing compatible and reg in the clockgen node, which even the old binding required). The pll/mux details in old device trees will be ignored, but "clocks" properties pointing at the old nodes will still work, and be directed at the corresponding new clock. Signed-off-by: Scott Wood <scottwood@freescale.com> Acked-by: Stephen Boyd <sboyd@codeaurora.org>
2015-09-20 12:29:54 +08:00
if (!(cg->info.pll_mask & (1 << idx)))
return;
if (cg->info.flags & CG_VER3) {
switch (idx) {
case PLATFORM_PLL:
reg = cg->regs + 0x60080;
break;
case CGA_PLL1:
reg = cg->regs + 0x80;
break;
case CGA_PLL2:
reg = cg->regs + 0xa0;
break;
case CGB_PLL1:
reg = cg->regs + 0x10080;
break;
case CGB_PLL2:
reg = cg->regs + 0x100a0;
break;
default:
WARN_ONCE(1, "index %d\n", idx);
return;
}
} else {
if (idx == PLATFORM_PLL)
reg = cg->regs + 0xc00;
else
reg = cg->regs + 0x800 + 0x20 * (idx - 1);
}
clk: qoriq: Move chip-specific knowledge into driver The device tree should describe the chips (or chip-like subblocks) in the system, but it generally does not describe individual registers -- it should identify, rather than describe, a programming interface. This has not been the case with the QorIQ clockgen nodes. The knowledge of what each bit setting of CLKCnCSR means is encoded in three places (binding, pll node, and mux node), and the last also needs to know which options are valid on a particular chip. All three of these locations are considered stable ABI, making it difficult to fix mistakes (of which I have found several), much less refactor the abstraction to be able to address problems, limitations, or new chips. Under the current binding, a pll clock specifier of 2 means that the PLL is divided by 4 -- and the driver implements this, unless there happen to be four clock-output-names rather than 3, in which case it interprets it as PLL divided by 3. This does not appear in the binding documentation at all. That hack is now considered stable ABI. The current device tree nodes contain errors, such as saying that T1040 can set a core clock to PLL/4 when only PLL and PLL/2 are options. The current binding also ignores some restrictions on clock selection, such as p5020's requirement that if a core uses the "wrong" PLL, that PLL must be clocked lower than the "correct" PLL and be at most 80% of the rated CPU frequency. Possibly because of the lack of the ability to express such nuance in the binding, some valid options are omitted from the device trees, such as the ability on p4080 to run cores 0-3 from PLL3 and cores 4-7 from PLL1 (again, only if they are at most 80% of rated CPU frequency). This omission, combined with excessive caution in the cpufreq driver (addressed in a subsequent patch), means that currently on a 1500 MHz p4080 with typical PLL configuration, cpufreq can lower the frequency to 1200 MHz on half the CPUs and do nothing on the others. With this patchset, all CPUs can be lowered to 1200 MHz on a rev2 p4080, and on a rev3 p4080 half can be lowered to 750 MHz and the other half to 600 MHz. The current binding only deals with CPU clocks. To describe FMan in the device tree, we need to describe its clock. Some chips have additional muxes that work like the CPU muxes, but are not described in the device tree. Others require inspecting the Reset Control Word to determine which PLL is used. Rather than continue to extend this mess, replace it. Have the driver bind to the chip-specific clockgen compatible, and keep the detailed description of quirky chip variations in the driver, where it can be easily fixed, refactored, and extended. Older device trees will continue to work (including a workaround for old ls1021a device trees that are missing compatible and reg in the clockgen node, which even the old binding required). The pll/mux details in old device trees will be ignored, but "clocks" properties pointing at the old nodes will still work, and be directed at the corresponding new clock. Signed-off-by: Scott Wood <scottwood@freescale.com> Acked-by: Stephen Boyd <sboyd@codeaurora.org>
2015-09-20 12:29:54 +08:00
/* Get the multiple of PLL */
mult = cg_in(cg, reg);
clk: qoriq: Move chip-specific knowledge into driver The device tree should describe the chips (or chip-like subblocks) in the system, but it generally does not describe individual registers -- it should identify, rather than describe, a programming interface. This has not been the case with the QorIQ clockgen nodes. The knowledge of what each bit setting of CLKCnCSR means is encoded in three places (binding, pll node, and mux node), and the last also needs to know which options are valid on a particular chip. All three of these locations are considered stable ABI, making it difficult to fix mistakes (of which I have found several), much less refactor the abstraction to be able to address problems, limitations, or new chips. Under the current binding, a pll clock specifier of 2 means that the PLL is divided by 4 -- and the driver implements this, unless there happen to be four clock-output-names rather than 3, in which case it interprets it as PLL divided by 3. This does not appear in the binding documentation at all. That hack is now considered stable ABI. The current device tree nodes contain errors, such as saying that T1040 can set a core clock to PLL/4 when only PLL and PLL/2 are options. The current binding also ignores some restrictions on clock selection, such as p5020's requirement that if a core uses the "wrong" PLL, that PLL must be clocked lower than the "correct" PLL and be at most 80% of the rated CPU frequency. Possibly because of the lack of the ability to express such nuance in the binding, some valid options are omitted from the device trees, such as the ability on p4080 to run cores 0-3 from PLL3 and cores 4-7 from PLL1 (again, only if they are at most 80% of rated CPU frequency). This omission, combined with excessive caution in the cpufreq driver (addressed in a subsequent patch), means that currently on a 1500 MHz p4080 with typical PLL configuration, cpufreq can lower the frequency to 1200 MHz on half the CPUs and do nothing on the others. With this patchset, all CPUs can be lowered to 1200 MHz on a rev2 p4080, and on a rev3 p4080 half can be lowered to 750 MHz and the other half to 600 MHz. The current binding only deals with CPU clocks. To describe FMan in the device tree, we need to describe its clock. Some chips have additional muxes that work like the CPU muxes, but are not described in the device tree. Others require inspecting the Reset Control Word to determine which PLL is used. Rather than continue to extend this mess, replace it. Have the driver bind to the chip-specific clockgen compatible, and keep the detailed description of quirky chip variations in the driver, where it can be easily fixed, refactored, and extended. Older device trees will continue to work (including a workaround for old ls1021a device trees that are missing compatible and reg in the clockgen node, which even the old binding required). The pll/mux details in old device trees will be ignored, but "clocks" properties pointing at the old nodes will still work, and be directed at the corresponding new clock. Signed-off-by: Scott Wood <scottwood@freescale.com> Acked-by: Stephen Boyd <sboyd@codeaurora.org>
2015-09-20 12:29:54 +08:00
/* Check if this PLL is disabled */
if (mult & PLL_KILL) {
clk: qoriq: Move chip-specific knowledge into driver The device tree should describe the chips (or chip-like subblocks) in the system, but it generally does not describe individual registers -- it should identify, rather than describe, a programming interface. This has not been the case with the QorIQ clockgen nodes. The knowledge of what each bit setting of CLKCnCSR means is encoded in three places (binding, pll node, and mux node), and the last also needs to know which options are valid on a particular chip. All three of these locations are considered stable ABI, making it difficult to fix mistakes (of which I have found several), much less refactor the abstraction to be able to address problems, limitations, or new chips. Under the current binding, a pll clock specifier of 2 means that the PLL is divided by 4 -- and the driver implements this, unless there happen to be four clock-output-names rather than 3, in which case it interprets it as PLL divided by 3. This does not appear in the binding documentation at all. That hack is now considered stable ABI. The current device tree nodes contain errors, such as saying that T1040 can set a core clock to PLL/4 when only PLL and PLL/2 are options. The current binding also ignores some restrictions on clock selection, such as p5020's requirement that if a core uses the "wrong" PLL, that PLL must be clocked lower than the "correct" PLL and be at most 80% of the rated CPU frequency. Possibly because of the lack of the ability to express such nuance in the binding, some valid options are omitted from the device trees, such as the ability on p4080 to run cores 0-3 from PLL3 and cores 4-7 from PLL1 (again, only if they are at most 80% of rated CPU frequency). This omission, combined with excessive caution in the cpufreq driver (addressed in a subsequent patch), means that currently on a 1500 MHz p4080 with typical PLL configuration, cpufreq can lower the frequency to 1200 MHz on half the CPUs and do nothing on the others. With this patchset, all CPUs can be lowered to 1200 MHz on a rev2 p4080, and on a rev3 p4080 half can be lowered to 750 MHz and the other half to 600 MHz. The current binding only deals with CPU clocks. To describe FMan in the device tree, we need to describe its clock. Some chips have additional muxes that work like the CPU muxes, but are not described in the device tree. Others require inspecting the Reset Control Word to determine which PLL is used. Rather than continue to extend this mess, replace it. Have the driver bind to the chip-specific clockgen compatible, and keep the detailed description of quirky chip variations in the driver, where it can be easily fixed, refactored, and extended. Older device trees will continue to work (including a workaround for old ls1021a device trees that are missing compatible and reg in the clockgen node, which even the old binding required). The pll/mux details in old device trees will be ignored, but "clocks" properties pointing at the old nodes will still work, and be directed at the corresponding new clock. Signed-off-by: Scott Wood <scottwood@freescale.com> Acked-by: Stephen Boyd <sboyd@codeaurora.org>
2015-09-20 12:29:54 +08:00
pr_debug("%s(): pll %p disabled\n", __func__, reg);
return;
}
if ((cg->info.flags & CG_VER3) ||
((cg->info.flags & CG_PLL_8BIT) && idx != PLATFORM_PLL))
clk: qoriq: Move chip-specific knowledge into driver The device tree should describe the chips (or chip-like subblocks) in the system, but it generally does not describe individual registers -- it should identify, rather than describe, a programming interface. This has not been the case with the QorIQ clockgen nodes. The knowledge of what each bit setting of CLKCnCSR means is encoded in three places (binding, pll node, and mux node), and the last also needs to know which options are valid on a particular chip. All three of these locations are considered stable ABI, making it difficult to fix mistakes (of which I have found several), much less refactor the abstraction to be able to address problems, limitations, or new chips. Under the current binding, a pll clock specifier of 2 means that the PLL is divided by 4 -- and the driver implements this, unless there happen to be four clock-output-names rather than 3, in which case it interprets it as PLL divided by 3. This does not appear in the binding documentation at all. That hack is now considered stable ABI. The current device tree nodes contain errors, such as saying that T1040 can set a core clock to PLL/4 when only PLL and PLL/2 are options. The current binding also ignores some restrictions on clock selection, such as p5020's requirement that if a core uses the "wrong" PLL, that PLL must be clocked lower than the "correct" PLL and be at most 80% of the rated CPU frequency. Possibly because of the lack of the ability to express such nuance in the binding, some valid options are omitted from the device trees, such as the ability on p4080 to run cores 0-3 from PLL3 and cores 4-7 from PLL1 (again, only if they are at most 80% of rated CPU frequency). This omission, combined with excessive caution in the cpufreq driver (addressed in a subsequent patch), means that currently on a 1500 MHz p4080 with typical PLL configuration, cpufreq can lower the frequency to 1200 MHz on half the CPUs and do nothing on the others. With this patchset, all CPUs can be lowered to 1200 MHz on a rev2 p4080, and on a rev3 p4080 half can be lowered to 750 MHz and the other half to 600 MHz. The current binding only deals with CPU clocks. To describe FMan in the device tree, we need to describe its clock. Some chips have additional muxes that work like the CPU muxes, but are not described in the device tree. Others require inspecting the Reset Control Word to determine which PLL is used. Rather than continue to extend this mess, replace it. Have the driver bind to the chip-specific clockgen compatible, and keep the detailed description of quirky chip variations in the driver, where it can be easily fixed, refactored, and extended. Older device trees will continue to work (including a workaround for old ls1021a device trees that are missing compatible and reg in the clockgen node, which even the old binding required). The pll/mux details in old device trees will be ignored, but "clocks" properties pointing at the old nodes will still work, and be directed at the corresponding new clock. Signed-off-by: Scott Wood <scottwood@freescale.com> Acked-by: Stephen Boyd <sboyd@codeaurora.org>
2015-09-20 12:29:54 +08:00
mult = (mult & GENMASK(8, 1)) >> 1;
else
mult = (mult & GENMASK(6, 1)) >> 1;
for (i = 0; i < ARRAY_SIZE(pll->div); i++) {
struct clk *clk;
snprintf(pll->div[i].name, sizeof(pll->div[i].name),
"cg-pll%d-div%d", idx, i + 1);
clk = clk_register_fixed_factor(NULL,
pll->div[i].name, "cg-sysclk", 0, mult, i + 1);
if (IS_ERR(clk)) {
pr_err("%s: %s: register failed %ld\n",
__func__, pll->div[i].name, PTR_ERR(clk));
continue;
}
pll->div[i].clk = clk;
}
clk: qoriq: Move chip-specific knowledge into driver The device tree should describe the chips (or chip-like subblocks) in the system, but it generally does not describe individual registers -- it should identify, rather than describe, a programming interface. This has not been the case with the QorIQ clockgen nodes. The knowledge of what each bit setting of CLKCnCSR means is encoded in three places (binding, pll node, and mux node), and the last also needs to know which options are valid on a particular chip. All three of these locations are considered stable ABI, making it difficult to fix mistakes (of which I have found several), much less refactor the abstraction to be able to address problems, limitations, or new chips. Under the current binding, a pll clock specifier of 2 means that the PLL is divided by 4 -- and the driver implements this, unless there happen to be four clock-output-names rather than 3, in which case it interprets it as PLL divided by 3. This does not appear in the binding documentation at all. That hack is now considered stable ABI. The current device tree nodes contain errors, such as saying that T1040 can set a core clock to PLL/4 when only PLL and PLL/2 are options. The current binding also ignores some restrictions on clock selection, such as p5020's requirement that if a core uses the "wrong" PLL, that PLL must be clocked lower than the "correct" PLL and be at most 80% of the rated CPU frequency. Possibly because of the lack of the ability to express such nuance in the binding, some valid options are omitted from the device trees, such as the ability on p4080 to run cores 0-3 from PLL3 and cores 4-7 from PLL1 (again, only if they are at most 80% of rated CPU frequency). This omission, combined with excessive caution in the cpufreq driver (addressed in a subsequent patch), means that currently on a 1500 MHz p4080 with typical PLL configuration, cpufreq can lower the frequency to 1200 MHz on half the CPUs and do nothing on the others. With this patchset, all CPUs can be lowered to 1200 MHz on a rev2 p4080, and on a rev3 p4080 half can be lowered to 750 MHz and the other half to 600 MHz. The current binding only deals with CPU clocks. To describe FMan in the device tree, we need to describe its clock. Some chips have additional muxes that work like the CPU muxes, but are not described in the device tree. Others require inspecting the Reset Control Word to determine which PLL is used. Rather than continue to extend this mess, replace it. Have the driver bind to the chip-specific clockgen compatible, and keep the detailed description of quirky chip variations in the driver, where it can be easily fixed, refactored, and extended. Older device trees will continue to work (including a workaround for old ls1021a device trees that are missing compatible and reg in the clockgen node, which even the old binding required). The pll/mux details in old device trees will be ignored, but "clocks" properties pointing at the old nodes will still work, and be directed at the corresponding new clock. Signed-off-by: Scott Wood <scottwood@freescale.com> Acked-by: Stephen Boyd <sboyd@codeaurora.org>
2015-09-20 12:29:54 +08:00
}
static void __init create_plls(struct clockgen *cg)
{
int i;
for (i = 0; i < ARRAY_SIZE(cg->pll); i++)
create_one_pll(cg, i);
}
clk: qoriq: Move chip-specific knowledge into driver The device tree should describe the chips (or chip-like subblocks) in the system, but it generally does not describe individual registers -- it should identify, rather than describe, a programming interface. This has not been the case with the QorIQ clockgen nodes. The knowledge of what each bit setting of CLKCnCSR means is encoded in three places (binding, pll node, and mux node), and the last also needs to know which options are valid on a particular chip. All three of these locations are considered stable ABI, making it difficult to fix mistakes (of which I have found several), much less refactor the abstraction to be able to address problems, limitations, or new chips. Under the current binding, a pll clock specifier of 2 means that the PLL is divided by 4 -- and the driver implements this, unless there happen to be four clock-output-names rather than 3, in which case it interprets it as PLL divided by 3. This does not appear in the binding documentation at all. That hack is now considered stable ABI. The current device tree nodes contain errors, such as saying that T1040 can set a core clock to PLL/4 when only PLL and PLL/2 are options. The current binding also ignores some restrictions on clock selection, such as p5020's requirement that if a core uses the "wrong" PLL, that PLL must be clocked lower than the "correct" PLL and be at most 80% of the rated CPU frequency. Possibly because of the lack of the ability to express such nuance in the binding, some valid options are omitted from the device trees, such as the ability on p4080 to run cores 0-3 from PLL3 and cores 4-7 from PLL1 (again, only if they are at most 80% of rated CPU frequency). This omission, combined with excessive caution in the cpufreq driver (addressed in a subsequent patch), means that currently on a 1500 MHz p4080 with typical PLL configuration, cpufreq can lower the frequency to 1200 MHz on half the CPUs and do nothing on the others. With this patchset, all CPUs can be lowered to 1200 MHz on a rev2 p4080, and on a rev3 p4080 half can be lowered to 750 MHz and the other half to 600 MHz. The current binding only deals with CPU clocks. To describe FMan in the device tree, we need to describe its clock. Some chips have additional muxes that work like the CPU muxes, but are not described in the device tree. Others require inspecting the Reset Control Word to determine which PLL is used. Rather than continue to extend this mess, replace it. Have the driver bind to the chip-specific clockgen compatible, and keep the detailed description of quirky chip variations in the driver, where it can be easily fixed, refactored, and extended. Older device trees will continue to work (including a workaround for old ls1021a device trees that are missing compatible and reg in the clockgen node, which even the old binding required). The pll/mux details in old device trees will be ignored, but "clocks" properties pointing at the old nodes will still work, and be directed at the corresponding new clock. Signed-off-by: Scott Wood <scottwood@freescale.com> Acked-by: Stephen Boyd <sboyd@codeaurora.org>
2015-09-20 12:29:54 +08:00
static void __init legacy_pll_init(struct device_node *np, int idx)
{
struct clockgen_pll *pll;
struct clk_onecell_data *onecell_data;
struct clk **subclks;
int count, rc;
legacy_init_clockgen(np);
pll = &clockgen.pll[idx];
count = of_property_count_strings(np, "clock-output-names");
clk: qoriq: Move chip-specific knowledge into driver The device tree should describe the chips (or chip-like subblocks) in the system, but it generally does not describe individual registers -- it should identify, rather than describe, a programming interface. This has not been the case with the QorIQ clockgen nodes. The knowledge of what each bit setting of CLKCnCSR means is encoded in three places (binding, pll node, and mux node), and the last also needs to know which options are valid on a particular chip. All three of these locations are considered stable ABI, making it difficult to fix mistakes (of which I have found several), much less refactor the abstraction to be able to address problems, limitations, or new chips. Under the current binding, a pll clock specifier of 2 means that the PLL is divided by 4 -- and the driver implements this, unless there happen to be four clock-output-names rather than 3, in which case it interprets it as PLL divided by 3. This does not appear in the binding documentation at all. That hack is now considered stable ABI. The current device tree nodes contain errors, such as saying that T1040 can set a core clock to PLL/4 when only PLL and PLL/2 are options. The current binding also ignores some restrictions on clock selection, such as p5020's requirement that if a core uses the "wrong" PLL, that PLL must be clocked lower than the "correct" PLL and be at most 80% of the rated CPU frequency. Possibly because of the lack of the ability to express such nuance in the binding, some valid options are omitted from the device trees, such as the ability on p4080 to run cores 0-3 from PLL3 and cores 4-7 from PLL1 (again, only if they are at most 80% of rated CPU frequency). This omission, combined with excessive caution in the cpufreq driver (addressed in a subsequent patch), means that currently on a 1500 MHz p4080 with typical PLL configuration, cpufreq can lower the frequency to 1200 MHz on half the CPUs and do nothing on the others. With this patchset, all CPUs can be lowered to 1200 MHz on a rev2 p4080, and on a rev3 p4080 half can be lowered to 750 MHz and the other half to 600 MHz. The current binding only deals with CPU clocks. To describe FMan in the device tree, we need to describe its clock. Some chips have additional muxes that work like the CPU muxes, but are not described in the device tree. Others require inspecting the Reset Control Word to determine which PLL is used. Rather than continue to extend this mess, replace it. Have the driver bind to the chip-specific clockgen compatible, and keep the detailed description of quirky chip variations in the driver, where it can be easily fixed, refactored, and extended. Older device trees will continue to work (including a workaround for old ls1021a device trees that are missing compatible and reg in the clockgen node, which even the old binding required). The pll/mux details in old device trees will be ignored, but "clocks" properties pointing at the old nodes will still work, and be directed at the corresponding new clock. Signed-off-by: Scott Wood <scottwood@freescale.com> Acked-by: Stephen Boyd <sboyd@codeaurora.org>
2015-09-20 12:29:54 +08:00
BUILD_BUG_ON(ARRAY_SIZE(pll->div) < 4);
subclks = kcalloc(4, sizeof(struct clk *), GFP_KERNEL);
if (!subclks)
clk: qoriq: Move chip-specific knowledge into driver The device tree should describe the chips (or chip-like subblocks) in the system, but it generally does not describe individual registers -- it should identify, rather than describe, a programming interface. This has not been the case with the QorIQ clockgen nodes. The knowledge of what each bit setting of CLKCnCSR means is encoded in three places (binding, pll node, and mux node), and the last also needs to know which options are valid on a particular chip. All three of these locations are considered stable ABI, making it difficult to fix mistakes (of which I have found several), much less refactor the abstraction to be able to address problems, limitations, or new chips. Under the current binding, a pll clock specifier of 2 means that the PLL is divided by 4 -- and the driver implements this, unless there happen to be four clock-output-names rather than 3, in which case it interprets it as PLL divided by 3. This does not appear in the binding documentation at all. That hack is now considered stable ABI. The current device tree nodes contain errors, such as saying that T1040 can set a core clock to PLL/4 when only PLL and PLL/2 are options. The current binding also ignores some restrictions on clock selection, such as p5020's requirement that if a core uses the "wrong" PLL, that PLL must be clocked lower than the "correct" PLL and be at most 80% of the rated CPU frequency. Possibly because of the lack of the ability to express such nuance in the binding, some valid options are omitted from the device trees, such as the ability on p4080 to run cores 0-3 from PLL3 and cores 4-7 from PLL1 (again, only if they are at most 80% of rated CPU frequency). This omission, combined with excessive caution in the cpufreq driver (addressed in a subsequent patch), means that currently on a 1500 MHz p4080 with typical PLL configuration, cpufreq can lower the frequency to 1200 MHz on half the CPUs and do nothing on the others. With this patchset, all CPUs can be lowered to 1200 MHz on a rev2 p4080, and on a rev3 p4080 half can be lowered to 750 MHz and the other half to 600 MHz. The current binding only deals with CPU clocks. To describe FMan in the device tree, we need to describe its clock. Some chips have additional muxes that work like the CPU muxes, but are not described in the device tree. Others require inspecting the Reset Control Word to determine which PLL is used. Rather than continue to extend this mess, replace it. Have the driver bind to the chip-specific clockgen compatible, and keep the detailed description of quirky chip variations in the driver, where it can be easily fixed, refactored, and extended. Older device trees will continue to work (including a workaround for old ls1021a device trees that are missing compatible and reg in the clockgen node, which even the old binding required). The pll/mux details in old device trees will be ignored, but "clocks" properties pointing at the old nodes will still work, and be directed at the corresponding new clock. Signed-off-by: Scott Wood <scottwood@freescale.com> Acked-by: Stephen Boyd <sboyd@codeaurora.org>
2015-09-20 12:29:54 +08:00
return;
onecell_data = kmalloc(sizeof(*onecell_data), GFP_KERNEL);
if (!onecell_data)
goto err_clks;
clk: qoriq: Move chip-specific knowledge into driver The device tree should describe the chips (or chip-like subblocks) in the system, but it generally does not describe individual registers -- it should identify, rather than describe, a programming interface. This has not been the case with the QorIQ clockgen nodes. The knowledge of what each bit setting of CLKCnCSR means is encoded in three places (binding, pll node, and mux node), and the last also needs to know which options are valid on a particular chip. All three of these locations are considered stable ABI, making it difficult to fix mistakes (of which I have found several), much less refactor the abstraction to be able to address problems, limitations, or new chips. Under the current binding, a pll clock specifier of 2 means that the PLL is divided by 4 -- and the driver implements this, unless there happen to be four clock-output-names rather than 3, in which case it interprets it as PLL divided by 3. This does not appear in the binding documentation at all. That hack is now considered stable ABI. The current device tree nodes contain errors, such as saying that T1040 can set a core clock to PLL/4 when only PLL and PLL/2 are options. The current binding also ignores some restrictions on clock selection, such as p5020's requirement that if a core uses the "wrong" PLL, that PLL must be clocked lower than the "correct" PLL and be at most 80% of the rated CPU frequency. Possibly because of the lack of the ability to express such nuance in the binding, some valid options are omitted from the device trees, such as the ability on p4080 to run cores 0-3 from PLL3 and cores 4-7 from PLL1 (again, only if they are at most 80% of rated CPU frequency). This omission, combined with excessive caution in the cpufreq driver (addressed in a subsequent patch), means that currently on a 1500 MHz p4080 with typical PLL configuration, cpufreq can lower the frequency to 1200 MHz on half the CPUs and do nothing on the others. With this patchset, all CPUs can be lowered to 1200 MHz on a rev2 p4080, and on a rev3 p4080 half can be lowered to 750 MHz and the other half to 600 MHz. The current binding only deals with CPU clocks. To describe FMan in the device tree, we need to describe its clock. Some chips have additional muxes that work like the CPU muxes, but are not described in the device tree. Others require inspecting the Reset Control Word to determine which PLL is used. Rather than continue to extend this mess, replace it. Have the driver bind to the chip-specific clockgen compatible, and keep the detailed description of quirky chip variations in the driver, where it can be easily fixed, refactored, and extended. Older device trees will continue to work (including a workaround for old ls1021a device trees that are missing compatible and reg in the clockgen node, which even the old binding required). The pll/mux details in old device trees will be ignored, but "clocks" properties pointing at the old nodes will still work, and be directed at the corresponding new clock. Signed-off-by: Scott Wood <scottwood@freescale.com> Acked-by: Stephen Boyd <sboyd@codeaurora.org>
2015-09-20 12:29:54 +08:00
if (count <= 3) {
subclks[0] = pll->div[0].clk;
subclks[1] = pll->div[1].clk;
subclks[2] = pll->div[3].clk;
} else {
subclks[0] = pll->div[0].clk;
subclks[1] = pll->div[1].clk;
subclks[2] = pll->div[2].clk;
subclks[3] = pll->div[3].clk;
}
onecell_data->clks = subclks;
onecell_data->clk_num = count;
rc = of_clk_add_provider(np, of_clk_src_onecell_get, onecell_data);
if (rc) {
clk: qoriq: Move chip-specific knowledge into driver The device tree should describe the chips (or chip-like subblocks) in the system, but it generally does not describe individual registers -- it should identify, rather than describe, a programming interface. This has not been the case with the QorIQ clockgen nodes. The knowledge of what each bit setting of CLKCnCSR means is encoded in three places (binding, pll node, and mux node), and the last also needs to know which options are valid on a particular chip. All three of these locations are considered stable ABI, making it difficult to fix mistakes (of which I have found several), much less refactor the abstraction to be able to address problems, limitations, or new chips. Under the current binding, a pll clock specifier of 2 means that the PLL is divided by 4 -- and the driver implements this, unless there happen to be four clock-output-names rather than 3, in which case it interprets it as PLL divided by 3. This does not appear in the binding documentation at all. That hack is now considered stable ABI. The current device tree nodes contain errors, such as saying that T1040 can set a core clock to PLL/4 when only PLL and PLL/2 are options. The current binding also ignores some restrictions on clock selection, such as p5020's requirement that if a core uses the "wrong" PLL, that PLL must be clocked lower than the "correct" PLL and be at most 80% of the rated CPU frequency. Possibly because of the lack of the ability to express such nuance in the binding, some valid options are omitted from the device trees, such as the ability on p4080 to run cores 0-3 from PLL3 and cores 4-7 from PLL1 (again, only if they are at most 80% of rated CPU frequency). This omission, combined with excessive caution in the cpufreq driver (addressed in a subsequent patch), means that currently on a 1500 MHz p4080 with typical PLL configuration, cpufreq can lower the frequency to 1200 MHz on half the CPUs and do nothing on the others. With this patchset, all CPUs can be lowered to 1200 MHz on a rev2 p4080, and on a rev3 p4080 half can be lowered to 750 MHz and the other half to 600 MHz. The current binding only deals with CPU clocks. To describe FMan in the device tree, we need to describe its clock. Some chips have additional muxes that work like the CPU muxes, but are not described in the device tree. Others require inspecting the Reset Control Word to determine which PLL is used. Rather than continue to extend this mess, replace it. Have the driver bind to the chip-specific clockgen compatible, and keep the detailed description of quirky chip variations in the driver, where it can be easily fixed, refactored, and extended. Older device trees will continue to work (including a workaround for old ls1021a device trees that are missing compatible and reg in the clockgen node, which even the old binding required). The pll/mux details in old device trees will be ignored, but "clocks" properties pointing at the old nodes will still work, and be directed at the corresponding new clock. Signed-off-by: Scott Wood <scottwood@freescale.com> Acked-by: Stephen Boyd <sboyd@codeaurora.org>
2015-09-20 12:29:54 +08:00
pr_err("%s: Couldn't register clk provider for node %s: %d\n",
__func__, np->name, rc);
goto err_cell;
}
return;
err_cell:
kfree(onecell_data);
err_clks:
kfree(subclks);
}
clk: qoriq: Move chip-specific knowledge into driver The device tree should describe the chips (or chip-like subblocks) in the system, but it generally does not describe individual registers -- it should identify, rather than describe, a programming interface. This has not been the case with the QorIQ clockgen nodes. The knowledge of what each bit setting of CLKCnCSR means is encoded in three places (binding, pll node, and mux node), and the last also needs to know which options are valid on a particular chip. All three of these locations are considered stable ABI, making it difficult to fix mistakes (of which I have found several), much less refactor the abstraction to be able to address problems, limitations, or new chips. Under the current binding, a pll clock specifier of 2 means that the PLL is divided by 4 -- and the driver implements this, unless there happen to be four clock-output-names rather than 3, in which case it interprets it as PLL divided by 3. This does not appear in the binding documentation at all. That hack is now considered stable ABI. The current device tree nodes contain errors, such as saying that T1040 can set a core clock to PLL/4 when only PLL and PLL/2 are options. The current binding also ignores some restrictions on clock selection, such as p5020's requirement that if a core uses the "wrong" PLL, that PLL must be clocked lower than the "correct" PLL and be at most 80% of the rated CPU frequency. Possibly because of the lack of the ability to express such nuance in the binding, some valid options are omitted from the device trees, such as the ability on p4080 to run cores 0-3 from PLL3 and cores 4-7 from PLL1 (again, only if they are at most 80% of rated CPU frequency). This omission, combined with excessive caution in the cpufreq driver (addressed in a subsequent patch), means that currently on a 1500 MHz p4080 with typical PLL configuration, cpufreq can lower the frequency to 1200 MHz on half the CPUs and do nothing on the others. With this patchset, all CPUs can be lowered to 1200 MHz on a rev2 p4080, and on a rev3 p4080 half can be lowered to 750 MHz and the other half to 600 MHz. The current binding only deals with CPU clocks. To describe FMan in the device tree, we need to describe its clock. Some chips have additional muxes that work like the CPU muxes, but are not described in the device tree. Others require inspecting the Reset Control Word to determine which PLL is used. Rather than continue to extend this mess, replace it. Have the driver bind to the chip-specific clockgen compatible, and keep the detailed description of quirky chip variations in the driver, where it can be easily fixed, refactored, and extended. Older device trees will continue to work (including a workaround for old ls1021a device trees that are missing compatible and reg in the clockgen node, which even the old binding required). The pll/mux details in old device trees will be ignored, but "clocks" properties pointing at the old nodes will still work, and be directed at the corresponding new clock. Signed-off-by: Scott Wood <scottwood@freescale.com> Acked-by: Stephen Boyd <sboyd@codeaurora.org>
2015-09-20 12:29:54 +08:00
/* Legacy node */
static void __init pltfrm_pll_init(struct device_node *np)
{
clk: qoriq: Move chip-specific knowledge into driver The device tree should describe the chips (or chip-like subblocks) in the system, but it generally does not describe individual registers -- it should identify, rather than describe, a programming interface. This has not been the case with the QorIQ clockgen nodes. The knowledge of what each bit setting of CLKCnCSR means is encoded in three places (binding, pll node, and mux node), and the last also needs to know which options are valid on a particular chip. All three of these locations are considered stable ABI, making it difficult to fix mistakes (of which I have found several), much less refactor the abstraction to be able to address problems, limitations, or new chips. Under the current binding, a pll clock specifier of 2 means that the PLL is divided by 4 -- and the driver implements this, unless there happen to be four clock-output-names rather than 3, in which case it interprets it as PLL divided by 3. This does not appear in the binding documentation at all. That hack is now considered stable ABI. The current device tree nodes contain errors, such as saying that T1040 can set a core clock to PLL/4 when only PLL and PLL/2 are options. The current binding also ignores some restrictions on clock selection, such as p5020's requirement that if a core uses the "wrong" PLL, that PLL must be clocked lower than the "correct" PLL and be at most 80% of the rated CPU frequency. Possibly because of the lack of the ability to express such nuance in the binding, some valid options are omitted from the device trees, such as the ability on p4080 to run cores 0-3 from PLL3 and cores 4-7 from PLL1 (again, only if they are at most 80% of rated CPU frequency). This omission, combined with excessive caution in the cpufreq driver (addressed in a subsequent patch), means that currently on a 1500 MHz p4080 with typical PLL configuration, cpufreq can lower the frequency to 1200 MHz on half the CPUs and do nothing on the others. With this patchset, all CPUs can be lowered to 1200 MHz on a rev2 p4080, and on a rev3 p4080 half can be lowered to 750 MHz and the other half to 600 MHz. The current binding only deals with CPU clocks. To describe FMan in the device tree, we need to describe its clock. Some chips have additional muxes that work like the CPU muxes, but are not described in the device tree. Others require inspecting the Reset Control Word to determine which PLL is used. Rather than continue to extend this mess, replace it. Have the driver bind to the chip-specific clockgen compatible, and keep the detailed description of quirky chip variations in the driver, where it can be easily fixed, refactored, and extended. Older device trees will continue to work (including a workaround for old ls1021a device trees that are missing compatible and reg in the clockgen node, which even the old binding required). The pll/mux details in old device trees will be ignored, but "clocks" properties pointing at the old nodes will still work, and be directed at the corresponding new clock. Signed-off-by: Scott Wood <scottwood@freescale.com> Acked-by: Stephen Boyd <sboyd@codeaurora.org>
2015-09-20 12:29:54 +08:00
legacy_pll_init(np, PLATFORM_PLL);
}
clk: qoriq: Move chip-specific knowledge into driver The device tree should describe the chips (or chip-like subblocks) in the system, but it generally does not describe individual registers -- it should identify, rather than describe, a programming interface. This has not been the case with the QorIQ clockgen nodes. The knowledge of what each bit setting of CLKCnCSR means is encoded in three places (binding, pll node, and mux node), and the last also needs to know which options are valid on a particular chip. All three of these locations are considered stable ABI, making it difficult to fix mistakes (of which I have found several), much less refactor the abstraction to be able to address problems, limitations, or new chips. Under the current binding, a pll clock specifier of 2 means that the PLL is divided by 4 -- and the driver implements this, unless there happen to be four clock-output-names rather than 3, in which case it interprets it as PLL divided by 3. This does not appear in the binding documentation at all. That hack is now considered stable ABI. The current device tree nodes contain errors, such as saying that T1040 can set a core clock to PLL/4 when only PLL and PLL/2 are options. The current binding also ignores some restrictions on clock selection, such as p5020's requirement that if a core uses the "wrong" PLL, that PLL must be clocked lower than the "correct" PLL and be at most 80% of the rated CPU frequency. Possibly because of the lack of the ability to express such nuance in the binding, some valid options are omitted from the device trees, such as the ability on p4080 to run cores 0-3 from PLL3 and cores 4-7 from PLL1 (again, only if they are at most 80% of rated CPU frequency). This omission, combined with excessive caution in the cpufreq driver (addressed in a subsequent patch), means that currently on a 1500 MHz p4080 with typical PLL configuration, cpufreq can lower the frequency to 1200 MHz on half the CPUs and do nothing on the others. With this patchset, all CPUs can be lowered to 1200 MHz on a rev2 p4080, and on a rev3 p4080 half can be lowered to 750 MHz and the other half to 600 MHz. The current binding only deals with CPU clocks. To describe FMan in the device tree, we need to describe its clock. Some chips have additional muxes that work like the CPU muxes, but are not described in the device tree. Others require inspecting the Reset Control Word to determine which PLL is used. Rather than continue to extend this mess, replace it. Have the driver bind to the chip-specific clockgen compatible, and keep the detailed description of quirky chip variations in the driver, where it can be easily fixed, refactored, and extended. Older device trees will continue to work (including a workaround for old ls1021a device trees that are missing compatible and reg in the clockgen node, which even the old binding required). The pll/mux details in old device trees will be ignored, but "clocks" properties pointing at the old nodes will still work, and be directed at the corresponding new clock. Signed-off-by: Scott Wood <scottwood@freescale.com> Acked-by: Stephen Boyd <sboyd@codeaurora.org>
2015-09-20 12:29:54 +08:00
/* Legacy node */
static void __init core_pll_init(struct device_node *np)
{
struct resource res;
int idx;
if (of_address_to_resource(np, 0, &res))
return;
clk: qoriq: Move chip-specific knowledge into driver The device tree should describe the chips (or chip-like subblocks) in the system, but it generally does not describe individual registers -- it should identify, rather than describe, a programming interface. This has not been the case with the QorIQ clockgen nodes. The knowledge of what each bit setting of CLKCnCSR means is encoded in three places (binding, pll node, and mux node), and the last also needs to know which options are valid on a particular chip. All three of these locations are considered stable ABI, making it difficult to fix mistakes (of which I have found several), much less refactor the abstraction to be able to address problems, limitations, or new chips. Under the current binding, a pll clock specifier of 2 means that the PLL is divided by 4 -- and the driver implements this, unless there happen to be four clock-output-names rather than 3, in which case it interprets it as PLL divided by 3. This does not appear in the binding documentation at all. That hack is now considered stable ABI. The current device tree nodes contain errors, such as saying that T1040 can set a core clock to PLL/4 when only PLL and PLL/2 are options. The current binding also ignores some restrictions on clock selection, such as p5020's requirement that if a core uses the "wrong" PLL, that PLL must be clocked lower than the "correct" PLL and be at most 80% of the rated CPU frequency. Possibly because of the lack of the ability to express such nuance in the binding, some valid options are omitted from the device trees, such as the ability on p4080 to run cores 0-3 from PLL3 and cores 4-7 from PLL1 (again, only if they are at most 80% of rated CPU frequency). This omission, combined with excessive caution in the cpufreq driver (addressed in a subsequent patch), means that currently on a 1500 MHz p4080 with typical PLL configuration, cpufreq can lower the frequency to 1200 MHz on half the CPUs and do nothing on the others. With this patchset, all CPUs can be lowered to 1200 MHz on a rev2 p4080, and on a rev3 p4080 half can be lowered to 750 MHz and the other half to 600 MHz. The current binding only deals with CPU clocks. To describe FMan in the device tree, we need to describe its clock. Some chips have additional muxes that work like the CPU muxes, but are not described in the device tree. Others require inspecting the Reset Control Word to determine which PLL is used. Rather than continue to extend this mess, replace it. Have the driver bind to the chip-specific clockgen compatible, and keep the detailed description of quirky chip variations in the driver, where it can be easily fixed, refactored, and extended. Older device trees will continue to work (including a workaround for old ls1021a device trees that are missing compatible and reg in the clockgen node, which even the old binding required). The pll/mux details in old device trees will be ignored, but "clocks" properties pointing at the old nodes will still work, and be directed at the corresponding new clock. Signed-off-by: Scott Wood <scottwood@freescale.com> Acked-by: Stephen Boyd <sboyd@codeaurora.org>
2015-09-20 12:29:54 +08:00
if ((res.start & 0xfff) == 0xc00) {
/*
* ls1021a devtree labels the platform PLL
* with the core PLL compatible
*/
pltfrm_pll_init(np);
} else {
idx = (res.start & 0xf0) >> 5;
legacy_pll_init(np, CGA_PLL1 + idx);
}
clk: qoriq: Move chip-specific knowledge into driver The device tree should describe the chips (or chip-like subblocks) in the system, but it generally does not describe individual registers -- it should identify, rather than describe, a programming interface. This has not been the case with the QorIQ clockgen nodes. The knowledge of what each bit setting of CLKCnCSR means is encoded in three places (binding, pll node, and mux node), and the last also needs to know which options are valid on a particular chip. All three of these locations are considered stable ABI, making it difficult to fix mistakes (of which I have found several), much less refactor the abstraction to be able to address problems, limitations, or new chips. Under the current binding, a pll clock specifier of 2 means that the PLL is divided by 4 -- and the driver implements this, unless there happen to be four clock-output-names rather than 3, in which case it interprets it as PLL divided by 3. This does not appear in the binding documentation at all. That hack is now considered stable ABI. The current device tree nodes contain errors, such as saying that T1040 can set a core clock to PLL/4 when only PLL and PLL/2 are options. The current binding also ignores some restrictions on clock selection, such as p5020's requirement that if a core uses the "wrong" PLL, that PLL must be clocked lower than the "correct" PLL and be at most 80% of the rated CPU frequency. Possibly because of the lack of the ability to express such nuance in the binding, some valid options are omitted from the device trees, such as the ability on p4080 to run cores 0-3 from PLL3 and cores 4-7 from PLL1 (again, only if they are at most 80% of rated CPU frequency). This omission, combined with excessive caution in the cpufreq driver (addressed in a subsequent patch), means that currently on a 1500 MHz p4080 with typical PLL configuration, cpufreq can lower the frequency to 1200 MHz on half the CPUs and do nothing on the others. With this patchset, all CPUs can be lowered to 1200 MHz on a rev2 p4080, and on a rev3 p4080 half can be lowered to 750 MHz and the other half to 600 MHz. The current binding only deals with CPU clocks. To describe FMan in the device tree, we need to describe its clock. Some chips have additional muxes that work like the CPU muxes, but are not described in the device tree. Others require inspecting the Reset Control Word to determine which PLL is used. Rather than continue to extend this mess, replace it. Have the driver bind to the chip-specific clockgen compatible, and keep the detailed description of quirky chip variations in the driver, where it can be easily fixed, refactored, and extended. Older device trees will continue to work (including a workaround for old ls1021a device trees that are missing compatible and reg in the clockgen node, which even the old binding required). The pll/mux details in old device trees will be ignored, but "clocks" properties pointing at the old nodes will still work, and be directed at the corresponding new clock. Signed-off-by: Scott Wood <scottwood@freescale.com> Acked-by: Stephen Boyd <sboyd@codeaurora.org>
2015-09-20 12:29:54 +08:00
}
clk: qoriq: Move chip-specific knowledge into driver The device tree should describe the chips (or chip-like subblocks) in the system, but it generally does not describe individual registers -- it should identify, rather than describe, a programming interface. This has not been the case with the QorIQ clockgen nodes. The knowledge of what each bit setting of CLKCnCSR means is encoded in three places (binding, pll node, and mux node), and the last also needs to know which options are valid on a particular chip. All three of these locations are considered stable ABI, making it difficult to fix mistakes (of which I have found several), much less refactor the abstraction to be able to address problems, limitations, or new chips. Under the current binding, a pll clock specifier of 2 means that the PLL is divided by 4 -- and the driver implements this, unless there happen to be four clock-output-names rather than 3, in which case it interprets it as PLL divided by 3. This does not appear in the binding documentation at all. That hack is now considered stable ABI. The current device tree nodes contain errors, such as saying that T1040 can set a core clock to PLL/4 when only PLL and PLL/2 are options. The current binding also ignores some restrictions on clock selection, such as p5020's requirement that if a core uses the "wrong" PLL, that PLL must be clocked lower than the "correct" PLL and be at most 80% of the rated CPU frequency. Possibly because of the lack of the ability to express such nuance in the binding, some valid options are omitted from the device trees, such as the ability on p4080 to run cores 0-3 from PLL3 and cores 4-7 from PLL1 (again, only if they are at most 80% of rated CPU frequency). This omission, combined with excessive caution in the cpufreq driver (addressed in a subsequent patch), means that currently on a 1500 MHz p4080 with typical PLL configuration, cpufreq can lower the frequency to 1200 MHz on half the CPUs and do nothing on the others. With this patchset, all CPUs can be lowered to 1200 MHz on a rev2 p4080, and on a rev3 p4080 half can be lowered to 750 MHz and the other half to 600 MHz. The current binding only deals with CPU clocks. To describe FMan in the device tree, we need to describe its clock. Some chips have additional muxes that work like the CPU muxes, but are not described in the device tree. Others require inspecting the Reset Control Word to determine which PLL is used. Rather than continue to extend this mess, replace it. Have the driver bind to the chip-specific clockgen compatible, and keep the detailed description of quirky chip variations in the driver, where it can be easily fixed, refactored, and extended. Older device trees will continue to work (including a workaround for old ls1021a device trees that are missing compatible and reg in the clockgen node, which even the old binding required). The pll/mux details in old device trees will be ignored, but "clocks" properties pointing at the old nodes will still work, and be directed at the corresponding new clock. Signed-off-by: Scott Wood <scottwood@freescale.com> Acked-by: Stephen Boyd <sboyd@codeaurora.org>
2015-09-20 12:29:54 +08:00
static struct clk *clockgen_clk_get(struct of_phandle_args *clkspec, void *data)
{
struct clockgen *cg = data;
struct clk *clk;
struct clockgen_pll *pll;
u32 type, idx;
if (clkspec->args_count < 2) {
pr_err("%s: insufficient phandle args\n", __func__);
return ERR_PTR(-EINVAL);
}
clk: qoriq: Move chip-specific knowledge into driver The device tree should describe the chips (or chip-like subblocks) in the system, but it generally does not describe individual registers -- it should identify, rather than describe, a programming interface. This has not been the case with the QorIQ clockgen nodes. The knowledge of what each bit setting of CLKCnCSR means is encoded in three places (binding, pll node, and mux node), and the last also needs to know which options are valid on a particular chip. All three of these locations are considered stable ABI, making it difficult to fix mistakes (of which I have found several), much less refactor the abstraction to be able to address problems, limitations, or new chips. Under the current binding, a pll clock specifier of 2 means that the PLL is divided by 4 -- and the driver implements this, unless there happen to be four clock-output-names rather than 3, in which case it interprets it as PLL divided by 3. This does not appear in the binding documentation at all. That hack is now considered stable ABI. The current device tree nodes contain errors, such as saying that T1040 can set a core clock to PLL/4 when only PLL and PLL/2 are options. The current binding also ignores some restrictions on clock selection, such as p5020's requirement that if a core uses the "wrong" PLL, that PLL must be clocked lower than the "correct" PLL and be at most 80% of the rated CPU frequency. Possibly because of the lack of the ability to express such nuance in the binding, some valid options are omitted from the device trees, such as the ability on p4080 to run cores 0-3 from PLL3 and cores 4-7 from PLL1 (again, only if they are at most 80% of rated CPU frequency). This omission, combined with excessive caution in the cpufreq driver (addressed in a subsequent patch), means that currently on a 1500 MHz p4080 with typical PLL configuration, cpufreq can lower the frequency to 1200 MHz on half the CPUs and do nothing on the others. With this patchset, all CPUs can be lowered to 1200 MHz on a rev2 p4080, and on a rev3 p4080 half can be lowered to 750 MHz and the other half to 600 MHz. The current binding only deals with CPU clocks. To describe FMan in the device tree, we need to describe its clock. Some chips have additional muxes that work like the CPU muxes, but are not described in the device tree. Others require inspecting the Reset Control Word to determine which PLL is used. Rather than continue to extend this mess, replace it. Have the driver bind to the chip-specific clockgen compatible, and keep the detailed description of quirky chip variations in the driver, where it can be easily fixed, refactored, and extended. Older device trees will continue to work (including a workaround for old ls1021a device trees that are missing compatible and reg in the clockgen node, which even the old binding required). The pll/mux details in old device trees will be ignored, but "clocks" properties pointing at the old nodes will still work, and be directed at the corresponding new clock. Signed-off-by: Scott Wood <scottwood@freescale.com> Acked-by: Stephen Boyd <sboyd@codeaurora.org>
2015-09-20 12:29:54 +08:00
type = clkspec->args[0];
idx = clkspec->args[1];
clk: qoriq: Move chip-specific knowledge into driver The device tree should describe the chips (or chip-like subblocks) in the system, but it generally does not describe individual registers -- it should identify, rather than describe, a programming interface. This has not been the case with the QorIQ clockgen nodes. The knowledge of what each bit setting of CLKCnCSR means is encoded in three places (binding, pll node, and mux node), and the last also needs to know which options are valid on a particular chip. All three of these locations are considered stable ABI, making it difficult to fix mistakes (of which I have found several), much less refactor the abstraction to be able to address problems, limitations, or new chips. Under the current binding, a pll clock specifier of 2 means that the PLL is divided by 4 -- and the driver implements this, unless there happen to be four clock-output-names rather than 3, in which case it interprets it as PLL divided by 3. This does not appear in the binding documentation at all. That hack is now considered stable ABI. The current device tree nodes contain errors, such as saying that T1040 can set a core clock to PLL/4 when only PLL and PLL/2 are options. The current binding also ignores some restrictions on clock selection, such as p5020's requirement that if a core uses the "wrong" PLL, that PLL must be clocked lower than the "correct" PLL and be at most 80% of the rated CPU frequency. Possibly because of the lack of the ability to express such nuance in the binding, some valid options are omitted from the device trees, such as the ability on p4080 to run cores 0-3 from PLL3 and cores 4-7 from PLL1 (again, only if they are at most 80% of rated CPU frequency). This omission, combined with excessive caution in the cpufreq driver (addressed in a subsequent patch), means that currently on a 1500 MHz p4080 with typical PLL configuration, cpufreq can lower the frequency to 1200 MHz on half the CPUs and do nothing on the others. With this patchset, all CPUs can be lowered to 1200 MHz on a rev2 p4080, and on a rev3 p4080 half can be lowered to 750 MHz and the other half to 600 MHz. The current binding only deals with CPU clocks. To describe FMan in the device tree, we need to describe its clock. Some chips have additional muxes that work like the CPU muxes, but are not described in the device tree. Others require inspecting the Reset Control Word to determine which PLL is used. Rather than continue to extend this mess, replace it. Have the driver bind to the chip-specific clockgen compatible, and keep the detailed description of quirky chip variations in the driver, where it can be easily fixed, refactored, and extended. Older device trees will continue to work (including a workaround for old ls1021a device trees that are missing compatible and reg in the clockgen node, which even the old binding required). The pll/mux details in old device trees will be ignored, but "clocks" properties pointing at the old nodes will still work, and be directed at the corresponding new clock. Signed-off-by: Scott Wood <scottwood@freescale.com> Acked-by: Stephen Boyd <sboyd@codeaurora.org>
2015-09-20 12:29:54 +08:00
switch (type) {
case 0:
if (idx != 0)
goto bad_args;
clk = cg->sysclk;
break;
case 1:
if (idx >= ARRAY_SIZE(cg->cmux))
goto bad_args;
clk = cg->cmux[idx];
break;
case 2:
if (idx >= ARRAY_SIZE(cg->hwaccel))
goto bad_args;
clk = cg->hwaccel[idx];
break;
case 3:
if (idx >= ARRAY_SIZE(cg->fman))
goto bad_args;
clk = cg->fman[idx];
break;
case 4:
pll = &cg->pll[PLATFORM_PLL];
if (idx >= ARRAY_SIZE(pll->div))
goto bad_args;
clk = pll->div[idx].clk;
break;
default:
goto bad_args;
}
if (!clk)
return ERR_PTR(-ENOENT);
return clk;
bad_args:
pr_err("%s: Bad phandle args %u %u\n", __func__, type, idx);
return ERR_PTR(-EINVAL);
}
clk: qoriq: Move chip-specific knowledge into driver The device tree should describe the chips (or chip-like subblocks) in the system, but it generally does not describe individual registers -- it should identify, rather than describe, a programming interface. This has not been the case with the QorIQ clockgen nodes. The knowledge of what each bit setting of CLKCnCSR means is encoded in three places (binding, pll node, and mux node), and the last also needs to know which options are valid on a particular chip. All three of these locations are considered stable ABI, making it difficult to fix mistakes (of which I have found several), much less refactor the abstraction to be able to address problems, limitations, or new chips. Under the current binding, a pll clock specifier of 2 means that the PLL is divided by 4 -- and the driver implements this, unless there happen to be four clock-output-names rather than 3, in which case it interprets it as PLL divided by 3. This does not appear in the binding documentation at all. That hack is now considered stable ABI. The current device tree nodes contain errors, such as saying that T1040 can set a core clock to PLL/4 when only PLL and PLL/2 are options. The current binding also ignores some restrictions on clock selection, such as p5020's requirement that if a core uses the "wrong" PLL, that PLL must be clocked lower than the "correct" PLL and be at most 80% of the rated CPU frequency. Possibly because of the lack of the ability to express such nuance in the binding, some valid options are omitted from the device trees, such as the ability on p4080 to run cores 0-3 from PLL3 and cores 4-7 from PLL1 (again, only if they are at most 80% of rated CPU frequency). This omission, combined with excessive caution in the cpufreq driver (addressed in a subsequent patch), means that currently on a 1500 MHz p4080 with typical PLL configuration, cpufreq can lower the frequency to 1200 MHz on half the CPUs and do nothing on the others. With this patchset, all CPUs can be lowered to 1200 MHz on a rev2 p4080, and on a rev3 p4080 half can be lowered to 750 MHz and the other half to 600 MHz. The current binding only deals with CPU clocks. To describe FMan in the device tree, we need to describe its clock. Some chips have additional muxes that work like the CPU muxes, but are not described in the device tree. Others require inspecting the Reset Control Word to determine which PLL is used. Rather than continue to extend this mess, replace it. Have the driver bind to the chip-specific clockgen compatible, and keep the detailed description of quirky chip variations in the driver, where it can be easily fixed, refactored, and extended. Older device trees will continue to work (including a workaround for old ls1021a device trees that are missing compatible and reg in the clockgen node, which even the old binding required). The pll/mux details in old device trees will be ignored, but "clocks" properties pointing at the old nodes will still work, and be directed at the corresponding new clock. Signed-off-by: Scott Wood <scottwood@freescale.com> Acked-by: Stephen Boyd <sboyd@codeaurora.org>
2015-09-20 12:29:54 +08:00
#ifdef CONFIG_PPC
#include <asm/mpc85xx.h>
static const u32 a4510_svrs[] __initconst = {
(SVR_P2040 << 8) | 0x10, /* P2040 1.0 */
(SVR_P2040 << 8) | 0x11, /* P2040 1.1 */
(SVR_P2041 << 8) | 0x10, /* P2041 1.0 */
(SVR_P2041 << 8) | 0x11, /* P2041 1.1 */
(SVR_P3041 << 8) | 0x10, /* P3041 1.0 */
(SVR_P3041 << 8) | 0x11, /* P3041 1.1 */
(SVR_P4040 << 8) | 0x20, /* P4040 2.0 */
(SVR_P4080 << 8) | 0x20, /* P4080 2.0 */
(SVR_P5010 << 8) | 0x10, /* P5010 1.0 */
(SVR_P5010 << 8) | 0x20, /* P5010 2.0 */
(SVR_P5020 << 8) | 0x10, /* P5020 1.0 */
(SVR_P5021 << 8) | 0x10, /* P5021 1.0 */
(SVR_P5040 << 8) | 0x10, /* P5040 1.0 */
};
#define SVR_SECURITY 0x80000 /* The Security (E) bit */
static bool __init has_erratum_a4510(void)
{
clk: qoriq: Move chip-specific knowledge into driver The device tree should describe the chips (or chip-like subblocks) in the system, but it generally does not describe individual registers -- it should identify, rather than describe, a programming interface. This has not been the case with the QorIQ clockgen nodes. The knowledge of what each bit setting of CLKCnCSR means is encoded in three places (binding, pll node, and mux node), and the last also needs to know which options are valid on a particular chip. All three of these locations are considered stable ABI, making it difficult to fix mistakes (of which I have found several), much less refactor the abstraction to be able to address problems, limitations, or new chips. Under the current binding, a pll clock specifier of 2 means that the PLL is divided by 4 -- and the driver implements this, unless there happen to be four clock-output-names rather than 3, in which case it interprets it as PLL divided by 3. This does not appear in the binding documentation at all. That hack is now considered stable ABI. The current device tree nodes contain errors, such as saying that T1040 can set a core clock to PLL/4 when only PLL and PLL/2 are options. The current binding also ignores some restrictions on clock selection, such as p5020's requirement that if a core uses the "wrong" PLL, that PLL must be clocked lower than the "correct" PLL and be at most 80% of the rated CPU frequency. Possibly because of the lack of the ability to express such nuance in the binding, some valid options are omitted from the device trees, such as the ability on p4080 to run cores 0-3 from PLL3 and cores 4-7 from PLL1 (again, only if they are at most 80% of rated CPU frequency). This omission, combined with excessive caution in the cpufreq driver (addressed in a subsequent patch), means that currently on a 1500 MHz p4080 with typical PLL configuration, cpufreq can lower the frequency to 1200 MHz on half the CPUs and do nothing on the others. With this patchset, all CPUs can be lowered to 1200 MHz on a rev2 p4080, and on a rev3 p4080 half can be lowered to 750 MHz and the other half to 600 MHz. The current binding only deals with CPU clocks. To describe FMan in the device tree, we need to describe its clock. Some chips have additional muxes that work like the CPU muxes, but are not described in the device tree. Others require inspecting the Reset Control Word to determine which PLL is used. Rather than continue to extend this mess, replace it. Have the driver bind to the chip-specific clockgen compatible, and keep the detailed description of quirky chip variations in the driver, where it can be easily fixed, refactored, and extended. Older device trees will continue to work (including a workaround for old ls1021a device trees that are missing compatible and reg in the clockgen node, which even the old binding required). The pll/mux details in old device trees will be ignored, but "clocks" properties pointing at the old nodes will still work, and be directed at the corresponding new clock. Signed-off-by: Scott Wood <scottwood@freescale.com> Acked-by: Stephen Boyd <sboyd@codeaurora.org>
2015-09-20 12:29:54 +08:00
u32 svr = mfspr(SPRN_SVR);
int i;
svr &= ~SVR_SECURITY;
for (i = 0; i < ARRAY_SIZE(a4510_svrs); i++) {
if (svr == a4510_svrs[i])
return true;
}
clk: qoriq: Move chip-specific knowledge into driver The device tree should describe the chips (or chip-like subblocks) in the system, but it generally does not describe individual registers -- it should identify, rather than describe, a programming interface. This has not been the case with the QorIQ clockgen nodes. The knowledge of what each bit setting of CLKCnCSR means is encoded in three places (binding, pll node, and mux node), and the last also needs to know which options are valid on a particular chip. All three of these locations are considered stable ABI, making it difficult to fix mistakes (of which I have found several), much less refactor the abstraction to be able to address problems, limitations, or new chips. Under the current binding, a pll clock specifier of 2 means that the PLL is divided by 4 -- and the driver implements this, unless there happen to be four clock-output-names rather than 3, in which case it interprets it as PLL divided by 3. This does not appear in the binding documentation at all. That hack is now considered stable ABI. The current device tree nodes contain errors, such as saying that T1040 can set a core clock to PLL/4 when only PLL and PLL/2 are options. The current binding also ignores some restrictions on clock selection, such as p5020's requirement that if a core uses the "wrong" PLL, that PLL must be clocked lower than the "correct" PLL and be at most 80% of the rated CPU frequency. Possibly because of the lack of the ability to express such nuance in the binding, some valid options are omitted from the device trees, such as the ability on p4080 to run cores 0-3 from PLL3 and cores 4-7 from PLL1 (again, only if they are at most 80% of rated CPU frequency). This omission, combined with excessive caution in the cpufreq driver (addressed in a subsequent patch), means that currently on a 1500 MHz p4080 with typical PLL configuration, cpufreq can lower the frequency to 1200 MHz on half the CPUs and do nothing on the others. With this patchset, all CPUs can be lowered to 1200 MHz on a rev2 p4080, and on a rev3 p4080 half can be lowered to 750 MHz and the other half to 600 MHz. The current binding only deals with CPU clocks. To describe FMan in the device tree, we need to describe its clock. Some chips have additional muxes that work like the CPU muxes, but are not described in the device tree. Others require inspecting the Reset Control Word to determine which PLL is used. Rather than continue to extend this mess, replace it. Have the driver bind to the chip-specific clockgen compatible, and keep the detailed description of quirky chip variations in the driver, where it can be easily fixed, refactored, and extended. Older device trees will continue to work (including a workaround for old ls1021a device trees that are missing compatible and reg in the clockgen node, which even the old binding required). The pll/mux details in old device trees will be ignored, but "clocks" properties pointing at the old nodes will still work, and be directed at the corresponding new clock. Signed-off-by: Scott Wood <scottwood@freescale.com> Acked-by: Stephen Boyd <sboyd@codeaurora.org>
2015-09-20 12:29:54 +08:00
return false;
}
#else
static bool __init has_erratum_a4510(void)
{
return false;
}
#endif
clk: qoriq: Move chip-specific knowledge into driver The device tree should describe the chips (or chip-like subblocks) in the system, but it generally does not describe individual registers -- it should identify, rather than describe, a programming interface. This has not been the case with the QorIQ clockgen nodes. The knowledge of what each bit setting of CLKCnCSR means is encoded in three places (binding, pll node, and mux node), and the last also needs to know which options are valid on a particular chip. All three of these locations are considered stable ABI, making it difficult to fix mistakes (of which I have found several), much less refactor the abstraction to be able to address problems, limitations, or new chips. Under the current binding, a pll clock specifier of 2 means that the PLL is divided by 4 -- and the driver implements this, unless there happen to be four clock-output-names rather than 3, in which case it interprets it as PLL divided by 3. This does not appear in the binding documentation at all. That hack is now considered stable ABI. The current device tree nodes contain errors, such as saying that T1040 can set a core clock to PLL/4 when only PLL and PLL/2 are options. The current binding also ignores some restrictions on clock selection, such as p5020's requirement that if a core uses the "wrong" PLL, that PLL must be clocked lower than the "correct" PLL and be at most 80% of the rated CPU frequency. Possibly because of the lack of the ability to express such nuance in the binding, some valid options are omitted from the device trees, such as the ability on p4080 to run cores 0-3 from PLL3 and cores 4-7 from PLL1 (again, only if they are at most 80% of rated CPU frequency). This omission, combined with excessive caution in the cpufreq driver (addressed in a subsequent patch), means that currently on a 1500 MHz p4080 with typical PLL configuration, cpufreq can lower the frequency to 1200 MHz on half the CPUs and do nothing on the others. With this patchset, all CPUs can be lowered to 1200 MHz on a rev2 p4080, and on a rev3 p4080 half can be lowered to 750 MHz and the other half to 600 MHz. The current binding only deals with CPU clocks. To describe FMan in the device tree, we need to describe its clock. Some chips have additional muxes that work like the CPU muxes, but are not described in the device tree. Others require inspecting the Reset Control Word to determine which PLL is used. Rather than continue to extend this mess, replace it. Have the driver bind to the chip-specific clockgen compatible, and keep the detailed description of quirky chip variations in the driver, where it can be easily fixed, refactored, and extended. Older device trees will continue to work (including a workaround for old ls1021a device trees that are missing compatible and reg in the clockgen node, which even the old binding required). The pll/mux details in old device trees will be ignored, but "clocks" properties pointing at the old nodes will still work, and be directed at the corresponding new clock. Signed-off-by: Scott Wood <scottwood@freescale.com> Acked-by: Stephen Boyd <sboyd@codeaurora.org>
2015-09-20 12:29:54 +08:00
static void __init clockgen_init(struct device_node *np)
{
int i, ret;
bool is_old_ls1021a = false;
clk: qoriq: Move chip-specific knowledge into driver The device tree should describe the chips (or chip-like subblocks) in the system, but it generally does not describe individual registers -- it should identify, rather than describe, a programming interface. This has not been the case with the QorIQ clockgen nodes. The knowledge of what each bit setting of CLKCnCSR means is encoded in three places (binding, pll node, and mux node), and the last also needs to know which options are valid on a particular chip. All three of these locations are considered stable ABI, making it difficult to fix mistakes (of which I have found several), much less refactor the abstraction to be able to address problems, limitations, or new chips. Under the current binding, a pll clock specifier of 2 means that the PLL is divided by 4 -- and the driver implements this, unless there happen to be four clock-output-names rather than 3, in which case it interprets it as PLL divided by 3. This does not appear in the binding documentation at all. That hack is now considered stable ABI. The current device tree nodes contain errors, such as saying that T1040 can set a core clock to PLL/4 when only PLL and PLL/2 are options. The current binding also ignores some restrictions on clock selection, such as p5020's requirement that if a core uses the "wrong" PLL, that PLL must be clocked lower than the "correct" PLL and be at most 80% of the rated CPU frequency. Possibly because of the lack of the ability to express such nuance in the binding, some valid options are omitted from the device trees, such as the ability on p4080 to run cores 0-3 from PLL3 and cores 4-7 from PLL1 (again, only if they are at most 80% of rated CPU frequency). This omission, combined with excessive caution in the cpufreq driver (addressed in a subsequent patch), means that currently on a 1500 MHz p4080 with typical PLL configuration, cpufreq can lower the frequency to 1200 MHz on half the CPUs and do nothing on the others. With this patchset, all CPUs can be lowered to 1200 MHz on a rev2 p4080, and on a rev3 p4080 half can be lowered to 750 MHz and the other half to 600 MHz. The current binding only deals with CPU clocks. To describe FMan in the device tree, we need to describe its clock. Some chips have additional muxes that work like the CPU muxes, but are not described in the device tree. Others require inspecting the Reset Control Word to determine which PLL is used. Rather than continue to extend this mess, replace it. Have the driver bind to the chip-specific clockgen compatible, and keep the detailed description of quirky chip variations in the driver, where it can be easily fixed, refactored, and extended. Older device trees will continue to work (including a workaround for old ls1021a device trees that are missing compatible and reg in the clockgen node, which even the old binding required). The pll/mux details in old device trees will be ignored, but "clocks" properties pointing at the old nodes will still work, and be directed at the corresponding new clock. Signed-off-by: Scott Wood <scottwood@freescale.com> Acked-by: Stephen Boyd <sboyd@codeaurora.org>
2015-09-20 12:29:54 +08:00
/* May have already been called by a legacy probe */
if (clockgen.node)
return;
clk: qoriq: Move chip-specific knowledge into driver The device tree should describe the chips (or chip-like subblocks) in the system, but it generally does not describe individual registers -- it should identify, rather than describe, a programming interface. This has not been the case with the QorIQ clockgen nodes. The knowledge of what each bit setting of CLKCnCSR means is encoded in three places (binding, pll node, and mux node), and the last also needs to know which options are valid on a particular chip. All three of these locations are considered stable ABI, making it difficult to fix mistakes (of which I have found several), much less refactor the abstraction to be able to address problems, limitations, or new chips. Under the current binding, a pll clock specifier of 2 means that the PLL is divided by 4 -- and the driver implements this, unless there happen to be four clock-output-names rather than 3, in which case it interprets it as PLL divided by 3. This does not appear in the binding documentation at all. That hack is now considered stable ABI. The current device tree nodes contain errors, such as saying that T1040 can set a core clock to PLL/4 when only PLL and PLL/2 are options. The current binding also ignores some restrictions on clock selection, such as p5020's requirement that if a core uses the "wrong" PLL, that PLL must be clocked lower than the "correct" PLL and be at most 80% of the rated CPU frequency. Possibly because of the lack of the ability to express such nuance in the binding, some valid options are omitted from the device trees, such as the ability on p4080 to run cores 0-3 from PLL3 and cores 4-7 from PLL1 (again, only if they are at most 80% of rated CPU frequency). This omission, combined with excessive caution in the cpufreq driver (addressed in a subsequent patch), means that currently on a 1500 MHz p4080 with typical PLL configuration, cpufreq can lower the frequency to 1200 MHz on half the CPUs and do nothing on the others. With this patchset, all CPUs can be lowered to 1200 MHz on a rev2 p4080, and on a rev3 p4080 half can be lowered to 750 MHz and the other half to 600 MHz. The current binding only deals with CPU clocks. To describe FMan in the device tree, we need to describe its clock. Some chips have additional muxes that work like the CPU muxes, but are not described in the device tree. Others require inspecting the Reset Control Word to determine which PLL is used. Rather than continue to extend this mess, replace it. Have the driver bind to the chip-specific clockgen compatible, and keep the detailed description of quirky chip variations in the driver, where it can be easily fixed, refactored, and extended. Older device trees will continue to work (including a workaround for old ls1021a device trees that are missing compatible and reg in the clockgen node, which even the old binding required). The pll/mux details in old device trees will be ignored, but "clocks" properties pointing at the old nodes will still work, and be directed at the corresponding new clock. Signed-off-by: Scott Wood <scottwood@freescale.com> Acked-by: Stephen Boyd <sboyd@codeaurora.org>
2015-09-20 12:29:54 +08:00
clockgen.node = np;
clockgen.regs = of_iomap(np, 0);
if (!clockgen.regs &&
of_device_is_compatible(of_root, "fsl,ls1021a")) {
/* Compatibility hack for old, broken device trees */
clockgen.regs = ioremap(0x1ee1000, 0x1000);
is_old_ls1021a = true;
}
if (!clockgen.regs) {
pr_err("%s(): %s: of_iomap() failed\n", __func__, np->name);
return;
}
clk: qoriq: Move chip-specific knowledge into driver The device tree should describe the chips (or chip-like subblocks) in the system, but it generally does not describe individual registers -- it should identify, rather than describe, a programming interface. This has not been the case with the QorIQ clockgen nodes. The knowledge of what each bit setting of CLKCnCSR means is encoded in three places (binding, pll node, and mux node), and the last also needs to know which options are valid on a particular chip. All three of these locations are considered stable ABI, making it difficult to fix mistakes (of which I have found several), much less refactor the abstraction to be able to address problems, limitations, or new chips. Under the current binding, a pll clock specifier of 2 means that the PLL is divided by 4 -- and the driver implements this, unless there happen to be four clock-output-names rather than 3, in which case it interprets it as PLL divided by 3. This does not appear in the binding documentation at all. That hack is now considered stable ABI. The current device tree nodes contain errors, such as saying that T1040 can set a core clock to PLL/4 when only PLL and PLL/2 are options. The current binding also ignores some restrictions on clock selection, such as p5020's requirement that if a core uses the "wrong" PLL, that PLL must be clocked lower than the "correct" PLL and be at most 80% of the rated CPU frequency. Possibly because of the lack of the ability to express such nuance in the binding, some valid options are omitted from the device trees, such as the ability on p4080 to run cores 0-3 from PLL3 and cores 4-7 from PLL1 (again, only if they are at most 80% of rated CPU frequency). This omission, combined with excessive caution in the cpufreq driver (addressed in a subsequent patch), means that currently on a 1500 MHz p4080 with typical PLL configuration, cpufreq can lower the frequency to 1200 MHz on half the CPUs and do nothing on the others. With this patchset, all CPUs can be lowered to 1200 MHz on a rev2 p4080, and on a rev3 p4080 half can be lowered to 750 MHz and the other half to 600 MHz. The current binding only deals with CPU clocks. To describe FMan in the device tree, we need to describe its clock. Some chips have additional muxes that work like the CPU muxes, but are not described in the device tree. Others require inspecting the Reset Control Word to determine which PLL is used. Rather than continue to extend this mess, replace it. Have the driver bind to the chip-specific clockgen compatible, and keep the detailed description of quirky chip variations in the driver, where it can be easily fixed, refactored, and extended. Older device trees will continue to work (including a workaround for old ls1021a device trees that are missing compatible and reg in the clockgen node, which even the old binding required). The pll/mux details in old device trees will be ignored, but "clocks" properties pointing at the old nodes will still work, and be directed at the corresponding new clock. Signed-off-by: Scott Wood <scottwood@freescale.com> Acked-by: Stephen Boyd <sboyd@codeaurora.org>
2015-09-20 12:29:54 +08:00
for (i = 0; i < ARRAY_SIZE(chipinfo); i++) {
if (of_device_is_compatible(np, chipinfo[i].compat))
break;
if (is_old_ls1021a &&
!strcmp(chipinfo[i].compat, "fsl,ls1021a-clockgen"))
break;
}
clk: qoriq: Move chip-specific knowledge into driver The device tree should describe the chips (or chip-like subblocks) in the system, but it generally does not describe individual registers -- it should identify, rather than describe, a programming interface. This has not been the case with the QorIQ clockgen nodes. The knowledge of what each bit setting of CLKCnCSR means is encoded in three places (binding, pll node, and mux node), and the last also needs to know which options are valid on a particular chip. All three of these locations are considered stable ABI, making it difficult to fix mistakes (of which I have found several), much less refactor the abstraction to be able to address problems, limitations, or new chips. Under the current binding, a pll clock specifier of 2 means that the PLL is divided by 4 -- and the driver implements this, unless there happen to be four clock-output-names rather than 3, in which case it interprets it as PLL divided by 3. This does not appear in the binding documentation at all. That hack is now considered stable ABI. The current device tree nodes contain errors, such as saying that T1040 can set a core clock to PLL/4 when only PLL and PLL/2 are options. The current binding also ignores some restrictions on clock selection, such as p5020's requirement that if a core uses the "wrong" PLL, that PLL must be clocked lower than the "correct" PLL and be at most 80% of the rated CPU frequency. Possibly because of the lack of the ability to express such nuance in the binding, some valid options are omitted from the device trees, such as the ability on p4080 to run cores 0-3 from PLL3 and cores 4-7 from PLL1 (again, only if they are at most 80% of rated CPU frequency). This omission, combined with excessive caution in the cpufreq driver (addressed in a subsequent patch), means that currently on a 1500 MHz p4080 with typical PLL configuration, cpufreq can lower the frequency to 1200 MHz on half the CPUs and do nothing on the others. With this patchset, all CPUs can be lowered to 1200 MHz on a rev2 p4080, and on a rev3 p4080 half can be lowered to 750 MHz and the other half to 600 MHz. The current binding only deals with CPU clocks. To describe FMan in the device tree, we need to describe its clock. Some chips have additional muxes that work like the CPU muxes, but are not described in the device tree. Others require inspecting the Reset Control Word to determine which PLL is used. Rather than continue to extend this mess, replace it. Have the driver bind to the chip-specific clockgen compatible, and keep the detailed description of quirky chip variations in the driver, where it can be easily fixed, refactored, and extended. Older device trees will continue to work (including a workaround for old ls1021a device trees that are missing compatible and reg in the clockgen node, which even the old binding required). The pll/mux details in old device trees will be ignored, but "clocks" properties pointing at the old nodes will still work, and be directed at the corresponding new clock. Signed-off-by: Scott Wood <scottwood@freescale.com> Acked-by: Stephen Boyd <sboyd@codeaurora.org>
2015-09-20 12:29:54 +08:00
if (i == ARRAY_SIZE(chipinfo)) {
pr_err("%s: unknown clockgen node %s\n", __func__,
np->full_name);
goto err;
}
clockgen.info = chipinfo[i];
if (clockgen.info.guts_compat) {
struct device_node *guts;
clk: qoriq: Move chip-specific knowledge into driver The device tree should describe the chips (or chip-like subblocks) in the system, but it generally does not describe individual registers -- it should identify, rather than describe, a programming interface. This has not been the case with the QorIQ clockgen nodes. The knowledge of what each bit setting of CLKCnCSR means is encoded in three places (binding, pll node, and mux node), and the last also needs to know which options are valid on a particular chip. All three of these locations are considered stable ABI, making it difficult to fix mistakes (of which I have found several), much less refactor the abstraction to be able to address problems, limitations, or new chips. Under the current binding, a pll clock specifier of 2 means that the PLL is divided by 4 -- and the driver implements this, unless there happen to be four clock-output-names rather than 3, in which case it interprets it as PLL divided by 3. This does not appear in the binding documentation at all. That hack is now considered stable ABI. The current device tree nodes contain errors, such as saying that T1040 can set a core clock to PLL/4 when only PLL and PLL/2 are options. The current binding also ignores some restrictions on clock selection, such as p5020's requirement that if a core uses the "wrong" PLL, that PLL must be clocked lower than the "correct" PLL and be at most 80% of the rated CPU frequency. Possibly because of the lack of the ability to express such nuance in the binding, some valid options are omitted from the device trees, such as the ability on p4080 to run cores 0-3 from PLL3 and cores 4-7 from PLL1 (again, only if they are at most 80% of rated CPU frequency). This omission, combined with excessive caution in the cpufreq driver (addressed in a subsequent patch), means that currently on a 1500 MHz p4080 with typical PLL configuration, cpufreq can lower the frequency to 1200 MHz on half the CPUs and do nothing on the others. With this patchset, all CPUs can be lowered to 1200 MHz on a rev2 p4080, and on a rev3 p4080 half can be lowered to 750 MHz and the other half to 600 MHz. The current binding only deals with CPU clocks. To describe FMan in the device tree, we need to describe its clock. Some chips have additional muxes that work like the CPU muxes, but are not described in the device tree. Others require inspecting the Reset Control Word to determine which PLL is used. Rather than continue to extend this mess, replace it. Have the driver bind to the chip-specific clockgen compatible, and keep the detailed description of quirky chip variations in the driver, where it can be easily fixed, refactored, and extended. Older device trees will continue to work (including a workaround for old ls1021a device trees that are missing compatible and reg in the clockgen node, which even the old binding required). The pll/mux details in old device trees will be ignored, but "clocks" properties pointing at the old nodes will still work, and be directed at the corresponding new clock. Signed-off-by: Scott Wood <scottwood@freescale.com> Acked-by: Stephen Boyd <sboyd@codeaurora.org>
2015-09-20 12:29:54 +08:00
guts = of_find_compatible_node(NULL, NULL,
clockgen.info.guts_compat);
if (guts) {
clockgen.guts = of_iomap(guts, 0);
if (!clockgen.guts) {
pr_err("%s: Couldn't map %s regs\n", __func__,
guts->full_name);
}
}
clk: qoriq: Move chip-specific knowledge into driver The device tree should describe the chips (or chip-like subblocks) in the system, but it generally does not describe individual registers -- it should identify, rather than describe, a programming interface. This has not been the case with the QorIQ clockgen nodes. The knowledge of what each bit setting of CLKCnCSR means is encoded in three places (binding, pll node, and mux node), and the last also needs to know which options are valid on a particular chip. All three of these locations are considered stable ABI, making it difficult to fix mistakes (of which I have found several), much less refactor the abstraction to be able to address problems, limitations, or new chips. Under the current binding, a pll clock specifier of 2 means that the PLL is divided by 4 -- and the driver implements this, unless there happen to be four clock-output-names rather than 3, in which case it interprets it as PLL divided by 3. This does not appear in the binding documentation at all. That hack is now considered stable ABI. The current device tree nodes contain errors, such as saying that T1040 can set a core clock to PLL/4 when only PLL and PLL/2 are options. The current binding also ignores some restrictions on clock selection, such as p5020's requirement that if a core uses the "wrong" PLL, that PLL must be clocked lower than the "correct" PLL and be at most 80% of the rated CPU frequency. Possibly because of the lack of the ability to express such nuance in the binding, some valid options are omitted from the device trees, such as the ability on p4080 to run cores 0-3 from PLL3 and cores 4-7 from PLL1 (again, only if they are at most 80% of rated CPU frequency). This omission, combined with excessive caution in the cpufreq driver (addressed in a subsequent patch), means that currently on a 1500 MHz p4080 with typical PLL configuration, cpufreq can lower the frequency to 1200 MHz on half the CPUs and do nothing on the others. With this patchset, all CPUs can be lowered to 1200 MHz on a rev2 p4080, and on a rev3 p4080 half can be lowered to 750 MHz and the other half to 600 MHz. The current binding only deals with CPU clocks. To describe FMan in the device tree, we need to describe its clock. Some chips have additional muxes that work like the CPU muxes, but are not described in the device tree. Others require inspecting the Reset Control Word to determine which PLL is used. Rather than continue to extend this mess, replace it. Have the driver bind to the chip-specific clockgen compatible, and keep the detailed description of quirky chip variations in the driver, where it can be easily fixed, refactored, and extended. Older device trees will continue to work (including a workaround for old ls1021a device trees that are missing compatible and reg in the clockgen node, which even the old binding required). The pll/mux details in old device trees will be ignored, but "clocks" properties pointing at the old nodes will still work, and be directed at the corresponding new clock. Signed-off-by: Scott Wood <scottwood@freescale.com> Acked-by: Stephen Boyd <sboyd@codeaurora.org>
2015-09-20 12:29:54 +08:00
}
clk: qoriq: Move chip-specific knowledge into driver The device tree should describe the chips (or chip-like subblocks) in the system, but it generally does not describe individual registers -- it should identify, rather than describe, a programming interface. This has not been the case with the QorIQ clockgen nodes. The knowledge of what each bit setting of CLKCnCSR means is encoded in three places (binding, pll node, and mux node), and the last also needs to know which options are valid on a particular chip. All three of these locations are considered stable ABI, making it difficult to fix mistakes (of which I have found several), much less refactor the abstraction to be able to address problems, limitations, or new chips. Under the current binding, a pll clock specifier of 2 means that the PLL is divided by 4 -- and the driver implements this, unless there happen to be four clock-output-names rather than 3, in which case it interprets it as PLL divided by 3. This does not appear in the binding documentation at all. That hack is now considered stable ABI. The current device tree nodes contain errors, such as saying that T1040 can set a core clock to PLL/4 when only PLL and PLL/2 are options. The current binding also ignores some restrictions on clock selection, such as p5020's requirement that if a core uses the "wrong" PLL, that PLL must be clocked lower than the "correct" PLL and be at most 80% of the rated CPU frequency. Possibly because of the lack of the ability to express such nuance in the binding, some valid options are omitted from the device trees, such as the ability on p4080 to run cores 0-3 from PLL3 and cores 4-7 from PLL1 (again, only if they are at most 80% of rated CPU frequency). This omission, combined with excessive caution in the cpufreq driver (addressed in a subsequent patch), means that currently on a 1500 MHz p4080 with typical PLL configuration, cpufreq can lower the frequency to 1200 MHz on half the CPUs and do nothing on the others. With this patchset, all CPUs can be lowered to 1200 MHz on a rev2 p4080, and on a rev3 p4080 half can be lowered to 750 MHz and the other half to 600 MHz. The current binding only deals with CPU clocks. To describe FMan in the device tree, we need to describe its clock. Some chips have additional muxes that work like the CPU muxes, but are not described in the device tree. Others require inspecting the Reset Control Word to determine which PLL is used. Rather than continue to extend this mess, replace it. Have the driver bind to the chip-specific clockgen compatible, and keep the detailed description of quirky chip variations in the driver, where it can be easily fixed, refactored, and extended. Older device trees will continue to work (including a workaround for old ls1021a device trees that are missing compatible and reg in the clockgen node, which even the old binding required). The pll/mux details in old device trees will be ignored, but "clocks" properties pointing at the old nodes will still work, and be directed at the corresponding new clock. Signed-off-by: Scott Wood <scottwood@freescale.com> Acked-by: Stephen Boyd <sboyd@codeaurora.org>
2015-09-20 12:29:54 +08:00
if (has_erratum_a4510())
clockgen.info.flags |= CG_CMUX_GE_PLAT;
clockgen.sysclk = create_sysclk("cg-sysclk");
create_plls(&clockgen);
create_muxes(&clockgen);
if (clockgen.info.init_periph)
clockgen.info.init_periph(&clockgen);
ret = of_clk_add_provider(np, clockgen_clk_get, &clockgen);
if (ret) {
pr_err("%s: Couldn't register clk provider for node %s: %d\n",
__func__, np->name, ret);
}
return;
clk: qoriq: Move chip-specific knowledge into driver The device tree should describe the chips (or chip-like subblocks) in the system, but it generally does not describe individual registers -- it should identify, rather than describe, a programming interface. This has not been the case with the QorIQ clockgen nodes. The knowledge of what each bit setting of CLKCnCSR means is encoded in three places (binding, pll node, and mux node), and the last also needs to know which options are valid on a particular chip. All three of these locations are considered stable ABI, making it difficult to fix mistakes (of which I have found several), much less refactor the abstraction to be able to address problems, limitations, or new chips. Under the current binding, a pll clock specifier of 2 means that the PLL is divided by 4 -- and the driver implements this, unless there happen to be four clock-output-names rather than 3, in which case it interprets it as PLL divided by 3. This does not appear in the binding documentation at all. That hack is now considered stable ABI. The current device tree nodes contain errors, such as saying that T1040 can set a core clock to PLL/4 when only PLL and PLL/2 are options. The current binding also ignores some restrictions on clock selection, such as p5020's requirement that if a core uses the "wrong" PLL, that PLL must be clocked lower than the "correct" PLL and be at most 80% of the rated CPU frequency. Possibly because of the lack of the ability to express such nuance in the binding, some valid options are omitted from the device trees, such as the ability on p4080 to run cores 0-3 from PLL3 and cores 4-7 from PLL1 (again, only if they are at most 80% of rated CPU frequency). This omission, combined with excessive caution in the cpufreq driver (addressed in a subsequent patch), means that currently on a 1500 MHz p4080 with typical PLL configuration, cpufreq can lower the frequency to 1200 MHz on half the CPUs and do nothing on the others. With this patchset, all CPUs can be lowered to 1200 MHz on a rev2 p4080, and on a rev3 p4080 half can be lowered to 750 MHz and the other half to 600 MHz. The current binding only deals with CPU clocks. To describe FMan in the device tree, we need to describe its clock. Some chips have additional muxes that work like the CPU muxes, but are not described in the device tree. Others require inspecting the Reset Control Word to determine which PLL is used. Rather than continue to extend this mess, replace it. Have the driver bind to the chip-specific clockgen compatible, and keep the detailed description of quirky chip variations in the driver, where it can be easily fixed, refactored, and extended. Older device trees will continue to work (including a workaround for old ls1021a device trees that are missing compatible and reg in the clockgen node, which even the old binding required). The pll/mux details in old device trees will be ignored, but "clocks" properties pointing at the old nodes will still work, and be directed at the corresponding new clock. Signed-off-by: Scott Wood <scottwood@freescale.com> Acked-by: Stephen Boyd <sboyd@codeaurora.org>
2015-09-20 12:29:54 +08:00
err:
iounmap(clockgen.regs);
clockgen.regs = NULL;
}
clk: qoriq: Move chip-specific knowledge into driver The device tree should describe the chips (or chip-like subblocks) in the system, but it generally does not describe individual registers -- it should identify, rather than describe, a programming interface. This has not been the case with the QorIQ clockgen nodes. The knowledge of what each bit setting of CLKCnCSR means is encoded in three places (binding, pll node, and mux node), and the last also needs to know which options are valid on a particular chip. All three of these locations are considered stable ABI, making it difficult to fix mistakes (of which I have found several), much less refactor the abstraction to be able to address problems, limitations, or new chips. Under the current binding, a pll clock specifier of 2 means that the PLL is divided by 4 -- and the driver implements this, unless there happen to be four clock-output-names rather than 3, in which case it interprets it as PLL divided by 3. This does not appear in the binding documentation at all. That hack is now considered stable ABI. The current device tree nodes contain errors, such as saying that T1040 can set a core clock to PLL/4 when only PLL and PLL/2 are options. The current binding also ignores some restrictions on clock selection, such as p5020's requirement that if a core uses the "wrong" PLL, that PLL must be clocked lower than the "correct" PLL and be at most 80% of the rated CPU frequency. Possibly because of the lack of the ability to express such nuance in the binding, some valid options are omitted from the device trees, such as the ability on p4080 to run cores 0-3 from PLL3 and cores 4-7 from PLL1 (again, only if they are at most 80% of rated CPU frequency). This omission, combined with excessive caution in the cpufreq driver (addressed in a subsequent patch), means that currently on a 1500 MHz p4080 with typical PLL configuration, cpufreq can lower the frequency to 1200 MHz on half the CPUs and do nothing on the others. With this patchset, all CPUs can be lowered to 1200 MHz on a rev2 p4080, and on a rev3 p4080 half can be lowered to 750 MHz and the other half to 600 MHz. The current binding only deals with CPU clocks. To describe FMan in the device tree, we need to describe its clock. Some chips have additional muxes that work like the CPU muxes, but are not described in the device tree. Others require inspecting the Reset Control Word to determine which PLL is used. Rather than continue to extend this mess, replace it. Have the driver bind to the chip-specific clockgen compatible, and keep the detailed description of quirky chip variations in the driver, where it can be easily fixed, refactored, and extended. Older device trees will continue to work (including a workaround for old ls1021a device trees that are missing compatible and reg in the clockgen node, which even the old binding required). The pll/mux details in old device trees will be ignored, but "clocks" properties pointing at the old nodes will still work, and be directed at the corresponding new clock. Signed-off-by: Scott Wood <scottwood@freescale.com> Acked-by: Stephen Boyd <sboyd@codeaurora.org>
2015-09-20 12:29:54 +08:00
CLK_OF_DECLARE(qoriq_clockgen_1, "fsl,qoriq-clockgen-1.0", clockgen_init);
CLK_OF_DECLARE(qoriq_clockgen_2, "fsl,qoriq-clockgen-2.0", clockgen_init);
CLK_OF_DECLARE(qoriq_clockgen_ls1021a, "fsl,ls1021a-clockgen", clockgen_init);
CLK_OF_DECLARE(qoriq_clockgen_ls1043a, "fsl,ls1043a-clockgen", clockgen_init);
CLK_OF_DECLARE(qoriq_clockgen_ls2080a, "fsl,ls2080a-clockgen", clockgen_init);
clk: qoriq: Move chip-specific knowledge into driver The device tree should describe the chips (or chip-like subblocks) in the system, but it generally does not describe individual registers -- it should identify, rather than describe, a programming interface. This has not been the case with the QorIQ clockgen nodes. The knowledge of what each bit setting of CLKCnCSR means is encoded in three places (binding, pll node, and mux node), and the last also needs to know which options are valid on a particular chip. All three of these locations are considered stable ABI, making it difficult to fix mistakes (of which I have found several), much less refactor the abstraction to be able to address problems, limitations, or new chips. Under the current binding, a pll clock specifier of 2 means that the PLL is divided by 4 -- and the driver implements this, unless there happen to be four clock-output-names rather than 3, in which case it interprets it as PLL divided by 3. This does not appear in the binding documentation at all. That hack is now considered stable ABI. The current device tree nodes contain errors, such as saying that T1040 can set a core clock to PLL/4 when only PLL and PLL/2 are options. The current binding also ignores some restrictions on clock selection, such as p5020's requirement that if a core uses the "wrong" PLL, that PLL must be clocked lower than the "correct" PLL and be at most 80% of the rated CPU frequency. Possibly because of the lack of the ability to express such nuance in the binding, some valid options are omitted from the device trees, such as the ability on p4080 to run cores 0-3 from PLL3 and cores 4-7 from PLL1 (again, only if they are at most 80% of rated CPU frequency). This omission, combined with excessive caution in the cpufreq driver (addressed in a subsequent patch), means that currently on a 1500 MHz p4080 with typical PLL configuration, cpufreq can lower the frequency to 1200 MHz on half the CPUs and do nothing on the others. With this patchset, all CPUs can be lowered to 1200 MHz on a rev2 p4080, and on a rev3 p4080 half can be lowered to 750 MHz and the other half to 600 MHz. The current binding only deals with CPU clocks. To describe FMan in the device tree, we need to describe its clock. Some chips have additional muxes that work like the CPU muxes, but are not described in the device tree. Others require inspecting the Reset Control Word to determine which PLL is used. Rather than continue to extend this mess, replace it. Have the driver bind to the chip-specific clockgen compatible, and keep the detailed description of quirky chip variations in the driver, where it can be easily fixed, refactored, and extended. Older device trees will continue to work (including a workaround for old ls1021a device trees that are missing compatible and reg in the clockgen node, which even the old binding required). The pll/mux details in old device trees will be ignored, but "clocks" properties pointing at the old nodes will still work, and be directed at the corresponding new clock. Signed-off-by: Scott Wood <scottwood@freescale.com> Acked-by: Stephen Boyd <sboyd@codeaurora.org>
2015-09-20 12:29:54 +08:00
/* Legacy nodes */
CLK_OF_DECLARE(qoriq_sysclk_1, "fsl,qoriq-sysclk-1.0", sysclk_init);
CLK_OF_DECLARE(qoriq_sysclk_2, "fsl,qoriq-sysclk-2.0", sysclk_init);
CLK_OF_DECLARE(qoriq_core_pll_1, "fsl,qoriq-core-pll-1.0", core_pll_init);
CLK_OF_DECLARE(qoriq_core_pll_2, "fsl,qoriq-core-pll-2.0", core_pll_init);
CLK_OF_DECLARE(qoriq_core_mux_1, "fsl,qoriq-core-mux-1.0", core_mux_init);
CLK_OF_DECLARE(qoriq_core_mux_2, "fsl,qoriq-core-mux-2.0", core_mux_init);
CLK_OF_DECLARE(qoriq_pltfrm_pll_1, "fsl,qoriq-platform-pll-1.0", pltfrm_pll_init);
CLK_OF_DECLARE(qoriq_pltfrm_pll_2, "fsl,qoriq-platform-pll-2.0", pltfrm_pll_init);