diff options
author | dmitry pervushin <dimka@nomadgs.com> | 2007-04-24 13:39:09 +0900 |
---|---|---|
committer | Paul Mundt <lethal@hera.kernel.org> | 2007-05-07 02:11:56 +0000 |
commit | 1929cb340b74904c130fdf3de3fe5bbedb68a5aa (patch) | |
tree | 2675de406fca78e6bfd5b29535bbb220a167bc2e /arch/sh | |
parent | 34a780a0afeb8f99c9ca9934f4cc0822541769c6 (diff) | |
download | op-kernel-dev-1929cb340b74904c130fdf3de3fe5bbedb68a5aa.zip op-kernel-dev-1929cb340b74904c130fdf3de3fe5bbedb68a5aa.tar.gz |
sh: SH7722 clock framework support.
This adds support for the SH7722 (MobileR) to the clock framework.
Signed-off-by: dmitry pervushin <dimka@nomadgs.com>
Signed-off-by: Paul Mundt <lethal@linux-sh.org>
Diffstat (limited to 'arch/sh')
-rw-r--r-- | arch/sh/kernel/cpu/clock.c | 28 | ||||
-rw-r--r-- | arch/sh/kernel/cpu/sh4/clock-sh4-202.c | 3 | ||||
-rw-r--r-- | arch/sh/kernel/cpu/sh4a/Makefile | 2 | ||||
-rw-r--r-- | arch/sh/kernel/cpu/sh4a/clock-sh7722.c | 600 |
4 files changed, 627 insertions, 6 deletions
diff --git a/arch/sh/kernel/cpu/clock.c b/arch/sh/kernel/cpu/clock.c index abb586b1..2075f90 100644 --- a/arch/sh/kernel/cpu/clock.c +++ b/arch/sh/kernel/cpu/clock.c @@ -98,13 +98,14 @@ int __clk_enable(struct clk *clk) if (clk->ops && clk->ops->init) clk->ops->init(clk); + kref_get(&clk->kref); + if (clk->flags & CLK_ALWAYS_ENABLED) return 0; if (likely(clk->ops && clk->ops->enable)) clk->ops->enable(clk); - kref_get(&clk->kref); return 0; } @@ -127,10 +128,15 @@ static void clk_kref_release(struct kref *kref) void __clk_disable(struct clk *clk) { + int count = kref_put(&clk->kref, clk_kref_release); + if (clk->flags & CLK_ALWAYS_ENABLED) return; - kref_put(&clk->kref, clk_kref_release); + if (!count) { /* count reaches zero, disable the clock */ + if (likely(clk->ops && clk->ops->disable)) + clk->ops->disable(clk); + } } void clk_disable(struct clk *clk) @@ -151,6 +157,15 @@ int clk_register(struct clk *clk) mutex_unlock(&clock_list_sem); + if (clk->flags & CLK_ALWAYS_ENABLED) { + pr_debug( "Clock '%s' is ALWAYS_ENABLED\n", clk->name); + if (clk->ops && clk->ops->init) + clk->ops->init(clk); + if (clk->ops && clk->ops->enable) + clk->ops->enable(clk); + pr_debug( "Enabled."); + } + return 0; } @@ -168,13 +183,18 @@ inline unsigned long clk_get_rate(struct clk *clk) int clk_set_rate(struct clk *clk, unsigned long rate) { + return clk_set_rate_ex(clk, rate, 0); +} + +int clk_set_rate_ex(struct clk *clk, unsigned long rate, int algo_id) +{ int ret = -EOPNOTSUPP; if (likely(clk->ops && clk->ops->set_rate)) { unsigned long flags; spin_lock_irqsave(&clock_lock, flags); - ret = clk->ops->set_rate(clk, rate); + ret = clk->ops->set_rate(clk, rate, algo_id); spin_unlock_irqrestore(&clock_lock, flags); } @@ -256,7 +276,6 @@ int __init clk_init(void) arch_init_clk_ops(&clk->ops, i); ret |= clk_register(clk); - clk_enable(clk); } /* Kick the child clocks.. */ @@ -298,3 +317,4 @@ EXPORT_SYMBOL_GPL(__clk_disable); EXPORT_SYMBOL_GPL(clk_get_rate); EXPORT_SYMBOL_GPL(clk_set_rate); EXPORT_SYMBOL_GPL(clk_recalc_rate); +EXPORT_SYMBOL_GPL(clk_set_rate_ex); diff --git a/arch/sh/kernel/cpu/sh4/clock-sh4-202.c b/arch/sh/kernel/cpu/sh4/clock-sh4-202.c index fa2019a..fcb2c41b 100644 --- a/arch/sh/kernel/cpu/sh4/clock-sh4-202.c +++ b/arch/sh/kernel/cpu/sh4/clock-sh4-202.c @@ -82,7 +82,8 @@ static void shoc_clk_init(struct clk *clk) for (i = 0; i < ARRAY_SIZE(frqcr3_divisors); i++) { int divisor = frqcr3_divisors[i]; - if (clk->ops->set_rate(clk, clk->parent->rate / divisor) == 0) + if (clk->ops->set_rate(clk, clk->parent->rate / + divisor, 0) == 0) break; } diff --git a/arch/sh/kernel/cpu/sh4a/Makefile b/arch/sh/kernel/cpu/sh4a/Makefile index 1efeda0..ab7422f 100644 --- a/arch/sh/kernel/cpu/sh4a/Makefile +++ b/arch/sh/kernel/cpu/sh4a/Makefile @@ -16,6 +16,6 @@ clock-$(CONFIG_CPU_SUBTYPE_SH7770) := clock-sh7770.o clock-$(CONFIG_CPU_SUBTYPE_SH7780) := clock-sh7780.o clock-$(CONFIG_CPU_SUBTYPE_SH7785) := clock-sh7785.o clock-$(CONFIG_CPU_SUBTYPE_SH7343) := clock-sh7343.o -clock-$(CONFIG_CPU_SUBTYPE_SH7722) := clock-sh7343.o +clock-$(CONFIG_CPU_SUBTYPE_SH7722) := clock-sh7722.o obj-y += $(clock-y) diff --git a/arch/sh/kernel/cpu/sh4a/clock-sh7722.c b/arch/sh/kernel/cpu/sh4a/clock-sh7722.c new file mode 100644 index 0000000..2909003 --- /dev/null +++ b/arch/sh/kernel/cpu/sh4a/clock-sh7722.c @@ -0,0 +1,600 @@ +/* + * arch/sh/kernel/cpu/sh4a/clock-sh7722.c + * + * SH7722 support for the clock framework + * + * Copyright (c) 2006-2007 Nomad Global Solutions Inc + * Based on code for sh7343 by Paul Mundt + * + * This file is subject to the terms and conditions of the GNU General Public + * License. See the file "COPYING" in the main directory of this archive + * for more details. + */ +#include <linux/init.h> +#include <linux/kernel.h> +#include <linux/io.h> +#include <linux/errno.h> +#include <asm/clock.h> +#include <asm/freq.h> + +#define SH7722_PLL_FREQ (32000000/8) +#define N (-1) +#define NM (-2) +#define ROUND_NEAREST 0 +#define ROUND_DOWN -1 +#define ROUND_UP +1 + +static int adjust_algos[][3] = { + {}, /* NO_CHANGE */ + { NM, N, 1 }, /* N:1, N:1 */ + { 3, 2, 2 }, /* 3:2:2 */ + { 5, 2, 2 }, /* 5:2:2 */ + { N, 1, 1 }, /* N:1:1 */ + + { N, 1 }, /* N:1 */ + + { N, 1 }, /* N:1 */ + { 3, 2 }, + { 4, 3 }, + { 5, 4 }, + + { N, 1 } +}; + +static unsigned long adjust_pair_of_clocks(unsigned long r1, unsigned long r2, + int m1, int m2, int round_flag) +{ + unsigned long rem, div; + int the_one = 0; + + pr_debug( "Actual values: r1 = %ld\n", r1); + pr_debug( "...............r2 = %ld\n", r2); + + if (m1 == m2) { + r2 = r1; + pr_debug( "setting equal rates: r2 now %ld\n", r2); + } else if ((m2 == N && m1 == 1) || + (m2 == NM && m1 == N)) { /* N:1 or NM:N */ + pr_debug( "Setting rates as 1:N (N:N*M)\n"); + rem = r2 % r1; + pr_debug( "...remainder = %ld\n", rem); + if (rem) { + div = r2 / r1; + pr_debug( "...div = %ld\n", div); + switch (round_flag) { + case ROUND_NEAREST: + the_one = rem >= r1/2 ? 1 : 0; break; + case ROUND_UP: + the_one = 1; break; + case ROUND_DOWN: + the_one = 0; break; + } + + r2 = r1 * (div + the_one); + pr_debug( "...setting r2 to %ld\n", r2); + } + } else if ((m2 == 1 && m1 == N) || + (m2 == N && m1 == NM)) { /* 1:N or N:NM */ + pr_debug( "Setting rates as N:1 (N*M:N)\n"); + rem = r1 % r2; + pr_debug( "...remainder = %ld\n", rem); + if (rem) { + div = r1 / r2; + pr_debug( "...div = %ld\n", div); + switch (round_flag) { + case ROUND_NEAREST: + the_one = rem > r2/2 ? 1 : 0; break; + case ROUND_UP: + the_one = 0; break; + case ROUND_DOWN: + the_one = 1; break; + } + + r2 = r1 / (div + the_one); + pr_debug( "...setting r2 to %ld\n", r2); + } + } else { /* value:value */ + pr_debug( "Setting rates as %d:%d\n", m1, m2); + div = r1 / m1; + r2 = div * m2; + pr_debug( "...div = %ld\n", div); + pr_debug( "...setting r2 to %ld\n", r2); + } + + return r2; +} + +static void adjust_clocks(int originate, int *l, unsigned long v[], + int n_in_line) +{ + int x; + + pr_debug( "Go down from %d...\n", originate); + /* go up recalculation clocks */ + for (x = originate; x>0; x -- ) + v[x-1] = adjust_pair_of_clocks(v[x], v[x-1], + l[x], l[x-1], + ROUND_UP); + + pr_debug( "Go up from %d...\n", originate); + /* go down recalculation clocks */ + for (x = originate; x<n_in_line - 1; x ++ ) + v[x+1] = adjust_pair_of_clocks(v[x], v[x+1], + l[x], l[x+1], + ROUND_UP); +} + + +/* + * SH7722 uses a common set of multipliers and divisors, so this + * is quite simple.. + */ + +/* + * Instead of having two separate multipliers/divisors set, like this: + * + * static int multipliers[] = { 1, 2, 1, 2, 1, 1, 1, 1, 1, 1, 1, 1, 1 }; + * static int divisors[] = { 1, 3, 2, 5, 3, 4, 5, 6, 8, 10, 12, 16, 20 }; + * + * I created the divisors2 array, which is used to calculate rate like + * rate = parent * 2 / divisors2[ divisor ]; +*/ +static int divisors2[] = { 2, 3, 4, 5, 6, 8, 10, 12, 16, 20, 24, 32, 40 }; + +static void master_clk_init(struct clk *clk) +{ + clk_set_rate(clk, clk_get_rate(clk)); +} + +static void master_clk_recalc(struct clk *clk) +{ + unsigned long frqcr = ctrl_inl(FRQCR); + + clk->rate = CONFIG_SH_PCLK_FREQ * (1 + (frqcr >> 24 & 0xF)); +} + +static int master_clk_setrate(struct clk *clk, unsigned long rate, int id) +{ + int div = rate / SH7722_PLL_FREQ; + int master_divs[] = { 2, 3, 4, 6, 8, 16 }; + int index; + unsigned long frqcr; + + if (rate < SH7722_PLL_FREQ * 2) + return -EINVAL; + + for (index = 1; index < ARRAY_SIZE(master_divs); index++) + if (div >= master_divs[index - 1] && div < master_divs[index]) + break; + + if (index >= ARRAY_SIZE(master_divs)) + index = ARRAY_SIZE(master_divs); + div = master_divs[index - 1]; + + frqcr = ctrl_inl(FRQCR); + frqcr &= ~(0xF << 24); + frqcr |= ( (div-1) << 24); + ctrl_outl(frqcr, FRQCR); + + return 0; +} + +static struct clk_ops sh7722_master_clk_ops = { + .init = master_clk_init, + .recalc = master_clk_recalc, + .set_rate = master_clk_setrate, +}; + +struct frqcr_context { + unsigned mask; + unsigned shift; +}; + +struct frqcr_context sh7722_get_clk_context(const char *name) +{ + struct frqcr_context ctx = { 0, }; + + if (!strcmp(name, "peripheral_clk")) { + ctx.shift = 0; + ctx.mask = 0xF; + } else if (!strcmp(name, "sdram_clk")) { + ctx.shift = 4; + ctx.mask = 0xF; + } else if (!strcmp(name, "bus_clk")) { + ctx.shift = 8; + ctx.mask = 0xF; + } else if (!strcmp(name, "sh_clk")) { + ctx.shift = 12; + ctx.mask = 0xF; + } else if (!strcmp(name, "umem_clk")) { + ctx.shift = 16; + ctx.mask = 0xF; + } else if (!strcmp(name, "cpu_clk")) { + ctx.shift = 20; + ctx.mask = 7; + } + return ctx; +} + +/** + * sh7722_find_divisors - find divisor for setting rate + * + * All sh7722 clocks use the same set of multipliers/divisors. This function + * chooses correct divisor to set the rate of clock with parent clock that + * generates frequency of 'parent_rate' + * + * @parent_rate: rate of parent clock + * @rate: requested rate to be set + */ +static int sh7722_find_divisors(unsigned long parent_rate, unsigned rate) +{ + unsigned div2 = parent_rate * 2 / rate; + int index; + + if (rate > parent_rate) + return -EINVAL; + + for (index = 1; index < ARRAY_SIZE(divisors2); index++) { + if (div2 > divisors2[index] && div2 <= divisors2[index]) + break; + } + if (index >= ARRAY_SIZE(divisors2)) + index = ARRAY_SIZE(divisors2) - 1; + return divisors2[index]; +} + +static void sh7722_frqcr_recalc(struct clk *clk) +{ + struct frqcr_context ctx = sh7722_get_clk_context(clk->name); + unsigned long frqcr = ctrl_inl(FRQCR); + int index; + + index = (frqcr >> ctx.shift) & ctx.mask; + clk->rate = clk->parent->rate * 2 / divisors2[index]; +} + +static int sh7722_frqcr_set_rate(struct clk *clk, unsigned long rate, + int algo_id) +{ + struct frqcr_context ctx = sh7722_get_clk_context(clk->name); + unsigned long parent_rate = clk->parent->rate; + int div; + unsigned long frqcr; + int err = 0; + + /* pretty invalid */ + if (parent_rate < rate) + return -EINVAL; + + /* look for multiplier/divisor pair */ + div = sh7722_find_divisors(parent_rate, rate); + if (div<0) + return div; + + /* calculate new value of clock rate */ + clk->rate = parent_rate * 2 / div; + frqcr = ctrl_inl(FRQCR); + + /* FIXME: adjust as algo_id specifies */ + if (algo_id != NO_CHANGE) { + int originator; + char *algo_group_1[] = { "cpu_clk", "umem_clk", "sh_clk" }; + char *algo_group_2[] = { "sh_clk", "bus_clk" }; + char *algo_group_3[] = { "sh_clk", "sdram_clk" }; + char *algo_group_4[] = { "bus_clk", "peripheral_clk" }; + char *algo_group_5[] = { "cpu_clk", "peripheral_clk" }; + char **algo_current = NULL; + /* 3 is the maximum number of clocks in relation */ + struct clk *ck[3]; + unsigned long values[3]; /* the same comment as above */ + int part_length = -1; + int i; + + /* + * all the steps below only required if adjustion was + * requested + */ + if (algo_id == IUS_N1_N1 || + algo_id == IUS_322 || + algo_id == IUS_522 || + algo_id == IUS_N11) { + algo_current = algo_group_1; + part_length = 3; + } + if (algo_id == SB_N1) { + algo_current = algo_group_2; + part_length = 2; + } + if (algo_id == SB3_N1 || + algo_id == SB3_32 || + algo_id == SB3_43 || + algo_id == SB3_54) { + algo_current = algo_group_3; + part_length = 2; + } + if (algo_id == BP_N1) { + algo_current = algo_group_4; + part_length = 2; + } + if (algo_id == IP_N1) { + algo_current = algo_group_5; + part_length = 2; + } + if (!algo_current) + goto incorrect_algo_id; + + originator = -1; + for (i = 0; i < part_length; i ++ ) { + if (originator >= 0 && !strcmp(clk->name, + algo_current[i])) + originator = i; + ck[i] = clk_get(NULL, algo_current[i]); + values[i] = clk_get_rate(ck[i]); + } + + if (originator >= 0) + adjust_clocks(originator, adjust_algos[algo_id], + values, part_length); + + for (i = 0; i < part_length; i ++ ) { + struct frqcr_context part_ctx; + int part_div; + + if (likely(!err)) { + part_div = sh7722_find_divisors(parent_rate, + rate); + if (part_div > 0) { + part_ctx = sh7722_get_clk_context( + ck[i]->name); + frqcr &= ~(part_ctx.mask << + part_ctx.shift); + frqcr |= part_div << part_ctx.shift; + } else + err = part_div; + } + + ck[i]->ops->recalc(ck[i]); + clk_put(ck[i]); + } + } + + /* was there any error during recalculation ? If so, bail out.. */ + if (unlikely(err!=0)) + goto out_err; + + /* clear FRQCR bits */ + frqcr &= ~(ctx.mask << ctx.shift); + frqcr |= div << ctx.shift; + + /* ...and perform actual change */ + ctrl_outl(frqcr, FRQCR); + return 0; + +incorrect_algo_id: + return -EINVAL; +out_err: + return err; +} + +static struct clk_ops sh7722_frqcr_clk_ops = { + .recalc = sh7722_frqcr_recalc, + .set_rate = sh7722_frqcr_set_rate, +}; + +/* + * clock ops methods for SIU A/B and IrDA clock + * + */ +static int sh7722_siu_which(struct clk *clk) +{ + if (!strcmp(clk->name, "siu_a_clk")) + return 0; + if (!strcmp(clk->name, "siu_b_clk")) + return 1; + if (!strcmp(clk->name, "irda_clk")) + return 2; + return -EINVAL; +} + +static unsigned long sh7722_siu_regs[] = { + [0] = SCLKACR, + [1] = SCLKBCR, + [2] = IrDACLKCR, +}; + +static int sh7722_siu_start_stop(struct clk *clk, int enable) +{ + int siu = sh7722_siu_which(clk); + unsigned long r; + + if (siu < 0) + return siu; + BUG_ON(siu > 2); + r = ctrl_inl(sh7722_siu_regs[siu]); + if (enable) + ctrl_outl(r & ~(1 << 8), sh7722_siu_regs[siu]); + else + ctrl_outl(r | (1 << 8), sh7722_siu_regs[siu]); + return 0; +} + +static void sh7722_siu_enable(struct clk *clk) +{ + sh7722_siu_start_stop(clk, 1); +} + +static void sh7722_siu_disable(struct clk *clk) +{ + sh7722_siu_start_stop(clk, 0); +} + +static void sh7722_video_enable(struct clk *clk) +{ + unsigned long r; + + r = ctrl_inl(VCLKCR); + ctrl_outl( r & ~(1<<8), VCLKCR); +} + +static void sh7722_video_disable(struct clk *clk) +{ + unsigned long r; + + r = ctrl_inl(VCLKCR); + ctrl_outl( r | (1<<8), VCLKCR); +} + +static int sh7722_video_set_rate(struct clk *clk, unsigned long rate, + int algo_id) +{ + unsigned long r; + + r = ctrl_inl(VCLKCR); + r &= ~0x3F; + r |= ((clk->parent->rate / rate - 1) & 0x3F); + ctrl_outl(r, VCLKCR); + return 0; +} + +static void sh7722_video_recalc(struct clk *clk) +{ + unsigned long r; + + r = ctrl_inl(VCLKCR); + clk->rate = clk->parent->rate / ((r & 0x3F) + 1); +} + +static int sh7722_siu_set_rate(struct clk *clk, unsigned long rate, int algo_id) +{ + int siu = sh7722_siu_which(clk); + unsigned long r; + int div; + + if (siu < 0) + return siu; + BUG_ON(siu > 2); + r = ctrl_inl(sh7722_siu_regs[siu]); + div = sh7722_find_divisors(clk->parent->rate, rate); + if (div < 0) + return div; + r = (r & ~0xF) | div; + ctrl_outl(r, sh7722_siu_regs[siu]); + return 0; +} + +static void sh7722_siu_recalc(struct clk *clk) +{ + int siu = sh7722_siu_which(clk); + unsigned long r; + + if (siu < 0) + return /* siu */ ; + BUG_ON(siu > 1); + r = ctrl_inl(sh7722_siu_regs[siu]); + clk->rate = clk->parent->rate * 2 / divisors2[r & 0xF]; +} + +static struct clk_ops sh7722_siu_clk_ops = { + .recalc = sh7722_siu_recalc, + .set_rate = sh7722_siu_set_rate, + .enable = sh7722_siu_enable, + .disable = sh7722_siu_disable, +}; + +static struct clk_ops sh7722_video_clk_ops = { + .recalc = sh7722_video_recalc, + .set_rate = sh7722_video_set_rate, + .enable = sh7722_video_enable, + .disable = sh7722_video_disable, +}; +/* + * and at last, clock definitions themselves + */ +static struct clk sh7722_umem_clock = { + .name = "umem_clk", + .ops = &sh7722_frqcr_clk_ops, +}; + +static struct clk sh7722_sh_clock = { + .name = "sh_clk", + .ops = &sh7722_frqcr_clk_ops, +}; + +static struct clk sh7722_peripheral_clock = { + .name = "peripheral_clk", + .ops = &sh7722_frqcr_clk_ops, +}; + +static struct clk sh7722_sdram_clock = { + .name = "sdram_clk", + .ops = &sh7722_frqcr_clk_ops, +}; + +/* + * these three clocks - SIU A, SIU B, IrDA - share the same clk_ops + * methods of clk_ops determine which register they should access by + * examining clk->name field + */ +static struct clk sh7722_siu_a_clock = { + .name = "siu_a_clk", + .ops = &sh7722_siu_clk_ops, +}; + +static struct clk sh7722_siu_b_clock = { + .name = "siu_b_clk", + .ops = &sh7722_siu_clk_ops, +}; + +static struct clk sh7722_irda_clock = { + .name = "irda_clk", + .ops = &sh7722_siu_clk_ops, +}; + +static struct clk sh7722_video_clock = { + .name = "video_clk", + .ops = &sh7722_video_clk_ops, +}; + +static struct clk *sh7722_clocks[] = { + &sh7722_umem_clock, + &sh7722_sh_clock, + &sh7722_peripheral_clock, + &sh7722_sdram_clock, + &sh7722_siu_a_clock, + &sh7722_siu_b_clock, + &sh7722_irda_clock, + &sh7722_video_clock, +}; + +/* + * init in order: master, module, bus, cpu + */ +struct clk_ops *onchip_ops[] = { + &sh7722_master_clk_ops, + &sh7722_frqcr_clk_ops, + &sh7722_frqcr_clk_ops, + &sh7722_frqcr_clk_ops, +}; + +void __init +arch_init_clk_ops(struct clk_ops **ops, int type) +{ + BUG_ON(type < 0 || type > ARRAY_SIZE(onchip_ops)); + *ops = onchip_ops[type]; +} + +int __init sh7722_clock_init(void) +{ + struct clk *master; + int i; + + master = clk_get(NULL, "master_clk"); + for (i = 0; i < ARRAY_SIZE(sh7722_clocks); i++) { + pr_debug( "Registering clock '%s'\n", sh7722_clocks[i]->name); + sh7722_clocks[i]->parent = master; + clk_register(sh7722_clocks[i]); + } + clk_put(master); + return 0; +} +arch_initcall(sh7722_clock_init); |