diff options
Diffstat (limited to 'arch/sh/kernel/cpu')
-rw-r--r-- | arch/sh/kernel/cpu/Makefile | 2 | ||||
-rw-r--r-- | arch/sh/kernel/cpu/clock-cpg.c | 94 | ||||
-rw-r--r-- | arch/sh/kernel/cpu/fpu.c | 84 | ||||
-rw-r--r-- | arch/sh/kernel/cpu/init.c | 98 | ||||
-rw-r--r-- | arch/sh/kernel/cpu/sh2a/fpu.c | 111 | ||||
-rw-r--r-- | arch/sh/kernel/cpu/sh3/ex.S | 2 | ||||
-rw-r--r-- | arch/sh/kernel/cpu/sh4/fpu.c | 159 | ||||
-rw-r--r-- | arch/sh/kernel/cpu/sh4/probe.c | 8 | ||||
-rw-r--r-- | arch/sh/kernel/cpu/sh4a/Makefile | 9 | ||||
-rw-r--r-- | arch/sh/kernel/cpu/sh4a/clock-sh7722.c | 25 | ||||
-rw-r--r-- | arch/sh/kernel/cpu/sh4a/clock-sh7786.c | 184 | ||||
-rw-r--r-- | arch/sh/kernel/cpu/sh4a/smp-shx3.c | 5 | ||||
-rw-r--r-- | arch/sh/kernel/cpu/sh4a/ubc.c | 133 | ||||
-rw-r--r-- | arch/sh/kernel/cpu/sh5/entry.S | 6 | ||||
-rw-r--r-- | arch/sh/kernel/cpu/sh5/fpu.c | 65 |
15 files changed, 560 insertions, 425 deletions
diff --git a/arch/sh/kernel/cpu/Makefile b/arch/sh/kernel/cpu/Makefile index d97c803719ec..0e48bc61c272 100644 --- a/arch/sh/kernel/cpu/Makefile +++ b/arch/sh/kernel/cpu/Makefile @@ -17,5 +17,7 @@ obj-$(CONFIG_ARCH_SHMOBILE) += shmobile/ obj-$(CONFIG_SH_ADC) += adc.o obj-$(CONFIG_SH_CLK_CPG) += clock-cpg.o +obj-$(CONFIG_SH_FPU) += fpu.o +obj-$(CONFIG_SH_FPU_EMU) += fpu.o obj-y += irq/ init.o clock.o hwblk.o diff --git a/arch/sh/kernel/cpu/clock-cpg.c b/arch/sh/kernel/cpu/clock-cpg.c index 6dfe2cced3fc..2827abb5d2ab 100644 --- a/arch/sh/kernel/cpu/clock-cpg.c +++ b/arch/sh/kernel/cpu/clock-cpg.c @@ -160,13 +160,81 @@ static unsigned long sh_clk_div4_recalc(struct clk *clk) return clk->freq_table[idx].frequency; } +static int sh_clk_div4_set_parent(struct clk *clk, struct clk *parent) +{ + struct clk_div_mult_table *table = clk->priv; + u32 value; + int ret; + + if (!strcmp("pll_clk", parent->name)) + value = __raw_readl(clk->enable_reg) & ~(1 << 7); + else + value = __raw_readl(clk->enable_reg) | (1 << 7); + + ret = clk_reparent(clk, parent); + if (ret < 0) + return ret; + + __raw_writel(value, clk->enable_reg); + + /* Rebiuld the frequency table */ + clk_rate_table_build(clk, clk->freq_table, table->nr_divisors, + table, &clk->arch_flags); + + return 0; +} + +static int sh_clk_div4_set_rate(struct clk *clk, unsigned long rate, int algo_id) +{ + unsigned long value; + int idx = clk_rate_table_find(clk, clk->freq_table, rate); + if (idx < 0) + return idx; + + value = __raw_readl(clk->enable_reg); + value &= ~0xf; + value |= idx; + __raw_writel(value, clk->enable_reg); + + return 0; +} + +static int sh_clk_div4_enable(struct clk *clk) +{ + __raw_writel(__raw_readl(clk->enable_reg) & ~(1 << 8), clk->enable_reg); + return 0; +} + +static void sh_clk_div4_disable(struct clk *clk) +{ + __raw_writel(__raw_readl(clk->enable_reg) | (1 << 8), clk->enable_reg); +} + static struct clk_ops sh_clk_div4_clk_ops = { .recalc = sh_clk_div4_recalc, + .set_rate = sh_clk_div4_set_rate, .round_rate = sh_clk_div_round_rate, }; -int __init sh_clk_div4_register(struct clk *clks, int nr, - struct clk_div_mult_table *table) +static struct clk_ops sh_clk_div4_enable_clk_ops = { + .recalc = sh_clk_div4_recalc, + .set_rate = sh_clk_div4_set_rate, + .round_rate = sh_clk_div_round_rate, + .enable = sh_clk_div4_enable, + .disable = sh_clk_div4_disable, +}; + +static struct clk_ops sh_clk_div4_reparent_clk_ops = { + .recalc = sh_clk_div4_recalc, + .set_rate = sh_clk_div4_set_rate, + .round_rate = sh_clk_div_round_rate, + .enable = sh_clk_div4_enable, + .disable = sh_clk_div4_disable, + .set_parent = sh_clk_div4_set_parent, +}; + +static int __init sh_clk_div4_register_ops(struct clk *clks, int nr, + struct clk_div_mult_table *table, struct clk_ops *ops) { struct clk *clkp; void *freq_table; @@ -185,7 +253,7 @@ int __init sh_clk_div4_register(struct clk *clks, int nr, for (k = 0; !ret && (k < nr); k++) { clkp = clks + k; - clkp->ops = &sh_clk_div4_clk_ops; + clkp->ops = ops; clkp->id = -1; clkp->priv = table; @@ -198,6 +266,26 @@ int __init sh_clk_div4_register(struct clk *clks, int nr, return ret; } +int __init sh_clk_div4_register(struct clk *clks, int nr, + struct clk_div_mult_table *table) +{ + return sh_clk_div4_register_ops(clks, nr, table, &sh_clk_div4_clk_ops); +} + +int __init sh_clk_div4_enable_register(struct clk *clks, int nr, + struct clk_div_mult_table *table) +{ + return sh_clk_div4_register_ops(clks, nr, table, + &sh_clk_div4_enable_clk_ops); +} + +int __init sh_clk_div4_reparent_register(struct clk *clks, int nr, + struct clk_div_mult_table *table) +{ + return sh_clk_div4_register_ops(clks, nr, table, + &sh_clk_div4_reparent_clk_ops); +} + #ifdef CONFIG_SH_CLK_CPG_LEGACY static struct clk master_clk = { .name = "master_clk", diff --git a/arch/sh/kernel/cpu/fpu.c b/arch/sh/kernel/cpu/fpu.c new file mode 100644 index 000000000000..f059ed62cf57 --- /dev/null +++ b/arch/sh/kernel/cpu/fpu.c @@ -0,0 +1,84 @@ +#include <linux/sched.h> +#include <asm/processor.h> +#include <asm/fpu.h> + +int init_fpu(struct task_struct *tsk) +{ + if (tsk_used_math(tsk)) { + if ((boot_cpu_data.flags & CPU_HAS_FPU) && tsk == current) + unlazy_fpu(tsk, task_pt_regs(tsk)); + return 0; + } + + /* + * Memory allocation at the first usage of the FPU and other state. + */ + if (!tsk->thread.xstate) { + tsk->thread.xstate = kmem_cache_alloc(task_xstate_cachep, + GFP_KERNEL); + if (!tsk->thread.xstate) + return -ENOMEM; + } + + if (boot_cpu_data.flags & CPU_HAS_FPU) { + struct sh_fpu_hard_struct *fp = &tsk->thread.xstate->hardfpu; + memset(fp, 0, xstate_size); + fp->fpscr = FPSCR_INIT; + } else { + struct sh_fpu_soft_struct *fp = &tsk->thread.xstate->softfpu; + memset(fp, 0, xstate_size); + fp->fpscr = FPSCR_INIT; + } + + set_stopped_child_used_math(tsk); + return 0; +} + +#ifdef CONFIG_SH_FPU +void __fpu_state_restore(void) +{ + struct task_struct *tsk = current; + + restore_fpu(tsk); + + task_thread_info(tsk)->status |= TS_USEDFPU; + tsk->fpu_counter++; +} + +void fpu_state_restore(struct pt_regs *regs) +{ + struct task_struct *tsk = current; + + if (unlikely(!user_mode(regs))) { + printk(KERN_ERR "BUG: FPU is used in kernel mode.\n"); + BUG(); + return; + } + + if (!tsk_used_math(tsk)) { + local_irq_enable(); + /* + * does a slab alloc which can sleep + */ + if (init_fpu(tsk)) { + /* + * ran out of memory! + */ + do_group_exit(SIGKILL); + return; + } + local_irq_disable(); + } + + grab_fpu(regs); + + __fpu_state_restore(); +} + +BUILD_TRAP_HANDLER(fpu_state_restore) +{ + TRAP_HANDLER_DECL; + + fpu_state_restore(regs); +} +#endif /* CONFIG_SH_FPU */ diff --git a/arch/sh/kernel/cpu/init.c b/arch/sh/kernel/cpu/init.c index 89b4b76c0d76..a5bb0550bbf3 100644 --- a/arch/sh/kernel/cpu/init.c +++ b/arch/sh/kernel/cpu/init.c @@ -24,22 +24,31 @@ #include <asm/elf.h> #include <asm/io.h> #include <asm/smp.h> -#ifdef CONFIG_SUPERH32 -#include <asm/ubc.h> + +#ifdef CONFIG_SH_FPU +#define cpu_has_fpu 1 +#else +#define cpu_has_fpu 0 +#endif + +#ifdef CONFIG_SH_DSP +#define cpu_has_dsp 1 +#else +#define cpu_has_dsp 0 #endif /* * Generic wrapper for command line arguments to disable on-chip * peripherals (nofpu, nodsp, and so forth). */ -#define onchip_setup(x) \ -static int x##_disabled __initdata = 0; \ - \ -static int __init x##_setup(char *opts) \ -{ \ - x##_disabled = 1; \ - return 1; \ -} \ +#define onchip_setup(x) \ +static int x##_disabled __initdata = !cpu_has_##x; \ + \ +static int __init x##_setup(char *opts) \ +{ \ + x##_disabled = 1; \ + return 1; \ +} \ __setup("no" __stringify(x), x##_setup); onchip_setup(fpu); @@ -207,6 +216,18 @@ static void detect_cache_shape(void) l2_cache_shape = -1; /* No S-cache */ } +static void __init fpu_init(void) +{ + /* Disable the FPU */ + if (fpu_disabled && (current_cpu_data.flags & CPU_HAS_FPU)) { + printk("FPU Disabled\n"); + current_cpu_data.flags &= ~CPU_HAS_FPU; + } + + disable_fpu(); + clear_used_math(); +} + #ifdef CONFIG_SH_DSP static void __init release_dsp(void) { @@ -244,28 +265,35 @@ static void __init dsp_init(void) if (sr & SR_DSP) current_cpu_data.flags |= CPU_HAS_DSP; + /* Disable the DSP */ + if (dsp_disabled && (current_cpu_data.flags & CPU_HAS_DSP)) { + printk("DSP Disabled\n"); + current_cpu_data.flags &= ~CPU_HAS_DSP; + } + /* Now that we've determined the DSP status, clear the DSP bit. */ release_dsp(); } +#else +static inline void __init dsp_init(void) { } #endif /* CONFIG_SH_DSP */ /** * sh_cpu_init * - * This is our initial entry point for each CPU, and is invoked on the boot - * CPU prior to calling start_kernel(). For SMP, a combination of this and - * start_secondary() will bring up each processor to a ready state prior - * to hand forking the idle loop. + * This is our initial entry point for each CPU, and is invoked on the + * boot CPU prior to calling start_kernel(). For SMP, a combination of + * this and start_secondary() will bring up each processor to a ready + * state prior to hand forking the idle loop. * - * We do all of the basic processor init here, including setting up the - * caches, FPU, DSP, kicking the UBC, etc. By the time start_kernel() is - * hit (and subsequently platform_setup()) things like determining the - * CPU subtype and initial configuration will all be done. + * We do all of the basic processor init here, including setting up + * the caches, FPU, DSP, etc. By the time start_kernel() is hit (and + * subsequently platform_setup()) things like determining the CPU + * subtype and initial configuration will all be done. * * Each processor family is still responsible for doing its own probing * and cache configuration in detect_cpu_and_cache_system(). */ - asmlinkage void __init sh_cpu_init(void) { current_thread_info()->cpu = hard_smp_processor_id(); @@ -302,18 +330,8 @@ asmlinkage void __init sh_cpu_init(void) detect_cache_shape(); } - /* Disable the FPU */ - if (fpu_disabled) { - printk("FPU Disabled\n"); - current_cpu_data.flags &= ~CPU_HAS_FPU; - } - - /* FPU initialization */ - disable_fpu(); - if ((current_cpu_data.flags & CPU_HAS_FPU)) { - current_thread_info()->status &= ~TS_USEDFPU; - clear_used_math(); - } + fpu_init(); + dsp_init(); /* * Initialize the per-CPU ASID cache very early, since the @@ -321,18 +339,12 @@ asmlinkage void __init sh_cpu_init(void) */ current_cpu_data.asid_cache = NO_CONTEXT; -#ifdef CONFIG_SH_DSP - /* Probe for DSP */ - dsp_init(); - - /* Disable the DSP */ - if (dsp_disabled) { - printk("DSP Disabled\n"); - current_cpu_data.flags &= ~CPU_HAS_DSP; - release_dsp(); - } -#endif - speculative_execution_init(); expmask_init(); + + /* + * Boot processor to setup the FP and extended state context info. + */ + if (raw_smp_processor_id() == 0) + init_thread_xstate(); } diff --git a/arch/sh/kernel/cpu/sh2a/fpu.c b/arch/sh/kernel/cpu/sh2a/fpu.c index d395ce5740e7..488d24e0cdf0 100644 --- a/arch/sh/kernel/cpu/sh2a/fpu.c +++ b/arch/sh/kernel/cpu/sh2a/fpu.c @@ -26,8 +26,7 @@ /* * Save FPU registers onto task structure. */ -void -save_fpu(struct task_struct *tsk) +void save_fpu(struct task_struct *tsk) { unsigned long dummy; @@ -52,7 +51,7 @@ save_fpu(struct task_struct *tsk) "fmov.s fr0, @-%0\n\t" "lds %3, fpscr\n\t" : "=r" (dummy) - : "0" ((char *)(&tsk->thread.fpu.hard.status)), + : "0" ((char *)(&tsk->thread.xstate->hardfpu.status)), "r" (FPSCR_RCHG), "r" (FPSCR_INIT) : "memory"); @@ -60,8 +59,7 @@ save_fpu(struct task_struct *tsk) disable_fpu(); } -static void -restore_fpu(struct task_struct *tsk) +void restore_fpu(struct task_struct *tsk) { unsigned long dummy; @@ -85,45 +83,12 @@ restore_fpu(struct task_struct *tsk) "lds.l @%0+, fpscr\n\t" "lds.l @%0+, fpul\n\t" : "=r" (dummy) - : "0" (&tsk->thread.fpu), "r" (FPSCR_RCHG) + : "0" (tsk->thread.xstate), "r" (FPSCR_RCHG) : "memory"); disable_fpu(); } /* - * Load the FPU with signalling NANS. This bit pattern we're using - * has the property that no matter wether considered as single or as - * double precission represents signaling NANS. - */ - -static void -fpu_init(void) -{ - enable_fpu(); - asm volatile("lds %0, fpul\n\t" - "fsts fpul, fr0\n\t" - "fsts fpul, fr1\n\t" - "fsts fpul, fr2\n\t" - "fsts fpul, fr3\n\t" - "fsts fpul, fr4\n\t" - "fsts fpul, fr5\n\t" - "fsts fpul, fr6\n\t" - "fsts fpul, fr7\n\t" - "fsts fpul, fr8\n\t" - "fsts fpul, fr9\n\t" - "fsts fpul, fr10\n\t" - "fsts fpul, fr11\n\t" - "fsts fpul, fr12\n\t" - "fsts fpul, fr13\n\t" - "fsts fpul, fr14\n\t" - "fsts fpul, fr15\n\t" - "lds %2, fpscr\n\t" - : /* no output */ - : "r" (0), "r" (FPSCR_RCHG), "r" (FPSCR_INIT)); - disable_fpu(); -} - -/* * Emulate arithmetic ops on denormalized number for some FPU insns. */ @@ -490,9 +455,9 @@ ieee_fpe_handler (struct pt_regs *regs) if ((finsn & 0xf1ff) == 0xf0ad) { /* fcnvsd */ struct task_struct *tsk = current; - if ((tsk->thread.fpu.hard.fpscr & FPSCR_FPU_ERROR)) { + if ((tsk->thread.xstate->hardfpu.fpscr & FPSCR_FPU_ERROR)) { /* FPU error */ - denormal_to_double (&tsk->thread.fpu.hard, + denormal_to_double (&tsk->thread.xstate->hardfpu, (finsn >> 8) & 0xf); } else return 0; @@ -507,9 +472,9 @@ ieee_fpe_handler (struct pt_regs *regs) n = (finsn >> 8) & 0xf; m = (finsn >> 4) & 0xf; - hx = tsk->thread.fpu.hard.fp_regs[n]; - hy = tsk->thread.fpu.hard.fp_regs[m]; - fpscr = tsk->thread.fpu.hard.fpscr; + hx = tsk->thread.xstate->hardfpu.fp_regs[n]; + hy = tsk->thread.xstate->hardfpu.fp_regs[m]; + fpscr = tsk->thread.xstate->hardfpu.fpscr; prec = fpscr & (1 << 19); if ((fpscr & FPSCR_FPU_ERROR) @@ -519,15 +484,15 @@ ieee_fpe_handler (struct pt_regs *regs) /* FPU error because of denormal */ llx = ((long long) hx << 32) - | tsk->thread.fpu.hard.fp_regs[n+1]; + | tsk->thread.xstate->hardfpu.fp_regs[n+1]; lly = ((long long) hy << 32) - | tsk->thread.fpu.hard.fp_regs[m+1]; + | tsk->thread.xstate->hardfpu.fp_regs[m+1]; if ((hx & 0x7fffffff) >= 0x00100000) llx = denormal_muld(lly, llx); else llx = denormal_muld(llx, lly); - tsk->thread.fpu.hard.fp_regs[n] = llx >> 32; - tsk->thread.fpu.hard.fp_regs[n+1] = llx & 0xffffffff; + tsk->thread.xstate->hardfpu.fp_regs[n] = llx >> 32; + tsk->thread.xstate->hardfpu.fp_regs[n+1] = llx & 0xffffffff; } else if ((fpscr & FPSCR_FPU_ERROR) && (!prec && ((hx & 0x7fffffff) < 0x00800000 || (hy & 0x7fffffff) < 0x00800000))) { @@ -536,7 +501,7 @@ ieee_fpe_handler (struct pt_regs *regs) hx = denormal_mulf(hy, hx); else hx = denormal_mulf(hx, hy); - tsk->thread.fpu.hard.fp_regs[n] = hx; + tsk->thread.xstate->hardfpu.fp_regs[n] = hx; } else return 0; @@ -550,9 +515,9 @@ ieee_fpe_handler (struct pt_regs *regs) n = (finsn >> 8) & 0xf; m = (finsn >> 4) & 0xf; - hx = tsk->thread.fpu.hard.fp_regs[n]; - hy = tsk->thread.fpu.hard.fp_regs[m]; - fpscr = tsk->thread.fpu.hard.fpscr; + hx = tsk->thread.xstate->hardfpu.fp_regs[n]; + hy = tsk->thread.xstate->hardfpu.fp_regs[m]; + fpscr = tsk->thread.xstate->hardfpu.fpscr; prec = fpscr & (1 << 19); if ((fpscr & FPSCR_FPU_ERROR) @@ -562,15 +527,15 @@ ieee_fpe_handler (struct pt_regs *regs) /* FPU error because of denormal */ llx = ((long long) hx << 32) - | tsk->thread.fpu.hard.fp_regs[n+1]; + | tsk->thread.xstate->hardfpu.fp_regs[n+1]; lly = ((long long) hy << 32) - | tsk->thread.fpu.hard.fp_regs[m+1]; + | tsk->thread.xstate->hardfpu.fp_regs[m+1]; if ((finsn & 0xf00f) == 0xf000) llx = denormal_addd(llx, lly); else llx = denormal_addd(llx, lly ^ (1LL << 63)); - tsk->thread.fpu.hard.fp_regs[n] = llx >> 32; - tsk->thread.fpu.hard.fp_regs[n+1] = llx & 0xffffffff; + tsk->thread.xstate->hardfpu.fp_regs[n] = llx >> 32; + tsk->thread.xstate->hardfpu.fp_regs[n+1] = llx & 0xffffffff; } else if ((fpscr & FPSCR_FPU_ERROR) && (!prec && ((hx & 0x7fffffff) < 0x00800000 || (hy & 0x7fffffff) < 0x00800000))) { @@ -579,7 +544,7 @@ ieee_fpe_handler (struct pt_regs *regs) hx = denormal_addf(hx, hy); else hx = denormal_addf(hx, hy ^ 0x80000000); - tsk->thread.fpu.hard.fp_regs[n] = hx; + tsk->thread.xstate->hardfpu.fp_regs[n] = hx; } else return 0; @@ -597,7 +562,7 @@ BUILD_TRAP_HANDLER(fpu_error) __unlazy_fpu(tsk, regs); if (ieee_fpe_handler(regs)) { - tsk->thread.fpu.hard.fpscr &= + tsk->thread.xstate->hardfpu.fpscr &= ~(FPSCR_CAUSE_MASK | FPSCR_FLAG_MASK); grab_fpu(regs); restore_fpu(tsk); @@ -607,33 +572,3 @@ BUILD_TRAP_HANDLER(fpu_error) force_sig(SIGFPE, tsk); } - -void fpu_state_restore(struct pt_regs *regs) -{ - struct task_struct *tsk = current; - - grab_fpu(regs); - if (unlikely(!user_mode(regs))) { - printk(KERN_ERR "BUG: FPU is used in kernel mode.\n"); - BUG(); - return; - } - - if (likely(used_math())) { - /* Using the FPU again. */ - restore_fpu(tsk); - } else { - /* First time FPU user. */ - fpu_init(); - set_used_math(); - } - task_thread_info(tsk)->status |= TS_USEDFPU; - tsk->fpu_counter++; -} - -BUILD_TRAP_HANDLER(fpu_state_restore) -{ - TRAP_HANDLER_DECL; - - fpu_state_restore(regs); -} diff --git a/arch/sh/kernel/cpu/sh3/ex.S b/arch/sh/kernel/cpu/sh3/ex.S index 46610c35c232..99b4d020179a 100644 --- a/arch/sh/kernel/cpu/sh3/ex.S +++ b/arch/sh/kernel/cpu/sh3/ex.S @@ -49,7 +49,7 @@ ENTRY(exception_handling_table) .long exception_error ! reserved_instruction (filled by trap_init) /* 180 */ .long exception_error ! illegal_slot_instruction (filled by trap_init) /*1A0*/ .long nmi_trap_handler /* 1C0 */ ! Allow trap to debugger - .long break_point_trap /* 1E0 */ + .long breakpoint_trap_handler /* 1E0 */ /* * Pad the remainder of the table out, exceptions residing in far diff --git a/arch/sh/kernel/cpu/sh4/fpu.c b/arch/sh/kernel/cpu/sh4/fpu.c index e97857aec8a0..447482d7f65e 100644 --- a/arch/sh/kernel/cpu/sh4/fpu.c +++ b/arch/sh/kernel/cpu/sh4/fpu.c @@ -85,14 +85,14 @@ void save_fpu(struct task_struct *tsk) "fmov.s fr1, @-%0\n\t" "fmov.s fr0, @-%0\n\t" "lds %3, fpscr\n\t":"=r" (dummy) - :"0"((char *)(&tsk->thread.fpu.hard.status)), + :"0"((char *)(&tsk->thread.xstate->hardfpu.status)), "r"(FPSCR_RCHG), "r"(FPSCR_INIT) :"memory"); disable_fpu(); } -static void restore_fpu(struct task_struct *tsk) +void restore_fpu(struct task_struct *tsk) { unsigned long dummy; @@ -135,62 +135,11 @@ static void restore_fpu(struct task_struct *tsk) "lds.l @%0+, fpscr\n\t" "lds.l @%0+, fpul\n\t" :"=r" (dummy) - :"0"(&tsk->thread.fpu), "r"(FPSCR_RCHG) + :"0" (tsk->thread.xstate), "r" (FPSCR_RCHG) :"memory"); disable_fpu(); } -/* - * Load the FPU with signalling NANS. This bit pattern we're using - * has the property that no matter wether considered as single or as - * double precision represents signaling NANS. - */ - -static void fpu_init(void) -{ - enable_fpu(); - asm volatile ( "lds %0, fpul\n\t" - "lds %1, fpscr\n\t" - "fsts fpul, fr0\n\t" - "fsts fpul, fr1\n\t" - "fsts fpul, fr2\n\t" - "fsts fpul, fr3\n\t" - "fsts fpul, fr4\n\t" - "fsts fpul, fr5\n\t" - "fsts fpul, fr6\n\t" - "fsts fpul, fr7\n\t" - "fsts fpul, fr8\n\t" - "fsts fpul, fr9\n\t" - "fsts fpul, fr10\n\t" - "fsts fpul, fr11\n\t" - "fsts fpul, fr12\n\t" - "fsts fpul, fr13\n\t" - "fsts fpul, fr14\n\t" - "fsts fpul, fr15\n\t" - "frchg\n\t" - "fsts fpul, fr0\n\t" - "fsts fpul, fr1\n\t" - "fsts fpul, fr2\n\t" - "fsts fpul, fr3\n\t" - "fsts fpul, fr4\n\t" - "fsts fpul, fr5\n\t" - "fsts fpul, fr6\n\t" - "fsts fpul, fr7\n\t" - "fsts fpul, fr8\n\t" - "fsts fpul, fr9\n\t" - "fsts fpul, fr10\n\t" - "fsts fpul, fr11\n\t" - "fsts fpul, fr12\n\t" - "fsts fpul, fr13\n\t" - "fsts fpul, fr14\n\t" - "fsts fpul, fr15\n\t" - "frchg\n\t" - "lds %2, fpscr\n\t" - : /* no output */ - :"r" (0), "r"(FPSCR_RCHG), "r"(FPSCR_INIT)); - disable_fpu(); -} - /** * denormal_to_double - Given denormalized float number, * store double float @@ -282,9 +231,9 @@ static int ieee_fpe_handler(struct pt_regs *regs) /* fcnvsd */ struct task_struct *tsk = current; - if ((tsk->thread.fpu.hard.fpscr & FPSCR_CAUSE_ERROR)) + if ((tsk->thread.xstate->hardfpu.fpscr & FPSCR_CAUSE_ERROR)) /* FPU error */ - denormal_to_double(&tsk->thread.fpu.hard, + denormal_to_double(&tsk->thread.xstate->hardfpu, (finsn >> 8) & 0xf); else return 0; @@ -300,9 +249,9 @@ static int ieee_fpe_handler(struct pt_regs *regs) n = (finsn >> 8) & 0xf; m = (finsn >> 4) & 0xf; - hx = tsk->thread.fpu.hard.fp_regs[n]; - hy = tsk->thread.fpu.hard.fp_regs[m]; - fpscr = tsk->thread.fpu.hard.fpscr; + hx = tsk->thread.xstate->hardfpu.fp_regs[n]; + hy = tsk->thread.xstate->hardfpu.fp_regs[m]; + fpscr = tsk->thread.xstate->hardfpu.fpscr; prec = fpscr & FPSCR_DBL_PRECISION; if ((fpscr & FPSCR_CAUSE_ERROR) @@ -312,18 +261,18 @@ static int ieee_fpe_handler(struct pt_regs *regs) /* FPU error because of denormal (doubles) */ llx = ((long long)hx << 32) - | tsk->thread.fpu.hard.fp_regs[n + 1]; + | tsk->thread.xstate->hardfpu.fp_regs[n + 1]; lly = ((long long)hy << 32) - | tsk->thread.fpu.hard.fp_regs[m + 1]; + | tsk->thread.xstate->hardfpu.fp_regs[m + 1]; llx = float64_mul(llx, lly); - tsk->thread.fpu.hard.fp_regs[n] = llx >> 32; - tsk->thread.fpu.hard.fp_regs[n + 1] = llx & 0xffffffff; + tsk->thread.xstate->hardfpu.fp_regs[n] = llx >> 32; + tsk->thread.xstate->hardfpu.fp_regs[n + 1] = llx & 0xffffffff; } else if ((fpscr & FPSCR_CAUSE_ERROR) && (!prec && ((hx & 0x7fffffff) < 0x00800000 || (hy & 0x7fffffff) < 0x00800000))) { /* FPU error because of denormal (floats) */ hx = float32_mul(hx, hy); - tsk->thread.fpu.hard.fp_regs[n] = hx; + tsk->thread.xstate->hardfpu.fp_regs[n] = hx; } else return 0; @@ -338,9 +287,9 @@ static int ieee_fpe_handler(struct pt_regs *regs) n = (finsn >> 8) & 0xf; m = (finsn >> 4) & 0xf; - hx = tsk->thread.fpu.hard.fp_regs[n]; - hy = tsk->thread.fpu.hard.fp_regs[m]; - fpscr = tsk->thread.fpu.hard.fpscr; + hx = tsk->thread.xstate->hardfpu.fp_regs[n]; + hy = tsk->thread.xstate->hardfpu.fp_regs[m]; + fpscr = tsk->thread.xstate->hardfpu.fpscr; prec = fpscr & FPSCR_DBL_PRECISION; if ((fpscr & FPSCR_CAUSE_ERROR) @@ -350,15 +299,15 @@ static int ieee_fpe_handler(struct pt_regs *regs) /* FPU error because of denormal (doubles) */ llx = ((long long)hx << 32) - | tsk->thread.fpu.hard.fp_regs[n + 1]; + | tsk->thread.xstate->hardfpu.fp_regs[n + 1]; lly = ((long long)hy << 32) - | tsk->thread.fpu.hard.fp_regs[m + 1]; + | tsk->thread.xstate->hardfpu.fp_regs[m + 1]; if ((finsn & 0xf00f) == 0xf000) llx = float64_add(llx, lly); else llx = float64_sub(llx, lly); - tsk->thread.fpu.hard.fp_regs[n] = llx >> 32; - tsk->thread.fpu.hard.fp_regs[n + 1] = llx & 0xffffffff; + tsk->thread.xstate->hardfpu.fp_regs[n] = llx >> 32; + tsk->thread.xstate->hardfpu.fp_regs[n + 1] = llx & 0xffffffff; } else if ((fpscr & FPSCR_CAUSE_ERROR) && (!prec && ((hx & 0x7fffffff) < 0x00800000 || (hy & 0x7fffffff) < 0x00800000))) { @@ -367,7 +316,7 @@ static int ieee_fpe_handler(struct pt_regs *regs) hx = float32_add(hx, hy); else hx = float32_sub(hx, hy); - tsk->thread.fpu.hard.fp_regs[n] = hx; + tsk->thread.xstate->hardfpu.fp_regs[n] = hx; } else return 0; @@ -382,9 +331,9 @@ static int ieee_fpe_handler(struct pt_regs *regs) n = (finsn >> 8) & 0xf; m = (finsn >> 4) & 0xf; - hx = tsk->thread.fpu.hard.fp_regs[n]; - hy = tsk->thread.fpu.hard.fp_regs[m]; - fpscr = tsk->thread.fpu.hard.fpscr; + hx = tsk->thread.xstate->hardfpu.fp_regs[n]; + hy = tsk->thread.xstate->hardfpu.fp_regs[m]; + fpscr = tsk->thread.xstate->hardfpu.fpscr; prec = fpscr & FPSCR_DBL_PRECISION; if ((fpscr & FPSCR_CAUSE_ERROR) @@ -394,20 +343,20 @@ static int ieee_fpe_handler(struct pt_regs *regs) /* FPU error because of denormal (doubles) */ llx = ((long long)hx << 32) - | tsk->thread.fpu.hard.fp_regs[n + 1]; + | tsk->thread.xstate->hardfpu.fp_regs[n + 1]; lly = ((long long)hy << 32) - | tsk->thread.fpu.hard.fp_regs[m + 1]; + | tsk->thread.xstate->hardfpu.fp_regs[m + 1]; llx = float64_div(llx, lly); - tsk->thread.fpu.hard.fp_regs[n] = llx >> 32; - tsk->thread.fpu.hard.fp_regs[n + 1] = llx & 0xffffffff; + tsk->thread.xstate->hardfpu.fp_regs[n] = llx >> 32; + tsk->thread.xstate->hardfpu.fp_regs[n + 1] = llx & 0xffffffff; } else if ((fpscr & FPSCR_CAUSE_ERROR) && (!prec && ((hx & 0x7fffffff) < 0x00800000 || (hy & 0x7fffffff) < 0x00800000))) { /* FPU error because of denormal (floats) */ hx = float32_div(hx, hy); - tsk->thread.fpu.hard.fp_regs[n] = hx; + tsk->thread.xstate->hardfpu.fp_regs[n] = hx; } else return 0; @@ -420,17 +369,17 @@ static int ieee_fpe_handler(struct pt_regs *regs) unsigned int hx; m = (finsn >> 8) & 0x7; - hx = tsk->thread.fpu.hard.fp_regs[m]; + hx = tsk->thread.xstate->hardfpu.fp_regs[m]; - if ((tsk->thread.fpu.hard.fpscr & FPSCR_CAUSE_ERROR) + if ((tsk->thread.xstate->hardfpu.fpscr & FPSCR_CAUSE_ERROR) && ((hx & 0x7fffffff) < 0x00100000)) { /* subnormal double to float conversion */ long long llx; - llx = ((long long)tsk->thread.fpu.hard.fp_regs[m] << 32) - | tsk->thread.fpu.hard.fp_regs[m + 1]; + llx = ((long long)tsk->thread.xstate->hardfpu.fp_regs[m] << 32) + | tsk->thread.xstate->hardfpu.fp_regs[m + 1]; - tsk->thread.fpu.hard.fpul = float64_to_float32(llx); + tsk->thread.xstate->hardfpu.fpul = float64_to_float32(llx); } else return 0; @@ -449,7 +398,7 @@ void float_raise(unsigned int flags) int float_rounding_mode(void) { struct task_struct *tsk = current; - int roundingMode = FPSCR_ROUNDING_MODE(tsk->thread.fpu.hard.fpscr); + int roundingMode = FPSCR_ROUNDING_MODE(tsk->thread.xstate->hardfpu.fpscr); return roundingMode; } @@ -461,16 +410,16 @@ BUILD_TRAP_HANDLER(fpu_error) __unlazy_fpu(tsk, regs); fpu_exception_flags = 0; if (ieee_fpe_handler(regs)) { - tsk->thread.fpu.hard.fpscr &= + tsk->thread.xstate->hardfpu.fpscr &= ~(FPSCR_CAUSE_MASK | FPSCR_FLAG_MASK); - tsk->thread.fpu.hard.fpscr |= fpu_exception_flags; + tsk->thread.xstate->hardfpu.fpscr |= fpu_exception_flags; /* Set the FPSCR flag as well as cause bits - simply * replicate the cause */ - tsk->thread.fpu.hard.fpscr |= (fpu_exception_flags >> 10); + tsk->thread.xstate->hardfpu.fpscr |= (fpu_exception_flags >> 10); grab_fpu(regs); restore_fpu(tsk); task_thread_info(tsk)->status |= TS_USEDFPU; - if ((((tsk->thread.fpu.hard.fpscr & FPSCR_ENABLE_MASK) >> 7) & + if ((((tsk->thread.xstate->hardfpu.fpscr & FPSCR_ENABLE_MASK) >> 7) & (fpu_exception_flags >> 2)) == 0) { return; } @@ -478,33 +427,3 @@ BUILD_TRAP_HANDLER(fpu_error) force_sig(SIGFPE, tsk); } - -void fpu_state_restore(struct pt_regs *regs) -{ - struct task_struct *tsk = current; - - grab_fpu(regs); - if (unlikely(!user_mode(regs))) { - printk(KERN_ERR "BUG: FPU is used in kernel mode.\n"); - BUG(); - return; - } - - if (likely(used_math())) { - /* Using the FPU again. */ - restore_fpu(tsk); - } else { - /* First time FPU user. */ - fpu_init(); - set_used_math(); - } - task_thread_info(tsk)->status |= TS_USEDFPU; - tsk->fpu_counter++; -} - -BUILD_TRAP_HANDLER(fpu_state_restore) -{ - TRAP_HANDLER_DECL; - - fpu_state_restore(regs); -} diff --git a/arch/sh/kernel/cpu/sh4/probe.c b/arch/sh/kernel/cpu/sh4/probe.c index d36f0c45f55f..cc02b3145cca 100644 --- a/arch/sh/kernel/cpu/sh4/probe.c +++ b/arch/sh/kernel/cpu/sh4/probe.c @@ -71,11 +71,11 @@ int __init detect_cpu_and_cache_system(void) boot_cpu_data.dcache.ways = 4; } else { /* And some SH-4 defaults.. */ - boot_cpu_data.flags |= CPU_HAS_PTEA; + boot_cpu_data.flags |= CPU_HAS_PTEA | CPU_HAS_FPU; boot_cpu_data.family = CPU_FAMILY_SH4; } - /* FPU detection works for everyone */ + /* FPU detection works for almost everyone */ if ((cvr & 0x20000000)) boot_cpu_data.flags |= CPU_HAS_FPU; @@ -124,6 +124,7 @@ int __init detect_cpu_and_cache_system(void) boot_cpu_data.type = CPU_SH7785; break; case 0x4004: + case 0x4005: boot_cpu_data.type = CPU_SH7786; boot_cpu_data.flags |= CPU_HAS_PTEAEX | CPU_HAS_L2_CACHE; break; @@ -160,6 +161,7 @@ int __init detect_cpu_and_cache_system(void) break; case 0x700: boot_cpu_data.type = CPU_SH4_501; + boot_cpu_data.flags &= ~CPU_HAS_FPU; boot_cpu_data.icache.ways = 2; boot_cpu_data.dcache.ways = 2; break; @@ -227,7 +229,7 @@ int __init detect_cpu_and_cache_system(void) * Size calculation is much more sensible * than it is for the L1. * - * Sizes are 128KB, 258KB, 512KB, and 1MB. + * Sizes are 128KB, 256KB, 512KB, and 1MB. */ size = (cvr & 0xf) << 17; diff --git a/arch/sh/kernel/cpu/sh4a/Makefile b/arch/sh/kernel/cpu/sh4a/Makefile index 33bab477d2e2..b144e8af89dc 100644 --- a/arch/sh/kernel/cpu/sh4a/Makefile +++ b/arch/sh/kernel/cpu/sh4a/Makefile @@ -41,7 +41,8 @@ pinmux-$(CONFIG_CPU_SUBTYPE_SH7757) := pinmux-sh7757.o pinmux-$(CONFIG_CPU_SUBTYPE_SH7785) := pinmux-sh7785.o pinmux-$(CONFIG_CPU_SUBTYPE_SH7786) := pinmux-sh7786.o -obj-y += $(clock-y) -obj-$(CONFIG_SMP) += $(smp-y) -obj-$(CONFIG_GENERIC_GPIO) += $(pinmux-y) -obj-$(CONFIG_PERF_EVENTS) += perf_event.o +obj-y += $(clock-y) +obj-$(CONFIG_SMP) += $(smp-y) +obj-$(CONFIG_GENERIC_GPIO) += $(pinmux-y) +obj-$(CONFIG_PERF_EVENTS) += perf_event.o +obj-$(CONFIG_HAVE_HW_BREAKPOINT) += ubc.o diff --git a/arch/sh/kernel/cpu/sh4a/clock-sh7722.c b/arch/sh/kernel/cpu/sh4a/clock-sh7722.c index ea38b554dc05..860ee2bf4bf0 100644 --- a/arch/sh/kernel/cpu/sh4a/clock-sh7722.c +++ b/arch/sh/kernel/cpu/sh4a/clock-sh7722.c @@ -117,12 +117,11 @@ static struct clk_div_mult_table div4_table = { .nr_multipliers = ARRAY_SIZE(multipliers), }; -enum { DIV4_I, DIV4_U, DIV4_SH, DIV4_B, DIV4_B3, DIV4_P, - DIV4_SIUA, DIV4_SIUB, DIV4_IRDA, DIV4_NR }; - #define DIV4(_str, _reg, _bit, _mask, _flags) \ SH_CLK_DIV4(_str, &pll_clk, _reg, _bit, _mask, _flags) +enum { DIV4_I, DIV4_U, DIV4_SH, DIV4_B, DIV4_B3, DIV4_P, DIV4_NR }; + struct clk div4_clks[DIV4_NR] = { [DIV4_I] = DIV4("cpu_clk", FRQCR, 20, 0x1fef, CLK_ENABLE_ON_INIT), [DIV4_U] = DIV4("umem_clk", FRQCR, 16, 0x1fff, CLK_ENABLE_ON_INIT), @@ -130,9 +129,19 @@ struct clk div4_clks[DIV4_NR] = { [DIV4_B] = DIV4("bus_clk", FRQCR, 8, 0x1fff, CLK_ENABLE_ON_INIT), [DIV4_B3] = DIV4("b3_clk", FRQCR, 4, 0x1fff, CLK_ENABLE_ON_INIT), [DIV4_P] = DIV4("peripheral_clk", FRQCR, 0, 0x1fff, 0), +}; + +enum { DIV4_IRDA, DIV4_ENABLE_NR }; + +struct clk div4_enable_clks[DIV4_ENABLE_NR] = { + [DIV4_IRDA] = DIV4("irda_clk", IRDACLKCR, 0, 0x1fff, 0), +}; + +enum { DIV4_SIUA, DIV4_SIUB, DIV4_REPARENT_NR }; + +struct clk div4_reparent_clks[DIV4_REPARENT_NR] = { [DIV4_SIUA] = DIV4("siua_clk", SCLKACR, 0, 0x1fff, 0), [DIV4_SIUB] = DIV4("siub_clk", SCLKBCR, 0, 0x1fff, 0), - [DIV4_IRDA] = DIV4("irda_clk", IRDACLKCR, 0, 0x1fff, 0), }; struct clk div6_clks[] = { @@ -189,6 +198,14 @@ int __init arch_clk_init(void) ret = sh_clk_div4_register(div4_clks, DIV4_NR, &div4_table); if (!ret) + ret = sh_clk_div4_enable_register(div4_enable_clks, + DIV4_ENABLE_NR, &div4_table); + + if (!ret) + ret = sh_clk_div4_reparent_register(div4_reparent_clks, + DIV4_REPARENT_NR, &div4_table); + + if (!ret) ret = sh_clk_div6_register(div6_clks, ARRAY_SIZE(div6_clks)); if (!ret) diff --git a/arch/sh/kernel/cpu/sh4a/clock-sh7786.c b/arch/sh/kernel/cpu/sh4a/clock-sh7786.c index a0e8869071ac..494c636012bb 100644 --- a/arch/sh/kernel/cpu/sh4a/clock-sh7786.c +++ b/arch/sh/kernel/cpu/sh4a/clock-sh7786.c @@ -3,11 +3,7 @@ * * SH7786 support for the clock framework * - * Copyright (C) 2008, 2009 Renesas Solutions Corp. - * Kuninori Morimoto <morimoto.kuninori@renesas.com> - * - * Based on SH7785 - * Copyright (C) 2007 Paul Mundt + * Copyright (C) 2010 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 @@ -15,127 +11,123 @@ */ #include <linux/init.h> #include <linux/kernel.h> +#include <linux/clk.h> +#include <linux/io.h> #include <asm/clock.h> #include <asm/freq.h> -#include <asm/io.h> - -static int ifc_divisors[] = { 1, 2, 4, 1 }; -static int sfc_divisors[] = { 1, 1, 4, 1 }; -static int bfc_divisors[] = { 1, 1, 1, 1, 1, 12, 16, 1, - 24, 32, 1, 1, 1, 1, 1, 1 }; -static int mfc_divisors[] = { 1, 1, 4, 1 }; -static int pfc_divisors[] = { 1, 1, 1, 1, 1, 1, 16, 1, - 24, 32, 1, 48, 1, 1, 1, 1 }; - -static void master_clk_init(struct clk *clk) -{ - clk->rate *= pfc_divisors[ctrl_inl(FRQMR1) & 0x000f]; -} -static struct clk_ops sh7786_master_clk_ops = { - .init = master_clk_init, +/* + * Default rate for the root input clock, reset this with clk_set_rate() + * from the platform code. + */ +static struct clk extal_clk = { + .name = "extal", + .id = -1, + .rate = 33333333, }; -static unsigned long module_clk_recalc(struct clk *clk) +static unsigned long pll_recalc(struct clk *clk) { - int idx = (ctrl_inl(FRQMR1) & 0x000f); - return clk->parent->rate / pfc_divisors[idx]; -} + int multiplier; -static struct clk_ops sh7786_module_clk_ops = { - .recalc = module_clk_recalc, -}; + /* + * Clock modes 0, 1, and 2 use an x64 multiplier against PLL1, + * while modes 3, 4, and 5 use an x32. + */ + multiplier = (sh_mv.mv_mode_pins() & 0xf) < 3 ? 64 : 32; -static unsigned long bus_clk_recalc(struct clk *clk) -{ - int idx = ((ctrl_inl(FRQMR1) >> 16) & 0x000f); - return clk->parent->rate / bfc_divisors[idx]; + return clk->parent->rate * multiplier; } -static struct clk_ops sh7786_bus_clk_ops = { - .recalc = bus_clk_recalc, +static struct clk_ops pll_clk_ops = { + .recalc = pll_recalc, }; -static unsigned long cpu_clk_recalc(struct clk *clk) -{ - int idx = ((ctrl_inl(FRQMR1) >> 28) & 0x0003); - return clk->parent->rate / ifc_divisors[idx]; -} - -static struct clk_ops sh7786_cpu_clk_ops = { - .recalc = cpu_clk_recalc, +static struct clk pll_clk = { + .name = "pll_clk", + .id = -1, + .ops = &pll_clk_ops, + .parent = &extal_clk, + .flags = CLK_ENABLE_ON_INIT, }; -static struct clk_ops *sh7786_clk_ops[] = { - &sh7786_master_clk_ops, - &sh7786_module_clk_ops, - &sh7786_bus_clk_ops, - &sh7786_cpu_clk_ops, +static struct clk *clks[] = { + &extal_clk, + &pll_clk, }; -void __init arch_init_clk_ops(struct clk_ops **ops, int idx) -{ - if (idx < ARRAY_SIZE(sh7786_clk_ops)) - *ops = sh7786_clk_ops[idx]; -} - -static unsigned long shyway_clk_recalc(struct clk *clk) -{ - int idx = ((ctrl_inl(FRQMR1) >> 20) & 0x0003); - return clk->parent->rate / sfc_divisors[idx]; -} +static unsigned int div2[] = { 1, 2, 4, 6, 8, 12, 16, 18, + 24, 32, 36, 48 }; -static struct clk_ops sh7786_shyway_clk_ops = { - .recalc = shyway_clk_recalc, +static struct clk_div_mult_table div4_table = { + .divisors = div2, + .nr_divisors = ARRAY_SIZE(div2), }; -static struct clk sh7786_shyway_clk = { - .name = "shyway_clk", - .flags = CLK_ENABLE_ON_INIT, - .ops = &sh7786_shyway_clk_ops, -}; +enum { DIV4_I, DIV4_SH, DIV4_B, DIV4_DDR, DIV4_DU, DIV4_P, DIV4_NR }; -static unsigned long ddr_clk_recalc(struct clk *clk) -{ - int idx = ((ctrl_inl(FRQMR1) >> 12) & 0x0003); - return clk->parent->rate / mfc_divisors[idx]; -} +#define DIV4(_str, _bit, _mask, _flags) \ + SH_CLK_DIV4(_str, &pll_clk, FRQMR1, _bit, _mask, _flags) -static struct clk_ops sh7786_ddr_clk_ops = { - .recalc = ddr_clk_recalc, +struct clk div4_clks[DIV4_NR] = { + [DIV4_P] = DIV4("peripheral_clk", 0, 0x0b40, 0), + [DIV4_DU] = DIV4("du_clk", 4, 0x0010, 0), + [DIV4_DDR] = DIV4("ddr_clk", 12, 0x0002, CLK_ENABLE_ON_INIT), + [DIV4_B] = DIV4("bus_clk", 16, 0x0360, CLK_ENABLE_ON_INIT), + [DIV4_SH] = DIV4("shyway_clk", 20, 0x0002, CLK_ENABLE_ON_INIT), + [DIV4_I] = DIV4("cpu_clk", 28, 0x0006, CLK_ENABLE_ON_INIT), }; -static struct clk sh7786_ddr_clk = { - .name = "ddr_clk", - .flags = CLK_ENABLE_ON_INIT, - .ops = &sh7786_ddr_clk_ops, -}; - -/* - * Additional SH7786-specific on-chip clocks that aren't already part of the - * clock framework - */ -static struct clk *sh7786_onchip_clocks[] = { - &sh7786_shyway_clk, - &sh7786_ddr_clk, +#define MSTPCR0 0xffc40030 +#define MSTPCR1 0xffc40034 + +static struct clk mstp_clks[] = { + /* MSTPCR0 */ + SH_CLK_MSTP32("scif_fck", 5, &div4_clks[DIV4_P], MSTPCR0, 29, 0), + SH_CLK_MSTP32("scif_fck", 4, &div4_clks[DIV4_P], MSTPCR0, 28, 0), + SH_CLK_MSTP32("scif_fck", 3, &div4_clks[DIV4_P], MSTPCR0, 27, 0), + SH_CLK_MSTP32("scif_fck", 2, &div4_clks[DIV4_P], MSTPCR0, 26, 0), + SH_CLK_MSTP32("scif_fck", 1, &div4_clks[DIV4_P], MSTPCR0, 25, 0), + SH_CLK_MSTP32("scif_fck", 0, &div4_clks[DIV4_P], MSTPCR0, 24, 0), + SH_CLK_MSTP32("ssi_fck", 3, &div4_clks[DIV4_P], MSTPCR0, 23, 0), + SH_CLK_MSTP32("ssi_fck", 2, &div4_clks[DIV4_P], MSTPCR0, 22, 0), + SH_CLK_MSTP32("ssi_fck", 1, &div4_clks[DIV4_P], MSTPCR0, 21, 0), + SH_CLK_MSTP32("ssi_fck", 0, &div4_clks[DIV4_P], MSTPCR0, 20, 0), + SH_CLK_MSTP32("hac_fck", 1, &div4_clks[DIV4_P], MSTPCR0, 17, 0), + SH_CLK_MSTP32("hac_fck", 0, &div4_clks[DIV4_P], MSTPCR0, 16, 0), + SH_CLK_MSTP32("i2c_fck", 1, &div4_clks[DIV4_P], MSTPCR0, 15, 0), + SH_CLK_MSTP32("i2c_fck", 0, &div4_clks[DIV4_P], MSTPCR0, 14, 0), + SH_CLK_MSTP32("tmu9_11_fck", -1, &div4_clks[DIV4_P], MSTPCR0, 11, 0), + SH_CLK_MSTP32("tmu678_fck", -1, &div4_clks[DIV4_P], MSTPCR0, 10, 0), + SH_CLK_MSTP32("tmu345_fck", -1, &div4_clks[DIV4_P], MSTPCR0, 9, 0), + SH_CLK_MSTP32("tmu012_fck", -1, &div4_clks[DIV4_P], MSTPCR0, 8, 0), + SH_CLK_MSTP32("sdif_fck", 1, &div4_clks[DIV4_P], MSTPCR0, 5, 0), + SH_CLK_MSTP32("sdif_fck", 0, &div4_clks[DIV4_P], MSTPCR0, 4, 0), + SH_CLK_MSTP32("hspi_fck", -1, &div4_clks[DIV4_P], MSTPCR0, 2, 0), + + /* MSTPCR1 */ + SH_CLK_MSTP32("usb_fck", -1, NULL, MSTPCR1, 12, 0), + SH_CLK_MSTP32("pcie_fck", 2, NULL, MSTPCR1, 10, 0), + SH_CLK_MSTP32("pcie_fck", 1, NULL, MSTPCR1, 9, 0), + SH_CLK_MSTP32("pcie_fck", 0, NULL, MSTPCR1, 8, 0), + SH_CLK_MSTP32("dmac_11_6_fck", -1, NULL, MSTPCR1, 5, 0), + SH_CLK_MSTP32("dmac_5_0_fck", -1, NULL, MSTPCR1, 4, 0), + SH_CLK_MSTP32("du_fck", -1, NULL, MSTPCR1, 3, 0), + SH_CLK_MSTP32("ether_fck", -1, NULL, MSTPCR1, 2, 0), }; int __init arch_clk_init(void) { - struct clk *clk; int i, ret = 0; - cpg_clk_init(); - - clk = clk_get(NULL, "master_clk"); - for (i = 0; i < ARRAY_SIZE(sh7786_onchip_clocks); i++) { - struct clk *clkp = sh7786_onchip_clocks[i]; - - clkp->parent = clk; - ret |= clk_register(clkp); - } + for (i = 0; i < ARRAY_SIZE(clks); i++) + ret |= clk_register(clks[i]); - clk_put(clk); + if (!ret) + ret = sh_clk_div4_register(div4_clks, ARRAY_SIZE(div4_clks), + &div4_table); + if (!ret) + ret = sh_clk_mstp32_register(mstp_clks, ARRAY_SIZE(mstp_clks)); return ret; } diff --git a/arch/sh/kernel/cpu/sh4a/smp-shx3.c b/arch/sh/kernel/cpu/sh4a/smp-shx3.c index 5863e0c4d02f..11bf4c1e25c0 100644 --- a/arch/sh/kernel/cpu/sh4a/smp-shx3.c +++ b/arch/sh/kernel/cpu/sh4a/smp-shx3.c @@ -78,7 +78,10 @@ void __init plat_prepare_cpus(unsigned int max_cpus) void plat_start_cpu(unsigned int cpu, unsigned long entry_point) { - __raw_writel(entry_point, RESET_REG(cpu)); + if (__in_29bit_mode()) + __raw_writel(entry_point, RESET_REG(cpu)); + else + __raw_writel(virt_to_phys(entry_point), RESET_REG(cpu)); if (!(__raw_readl(STBCR_REG(cpu)) & STBCR_MSTP)) __raw_writel(STBCR_MSTP, STBCR_REG(cpu)); diff --git a/arch/sh/kernel/cpu/sh4a/ubc.c b/arch/sh/kernel/cpu/sh4a/ubc.c new file mode 100644 index 000000000000..efb2745bcb36 --- /dev/null +++ b/arch/sh/kernel/cpu/sh4a/ubc.c @@ -0,0 +1,133 @@ +/* + * arch/sh/kernel/cpu/sh4a/ubc.c + * + * On-chip UBC support for SH-4A CPUs. + * + * Copyright (C) 2009 - 2010 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/err.h> +#include <linux/clk.h> +#include <linux/io.h> +#include <asm/hw_breakpoint.h> + +#define UBC_CBR(idx) (0xff200000 + (0x20 * idx)) +#define UBC_CRR(idx) (0xff200004 + (0x20 * idx)) +#define UBC_CAR(idx) (0xff200008 + (0x20 * idx)) +#define UBC_CAMR(idx) (0xff20000c + (0x20 * idx)) + +#define UBC_CCMFR 0xff200600 +#define UBC_CBCR 0xff200620 + +/* CRR */ +#define UBC_CRR_PCB (1 << 1) +#define UBC_CRR_BIE (1 << 0) + +/* CBR */ +#define UBC_CBR_CE (1 << 0) + +static struct sh_ubc sh4a_ubc; + +static void sh4a_ubc_enable(struct arch_hw_breakpoint *info, int idx) +{ + __raw_writel(UBC_CBR_CE | info->len | info->type, UBC_CBR(idx)); + __raw_writel(info->address, UBC_CAR(idx)); +} + +static void sh4a_ubc_disable(struct arch_hw_breakpoint *info, int idx) +{ + __raw_writel(0, UBC_CBR(idx)); + __raw_writel(0, UBC_CAR(idx)); +} + +static void sh4a_ubc_enable_all(unsigned long mask) +{ + int i; + + for (i = 0; i < sh4a_ubc.num_events; i++) + if (mask & (1 << i)) + __raw_writel(__raw_readl(UBC_CBR(i)) | UBC_CBR_CE, + UBC_CBR(i)); +} + +static void sh4a_ubc_disable_all(void) +{ + int i; + + for (i = 0; i < sh4a_ubc.num_events; i++) + __raw_writel(__raw_readl(UBC_CBR(i)) & ~UBC_CBR_CE, + UBC_CBR(i)); +} + +static unsigned long sh4a_ubc_active_mask(void) +{ + unsigned long active = 0; + int i; + + for (i = 0; i < sh4a_ubc.num_events; i++) + if (__raw_readl(UBC_CBR(i)) & UBC_CBR_CE) + active |= (1 << i); + + return active; +} + +static unsigned long sh4a_ubc_triggered_mask(void) +{ + return __raw_readl(UBC_CCMFR); +} + +static void sh4a_ubc_clear_triggered_mask(unsigned long mask) +{ + __raw_writel(__raw_readl(UBC_CCMFR) & ~mask, UBC_CCMFR); +} + +static struct sh_ubc sh4a_ubc = { + .name = "SH-4A", + .num_events = 2, + .trap_nr = 0x1e0, + .enable = sh4a_ubc_enable, + .disable = sh4a_ubc_disable, + .enable_all = sh4a_ubc_enable_all, + .disable_all = sh4a_ubc_disable_all, + .active_mask = sh4a_ubc_active_mask, + .triggered_mask = sh4a_ubc_triggered_mask, + .clear_triggered_mask = sh4a_ubc_clear_triggered_mask, +}; + +static int __init sh4a_ubc_init(void) +{ + struct clk *ubc_iclk = clk_get(NULL, "ubc0"); + int i; + + /* + * The UBC MSTP bit is optional, as not all platforms will have + * it. Just ignore it if we can't find it. + */ + if (IS_ERR(ubc_iclk)) + ubc_iclk = NULL; + + clk_enable(ubc_iclk); + + __raw_writel(0, UBC_CBCR); + + for (i = 0; i < sh4a_ubc.num_events; i++) { + __raw_writel(0, UBC_CAMR(i)); + __raw_writel(0, UBC_CBR(i)); + + __raw_writel(UBC_CRR_BIE | UBC_CRR_PCB, UBC_CRR(i)); + + /* dummy read for write posting */ + (void)__raw_readl(UBC_CRR(i)); + } + + clk_disable(ubc_iclk); + + sh4a_ubc.clk = ubc_iclk; + + return register_sh_ubc(&sh4a_ubc); +} +arch_initcall(sh4a_ubc_init); diff --git a/arch/sh/kernel/cpu/sh5/entry.S b/arch/sh/kernel/cpu/sh5/entry.S index 8f13f73cb2cb..6b80295dd7a4 100644 --- a/arch/sh/kernel/cpu/sh5/entry.S +++ b/arch/sh/kernel/cpu/sh5/entry.S @@ -187,7 +187,7 @@ trap_jtable: .rept 6 .long do_exception_error /* 0x880 - 0x920 */ .endr - .long do_software_break_point /* 0x940 */ + .long breakpoint_trap_handler /* 0x940 */ .long do_exception_error /* 0x960 */ .long do_single_step /* 0x980 */ @@ -1124,7 +1124,7 @@ fpu_error_or_IRQA: pta its_IRQ, tr0 beqi/l r4, EVENT_INTERRUPT, tr0 #ifdef CONFIG_SH_FPU - movi do_fpu_state_restore, r6 + movi fpu_state_restore_trap_handler, r6 #else movi do_exception_error, r6 #endif @@ -1135,7 +1135,7 @@ fpu_error_or_IRQB: pta its_IRQ, tr0 beqi/l r4, EVENT_INTERRUPT, tr0 #ifdef CONFIG_SH_FPU - movi do_fpu_state_restore, r6 + movi fpu_state_restore_trap_handler, r6 #else movi do_exception_error, r6 #endif diff --git a/arch/sh/kernel/cpu/sh5/fpu.c b/arch/sh/kernel/cpu/sh5/fpu.c index 4648ccee6c4d..4b3bb35e99f3 100644 --- a/arch/sh/kernel/cpu/sh5/fpu.c +++ b/arch/sh/kernel/cpu/sh5/fpu.c @@ -15,24 +15,6 @@ #include <linux/sched.h> #include <linux/signal.h> #include <asm/processor.h> -#include <asm/user.h> -#include <asm/io.h> -#include <asm/fpu.h> - -/* - * Initially load the FPU with signalling NANS. This bit pattern - * has the property that no matter whether considered as single or as - * double precision, it still represents a signalling NAN. - */ -#define sNAN64 0xFFFFFFFFFFFFFFFFULL -#define sNAN32 0xFFFFFFFFUL - -static union sh_fpu_union init_fpuregs = { - .hard = { - .fp_regs = { [0 ... 63] = sNAN32 }, - .fpscr = FPSCR_INIT - } -}; void save_fpu(struct task_struct *tsk) { @@ -72,12 +54,11 @@ void save_fpu(struct task_struct *tsk) "fgetscr fr63\n\t" "fst.s %0, (32*8), fr63\n\t" : /* no output */ - : "r" (&tsk->thread.fpu.hard) + : "r" (&tsk->thread.xstate->hardfpu) : "memory"); } -static inline void -fpload(struct sh_fpu_hard_struct *fpregs) +void restore_fpu(struct task_struct *tsk) { asm volatile("fld.p %0, (0*8), fp0\n\t" "fld.p %0, (1*8), fp2\n\t" @@ -116,16 +97,11 @@ fpload(struct sh_fpu_hard_struct *fpregs) "fld.p %0, (31*8), fp62\n\t" : /* no output */ - : "r" (fpregs) ); -} - -void fpinit(struct sh_fpu_hard_struct *fpregs) -{ - *fpregs = init_fpuregs.hard; + : "r" (&tsk->thread.xstate->hardfpu) + : "memory"); } -asmlinkage void -do_fpu_error(unsigned long ex, struct pt_regs *regs) +asmlinkage void do_fpu_error(unsigned long ex, struct pt_regs *regs) { struct task_struct *tsk = current; @@ -133,35 +109,6 @@ do_fpu_error(unsigned long ex, struct pt_regs *regs) tsk->thread.trap_no = 11; tsk->thread.error_code = 0; - force_sig(SIGFPE, tsk); -} - - -asmlinkage void -do_fpu_state_restore(unsigned long ex, struct pt_regs *regs) -{ - void die(const char *str, struct pt_regs *regs, long err); - - if (! user_mode(regs)) - die("FPU used in kernel", regs, ex); - regs->sr &= ~SR_FD; - - if (last_task_used_math == current) - return; - - enable_fpu(); - if (last_task_used_math != NULL) - /* Other processes fpu state, save away */ - save_fpu(last_task_used_math); - - last_task_used_math = current; - if (used_math()) { - fpload(¤t->thread.fpu.hard); - } else { - /* First time FPU user. */ - fpload(&init_fpuregs.hard); - set_used_math(); - } - disable_fpu(); + force_sig(SIGFPE, tsk); } |