r4xx0.c

移植到2410开发板上的源代码

	if (size >= (unsigned long)scache_size) {
		flush_cache_l1();
		return;
	}

	a = addr & ~((unsigned long)sc_lsize - 1);
	end = (addr + size) & ~((unsigned long)sc_lsize - 1);
	while (1) {
		flush_scache_line(a); /* Hit_Writeback_Inv_SD */
		if (a == end) break;
		a += sc_lsize;
	}
}

static void r4k_dma_cache_wback(unsigned long addr, unsigned long size)
{
	panic("r4k_dma_cache called - should not happen.\n");
}

/*
 * While we're protected against bad userland addresses we don't care
 * very much about what happens in that case.  Usually a segmentation
 * fault will dump the process later on anyway ...
 */
static void r4k_flush_cache_sigtramp(unsigned long addr)
{
	__asm__ __volatile__("nop;nop;nop;nop");	/* R4600 V1.7 */

	protected_writeback_dcache_line(addr & ~(dc_lsize - 1));
	protected_flush_icache_line(addr & ~(ic_lsize - 1));
}

static void r4600v20k_flush_cache_sigtramp(unsigned long addr)
{
	unsigned int flags;

	__save_and_cli(flags);

	/* Clear internal cache refill buffer */
	*(volatile unsigned int *)KSEG1;

	protected_writeback_dcache_line(addr & ~(dc_lsize - 1));
	protected_flush_icache_line(addr & ~(ic_lsize - 1));

	__restore_flags(flags);
}

#undef DEBUG_TLB

#define NTLB_ENTRIES       48  /* Fixed on all R4XX0 variants... */

#define NTLB_ENTRIES_HALF  24  /* Fixed on all R4XX0 variants... */

static inline void r4k_flush_tlb_all(void)
{
	unsigned long flags;
	unsigned long old_ctx;
	int entry;

#ifdef DEBUG_TLB
	printk("[tlball]");
#endif

	__save_and_cli(flags);
	/* Save old context and create impossible VPN2 value */
	old_ctx = (get_entryhi() & 0xff);
	set_entryhi(KSEG0);
	set_entrylo0(0);
	set_entrylo1(0);
	BARRIER;

	entry = get_wired();

	/* Blast 'em all away. */
	while(entry < NTLB_ENTRIES) {
		set_index(entry);
		BARRIER;
		tlb_write_indexed();
		BARRIER;
		entry++;
	}
	BARRIER;
	set_entryhi(old_ctx);
	__restore_flags(flags);
}

static void r4k_flush_tlb_mm(struct mm_struct *mm)
{
	if (CPU_CONTEXT(smp_processor_id(), mm) != 0) {
		unsigned long flags;

#ifdef DEBUG_TLB
		printk("[tlbmm<%d>]", mm->context);
#endif
		__save_and_cli(flags);
		get_new_cpu_mmu_context(mm, smp_processor_id());
		if(mm == current->mm)
			set_entryhi(CPU_CONTEXT(smp_processor_id(), mm) & 0xff);
		__restore_flags(flags);
	}
}

static void r4k_flush_tlb_range(struct mm_struct *mm, unsigned long start,
				unsigned long end)
{
	if (CPU_CONTEXT(smp_processor_id(), mm) != 0) {
		unsigned long flags;
		int size;

#ifdef DEBUG_TLB
		printk("[tlbrange<%02x,%08lx,%08lx>]", (mm->context & 0xff),
		       start, end);
#endif
		__save_and_cli(flags);
		size = (end - start + (PAGE_SIZE - 1)) >> PAGE_SHIFT;
		size = (size + 1) >> 1;
		if(size <= NTLB_ENTRIES_HALF) {
			int oldpid = (get_entryhi() & 0xff);
			int newpid = (CPU_CONTEXT(smp_processor_id(), mm) & 0xff);

			start &= (PAGE_MASK << 1);
			end += ((PAGE_SIZE << 1) - 1);
			end &= (PAGE_MASK << 1);
			while(start < end) {
				int idx;

				set_entryhi(start | newpid);
				start += (PAGE_SIZE << 1);
				BARRIER;
				tlb_probe();
				BARRIER;
				idx = get_index();
				set_entrylo0(0);
				set_entrylo1(0);
				set_entryhi(KSEG0);
				BARRIER;
				if(idx < 0)
					continue;
				tlb_write_indexed();
				BARRIER;
			}
			set_entryhi(oldpid);
		} else {
			get_new_cpu_mmu_context(mm, smp_processor_id());
			if(mm == current->mm)
				set_entryhi(CPU_CONTEXT(smp_processor_id(), 
								mm) & 0xff);
		}
		__restore_flags(flags);
	}
}

static void r4k_flush_tlb_page(struct vm_area_struct *vma, unsigned long page)
{
	if (CPU_CONTEXT(smp_processor_id(), vma->vm_mm) != 0) {
		unsigned long flags;
		int oldpid, newpid, idx;

#ifdef DEBUG_TLB
		printk("[tlbpage<%d,%08lx>]", vma->vm_mm->context, page);
#endif
		newpid = (CPU_CONTEXT(smp_processor_id(), vma->vm_mm) & 0xff);
		page &= (PAGE_MASK << 1);
		__save_and_cli(flags);
		oldpid = (get_entryhi() & 0xff);
		set_entryhi(page | newpid);
		BARRIER;
		tlb_probe();
		BARRIER;
		idx = get_index();
		set_entrylo0(0);
		set_entrylo1(0);
		set_entryhi(KSEG0);
		if(idx < 0)
			goto finish;
		BARRIER;
		tlb_write_indexed();

	finish:
		BARRIER;
		set_entryhi(oldpid);
		__restore_flags(flags);
	}
}

static void r4k_flush_cache_l2(void)
{
}

/* We will need multiple versions of update_mmu_cache(), one that just
 * updates the TLB with the new pte(s), and another which also checks
 * for the R4k "end of page" hardware bug and does the needy.
 */
static void r4k_update_mmu_cache(struct vm_area_struct * vma,
				 unsigned long address, pte_t pte)
{
	unsigned long flags;
	pgd_t *pgdp;
	pmd_t *pmdp;
	pte_t *ptep;
	int idx, pid;

	/*
	 * Handle debugger faulting in for debugee.
	 */
	if (current->active_mm != vma->vm_mm)
		return;

	__save_and_cli(flags);
	pid = (get_entryhi() & 0xff);

#ifdef DEBUG_TLB
	if((pid != (CPU_CONTEXT(smp_processor_id(), vma->vm_mm) & 0xff)) ||
	   (CPU_CONTEXT(smp_processor_id(), vma->vm_mm) == 0)) {
		printk("update_mmu_cache: Wheee, bogus tlbpid mmpid=%d
			tlbpid=%d\n", (int) (CPU_CONTEXT(smp_processor_id(),
			vma->vm_mm) & 0xff), pid);
	}
#endif

	address &= (PAGE_MASK << 1);
	set_entryhi(address | (pid));
	pgdp = pgd_offset(vma->vm_mm, address);
	BARRIER;
	tlb_probe();
	BARRIER;
	pmdp = pmd_offset(pgdp, address);
	idx = get_index();
	ptep = pte_offset(pmdp, address);
	BARRIER;
	set_entrylo0(pte_val(*ptep++) >> 6);
	set_entrylo1(pte_val(*ptep) >> 6);
	set_entryhi(address | (pid));
	BARRIER;
	if(idx < 0) {
		tlb_write_random();
	} else {
		tlb_write_indexed();
	}
	BARRIER;
	set_entryhi(pid);
	BARRIER;
	__restore_flags(flags);
}

#if 0
static void r4k_update_mmu_cache_hwbug(struct vm_area_struct * vma,
				       unsigned long address, pte_t pte)
{
	unsigned long flags;
	pgd_t *pgdp;
	pmd_t *pmdp;
	pte_t *ptep;
	int idx;

	__save_and_cli(flags);
	address &= (PAGE_MASK << 1);
	set_entryhi(address | (get_entryhi() & 0xff));
	pgdp = pgd_offset(vma->vm_mm, address);
	tlb_probe();
	pmdp = pmd_offset(pgdp, address);
	idx = get_index();
	ptep = pte_offset(pmdp, address);
	set_entrylo0(pte_val(*ptep++) >> 6);
	set_entrylo1(pte_val(*ptep) >> 6);
	BARRIER;
	if(idx < 0)
		tlb_write_random();
	else
		tlb_write_indexed();
	BARRIER;
	__restore_flags(flags);
}
#endif

static void r4k_show_regs(struct pt_regs *regs)
{
	/* Saved main processor registers. */
	printk("$0      : %016lx %016lx %016lx %016lx\n",
	       0UL, regs->regs[1], regs->regs[2], regs->regs[3]);
	printk("$4      : %016lx %016lx %016lx %016lx\n",
               regs->regs[4], regs->regs[5], regs->regs[6], regs->regs[7]);
	printk("$8      : %016lx %016lx %016lx %016lx\n",
	       regs->regs[8], regs->regs[9], regs->regs[10], regs->regs[11]);
	printk("$12     : %016lx %016lx %016lx %016lx\n",
               regs->regs[12], regs->regs[13], regs->regs[14], regs->regs[15]);
	printk("$16     : %016lx %016lx %016lx %016lx\n",
	       regs->regs[16], regs->regs[17], regs->regs[18], regs->regs[19]);
	printk("$20     : %016lx %016lx %016lx %016lx\n",
               regs->regs[20], regs->regs[21], regs->regs[22], regs->regs[23]);
	printk("$24     : %016lx %016lx\n",
	       regs->regs[24], regs->regs[25]);
	printk("$28     : %016lx %016lx %016lx %016lx\n",
	       regs->regs[28], regs->regs[29], regs->regs[30], regs->regs[31]);
	printk("Hi      : %016lx\n", regs->hi);
	printk("Lo      : %016lx\n", regs->lo);

	/* Saved cp0 registers. */
	printk("epc     : %016lx    %s\nbadvaddr: %016lx\n",
	       regs->cp0_epc, print_tainted(), regs->cp0_badvaddr);
	printk("Status  : %08x\nCause   : %08x\n",
	       (unsigned int) regs->cp0_status, (unsigned int) regs->cp0_cause);
}

/* Detect and size the various r4k caches. */
static void __init probe_icache(unsigned long config)
{
	icache_size = 1 << (12 + ((config >> 9) & 7));
	ic_lsize = 16 << ((config >> 5) & 1);

	printk("Primary instruction cache %dkb, linesize %d bytes)\n",
	       icache_size >> 10, ic_lsize);
}

static void __init probe_dcache(unsigned long config)
{
	dcache_size = 1 << (12 + ((config >> 6) & 7));
	dc_lsize = 16 << ((config >> 4) & 1);

	printk("Primary data cache %dkb, linesize %d bytes)\n",
	       dcache_size >> 10, dc_lsize);
}


/* If you even _breathe_ on this function, look at the gcc output
 * and make sure it does not pop things on and off the stack for
 * the cache sizing loop that executes in KSEG1 space or else
 * you will crash and burn badly.  You have been warned.
 */
static int __init probe_scache(unsigned long config)
{
	extern unsigned long stext;
	unsigned long flags, addr, begin, end, pow2;
	int tmp;

	tmp = ((config >> 17) & 1);
	if(tmp)
		return 0;
	tmp = ((config >> 22) & 3);
	switch(tmp) {
	case 0:
		sc_lsize = 16;
		break;
	case 1:
		sc_lsize = 32;
		break;
	case 2:
		sc_lsize = 64;
		break;
	case 3:
		sc_lsize = 128;
		break;
	}

	begin = (unsigned long) &stext;
	begin &= ~((4 * 1024 * 1024) - 1);
	end = begin + (4 * 1024 * 1024);

	/* This is such a bitch, you'd think they would make it
	 * easy to do this.  Away you daemons of stupidity!
	 */
	__save_and_cli(flags);

	/* Fill each size-multiple cache line with a valid tag. */
	pow2 = (64 * 1024);
	for(addr = begin; addr < end; addr = (begin + pow2)) {
		unsigned long *p = (unsigned long *) addr;
		__asm__ __volatile__("nop" : : "r" (*p)); /* whee... */
		pow2 <<= 1;
	}

	/* Load first line with zero (therefore invalid) tag. */
	set_taglo(0);
	set_taghi(0);
	__asm__ __volatile__("nop; nop; nop; nop;"); /* avoid the hazard */
	__asm__ __volatile__("\n\t.set noreorder\n\t"
			     "cache 8, (%0)\n\t"
			     ".set reorder\n\t" : : "r" (begin));
	__asm__ __volatile__("\n\t.set noreorder\n\t"
			     "cache 9, (%0)\n\t"
			     ".set reorder\n\t" : : "r" (begin));
	__asm__ __volatile__("\n\t.set noreorder\n\t"
			     "cache 11, (%0)\n\t"
			     ".set reorder\n\t" : : "r" (begin));

	/* Now search for the wrap around point. */
	pow2 = (128 * 1024);
	tmp = 0;
	for(addr = (begin + (128 * 1024)); addr < (end); addr = (begin + pow2)) {
		__asm__ __volatile__("\n\t.set noreorder\n\t"
				     "cache 7, (%0)\n\t"
				     ".set reorder\n\t" : : "r" (addr));
		__asm__ __volatile__("nop; nop; nop; nop;"); /* hazard... */
		if(!get_taglo())
			break;
		pow2 <<= 1;
	}
	__restore_flags(flags);
	addr -= begin;
	printk("Secondary cache sized at %dK linesize %d\n",
	       (int) (addr >> 10), sc_lsize);
	scache_size = addr;
	return 1;
}

static void __init setup_noscache_funcs(void)
{
	unsigned int prid;

	switch(dc_lsize) {
	case 16:
		_clear_page = r4k_clear_page_d16;
		_copy_page = r4k_copy_page_d16;
		_flush_cache_l1 = r4k_flush_cache_all_d16i16;
		_flush_cache_mm = r4k_flush_cache_mm_d16i16;
		_flush_cache_range = r4k_flush_cache_range_d16i16;
		_flush_cache_page = r4k_flush_cache_page_d16i16;
		break;
	case 32:
		prid = read_32bit_cp0_register(CP0_PRID) & 0xfff0;
		if (prid == 0x2010) {			/* R4600 V1.7 */
			_clear_page = r4k_clear_page_r4600_v1;
			_copy_page = r4k_copy_page_r4600_v1;
		} else if (prid == 0x2020) {		/* R4600 V2.0 */
			_clear_page = r4k_clear_page_r4600_v2;
			_copy_page = r4k_copy_page_r4600_v2;
		} else {
			_clear_page = r4k_clear_page_d32;
			_copy_page = r4k_copy_page_d32;
		}
		_flush_cache_l1 = r4k_flush_cache_all_d32i32;
		_flush_cache_mm = r4k_flush_cache_mm_d32i32;
		_flush_cache_range = r4k_flush_cache_range_d32i32;
		_flush_cache_page = r4k_flush_cache_page_d32i32;
		break;
	}

	switch(ic_lsize) {
	case 16:
		_flush_page_to_ram = r4k_flush_page_to_ram_d16;
		break;
	case 32:
		_flush_page_to_ram = r4k_flush_page_to_ram_d32;
		break;
	}

	_dma_cache_wback_inv = r4k_dma_cache_wback_inv_pc;
	_dma_cache_wback = r4k_dma_cache_wback;
	_dma_cache_inv = r4k_dma_cache_inv_pc;
}

static void __init setup_scache_funcs(void)
{
	switch(sc_lsize) {
	case 16:
		switch(dc_lsize) {
		case 16:
			_flush_cache_l1 = r4k_flush_cache_all_s16d16i16;
			_flush_cache_mm = r4k_flush_cache_mm_s16d16i16;
			_flush_cache_range = r4k_flush_cache_range_s16d16i16;
			_flush_cache_page = r4k_flush_cache_page_s16d16i16;
			break;
		case 32:
			panic("Invalid cache configuration detected");
		};
		_flush_page_to_ram = r4k_flush_page_to_ram_s16;
		_clear_page = r4k_clear_page_s16;
		_copy_page = r4k_copy_page_s16;
		break;
	case 32:
		switch(dc_lsize) {
		case 16:
			_flush_cache_l1 = r4k_flush_cache_all_s32d16i16;
			_flush_cache_mm = r4k_flush_cache_mm_s32d16i16;
			_flush_cache_range = r4k_flush_cache_range_s32d16i16;
			_flush_cache_page = r4k_flush_cache_page_s32d16i16;
			break;
		case 32:
			_flush_cache_l1 = r4k_flush_cache_all_s32d32i32;
			_flush_cache_mm = r4k_flush_cache_mm_s32d32i32;
			_flush_cache_range = r4k_flush_cache_range_s32d32i32;
			_flush_cache_page = r4k_flush_cache_page_s32d32i32;
			break;
		};
		_flush_page_to_ram = r4k_flush_page_to_ram_s32;
		_clear_page = r4k_clear_page_s32;
		_copy_page = r4k_copy_page_s32;
		break;
	case 64:
		switch(dc_lsize) {
		case 16:
			_flush_cache_l1 = r4k_flush_cache_all_s64d16i16;
			_flush_cache_mm = r4k_flush_cache_mm_s64d16i16;
			_flush_cache_range = r4k_flush_cache_range_s64d16i16;
			_flush_cache_page = r4k_flush_cache_page_s64d16i16;
			break;
		case 32:
			_flush_cache_l1 = r4k_flush_cache_all_s64d32i32;
			_flush_cache_mm = r4k_flush_cache_mm_s64d32i32;
			_flush_cache_range = r4k_flush_cache_range_s64d32i32;
			_flush_cache_page = r4k_flush_cache_page_s64d32i32;
			break;
		};
		_flush_page_to_ram = r4k_flush_page_to_ram_s64;
		_clear_page = r4k_clear_page_s64;
		_copy_page = r4k_copy_page_s64;
		break;
	case 128:
		switch(dc_lsize) {
		case 16:
			_flush_cache_l1 = r4k_flush_cache_all_s128d16i16;
			_flush_cache_mm = r4k_flush_cache_mm_s128d16i16;
			_flush_cache_range = r4k_flush_cache_range_s128d16i16;
			_flush_cache_page = r4k_flush_cache_page_s128d16i16;
			break;
		case 32:
			_flush_cache_l1 = r4k_flush_cache_all_s128d32i32;
			_flush_cache_mm = r4k_flush_cache_mm_s128d32i32;
			_flush_cache_range = r4k_flush_cache_range_s128d32i32;
			_flush_cache_page = r4k_flush_cache_page_s128d32i32;
			break;
		};
		_flush_page_to_ram = r4k_flush_page_to_ram_s128;
		_clear_page = r4k_clear_page_s128;
		_copy_page = r4k_copy_page_s128;
		break;
	}
	_dma_cache_wback_inv = r4k_dma_cache_wback_inv_sc;
	_dma_cache_wback = r4k_dma_cache_wback;
	_dma_cache_inv = r4k_dma_cache_inv_sc;
}

typedef int (*probe_func_t)(unsigned long);

static inline void __init setup_scache(unsigned int config)
{
	probe_func_t probe_scache_kseg1;
	int sc_present = 0;

	/* Maybe the cpu knows about a l2 cache? */
	probe_scache_kseg1 = (probe_func_t) (KSEG1ADDR(&probe_scache));
	sc_present = probe_scache_kseg1(config);

	if (sc_present) {
		setup_scache_funcs();
		return;
	}

	setup_noscache_funcs();
}

void __init ld_mmu_r4xx0(void)
{
	unsigned long config = read_32bit_cp0_register(CP0_CONFIG);

	printk("CPU revision is: %08x\n", read_32bit_cp0_register(CP0_PRID));

#ifdef CONFIG_MIPS_UNCACHED
	set_cp0_config(CONF_CM_CMASK, CONF_CM_UNCACHED);
#else
	set_cp0_config(CONF_CM_CMASK, CONF_CM_CACHABLE_NONCOHERENT);
#endif /* UNCACHED */

	probe_icache(config);
	probe_dcache(config);
	setup_scache(config);

	switch(mips_cputype) {
	case CPU_R4600:			/* QED style two way caches? */
	case CPU_R4700:
	case CPU_R5000:
	case CPU_NEVADA:
		_flush_cache_page = r4k_flush_cache_page_d32i32_r4600;
	}

	_flush_cache_sigtramp = r4k_flush_cache_sigtramp;
	if ((read_32bit_cp0_register(CP0_PRID) & 0xfff0) == 0x2020) {
		_flush_cache_sigtramp = r4600v20k_flush_cache_sigtramp;
	}

	_flush_tlb_all = r4k_flush_tlb_all;
	_flush_tlb_mm = r4k_flush_tlb_mm;
	_flush_tlb_range = r4k_flush_tlb_range;
	_flush_tlb_page = r4k_flush_tlb_page;
	_flush_cache_l2 = r4k_flush_cache_l2;

	update_mmu_cache = r4k_update_mmu_cache;

	_show_regs = r4k_show_regs;

	flush_cache_l1();

	/*
	 * You should never change this register:
	 *   - On R4600 1.7 the tlbp never hits for pages smaller than
	 *     the value in the c0_pagemask register.
	 *   - The entire mm handling assumes the c0_pagemask register to
	 *     be set for 4kb pages.
	 */
	write_32bit_cp0_register(CP0_PAGEMASK, PM_4K);
	_flush_tlb_all();
}