srmmu.c

linux-2.6.15.6

/*
 * srmmu.c:  SRMMU specific routines for memory management.
 *
 * Copyright (C) 1995 David S. Miller  ([e-mail])
 * Copyright (C) 1995,2002 Pete Zaitcev ([e-mail])
 * Copyright (C) 1996 Eddie C. Dost    ([e-mail])
 * Copyright (C) 1997,1998 Jakub Jelinek ([e-mail])
 * Copyright (C) 1999,2000 Anton Blanchard ([e-mail])
 */

#include <linux/config.h>
#include <linux/kernel.h>
#include <linux/mm.h>
#include <linux/slab.h>
#include <linux/vmalloc.h>
#include <linux/pagemap.h>
#include <linux/init.h>
#include <linux/spinlock.h>
#include <linux/bootmem.h>
#include <linux/fs.h>
#include <linux/seq_file.h>

#include <asm/bitext.h>
#include <asm/page.h>
#include <asm/pgalloc.h>
#include <asm/pgtable.h>
#include <asm/io.h>
#include <asm/kdebug.h>
#include <asm/vaddrs.h>
#include <asm/traps.h>
#include <asm/smp.h>
#include <asm/mbus.h>
#include <asm/cache.h>
#include <asm/oplib.h>
#include <asm/sbus.h>
#include <asm/asi.h>
#include <asm/msi.h>
#include <asm/a.out.h>
#include <asm/mmu_context.h>
#include <asm/io-unit.h>
#include <asm/cacheflush.h>
#include <asm/tlbflush.h>

/* Now the cpu specific definitions. */
#include <asm/viking.h>
#include <asm/mxcc.h>
#include <asm/ross.h>
#include <asm/tsunami.h>
#include <asm/swift.h>
#include <asm/turbosparc.h>

#include <asm/btfixup.h>

enum mbus_module srmmu_modtype;
unsigned int hwbug_bitmask;
int vac_cache_size;
int vac_line_size;

extern struct resource sparc_iomap;

extern unsigned long last_valid_pfn;

extern unsigned long page_kernel;

pgd_t *srmmu_swapper_pg_dir;

#ifdef CONFIG_SMP
#define FLUSH_BEGIN(mm)
#define FLUSH_END
#else
#define FLUSH_BEGIN(mm) if((mm)->context != NO_CONTEXT) {
#define FLUSH_END	}
#endif

BTFIXUPDEF_CALL(void, flush_page_for_dma, unsigned long)
#define flush_page_for_dma(page) BTFIXUP_CALL(flush_page_for_dma)(page)

int flush_page_for_dma_global = 1;

#ifdef CONFIG_SMP
BTFIXUPDEF_CALL(void, local_flush_page_for_dma, unsigned long)
#define local_flush_page_for_dma(page) BTFIXUP_CALL(local_flush_page_for_dma)(page)
#endif

char *srmmu_name;

ctxd_t *srmmu_ctx_table_phys;
ctxd_t *srmmu_context_table;

int viking_mxcc_present;
static DEFINE_SPINLOCK(srmmu_context_spinlock);

int is_hypersparc;

/*
 * In general all page table modifications should use the V8 atomic
 * swap instruction.  This insures the mmu and the cpu are in sync
 * with respect to ref/mod bits in the page tables.
 */
static inline unsigned long srmmu_swap(unsigned long *addr, unsigned long value)
{
	__asm__ __volatile__("swap [%2], %0" : "=&r" (value) : "0" (value), "r" (addr));
	return value;
}

static inline void srmmu_set_pte(pte_t *ptep, pte_t pteval)
{
	srmmu_swap((unsigned long *)ptep, pte_val(pteval));
}

/* The very generic SRMMU page table operations. */
static inline int srmmu_device_memory(unsigned long x)
{
	return ((x & 0xF0000000) != 0);
}

int srmmu_cache_pagetables;

/* these will be initialized in srmmu_nocache_calcsize() */
unsigned long srmmu_nocache_size;
unsigned long srmmu_nocache_end;

/* 1 bit <=> 256 bytes of nocache <=> 64 PTEs */
#define SRMMU_NOCACHE_BITMAP_SHIFT (PAGE_SHIFT - 4)

/* The context table is a nocache user with the biggest alignment needs. */
#define SRMMU_NOCACHE_ALIGN_MAX (sizeof(ctxd_t)*SRMMU_MAX_CONTEXTS)

void *srmmu_nocache_pool;
void *srmmu_nocache_bitmap;
static struct bit_map srmmu_nocache_map;

static unsigned long srmmu_pte_pfn(pte_t pte)
{
	if (srmmu_device_memory(pte_val(pte))) {
		/* Just return something that will cause
		 * pfn_valid() to return false.  This makes
		 * copy_one_pte() to just directly copy to
		 * PTE over.
		 */
		return ~0UL;
	}
	return (pte_val(pte) & SRMMU_PTE_PMASK) >> (PAGE_SHIFT-4);
}

static struct page *srmmu_pmd_page(pmd_t pmd)
{

	if (srmmu_device_memory(pmd_val(pmd)))
		BUG();
	return pfn_to_page((pmd_val(pmd) & SRMMU_PTD_PMASK) >> (PAGE_SHIFT-4));
}

static inline unsigned long srmmu_pgd_page(pgd_t pgd)
{ return srmmu_device_memory(pgd_val(pgd))?~0:(unsigned long)__nocache_va((pgd_val(pgd) & SRMMU_PTD_PMASK) << 4); }


static inline int srmmu_pte_none(pte_t pte)
{ return !(pte_val(pte) & 0xFFFFFFF); }

static inline int srmmu_pte_present(pte_t pte)
{ return ((pte_val(pte) & SRMMU_ET_MASK) == SRMMU_ET_PTE); }

static inline int srmmu_pte_read(pte_t pte)
{ return !(pte_val(pte) & SRMMU_NOREAD); }

static inline void srmmu_pte_clear(pte_t *ptep)
{ srmmu_set_pte(ptep, __pte(0)); }

static inline int srmmu_pmd_none(pmd_t pmd)
{ return !(pmd_val(pmd) & 0xFFFFFFF); }

static inline int srmmu_pmd_bad(pmd_t pmd)
{ return (pmd_val(pmd) & SRMMU_ET_MASK) != SRMMU_ET_PTD; }

static inline int srmmu_pmd_present(pmd_t pmd)
{ return ((pmd_val(pmd) & SRMMU_ET_MASK) == SRMMU_ET_PTD); }

static inline void srmmu_pmd_clear(pmd_t *pmdp) {
	int i;
	for (i = 0; i < PTRS_PER_PTE/SRMMU_REAL_PTRS_PER_PTE; i++)
		srmmu_set_pte((pte_t *)&pmdp->pmdv[i], __pte(0));
}

static inline int srmmu_pgd_none(pgd_t pgd)          
{ return !(pgd_val(pgd) & 0xFFFFFFF); }

static inline int srmmu_pgd_bad(pgd_t pgd)
{ return (pgd_val(pgd) & SRMMU_ET_MASK) != SRMMU_ET_PTD; }

static inline int srmmu_pgd_present(pgd_t pgd)
{ return ((pgd_val(pgd) & SRMMU_ET_MASK) == SRMMU_ET_PTD); }

static inline void srmmu_pgd_clear(pgd_t * pgdp)
{ srmmu_set_pte((pte_t *)pgdp, __pte(0)); }

static inline pte_t srmmu_pte_wrprotect(pte_t pte)
{ return __pte(pte_val(pte) & ~SRMMU_WRITE);}

static inline pte_t srmmu_pte_mkclean(pte_t pte)
{ return __pte(pte_val(pte) & ~SRMMU_DIRTY);}

static inline pte_t srmmu_pte_mkold(pte_t pte)
{ return __pte(pte_val(pte) & ~SRMMU_REF);}

static inline pte_t srmmu_pte_mkwrite(pte_t pte)
{ return __pte(pte_val(pte) | SRMMU_WRITE);}

static inline pte_t srmmu_pte_mkdirty(pte_t pte)
{ return __pte(pte_val(pte) | SRMMU_DIRTY);}

static inline pte_t srmmu_pte_mkyoung(pte_t pte)
{ return __pte(pte_val(pte) | SRMMU_REF);}

/*
 * Conversion functions: convert a page and protection to a page entry,
 * and a page entry and page directory to the page they refer to.
 */
static pte_t srmmu_mk_pte(struct page *page, pgprot_t pgprot)
{ return __pte((page_to_pfn(page) << (PAGE_SHIFT-4)) | pgprot_val(pgprot)); }

static pte_t srmmu_mk_pte_phys(unsigned long page, pgprot_t pgprot)
{ return __pte(((page) >> 4) | pgprot_val(pgprot)); }

static pte_t srmmu_mk_pte_io(unsigned long page, pgprot_t pgprot, int space)
{ return __pte(((page) >> 4) | (space << 28) | pgprot_val(pgprot)); }

/* XXX should we hyper_flush_whole_icache here - Anton */
static inline void srmmu_ctxd_set(ctxd_t *ctxp, pgd_t *pgdp)
{ srmmu_set_pte((pte_t *)ctxp, (SRMMU_ET_PTD | (__nocache_pa((unsigned long) pgdp) >> 4))); }

static inline void srmmu_pgd_set(pgd_t * pgdp, pmd_t * pmdp)
{ srmmu_set_pte((pte_t *)pgdp, (SRMMU_ET_PTD | (__nocache_pa((unsigned long) pmdp) >> 4))); }

static void srmmu_pmd_set(pmd_t *pmdp, pte_t *ptep)
{
	unsigned long ptp;	/* Physical address, shifted right by 4 */
	int i;

	ptp = __nocache_pa((unsigned long) ptep) >> 4;
	for (i = 0; i < PTRS_PER_PTE/SRMMU_REAL_PTRS_PER_PTE; i++) {
		srmmu_set_pte((pte_t *)&pmdp->pmdv[i], SRMMU_ET_PTD | ptp);
		ptp += (SRMMU_REAL_PTRS_PER_PTE*sizeof(pte_t) >> 4);
	}
}

static void srmmu_pmd_populate(pmd_t *pmdp, struct page *ptep)
{
	unsigned long ptp;	/* Physical address, shifted right by 4 */
	int i;

	ptp = page_to_pfn(ptep) << (PAGE_SHIFT-4);	/* watch for overflow */
	for (i = 0; i < PTRS_PER_PTE/SRMMU_REAL_PTRS_PER_PTE; i++) {
		srmmu_set_pte((pte_t *)&pmdp->pmdv[i], SRMMU_ET_PTD | ptp);
		ptp += (SRMMU_REAL_PTRS_PER_PTE*sizeof(pte_t) >> 4);
	}
}

static inline pte_t srmmu_pte_modify(pte_t pte, pgprot_t newprot)
{ return __pte((pte_val(pte) & SRMMU_CHG_MASK) | pgprot_val(newprot)); }

/* to find an entry in a top-level page table... */
static inline pgd_t *srmmu_pgd_offset(struct mm_struct * mm, unsigned long address)
{ return mm->pgd + (address >> SRMMU_PGDIR_SHIFT); }

/* Find an entry in the second-level page table.. */
static inline pmd_t *srmmu_pmd_offset(pgd_t * dir, unsigned long address)
{
	return (pmd_t *) srmmu_pgd_page(*dir) +
	    ((address >> PMD_SHIFT) & (PTRS_PER_PMD - 1));
}

/* Find an entry in the third-level page table.. */ 
static inline pte_t *srmmu_pte_offset(pmd_t * dir, unsigned long address)
{
	void *pte;

	pte = __nocache_va((dir->pmdv[0] & SRMMU_PTD_PMASK) << 4);
	return (pte_t *) pte +
	    ((address >> PAGE_SHIFT) & (PTRS_PER_PTE - 1));
}

static unsigned long srmmu_swp_type(swp_entry_t entry)
{
	return (entry.val >> SRMMU_SWP_TYPE_SHIFT) & SRMMU_SWP_TYPE_MASK;
}

static unsigned long srmmu_swp_offset(swp_entry_t entry)
{
	return (entry.val >> SRMMU_SWP_OFF_SHIFT) & SRMMU_SWP_OFF_MASK;
}

static swp_entry_t srmmu_swp_entry(unsigned long type, unsigned long offset)
{
	return (swp_entry_t) {
		  (type & SRMMU_SWP_TYPE_MASK) << SRMMU_SWP_TYPE_SHIFT
		| (offset & SRMMU_SWP_OFF_MASK) << SRMMU_SWP_OFF_SHIFT };
}

/*
 * size: bytes to allocate in the nocache area.
 * align: bytes, number to align at.
 * Returns the virtual address of the allocated area.
 */
static unsigned long __srmmu_get_nocache(int size, int align)
{
	int offset;

	if (size < SRMMU_NOCACHE_BITMAP_SHIFT) {
		printk("Size 0x%x too small for nocache request\n", size);
		size = SRMMU_NOCACHE_BITMAP_SHIFT;
	}
	if (size & (SRMMU_NOCACHE_BITMAP_SHIFT-1)) {
		printk("Size 0x%x unaligned int nocache request\n", size);
		size += SRMMU_NOCACHE_BITMAP_SHIFT-1;
	}
	BUG_ON(align > SRMMU_NOCACHE_ALIGN_MAX);

	offset = bit_map_string_get(&srmmu_nocache_map,
		       			size >> SRMMU_NOCACHE_BITMAP_SHIFT,
					align >> SRMMU_NOCACHE_BITMAP_SHIFT);
	if (offset == -1) {
		printk("srmmu: out of nocache %d: %d/%d\n",
		    size, (int) srmmu_nocache_size,
		    srmmu_nocache_map.used << SRMMU_NOCACHE_BITMAP_SHIFT);
		return 0;
	}

	return (SRMMU_NOCACHE_VADDR + (offset << SRMMU_NOCACHE_BITMAP_SHIFT));
}

unsigned inline long srmmu_get_nocache(int size, int align)
{
	unsigned long tmp;

	tmp = __srmmu_get_nocache(size, align);

	if (tmp)
		memset((void *)tmp, 0, size);

	return tmp;
}

void srmmu_free_nocache(unsigned long vaddr, int size)
{
	int offset;

	if (vaddr < SRMMU_NOCACHE_VADDR) {
		printk("Vaddr %lx is smaller than nocache base 0x%lx\n",
		    vaddr, (unsigned long)SRMMU_NOCACHE_VADDR);
		BUG();
	}
	if (vaddr+size > srmmu_nocache_end) {
		printk("Vaddr %lx is bigger than nocache end 0x%lx\n",
		    vaddr, srmmu_nocache_end);
		BUG();
	}
	if (size & (size-1)) {
		printk("Size 0x%x is not a power of 2\n", size);
		BUG();
	}
	if (size < SRMMU_NOCACHE_BITMAP_SHIFT) {
		printk("Size 0x%x is too small\n", size);
		BUG();
	}
	if (vaddr & (size-1)) {
		printk("Vaddr %lx is not aligned to size 0x%x\n", vaddr, size);
		BUG();
	}

	offset = (vaddr - SRMMU_NOCACHE_VADDR) >> SRMMU_NOCACHE_BITMAP_SHIFT;
	size = size >> SRMMU_NOCACHE_BITMAP_SHIFT;

	bit_map_clear(&srmmu_nocache_map, offset, size);
}

void srmmu_early_allocate_ptable_skeleton(unsigned long start, unsigned long end);

extern unsigned long probe_memory(void);	/* in fault.c */

/*
 * Reserve nocache dynamically proportionally to the amount of
 * system RAM. -- Tomas Szepe <[e-mail]>, June 2002
 */
void srmmu_nocache_calcsize(void)
{
	unsigned long sysmemavail = probe_memory() / 1024;
	int srmmu_nocache_npages;

	srmmu_nocache_npages =
		sysmemavail / SRMMU_NOCACHE_ALCRATIO / 1024 * 256;

 /* P3 XXX The 4x overuse: corroborated by /proc/meminfo. */
	// if (srmmu_nocache_npages < 256) srmmu_nocache_npages = 256;
	if (srmmu_nocache_npages < SRMMU_MIN_NOCACHE_PAGES)
		srmmu_nocache_npages = SRMMU_MIN_NOCACHE_PAGES;

	/* anything above 1280 blows up */
	if (srmmu_nocache_npages > SRMMU_MAX_NOCACHE_PAGES)
		srmmu_nocache_npages = SRMMU_MAX_NOCACHE_PAGES;

	srmmu_nocache_size = srmmu_nocache_npages * PAGE_SIZE;
	srmmu_nocache_end = SRMMU_NOCACHE_VADDR + srmmu_nocache_size;
}

void srmmu_nocache_init(void)
{
	unsigned int bitmap_bits;
	pgd_t *pgd;
	pmd_t *pmd;
	pte_t *pte;
	unsigned long paddr, vaddr;
	unsigned long pteval;

	bitmap_bits = srmmu_nocache_size >> SRMMU_NOCACHE_BITMAP_SHIFT;

	srmmu_nocache_pool = __alloc_bootmem(srmmu_nocache_size,
		SRMMU_NOCACHE_ALIGN_MAX, 0UL);
	memset(srmmu_nocache_pool, 0, srmmu_nocache_size);

	srmmu_nocache_bitmap = __alloc_bootmem(bitmap_bits >> 3, SMP_CACHE_BYTES, 0UL);
	bit_map_init(&srmmu_nocache_map, srmmu_nocache_bitmap, bitmap_bits);

	srmmu_swapper_pg_dir = (pgd_t *)__srmmu_get_nocache(SRMMU_PGD_TABLE_SIZE, SRMMU_PGD_TABLE_SIZE);
	memset(__nocache_fix(srmmu_swapper_pg_dir), 0, SRMMU_PGD_TABLE_SIZE);
	init_mm.pgd = srmmu_swapper_pg_dir;

	srmmu_early_allocate_ptable_skeleton(SRMMU_NOCACHE_VADDR, srmmu_nocache_end);

	paddr = __pa((unsigned long)srmmu_nocache_pool);
	vaddr = SRMMU_NOCACHE_VADDR;

	while (vaddr < srmmu_nocache_end) {
		pgd = pgd_offset_k(vaddr);
		pmd = srmmu_pmd_offset(__nocache_fix(pgd), vaddr);
		pte = srmmu_pte_offset(__nocache_fix(pmd), vaddr);

		pteval = ((paddr >> 4) | SRMMU_ET_PTE | SRMMU_PRIV);

		if (srmmu_cache_pagetables)
			pteval |= SRMMU_CACHE;