/**

  * $Id$

  * 

  * Translation cache management. This part is architecture independent.

  *

  * Copyright (c) 2005 Nathan Keynes.

  *

  * This program is free software; you can redistribute it and/or modify

  * it under the terms of the GNU General Public License as published by

  * the Free Software Foundation; either version 2 of the License, or

  * (at your option) any later version.

  *

  * This program is distributed in the hope that it will be useful,

  * but WITHOUT ANY WARRANTY; without even the implied warranty of

  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the

  * GNU General Public License for more details.

  */

 #include <sys/types.h>

 #include <sys/mman.h>

 #include <assert.h>

 #include "dreamcast.h"

 #include "sh4/sh4core.h"

 #include "sh4/xltcache.h"

 #include "x86dasm/x86dasm.h"

 #define XLAT_LUT_PAGE_BITS 12

 #define XLAT_LUT_TOTAL_BITS 28

 #define XLAT_LUT_PAGE(addr) (((addr)>>13) & 0xFFFF)

 #define XLAT_LUT_ENTRY(addr) (((addr)&0x1FFE) >> 1)

 #define XLAT_LUT_PAGES (1<<(XLAT_LUT_TOTAL_BITS-XLAT_LUT_PAGE_BITS))

 #define XLAT_LUT_PAGE_ENTRIES (1<<XLAT_LUT_PAGE_BITS)

 #define XLAT_LUT_PAGE_SIZE (XLAT_LUT_PAGE_ENTRIES * sizeof(void *))

 #define XLAT_LUT_ENTRY_EMPTY (void *)0

 #define XLAT_LUT_ENTRY_USED  (void *)1

 #define NEXT(block) ( (xlat_cache_block_t)&((block)->code[(block)->size]))

 #define BLOCK_FOR_CODE(code) (((xlat_cache_block_t)code)-1)

 #define IS_ENTRY_POINT(ent) (ent > XLAT_LUT_ENTRY_USED)

 #define IS_ENTRY_USED(ent) (ent != XLAT_LUT_ENTRY_EMPTY)

 #define MIN_BLOCK_SIZE 32

 #define MIN_TOTAL_SIZE (sizeof(struct xlat_cache_block)+MIN_BLOCK_SIZE)

 #define BLOCK_INACTIVE 0

 #define BLOCK_ACTIVE 1

 #define BLOCK_USED 2

 xlat_cache_block_t xlat_new_cache;

 xlat_cache_block_t xlat_new_cache_ptr;

 xlat_cache_block_t xlat_new_create_ptr;

 xlat_cache_block_t xlat_temp_cache;

 xlat_cache_block_t xlat_temp_cache_ptr;

 xlat_cache_block_t xlat_old_cache;

 xlat_cache_block_t xlat_old_cache_ptr;

 static void ***xlat_lut;

 static gboolean xlat_initialized = FALSE;

 void xlat_cache_init(void) 

 {

     if( !xlat_initialized ) {

         xlat_initialized = TRUE;

         xlat_new_cache = mmap( NULL, XLAT_NEW_CACHE_SIZE, PROT_EXEC|PROT_READ|PROT_WRITE,

                 MAP_PRIVATE|MAP_ANON, -1, 0 );

         xlat_temp_cache = mmap( NULL, XLAT_TEMP_CACHE_SIZE, PROT_EXEC|PROT_READ|PROT_WRITE,

                 MAP_PRIVATE|MAP_ANON, -1, 0 );

         xlat_old_cache = mmap( NULL, XLAT_OLD_CACHE_SIZE, PROT_EXEC|PROT_READ|PROT_WRITE,

                 MAP_PRIVATE|MAP_ANON, -1, 0 );

         xlat_new_cache_ptr = xlat_new_cache;

         xlat_temp_cache_ptr = xlat_temp_cache;

         xlat_old_cache_ptr = xlat_old_cache;

         xlat_new_create_ptr = xlat_new_cache;

         xlat_lut = mmap( NULL, XLAT_LUT_PAGES*sizeof(void *), PROT_READ|PROT_WRITE,

                 MAP_PRIVATE|MAP_ANON, -1, 0);

         memset( xlat_lut, 0, XLAT_LUT_PAGES*sizeof(void *) );

     }

     xlat_flush_cache();

 }

 void xlat_print_free( FILE *out )

 {

     fprintf( out, "New space: %d\nTemp space: %d\nOld space: %d\n", 

             xlat_new_cache_ptr->size, xlat_temp_cache_ptr->size, xlat_old_cache_ptr->size );

 }

 /**

  * Reset the cache structure to its default state

  */

 void xlat_flush_cache() 

 {

     xlat_cache_block_t tmp;

     int i;

     xlat_new_cache_ptr = xlat_new_cache;

     xlat_new_cache_ptr->active = 0;

     xlat_new_cache_ptr->size = XLAT_NEW_CACHE_SIZE - 2*sizeof(struct xlat_cache_block);

     tmp = NEXT(xlat_new_cache_ptr);

     tmp->active = 1;

     tmp->size = 0;

     xlat_temp_cache_ptr = xlat_temp_cache;

     xlat_temp_cache_ptr->active = 0;

     xlat_temp_cache_ptr->size = XLAT_TEMP_CACHE_SIZE - 2*sizeof(struct xlat_cache_block);

     tmp = NEXT(xlat_temp_cache_ptr);

     tmp->active = 1;

     tmp->size = 0;

     xlat_old_cache_ptr = xlat_old_cache;

     xlat_old_cache_ptr->active = 0;

     xlat_old_cache_ptr->size = XLAT_OLD_CACHE_SIZE - 2*sizeof(struct xlat_cache_block);

     tmp = NEXT(xlat_old_cache_ptr);

     tmp->active = 1;

     tmp->size = 0;

     for( i=0; i<XLAT_LUT_PAGES; i++ ) {

         if( xlat_lut[i] != NULL ) {

             memset( xlat_lut[i], 0, XLAT_LUT_PAGE_SIZE );

         }

     }

 }

 static void xlat_flush_page_by_lut( void **page )

 {

     int i;

     for( i=0; i<XLAT_LUT_PAGE_ENTRIES; i++ ) {

         if( IS_ENTRY_POINT(page[i]) ) {

             BLOCK_FOR_CODE(page[i])->active = 0;

         }

         page[i] = NULL;

     }

 }

 void xlat_invalidate_word( sh4addr_t addr )

 {

     if( xlat_lut ) {

         void **page = xlat_lut[XLAT_LUT_PAGE(addr)];

         if( page != NULL ) {

             int entry = XLAT_LUT_ENTRY(addr);

             if( page[entry] != NULL ) {

                 xlat_flush_page_by_lut(page);

             }

         }

     }

 }

 void xlat_invalidate_long( sh4addr_t addr )

 {

     if( xlat_lut ) {

         void **page = xlat_lut[XLAT_LUT_PAGE(addr)];

         if( page != NULL ) {

             int entry = XLAT_LUT_ENTRY(addr);

             if( page[entry] != NULL || page[entry+1] != NULL ) {

                 xlat_flush_page_by_lut(page);

             }

         }

     }

 }

 void xlat_invalidate_block( sh4addr_t address, size_t size )

 {

     int i;

     int entry_count = size >> 1; // words;

     uint32_t page_no = XLAT_LUT_PAGE(address);

     int entry = XLAT_LUT_ENTRY(address);

     if( xlat_lut ) {

         do {

             void **page = xlat_lut[page_no];

             int page_entries = XLAT_LUT_PAGE_ENTRIES - entry;

             if( entry_count < page_entries ) {

                 page_entries = entry_count;

             }

             if( page != NULL ) {

                 if( page_entries == XLAT_LUT_PAGE_ENTRIES ) {

                     /* Overwriting the entire page anyway */

                     xlat_flush_page_by_lut(page);

                 } else {

                     for( i=entry; i<entry+page_entries; i++ ) {

                         if( page[i] != NULL ) {

                             xlat_flush_page_by_lut(page);

                             break;

                         }

                     }

                 }

                 entry_count -= page_entries;

             }

             page_no ++;

             entry_count -= page_entries;

             entry = 0;

         } while( entry_count > 0 );

     }

 }

 void xlat_flush_page( sh4addr_t address )

 {

     void **page = xlat_lut[XLAT_LUT_PAGE(address)];

     if( page != NULL ) {

         xlat_flush_page_by_lut(page);

     }

 }

 void *xlat_get_code( sh4addr_t address )

 {

     void *result = NULL;

     void **page = xlat_lut[XLAT_LUT_PAGE(address)];

     if( page != NULL ) {

         result = (void *)(((uintptr_t)(page[XLAT_LUT_ENTRY(address)])) & (~((uintptr_t)0x03)));

     }

     return result;

 }

 xlat_recovery_record_t xlat_get_post_recovery( void *code, void *native_pc, gboolean with_terminal )

 {

     if( code != NULL ) {

         uintptr_t pc_offset = ((uint8_t *)native_pc) - ((uint8_t *)code);

         xlat_cache_block_t block = BLOCK_FOR_CODE(code);

         uint32_t count = block->recover_table_size;

         xlat_recovery_record_t records = (xlat_recovery_record_t)(&block->code[block->recover_table_offset]);

         uint32_t posn;

         if( count > 0 && !with_terminal )

         	count--;

         if( records[count-1].xlat_offset < pc_offset ) {

         	return NULL;

         }

         for( posn=count-1; posn > 0; posn-- ) {

         	if( records[posn-1].xlat_offset < pc_offset ) {

         		return &records[posn];

         	}

         }

         return &records[0]; // shouldn't happen

     }

     return NULL;

 }

 xlat_recovery_record_t xlat_get_pre_recovery( void *code, void *native_pc )

 {

     if( code != NULL ) {

         uintptr_t pc_offset = ((uint8_t *)native_pc) - ((uint8_t *)code);

         xlat_cache_block_t block = BLOCK_FOR_CODE(code);

         uint32_t count = block->recover_table_size;

         xlat_recovery_record_t records = (xlat_recovery_record_t)(&block->code[block->recover_table_offset]);

         uint32_t posn;

         for( posn = 1; posn < count; posn++ ) {

         	if( records[posn].xlat_offset >= pc_offset ) {

         		return &records[posn-1];

         	}

         }

         return &records[count-1];

     }

     return NULL;	

 }

 void **xlat_get_lut_entry( sh4addr_t address )

 {

     void **page = xlat_lut[XLAT_LUT_PAGE(address)];

     /* Add the LUT entry for the block */

     if( page == NULL ) {

         xlat_lut[XLAT_LUT_PAGE(address)] = page =

             mmap( NULL, XLAT_LUT_PAGE_SIZE, PROT_READ|PROT_WRITE,

                     MAP_PRIVATE|MAP_ANON, -1, 0 );

         memset( page, 0, XLAT_LUT_PAGE_SIZE );

     }

     return &page[XLAT_LUT_ENTRY(address)];

 }

 uint32_t xlat_get_block_size( void *block )

 {

     xlat_cache_block_t xlt = (xlat_cache_block_t)(((char *)block)-sizeof(struct xlat_cache_block));

     return xlt->size;

 }

 uint32_t xlat_get_code_size( void *block )

 {

     xlat_cache_block_t xlt = (xlat_cache_block_t)(((char *)block)-sizeof(struct xlat_cache_block));

     if( xlt->recover_table_offset == 0 ) {

         return xlt->size;

     } else {

         return xlt->recover_table_offset;

     }

 }

 /**

  * Cut the specified block so that it has the given size, with the remaining data

  * forming a new free block. If the free block would be less than the minimum size,

  * the cut is not performed.

  * @return the next block after the (possibly cut) block.

  */

 static inline xlat_cache_block_t xlat_cut_block( xlat_cache_block_t block, int cutsize )

 {

     cutsize = (cutsize + 3) & 0xFFFFFFFC; // force word alignment

     assert( cutsize <= block->size );

     if( block->size > cutsize + MIN_TOTAL_SIZE ) {

         int oldsize = block->size;

         block->size = cutsize;

         xlat_cache_block_t next = NEXT(block);

         next->active = 0;

         next->size = oldsize - cutsize - sizeof(struct xlat_cache_block);

         return next;

     } else {

         return NEXT(block);

     }

 }

 /**

  * Promote a block in temp space (or elsewhere for that matter) to old space.

  *

  * @param block to promote.

  */

 static void xlat_promote_to_old_space( xlat_cache_block_t block )

 {

     int allocation = (int)-sizeof(struct xlat_cache_block);

     int size = block->size;

     xlat_cache_block_t curr = xlat_old_cache_ptr;

     xlat_cache_block_t start_block = curr;

     do {

         allocation += curr->size + sizeof(struct xlat_cache_block);

         curr = NEXT(curr);

         if( allocation > size ) {

             break; /* done */

         }

         if( curr->size == 0 ) { /* End-of-cache Sentinel */

             /* Leave what we just released as free space and start again from the

              * top of the cache

              */

             start_block->active = 0;

             start_block->size = allocation;

             allocation = (int)-sizeof(struct xlat_cache_block);

             start_block = curr = xlat_old_cache;

         }

     } while(1);

     start_block->active = 1;

     start_block->size = allocation;

     start_block->lut_entry = block->lut_entry;

     start_block->recover_table_offset = block->recover_table_offset;

     start_block->recover_table_size = block->recover_table_size;

     *block->lut_entry = &start_block->code;

     memcpy( start_block->code, block->code, block->size );

     xlat_old_cache_ptr = xlat_cut_block(start_block, size );

     if( xlat_old_cache_ptr->size == 0 ) {

         xlat_old_cache_ptr = xlat_old_cache;

     }

 }

 /**

  * Similarly to the above method, promotes a block to temp space.

  * TODO: Try to combine these - they're nearly identical

  */

 void xlat_promote_to_temp_space( xlat_cache_block_t block )

 {

     int size = block->size;

     int allocation = (int)-sizeof(struct xlat_cache_block);

     xlat_cache_block_t curr = xlat_temp_cache_ptr;

     xlat_cache_block_t start_block = curr;

     do {

         if( curr->active == BLOCK_USED ) {

             xlat_promote_to_old_space( curr );

         } else if( curr->active == BLOCK_ACTIVE ) {

             // Active but not used, release block

             *((uintptr_t *)curr->lut_entry) &= ((uintptr_t)0x03);

         }

         allocation += curr->size + sizeof(struct xlat_cache_block);

         curr = NEXT(curr);

         if( allocation > size ) {

             break; /* done */

         }

         if( curr->size == 0 ) { /* End-of-cache Sentinel */

             /* Leave what we just released as free space and start again from the

              * top of the cache

              */

             start_block->active = 0;

             start_block->size = allocation;

             allocation = (int)-sizeof(struct xlat_cache_block);

             start_block = curr = xlat_temp_cache;

         }

     } while(1);

     start_block->active = 1;

     start_block->size = allocation;

     start_block->lut_entry = block->lut_entry;

     start_block->recover_table_offset = block->recover_table_offset;

     start_block->recover_table_size = block->recover_table_size;

     *block->lut_entry = &start_block->code;

     memcpy( start_block->code, block->code, block->size );

     xlat_temp_cache_ptr = xlat_cut_block(start_block, size );

     if( xlat_temp_cache_ptr->size == 0 ) {

         xlat_temp_cache_ptr = xlat_temp_cache;

     }

 }

 /**

  * Returns the next block in the new cache list that can be written to by the

  * translator. If the next block is active, it is evicted first.

  */

 xlat_cache_block_t xlat_start_block( sh4addr_t address )

 {

     if( xlat_new_cache_ptr->size == 0 ) {

         xlat_new_cache_ptr = xlat_new_cache;

     }

     if( xlat_new_cache_ptr->active ) {

         xlat_promote_to_temp_space( xlat_new_cache_ptr );

     }

     xlat_new_create_ptr = xlat_new_cache_ptr;

     xlat_new_create_ptr->active = 1;

     xlat_new_cache_ptr = NEXT(xlat_new_cache_ptr);

     /* Add the LUT entry for the block */

     if( xlat_lut[XLAT_LUT_PAGE(address)] == NULL ) {

         xlat_lut[XLAT_LUT_PAGE(address)] =

             mmap( NULL, XLAT_LUT_PAGE_SIZE, PROT_READ|PROT_WRITE,

                     MAP_PRIVATE|MAP_ANON, -1, 0 );

         memset( xlat_lut[XLAT_LUT_PAGE(address)], 0, XLAT_LUT_PAGE_SIZE );

     }

     if( IS_ENTRY_POINT(xlat_lut[XLAT_LUT_PAGE(address)][XLAT_LUT_ENTRY(address)]) ) {

         xlat_cache_block_t oldblock = BLOCK_FOR_CODE(xlat_lut[XLAT_LUT_PAGE(address)][XLAT_LUT_ENTRY(address)]);

         oldblock->active = 0;

     }

     xlat_lut[XLAT_LUT_PAGE(address)][XLAT_LUT_ENTRY(address)] = 

         &xlat_new_create_ptr->code;

     xlat_new_create_ptr->lut_entry = xlat_lut[XLAT_LUT_PAGE(address)] + XLAT_LUT_ENTRY(address);

     return xlat_new_create_ptr;

 }

 xlat_cache_block_t xlat_extend_block( uint32_t newSize )

 {

     while( xlat_new_create_ptr->size < newSize ) {

         if( xlat_new_cache_ptr->size == 0 ) {

             /* Migrate to the front of the cache to keep it contiguous */

             xlat_new_create_ptr->active = 0;

             sh4ptr_t olddata = xlat_new_create_ptr->code;

             int oldsize = xlat_new_create_ptr->size;

             int size = oldsize + MIN_BLOCK_SIZE; /* minimum expansion */

             void **lut_entry = xlat_new_create_ptr->lut_entry;

             int allocation = (int)-sizeof(struct xlat_cache_block);

             xlat_new_cache_ptr = xlat_new_cache;

             do {

                 if( xlat_new_cache_ptr->active ) {

                     xlat_promote_to_temp_space( xlat_new_cache_ptr );

                 }

                 allocation += xlat_new_cache_ptr->size + sizeof(struct xlat_cache_block);

                 xlat_new_cache_ptr = NEXT(xlat_new_cache_ptr);

             } while( allocation < size );

             xlat_new_create_ptr = xlat_new_cache;

             xlat_new_create_ptr->active = 1;

             xlat_new_create_ptr->size = allocation;

             xlat_new_create_ptr->lut_entry = lut_entry;

             *lut_entry = &xlat_new_create_ptr->code;

             memmove( xlat_new_create_ptr->code, olddata, oldsize );

         } else {

             if( xlat_new_cache_ptr->active ) {

                 xlat_promote_to_temp_space( xlat_new_cache_ptr );

             }

             xlat_new_create_ptr->size += xlat_new_cache_ptr->size + sizeof(struct xlat_cache_block);

             xlat_new_cache_ptr = NEXT(xlat_new_cache_ptr);

         }

     }

     return xlat_new_create_ptr;

 }

 void xlat_commit_block( uint32_t destsize, uint32_t srcsize )

 {

     void **ptr = xlat_new_create_ptr->lut_entry;

     void **endptr = ptr + (srcsize>>2);

     while( ptr < endptr ) {

         if( *ptr == NULL ) {

             *ptr = XLAT_LUT_ENTRY_USED;

         }

         ptr++;

     }

     xlat_new_cache_ptr = xlat_cut_block( xlat_new_create_ptr, destsize );

 }

 void xlat_delete_block( xlat_cache_block_t block ) 

 {

     block->active = 0;

     *block->lut_entry = NULL;

 }

 void xlat_check_cache_integrity( xlat_cache_block_t cache, xlat_cache_block_t ptr, int size )

 {

     int foundptr = 0;

     xlat_cache_block_t tail = 

         (xlat_cache_block_t)(((char *)cache) + size - sizeof(struct xlat_cache_block));

     assert( tail->active == 1 );

     assert( tail->size == 0 ); 

     while( cache < tail ) {

         assert( cache->active >= 0 && cache->active <= 2 );

         assert( cache->size >= 0 && cache->size < size );

         if( cache == ptr ) {

             foundptr = 1;

         }

         cache = NEXT(cache);

     }

     assert( cache == tail );

     assert( foundptr == 1 || tail == ptr );

 }

 void xlat_check_integrity( )

 {

     xlat_check_cache_integrity( xlat_new_cache, xlat_new_cache_ptr, XLAT_NEW_CACHE_SIZE );

     xlat_check_cache_integrity( xlat_temp_cache, xlat_temp_cache_ptr, XLAT_TEMP_CACHE_SIZE );

     xlat_check_cache_integrity( xlat_old_cache, xlat_old_cache_ptr, XLAT_OLD_CACHE_SIZE );

 }