- Round up modification time to the next even second because of FAT's

[AROS.git] / rom / filesys / fat / fat.c

/*
 * fat-handler - FAT12/16/32 filesystem handler
 *
 * Copyright © 2006 Marek Szyprowski
 * Copyright © 2007-2015 The AROS Development Team
 *
 * This program is free software; you can redistribute it and/or modify it
 * under the same terms as AROS itself.
 *
 * $Id$
 */

#include <aros/macros.h>
#include <exec/errors.h>
#include <exec/types.h>
#include <dos/dos.h>
#include <dos/dosextens.h>
#include <dos/filehandler.h>

#include <proto/exec.h>
#include <proto/dos.h>
#include <proto/utility.h>

#include <clib/macros.h>

#include <string.h>   
#include <ctype.h>

#include "fat_fs.h"
#include "fat_protos.h"

#define DEBUG DEBUG_MISC
#include "debug.h"

/* helper function to get the location of a fat entry for a cluster. it used
 * to be a define until it got too crazy */
static UBYTE *GetFatEntryPtr(struct FSSuper *sb, ULONG offset, APTR *rb,
    UWORD fat_no) {
    ULONG entry_cache_block = offset >> sb->fat_cachesize_bits;
    ULONG entry_cache_offset = offset & (sb->fat_cachesize - 1);
    ULONG num;
    UWORD i;

    /* if the target cluster is not within the currently loaded chunk of fat,
     * we need to get the right data in */
    if (sb->fat_cache_block != entry_cache_block
        || sb->fat_cache_no != fat_no) {
        D(bug("[fat] loading %ld FAT sectors starting at sector %ld\n", sb->fat_blocks_count, entry_cache_block << (sb->fat_cachesize_bits - sb->sectorsize_bits)));
        /* put the old ones back */
        if (sb->fat_cache_block != 0xffffffff)
            for (i = 0; i < sb->fat_blocks_count; i++)
                Cache_FreeBlock(sb->cache, sb->fat_blocks[i]);

        /* load some more */
        num = sb->first_device_sector + sb->first_fat_sector
            + sb->fat_size * fat_no + (entry_cache_block
            << (sb->fat_cachesize_bits - sb->sectorsize_bits));
        for (i = 0; i < sb->fat_blocks_count; i++)
            sb->fat_blocks[i] =
                Cache_GetBlock(sb->cache, num + i, &sb->fat_buffers[i]);

        /* remember where we are for next time */
        sb->fat_cache_block = entry_cache_block;
        sb->fat_cache_no = fat_no;
    }

    /* give the block back if they asked for it (needed to mark the block
     * dirty if they're writing) */
    if (rb != NULL)
        *rb = sb->fat_blocks[entry_cache_offset >> sb->sectorsize_bits];

    /* compute the pointer location and return it */
    return sb->fat_buffers[entry_cache_offset >> sb->sectorsize_bits] +
        (entry_cache_offset & (sb->sectorsize - 1));
}

/* FAT12 has, as the name suggests, 12-bit FAT entries. This means that two
 * entries are condensed into three bytes, like so:
 *
 * entry: aaaaaaaa aaaabbbb bbbbbbbb
 * bytes: xxxxxxxx xxxxxxxx xxxxxxxx
 *
 * To get at the entry we want, we find and grab the word starting at either
 * byte 0 or 1 of the three-byte set, then shift up or down as needed. FATdoc
 * 1.03 p16-17 describes the method
 * 
 * The only tricky bit is if the word falls such that the first byte is the
 * last byte of the block and the second byte is the first byte of the next
 * block. Since our block data are stored within cache block structures, a
 * simple cast won't do (hell, the second block may not even be in memory if
 * we're at the end of the FAT cache). So we get it a byte at a time, and
 * build the word ourselves.
 */
static ULONG GetFat12Entry(struct FSSuper *sb, ULONG n) {
    ULONG offset = n + n/2;
    UWORD val;

    if ((offset & (sb->sectorsize-1)) == sb->sectorsize-1) {
        D(bug("[fat] fat12 cluster pair on block boundary, compensating\n"));

        val = *GetFatEntryPtr(sb, offset + 1, NULL, 0) << 8;
        val |= *GetFatEntryPtr(sb, offset, NULL, 0);
    }
    else
        val = AROS_LE2WORD(*((UWORD *) GetFatEntryPtr(sb, offset, NULL, 0)));

    if (n & 1)
        val >>= 4;
    else
        val &= 0xfff;

    return val;
}

/*
 * FAT16 and FAT32, on the other hand, have nice neat entry widths, so simple
 * word/long casts are fine. There's also no chance that the entry can be
 * split across blocks. Why can't everything be this simple?
 */
static ULONG GetFat16Entry(struct FSSuper *sb, ULONG n) {
    return AROS_LE2WORD(*((UWORD *) GetFatEntryPtr(sb, n << 1, NULL, 0)));
}

static ULONG GetFat32Entry(struct FSSuper *sb, ULONG n) {
    return AROS_LE2LONG(*((ULONG *) GetFatEntryPtr(sb, n << 2, NULL, 0)))
        & 0x0fffffff;
}

static void SetFat12Entry(struct FSSuper *sb, ULONG n, ULONG val) {
    APTR b;
    ULONG offset = n + n/2;
    BOOL boundary = FALSE;
    UWORD *fat = NULL, newval, i;

    for (i = 0; i < sb->fat_count; i++)
    {
        if ((offset & (sb->sectorsize-1)) == sb->sectorsize-1) {
            boundary = TRUE;

            D(bug("[fat] fat12 cluster pair on block boundary, compensating\n"));

            newval = *GetFatEntryPtr(sb, offset + 1, NULL, i) << 8;
            newval |= *GetFatEntryPtr(sb, offset, NULL, i);
        }
        else {
            fat = (UWORD *) GetFatEntryPtr(sb, offset, &b, i);
            newval = AROS_LE2WORD(*fat);
        }

        if (n & 1) {
            val <<= 4;
            newval = (newval & 0xf) | val;
        }
        else {
            newval = (newval & 0xf000) | val;
        }

        if (boundary) {
            /* XXX ideally we'd mark both blocks dirty at the same time or
             * only do it once if they're the same block. unfortunately any 
             * old value of b is invalid after a call to GetFatEntryPtr, as
             * it may have swapped the previous cache out. This is probably
             * safe enough. */
            *GetFatEntryPtr(sb, offset+1, &b, i) = newval >> 8;
            Cache_MarkBlockDirty(sb->cache, b);
            *GetFatEntryPtr(sb, offset, &b, i) = newval & 0xff;
            Cache_MarkBlockDirty(sb->cache, b);
        }
        else {
            *fat = AROS_WORD2LE(newval);
            Cache_MarkBlockDirty(sb->cache, b);
        }
    }
}

static void SetFat16Entry(struct FSSuper *sb, ULONG n, ULONG val) {
    APTR b;
    UWORD i;

    for (i = 0; i < sb->fat_count; i++)
    {
        *((UWORD *) GetFatEntryPtr(sb, n << 1, &b, i)) =
            AROS_WORD2LE((UWORD) val);
        Cache_MarkBlockDirty(sb->cache, b);
    }
}

static void SetFat32Entry(struct FSSuper *sb, ULONG n, ULONG val) {
    APTR b;
    ULONG *fat;
    UWORD i;

    for (i = 0; i < sb->fat_count; i++)
    {
        fat = (ULONG *) GetFatEntryPtr(sb, n << 2, &b, i);

        *fat = (*fat & 0xf0000000) | val;

        Cache_MarkBlockDirty(sb->cache, b);
    }
}

LONG ReadFATSuper(struct FSSuper *sb ) {
    struct DosEnvec *de = BADDR(glob->fssm->fssm_Environ);
    LONG err;
    ULONG bsize = de->de_SizeBlock * 4;
    struct FATBootSector *boot;
    struct FATFSInfo *fsinfo;
    BOOL invalid = FALSE;
    ULONG end;

    D(bug("[fat] reading boot sector\n"));

    boot = AllocMem(bsize, MEMF_ANY);
    if (!boot)
        return ERROR_NO_FREE_STORE;

    /*
     * Read the boot sector. We go direct because we don't have a cache yet,
     * and can't create one until we know the sector size, which is held in
     * the boot sector. In practice it doesn't matter - we're going to use
     * this once and once only.
     */
    sb->first_device_sector =
        de->de_BlocksPerTrack * de->de_Surfaces * de->de_LowCyl;

    D(bug("[fat] boot sector at sector %ld\n", sb->first_device_sector));

    if ((err = AccessDisk(FALSE, sb->first_device_sector, 1, bsize, (UBYTE *)boot)) != 0) {
        D(bug("[fat] couldn't read boot block (%ld)\n", err));
        FreeMem(boot, bsize);
        return err;
    }

    D(bug("\tBoot sector:\n"));

    sb->sectorsize = AROS_LE2WORD(boot->bpb_bytes_per_sect);
    sb->sectorsize_bits = log2(sb->sectorsize);
    D(bug("\tSectorSize = %ld\n", sb->sectorsize));
    D(bug("\tSectorSize Bits = %ld\n", sb->sectorsize_bits));

    sb->cluster_sectors = boot->bpb_sect_per_clust;
    sb->clustersize = sb->sectorsize * boot->bpb_sect_per_clust;
    sb->clustersize_bits = log2(sb->clustersize);
    sb->cluster_sectors_bits = sb->clustersize_bits - sb->sectorsize_bits;

    D(bug("\tSectorsPerCluster = %ld\n", (ULONG)boot->bpb_sect_per_clust));
    D(bug("\tClusterSize = %ld\n", sb->clustersize));
    D(bug("\tClusterSize Bits = %ld\n", sb->clustersize_bits));
    D(bug("\tCluster Sectors Bits = %ld\n", sb->cluster_sectors_bits));

    sb->first_fat_sector = AROS_LE2WORD(boot->bpb_rsvd_sect_count);
    D(bug("\tFirst FAT Sector = %ld\n", sb->first_fat_sector));

    sb->fat_count = boot->bpb_num_fats;
    D(bug("\tNumber of FATs = %d\n", sb->fat_count));

    if (boot->bpb_fat_size_16 != 0)
        sb->fat_size = AROS_LE2WORD(boot->bpb_fat_size_16);
    else
        sb->fat_size = AROS_LE2LONG(boot->type.fat32.bpb_fat_size_32);
    D(bug("\tFAT Size = %ld\n", sb->fat_size));

    if (boot->bpb_total_sectors_16 != 0)
        sb->total_sectors = AROS_LE2WORD(boot->bpb_total_sectors_16);
    else
        sb->total_sectors = AROS_LE2LONG(boot->bpb_total_sectors_32);
    D(bug("\tTotal Sectors = %ld\n", sb->total_sectors));

    sb->rootdir_sectors = ((AROS_LE2WORD(boot->bpb_root_entries_count) * sizeof(struct FATDirEntry)) + (sb->sectorsize - 1)) >> sb->sectorsize_bits;
    D(bug("\tRootDir Sectors = %ld\n", sb->rootdir_sectors));

    sb->data_sectors = sb->total_sectors - (sb->first_fat_sector + (sb->fat_count * sb->fat_size) + sb->rootdir_sectors);
    D(bug("\tData Sectors = %ld\n", sb->data_sectors));

    sb->clusters_count = sb->data_sectors >> sb->cluster_sectors_bits;
    D(bug("\tClusters Count = %ld\n", sb->clusters_count));

    sb->first_rootdir_sector = sb->first_fat_sector + (sb->fat_count * sb->fat_size);
    D(bug("\tFirst RootDir Sector = %ld\n", sb->first_rootdir_sector));

    sb->first_data_sector = sb->first_fat_sector + (sb->fat_count * sb->fat_size) + sb->rootdir_sectors;
    D(bug("\tFirst Data Sector = %ld\n", sb->first_data_sector));

    /* check if disk is in fact a FAT filesystem */

    /* valid sector size: 512, 1024, 2048, 4096 */
    if (sb->sectorsize != 512 && sb->sectorsize != 1024 && sb->sectorsize != 2048 && sb->sectorsize != 4096)
        invalid = TRUE;

    /* valid bpb_sect_per_clust: 1, 2, 4, 8, 16, 32, 64, 128 */
    if ((boot->bpb_sect_per_clust & (boot->bpb_sect_per_clust - 1)) != 0 || boot->bpb_sect_per_clust == 0 || boot->bpb_sect_per_clust > 128)
        invalid = TRUE;

    /* valid cluster size: 512, 1024, 2048, 4096, 8192, 16k, 32k, 64k */
    if (sb->clustersize > 64 * 1024)
        invalid = TRUE;

    if (sb->first_fat_sector == 0)
        invalid = TRUE;

    if (sb->fat_count == 0)
        invalid = TRUE;

    if (boot->bpb_media < 0xF0)
        invalid = TRUE;

    /* FAT "signature" */
    if (boot->bpb_signature[0] != 0x55 || boot->bpb_signature[1] != 0xaa)
        invalid = TRUE;
 
    if (invalid) {
        D(bug("\tInvalid FAT Boot Sector\n"));
        FreeMem(boot, bsize);
        return ERROR_NOT_A_DOS_DISK;
    }
    end = 0xFFFFFFFF / sb->sectorsize;
    if ((sb->first_device_sector + sb->total_sectors - 1 > end) && (glob->readcmd == CMD_READ)) {
        D(bug("\tDevice is too large\n"));
        FreeMem(boot, bsize);
        return IOERR_BADADDRESS;
    }

    sb->cache = Cache_CreateCache(64, 64, sb->sectorsize);

    if (sb->clusters_count < 4085) {
        D(bug("\tFAT12 filesystem detected\n"));
        sb->type = 12;
        sb->eoc_mark = 0x0FFF;
        sb->func_get_fat_entry = GetFat12Entry;
        sb->func_set_fat_entry = SetFat12Entry;
    }
    else if (sb->clusters_count < 65525) {
        D(bug("\tFAT16 filesystem detected\n"));
        sb->type = 16;
        sb->eoc_mark = 0xFFFF;
        sb->func_get_fat_entry = GetFat16Entry;
        sb->func_set_fat_entry = SetFat16Entry;
    }
    else {
        D(bug("\tFAT32 filesystem detected\n"));
        sb->type = 32;
        sb->eoc_mark = 0x0FFFFFFF;
        sb->func_get_fat_entry = GetFat32Entry;
        sb->func_set_fat_entry = SetFat32Entry;
    }

    /* setup the FAT cache and load the first blocks */
    sb->fat_cachesize = 4096;
    sb->fat_cachesize_bits = log2(sb->fat_cachesize);
    sb->fat_cache_block = 0xffffffff;

    sb->fat_blocks_count =
        MIN(sb->fat_size, sb->fat_cachesize >> sb->sectorsize_bits);
    sb->fat_blocks = AllocVecPooled(glob->mempool,
        sizeof(APTR) * sb->fat_blocks_count);
    sb->fat_buffers = AllocVecPooled(glob->mempool,
        sizeof(APTR) * sb->fat_blocks_count);

    if (sb->type != 32) { /* FAT 12/16 */
        /* setup volume id */
        sb->volume_id = AROS_LE2LONG(boot->type.fat16.bs_volid);

        /* location of root directory */
        sb->rootdir_cluster = 0;
        sb->rootdir_sector = sb->first_rootdir_sector;
    }
    else {
        /* setup volume id */
        sb->volume_id = AROS_LE2LONG(boot->type.fat32.bs_volid);
 
        /* location of root directory */
        sb->rootdir_cluster = AROS_LE2LONG(boot->type.fat32.bpb_root_cluster);
        sb->rootdir_sector = 0;
    }

    D(bug("[fat] rootdir at cluster %ld sector %ld\n", sb->rootdir_cluster, sb->rootdir_sector));

    if (GetVolumeIdentity(sb, &(sb->volume)) != 0) {
        LONG i;
        UBYTE *uu = (void *)&sb->volume_id;

        for (i=1; i<10;) {
            int d;

            if (i==5)
                sb->volume.name[i++]='-';

            d = (*uu) & 0x0f;
            sb->volume.name[i++] = (d < 10) ? '0' + d : 'A' - 10 + d;
            d = ((*uu) & 0xf0)>>4;
            sb->volume.name[i++] = (d < 10) ? '0' + d : 'A' - 10 + d;

            uu++;
        }

        sb->volume.name[i] = '\0';
        sb->volume.name[0] = 9;
    }

    /* get initial number of free clusters */
    sb->free_clusters = -1;
    sb->next_cluster = -1;
    if (sb->type == 32) {
        sb->fsinfo_block = Cache_GetBlock(sb->cache, sb->first_device_sector
            + AROS_LE2WORD(boot->type.fat32.bpb_fs_info), (UBYTE **)&fsinfo);
        if (sb->fsinfo_block != NULL) {
            if (fsinfo->lead_sig == AROS_LONG2LE(FSI_LEAD_SIG)
                && fsinfo->struct_sig == AROS_LONG2LE(FSI_STRUCT_SIG)
                && fsinfo->trail_sig == AROS_LONG2LE(FSI_TRAIL_SIG)) {
                sb->free_clusters = AROS_LE2LONG(fsinfo->free_count);
                sb->next_cluster = AROS_LE2LONG(fsinfo->next_free);
                D(bug("[fat] valid FATFSInfo block found\n"));
                sb->fsinfo_buffer = fsinfo;
            }
            else
                Cache_FreeBlock(sb->cache, sb->fsinfo_block);
        }
    }
    if (sb->free_clusters == -1)
        CountFreeClusters(sb);
    if (sb->next_cluster == -1)
        sb->next_cluster = 2;

    D(bug("\tFAT Filesystem successfully detected.\n"));
    D(bug("\tFree Clusters = %ld\n", sb->free_clusters));
    D(bug("\tNext Free Cluster = %ld\n", sb->next_cluster));
    FreeMem(boot, bsize);
    return 0;
}

LONG GetVolumeIdentity(struct FSSuper *sb, struct VolumeIdentity *volume) {
    struct DirHandle dh;
    struct DirEntry de;
    LONG err;
    int i;

    D(bug("[fat] searching root directory for volume name\n"));

    /* search the directory for the volume id entry. it would've been nice to
     * just use GetNextDirEntry but I didn't want a flag or something to tell
     * it not to skip the volume name */
    InitDirHandle(sb, sb->rootdir_cluster, &dh, FALSE);

    while ((err = GetDirEntry(&dh, dh.cur_index + 1, &de)) == 0) {

        /* match the volume id entry */
        if ((de.e.entry.attr & ATTR_VOLUME_ID_MASK) == ATTR_VOLUME_ID
            && de.e.entry.name[0] != 0xe5) {
            D(bug("[fat] found volume id entry %ld\n", dh.cur_index));

            /* copy the name in. volume->name is a BSTR */

            volume->name[1] = de.e.entry.name[0];

            for (i = 1; i < 11; i++) {
                if (volume->name[i] == ' ')
                    volume->name[i+1] = de.e.entry.name[i];
                else
                    volume->name[i+1] = tolower(de.e.entry.name[i]);
            }

            for (i = 10; volume->name[i+1] == ' '; i--);
            volume->name[i+2] = '\0';
            volume->name[0] = strlen(&(volume->name[1]));

            /* get the volume creation date too */
            ConvertFATDate(de.e.entry.create_date, de.e.entry.create_time, &volume->create_time);

            D(bug("[fat] volume name is '%s'\n", &(volume->name[1])));

            break;
        }

        /* bail out if we hit the end of the dir */
        if (de.e.entry.name[0] == 0x00) {
            D(bug("[fat] found end-of-directory marker, volume name entry not found\n"));
            err = ERROR_OBJECT_NOT_FOUND;
            break;
        }
    }

    ReleaseDirHandle(&dh);
    return err;
}

LONG SetVolumeName(struct FSSuper *sb, UBYTE *name) {
    struct DirHandle dh;
    struct DirEntry de;
    LONG err;
    int i;
    struct DosEnvec *dosenv = BADDR(glob->fssm->fssm_Environ);
    ULONG bsize = dosenv->de_SizeBlock * 4;
    struct FATBootSector *boot;

    /* read boot block */
    boot = AllocMem(bsize, MEMF_ANY);
    if (!boot)
        return ERROR_NO_FREE_STORE;

    if ((err = AccessDisk(FALSE, sb->first_device_sector, 1, bsize, (UBYTE *)boot)) != 0) {
        D(bug("[fat] couldn't read boot block (%ld)\n", err));
        FreeMem(boot, bsize);
        return err;
    }

    D(bug("[fat] searching root directory for volume name\n"));

    /* search the directory for the volume id entry. it would've been nice to
     * just use GetNextDirEntry but I didn't want a flag or something to tell
     * it not to skip the volume name */
    InitDirHandle(sb, 0, &dh, FALSE);

    while ((err = GetDirEntry(&dh, dh.cur_index + 1, &de)) == 0) {

        /* match the volume id entry */
        if ((de.e.entry.attr & ATTR_VOLUME_ID_MASK) == ATTR_VOLUME_ID
            && de.e.entry.name[0] != 0xe5) {
            D(bug("[fat] found volume id entry %ld\n", dh.cur_index));
            err = 0;
            break;
        }

        /* bail out if we hit the end of the dir */
        if (de.e.entry.name[0] == 0x00) {
            D(bug("[fat] found end-of-directory marker, volume name entry not found\n"));
            err = ERROR_OBJECT_NOT_FOUND;
            break;
        }
    }

    /* create a new volume id entry if there wasn't one */
    if (err != 0) {
        err = AllocDirEntry(&dh, 0, &de);
        if (err == 0) {
            memset(&de.e.entry, 0, sizeof(struct FATDirEntry));
            de.e.entry.attr = ATTR_VOLUME_ID;
        }
    }

    /* copy the name in. name is a BSTR */
    if (err == 0) {
        de.e.entry.name[0] = name[1];
        for (i = 0; i < 11; i++)
            if (i < name[0])
                de.e.entry.name[i] = toupper(name[i+1]);
            else
                de.e.entry.name[i] = ' ';

        if ((err = UpdateDirEntry(&de)) != 0) {
            D(bug("[fat] couldn't change volume name\n"));
            return err;
        }
    }

    /* copy name to boot block as well, and save */
    if (sb->type == 32)
        CopyMem(de.e.entry.name, boot->type.fat32.bs_vollab, 11);
    else
        CopyMem(de.e.entry.name, boot->type.fat16.bs_vollab, 11);

    if ((err = AccessDisk(TRUE, sb->first_device_sector, 1, bsize,
        (UBYTE *)boot)) != 0)
        D(bug("[fat] couldn't write boot block (%ld)\n", err));
    FreeMem(boot, bsize);

    /* update name in sb */
    sb->volume.name[0] = name[0] <= 11 ? name[0] : 11;
    CopyMem(&name[1], &(sb->volume.name[1]), sb->volume.name[0]);
    sb->volume.name[sb->volume.name[0]+1] = '\0';

    D(bug("[fat] new volume name is '%s'\n", &(sb->volume.name[1])));

    ReleaseDirHandle(&dh);
    return err;
}

LONG FindFreeCluster(struct FSSuper *sb, ULONG *rcluster) {
    ULONG cluster = 0;
    BOOL found = FALSE;

    for (cluster = sb->next_cluster;
        cluster < 2 + sb->clusters_count && !found;
        cluster++)
    {
        if (GET_NEXT_CLUSTER(sb, cluster) == 0)
        {
            *rcluster = cluster;
            found = TRUE;
        }
    }

    if (!found)
    {
        for (cluster = 2; cluster < sb->next_cluster && !found;
            cluster++)
        {
            if (GET_NEXT_CLUSTER(sb, cluster) == 0)
            {
                *rcluster = cluster;
                found = TRUE;
            }
        }
    }

    if (!found) {
        D(bug("[fat] no more free clusters, we're out of space\n"));
        return ERROR_DISK_FULL;
    }

    sb->next_cluster = *rcluster;

    D(bug("[fat] found free cluster %ld\n", *rcluster));

    return 0;
}

void FreeFATSuper(struct FSSuper *sb) {
    D(bug("\tRemoving Super Block from memory\n"));
    Cache_DestroyCache(sb->cache);
    FreeVecPooled(glob->mempool, sb->fat_buffers);
    sb->fat_buffers = NULL;
    FreeVecPooled(glob->mempool, sb->fat_blocks);
    sb->fat_blocks = NULL;
}

/* see how many unused clusters are available */
void CountFreeClusters(struct FSSuper *sb) {
    ULONG cluster = 0;
    ULONG free = 0;

    /* loop over all the data clusters */
    for (cluster = 2; cluster < sb->clusters_count + 2; cluster++)
    {
        /* record the free ones */
        if (GET_NEXT_CLUSTER(sb, cluster) == 0)
            free++;
    }

    /* put the value away for later */
    sb->free_clusters = free;

    D(bug("\tfree clusters: %ld\n", free));
}

void AllocCluster(struct FSSuper *sb, ULONG cluster) {
    SET_NEXT_CLUSTER(sb, cluster, sb->eoc_mark);
    sb->free_clusters--;
    if (sb->fsinfo_buffer != NULL) {
        sb->fsinfo_buffer->free_count = AROS_LONG2LE(sb->free_clusters);
        sb->fsinfo_buffer->next_free = AROS_LONG2LE(sb->next_cluster);
        Cache_MarkBlockDirty(sb->cache, sb->fsinfo_block);
    }
}

void FreeCluster(struct FSSuper *sb, ULONG cluster) {
    SET_NEXT_CLUSTER(sb, cluster, 0);
    sb->free_clusters++;
    if (sb->fsinfo_buffer != NULL) {
        sb->fsinfo_buffer->free_count = AROS_LONG2LE(sb->free_clusters);
        Cache_MarkBlockDirty(sb->cache, sb->fsinfo_block);
    }
}

void ConvertFATDate(UWORD date, UWORD time, struct DateStamp *ds) {
    ULONG year, month, day, hours, mins, secs;
    struct ClockData clock_data;

    /* date bits: yyyy yyym mmmd dddd */
    year = (date & 0xfe00) >> 9;    /* bits 15-9 */
    month = (date & 0x01e0) >> 5;   /* bits 8-5 */
    day = date & 0x001f;            /* bits 4-0 */

    /* time bits: hhhh hmmm mmms ssss */
    hours = (time & 0xf800) >> 11;  /* bits 15-11 */
    mins = (time & 0x07e0) >> 5;    /* bits 10-5 */
    secs = time & 0x001f;           /* bits 4-0 */

    D(bug("[fat] converting fat date: year %d month %d day %d hours %d mins %d secs %d\n", year, month, day, hours, mins, secs));

    clock_data.year = 1980 + year;
    clock_data.month = month;
    clock_data.mday = day;
    clock_data.hour = hours;
    clock_data.min = mins;
    clock_data.sec = secs << 1;
    secs = Date2Amiga(&clock_data);

    /* calculate days since 1978-01-01 (DOS epoch) */
    ds->ds_Days = secs / (60 * 60 * 24);

    /* minutes since midnight */
    ds->ds_Minute = secs / 60 % (24 * 60);

    /* 1/50 sec ticks since last minute */
    ds->ds_Tick = secs % 60 * TICKS_PER_SECOND;

    D(bug("[fat] converted fat date: days %ld minutes %ld ticks %ld\n", ds->ds_Days, ds->ds_Minute, ds->ds_Tick));
}

void ConvertAROSDate(struct DateStamp *ds, UWORD *date, UWORD *time) {
    ULONG year, month, day, hours, mins, secs;
    struct ClockData clock_data;

    /* convert datestamp to seconds since 1978 */
    secs = ds->ds_Days * 60 * 60 * 24 + ds->ds_Minute * 60
        + ds->ds_Tick / TICKS_PER_SECOND;

    /* Round up to next even second because of FAT's two-second granularity */
    secs = (secs & ~1) + 2;

    /* convert seconds since 1978 to calendar/time data */
    Amiga2Date(secs, &clock_data);

    /* get values used in FAT dates */
    year = clock_data.year - 1980;
    month = clock_data.month - 0;
    day = clock_data.mday;
    hours = clock_data.hour;
    mins = clock_data.min;
    secs = clock_data.sec >> 1;

    /* all that remains is to bit-encode the whole lot */

    /* date bits: yyyy yyym mmmd dddd */
    *date = (((ULONG) year) << 9) | (((ULONG) month) << 5) | day;

    /* time bits: hhhh hmmm mmms ssss */
    *time = (((ULONG) hours) << 11) | (((ULONG) mins) << 5) | secs;
}