exFAT Filesystem

1. Introduction
2. exFAT Volume
3. Boot Sector
4. FAT and Allocation Bitmap
5. Up-case Table
6. Directory Entry
7. Directory Operations
8. Disk Partitioning

Introduction

The exFAT filesystem (Microsoft Extended FAT Filesystem) was developped as a succession of industry standard FAT filesystem widely used for removable media and built-in storages in embedded systems. For this reason, it is designed based on the FAT filesystem to be easily implemented to existng systems. This documentation describes mainly the differences between these filesystems. exFAT filesystem has some advantages of FAT filesystem as follows.

• Supports file size of over 4 GB (virtually unlimited).
• Supports volume size of over 2 TB (virtually unlimited).
• Improved performance of cluster allocation by allocation bitmap.
• Improved performance of data transfer by extended cluster size (up to 32 MB).
• Supports UTC timestamp (optional).
• Supports transaction safe filesystem (optional for WindowsCE).

About Notations in this Document

Numbers with heading 0x are assumed to be hexadecimal numbers, and others are decimal numbers.

The prefix K, M, G and T of each unit is assumed to be 2¹⁰, 2²⁰, 2³⁰ and 2⁴⁰ respectively.

The fragment of program codes contained in this document are written in C language, but it is not strict to the syntax.

32-bit value and 16-bit value are arbitrary mixed in the fragment of the program codes. The programmer needs be aware of the data loss due to the type conversion and how to avoid it. Also, every data type is assumed to be unsigned. Do not calculate in signed, or it may result unintended results.

exFAT Volume

The exFAT volume is composed of three areas. Each area is located on the volume in order of as follows: (exFAT Volume Map)

1. Boot Area (volume configuration parameters)
2. FAT Area (for only fragmented cluster chains)
3. Data Area (file, directroy and some system structures)

Boot Sector

Volume configuration parameters are recoreded in the boot sector as FAT filesystem does. The exFAT boot sector is actually in 12 contiguous sectors block, and two set of boot sectors (main and backup) are set in the boot area where the top of the exFAT volume. Each boundaries, sizes and count of clusters are explicitly recorded in the boot sector and there is nothing unclear in volume recognization.

Sector 9 contains OEM parameter and sector 10 is reserved for future use. The content of these sectors is depends on the system and filled with zero when not used. Sector 11 is filled with 32-bit check sum values of from sector 0 to sector 10 inclusive. These 11 sectors are calculated in byte-by-byte as a simple byte array, however VolumeFlags and PercentInUse are excluded from the calculation.

/* Calculation of 32-bit check sum */
uint32_t sum32 (const void* p, uint32_t n)
{
    uint32_t sum = 0;
    const uint8_t *dp = (const uint8_t*)p;

    do {
        sum = ((sum & 1) ? 0x80000000 : 0) + (sum >> 1) + *dp++;
    } while (--n);

    return sum;
}

/* Calculation of 16-bit check sum */
uint16_t sum16 (const void* p, uint32_t n)
{
    uint16_t sum = 0;
    const uint8_t *dp = (const uint8_t*)p;

    do {
        sum = ((sum & 1) ? 0x8000 : 0) + (sum >> 1) + *dp++;
    } while (--n);

    return sum;
}

FAT and Allocation Bitmap

The exFAT FAT entry is 32-bit in size as FAT32 FAT entry except for all bits in each entries are used. exFAT has adopted allocation bitmap to manage the cluster allocation status, free or not. The first bit of the allocation bitmap (bit0 of the first byte) corresponds to the cluseter 2. The bit value 1 means the cluster is in-use and not free for new allocation. The bit value 0 means the cluster is free and the value of FAT entry has no meaning. This is a significant difference of exFAT filesystem from FAT filesystem which indicates free cluster by a zero in the FAT entry. Following table shows the status of a cluster and corresponding entries. In the special case, the value in the FAT entry can have no meaning even if the cluster is in-use. This is described in Directory Operations.

The allocation bitmap is recored in the data area and its allocation information (start cluster and size) is recorded as a special directory entry (described in Directory Entry) on the root directory. The sizeof allocation bitmap becomes (ClusterCount + 7) / 8 bytes.

Up-case Table

In the exFAT filesystem, file names are handled as LFN extension of FAT filesystem does. Case information is preserved on recording, and matched in case insensitive. exFAT volume has up-case table on the volume for up-case conversion while it is done with on-system conversion table at the FAT filesystem. Perhaps, this is for new character definitions in the future. However what occures on the compatibility between the changes and existing systems is not clear. The up-case table needs to support ASCII alhpabetic character at least, and non-ASCII characters are optional. According to a research into the exFAT volume formatted by Windows, it seems to be same as FAT filesystem on the Windows does, but it is not that kept unchanged in the future.

The up-case table seems to support only BMP where corresponds to U+0000 - U+FFFF. It is recorded on the volume in compressed form. The following function is to extract the compressed table and create a case conversion table.

void load_upcase (
    uint16_t dst[],  /* Output table for U+0000 - U+FFFF */
    uint16_t src[],  /* Compressed table to input */
    uint16_t n_src   /* Size of compressed table [items] */
)
{
    uint16_t c, si, di;

    /* Fill output table with default values */
    c = 0; do dst[c] = c; while (++c);

    si = di = 0;
    do {
        c = src[si++];        /* Get an up-case character */
        if (c == 0xFFFF && si < n_src) {
            di += src[si++];  /* When U+FFFF appeared, skip the codes indicated by next item */
        } else {
            dst[di++] = c;    /* Store the up-case character */
        }
    } while (si < n_src);
}

Up-case table is recorded in the data area and its allocation information (start cluster and size) is recorded in a special directory entry (this is described below) on the root directory.

Directory Entry

Entry Type

Root directory is recorded in the data area. The first cluster number is stored in FirstClusterOfRootDirectory. The size of directory entry is 32 bytes. These are saeme as FAT filesystem does. Maximum length of a directory is 256 MB (8 M enrtries) while 2 MB at FAT filesystem. The first byte of each entry indicates the type of the entry. It is divided into four bit fields as shown below

Therefore the EntryType byte provides the information about the entry, but in fact the inportant thing is its value to identify the entry type. Following table shows the defined value and entry type. Last four entry types in the table are optional for generic use or not defined for exFAT 1.00, so that they are not described in this documentation.

EntryType is the only common field for every type of entries and other fields are defined for each entry types individually. To delete an entry, clear InUse bit of EntryType instead of storeing 0xE5 what FAT filessytem does. If EntryType is zero, all following enrties in the directory are guaranteed zero.

Allocation Bitmap Entry

Contains allocation information of allocation bitmap. This entry is recorded in the root directory independently.

Up-case Table Entry

Contains allocation information of up-case table. This entry is recorded in the root directory independently.

Volume Label Entry

Contains volume label of this volume. This entry is recorded in the root directory independently. If the size field is zero or this entry is not exist, the volume has no volume label. Any character allowed for file name includes dot can be used for the volume label.

File and Directory Entry

This is one of the entry types which composes an entry set. It indicates start of an entry set and contains file attribute and timestamp. The entry set is a block of consecutive directory entries to record meta data of a file or sub-directory. It is set on the direcotry in order of this entry (1 entry), stream extension entry (1 entry), and name extensiton entries (1-17 entries), so that an entry set spans from 3 to 19 entries.

Stream Extension Enrty

This is one of the entry types which composes an entry set. It contains the file allocation information.

File Name Entry

This is one of the entry types which composes an entry set. It contains the file name string. Any character except for control characters (U+0000 to U+001F, U+007F) and " * / : < > ? \ | is allowable for file name as LFN extensiton of FAT filesystem. exFAT filesystem does not support alternative file name (SFN) what FAT filesystem supports.

This is an example of exFAT root directory after formatting and a file creation. You will able to see how the entries are stored on the directory.

Directory Operations

Creating a File

To create a file, find a block of free entries in the directory and create the entry set of the file. The entry set has Archive attrribute and set DataLength field to zero as initial value. AllocationPossible bit is always 1 and NoFatChain bit is initially cleared.

When any data is written to the file and a new cluster is allocated, the cluster number is set to FirstCluster and set NoFatChain bit. Thereafter no data is written to the FAT entries as long as the cluter chain is contiguous. When the cluster chain gets fragmented on a cluster allocation, create valid cluster chain on the FAT and clear NoFatChain bit. This is applied to only file and sub-directory. Any other data (root directory, allocation bitmap and up-case table) always have valid cluster chain on the FAT.

Creating a Sub-directory

To create a sub-directory, find a block of free entries in the directory and create the entry set of the sub-directory. The entry set has Directory attrribute. A cluster is initially allocated to the directory and each entry is filled with zero. NoFatChain bit is as described above. When the sub-directory gets full, the cluster chain is stretched a cluster a time and the cluster is initialized to zero. The maximum length of a direcotry is 256 MB (8 M entries).

There are significant difference from FAT sub-directory. DataLength field has a valid value. Dot entries (".", "..") on the FAT sub-directory are not exist on the exFAT sub-directory. It may be logically generated and appeared in the directory listing.

Deleting a File

To remove a file, clear InUse bit in each entries of the file entry set. If the file has a cluster chain, also the chain needs to be removed from the FAT. In this case the FAT entries do not need to be changed because the clusters get free by clearing the allocation bitmap.

Deleting a Sub-directorty

It is same as deleting a file. All nodes below the directory needs to be scanned and all files and directories in the directory need to be deleted prior to delete the directory, or those objects' clusters get lost clusters.

Disk Partitioning

Same as the description of FAT filesystem. The value of system type in the partition table for exFAT volume is 0x07, which is same as NTFS. However MBR format does not support the storage device larger than 2 TB, because the allocation information in the partition table is recorded in 32-bit LBA. To use such storage device, it needs to be in SFD format or GPT format.