The entire raw NCAs are encrypted.
The only known area which is not encrypted in the raw NCA is the logo section, when the NCA includes that section. Everything else documented on this page is for the plaintext version of that data.
Encryption
The first 0xC00 bytes are encrypted with AES-XTS with sector size 0x200 with a non-standard "tweak" (endianness is reversed, see here), this encrypted data is an 0x400 NCA header + an 0x200 header for each section in the section table.
For pre-1.0.0 "NCA2" NCAs, the first 0x400 byte are encrypted the same way as in NCA3. However, each section header is individually encrypted as though it were sector 0, instead of the appropriate sector as in NCA3.
Header
Offset | Size | Description |
---|---|---|
0x0 | 0x100 | RSA-2048 signature over the 0x200-bytes starting at offset 0x200 using fixed key. |
0x100 | 0x100 | RSA-2048 signature over the 0x200-bytes starting at offset 0x200 using key from NPDM, or zeroes if not a program. |
0x200 | 0x4 | Magicnum (Normally "NCA3", "NCA2" for some pre-1.0.0 NCAs) |
0x204 | 0x1 | 0 for system NCAs. 1 for a NCA from gamecard. |
0x205 | 0x1 | Content Type (0=Program, 1=Meta, 2=Control, 3=Manual, 4=Data, 5=PublicData) |
0x206 | 0x1 | Crypto Type. Only used stating with 3.0.0. Normally 0. 2 = Crypto supported starting with 3.0.0. |
0x207 | 0x1 | Key index, must be 0-2. |
0x208 | 0x8 | Size of the entire NCA. |
0x210 | 0x8 | titleID |
0x218 | 0x4 | contentIndex |
0x21C | 0x4 | sdk_version. byte0 is normally 0? Compared with a required minimum-value(0x000B0000). Matches this string from codebin: "FS_ACCESS: { sdk_version: {byte3}.{byte2}.{byte1}, ...". |
0x220 | 0x1 | Crypto-Type2. Selects which crypto-sysver to use. 0x3 = 3.0.1, 0x4 = 4.0.0, 0x5 = 5.0.0. |
0x230 | 0x10 | Rights ID (Ticket) |
0x240 | 0x10*0x4(0x40) | Table for each section, see below. |
0x280 | 0x20*0x4(0x80) | Table of SHA256 hashes, over each 0x200-byte Section Header Block. |
0x300 | 0x10*0x4(0x40) | Key area |
The header is 0x400-bytes, at NCA+0.
When the above "Crypto Type" field is 0x2 on >=v3.0, different {crypto/keydata} is used for the sections' data. With system content, this is used with every ncatype except ncatype0. The only other exception is {data-content} for the firm titles: this is required in order for older-system-versions to install it.
Crypto-Type2 0x3(with the above crypto-type value) is used for all 3.0.1 sysmodules and the System_Version_Title.
With 3.0.2, all updated titles use the crypto from 3.0.1 for non-ncatype0, except for firm {data-content}.
Note: in some cases various game content uses the above newer crypto as well.
The keyindex passed to <key-generation-related code> is determined as follows:
- Pre-3.0.0: The ncahdr keyindex field(0x207) is passed directly.
- 3.0.0+: It's determined using ncahdr keyindex(0x207) and "Crypto Type"(0x206). The latter field must be 0-2. In each ncahdr_keyindex block, it executes "if(ncahdr_x206>=3)<panic>", but that won't trigger due to the earlier check. The end result is basically the same as pre-3.0.0, except when ncahdr_x206 == 0x2, final_index is new_base_index+ncahdr_keyindex. Actual implementation loads index from u32_array[ncahdr_crypto_type], where the address of u32_array is different for each ncahdr_keyindex.
- 3.0.1+: The dedicated range check for ncahdr_x206("Crypto Type") was removed, since the updated code no longer needs it. The output from a function masked with 0xFF is now used instead of ncahdr_x206(ncahdr_crypto_type). The range check for that field was changed from {ncahdr_x206 check with panic described above}, to "if(index>=4)final_index=10;"(skips accessing the array and uses 10 directly). The arrays were updated with an additional entry: final_index=v301_base_index+ncahdr_keyindex.
- The keydata for the above index10 is not(?) known to be initialized.
- The new function called by the code described above does:
if(ncahdr_x206 < ncahdr_x220){ret = ncahdr_x220; } else { ret = ncahdr_x206; } return ret;
Section Table Entry
Offset | Size | Description |
---|---|---|
0x0 | 0x4 | Media offset |
0x4 | 0x4 | Media end-offset |
0x8 | 0x4 | Unknown |
0xC | 0x4 | Unknown |
Entry size is 0x10-bytes.
Media offset is absoluteoffset/{mediasize}, where mediasize is hard-coded to 0x200.
Section Header Block
Offset | Size | Description |
---|---|---|
0x0 | 0x2 | Version. Always 0x2 |
0x2 | 0x1 | Filesystem Type. 0x0 = RomFS. 0x1 = PFS0 |
0x3 | 0x1 | Hash Type. 0x2 = PFS0. 0x3 = RomFS |
0x4 | 0x1 | Crypto type. 0 and >4 are invalid. 1 = none(plaintext from raw NCA). 2 = other crypto. 3 = regular crypto. 4 = unknown. |
0x5 | 0x1 | Padding? |
0x8 | 0xF8 | FS-specific superblock. |
0x100 | ? | Optional BKTR header. Can be used with any section, but only known to be used with game-updates RomFS. |
The Section Header Block for each section is at absoluteoffset+0x400+(sectionid*0x200), where sectionid corresponds to the index used with the entry/hash tables.
The total size is 0x200-bytes.
PFS0 superblock
Offset | Size | Description |
---|---|---|
0x0 | 0x20 | SHA256 hash over the hash-table at section-start+0 with the below hash-table size. |
0x20 | 0x4 | Block size in bytes. |
0x24 | 0x4 | Must be 0x2. |
0x28 | 0x8 | Offset of hash-table. Normally zero? |
0x30 | 0x8 | Size of hash-table. |
0x38 | 0x8 | Offset relative to section-start where the PFS0 header is located. |
0x40 | 0x8 | Actual byte-size of the PFS0 filesystem relative to the PFS0 header. |
0x48 | 0xB0 | Normally zeros. |
RomFS superblock
Offset | Size | Description |
---|---|---|
0x8 | 0xE0 | IVFC header. Basically the same as Savegames IVFC except with 2 more levels and +0x0C is non-zero, see below. |
This documents the structure of Section Header Block +0 for RomFS.
IVFC
Start | Length | Description |
---|---|---|
0x00 | 4 | Magic ("IVFC") |
0x04 | 4 | Magic Number (0x20000) |
0x08 | 4 | Master hash size? |
0x0C | 4 | Usually 7? Unknown, could be related to total number of levels maybe? |
0x10 | 8 | Level 1 offset |
0x18 | 8 | Level 1 size |
0x20 | 4 | Level 1 block size, in log2 |
0x24 | 4 | Reserved |
0x28 | 8 | Level 2 offset |
0x30 | 8 | Level 2 size |
0x38 | 4 | Level 2 block size, in log2. |
0x3C | 4 | Reserved |
0x40 | 8 | Level 3 offset |
0x48 | 8 | Level 3 size |
0x50 | 4 | Level 3 block size, in log2. |
0x54 | 4 | Reserved |
0x58 | 8 | Level 4 offset |
0x60 | 8 | Level 4 size |
0x68 | 4 | Level 4 block size, in log2. |
0x6C | 4 | Reserved |
0x70 | 8 | Level 5 offset |
0x78 | 8 | Level 5 size |
0x80 | 4 | Level 5 block size, in log2. |
0x84 | 4 | Reserved |
0x88 | 8 | Level 6 offset |
0x90 | 8 | Level 6 size |
0x98 | 4 | Level 6 block size, in log2. |
0x9C | 4 | Reserved |
0xA0 | 32 | Unknown, reserved? |
0xC0 | 32 | Hash |
BKTR
Start | Length | Description |
---|---|---|
0x0 | 0x8 | Offset |
0x8 | 0x8 | Size |
0x10 | 0x4 | "BKTR" |
0x14 | 0x4 | u32, must be <=1. |
0x18 | 0x4 | s32, must be >=1. |
0x1C | 0x4 | ? |
0x20 | 0x20 | Same as the above 0x20-bytes except with different data. |
0x40 | 0x4? | ? |
0x44 | 0x4? | ? |
Using this header is enabled when offset 0x8 in this header is non-zero.
The above byte-offsets are relative to the start of the section-data.
The two sections specified by the two BKTR entries are usually(?) at the very end of the section data(section_endoffset-{size of BKTR sections}).
RomFS Patching
The BKTR section enables combining data from an update NCA with the RomFS from a base NCA to create a single patched RomFS image.
The first BKTR entry describes how to map regions of the two RomFS images to create the patched RomFS. It has the following format:
Start | Length | Description |
---|---|---|
0x0 | 0x4 | Padding/Unused? |
0x4 | 0x4 | Number of Buckets |
0x8 | 0x8 | Total Size of the Virtual RomFS Image |
0x10 | 0x3FF0 | Base Virtual Offset for each Bucket (u64s, padded with 0s until end) |
0x4000 | 0x4000*X | Relocation Buckets |
Where relocation buckets are as follows:
Start | Length | Description |
---|---|---|
0x0 | 0x4 | Padding/Unused? |
0x4 | 0x4 | Number of Entries |
0x8 | 0x8 | End offset for this Bucket |
0x10 | 0x3FF0 | Relocation Entries |
Where relocation entries are as follows:
Start | Length | Description |
---|---|---|
0x0 | 0x8 | Address in Patched RomFS |
0x8 | 0x8 | Address in Source RomFS |
0x10 | 0x4 | 1=Is from Patch RomFS, 0=Is from Base RomFS |
The second BKTR entry describes the subsections within the Patch RomFS. It has the following format:
Start | Length | Description |
---|---|---|
0x0 | 0x4 | Padding/Unused? |
0x4 | 0x4 | Number of Buckets |
0x8 | 0x8 | Total Size of the Physical Patch Image |
0x10 | 0x3FF0 | Base Physical Offset for each Bucket (u64s, padded with 0s until end) |
0x4000 | 0x4000*X | Subsection Buckets |
Where subsection buckets are as follows:
Start | Length | Description |
---|---|---|
0x0 | 0x4 | Padding/Unused? |
0x4 | 0x4 | Number of Entries |
0x8 | 0x8 | End offset for this Bucket |
0x10 | 0x3FF0 | Subsection Entries |
Where subsection entries are as follows:
Start | Length | Description |
---|---|---|
0x0 | 0x8 | Address in Patch RomFS |
0x8 | 0x4 | Padding/Unused? |
0xC | 0x4 | Value for subsection AES-CTR |
Official code assumes the relocation entries are sorted, and performs a binary search when determining where to read from. Each subsection in the Patch RomFS has its CTR calculated separately from the others based on the value in its entry (the BKTR entries use normal crypto). Thus decrypting a Patch RomFS requires decrypting and parsing the BKTR entries before anything else.
Logo section
This is a PFS0. See here for the mounted-FS logo contents.
ExeFS section
This is a PFS0.
See here for mounted-FS ExeFS contents.
Game-updates
The section-data for ncatype1 RomFS section(section1) uses section-crypto-type 0x4.
Game-updates also contain multiple ncatype6 content, which contain "section0_pfs0/fragment". Some of these are just NCAs, unknown for the rest(presumably NCAs with additional crypto?). The first ncatype6 content fragment file has a NDV0 header, with the NCA starting at offset 0x44.
PFS0
Offset | Size | Description |
---|---|---|
{Hash-table offset from superblock} | {Hash-table size from superblock} | Table of SHA256 hashes. |
{Hash-table <offset+size> from superblock} | Zeros for alignment to {alignment size}. | |
{PFS0 offset from superblock} | {PFS0 size from superblock} | The actual PFS0. |
This is the FS which has magicnum "PFS0" at header+0. This is very similar to HFS0. A tool for extracting this FS is available here.
The hash table is hashes for every {Block size from superblock} starting at the PFS0 header. The size used for the last hash is {PFS0 filesystem size from superblock} - offset_relativeto_header.
See also the PFS0 superblock above.
Offset | Size | Description |
---|---|---|
0x0 | 0x4 | "PFS0" Magic |
0x4 | 0x4 | Number of files |
0x8 | 0x4 | Size of the string table |
0xC | 0x4 | Zero/Reserved |
0x10 | X | File Entry Table |
0x10 + X | Y | String Table |
0x10 + X + Y | Z | Raw File Data |
Where File Entry Table consists of Number of Files FileEntries:
Offset | Size | Description |
---|---|---|
0x0 | 0x8 | Offset of file in Data |
0x8 | 0x8 | Size of file in Data |
0x10 | 0x4 | Offset of filename in String Table |
0x14 | 0x4 | Normally zero? |