Mauro Carvalho Chehab | 447d6fb | 2006-05-22 10:31:37 -0300 | [diff] [blame] | 1 | This document describes the cx2341x memory map and documents some of the register |
| 2 | space. |
| 3 | |
| 4 | Warning! This information was figured out from searching through the memory and |
| 5 | registers, this information may not be correct and is certainly not complete, and |
| 6 | was not derived from anything more than searching through the memory space with |
| 7 | commands like: |
| 8 | |
| 9 | ivtvctl -O min=0x02000000,max=0x020000ff |
| 10 | |
| 11 | So take this as is, I'm always searching for more stuff, it's a large |
| 12 | register space :-). |
| 13 | |
| 14 | Memory Map |
| 15 | ========== |
| 16 | |
| 17 | The cx2341x exposes its entire 64M memory space to the PCI host via the PCI BAR0 |
| 18 | (Base Address Register 0). The addresses here are offsets relative to the |
| 19 | address held in BAR0. |
| 20 | |
| 21 | 0x00000000-0x00ffffff Encoder memory space |
| 22 | 0x00000000-0x0003ffff Encode.rom |
| 23 | ???-??? MPEG buffer(s) |
| 24 | ???-??? Raw video capture buffer(s) |
| 25 | ???-??? Raw audio capture buffer(s) |
| 26 | ???-??? Display buffers (6 or 9) |
| 27 | |
| 28 | 0x01000000-0x01ffffff Decoder memory space |
| 29 | 0x01000000-0x0103ffff Decode.rom |
| 30 | ???-??? MPEG buffers(s) |
| 31 | 0x0114b000-0x0115afff Audio.rom (deprecated?) |
| 32 | |
| 33 | 0x02000000-0x0200ffff Register Space |
| 34 | |
| 35 | Registers |
| 36 | ========= |
| 37 | |
| 38 | The registers occupy the 64k space starting at the 0x02000000 offset from BAR0. |
| 39 | All of these registers are 32 bits wide. |
| 40 | |
| 41 | DMA Registers 0x000-0xff: |
| 42 | |
| 43 | 0x00 - Control: |
| 44 | 0=reset/cancel, 1=read, 2=write, 4=stop |
| 45 | 0x04 - DMA status: |
| 46 | 1=read busy, 2=write busy, 4=read error, 8=write error, 16=link list error |
| 47 | 0x08 - pci DMA pointer for read link list |
| 48 | 0x0c - pci DMA pointer for write link list |
| 49 | 0x10 - read/write DMA enable: |
| 50 | 1=read enable, 2=write enable |
| 51 | 0x14 - always 0xffffffff, if set any lower instability occurs, 0x00 crashes |
| 52 | 0x18 - ?? |
| 53 | 0x1c - always 0x20 or 32, smaller values slow down DMA transactions |
| 54 | 0x20 - always value of 0x780a010a |
| 55 | 0x24-0x3c - usually just random values??? |
| 56 | 0x40 - Interrupt status |
| 57 | 0x44 - Write a bit here and shows up in Interrupt status 0x40 |
| 58 | 0x48 - Interrupt Mask |
| 59 | 0x4C - always value of 0xfffdffff, |
| 60 | if changed to 0xffffffff DMA write interrupts break. |
| 61 | 0x50 - always 0xffffffff |
| 62 | 0x54 - always 0xffffffff (0x4c, 0x50, 0x54 seem like interrupt masks, are |
Trent Piepho | 657de3c | 2006-06-20 00:30:57 -0300 | [diff] [blame^] | 63 | 3 processors on chip, Java ones, VPU, SPU, APU, maybe these are the |
Mauro Carvalho Chehab | 447d6fb | 2006-05-22 10:31:37 -0300 | [diff] [blame] | 64 | interrupt masks???). |
| 65 | 0x60-0x7C - random values |
| 66 | 0x80 - first write linked list reg, for Encoder Memory addr |
| 67 | 0x84 - first write linked list reg, for pci memory addr |
| 68 | 0x88 - first write linked list reg, for length of buffer in memory addr |
| 69 | (|0x80000000 or this for last link) |
| 70 | 0x8c-0xcc - rest of write linked list reg, 8 sets of 3 total, DMA goes here |
| 71 | from linked list addr in reg 0x0c, firmware must push through or |
| 72 | something. |
| 73 | 0xe0 - first (and only) read linked list reg, for pci memory addr |
| 74 | 0xe4 - first (and only) read linked list reg, for Decoder memory addr |
| 75 | 0xe8 - first (and only) read linked list reg, for length of buffer |
| 76 | 0xec-0xff - Nothing seems to be in these registers, 0xec-f4 are 0x00000000. |
| 77 | |
| 78 | Memory locations for Encoder Buffers 0x700-0x7ff: |
| 79 | |
| 80 | These registers show offsets of memory locations pertaining to each |
| 81 | buffer area used for encoding, have to shift them by <<1 first. |
| 82 | |
| 83 | 0x07F8: Encoder SDRAM refresh |
| 84 | 0x07FC: Encoder SDRAM pre-charge |
| 85 | |
| 86 | Memory locations for Decoder Buffers 0x800-0x8ff: |
| 87 | |
| 88 | These registers show offsets of memory locations pertaining to each |
| 89 | buffer area used for decoding, have to shift them by <<1 first. |
| 90 | |
| 91 | 0x08F8: Decoder SDRAM refresh |
| 92 | 0x08FC: Decoder SDRAM pre-charge |
| 93 | |
| 94 | Other memory locations: |
| 95 | |
| 96 | 0x2800: Video Display Module control |
| 97 | 0x2D00: AO (audio output?) control |
| 98 | 0x2D24: Bytes Flushed |
| 99 | 0x7000: LSB I2C write clock bit (inverted) |
| 100 | 0x7004: LSB I2C write data bit (inverted) |
| 101 | 0x7008: LSB I2C read clock bit |
| 102 | 0x700c: LSB I2C read data bit |
| 103 | 0x9008: GPIO get input state |
| 104 | 0x900c: GPIO set output state |
| 105 | 0x9020: GPIO direction (Bit7 (GPIO 0..7) - 0:input, 1:output) |
| 106 | 0x9050: SPU control |
| 107 | 0x9054: Reset HW blocks |
| 108 | 0x9058: VPU control |
| 109 | 0xA018: Bit6: interrupt pending? |
| 110 | 0xA064: APU command |
| 111 | |
| 112 | |
| 113 | Interrupt Status Register |
| 114 | ========================= |
| 115 | |
| 116 | The definition of the bits in the interrupt status register 0x0040, and the |
| 117 | interrupt mask 0x0048. If a bit is cleared in the mask, then we want our ISR to |
| 118 | execute. |
| 119 | |
| 120 | Bit |
| 121 | 31 Encoder Start Capture |
| 122 | 30 Encoder EOS |
| 123 | 29 Encoder VBI capture |
| 124 | 28 Encoder Video Input Module reset event |
| 125 | 27 Encoder DMA complete |
| 126 | 26 |
| 127 | 25 Decoder copy protect detection event |
| 128 | 24 Decoder audio mode change detection event |
| 129 | 23 |
| 130 | 22 Decoder data request |
| 131 | 21 Decoder I-Frame? done |
| 132 | 20 Decoder DMA complete |
| 133 | 19 Decoder VBI re-insertion |
| 134 | 18 Decoder DMA err (linked-list bad) |
| 135 | |
| 136 | Missing |
| 137 | Encoder API call completed |
| 138 | Decoder API call completed |
| 139 | Encoder API post(?) |
| 140 | Decoder API post(?) |
| 141 | Decoder VTRACE event |