tillitis-key/doc/system_description/software.md

366 lines
15 KiB
Markdown
Raw Normal View History

2022-09-19 03:43:53 -04:00
# Tillitis Key software
2022-09-19 02:51:11 -04:00
2022-09-19 03:43:53 -04:00
## Definitions
2022-09-19 02:51:11 -04:00
2022-09-19 03:43:53 -04:00
* Firmware -- software that is part of ROM, and is currently
supplied via the FPGA bit stream.
* Application -- software supplied by the host machine, which is
received, loaded, and measured by the firmware (by hashing a
digest over the binary).
Learn more about the concepts in the
[system_description.md](system_description.md).
2022-09-19 02:51:11 -04:00
2022-09-19 03:43:53 -04:00
## CPU
2022-09-19 02:51:11 -04:00
2022-09-19 03:43:53 -04:00
We use a PicoRV32, a 32-bit RISC-V system, as the CPU for running the
firmware and the loaded app. All types are little-endian.
2022-09-19 02:51:11 -04:00
## Constraints
2022-09-19 03:43:53 -04:00
The application FPGA is a Lattice ICE40 UP5K, with the following
2022-09-19 02:51:11 -04:00
specifications:
2022-09-19 03:43:53 -04:00
* 30 EBR[^1] x 4 Kbit => 120 Kbit. PicoRV32 uses ~4 EBRs internally
=> 13 KB for Firmware. We should probably aim for less; 8 KB
should be the target.
2022-10-24 05:58:54 -04:00
* 4 SPRAM[^2] x 32 KB => 128 KB RAM for application/software
2022-10-11 11:25:00 -04:00
[^1]: Embedded Block RAM (also BRAM) residing in the FPGA, can
2022-10-12 04:10:48 -04:00
be configured as RAM or ROM.
2022-10-24 05:58:54 -04:00
[^2]: Single Port RAM (also SRAM).
2022-09-19 02:51:11 -04:00
## Introduction
2022-09-19 03:43:53 -04:00
The Tillitis Key has two modes of operation; firmware/loader mode and
application mode. The firmware mode has the responsibility of
receiving, measuring, and loading the application.
2022-09-19 02:51:11 -04:00
The firmware and application uses a memory mapped IO for SoC
2022-10-25 11:27:26 -04:00
communication. This MMIO resides at `0xc000_0000`. *Nota bene*: Almost
all access to MMIO should be word (32 bit) aligned. See table below.
2022-09-19 02:51:11 -04:00
2022-09-19 03:43:53 -04:00
The application has a constrained variant of the firmware memory map,
which is outlined below. E.g. UDS isn't readable, and the `APP_{ADDR,
SIZE}` are not writable for the application.
The software on the Tillitis Key communicates to the host via the
`UART_{RX,TX}_{STATUS,DATA}` registers, using the framing protocol
described in [Framing
Protocol](../framing_protocol/framing_protocol.md).
2022-09-19 02:51:11 -04:00
The firmware defines a protocol (command/response interface) on top of
the framing layer, which is used to bootstrap the application onto the
device.
On the framing layer, it's required that each frame the device
receives, a responding frame must be sent back to the host, in a
ping-pong manner.
Applications define a per-application protocol, which is the contract
between the host and the device.
## Firmware
The device has 128 KB RAM. The current firmware loads the app at the
upper 100 KB. The lower 28 KB is set up as stack for the app. A
2022-11-03 11:11:31 -04:00
smaller app that wants continuous memory may want to relocate itself
when starting.
2022-09-19 04:59:29 -04:00
2022-09-19 03:43:53 -04:00
The firmware is part of FPGA bitstream (ROM), and is loaded at
`0x0000_0000`.
2022-09-19 02:51:11 -04:00
### Reset
2022-09-19 03:43:53 -04:00
The PicoRV32 starts executing at `0x0000_0000`. Our firmware starts at
`_start` from `start.S` which initializes the `.data`, and `.bss` at
`0x4000_0000` and upwards. A stack is also initialized, starting at
0x4000_6ff0 and downwards. When the initialization is finished, the
2022-09-19 03:43:53 -04:00
firmware waits for incoming commands from the host, by busy-polling
the `UART_RX_{STATUS,DATA}` registers. When a complete command is
2022-09-19 03:43:53 -04:00
read, the firmware executes the command.
2022-09-19 02:51:11 -04:00
### Loading an application
2022-09-19 03:43:53 -04:00
The purpose of the firmware is to bootstrap an application. The host
2022-11-03 11:11:31 -04:00
will send a raw binary targeted to be loaded at `0x4000_7000` in the
2022-09-19 03:43:53 -04:00
device.
1. The host sends the User Supplied Secret (USS) by using the
2022-10-12 09:10:10 -04:00
`FW_CMD_LOAD_USS` command and gets a `FW_RSP_LOAD_USS` back.
2. The host sends the size of the app by using the
`FW_CMD_LOAD_APP_SIZE` command.
3. The firmware executes `FW_CMD_LOAD_APP_SIZE` command, which
2022-09-19 02:51:11 -04:00
stores the application size into `APP_SIZE`, and sets `APP_ADDR`
to zero. A `FW_RSP_LOAD_APP_SIZE` reponse is sent back to the
host, with the status of the action (ok/fail).
2022-10-12 09:10:10 -04:00
4. If the the host receive a sucessful response, it will send
multiple `FW_CMD_LOAD_APP_DATA` commands, together containing the
full application.
5. For each received `FW_CMD_LOAD_APP_DATA` command the firmware
2022-11-03 11:11:31 -04:00
places the data into `0x4000_7000` and upwards. The firmware
2022-10-12 09:10:10 -04:00
replies with a `FW_RSP_LOAD_APP_DATA` response to the host for
2022-09-19 03:43:53 -04:00
each received block.
6. When the final block of the application image is received, we
2022-09-19 03:43:53 -04:00
measure the application by computing a BLAKE2s digest over the
entire application,
The Compound Device Identifier is computed by using the `UDS`,
the measurement of the application, and the `USS`, and placed in
2022-11-03 11:11:31 -04:00
the `CDI` register. Then `0x4000_7000` is written to `APP_ADDR`.
The final `FW_RSP_LOAD_APP_DATA` response is sent to the host,
completing the loading.
2022-09-19 02:51:11 -04:00
NOTE: The firmware uses SPRAM for data and stack. We need to make sure
that the application image does not overwrite the firmware's running
state. The application should probably do a similar relocation for
stack/data at reset, as the firmware does. Further; the firmware need
to check application image is sane. The shared firmware data area
(e.g. `.data` and the stack must be cleared prior launching the
application.
2022-10-25 11:27:26 -04:00
### CDI computation
The CDI is computed by:
```
CDI = blake2s(UDS, blake2s(app), USS)
```
In an ideal world, software would never be able to read UDS at all and
we would have a BLAKE2s function in hardware that would be the only
thing able to read the UDS. Unfortunately, we couldn't fit a BLAKE2s
implementation in the FPGA at this time.
The firmware instead does the CDI computation using the special
firmware-only `FW_RAM` which is invisible after switching to app mode.
The `blake2s()` function in the firmware is fed with a buffer in
`FW_RAM` containing the UDS, the app digest, and the USS. It is also
fed with a context used for computations that is also part of the
`FW_RAM`.
### Loading the User Supplied Secret (USS)
The host program may send `FW_CMD_LOAD_USS` and `FW_CMD_LOAD_APP_SIZE`
in any order. But it *should* always send both `FW_CMD_LOAD_USS` and
`FW_CMD_LOAD_APP_SIZE` before sending the multiple
`FW_CMD_LOAD_APP_DATA`. If it does not, the USS will not be
predictable because somebody could have send `FW_CMD_LOAD_USS` before,
and the last `FW_CMD_LOAD_APP_DATA` (on whichever iteration) will
cause the currently loaded USS to be used for calculating CDI.
2022-09-19 02:51:11 -04:00
### Starting an application
Starting an application includes the "switch to application mode"
2022-09-19 03:43:53 -04:00
step, which is done by writing to the `SWITCH_APP` register. The
2022-09-19 02:51:11 -04:00
switch from firmware mode to application mode is a mode switch, and
2022-09-19 03:43:53 -04:00
context switch. When entering application mode the MMIO region is
restricted; e.g. some registers are removed (`UDS`), and some are
2022-09-19 02:51:11 -04:00
switched from read/write to read-only. This is outlined in the memory
map below.
There is no other means of getting back from application mode to
2022-09-19 03:43:53 -04:00
firmware mode than resetting/power cycling the device.
2022-09-19 02:51:11 -04:00
Prerequisites: `APP_SIZE` and `APP_ADDR` has to be non-zero.
2022-09-19 03:43:53 -04:00
Procedure:
2022-09-19 02:51:11 -04:00
1. The host sends `FW_CMD_RUN_APP` to the device.
2022-09-19 03:43:53 -04:00
2. The firmware responds with `FW_RSP_RUN_APP`
3. The firmware writes to `SWITCH_APP`, and executes assembler code
that writes zeros to stack and data of the firmware, then jumps
to what's in APP_ADDR.
2022-09-19 03:43:53 -04:00
4. The device is now in application mode and is executing the
application.
2022-09-19 02:51:11 -04:00
### Protocol definition
Available commands/reponses:
2022-09-19 03:43:53 -04:00
#### `FW_{CMD,RSP}_LOAD_USS`
2022-09-19 02:51:11 -04:00
#### `FW_{CMD,RSP}_LOAD_APP_SIZE`
#### `FW_{CMD,RSP}_LOAD_APP_DATA`
#### `FW_{CMD,RSP}_RUN_APP`
#### `FW_{CMD,RSP}_NAME_VERSION`
2022-10-24 09:35:39 -04:00
#### `FW_{CMD,RSP}_UDI`
2022-09-19 03:43:53 -04:00
2022-09-19 02:51:11 -04:00
#### `FW_{CMD,RSP}_VERIFY_DEVICE`
2022-09-19 03:43:53 -04:00
2022-10-20 08:50:21 -04:00
Verification that the device is an authentic Tillitis
2022-09-19 02:51:11 -04:00
device. Implemented using challenge/response.
#### `FW_{CMD,RSP}_GET_APP_DIGEST`
2022-09-19 03:43:53 -04:00
2022-09-19 02:51:11 -04:00
This command returns the un-keyed hash digest for the application that
was loaded. It allows the host to verify that the application was
correctly loaded. This means that the CDI calculated will be correct
given that the UDS has not been modified.
#### Get the name and version of the device
```
host ->
u8 CMD[1 + 1];
CMD[0].len = 1 // command frame format
CMD[1] = 0x01 // FW_CMD_NAME_VERSION
host <-
u8 RSP[1 + 32]
RSP[0].len = 32 // command frame format
2022-09-19 02:51:11 -04:00
RSP[1] = 0x02 // FW_RSP_NAME_VERSION
RSP[2..6] = NAME0
RSP[6..10] = NAME1
RSP[10..14] = VERSION
RSP[14..] = 0
```
#### Load an application
2022-09-19 03:43:53 -04:00
2022-09-19 02:51:11 -04:00
```
host ->
u8 CMD[1 + 128];
CMD[0].len = 128 // command frame format
CMD[1] = 0x0a // FW_CMD_LOAD_USS
CMD[2..6] = User Supplied Secret
CMD[6..] = 0
host <-
u8 RSP[1 + 4];
RSP[0].len = 4 // command frame format
RSP[1] = 0x0b // FW_RSP_LOAD_USS
RSP[2] = STATUS
RSP[3..] = 0
2022-09-19 02:51:11 -04:00
host ->
u8 CMD[1 + 32];
CMD[0].len = 32 // command frame format
2022-09-19 02:51:11 -04:00
CMD[1] = 0x03 // FW_CMD_LOAD_APP_SIZE
CMD[2..6] = APP_SIZE
CMD[6..] = 0
host <-
u8 RSP[1 + 4];
RSP[0].len = 4 // command frame format
2022-09-19 02:51:11 -04:00
RSP[1] = 0x04 // FW_RSP_LOAD_APP_SIZE
RSP[2] = STATUS
RSP[3..] = 0
2022-09-19 03:43:53 -04:00
repeat ceil(APP_SIZE / 127) times:
2022-09-19 02:51:11 -04:00
host ->
2022-09-19 03:43:53 -04:00
u8 CMD[1 + 128];
2022-09-19 02:51:11 -04:00
2022-09-19 03:43:53 -04:00
CMD[0].len = 128 // command frame format
2022-09-19 02:51:11 -04:00
CMD[1] = 0x05 // FW_CMD_LOAD_APP_DATA
2022-09-19 03:43:53 -04:00
CMD[2..] = APP_DATA (127 bytes of app data, pad with zeros)
2022-09-19 02:51:11 -04:00
host <-
u8 RSP[1 + 4]
2022-09-19 03:43:53 -04:00
RSP[0].len = 4 // command frame format
2022-09-19 02:51:11 -04:00
RSP[1] = 0x06 // FW_RSP_LOAD_APP_DATA
RSP[2] = STATUS
RSP[3..] = 0
```
### Memory map
2022-09-19 03:43:53 -04:00
Assigned top level prefixes:
2022-09-19 02:51:11 -04:00
2022-09-19 03:43:53 -04:00
| *name* | *prefix* | *address length* |
|----------|----------|--------------------------------------|
| ROM | 0b00 | 30 bit address |
| RAM | 0b01 | 30 bit address |
| reserved | 0b10 | |
| MMIO | 0b11 | 6 bits for core select, 24 bits rest |
2022-09-19 02:51:11 -04:00
2022-09-19 03:43:53 -04:00
Addressing:
2022-09-19 02:51:11 -04:00
2022-09-19 03:43:53 -04:00
```
31st bit 0th bit
v v
0000 0000 0000 0000 0000 0000 0000 0000
- Bits [31 .. 30] (2 bits): Top level prefix (described above)
- Bits [29 .. 24] (6 bits): Core select. We want to support at least 16 cores
- Bits [23 .. 0] (24 bits): Memory/in-core address.
```
The memory exposes SoC functionality to the software when in firmware
mode. It is a set of memory mapped registers (MMIO), starting at base
address `0xc000_0000`. For specific offsets/bitmasks, see the file
2022-10-20 08:50:21 -04:00
[tk1_mem.h](../../hw/application_fpga/fw/tk1_mem.h) (in this repo).
2022-09-19 03:43:53 -04:00
Assigned core prefixes:
| *name* | *prefix* |
|--------|----------|
| ROM | 0x00 |
| RAM | 0x40 |
| TRNG | 0xc0 |
| TIMER | 0xc1 |
| UDS | 0xc2 |
| UART | 0xc3 |
| TOUCH | 0xc4 |
2022-10-25 11:27:26 -04:00
| FW_RAM | 0xd0 |
2022-10-20 08:50:21 -04:00
| TK1 | 0xff |
2022-10-25 11:27:26 -04:00
*Nota bene*: MMIO accesses should be 32 bit wide, e.g use `lw` and
`sw`. Exceptions are `FW_RAM` and `QEMU_DEBUG`.
| *name* | *fw* | *app* | *size* | *type* | *content* | *description* |
|-------------------|-------|-----------|--------|----------|-----------|------------------------------------------------------------------------|
| `TRNG_STATUS` | r | r | | | | TRNG_STATUS_READY_BIT is 1 when an entropy word is available. |
| `TRNG_ENTROPY` | r | r | 4B | u32 | | Entropy word. Reading a word will clear status. |
| `TIMER_CTRL` | r/w | r/w | | | | If TIMER_STATUS_READY_BIT in TIMER_STATUS is 1, writing anything here |
| | | | | | | starts the timer. If the same bit is 0 then writing stops the timer. |
| `TIMER_STATUS` | r | r | | | | TIMER_STATUS_READY_BIT is 1 when timer is ready to start running. |
| `TIMER_PRESCALER` | r/w | r/w | 4B | | | Prescaler init value. Write blocked when running. |
| `TIMER_TIMER` | r/w | r/w | 4B | | | Timer init or current value while running. Write blocked when running. |
| `UDS_FIRST` | r[^3] | invisible | 4B | u8[32] | | First word of Unique Device Secret key. |
| `UDS_LAST` | | invisible | | | | The last word of the UDS |
| `UART_BITRATE` | r/w | | | | | TBD |
| `UART_DATABITS` | r/w | | | | | TBD |
| `UART_STOPBITS` | r/w | | | | | TBD |
| `UART_RX_STATUS` | r | r | 1B | u8 | | Non-zero when there is data to read |
| `UART_RX_DATA` | r | r | 1B | u8 | | Data to read. Only LSB contains data |
| `UART_TX_STATUS` | r | r | 1B | u8 | | Non-zero when it's OK to write data |
| `UART_TX_DATA` | w | w | 1B | u8 | | Data to send. Only LSB contains data |
| `TOUCH_STATUS` | r/w | r/w | | | | TOUCH_STATUS_EVENT_BIT is 1 when touched. After detecting a touch |
| | | | | | | event (reading a 1), write anything here to acknowledge it. |
| `FW_RAM` | r/w | invisible | 1 kiB | u8[1024] | | Firmware-only RAM. |
| `UDA` | r | invisible | 16B | u8[16] | | Unique Device Authentication key. |
| `UDI` | r | r | 8B | u64 | | Unique Device ID (UDI). |
| `QEMU_DEBUG` | w | w | | u8 | | Debug console (only in QEMU) |
| `NAME0` | r | r | 4B | char[4] | "tk1 " | ID of core/stick |
| `NAME1` | r | r | 4B | char[4] | "mkdf" | ID of core/stick |
| `VERSION` | r | r | 4B | u32 | 1 | Current version. |
| `SWITCH_APP` | r/w | r | 1B | u8 | | Write anything here to trigger the switch to application mode. Reading |
| | | | | | | returns 0 if device is in firmware mode, 0xffffffff if in app mode. |
| `LED` | w | w | 1B | u8 | | |
| `GPIO` | | | | | | |
| `APP_ADDR` | r/w | r | 4B | u32 | | Firmware stores app load address here, so app can read its own location|
| `APP_SIZE` | r/w | r | 4B | u32 | | Firmware stores app app size here, so app can read its own size |
| `CDI_FIRST` | r/w | r | 32B | u8[32] | | Compound Device Identifier (CDI). UDS+measurement... |
| `CDI_LAST` | | r | | | | Last word of CDI |
2022-09-19 03:43:53 -04:00
2022-10-24 05:58:54 -04:00
[^3]: The UDS can only be read *once* per power-cycle.