mirror of
https://github.com/saymrwulf/xous-core.git
synced 2026-11-15 08:30:02 +00:00
firmware updater strategy and thoughts jotted into a file
This commit is contained in:
parent
cddd971aa9
commit
0733b65ead
1 changed files with 45 additions and 0 deletions
|
|
@ -15,6 +15,51 @@ use utralib::generated::*;
|
|||
// This puts the byte-address hex offset at (6 * 256 + 8 * 8) / 8 = 0xC8
|
||||
// within the IFR region. Total IFR region size is 0x200.
|
||||
|
||||
/*
|
||||
Thoughts on where to put the updating routine.
|
||||
|
||||
As a Xous runtime program:
|
||||
Pros:
|
||||
- Full security of MMU
|
||||
- Updater primitives are available as Xous primitives
|
||||
Cons:
|
||||
- Less RAM available to stage data
|
||||
- ReRAM is XIP; makes kernel update tricky
|
||||
As a loader program:
|
||||
Pros:
|
||||
- Higher performance (no OS overhead)
|
||||
- Faster dev time
|
||||
- Can stage full images in PSRAM before committing
|
||||
- Full overwrite of ReRAM OS possible
|
||||
Cons:
|
||||
- Larger loader image
|
||||
- No MMU security - bugs are more brittle/scary
|
||||
- Loader becomes a primary attack surface due to its size and complexity
|
||||
- Less code re-use with main code base
|
||||
- Loader update needs a special path to hand-off an image to Xous to avoid XIP conflict
|
||||
|
||||
I think the winning argument is that we could stage the full image in PSRAM before
|
||||
committing to either ReRAM or SPI RAM. This allows us to do full signature checking prior
|
||||
to committing any objects.
|
||||
|
||||
We could structure this so that when going into update mode, the secret key lifecycle
|
||||
bits are pushed forward, so we only have derived keys available. This would make any
|
||||
break into the loader updater less able to get at any root keys?
|
||||
*/
|
||||
|
||||
/*
|
||||
To-do:
|
||||
-[x] Add dummy lifecycle gate call
|
||||
-[ ] New OLED base driver
|
||||
-[ ] I2C driver for AXP2101
|
||||
-[ ] camera base driver (maybe loopback to OLED as demo?)
|
||||
-[ ] USB stack into loader; debugging there. Present as bulk transfer to emulated disk on
|
||||
PSRAM using ghostFS
|
||||
-[ ] bring mbox routine into loader so we can have access to ReRAM write primitive; structure so
|
||||
that we can improve this easily as the chip bugs are fixed
|
||||
-[ ] Image validation & burning routine
|
||||
*/
|
||||
|
||||
pub const RAM_SIZE: usize = utralib::generated::HW_SRAM_MEM_LEN;
|
||||
pub const RAM_BASE: usize = utralib::generated::HW_SRAM_MEM;
|
||||
pub const FLASH_BASE: usize = utralib::generated::HW_RERAM_MEM;
|
||||
|
|
|
|||
Loading…
Reference in a new issue