Allow SP to set active slot #1808
Conversation
It's useful for the SP to be able to change the active slot. Pull this through. The final finish step no longer implicitly swaps the bank.
| &mut self, | ||
| _: &RecvMessage, | ||
| ) -> Result<SlotId, RequestError<Infallible>> { | ||
| Ok(self.pending) |
There was a problem hiding this comment.
Is this state retrievable from a HW register? Just thinking of the corner case where update_server restarts and having two sources of truth.
There was a problem hiding this comment.
Ok... I see the fetch of HW state below. Looks ok.
| @@ -69,5 +69,24 @@ Interface( | |||
| ), | |||
| idempotent: true, | |||
| ), | |||
| "get_pending_boot_slot": ( | |||
There was a problem hiding this comment.
The idl/lpc55-update-server.idol switch_default_image API doesn't fit well enough to reuse here.
Someday we could unify the boot policy APIs.
|
Fixes #1755 |
| Inactive = 1, | ||
| } | ||
|
|
||
| impl TryFrom<u16> for SlotId { |
There was a problem hiding this comment.
If you're using this for something other than IPC, carry on.
For IPC specifically, your use of encoding: Hubpack relieves you from needing these instances (and explicit = X discriminator values). FYI.
| @@ -519,9 +543,24 @@ impl NotificationHandler for ServerImpl<'_> { | |||
| fn main() -> ! { | |||
| let flash = unsafe { &*device::FLASH::ptr() }; | |||
|
|
|||
| // If the server restarts we need to fix our pending state | |||
There was a problem hiding this comment.
Nice, this looks like correct handling of task restart.
| SerializedSize, | ||
| )] | ||
| pub enum SlotId { | ||
| Active = 0, |
There was a problem hiding this comment.
question: From the peanut gallery: Does the naming of the slot IDs come from an outside source, or is it otherwise well understood within Oxide?
From an outsider perspective, it seems weird that apparently there are two banks that you can boot to, so either can be the "actively booted bank" but one is still called "inactive" and the other is "active", regardless of which one is actually currently booted into.
I guess maybe these aren't actually slot/bank IDs but rather just a value that can be set as 1 to say "after restart, switch from current bank to other bank"? If that is the case, would this rather be a "pending boot slot value"?
It's useful for the SP to be able to change the active slot. Pull this through. The final finish step no longer implicitly swaps the bank.